Positivist realism

PDF

Download

Open document

Transcript (if available)

Content Positivist Realism Caleb Perl School of Philosophy Doctor of Philosophy (PHILOSOPHY) FACULTY OF THE USC GRADUATE SCHOOL University of Southern California Projected Conferral Date: August 8, 2017 1 Contents 0 Introduction 7 I Part I: Introducing positivist realism 12 1 Moral skepticism as pseudo-problem 13 1.1 Introducing a new picture . . . . . . . . . . . . . . . . . . . . 14 1.1.1 Background facts about presupposition . . . . . . . . 14 1.1.2 Why the Standard Assumption might be false . . . . . 19 1.2 Global Skepticism I: A Humean argument . . . . . . . . . . . 21 1.3 Global Skepticism II: arguments from disagreement . . . . . . 25 1.4 Global Skepticism III: Harman's dispensability argument . . 29 1.5 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 2 Moral debunking as linguistic mistake 34 2.1 Debunking, introduced . . . . . . . . . . . . . . . . . . . . . . 34 2.2 The positivist answers epistemic debunking . . . . . . . . . . 38 2.2.1 The basic answer . . . . . . . . . . . . . . . . . . . . . 39 2.2.2 The positivist's answer in a broader context . . . . . . 44 2.3 Does evolution undermine acceptance? . . . . . . . . . . . . . 46 2.4 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3 Moral naturalism without naturalized moral epistemology 51 3.1 A baseline about the naturalist's distinctive commitments . . 53 3.2 Back to the Standard Assumption . . . . . . . . . . . . . . . 56 3.2.1 Central hallmarks . . . . . . . . . . . . . . . . . . . . 56 3.2.2 Contrasts between moral knowledge and scientic knowl- edge . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 3.3 Questions about truth . . . . . . . . . . . . . . . . . . . . . . 61 3.4 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 2 4 De re belief 66 4.1 Basic issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.1.1 A companion in innocence . . . . . . . . . . . . . . . . 67 4.1.2 Turning to the moral case . . . . . . . . . . . . . . . . 70 4.2 Rening the rst pass . . . . . . . . . . . . . . . . . . . . . . 71 4.2.1 Sets of rules and orderings . . . . . . . . . . . . . . . . 71 4.2.2 Forms and limits of positivist realism . . . . . . . . . 73 4.3 Minor renements . . . . . . . . . . . . . . . . . . . . . . . . 75 4.3.1 What do we Ramsify over? . . . . . . . . . . . . . . . 75 4.3.2 Transparency's revenge . . . . . . . . . . . . . . . . . 76 4.4 Contrast with \moral functionalism" . . . . . . . . . . . . . . 78 4.5 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 4.6 Appendix: Semantics for partial orderings . . . . . . . . . . . 82 4.6.1 Supervaluationist semantics for partial orderings . . . 82 4.6.2 Why supervaluationist semantics for partial orderings? 86 5 De re knowledge 90 5.1 Performance errors, how . . . . . . . . . . . . . . . . . . . . . 90 5.2 Moral knowledge . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.2.1 The good cases: when mental representations support knowledge . . . . . . . . . . . . . . . . . . . . . . . . . 94 5.2.2 The bad cases: where mental representations don't support knowledge . . . . . . . . . . . . . . . . . . . . 97 5.3 Explaining the agent's beliefs . . . . . . . . . . . . . . . . . . 100 5.4 New horizons . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 II Part II: Arguing for positivist realism 105 6 Presuppositions are not entailments 108 6.1 Background about presupposition . . . . . . . . . . . . . . . . 109 6.1.1 Pragmatic explanations of presupposition . . . . . . . 110 6.1.2 Generalizing to attitude reports . . . . . . . . . . . . . 112 6.2 Why the Atom theory is false . . . . . . . . . . . . . . . . . . 113 6.2.1 A puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . 114 6.2.2 Attitude ascriptions without privileged commitments . 115 6.2.3 Presuppositions are dierent . . . . . . . . . . . . . . 118 6.3 Fission explanations of the Core Asymmetry . . . . . . . . . 122 6.3.1 A purely pragmatic Fission theory . . . . . . . . . . . 123 6.3.2 A hybridized pragmatic explanation . . . . . . . . . . 125 3 6.4 A problem: local accommodation . . . . . . . . . . . . . . . . 128 6.4.1 Explanations of the Fission theory . . . . . . . . . . . 129 6.4.2 Back to the Core Asymmetry . . . . . . . . . . . . . . 131 6.5 Singular thought and thought by description . . . . . . . . . 132 6.6 Applications in epistemology . . . . . . . . . . . . . . . . . . 135 6.7 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 7 Four arguments for positivist realism 139 7.1 The Asymmetry Argument for the hybrid theory . . . . . . . 140 7.1.1 Unidimensional accounts of attitude asymmetries . . . 141 7.2 The argument from local accommodation . . . . . . . . . . . 143 7.3 Monism and Pluralism . . . . . . . . . . . . . . . . . . . . . . 146 7.3.1 Why the hybrid theory leaves this choice open . . . . 147 7.3.2 Why Pluralism is so important . . . . . . . . . . . . . 150 7.4 A hermeneutical argument for Pluralism . . . . . . . . . . . . 152 7.5 The anthropological use of moral language . . . . . . . . . . . 154 7.6 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 8 Shifty contextualism about epistemics 160 8.1 The new theory . . . . . . . . . . . . . . . . . . . . . . . . . . 161 8.1.1 How Pluralism explains both disagreement and rea- sonableness . . . . . . . . . . . . . . . . . . . . . . . . 162 8.1.2 Contrasting the Pluralist with other contextualists . . 164 8.2 A systematic picture of attitudes towards epistemics . . . . . 167 8.2.1 A central but underappreciated question . . . . . . . . 168 8.2.2 Back to the epistemic case . . . . . . . . . . . . . . . . 170 8.2.3 Consequences of this explanation of Pluralism . . . . . 171 8.3 Yalcin's \epistemic contradictions" . . . . . . . . . . . . . . . 173 8.4 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 8.5 Appendix: The account of context-sensitivity . . . . . . . . . 180 III Part III: Disagreement 184 9 Ascriptions of agreement and disagreement 186 9.1 The positivist's basic advantage . . . . . . . . . . . . . . . . . 187 9.2 Ascriptions of disagreement and agreement . . . . . . . . . . 189 9.3 A companion in innocence . . . . . . . . . . . . . . . . . . . . 191 9.3.1 Liberalized singular thought . . . . . . . . . . . . . . . 192 9.3.2 Unanchored discourse referents . . . . . . . . . . . . . 194 4 9.4 The positivist's problem is a problem about intensional anaphora198 9.5 The positivist's problem in a unidimensional semantics . . . . 200 9.6 A master argument . . . . . . . . . . . . . . . . . . . . . . . . 202 10 A theory of acceptance 205 10.1 Acceptance and disagreement . . . . . . . . . . . . . . . . . . 206 10.2 The basic account of acceptance . . . . . . . . . . . . . . . . 207 10.3 Connecting the basic account to the target phenomenon . . . 210 10.4 Returning to positivist realism . . . . . . . . . . . . . . . . . 215 10.5 How the moral case works . . . . . . . . . . . . . . . . . . . . 218 11 Disagreements about knowledge 221 11.1 Knowledge and unanchored discourse referents . . . . . . . . 221 11.2 Is moral knowledge too easy? . . . . . . . . . . . . . . . . . . 224 11.3 Other kinds of knowledge attributions . . . . . . . . . . . . . 229 11.4 Wrapping up . . . . . . . . . . . . . . . . . . . . . . . . . . . 230 IV Part IV: Normative upshots 231 12 A positivist vindication of the method of re ective equilib- rium 232 12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232 12.2 Normative theorizing: its nature and method . . . . . . . . . 233 12.3 In ationary versus de ationary vindications of the method . . 235 12.4 Is a de ationist attitude psychologically possible? . . . . . . . 237 12.5 Looking forward . . . . . . . . . . . . . . . . . . . . . . . . . 241 13 Rule consequentialists should be positivists 243 13.1 Rule consequentialism in re ective equilibrium? Not on stan- dard assumptions! . . . . . . . . . . . . . . . . . . . . . . . . 244 13.1.1 Ignorance of the consequences . . . . . . . . . . . . . . 246 13.1.2 A more general diagnostic tool . . . . . . . . . . . . . 248 13.2 More general questions . . . . . . . . . . . . . . . . . . . . . . 250 13.2.1 Background epistemic assumptions . . . . . . . . . . . 250 13.2.2 The scope of the problem . . . . . . . . . . . . . . . . 251 13.3 The baseline is very dierent given positivist realism . . . . . 253 13.4 The nature of the positivist's proposal . . . . . . . . . . . . . 255 13.4.1 Rough and ready knowledge . . . . . . . . . . . . . . . 255 13.4.2 What is the aim of normative inquiry, again? . . . . 257 13.5 Rule consequentialists should be positivists . . . . . . . . . . 259 5 14 The normative signicance of sociobiology 262 14.1 A central problem . . . . . . . . . . . . . . . . . . . . . . . . 263 14.2 Evolutionary arguments in normative ethics, without invidi- ous distinctions . . . . . . . . . . . . . . . . . . . . . . . . . . 265 14.3 Wait { what are the data for normative ethics? . . . . . . . . 268 14.4 Does the no-evidence argument discriminate between norma- tive theories? . . . . . . . . . . . . . . . . . . . . . . . . . . . 270 14.5 Isn't the simplest normative theory highly implausible? . . . 272 14.6 Yay for consequentialists! . . . . . . . . . . . . . . . . . . . . 274 14.7 Against trolleyology . . . . . . . . . . . . . . . . . . . . . . . 276 Bibliography 278 6 Chapter 0: Introduction This dissertation introduces a new picture of moral language and thought. This picture has signicant upshots for moral epistemology, for moral meta- physics, and for normative ethics. I will call my new picture positivist realism. I develop it in four Parts. Part I introduces the view, and explains why it would be nice for non- skeptical moral realists if it is true. The goal in this part won't be to argue that the view is actually true. Rather, I'll try to show that it would be nice if it were. Part II will argue that positivist realism is true, by marshaling a range of linguistic evidence. Part III will answer a central problem for the positivist. And then Part IV will introduce some upshots that positivist realism has for normative ethics. The positivist holds that a range of central problems in moral epistemol- ogy are pseudo-problems. They seem forceful only because we've misunder- stood the way our language works. In particular, they seem forceful only because we've failed to appreciate that moral discourse carries substantial presuppositions. The positivist makes non-skeptical moral realism easier to defend, in two dierent ways. First, it's much easier to answer moral skeptics if the positivist right. It's just not possible to formulate the skeptics' arguments in this picture. Second, the positivist increases the range of realist views that are worth taking seriously. An important class of metaethical arguments start from our conviction that we do have moral knowledge. Arguments in this class leverage that conviction as evidence against some realists. Sharon Street (2006) is one example. She argues that her non-realist conception of moral belief is the only way to vindicate that conviction. In doing that, she's taking that conviction as evidence for her view and as evidence against her realist rivals. This style of argument also gures in intramural debates among realists. Non-naturalists like T. M. Scanlon (2014) are one example. They use our conviction as evidence against their naturalist rivals. The positivist systematically disarms this kind of argument. The positivist thinks that metaethicists have made a very important 7 mistake about the semantics and pragmatics of moral language. This mistake has led them to try to address some problems that are in fact only pseudo- problems. Moreover, attempting to answer these pseudo-problems has had bad eects in other areas of moral philosophy. In particular, it has had bad eects in moral metaphysics and in normative ethics. 8 I'm very lucky to have gotten to spend the last six years working at USC. It's an extraordinary and inspiring place to do philosophy; getting to come here is one of the best things that has happened to me. (Or rather: one of the best things outside of meeting Ashley and being born into my family.) I'm grateful for the special philosophical community that USC has fos- tered, for all the people here who combine philosophical rigor with kindness and concern. My dissertation has beneted enormously from conversations with the people here. When I look back at earlier versions of this mate- rial, I see lots of claims that now look mistaken, or fatally underdeveloped. In lots of those cases, it took a long time for someone else to convince me of those mistakes, or to get me to appreciate how the idea was underdevel- oped. I owe so much to the community here at USC { to Mike Asheld, Rima Basu, Stephen Bero, Renee Bolinger, Alex Dietz, Eric Encarnacion, Maegan Fairchild, August Gorman, Joe Horton, Nathan Howard, Robin Jeshion, Tanya Kostochka, Nicholas Laskowski, Woo Ram Lee, Ben Lennertz, Alida Liberman, Michael Milona, Abelard Podgorski, Jon Quong, Kenneth Silver, David Wallace, and Aness Webster. I also owe a great deal to philosophers outside of USC. I've learned lots of dierent lessons from comments and conversations with Selim Berker, Gunnar Bj ornsson, Laura Callahan, Sam Carter, Ruth Chang, Jonathan Drake, Dan Fogal, Michael Hannon, Dan Immerman, Alex Kaiserman, Dan Korman, David Sobel, David Sosa, Je Tolly, Sungwoo Um, and Preston Werner. I'm especially grateful to my undergrad advisor, Allen Wood. Some of my earliest sensibilities about moral philosophy are traceable back to him { in some cases because his philosophical views were inspiring, and in some cases because they were frustrating or challenging. And those earlier sen- sibilities have mostly lasted. (Maybe because I'm stubborn, but hopefully because I've come to re ectively endorse them.) This dissertation is in part an attempt to make good on those philosophical sensibilities. It develops a tech-y suggestion, about moral language. But I would have lost my enthu- siasm for that suggestion if I ever became convinced that it wasn't a way of vindicating some of the philosophical commitments I started to care about as an undergrad. I'm especially grateful to seven people at USC. I'm grateful to John Hawthorne for lots of conversations about this material { all the more so because he's not even on my committee! His synoptic command of the eld revealed entirely new groups of problems, as well as totally new applications. And he helped me better inhabit some alien territory, especially knowledge- 9 rst epistemology and MIT tricks with attitude ascriptions. I'm also grateful to Jon Wright for many conversations about this material. Those conver- sations have pervasively shaped the nal product; only a handful of people have forced me to rethink as much as he has. I'm grateful for how much of the idea he's internalized, and for his ability to see subtle contradictions between radically dierent parts of the project. In one example of many: he saw that one of my early claims in normative ethics ended up contradicting some of my foundational assumptions about presupposition. I'm also grateful to my committee members. I'm grateful for the high standards that Scott sets. He and Mark made me into into the philosopher I am today. And I'm grateful for his command of what's likely to be peripheral and what's likely to be central. Three years ago, he suggested that one of the rst things I needed to do was to develop a better account of some data about presupposition. Attempting to do that has borne all kinds of fruit. Most of the details in Parts II-III come from that attempt. My idea has always had lots of complicated moving parts. But the moving part that Scott identied three years ago is plausibly the most fundamental; it's striking that he identied it so long ago. Je Speaks says in his dissertation that whole sections of chapters emerged from Scott's comments. That's true for me, too. Something stronger is true, as well { a whole chapter also emerged from Scott's comments: Chapter 6. And that chapter may be the central pivot for the whole project, in a way that I didn't expect at all when I started writing. I've been very fortunate with my other committee members, too. Ralph found several central clusters of problems, early enough that he helped me avoid some major dead-ends. He's also been incredibly constructive. For example, he has suggested that I model Chapter 4 on knowledge of idiolects, and that Chapter 14 focus on debunking arguments about sex. Those sugges- tions make both chapters much more eective. Steve has been fantastically helpful. Working with him has made the material much more intelligible. It's no surprise that the only chapter that I've published so far is the one that he has worked on most closely. He's also been very helpful in posing chal- lenges to the idea { Chapter 8 is, for example, an extended response to one of his challenges. And Barry Schein has been an invaluable resource, both in pointing out signicant problems and pointing me to relevant literature. Mark has been the Platonic form of an advisor. He often seems to under- stand my idea better than I do, both its upsides and its costs. And he's of course fantastic at getting me to appreciate the costs, and doing so in a way that reveals the features of the view that force those costs. And he's been an astonishingly dedicated advisor. It's common ground in the eld that 10 he's one of the best. (Just ask an arbitrary grad student at a well-connected program). But even given what's common ground, the amount of time and eort he spends on his students is still amazing. I couldn't have been luckier. 11 Part I Part I: Introducing positivist realism 12 Chapter 1: Moral skepticism as pseudo-problem I disagree with Jeerson Davis about the permissibility of slavery. He thought it permissible. I disagree. There's something wrong with believ- ing as he does { his belief has some epistemic aw. Maybe the aw is that the belief is unjustied. Some philosophers agree but go further. They think that my moral beliefs are also epistemically awed in some way. They argue that neither Davis nor I know any (substantive) moral propositions. 1 That is, they argue for global moral skepticism. (GlobalMoralSkepticism) For any substantive moral propo- sition p, no one knows p, and no one knows:p either. Global skeptics ordinarily assume a realist conception of morality. They assume that moral judgments are about facts that are independent of and more fundamental than our individual evaluative states. Global skeptics have developed a battery of arguments that we don't know those facts. Some argue from purported constraints on knowledge in general. For example, Hume claimed that that all knowledge is grounded in either knowledge of matters of facts or knowledge of relations of ideas. And that claim might well lead to global moral skepticism. Other skeptics argue from more particular facts about morality. For example, some ar- gue that Global Skepticism is the best explanation of pervasive moral disagreement. I will argue that any argument forGlobalSkepticism rests on a misun- derstanding of moral language and thought. Now in making this argument, I will be assuming a realist conception of moral judgment. Global skepticism is most pressing given that realist conception. So assuming it puts the global skeptic in the best position to make her argument. I will claim that skeptics and realists have both been relying on a mistaken picture of moral language. 1 In focusing on substantive propositions, I'm excluding propositions like the proposi- tion that if an act is not impermissible, then it's permissible. 13 They've been assuming that moral discourse carries less substantial presup- positions than it in fact does. Arguments for global skepticism lose their force when when we recognize this fact. Now global skeptics are likely to nd this response unsatisfying. They'll insist that it doesn't answer the deep and unsettling challenges that ani- mate them. I'll take this reaction to suggest that they're really interested in pseudo-problems. Given an adequate understanding of moral discourse, arguments for global skepticism are neither deep nor unsettling . 1.1 Introducing a new picture Global skeptics and their realist rivals both agree on a very orthodox picture of moral language. At the heart of that orthodox picture is what I will call the Standard Assumption. (Standard Assumption) If moral realism is true, moral knowl- edge requires knowledge about a property that is independent of and more fundamental than our individual evaluative attitudes. Global skeptics have relied on this assumption. They try to show that our attitudes towards the realist's property don't have whatever features dis- tinguish knowledge from mere true belief. Realists have argued against the skeptic by trying to show that our attitudes do have those features. They've also made this assumption in trying to answer skeptics. And it's no surprise that they have! TheStandard Assumption is highly plausible. For exam- ple, it follows if moral properties are the semantic values of moral terms. Jill knows that murder is morally wrong then expresses that Jill has knowledge about the realist's property. This paper develops and defends a new picture of moral discourse. The Standard Assumption is false if this picture is right. So this picture is philosophically interesting. If it's true, a range of central debates in moral epistemology would look dierent. And I'll go on to argue later in the dis- sertation that this picture is also highly plausible. It follows from a simple conjecture about the presuppositions of moral discourse. And that conjec- ture is elegant and highly principled, when considered on its own merits as a purely linguistic thesis. 1.1.1 Background facts about presupposition As a warmup to introducing this new picture, I'll review some important facts about presuppositions. 14 As a rst pass, I'll understand the presuppositions of an utterance of a sentence as information that that utterance (i) is somehow associated with, but (ii) is somehow interpreted as being part of the background assumptions in the conversation. It's easier to appreciate this notion from particular examples. (i) It wasn't Tom who stole the tarts. (ii) I'm not going to the party. (iii) John was also in NY last night. Each of these sentences are somehow associated with two propositions, with one proposition interpreted as backgrounded and not-at-issue. (i) is associ- ated with the backgrounded, not-at-issue proposition that someone did steal the tarts. You wouldn't use (i) unless you accept that someone stole the tarts, but the main point of using (i) is to communicate that Tom didn't steal the tarts. 2 Backgrounded, not-at-issue content interacts with attitude ascriptions in a very distinctive way. Consider, for example, the way that (i) composes with attitude ascriptions. (1a) John hopes that it wasn't Tom who took out the trash. (1b) John fears that it wasn't Tom who took out the trash. (1c) John conjectures that it wasn't Tom who took out the trash. 2 I'm taking the term not-at-issue from the research tradition descending from the important work by Craige Roberts (2012), though I don't mean to import all her back- ground assumptions here. The other examples are similar { (ii) is associated with the backgrounded, not-at-issue proposition that there is a party, and (Iii) is associated with the backgrounded, not-at-issue proposition that someone else was in NY last night. (The characterization of (ii) isn't quite right; see Saul Kripke (2009) for a more careful account.) Now there are other kinds of backgrounded, not-at-issue content, besides presuppo- sitions. But this paper will restrict its attention to presuppositions, because some of the distinctive hallmarks of presupposition are what make the phenomenon particularly interesting. Other kinds of constructions associated with not-at-issue content includes non-restrictive relatives, and honorics. Think, for example, of the sentence Mary, who likes her mother, lives near her. Uttering that sentence communicates that Mary likes her mother, but we interpret that proposition as backgrounded, not as part of the main point of the utterance. Christopher Potts (2005) has an in uential discussion of some other candidates, and Judith Tonhauser et al. (2013) survey some other important relevant data. 15 (1a) can be true even if John doesn't hope that someone took out the trash. (Maybe he knows that someone took out the trash, and regrets that the trash was taken out.) The other examples are the same. (1b) can be true even if John doesn't fear that someone took out the trash; and (1c) can be true even if John doesn't conjecture that someone took out the trash. More generally, utterances of pA V-es that Sq can often be appropriate even if A doesn't V the presuppositions of S's complement. 3 Do utterances of pA V-es that Sq require that A has positive some attitude towards S's presupposition(s)? Yes. It's normally odd to use (1a)/(1b)/(1c) unless John has some positive attitude toward the presup- position of the complement. I'll adopt a proposal from Robert Stalnaker to capture this fact. He holds that attitude reports usually communicate that the subject accepts the presupposition of the complement, where \to accept a proposition is to treat it as true for some reason. One ignores, at least temporarily, and perhaps in a limited context, the possibility that it is false" (Stalnaker, 2002, 716). Importantly, then, there are lots of dierent grounds for acceptance. You might accept that S because you believe that S, or you know that S. But this picture allows for other kinds of reasons for accepting that S. This Stalnakerian picture thus contrasts with a more restrictive picture, where pA V-es that Sq requires A to know S's presupposition(s). It will be very important in this paper that the Stalnakerian picture is right and the more restrictive picture is wrong. So it's important for me to motivate the Stalnakerian picture. In fact, though, it's easier to motivate that picture 3 Now you might wonder whether these examples illustrate the point I take them to. In general, it's possible to hope in some content without hoping in everything that the content entails. You know that Bob is coming to the party, and think him a bore { you'd rather that he not be there. But Mary would redeem Bob's presence. (Mary's ris on Bob's personality are really funny.) You hope that Bob and Mary come can then be true, even though you hope that Bob comes isn't. Maybe (1a) is appropriate for the same reason. (1a)'s complement really does express the conjunctive content that it seems to. It's just that you don't need to hope in both conjuncts for (1a) to be appropriate. Maybe that's all these examples show { that you can hope in a content without hoping in everything the content entails. Clefts express conjunctive propositions: that someone took out the trash, and that Tom didn't. So there isn't anything special about these clefts; they just illustrate the general point that you can hope in a content without hoping in all its entailments. But that diagnosis is implausible, because it overgenerates. Suppose that John [hopes/ fears/ conjectures] that someone took out the trash. The present idea predicts that (1a) { (1c) should be appropriate then. After all, he's hoping/ fearing/ conjecturing one of the two conjuncts, and that's imagined to be enough to make the attitude reports appropriate. But it's very hard to understand the attitude reports as appropriate for that reason. The present idea overgenerates. 16 than you might expect. Note rst that the two theories agree whenever A does know the presupposition(s) of the complement. In that case, the Stalnakerian picture will automatically take A to accept the presuppositions. Knowing some proposition is a great reason for accepting it! The two theories make dierent predictions only when someone accepts a presupposition that they don't know. The Stalnakerian picture predicts that attitude reports are appropriate in that case, while the more restric- tive picture predicts that they're not. The Stalnakerian prediction is more plausible. (2a) I don't know whether someone took out the trash. But I'll accept for now that someone did. I denitely know that it wasn't Tom who took it out. I've been watching him all day, and he went nowhere near it. An utterance of it wasn't Tom who took it out presupposes that someone took it out. But (2a) is still appropriate, even though I acknowledge that I don't know that presupposition. So the attitude reports in (2a) can't require me to know the presupposition triggered from the complement. And that's just the prediction that the Stalnakerian picture makes. It's worth double-checking this data point. We should make sure that the felicity of (2a) is probative about presupposition, rather than probative about this particular linguistic environment. In order for the felicity of (2a) to be probative about presupposition, a discourse that asserts ignorance of the at-issue content and knowledge of the presupposed, not-at-issue content should sound defective. Does it? (2b) I don't know whether Tom took out the trash. But I'll accept for now that he didn't. * I denitely know that it wasn't Tom who took it out. The trash is empty, so someone took out out. (2b) is defective, in just the way the Stalnakerian should hope for. The third sentence asserts that I know that Tom didn't take out the trash, and the rst sentence asserts that I don't know whether he did. So the pattern in (2a) does reveal something about presupposition, rather than something about the particular linguistic environment. It's worth noting, too, that the same pattern happens with other pre- supposition triggers. 17 (2c) I don't know whether there'll be a party tonight. But I'll accept for now that there will be. I denitely know that I won't be going to the party. I'm too sick to leave my room. (2d) I don't know whether Mary ate in NYC tonight. But I'll accept for now that she did. I denitely know that Jane also ate in NYC tonight; I had dinner with her. We should accept the Stalnakerian picture, where acceptance of the presup- position is what matters. So we should hold that: knowledge is not a norm on presupposition. This conclusion will be very important throughout the rest of the disserta- tion. I'll defend it in more detail later. But for now, I'll just adopt a picture of presupposition triggers that cleanly explains this point. I'll use a one-many picture of presupposition triggers to that end. On this one-many picture, utterances of one sentence containing a presupposition trigger will communicate two dierent proposi- tions. One of the propositions is interpreted as at-issue, and the other as not-at-issue. 4 (*) I know that it wasn't Tom who took out the trash At-issue: I know that Tom didn't take out the trash and accept that someone took out the trash Not-at-issue: Someone took out the trash This one-many conception is a simple way of capturing the data the we've just seen. Why are utterances of (*) appropriate even if you don't know that someone took out the trash? Answer: none of the propositions communi- cated require you to know that someone took out the trash. 5 4 There is broad agreement that attitude reports communicate something about the matrix subject's attitude towards presuppositions; Paul Elbourne (2005), Bart Geurts (1998), Irene Heim (1992), and Lauri Karttunen (1974) all agree on this basic point. They don't consider the choice between the attitude being belief and its being acceptance. So I take the present Stalnakerian consideration to be compelling enough to motivate the conclusion that it's acceptance and not belief that matters. A range of philosophers have drawn on this kind of attitude of acceptance: Robert Stal- naker (2002) relies on it in explaining cooperative communication, and Bas van Fraassen (4980) takes scientists to adopt something like this attitude towards at least some theories. You might even think that the attitude described in the main text is very much like our ordinary notion of belief, if you think that the evidential standards for belief are very low. (Hawthorne, Rothschild, and Spectre (2016) have recently argued for that last possibility.) 5 Unfortunately, though, this one-many explanation of (*) is the minority view about 18 1.1.2 Why the Standard Assumption might be false Let's return to moral skepticism. I promised to introduce a new picture, a picture where the Standard Assumption is false. (Standard Assumption) If moral realism is true, moral knowl- edge requires knowledge about a property that is independent of and more fundamental than our individual evaluative attitudes. And I advertised my new picture as building from a claim about presup- position. We're now in a better position to appreciate why a claim about presupposition might be incompatible with the Standard Assumption. knowledge is not the norm on presupposition. So the Standard Assumption would be false if propositions about the realist's fundamental property are only presuppositions of the moral use of modals. Throughout the rest of this paper, I'll use the predicate is the moral standard to refer to the realist's fundamental property. So we can formulate the possibility where the Standard Assumption is false as the possibility that: propositions about being-the-moral-standard are presuppositions of the moral use of modals. These two indented claims are at the heart of my picture. I suggest that moral utterances assert propositions about sets of rules. An utterance of (3) communicates two propositions about some particular set of rules s 1 : (3) You ought to keep this promise. presupposition triggers under attitude verbs. (Theorists who defend it include at least Paul Dekker (2008), Karttunen and Peters (1979), Robert van Rooij (2005, 2010), David Oshima (2006), and Yasutada Sudo (2012).) The majority view takes utterances of (*) to communicate a single proposition where the presupposition is embedded under the attitude verb. (See, for example, Bart Geurts (1998), Irene Heim (1992) and Judith Tonhauser et al. (2013).) For our purposes, though, the minority view is a helpful guide to the facts about pre- supposition. The majority view is plausible only if it makes the same predictions as the one-many view does. It needs to explain why (*) is appropriately assertable in the con- text described, and in a range of similar contexts. Proponents of the majority view don't contest this point. They try to develop tools to explain it. If they succeed, the one-many view is an adequate guide to their view. If they don't succeed, their view isn't viable. 19 At-issue: s 1 demands that you keep this promise Not-at-issue: s 1 is the moral standard The foregrounded, at-issue content and the backgrounded, not-at-issue con- tent are linked together, because they're about the same set of rules. The at-issue proposition describes the demands of the set of rules, and the not- at-issue proposition describes that set of rules as the moral standard. A dierent utterance (of you must lie now, say) is associated with the same kind of not-at-issue proposition (that s 1 is the moral standard), and a dif- ferent at-issue one (that s 1 permits your lying now). Given this picture, a knowledge ascriptions communicate that the subject knows the at-issue content and accepts the not-at-issue content. (4) Jane knows that she ought to keep this promise. At-issue: Jane knows that s 1 requires keeping this promise, and accepts that s 1 is the moral standard Not-at-issue: s 1 is the moral standard. So this picture is incompatible with the Standard Assumption. Ordinary moral knowledge does not depend on knowledge of the fundamental realist property. But that property still plays an important role in knowledge as- criptions: it gures in what they presuppose. I will call this view positivist realism. Now one important piece of this conjecture is the suggestion that moral utterances are about a particular set of rules. And this suggestion might strike you as implausible. For one thing, you might think that the moral uses of ought just express a simple monadic property of propositions, or a simple dyadic relation between agents and actions. And I'm suggesting, by contrast, that those predicates express either a dyadic relation between sets of rules and propositions, or a triadic relation between sets of rules, agents, and actions. But I won't argue for this suggestion here. It's already standard to take modals like ought to have an argument place for what I've just called a set of rules. (It's more standard to use the term \ordering" for what saturates that argument place, but that's just a terminological dierence.) Linguists call this view contextualism about the moral use. 6 6 Angelika Kratzer (1977, 1981, 2012) is the locus classicus for positing another argu- ment place, but it's also common ground in a wide range of otherwise heterodox frame- works, including dynamic approaches like that of Malte Willer (2014) and relativist ap- proaches like Niko Kolodny and John MacFarlane (2010). 20 Importantly, though, this sort of contextualism is perfectly compatible with moral realism. The realist can accept it by holding that moral uses of modals communicate something about her fundamental moral property. In particular, she can claim that moral uses communicate that the set of rules (ordering) under discussion is the moral standard. But this claim doesn't settle whether that proposition is part of the foregrounded, at-issue content associated with the utterance, or part of the backgrounded, not-at-issue content. Now the conjecture that moral utterances carry this sort of presupposi- tion is a substantive conjecture that needs to be defended. But remember that we're just in Part I of the dissertation. And the goal of Part I is to show that positivist realism has nice upshots for non-skeptical realists. So I won't defend this substantive conjecture here. I'll rather show what follows if it is true, and save the task of defending it for Part II. 1.2 Global Skepticism I: A Humean argument My overarching goal is to show that no argument for global skepticism could succeed if the positivist is right. This chapter tackles that goal with three cases studies. I examine three arguments for global skepticism, and show how they fail if the positivist is right. Remember that the global skeptic is issuing a challenge to someone who believes that slavery is impermissible, as much to someone who believes that it's permissible. (That's just what it is to issue a global challenge.) So the global skeptic will insist that (8) is true. (8) Neither Jeerson Davis nor Frederick Douglass know what they believe about slavery. To answer the global skeptic, you need to show that (8) is false, and the (9) is true. (9) Either Jeerson Davis or Frederick Douglass know what they believe about slavery. That's what I'll do in this section. My rst example of an argument for global skepticism starts from some demanding strictures that Hume placed on knowledge. These strictures are probably too demanding to be plausible. But they're helpful for my purposes here, as a way of introducing the epistemic upshots of my presuppositional 21 picture. Hume require all knowledge to be either knowledge of \relations of ideas" or \matters of fact". Knowledge of relations of ideas is concep- tual knowledge. It is grounded in understanding of the concepts involved. Knowledge of matters of facts is empirical knowledge, about the contingent world we inhabit. Hume argues that our moral knowledge cannot be explained in either way. Consider a simple inference that seems to extend our moral knowledge. 1. This action would be an intentional telling of a falsehood. 2. It doesn't harm anyone. 3. So this action is wrong. Hume would deny that knowledge of these premises can support knowledge of the conclusion. There won't be any conceptual connection between the descriptive properties of the action and its moral properties. (For one thing, there isn't any contradiction in arming that some action has those de- scriptive properties while denying it has the moral property. Hume takes this point to show that there isn't a conceptual connection between the two properties.) But he also denies that knowledge of the conclusion is knowl- edge of matters of fact. We don't form our moral beliefs in the way we form beliefs about matters of facts. 7 Now if the positivist is right, the conclusion communicates two proposi- tions. 3. So I know that this action is morally wrong. At-issue: I know that o 1 forbids this action, and I accept that o 1 is the moral standard Not-at-issue: o 1 is the moral standard I'll show that 3. can follow from these premises. But to do that, we need a more complete picture what it takes to have knowledge about a set of rules { the sort of de re knowledge that the at-issue content requires. And I haven't said much yet about that sort of knowledge. It's helpful to have an articial example in mind, to illustrate how that knowledge would be possible. Suppose that Jane found a book describing the considerations that s 1 is sensitive to: for example, that s 1 takes the fact 7 This argument is one of the two arguments in the opening discussion of morals in the Treatise (III.1.1.26). 22 that I've promised to to count in favor of-ing being required by s 1 , that s 1 takes that consideration to be undercut if I can save a life by breaking the promise. Suppose further that she memorized that book. Because she has memorized that set of rules, she knows what s 1 demands. Now explicit memorization is an articial example of how we could come to know about that set of rules. A more realistic suggestion is that we know about our own set of rules mostly because we've been habituated into the way that our social group does things, and maybe also because we've modied that set of rules in our own idiosyncratic ways. That habituation gives us knowledge about our set of rules. The positivist then vindicates the knowledge that that Humean argument threatens. The basic reason you can come to know the moral conclusion from these descriptive premises is that the only kind of knowledge required is knowledge about how the relevant set of rules o 1 classies the action. You don't need know any propositions containing the realist's fundamental property. So the positivist reduces the task of explaining how this inference can extend our knowledge to the task of explaining how we can come to know how o 1 classies the action. And knowledge about the set of rules o 1 is purely descriptive knowledge about something natural, about a matter of fact. You can acquire knowledge of o 1 by memorizing a book listing the signicance of dierent features of actions, and how they weigh together. That memorization only gives you descriptive knowledge of the set of rules in that book; it gives you knowledge of matters of fact. But the positivist holds that that kind of descriptive knowledge is the only kind of knowledge necessary for moral knowledge. You do need to accept that o 1 is the moral standard, but you don't need to know that it is. That acceptance doesn't need to be traced back to matters of fact, or to relations of idea. The answer thus lowers the bar for moral knowledge: it makes the ingredients of moral knowledge much cheaper than Hume assumed. The positivist's answer works by rejecting a very natural picture of moral knowledge. On that natural picture, sentential knowledge reports are a good guide to propositional knowledge attributions. Sentential knowledge reports have a single object (a sentence). So the corresponding propositional knowl- edge ascription should only require knowledge of one proposition. For the positivist, by contrast, sentential knowledge reports are not a good guide. Moral sentences are associated with two propositions, and you can have moral knowledge even if you only know one of the two propositions. The positivist's idea is thus like the hybrid theories of moral judgment that match 23 moral belief ascriptions with pairs of beliefs and pro-attitudes. 8 The dier- ence is that the the new secondary attitude is of accepting some set of rules as the moral standard, rather than a pro-attitude. It's then possible to see how there can be inferences from descriptive observations to a normative conclusion. One of the constituents of the normative conclusion is itself a purely descriptive proposition: just the right sort of thing to infer from de- scriptive premises! The other constituent of the conclusion is not descriptive, but normative. But you don't need to know that proposition. Go back to the disagreement about slavery { whether (9) is true. (9) Either Jeerson Davis or Frederick Douglass know what they believe about slavery. Global moral skeptics deny that (9) is true. We're considering an argument for global skepticism that relies on Humean strictures to try to show that we don't know about the realist's fundamental property being-the-moral- standard. For the positivist, though, the disjunctive knowledge report in (9) doesn't require knowledge of that property. It only requires acceptance of some set of rules as having that property. The positivist thus vindicates the disjunctive knowledge report (9). Since the global skeptic denies that knowledge report, the positivist decisively answers the global skeptic. The global skeptic is just worried about a pseudo-problem. At this point, we can generalize the discussion of Hume's challenge in one further way. He famously denied that we can infer moral conclusions from descriptive premises. I am surprised to nd, that instead of the usual copulations of propositions, is, and is not, I meet with no proposition that is not connected with an ought, or an ought not. ... at the same time that a reason should be given, for what seems altogether inconceivable, how this new relation can be a deduction from others, which are entirely dierent from it' (Hume, 1739, III.i.i) This passage has several dierent interpretations. On one interpretation, it's continuous with the challenge this section has explored. You can't infer an 8 Contemporary hybrid theorists include David Alm (2000), Stephen Barker (2000), Dorit Bar-On and Matthew Chrisman (2009), Daniel Boisvert (2008), David Copp (2001), Frank Jackson (1999), Richard Joyce (2006b), Michael Ridge (2006a, 2014), and Caj Strandberg (2012, 2015). Stephen Finlay (2004, 2014) has shown how traditional cog- nitivists can use traditional non-cognitivist resources in surprising ways, and Allan Gib- bard (2003) has done something similar in the opposite direction. Mark Schroeder (2009) helpfully surveys the costs and benets of the dierent options. 24 ought from an is, because such inference can only involve relations of ideas or matters of fact, and neither option makes the inference appropriate. This interpretation makes sense of Hume's further claim to \subvert all the vulgar systems of morality" (III.i.i). Those vulgar systems assume that descriptive knowledge can ground moral knowledge, and Hume is arguing that they're wrong. The positivist crisply defends vulgar systems of morality. The ingredients of moral knowledge come so cheap that the vulgar moralists could be right. 1.3 Global Skepticism II: arguments from disagreement I'll now turn to my second case study, which is about moral disagreement. Persistent moral disagreement can cast doubt on our reliability. Imagine that you and I are reasoning in the same way about moral questions. We're both forming our beliefs by relying on the same methods. It's natural to think that our methods just aren't very reliable if they don't settle our disagreements in the long run. And lots of moral disagreement does seem to persist in the long run. But in those cases where we're not reliable, we don't know. So we have less moral knowledge than we think we do. This kind of challenge has a hoary history. It was important for skeptics like Sextus Empiricus and Montaigne. And it has signicantly shaped contemporary metaethical theorizing. The positivist has an elegant answer. Our methods of reasoning about moral questions are individuated more nely than this objection assumes. Persistent moral disagreement shows that the parties accept dierent sets of rules to be the moral standard. You take the fact that A is an inten- tional killing of an innocent to count decisively against it. I agree in most cases, but allow exceptions. Our disagreement persists beyond all attempts to persuade. The positivist takes this situation to reveal that we accept dierent sets of rules as the moral standard. You accept an set of rules o Y that doesn't permit any intentional killing of an innocent. I accept an set of rules o 1 that does, in certain exceptional cases. We form our moral beliefs by relying on our sets of rules. So we have dierent ways of forming our moral beliefs. Because we have dierent methods for forming our moral beliefs, each of us can be perfectly reliable in our own method. And we can be perfectly reliable even while we continue to disagree. Moral knowledge doesn't depend on knowledge directly about the property being-the-moral-standard. So it just doesn't matter whether we're reliable in forming beliefs directly about that property. 25 That's the basic answer. Now let's consider an important objection. Can the positivist really explain the intelligibility of moral disagreement? She allows that dierent agents can be thinking and talking about dierent sets of rules! So they seem to be talking past each other. They don't really seem to be disagreeing, any more than we're disagreeing if I say I'm wearing green and you say I'm not. We're just talking about dierent people. Ditto for the moral case: it seems like the positivist predicts that we're not even talking about the same set of rules. The positivist will answer this objection by construing moral disagree- ment as disagreement about the realist's fundamental property being-the- moral-standard. And she has a robust license for optimism that this con- strual will capture all cases of moral disagreement. License-for-Optimism: the traditional realist's propositions are obvious apriori consequences of the positivist's propositions. For example, traditional realists associate you ought to keep this promise with the proposition that the moral standard requires keeping this promise. The positivist associates it with two propositions: that some set of rules o i requires keeping the moral standard and that o i is the moral standard. The traditionalist's proposition is an obvious apriori consequence of the posi- tivist's propositions, whatever set of rules o i happens to be. Since we're committed to obvious apriori consequences of our commitments, the pos- itivist agrees with traditional realists that ordinary agents are committed to the traditionalist's propositions about the moral standard. So the posi- tivist can be optimistic that she has the traditional realist's account of moral disagreement: that it's disagreement about the demands of the moral stan- dard. And that account of moral disagreement is widely recognized to be adequate. 9 Part III of this dissertation will defend this License for Optimism 9 There are lots of further questions to explore here. One is whether the license for optimism is born out: whether it works for every construction. Another is if one of the propositions being presupposed and not-at-issue creates further problems about dis- agreement. We shouldn't be too pessimistic about either point. Disagreement can be disagreement in presupposition. Imagine A saying: Bill started drinking at noon, and B saying: [no/ what you said is false] { Bill continued drinking at noon! It is dicult to explain why we can disagree about the presuppositions of the utterance. But the diculty of explaining it shouldn't distract from the fact that disagreement in presupposition is a genuine empirical phenomenon. Fully adjudicating the license for optimism takes us beyond the scope of this paper. I'm trying to convince you that we should focus on investigating whether moral discourse does carry substantial not-at-issue commitments. And guring out if the License for Optimism holds is one part of that investigation. 26 in more detail. Let's now turn to contemporary discussion of arguments from moral dis- agreement. Start with disagreement with an epistemic peer. (Epistemic peers are roughly those who you take to be as epistemically well-situated as you are.) Some philosophers hold that disagreement with a peer should reduce your condence in what you believe. These philosophers are called conciliationists. 10 Conciliationism is just irrelevant for the moral case, if the positivist is right. Fix on the case where two agents accept dierent sets of rules to be the moral standard. Conciliationists often take peer disagree- ment to be important as evidence that we're unreliable in forming beliefs in the relevant domain, and take this evidence to create rational pressure to conciliate. But we don't need to be reliable in discriminating what set of rules is the moral standard. As a result, there is no rational pressure to conciliate in the moral case. Now this point needn't be an objection to conciliationist views. In fact, it can be a way of defending conciliationist views. Some philosophers ob- ject that it's a mistake to conciliate in the moral case; a morally sensitive agent will stand her ground, at least sometimes. 11 (Disagreement shouldn't reduce your condence that, say, intentional killings of the innocent are wrong.) If the positivist is right, though, the conciliationist can accept this point without abandoning their claim. They can continue to insist that peer disagreement always matters for epistemic evaluation. But they can allow that the moral case is dierent, because of the distinctive structure of moral judgment. Certain parts of our moral commitments aren't to be evaluated epistemically. Since the conciliationist is only making a claim about the attitudes that are to be evaluated epistemically, those parts fall outside the scope of the claims the conciliationists are defending. This challenge from disagreement has also shaped constructive theorizing in contemporary metaethics. Kieran Setiya (2012) helpfully characterizes an important range of metaethical views as accepting a claim he calls Depen- dence. (Dependence) If ethical knowledge is possible, the facts in ethics are constitutively bound to us. (Setiya, 2012, 115) Dependence is plausible as a way of explaining why persistent disagreement 10 David Christensen (2007), Adam Elga (2007), and Richard Feldman (2007) gave early statements of this sort of view. 11 Kieran Setiya (2012) makes this claim; Katia Vavova (2014b) discusses the issues in helpful detail. 27 doesn't threaten our reliability. Even if we persistently disagree about some questions, we're close enough to the truth on most of them, because the facts about them are constitutively bound to us. 12 Theorists who accept Dependence all do explain why our methods for forming moral beliefs are reliable. 13 ButDependence carries substantial philosophical costs, as Setiya notes. Views that vindicate it threaten to predict an implausible amount of con- vergence in judgments about what we ought to do. If the facts are constitu- tively bound to us, it's hard to see how we can collectively be too far from the facts. But this claim makes it harder to explain the full range of moral disagreement. Moral disagreement is pervasive and persistent, at least in some cases. That's exactly the point that Sextus Empiricus and Montaigne were concerned to emphasize. (Setiya develops this point in more detail. He has his own proposal to vindicate Dependence, without predicting nearly as much convergence as the alternatives.) Setiya thinks that moral epistemology is pretty hard. It requires balanc- ing on the edge of a knife: we risk either unreliability in moral judgment, or implausible predictions of convergence. Things are much easier for the positivist. Persistent moral disagreement shows only that we're accepting dierent sets of rules as the moral standard. No puzzles about reliability here. We can each be perfectly reliable about our own sets of rules. And that's the only kind of reliability that matters for moral knowledge. The positivist has a crisp and novel answer to a classical challenge about our reliability. Moreover, contemporary metaethicists have often recognized the force of this classical challenge, and taken on striking and costly commit- ments to answer it. The positivist avoids those costs. It's much less costly to vindicate our ordinary moral knowledge than others have assumed. 12 This sort of answer is given by constructivists like Sharon Street (2008a). Depen- dence is true for her because we've constructed the facts in ethics. But realists also do accept Dependence. Ralph Wedgwood (2007), for example, links concept-possession to dispositions to conform to the basic principles of rationality appropriate to that concept. On his view, you have the concept only if you're disposed to conform to those basic prin- ciples. Cornell realists like Richard Boyd (1988) and David Brink (1989) take our modes of moral inquiry to x the normative properties that our thought and language describe. And `moral functionalists' like Frank Jackson (1998b) take platitudes about morality to x those normative properties. Not all metaethicists accept Dependence: for example, non-naturalists like Derek Part (2011a) and T. M. Scanlon (2014) don't. 13 They're reliable for a constructivist because our methods for forming moral beliefs determine what it is to be a virtue. And they're reliable for an externalist because our beliefs wouldn't be about what's virtuous unless we had reliable enough means for forming true beliefs. 28 1.4 Global Skepticism III: Harman's dispensability argument This section turns to the nal case study. This case study is about Gilbert Harman's argument from the dispensability of moral explanations. Harman contrasts moral explanations with scientic explanations. Sup- pose a scientist observes some phenomenon: a vapor trail in a cloud cham- ber. She infers that there is a proton moving through the cloud chamber. In making this inference, she can come to know that there is a proton mov- ing through there. Harman suggests that this inference supports knowledge only because the best explanation of her observation is that there is a proton there. He argues that the moral case does not pattern in the same way. If I see someone lighting a cat on re, I take myself to observe that they are doing something wrong. The best explanation of my making this observa- tion does not appeal to moral facts. The best explanation instead appeals to facts about my psychology. Observational evidence plays a part in science it does not appear to play in ethics, because scientic principles can be justied ulti- mately by their role in explaining observations...by their explana- tory role. Apparently, moral principles cannot be justied in the same way. It appears to be true that there can be no explanatory chain between moral principles and particular observings in the way that there can be such a chain between scientic principles and particular observings. Conceived as an explanatory theory, morality, unlike science, seems to be cut o from observation. (Harman, 1977, 9) This argument is an argument for global skepticism. It applies to Jeerson Davis' moral beliefs as much as it applies to mine. 14 Positivist realism disarms this argument. The standards for justication are just radically dierent if the positivist is right than if more traditional moral realists are right. The only sort of belief involved in moral belief is belief about your own set of rules. So in order to be justied in some moral belief, you only have to be justied in what you believe about your own set of rules. You also need to accept that those rules are the moral standard { but acceptance is dierent from belief, and is not normed in the way that belief is. So we should expect Harman's argument to fail against the positivist. Harman only intends his argument to apply to morality realistically con- 14 See Sturgeon (1988) for an in uential discussion of this argument. 29 strued { that is, construed as about properties that are independent of and more fundamental than our individual evaluative states. Even though the positivist does construe morality as about that sort of property, she denies that moral knowledge requires knowledge about that property. In eect, then, the standards for moral justication are the standards that a relativist or a constructivist imposes. The justication of an agent's moral beliefs depends on whether her beliefs about her own sets of rules are justied. Inasmuch as you expect relativists or constructivists to vindicate moral jus- tication, you should have the same expectation of the positivist. For ex- ample, someone who thinks that the relativism that Harman (1975, 1996) has defended avoids skeptical conclusions should also see the positivist as avoiding skeptical consequences. Now there are important similarities between Harman's dispensability argument and the recent wave of \debunking" arguments against moral re- alists, of the sort that Sharon Street (2006) in uentially pioneered. 15 The next chapter will take them up. But we should anticipate that the positivist will disarm a central class of those debunking arguments. In particular, the positivist will disarm those versions of the debunking argument that threaten to lead global skepticism in the way that Harman's challenge threatened to. We've seen enough to see how the positivist will disarm them too. Global moral skepticism entails that disjunctions like (9) are false. (9) Either Jeerson Davis or Frederick Douglass know what they believe about slavery. Since disjunctions like (9) are true if the positivist is right, positivist realism answers global skeptics. Now some philosophers may want to raise other problems even while they concede that (9) is true. For example, they might argue that it's not psychologically possible to accept a set of rules as the moral standard once we recognize the evolutionary backstory. And this argument might undermine moral knowledge for those who are aware of it. After all, knowing that slavery is wrong involves accepting a set of rules as the moral standard, for the positivist. If you don't accept a set of rules as the moral standard, you don't know that slavery is wrong. Now, importantly, this argument does not 15 For dierent versions of this general kind of challenge, see Matt Bedke (2009, 2014), Joshua Greene (2008), Richard Joyce (2006a,b), Phillip Kitcher (2007), Michael Ruse (1998), Michael Ruse and E. O. Willson (1986), and Sharon Street (2006). For discussion of the dierences between these challenges, see Russ Shafer-Landau (2012) and Katia Vavova (2014a). 30 threaten (9). It's just a fact that Davis and Douglass accepted their own sets of rules as the moral standard. That fact is in the past, and it won't change. Moreover, you can accept (9) without accepting that some set of rules is the moral standard. 16 So arguments about what's psychologically possible are just irrelevant for evaluating global moral skepticism. Any global skeptic has to deny that (9) is true. Since (9) is true if the positivist is right, global skepticism is false if the positivist is right. Now philosophers who make debunking arguments tend to think that there's something really wrong with realists who retain moral beliefs after recognizing their evolutionary origin. 17 They're likely to think that the positivist's vindication of (9) doesn't answer all their concerns. For the purposes in this chapter, though, it just doesn't matter one way or another if they're right. I'm concerned to show that a philosophically interesting picture (global moral skepticism) is incompatible with a highly plausible linguistic thesis, about the presuppositions of moral discourse. A wide range of important arguments in moral epistemology are arguments for global skepticism, in- cluding Harman's argument from this section. There are denitely inter- esting arguments in moral epistemology that aren't arguments for global skepticism. And some debunkers may well intend to make such arguments. Inasmuch as they do, they fall outside of the scope of this paper. 16 The presupposition triggered from the left disjunct is incompatible with the presup- position triggered from the right disjunct. And when that happens, the presuppositions are locally accommodated { they are interpreted as embedded in each disjunct. (See Mandy Simons (2000) for further discussion.) (10) is one example. (10) Either Jane knows that she won, or she's upset that she lost. At-issue: (Jane won and she knows it), or (Jane lost and she's upset that she lost.) (9) receives the same kind of gloss. At-issue: Either (oj is the moral standard and JD knows that oj forbids slavery and accepts that oj is the moral standard), or (o f is the moral standard and FD knows thato f permits slavery and accepts that o f is the moral standard) So (9) doesn't require accepting a set of rules as the moral standard. 17 See Street (2008b) for a particularly clear articulation of this conviction. 31 1.5 Wrapping up I have in eect been arguing that global moral skeptics are worried about a pseudo-problem. The bar for moral knowledge is much lower than they expect it to be. Or at least the bar for moral knowledge is much lower if positivist realism is true, and there is good linguistic evidence that positivist realism is true. I want to close by situating the positivist's account in the broader philo- sophical landscape, by comparing it with the contextualism that Stewart Cohen (1999) and Keith DeRose (2002) have defended. There are some im- portant similarities between the two kinds of approaches, though it takes a bit of background to appreciate why. The background is about an impor- tant and general fact about presupposition. Utterances of it stopped raining presuppose that it used to be raining. And presuppositions normally project from embeddings. If an utterance of S presupposes p, utterances of p:Sq normally do too. But I can also intelligibly say because it never rained today, it never stopped raining, either. The second sentence is interpretable because the presupposition is locally accommodated: that is, it's interpreted as oc- curring under the scope of the negation. The second half of that sentence communicates that it's not true that (it was raining and isn't now). Remember that the positivist builds her picture around a claim about presupposition. She holds that the realist's property is a presupposition. But presuppositions can be locally accommodated. So the positivist has to predict that there are some contexts where the presupposition is locally accommodated, including contexts where we attribute knowledge. (11) Jane knows that killing is usually wrong. At-issue: Jane knows that (o 1 usually forbids killing and o 1 is the moral standard) And any context where the presupposition is locally accommodated is a context where moral knowledge requires knowledge about the fundamental realist property. And all the standard skeptical arguments do apply to this sort of knowledge. So the positivist has to acknowledge that there are some special contexts where the skeptical challenge could still arise. Acknowledging this point makes the positivist look like a contextualist about moral knowledge. In particular, she looks like a contextualist who holds that there are only two kinds of stakes in moral contexts: low stakes contexts, where moral knowledge comes really cheap, and high stakes con- 32 texts, where it doesn't. Does anything distinguish the positivist from this kind of contextualist? Yes { there are at least two important dierences. Less importantly, the positivist is motivated by a highly conservative linguistic thesis, about the presuppositions of moral utterances. This conjecture explains why the bar for moral knowledge is so low in some contexts, and why there are only two kinds of contexts. There are only two kinds, because the contexts where the presupposition isn't locally accommodated under knows are all low stakes contexts. More importantly, though, positivist realism vindicates moral knowledge in a much more stable way. The presuppositions of an utterance aren't lo- cally accommodated when the speaker herself accepts those presup- positions. Imagine a speaker who accepts that o 1 is the moral standard, and who we know to accept o 1 as the moral standard. If she says I know that killing is usually wrong, the only reading of her sentence is the one where the presupposition is globally accommodated. In other words, the only reading of her utterance is the low-stakes one. This fact distinguishes the positivist from the two-stakes contextualist. For the positivist, psycho- logical facts about what the speaker accepts can determine that the agent does have moral knowledge. Positivist realism in general has this upshot. If the positivist is right, moral epistemology proper isn't very interesting. Once we nd ourselves accepting propositions about the moral standard, the resulting beliefs will have whatever epistemic goodies we could hope they would have. So general arguments for moral skepticism, of the sort discussed in this paper, are just posing pseudo-problems. The interesting questions are psychological questions. That is, they are questions about under what conditions it's psychologically possible for agents to stably accept substantive propositions about the moral standard. So if the positivist's linguistic thesis is right, we should reorient metaeth- ical debates around those questions. We should focus on understanding the psychological role of moral attitudes in our lives together. The task of the realist is not to vindicate the epistemic credentials of those attitudes. Those attitudes have their epistemic credentials as long as soon as we have the relevant psychological states. The realist's task is rather to make our moral attitudes jointly stable { to show how it's psychologically possible for us to continue having those attitudes. 33 Chapter 2: Moral debunking as linguistic mistake This chapter focuses on the debunking arguments that philosophers like Richard Joyce (2006a,b) and Sharon Street (3006, 2008a, 2008b, ms) have made. 1 Many such arguments can work only by establishing the sort of global skepticism that we've encountered in the last section. In part, then, this chapter constitutes a fourth case study, in addition to the three from the previous chapter. It's another illustration of just how systematically positivist realism disarms skeptical arguments. But this case study will be more involved. There is much more controversy about just what conclusion debunking arguments aim at establishing. I'll rst argue that many kinds of debunking arguments do aim at estab- lishing global skepticism, and show how the positivist systematically disarms those arguments. Then we'll consider whether there is any way to reformu- late the debunker's argument in a way that is forceful against the positivist. I'll suggest that there isn't. 2.1 Debunking, introduced The debunker starts with the observation that our moral attitudes aorded our ancestors an adaptive advantage. Believing that there's a moral reason to care for your ospring will be more adaptive than believing the opposite. So we can explain why many agents tend to judge that there is a reason to care for your ospring. The debunker then argues that we can explain moral judgments independently of the truths about our reasons. 2 Debunkers argue 1 For other versions of this general kind of challenge, see Joshua Greene (2008), Phillip Kitcher (2007), Michael Ruse (1998), and Michael Ruse and E. O. Willson (2986). For discussion of the dierences between these challenges, see Russ Shafer-Landau (2012) and Katia Vavova (2014a). 2 More careful versions of the challenge start with an evolutionary explanation of dispo- sitions to form certain kinds of beliefs, or perhaps proto-evaluative attitudes. But we don't lose any important detail by focusing on the moral beliefs formed by those dispositions. Sharon Street (2006) helpfully discusses this point (219-20). For further discussion of the scientic issues here, see William FitzPatrick (2014), Benjamin Fraser (2014), Richard 34 that moral realists should take this fact to undercut moral knowledge. The heart of the global debunkers' challenge is the claim that evolu- tionary explanation of moral belief patterns with other kinds of explanation that undercut knowledge. If someone hypnotized me into having some be- liefs, say, without caring whether those beliefs are true, those beliefs won't amount to knowledge. Not even if they're true; the explanation undercuts knowledge. The hypnotist doesn't even need to be malicious. It's enough if she's indierent to the truth. Debunkers hold that evolution is importantly like an indierent hypnotist. Evolution doesn't \care" about lining up with the realist's fundamental properties, any more than our imagined hypnotist cares about the truth of the beliefs implanted. The task for the debunker is to make good on this metaphor { to show that the two kinds of cases do pattern together. 3 Any debunker who tries to make good on this metaphor has to be as- suming the Standard Assumption. (Standard Assumption) If moral realism is true, moral knowl- edge requires knowledge about a property that is independent of and more fundamental than our individual evaluative attitudes. The debunkers from the last paragraph rely on general strictures on what it takes to have to knowledge, motivated by cases like the indierent hypno- tist. They argue from those strictures to the conclusion that evolutionary explanations of of moral belief prevent us from knowing about the realist's fundamental property. But if theStandardAssumption is false, we should expect those strictures to be irrelevant. Distinguish two goals a debunker might have. They are targeting non- skeptical moral realists, who take many ordinary agents to know what they ought to do. The debunker can challenge the realist's claim that lots of ordinary people have true beliefs about what they ought to do. I'll call this the alethic challenge. Or she can challenge the realist's claim that lots of ordinary people have whatever features distinguish knowledge from mere true belief. I'll call this second challenge the epistemic challenge. The epistemic challenge itself divides in several dierent challenges. It might Joyce (2013), and Guy Kahane (2011). 3 For dierent versions of this general kind of challenge, see Matt Bedke (2009, 2014), Joshua Greene (2008), Richard Joyce (2006a,b), Phillip Kitcher (2007), Michael Ruse (1998), Michael Ruse and E. O. Willson (2986), and Sharon Street (2006). For discussion of the dierences between these challenges, see Russ Shafer-Landau (2012) and Katia Vavova (2014a). 35 threaten those who are aware of the facts of evolution. Or it might threaten those who aren't aware. I will understand the epistemic challenge very broadly, and the alethic challenge very narrowly. Take my belief that intentional killings of the in- nocent are usually wrong. If you think the alethic challenge applies to that belief, you are thinking that that belief is false. Moreover, that's the only way to make the alethic challenge. For example, claiming that there is no reason to expect that belief to be true is not a way of making that challenge. See this point as a terminological stipulation. Given my stipulation, the alethic challenge just can't apply to someone who believes that p and also to someone who believes that not-p. If you're interested in a challenge that applies to both, your challenge is epistemic, not alethic. As a result, only philosophers with a particular normative axe to grind can intelligibly make the alethic challenge. For example, it would be intelligible for a consequen- tialists like Peter Singer to try to make the alethic challenge. 4 But many philosophers who make debunking arguments do not make them in support of a particular normative theory. Philosophers like Street or Joyce have more global aspirations. They want to show that there is something wrong with the beliefs of both parties to a normative dispute, like over consequentialism. I understand these philosophers as issuing the epistemic challenge. In fact, the only way to think of global debunkers is as issuing the global challenge. For one thing, it's the challenge that you have to make in drawing analogies with indierent hypnotists. After all, those cases only motivate conclusions about the features that distinguish knowledge from mere true belief. More importantly, though, anyone who wants to make a global challenge to all our moral beliefs has to be issuing the epistemic challenge. Fix on a normative disagreement about logical contradictories, like: (1a) Jeerson Davis believes that slavery is permissible. (1b) I, by contrast, believe that slavery is not permissible. In this case, what Davis believes is true i what I believe is false. 5 So one of us has to have a true belief. Now the global debunker thinks that her 4 Now these philosophers may think of themselves as issuing the epistemic challenge { as challenging the epistemic credentials of the relevant intuition. For my purposes, though, the important point is that they're the only philosophers who can issue issue the alethic challenge. Philosophers who can oer the alethic challenge thus include Roger Crisp (2006) (on whether we value accomplishment independently of pleasure), Tim Mulgan (2006) (on the experience machine), and Peter Singer (2005) (about the scope of benecence). 5 This claim is false given some recharch e views about the logic of moral beliefs. It'd be 36 challenge applies to Davis and me both. As a result, whatever challenge she intends has to apply to someone who has true beliefs, in virtue of her ambitions to articulate a challenge to my beliefs as much as to Davis'. That is, her ambitions force her to be making the epistemic challenge, not the alethic challenge. 6 false, for example, if (1)'s complement is neither true nor false. I don't think that global debunkers want to assume these sorts of recharch e views in developing their challenge. So it's legitimate to set those sorts of views aside. 6 Sharon Street and Katia Vavova (2015) are both particularly clear about the fact that the global debunker is an epistemic challenge. Kieran Setiya (2012) discusses some related points in illuminating detail (69 { 84). This characterization of the debunker's goals is important for some responses to the debunker. Walter Sinnott-Armstrong (2006) Roger White (2010) both suggest that evolutionary explanations fail to vindicate our moral faculty, rather than showing that it's positively unreliable. But this response is somewhat less telling against the debunker who emphasizes the analogy with the hypnotist. Discovering that your beliefs are formed by a hypnotist undercuts your justication for them, even if you don't regard the hypnotist as positively unreliable. It's worth noting, too, that this interpretation of global debunking ts the literature, as well. Sharon Street, for example, characterizes the realist as pushed towards an epis- temic conclusion: \the realist is forced to embrace a skeptical conclusion { acknowledging that our normative judgments are in all likelihood hopelessly o track, having been fundamentally shaped in their content by forces that bear no relation to the independent normative truth" (Street, 2008b, 208). Richard Joyce characterizes the upshot of one of his arguments as \the discovery that our moral beliefs are products of a process that is entirely independent of their truth, which forces the recognition that we have no grounds one way or the other for maintaining these beliefs" (Joyce, 2006a, 145). 7 Matt Bedke is similar: \Alignment of ethical intuition and belief, on the one hand, and ethical fact, on the other, requires that we posit a cosmically unlikely coincidence between the two. In other domains, this pattern supplies a defeater" (Bedke, 2009, 199). And Guy Kahane interpret both Joyce and Street as trying to show that \none of our evaluative beliefs is justied" (Kahane, 2011, 115), and Erik Wielenberg (2010) gives the same kind of interpretation (458, 462). Ramon Das (2016) has a further discussion of these issues. We can tell that the debunkers are interested in the epistemic challenge more indirectly, too, by their reactions to realist responses. Some realists have tried to answer the alethic challenge. They explain how we have developed true beliefs about what we ought to do. David Copp (2008), for example, has shown how his distinctive metaethical account explains how we would develop true beliefs about what we ought to do. The reaction to his work suggests that debunkers are interested in more than just the alethic challenge. The responses usually grant Copp his claims, but insist that he has missed their point. Thus Street: But merely gathering together, pruning and systematizing our substantive normative views does nothing to save the alleged explanation from being trivially question-begging { any more so than gathering together, pruning and systematizing ones substantive views about Jupiters size, surface, num- ber of moons, and so on saves the analogous explanation from being trivially question-begging in the case where one has learned that those views were implanted in one by a hypnotist who drew them out of a hat. (Street, 2008b, 37 This paper will focus on the epistemic challenge. Now there is signicant controversy about what assumptions are necessary to drive the epistemic challenge. 8 At this point, though, we've seen enough to get the avor of the debunker's challenge. They think that evolution is somehow like an indierent hypnotist. Having acquired the avor of the challenge, though, it's easy to lose track of the challenge by getting bogged down in all the dierent ways of making it more precise. I don't want that to happen. So I'll give my distinctive answer to epistemic debunking rst, and only afterwards work through all the dierent ways of making the idea more precise. 2.2 The positivist answers epistemic debunking The last chapter introduced a picture where the Standard Assumption is false. (Standard Assumption) If moral realism is true, moral knowl- edge requires knowledge about the property being-the-moral-standard. We should expect the debunker's epistemic argument to fail if that picture 215-6) The fact that my belief was implanted by a hypnotist doesn't show that the belief is false. But the fact that the hypnotist implanted the beliefs aects whether those beliefs are justied. It then looks like Street intends a challenge about justication, or about knowledge, or something like that, rather than a challenge about truth. She is interested, in other words, in the epistemic and not the alethic challenge. 8 Katia Vavova (2014a) has an especially clear discussion of this point. Maybe our beliefs need to be safe in order to be knowledge? (We know that S only if we'd retain the belief that S across a certain range of cases.) It is tempting, after all, to diagnose the hypnotist in this way. The beliefs that the hypnotist implants are not safe, because the hypnotist could have easily implanted other beliefs. But subtle issues come up when trying to make good on this thought. It's hard to explain why the evolutionary explanation of our moral beliefs prevents them from being safe. (Justin Clarke-Doane (2012) articulates some of the problems that come up.) Perhaps, then, the debunker should appeal to some other feature of knowledge, and try to show that evolutionary in uences are incompatible with that feature. One way to respond to the debunker's puzzle is to wind through all the subtle epistemic claims that the debunker might be making, and show that none of them are plausible. Now my response will be even more direct than the so called `third factor' accounts, as developed in dierent ways by David Enoch (2010), Knut Skarsaune (2011), and Erik Wielenberg (2010). Those philosophers are, I think, conceding too much to the debunker. My approach will also more direct than accounts that appeal to what purport to be conceptual truths about morality, like those that Terrence Cuneo and Russ Shafer-Landau (2014) and Katia Vavova (2014a) develop. 38 is right. Given that picture, it just doesn't matter if our attitudes towards the realist's fundamental property have those epistemic features. This section makes good on this high level expectation. It shows in detail how the epistemic challenge fails if the Standard Assumption is false. Suppose I assert that Jane knows that she ought to keep this promise. The positivist takes me to be doing two things in making that assertion. I'm asserting that Jane knows that o 1 requires keeping this promise and accepts that o 1 is the moral standard. I'm accepting that o 1 is the moral standard In giving this account, the positivist is holding that moral knowledge only ever depends on knowledge de re about the set of rules that is the moral stan- dard. This is a change from traditional moral realism, where moral knowl- edge depends on knowledge about the property being-the-moral-standard. This feature of traditional moral realism is what makes it vulnerable to epistemic debunking. Max Barkhausen has recently developed a represen- tative debunking argument that serves to illustrate this point. He claims that \there are nearby worlds where our evolutionary history took slightly dierent turns and we arrived at radically dierent moral views using the same cognitive faculties that we used in the actual world" (Barkhausen, 2016, 681). At those worlds, we would have had false beliefs about the prop- erty being-the-moral-standard if we continued to form our moral beliefs in the same way that we actually do. For example, maybe we form our moral beliefs by a mix of relying on what seems intuitively correct, plus some def- erence to others, plus other parts of the method of re ective equilibrium. If we had evolved to have dierent intuitions, then the same method would produce dierent beliefs. Barkhausen takes his claim to show that we don't know, because our beliefs aren't safe enough to amount to knowledge. 9 This argument is a paradigmatic epistemic challenge. It tries to show that even those people who have landed on the moral standard don't have any moral knowledge. 2.2.1 The basic answer No argument like this one will work against the positivist. The argument assumes that ordinary moral knowledge depends knowledge about the prop- erty being-the-moral-standard, and the positivist denies that assumption. 9 He relies on this claim in the course of arguing that the debunking challenge arises for moral naturalists as much as moral non-naturalists. 39 The only kind of knowledge the positivist requires for ordinary moral knowl- edge is knowledge de re about o 1 . Up to this point, though, I haven't said much knowledge de re about a set of rules. It's helpful to have an articial example in mind, to illustrate how that knowledge would be possible. Suppose that Jane found a book describing the considerations that s 1 is sensitive to: for example, that s 1 takes the fact that I've promised to to count in favor of-ing being required by s 1 , that s 1 takes that consideration to be undercut if I can save a life by breaking the promise. Suppose further that she memorized that book. Because she has memorized that set of rules, she knows what s 1 demands. Now explicit memorization is an articial example of how we could come to know about that set of rules. A full development of the positivist's idea will build from a more realistic picture. Maybe we know about our own set of rules mostly because we've been habituated into the way that our social group does things, and maybe also because we've modied that set of rules in our own idiosyncratic ways. That habituation gives us knowledge about our set of rules. (This idea makes the label \positivist" realism particularly appropriate.) If the epistemic challenge is to have any force against the positivist, it needs to show that Jane doesn't have knowledge de re about her set of rules o 1 . To gure out if Jane has that de re knowledge, we ask if she has true beliefs about o 1 across all the nearby worlds where she continues to form her beliefs about o 1 in the same way. For concreteness, use the articial assumption that we memorize sets of rules. The worlds where we form our beliefs in the same way are all worlds where we're relying on what we memorized. The fact that Jane has memorized the set of rules is what explains why she has true beliefs about o 1 across all those worlds. As a result, she does have de re knowledge about o 1 . I've used the articial example of memorization to illustrate the impor- tant shift that distinguishes the positivist from more traditional kinds of realists. The general point is that the positivist's de re knowledge is sen- sitive to a narrower range of nearby worlds than Barkhausen cares about. The positivist only considers worlds where we form our moral beliefs in the same way we actually do (e.g. by relying on what we've actually memo- rized). Barkhausen considers more worlds. Suppose, for concreteness, that we form our moral beliefs largely by memorizing what our community thinks. Barkhausen would consider worlds where our community believes very dif- ferent things, as among the worlds that help determine whether we have moral knowledge. Since we would have false beliefs at least at some worlds where our community had very dierent beliefs, we don't know about the 40 property being-the-moral-standard. But those worlds are totally irrelevant for the positivist. The positivist only cares about the way that Jane actually forms her beliefs, by relying on the actual thing she memorized. And Barkhausen's other worlds are worlds where she's forming her beliefs in a dierent way, because she's memorized something dierent. As long as Jane is forming her beliefs by relying on what she actually memorized, she'll be forming beliefs about the same set of rules, o 1 . This assumption is plausible for memorization in general. Compare my memorizing a poem. I know that poem only if the range of worlds where I rely on what I actually memorized are worlds where I have true beliefs about the poem. We do not consider worlds where I've memorized a dierent poem or no poem at all. The same goes for Jane's de re knowledge. We hold xed her ways of forming beliefs, by relying on what she actually memorized. Given that way of forming her beliefs, she will continue to have true beliefs about o 1 . That is the basic positivist answer to epistemic debunking. The only kind of knowledge necessary for ordinary moral knowledge is de re knowledge about a particular set of rules. The considerations that debunkers highlight do not threaten that de re knowledge. However, this basic positivist answer needs to be unpacked in one more stage, before it's fully adequate. Nothing about this explanation of our moral beliefs guarantees that those beliefs will line up with the moral stan- dard. And the debunker might insist that our moral knowledge depends on that sort of guarantee. The debunker might develop this challenge by in- sisting that the positivist has to acknowledge the existence of (conceptually possible) worlds where we continue to form beliefs about o 1 in the way we actually do, but where o 1 is not the moral standard. (Sharon Street insists on this kind of point: \as a purely conceptual matter, these independent normative truths might be anything" (Street, 2008b, 208). 10 ) But we don't know what we ought to do at those worlds where o 1 isn't the moral standard. So our beliefs are not safe, after all: there are some worlds where they're false. The positivist denies that our ordinary moral knowledge depends on a modally robust guarantee that the set of rules we accept to be the moral standard is in fact the moral standard. The positivist will spot the objector that there are (conceptually possible) worlds where Jane is forming beliefs about o 1 but where o 1 is not the moral standard. But those worlds are no 10 Justin Clarke-Doane (2012) has a helpful discussion of the importance of this point in debunking arguments { see pp. 320. 41 threat to the positivist's vindication of our ordinary moral knowledge. Even at those worlds, Jane still has de re knowledge about o 1 . Now she doesn't haven't any moral knowledge at those worlds; it's false that Jane knows that she ought to keep this promise there. In order for her to know, there needs to be some set of rules that (i) she accepts to be the moral standard and (ii) is in fact the moral standard. No such set of rules exists at the counterfactual world we're envisioning. In other words, moral knowledge takes an additional ingredient, in addition to de re knowledge of o 1 . It also depends on acceptance of what's in fact the moral standard as the moral standard. That extra ingredient is missing at the counterfactual world. It's thus important to distinguish two ingredients in the positivist's con- ception of moral knowledge: knowledge de re of some set of rules, plus acceptance of what's in fact the moral standard as the moral standard. But the two ingredients are crucially independent of each other. You can have the rst without having the second. Nearby worlds where dierent sets of rules are the moral standard might be important for testing for the rst ingredient. In asking whether Jane continues to have true beliefs about o 1 across those counterfactual worlds, we're exploring whether she in fact has the de re knowledge that is the rst ingredient of moral knowledge. I've been explaining why her beliefs about o 1 are safe across that range of worlds, to defend the positivist's claim that she does have de re knowledge about o 1 . But those counterfactual worlds are irrelevant for the second ingredi- ent, acceptance of what's in fact the moral standard as the moral standard. Moral knowledge at the actual world depends only on her having the second ingredient of moral knowledge at the actual world. It's totally irrelevant whether she also has that ingredient at counterfactual worlds. The posi- tivist proposes a radical shift in our theorizing about moral knowledge: to this two-ingredient account of moral knowledge. This same answer works for any version of the epistemic challenge. Take the challenge that there is no reason to think that our moral beliefs line up with the moral standard. I suggested earlier that this challenge is a species of the epistemic challenge. So we should expect the positivist to answer it. And indeed she does. Note that this challenge entails that our beliefs are unlikely to be true. So we can evaluate this challenge by asking whether we can know that our beliefs are likely to be true. Now the positivist holds that judgments about likelihood are judgments about our own set of rules. She gives the following gloss on asserting (3). (3) It's likely that both I ought to keep this promise and that I ought to tell the truth now. 42 I'm asserting that it's likely that o 1 requires keeping this promise and that o 1 requires telling the truth now. I'm accepting that o 1 is the moral standard And the speaker can know what she would be asserting here is true. She can know how her set of rules classies both actions. If she does, she knows that it's likely that it classies them in that way. Once we've vindicated her rst-order knowledge, we can also explain why she's perfectly entitled to believe that her beliefs are likely to be true. Now this response assumes that moral vocabulary embeds under likely in the way just illustrated. But this assumption follows straightforwardly from the presuppositional conjecture defended inx3. Presuppositions in general embed under likely in just that way. 11 11 Does the same strategy generalize to iterated knowledge ascriptions, like (i)? (i) I know that Jane knows that she ought to keep this promise. You might reasonably worry about these examples. The positivist takes Jane's knowledge to depend on her actually having landed on the moral standard. Then it looks like (i) is true only if I myself know that she has landed on the moral standard. And that's exactly the sort of knowledge that the debunker is interested in undermining. So the present response does not succeed in full generality: it doesn't extend to examples like (i). (Moreover, there may be downstream problems if the positivist doesn't vindicate (i) while trying to vindicate Jane's ordinary moral knowledge.) Fortunately, though, the positivist has a simple and elegant explanation of iterated knowledge ascriptions. Iterated knowledge ascriptions are like rst-order knowledge, and only require knowledge de re about the salient set of rules. In asserting (i), I am: asserting that I know that Jane knows that o1 requires keeping this problem and accepts that o1 is the moral standard, and accepting that o1 is the moral standard. So the truth of (i) does not depend on my knowing that o1 is the moral standard; it only depends on my accepting that it is. This piece of the positivist's proposal is independent of what I've said before; it's an additional feature that I haven't highlighted until this foot- note. Complications like this one are one reason it's important to explain the positivist's account as following from one simple linguistic claim. Even the complicated suggestion in this footnote turn out to ow straightforwardly from that one claim, rather than being ad hoc. These iterated examples raise another issue. The account of (i) requires me and Jane to accept the same set of rules as the moral standard. But if Jane and I don't agree about killing in self-defense, there is no complete set of rules that we both accept to be the moral standard. Importantly, though, (i) can be true as long as Jane to know when she ought to keep her promises. It doesn't also require her to know about other moral questions, 43 Now you might think that the debunker should issue a dierent chal- lenge: that evolutionary explanations show that there's something wrong with accepting these presuppositions.x3 will take up that challenge. Up to this point, though, I've been arguing that all the standard ways of pressing the debunkers' challenge fail. Those standard ways all make the Standard Assumption. And debunking fails without that Assumption. (Standard Assumption) If moral realism is true, moral knowl- edge requires knowledge about the property being-the-moral-standard. This section has made good on this claim. It has shown in detail how epistemic debunking fails without the Standard Assumption. 2.2.2 The positivist's answer in a broader context It's helpful to see the positivist as exploiting the same loophole in epistemic debunking that a constructivist like Sharon Street (2008a) tries to exploit. We know what we ought to do, for a constructivist, because our judgments about what we ought to do are about something that we've made up. The positivist can use the same idea. We latch onto the particular set of rules that we do because that's the set of rules that models the expectations that we've made up for each other. Our constructing the expectations are what explains our de re knowledge; the positivist can just take over the construc- tivist's account of this point. But the positivist also disagrees with pure constructivists like Street, because the positivist insists that there are inde- pendent truths about what we ought to do. Those truths are truths about the moral standard. The positivist just denies that our moral knowledge de- pends on our knowing those independent truths. What's important is that our beliefs line up with that independent truth, not whether we know that they line up. That's why the positivist's account does not answer the alethic challenge. The positivist only explains how to vindicate the ordinary moral knowledge of those who have landed on the moral standard. Now moral realists think of themselves as having important advantages over Humean constructivists like Street. For one thing, they think that they can explain moral disagreement in a way that those constructivists cannot. like when she can kill in self-defense. There is an initial problem for the positivist in explaining how that is possible. To see how the positivist would capture this example, note that nothing in the positivist's basic approach requires sets of rules to be complete in classifying every action. The sets of rules that matter may be partial, in only classifying promise-keeping while being silent on self-defense. (i) can then incorporate a partial set of rules that is silent on self-defense but opinionated about promise-keeping. 44 (Imagine two people who've constructed dierent things. There aren't any independent facts for them to be disagreeing about, beyond what they've constructed.) The positivist has the very same advantage that the realist claims. Traditional realists can explain disagreement because they think that moral disagreement is always disagreement about the same property, being- the-moral-standard. When you say Mary ought to keep that promise, and I disagree, we're accepting inconsistent propositions about what the moral standard demands. Importantly, though, the positivist has the same explanation of disagree- ment. After all, those traditional realist propositions are obvious apriori consequences of the propositions that the positivist associates with the ut- terances. (The traditional realist associates Mary ought to keep that promise with the proposition that the moral standard requires Mary's keeping that promise. The positivist associates that sentence with the proposition that o 1 requires Mary's keeping the promise and the proposition that o 1 is the moral standard. The traditional realist proposition is an obvious apriori consequence of those two propositions.) Since speakers are committed to obvious apriori consequences of what they accept, they're committed to the exact same propositions that traditional realists use to explain disagreement. The positivist does have the realist's account of disagreement. 12 In general, then, the positivist denies that ordinary moral knowledge depends on knowledge about the property being-the-moral-standard. So a fortiori it doesn't depend on safe or reliable beliefs about that property. That is what distinguishes positivist realism from more traditional kinds of moral realism. Traditional moral realists hold that ordinary moral knowledge is knowledge about the property being-the-moral-standard ordinary moral knowledge requires reliable (or safe) belief about that property ordinary agents lack moral knowledge if evolutionary history would have taken dierent turns at nearby worlds moral disagreement is always disagreement about de dicto propositions about the moral standard 12 A full evaluation of this point would take an entire paper. This paragraph is indicat- ing the fundamental license for optimism that the positivist has the realist's account of disagreement. Making good on this license for optimism would require working through a range of cases there isn't space to consider here. 45 The positivist realist holds that ordinary moral knowledge is knowledge de re about the moral standard ordinary moral knowledge does not require reliable (or safe) belief about whatever is the moral standard ordinary moral knowledge only requires reliable or safe be- lief about the actual moral standard ordinary agents have moral knowledge even if evolutionary history would have taken dierent turns at nearby worlds moral disagreement arises because the traditional realist's propositions are obvious apriori consequences of the propo- sitions that the positivist associates with each moral judg- ment This departure from traditional moral realism is how the positivist disarms epistemic debunking. 2.3 Does evolution undermine acceptance? I've argued that epistemic debunking requires the Standard Assumption. (Standard Assumption) If moral realism is true, moral knowl- edge requires knowledge about the property being-the-moral-standard. I've introduced a picture where the Standard Assumption is false. And I've shown that this picture crisply answers the epistemic challenge. But you might think that this picture only pushes the debunker's chal- lenge back a stage. Debunkers are interested in our attitude towards the realist's fundamental property. And my answer to the debunker is that moral knowledge requires a less demanding attitude towards that property than the debunkers assume. It only requires acceptance, not knowledge. And debunkers might argue that their arguments also show that there is something wrong with accepting propositions about the realist's fundamen- tal property. This objection is connected with an odd feature of my presen- tation. The paper opened by suggesting that debunkers are really interested in a pseudo-problem. But the discussion in the last section didn't look like it was addressing a pseudo-problem. It looked like it was giving a straight answer to the debunker. 46 I agree that debunkers are likely to nd the answer in the last section un- satisfactory. And I agree that they will want to reformulate their argument as an argument about acceptance. But their nding that answer unsatis- factory is what reveals that they're interested in a pseudo-problem. This section explains why. Remember that the basic gloss on acceptance is Stalnakerian: \to accept a proposition is to treat it as true for some reason. One ignores, at least temporarily, and perhaps in a limited context, the possibility that it is false" (Stalnaker, 2002, 716). Importantly, though, knowledge is not the norm for acceptance. It's perfectly legitimate to accept things that you acknowledge that you don't know. Now the positivist holds that ordinary moral belief involves accepting a particular set of rules as the moral standard. The de- bunker who wants to challenge this ingredient of moral knowledge needs to argue that there's something wrong with doing that. The debunker might argue that it's irrational to accept substantive propositions about the moral standard. But no argument for this conclusion could work. The norms on acceptance aren't stringent enough to sustain it. For example, one way for the debunker to make this argument is to argue that anyone who recognizes an evolutionary explanation of her moral beliefs is rationally required to accept that her set of rules is no more likely to be the moral standard than some other set of rules. There is, they might insist, no reason to think that evolution pushed us towards the moral standard. And we should reason by the principle of indierence, and conclude that any two sets of rules are equally likely to be the moral standard. No plausible norm on acceptance can support this argument. Consider a purely linguistic case. You assume that someone stole the tarts. I myself think it's unlikely that the tarts were stolen. (I think it's much more likely that you or your colleague absent-mindedly misplaced them, and are trying to nd a victim to blame.) But I can still accept that someone did steal the tarts, if I want to have a cooperative conversation with you. Maybe I am still interested in guring out who would have stolen the tarts if anyone did, so I'm happy enough to go on with your assumption in this conversation. I still don't believe that somebody did. I'm just accepting it. I can do that and still obey all the plausible norms on acceptance. In fact, we should see what I'm doing in this case as constraining what norms are plausible for acceptance. However, we can reason by the principle of indierence in the linguistic case. That reasoning might lead us to accept that three possibilities are equally likely: (i) that you misplaced the tarts, (ii) that your colleague did, and (iii) that they were stolen. And if you accept that these three possibilities 47 are equally likely, then you don't accept that any particular one happened. (They're mutually exclusive.) So if the principle of indierence is a norm on acceptance, you can't be rational in accepting that someone stole the tarts. But this conclusion is a reductio. You can be rational in accepting that someone stole the tarts in the envisaged context. The debunker is appealing to a norm on acceptance that is too demanding to be plausible. In fact, something more general is true. It's hard to see any norm on acceptance that allows me to be rational in the tart case that is still stringent enough to do what the debunker needs. The debunker might then take a take a dierent tack. She might deny that it's psychologically possible to continue accepting substantive proposi- tions about the moral standard, once you recognize the evolutionary expla- nation of those states. Now there are lots of dierent ways for the positivist to answer this challenge. One option is to follow P. F. Strawson (1974) and hold that reactive attitudes like resentment and indignation are an in- dispensable part of human life. She could then add that moral beliefs are necessary for the stability of those attitudes. 13 And she could conclude that creatures like us will continue to have moral beliefs because of their connec- tion with our ineliminable reactive attitudes. Now the positivist also holds moral beliefs involve accepting particular sets of rules as the moral standard. For her, then, this Strawsonian line is an answer to the psychological chal- lenge. It explains how it's psychologically possible for us continue accepting substantive propositions about the moral standard, even when we recognize the evolutionary explanation of our psychological states. Of course, there are other ways to answer this psychological challenge, too { other ways of articulating the role of moral judgment in our social or individual life. But note that positivist realism is essential to the payos of these answers, in the present context. The Strawsonian line is plausi- ble only in establishing a psychological point: that our moral beliefs are linked to an ineliminable part of our psychology. In general, though, this sort of psychological point doesn't illuminate epistemological questions. A propensity to make certain probabilistic errors also appears to be an inelim- inable part of our psychology. 14 And that observation doesn't show that that propensity supports knowledge; more generally, the fact that some process is ineliminable simply doesn't tell us about its epistemic credentials. This 13 For discussion and defense of this point, see Joseph Butler (1729), John Rawls (1971), and R. Jay Wallace (1998). Moral beliefs might only be necessary for the rationality of our reactive attitudes { but belief that a reactive attitude is irrational tends to destabilize the attitude. 14 See, for example, Kahneman and Tversky 1974Kahneman and Tversky (1974). 48 Strawsonian point is important within the positivist's framework only be- cause the positivist has already vindicated the epistemic credentials of moral belief. She has shown how those beliefs can amount to knowledge. We've taken up the Strawsonian point only to address one last desperate ploy on behalf of the debunker. Given the positivist's framework, the debunker is articulating a pseudo- problem. There are only three ways of understanding what problem animates her. The debunker can try to show that evolutionary explanations of moral belief prevent us from knowing about our own sets of rules. But that chal- lenge is unpromising, asx2 showed at length. Alternatively, she can try to show that we're irrational in accepting substantive propositions about the moral standard. This challenge assumes rational strictures on acceptance that are too demanding to be plausible. Finally, the debunker can try to show that it's not psychologically possible to accept substantive propositions about the moral standard, once we recognize the salient facts about evolu- tion. And this attempt is implausible, in light of the psychological role of moral belief. 2.4 Wrapping up The positivist disarms a central class of evolutionary debunking arguments. At the same time, though, the positivist's answer works in a very unexpected way. Consider a comparison. Alexius Meinong (1904) argued that we need to posit non-existent objects to explain some central problems. Bertrand Russell (1905) famously argued that those problems were pseudo-problems { that they resulted from misunderstanding the structure of ordinary lan- guage. The positivist similarly holds that global debunkers are posing a pseudo-problem. And her argument is very much like Russell's. It holds that their problem results from misunderstanding the structure of ordinary language. Now we haven't had space here to fully argue for the positivist's core linguistic claim, any more than \On Denoting" fully argued for Russell's account of denite descriptions. I've rather focused on establishing that that linguistic claim is a central and unappreciated pivot point for problems in metaethics. Those problems turn out to be pseudo-problems if that linguistic claim is correct. Now some moral realists may also nd the positivist's answer unsatis- fying. But if the positivist is right, those realists are also worried about a pseudo-problem. Realists should instead focus on developing views that t with the best work in normative ethics, on the one hand, and on the best 49 work on the metaphysics of fundamentality, on the other. 15 That's it. Posi- tivist realism thus has similar upshots to the de ationary non-naturalism of philosophers like T. M. Scanlon. 16 The positivist agrees with them about the comparative centrality of rst-order normative theorizing, and the compar- ative marginality of general epistemology. The positivist diers from those de ationary non-naturalists in her argument for this point. She argues for it from a linguistic conjecture, rather than a conjecture about the metaphysics of value. In fact, though, the contrast between the positivist and the de ationary non-naturalist is even more pointed. Some philosophers seem to not even understand why the non-naturalist's metaphysical conjecture calls out for a de ationary attitude. 17 By contrast, it's perfectly clear why the positivist's linguistic claim has de ationary upshots. So the positivist may well have a better claim to rationalize a de ationary attitude towards moral epistemol- ogy than the traditional non-naturalists do. 15 The latter might include work by Kit Fine (1994), David Lewis (1983), Gideon Rosen (2010), and Ted Sider (2014). 16 A more extensive list of philosophers with sympathies for this kind of de ationary non-naturalism might include W. D. Ross (1930), Thomas Nagel (1986), Ronald Dworkin (1996), Jean Hampton (1998), Derek Part (2011b), and T. M. Scanlon (1998, 2014). 17 See Barry Maguire (2014) and Ralph Wedgwood (2016) for related discussion. 50 Chapter 3: Moral naturalism without naturalized moral epistemology The overarching goal of Part I in this dissertation is to argue that it would nice for non-skeptical moral realists if positivist realism is true. We've seen in particular that positivist realism is incompatible with global moral skepti- cism. And we've explored four case studies that illustrate how the positivist disarms arguments for global skepticism. These upshots of positivist realism are perfectly general. They're helpful for any moral realist. But it turns out that positivist realism is helpful in another, further way for one particular class of moral realists: moral naturalists. Moral naturalists try to locate moral properties within the natural world. They do that by claiming that moral properties bear some metaphysically central relation to natural properties { that is, properties that gure in some natural science. Some naturalists will take the metaphysically central rela- tion to be the relation of being-the-real-denition-of. Those naturalists will hold that the real denition of moral properties includes only natural prop- erties. Other naturalists take the metaphysically central relation to be the relation of being-reducible-to, and others the relation of being-composed-of, and others will favor other relations. 1 The naturalist's claim has been taken 1 I intend to be ecumenical in my characterization of naturalists, including both reduc- tive and non-reductive naturalists. I will assume that there is some genuine distinction between these naturalists and self-conceived non-naturalists, though I won't try to char- acterize the dierence in detail. I'll use the copula in describing naturalist views, and be open to dierent interpretations of the copula. Tristram McPherson has developed one attractive conception of this dierence: Reductive Naturalism There are actual elite normative properties, and the real def- inition of every elite [roughly, joint-carving] normative property is an elite (non-identity) function of only natural properties. Non-reductive Naturalism There are actual elite normative properties that are members of the naturalistic similarity class. All elite normative properties are either members of this class, or satisfy Reductive Naturalism. (McPherson, 2015, 139) The moral naturalist adopts one of these frameworks, and is trying to t moral properties 51 to play a number of theoretical roles: metaphysical, semantic, epistemic, and alethic. Metaphysical: the claim illuminates the fundamental structure of the world { Supervenience: the claim explains why moral properties super- vene on natural properties. (Brown (2011); Dreier (1992); Dun- away (2015); Jackson (1998b); Ridge (2007); Schroeder (2005)) { Location: the claim locates moral properties in the natural world. (Boyd (1988); Brandt (1979); Brink (1989); Copp (2007); Foot (2001); Hursthouse (1999); Jackson (1998b); Railton (1986); Schroeder (2007); Sturgeon (2006)) Semantic: the claim illuminates our use of moral language { Reference: the claim explains how moral terms manage to refer to moral properties. (Boyd (1988); Brink (2001); Copp (2000); Dow- ell (2016); Dunaway and McPherson (2016); Jackson (1998b)) (Maybe, for example, natural properties are reference magnets.) { Disagreement: the claim explains how moral disagreement is in- telligible. Disagreeing parties are disagreeing about the same nat- ural property. 2 Epistemic: the claim explains how true moral beliefs can also have whatever features distinguish mere true belief from knowledge. { Vindicatory: the claim helps explain how ordinary agents know what they ought to do. (Boyd (1988); Brink (1989); Copp (2007); Jackson (1998b); Hursthouse (1999); Railton (1986); Schroeder (2007); Sturgeon (2006)) { Methodological: the claim has important upshots for the way that normative ethics is done. (Boyd (1988); Foot (2001); Hurst- house (1999); Railton (1986); Sturgeon (2006); Thomson (1997); Thompson (2012)) Alethic: the claim explains how we came to have true moral beliefs within it. 2 This advantage is a standard enough advantage of realist views over anti-realist views that it tends not to be foregrounded in discussions of naturalism. 52 { (Boyd (1988); Brink (1989); Copp (2007); Jackson (1998b); Hurst- house (1999); Railton (1986); Schroeder (2007); Nicholas Stur- geon (2006)) That's a lot of dierent theoretical roles! The positivist has a novel account of the nal two roles: the epistemic and the alethic. She crisply vindicates ordinary moral knowledge, in a way that any naturalist can accept. I focus on developing this new account because I want to make naturalist views more defensible. In general, the more work a view needs to accomplish, the more likely it is to fail. That is, there are purely general reasons to expect that moral naturalism is more defensible if it doesn't require a distinctive account of the nal two roles. But there are also more particular reasons. Much of the criticism of moral naturalism concerns its epistemology. A range of philosophers reject naturalist views because they think that they force implausible conclusions about moral knowledge. Positivist realism smoothly disarms these criticisms. That thesis thus makes moral naturalism far more plausible than contemporary philosophers tend to assume. Moral naturalism doesn't lead to any objectionable con- clusions about moral epistemology; it doesn't require us to naturalize moral epistemology. 3.1 A baseline about the naturalist's distinctive commitments My thesis should look very surprising. Of course moral naturalism will lead to naturalized moral epistemology! (Where else could it lead?) My thesis is surprising because a Standard Assumption looks very compelling. (Standard Assumption) If moral naturalism is true, moral knowledge requires knowledge of a natural property. And this Standard Assumption is itself highly plausible. For example, it follows if moral properties are the semantic values of moral terms. Jill knows that murder is morally wrong then expresses that Jill has knowledge about a natural property. Unfortunately, though, if the Standard Assumption is correct, it's not possible to do what I'll try to do in this paper. The ambition of this paper is in eect to pry apart the last two theoretical roles from the rst two roles: to vindicate ordinary moral knowledge, without being constrained by the naturalist's metaphysical claims. The Standard Assumption will prevent me from prying apart the theoretical roles in the 53 way that I want to do. Given that assumption, moral naturalism constrains moral epistemology. I'll argue that the naturalist is better o without the Standard As- sumption than with it. In the course of making that argument, I'll intro- duce and defend a picture where the Standard Assumption is false. But to appreciate the signicance of this picture, we need to start with a baseline. We need to appreciate how things look with the Standard Assumption. That's what this section does. Then we can better appreciate why it's worth leaving the baseline behind. The baseline is about diculties that arise for naturalists in explaining several distinctive hallmarks of moral knowledge. One central hallmark of moral inquiry is that you can draw substantive moral conclusions by a priori reasoning. For example: if you think about the nature of pain long enough, you can come to know that the fact that some action in icts pain is a moral reason for not doing it. But we can't come to know substantive conclusions about natural properties by a priori reasoning. To learn about a natural property, you have to interact with the natural world. Nicholas Laskowski and Stephen Finlay have suggested that this kind of consideration has moved many philosophers away from moral naturalism: \over recent years, however, the popularity of this kind of naturalism in metaethics seems to have faded. ... A primary problem is the already-mentioned and generally acknowledged intuitionist or apriori epistemology of a robust range of normative truths, which is a primary motivation for Moore and other primitivists' denials that normative properties could be `naturalistic' " (Laskowski and Finlay 2017, 6). 3 Another closely related hallmark is that ethical inquiry seems to be au- tonomous from scientic inquiry. Ethical inquiry is legitimate whether or not it ends up lining up with methods of reasoning in any natural science. For example, Russ Shafer-Landau rejects naturalist views, because \ethics is an autonomous domain of inquiry; ethical concerns are not reducible with- out remainder to any other kind of concern, and ethical investigation is not properly conceived of as a subset of any other recognized discipline" (Shafer-Landau, 2003, 52). This complaint is similar to the complaint about aprioricity, in part because both explicitly target the naturalist's conclu- sions about moral knowledge. They try to show that those conclusions are implausible. Attempting to naturalize moral epistemology also constrains which kinds of naturalistic views are plausible. For example, Mark Schroeder (2007) has 3 Charles Pigden (2012) is also helpfully clear on this point. 54 argued for a Humean view about normative reasons as the best way for a naturalist to vindicate ordinary knowledge. In particular, he argues that that Humean view is in the best position to explain how our methods of forming beliefs about our normative reasons reliably leads us to the truth. Such a view explains how the kind of psychological state which is likely to lead us to form beliefs about reasons counts as having a content about reasons precisely because its genesis is structurally well enough connected up with the conditions under which there ac- tually are such reasons. So when you want a cup of coee, and your belief that there is coee in the lounge is true, and so are your background beliefs that the coee in the lounge is not being guarded by someone who will keep you from getting any if you go there, then it is in fact true that this fact is a reason for you to go to the lounge. And so forming beliefs about your reasons in this way can lead you to true beliefs about your reasons. (Schroeder, 2007, 174) And lots of philosophers nd a Humean conception of normative reasons independently objectionable. They worry that it leads to an objectionable conclusion. In particular, they worry that it either leads to an error theory (as J. L. Mackie (1977) or Richard Joyce (2001) suggest), or to relativism (as Gilbert Harman (1975) suggests), or to a picture where moral norms apply without giving people normative reasons (as Philippa Foot (1972) suggested at one point.) Though Schroeder tries to avoid those consequences, he acknowledges that his framework may in the end lead to them. 4 It is harder for some other naturalists to vindicate our normative knowl- edge. A prominent naturalist competitor to Humean views takes moral prop- erties to be broadly consequentialist properties. 5 And it is notoriously di- cult for such consequentialists to vindicate ordinary moral knowledge. (As James Lenman (2000) has argued, ordinary agents are too clueless about the consequences to have much moral knowledge, if consequentialists are right.) Our belief that ordinary agents do have a lot of moral knowledge thus con- strains which naturalist views we take seriously. And some philosophers nd these constraints too severe. Non-naturalism looks like the only option that makes good on the full range of their substantive convictions. 4 See pp. 120-1. 5 Richard Boyd (1988), David Brink (1989), David Copp (2007), Nicholas Sturgeon (2006), and Peter Railton (1986) all fall broadly in this tradition. 55 So if naturalists need to naturalize moral epistemology, they are likely to run into two kinds of costs. First, it is it hard for them to vindicate some of the central hallmarks of moral knowledge: that a priori reasoning can sup- port moral knowledge, and that moral inquiry is autonomous from scientic inquiry. Second, this need substantially constrains what naturalist views to take seriously. And the views left on the table may be independently ob- jectionable. A Humean conception of normative reasons is a good example. Trying to vindicate ordinary moral knowledge can back the naturalist into a corner, and force her into accepting costly theses. 3.2 Back to the Standard Assumption Moral naturalists claim that moral properties are natural properties. This claim usually plays several dierent kinds of theoretical roles. Metaphysical: the claim illuminates the fundamental structure of the world Semantic: the claim illuminates our use of moral language Epistemic: the claim explains how true moral beliefs can also have whatever features distinguish mere true belief from knowledge. Alethic: the claim explains how we came to have true moral beliefs And the overarching goal of this paper is to pry the rst two roles apart from the second two roles. If I'm right, moral naturalism constrains and helps answer metaphysical and semantic questions. But it does not constrain our approach to epistemic or alethic questions. I care about accomplishing this goal because I want make moral naturalism more defensible. And many philosophers have found it indefensible because of what it predicts about the epistemic questions. If positivist realism is true, moral naturalism just doesn't made the predictions that other philosophers have found indefensible. 3.2.1 Central hallmarks We saw earlier that some philosophers argue against the naturalist's epis- temology from central hallmarks of moral knowledge. They try to show that naturalists can't explain those hallmarks, and conclude that moral nat- uralism is implausible. One central hallmark is the aprioricity of moral knowledge { that is, that you can come to know substantive moral claims 56 just by a priori reasoning. For example: if you think about the nature of pain long enough, you can come to know that the fact that some action in icts pain is a moral reason for not doing it. And there is a simple and powerful argument that naturalists can't to explain this hallmark. A priori reasoning doesn't deliver substantive truths about the world. In order to know substantive natural facts, you have to interact with the natural world. Note rst that the positivist disarms this sort of argument. For the positivist, moral knowledge doesn't require any sort of knowledge about the fundamental naturalist property. So it just doesn't matter whether we can know about that property in an a priori way. Moral knowledge can be a priori if we can know about the demands of a set of rules a priori. And the argument from the last paragraph doesn't tell us anything about whether we can know about the demands of of a set of rules in an a priori way. Moreover, the positivist also has the resources to give a constructive explanation of the aprioricity of moral knowledge. We can come to know about a set of rules in the same kind of way that we acquire logical or mathematical concepts. I can learn the rules for the material conditional `É' by memorizing the rules governing it from a logic book. But once I've come to be competent with those rules, I can know substantive logical truths. For example, I can then know that if p and (p É q), then q. And I can know that fact a priori { by my understanding of the logical connective. Now it's important for this point that the knowledge ascription used rather than mentioned the connective. Knowledge that the sentence \p and (pÉ q), then q" is true isn't plausibly a priori, because you can't know a priori what `É' means. The positivist will oer the same account of knowledge of a set of rules like s 1 . You somehow come to understand that set of rules. In my articial example, you come to understand with them by memorizing some book that describes them. (More plausibly, you come to understand them by being habituated into them.) Once you understand s 1 , you can know a priori what s 1 forbids, permits, or requires. As in the logical case, it's important to distinguish knowledge of the set of rules themselves from knowledge of the set of rules in the book. The latter kind of knowledge is available only a posteriori. It requires knowledge that s 1 is the set of rules in the book, and that knowledge is only available a posteriori. (Similarly, knowledge \p and (pÉ q), then q" is true is available only a posteriori, because it requires knowledge that `É' meansÉ.) Having made this distinction, we can see how you can know about a set of rules in an a priori. You can know about them a priori because you understand them, in the way that you can know logical truths because you understand the logical connectives. The etiology of our 57 understanding doesn't prevent that understanding from supporting apriori knowledge, any more than it does in the logic case. It's comparatively easy for the positivist to vindicate the aprioricity of moral knowledge. It's comparatively easy because she doesn't need to ex- plain how a priori reasoning delivers substantive knowledge of a natural property. It only needs to deliver knowledge of a set of rules, and it's much easier to do that. Other philosophers argue against moral naturalism by appeal to what they take to be another hallmark of moral inquiry. They hold that moral inquiry is autonomous from the natural sciences. For example, Russ Shafer- Landau claims that \ethics is an autonomous domain of inquiry; ethical concerns are not reducible without remainder to any other kind of concern, and ethical investigation is not properly conceived of as a subset of any other recognized discipline" (Shafer-Landau, 2003, 52). The positivist is well- positioned to vindicate this hallmark. Moral inquiry is autonomous because it is an investigation into the structure of the set of rules that we accept to be the moral standard. Now that inquiry extends our moral knowledge only if ends up aligning what we accept more closely with a natural property. But this requirement doesn't constraint the methods that we use when doing moral inquiry. It's perfectly legitimate to reason from the armchair, without engagement with any other recognized discipline. The positivist thus predicts that moral knowledge is very dierent from scientic knowledge, even if moral properties are identical to properties stud- ied in some science. To appreciate the positivist's distinctive contribution, it's helpful to work through the dierences in more detail. 3.2.2 Contrasts between moral knowledge and scientic knowl- edge Ordinary moral knowledge is quite dierent from scientic knowledge about that very same property, if the positivist is right. Now this claim is very ab- stract. To make it more concrete, I'll x on one particular kind of naturalist: a consequentialist naturalist who reduces moral truths to truths about what maximize overall well-being. The positivist predicts that moral knowledge still comes much easier than scientic knowledge about well-being. It's hard to know what maximizes overall well-being; sociologists and economists have to expend a lot of work to gain that knowledge! And the reasons why it's hard for them to gain that knowledge appears to have important upshots for the autonomy of morality. But it takes a signicant amount of work to fully appreciate why. Let's start by noting how hard it 58 is to know about the consequences of individual actions. As James Lenman (2000) has noted, many of our actions are identity-aecting { they aect which people happen to exist. Identity-aecting actions have signicant eects. The person who otherwise wouldn't have existed could do a great deal of good, or a great deal of harm. Since the good or bad done is part of the total consequences of our identity-aecting actions, those consequences are part of what determines whether the action was right or wrong. Identity- aecting actions are a particularly vivid illustration of an important general point. Our actions ramify in ways that nite creatures like us aren't in a position to appreciate. And we're usually unable to discriminate these cases. As Lenman notes, this point threatens to undermine a great deal of or- dinary moral knowledge if consequentialists are right. Now to make this point precise, we need to situate it in a picture about knowledge in general. A natural picture holds that knowledge requires safe belief. Williamson's formulation is characteristic: \if one knows, one could not easily have been wrong in a similar case" (Williamson, 2000, 147). 6 The idea is to compare similar cases where you still form your beliefs in the way that you actually form them. A belief is safe if all those similar cases are cases where your belief is still true. Safety illuminates the structure of Lenman's reasoning. Take an act of scapegoating. Ordinary agents will continue to have the same beliefs about the wrongness of scapegoating whether this act of scapegoat- ing leads to the birth of some mass murder or prevents that birth. If act consequentialism is true, ordinary agents could easily been incorrect about this act of scapegoating in similar cases. For example, ordinary agents would still think that this act of scapegoating is wrong even in a case where the scapegoating prevents the birth of a terrible mass murderer. A similar point holds for the consequences of other things, besides indi- vidual acts. Take embeddings of rules { that is, social arrangements where the rules are widely or universally followed, or accepted, or internalized, by some or all or most of the community. Dierent (embeddings of) rules would have dierent consequences if something cataclysmic happens { if, for exam- ple, the globe warms past a certain point. (Theorists in the social contract tradition are presumably right to assume moderate scarcity of resources in their theorizing, and to take this assumption to be important in the rules that would be chosen.) Now if something cataclysmic actually happens, it could aect the truth of our moral beliefs, and so their amounting to knowl- edge. But focus on cataclysmic events that don't happen, but that could 6 This formulation is a characterization of the structure of our reasoning about what is known, not something that that takes us from more basic primitives to conclusions about what is known. 59 easily happen, given what we know right now. Supposing that the earth actually warms n degrees, focus on worlds where it warms 2n or 5n degrees. Rules are likely to have dierent consequences at those worlds. This point helps explain why sociologists and economists have to work so hard to gain their knowledge. The methods that they use need to produce safe beliefs about the properties they're investigating. And safe beliefs about these properties are appropriately attuned to the possibilities of cataclysmic events, like global warming far more catastrophic than actually occurs. So the scientists need to be relying on methods that are appropriately attuned to those possibilities. Consequentialist naturalists are thus an important example of the auton- omy concern that animates Shafer-Landau. If they're right, moral inquiry looks dependent on scientic inquiry. After all, moral inquiry aims at deliver- ing knowledge about a natural property. And the consequentialist naturalist should acknowledge that it's very hard to have safe beliefs about the natural property that's identical with that moral property. It seems like scientic methods are the only methods that will support that sort of safe belief. But then it looks like moral inquiry is properly conceived of as a subset of another recognized discipline, like sociology or economics { or at least its legitimacy depends on conceived of it in that way! The positivist disarms this autonomy concern. Moral inquiry is au- tonomous from scientic inquiry, because it does not need to deliver knowl- edge of a natural property. In fact, the positivist has a general and principled account of the dierence between moral knowledge and scientic knowledge. Let N be the natural property that constitutes being the moral standard. Moral knowledge doesn't require safe beliefs about that property N. By con- trast, scientic methods for coming to know about N do need to support safe belief about it. That's why moral inquiry does not depend for its le- gitimacy on vindication by scientic methods. The epistemic standards for moral inquiry are just very dierent from the epistemic standards for sci- entic inquiry, even though the fundamental moral property is a natural property. Scientific knowledge: is knowledge directly about the property being-N requires safe belief about N goes missing for agents who can't discriminate what has N in nearby worlds 60 Moral knowledge is knowledge de re about a set of rules that happens to have N does not require safe belief about whatever has N only requires safe belief about the set of rules that actually has N is still present for agents can't discriminate what has N in nearby worlds. It's much easier to gain moral knowledge than to gain the corresponding scientic knowledge. And it's easier because moral belief and knowledge has a distinctive hybrid structure that scientic knowledge does not. Scientic knowledge just involves knowledge of a single proposition about the property N. Moral knowledge, by contrast, involves a hybrid of knowing about a set of rules, plus accepting that those rules have the property N. More generally, the positivist has given a novel vindication of the method of re ective equilibrium. That is, she has to explain why it makes sense for normative ethicists to defend their theory by showing that it best ties our particular moral convictions into a coherent whole. The positivist interprets a normative ethicist who does that as trying to rene her set of rules into a more coherent whole. And it makes perfect sense to do that within the positivist's framework. Moreover, it makes perfect sense whether or not the method of re ective equilibrium delivers safe beliefs about a natural property { even if the fundamental moral property is a natural property. Moral inquiry is genuinely autonomous from scientic inquiry. 3.3 Questions about truth I've distinguished four dierent kinds of roles that the naturalist's charac- teristic claim might play. Metaphysical: the claim illuminates the fundamental structure of the world Semantic: the claim illuminates our use of moral language Epistemic: the claim explains how true moral beliefs can also have whatever features distinguish mere true belief from knowledge. Alethic: the claim explains how we came to have true moral beliefs 61 And the over-arching goal of this paper is to pry the rst two theoretical roles apart from the second two. The last section has shown how the natural- ist's claim doesn't need to play the third theoretical role. Moral knowledge doesn't require knowledge of the naturalist's fundamental property. It only requires a much more minimal attitude. So the naturalist's claim doesn't constrain the way we vindicate moral knowledge. This section turns to the fourth theoretical role that the naturalist's claim might play. Now it's very natural to expect that the same account should play both the third and the fourth theoretical roles. In explaining how a belief amounts to knowledge, you're explaining how it's not just a uke that it's true. (Maybe it's the result of a reliable process, for example.) And once you've explained how it's not just a uke, you've explained how we came to have true beliefs. So it's very natural to expect the positivist's account to explain how we ended up with true beliefs, if it is to explain how we know what we believe. However, this expectation is misguided. The positivist's account does not tell us anything about how we came to have true beliefs. Let's work through this point more slowly. The positivist is not a relativist. Moral knowledge does require landing on the moral standard. And she has a realist conception of the aims of normative theorizing. Normative theorizing aims at aligning our beliefs more closely with the moral standard. The positivist has given an account of how moral beliefs usually have whatever features distinguish knowledge from mere true belief. But this account is very striking and unusual. It doesn't also explain how we've come to have true beliefs. It just takes true beliefs as given. So the positivist's theory of what I called the epistemic role cannot also be an account of what I called the alethic role. Given everything I've said so far, then, it is still possible to argue for a particular naturalist view as playing the alethic role better than its rivals. In fact, we can repurpose arguments that are traditionally framed as being about the epistemic role in this way. Go back to Mark Schroeder's Humean view about normative reasons. As Schroeder notes, that view explains how the kind of psychological state which is likely to lead us to form beliefs about reasons counts as having a content about reasons precisely because its genesis is structurally well enough connected up with the conditions under which there actually are such reasons. (Schroeder, 2007, 174) He leverages this observation to argue that his Humean view is better posi- 62 tioned to vindicate ordinary knowledge than some rivals. However, he could also frame this observation as evidence that the Humean view is in a better position to explain how we have true beliefs. Then we could draw on our conviction that we do have true beliefs, as evidence for the Humean view and against other kinds of naturalism. This kind of argument is a dagger at the heart of the positivist's am- bitions. It suggests that the questions that animate traditional moral epis- temologists are still centrally important. The positivist has been trying to de ate the importance of those questions { to show that we can't use them to argue against moral naturalists. And it looks like we've just found a way of restating these questions, within the positivist's framework. But this impression does not stand up to further scrutiny. It is not possible to formulate this kind of argument in the positivist's framework. To see why, start by supposing that I accept a particular set of rules as the moral standard. We can ask why I believe what is true about the wrongness of unprovoked punches. Because presuppositions project from questions, we're thereby asking: At-issue: Why do I believe thats 1 forbids an unprovoked punch when s 1 does in fact forbid an unprovoked punch? Not-at-issue: s 1 is the moral standard To answer this question, we would list the considerations that s 1 takes to matter, and show that I believe that those considerations have the weight that they do. And that answer is equally available to all naturalists, what- ever their conception of the fundamental normative property. The positivist will generalize from this answer in explaining why I mostly have true beliefs. She will start by extending her answer to our initial question to my beliefs about killings in self-defense, promise-keeping, and so on. And she'll gen- eralize from those answers. Moreover, she will give a similar answer in the third-personal case. To explain why someone else has true beliefs, I list the considerations that my own set of rules s 1 takes to matter. And I show that the other person believes that those considerations have the weight they do. Of course, traditional moral epistemologists will nd this answer unsat- isfying. They want to know how we landed on the mind-independent truth about the moral standard. But for someone who accepts a particular set of rules as the moral standard, this question has an easy answer, as we just saw. We came to believe the considerations that s 1 takes to matter. And it's just a psychological fact about us, that we do accept that a particular set of rules as the moral standard. That's constitutive of having moral beliefs at all, on 63 the positivist's picture. Given that psychological fact, the traditional epis- temologist's question resists formulation within the positivist's framework. We can't use that question as a source of evidence in choosing among nat- uralist views, or in choosing between a naturalist view and a non-naturalist view. There is, however, a somewhat related kind of evidence that the positivist acknowledges to be genuine. Schroeder is convinced that it's more likely that we would have developed true normative beliefs if his Humean view is right than if some consequentialist view is right. If the positivist is right, though, this conviction is totally irrelevant for comparing Humean naturalism with consequentialist naturalism. But Schroeder might also argue that some of our normative beliefs are in fact false if the consequentialist is right but true if his Humean view is true. This argument provides genuine evidence for his Humean view, even in the positivist's framework. Ordinary moral belief involves accepting particular sets of rules as the moral standard. And it's irrational to accept incompatible propositions about the moral standard. If Schroeder can show that more of our normative beliefs are in fact false if the consequentialist is right, then he can conclude that it's less rational to accept consequentialist naturalism than to accept his Humean picture. Crucially, though, Schroeder would have started doing rst-order nor- mative ethics in making that argument. The argument tacitly draws on the method of re ective equilibrium { it tries to show that the Humean view ties our substantive moral convictions into a coherent whole more consistently than the consequentialist alternative. The positivist acknowledges that this argument would be compelling. But she still insists that it's a mistake to reject a naturalistic view because it makes it hard to see how we could have developed true beliefs. It is, however, perfectly correct to reject a naturalist view because it suggests that we have in fact developed a wide range of false beliefs. In other words: we can't discriminate among naturalist theories in their ability to make sense of the method of re ective equilibrium. Natu- ralists are all on a par in making sense of that method. But we can still distinguish between them by using the method. 3.4 Wrapping up Moral naturalists claim that fundamental moral properties are natural prop- erties. This claim has been taken to play a number of theoretical roles: metaphysical, semantic, alethic, and epistemic. Metaphysical: the claim illuminates the fundamental structure of 64 the world { Supervenience: the claim explains why moral properties super- vene on natural properties. { Location: the claim locates moral properties in the natural world. Semantic: the claim illuminates our use of moral language { Reference: the claim explains how moral terms manage to refer to moral properties. (On one prominent line, natural properties are reference magnets.) { Disagreement: the claim explains how moral disagreement is in- telligible. Disagreeing parties are disagreeing about the same nat- ural property. Epistemic: the claim explains how true moral beliefs can also have whatever features distinguish mere true belief from knowledge. Alethic: the claim explains how we came to have true moral beliefs My goal has been to pry the rst two theoretical roles apart from the sec- ond two roles. If I've succeeded, it's much easier to defend moral naturalism. It's a mistake to reject moral naturalism for its distinctive epistemic commit- ments. It just doesn't have distinctive epistemic commitments. Moreover, preferring one kind of naturalism to another for its epistemic commitments is also misguided. Naturalists are all on a par. If the positivist is right, naturalists should focus on developing views that t with the best work in normative ethics, on the one hand, and on the best work on the metaphysics of fundamentality, on the other. 7 The best work in general epistemology is just irrelevant. The last chapter compared posi- tivist realism with the de ationary non-naturalism of philosophers like T. M. Scanlon. 8 The positivist agrees with them about the comparative centrality of rst-order normative theorizing, and the comparative marginality of gen- eral epistemology. But the positivist defends this picture from a linguistic conjecture, rather than from a metaphysical claim. So she shows how moral naturalists can have the same de ationary attitude that Scanlon and others take. 7 The latter might include work by Kit Fine (1994), David Lewis (1983), Gideon Rosen (2010), and Ted Sider (2014). 8 A more extensive list of philosophers with sympathies for this kind of de ationary non-naturalism might include W. D. Ross (1930), Thomas Nagel (1986), Ronald Dworkin (1996), Jean Hampton (1998), Derek Part (2011b), and T. M. Scanlon (1998, 2014). 65 Chapter 4: De re belief At this point in the dissertation, I've introduced a view { positivist realism { and I've described how it would be nice for a range of non-skeptical realists if that view is true. This view builds from a particular linguistic claim about presupposition, which I haven't defended yet. Parts II and III of the dissertation will defend that linguistic claim. But the arguments in those parts are subtle, and involve complicated issues. So before turning to them, I want to try and convince you that that linguistic claim really is the crucial pivot in evaluating positivist realism. If that linguistic claim is true, positivist realism is true, and has the upshots that I've described. Now the positivist does have some other commitments, as well. For one thing, she has striking psychological commitments. She thinks that ordinary agents are thinking about particular sets of rules when they're thinking about what they ought to do. Positivist realism (1) Jane knows that she ought to keep this promise. (b) Jane knows that o 1 requires keeping this promise, and accepts that o 1 is the moral standard (where o 1 is the moral standard) This view contrasts with traditional kinds of moral realism. Traditional Kratzerianism (1) Jane knows that she ought to keep this promise. (a) Jane knows that [the x: x is the moral standard] ought x (she keeps that promise) So the positivist's picture is much more psychologically demanding than the tradi- tional alternative. Like the traditional picture, it requires ordinary agents to be in a position to think about the fundamental realist property. And like the traditional picture, it takes ordinary agents to be competent with modals. In addition to all that, though, the positivist also requires de re attitudes about some particular set of rules { including de re knowledge about that set of rules! And traditional realists don't need ordinary agents to have those de re attitudes. 66 This chapter aims at defending the positivist's psychological commitments. It explains why ordinary agents could think about particular sets of rules. The next chapter will add an explanation of explain how ordinary agents can also know about their sets of rules. The overarching goal here is to show that the linguistic claim discussed in the next Parts really is the crucial pivot. The positivist's other com- mitments are plausible enough that we should accept her view once we accept that linguistic claim. 4.1 Basic issues The positivist will take the sets of rules that model our moral beliefs to be quite complicated. A wide range of considerations can bear on the in iction of harm, say, and those considerations weigh together in subtle ways. Earlier chapters have used an articial heuristic to model an set of rules. They imagined agents memorizing a list of considerations, and how they weigh together. That act of memorization is how those agents are able to think about complicated sets of rules. This heuristic has three disadvantages. First, it's wildly articial. Acquiring moral knowledge is nothing like memorizing considerations in a book. Second, it rst makes moral knowledge too easy. To gure out what to do, you just look things up on a list you memorized. And nally, this heuristic also makes our moral beliefs more transparent than they actually are. If you've memorized a list of the relevant considerations, you're in a position to know immediately what those considerations are. Unfortunately, though, the considerations that matter are often not transparent to us. I think it's impermissible for the surgeon to cut up one to save ve, but I also think it's permissible to switch the trolley so that it only kills one rather than ve. I'm not sure why, just that there is some dierence. That case is hard to explain if my belief is the result of memorizing a list of considerations. My belief would then be the result of me calculating what my list says about these cases. And that calculation seems to require awareness of the considerations that the list takes to matter. This chapter gives the ocial account of de re attitudes about sets of rules, which replaces the articial heuristic of memorization. The goal is to explain how we can think about these sets of rules, without making our attitudes too easy to acquire, or too transparent. 4.1.1 A companion in innocence I pursue this goal by developing a companion in innocence. The companion in in- nocence will be the language that we each speak. Now talk about \language" is ambiguous. It can refer to whatever best ts my own production and interpreta- tion of sentences, or whatever best ts the way that my community produces and interprets sentences. It's the former and not the later that is my companion in innocence. To highlight this focus, I will talk about my idiolect, rather than my language, throughout what follows. 67 I will characterize idiolects formally: with a function from sentences and con- texts to propositions, or with something similar. If we think about idiolect this way, they have their syntactic and semantic properties essentially. (If you start with some idiolect and change the semantics for some particular expression, you've changed it into a dierent idiolect.) It is then contingent which of those idiolects that we each speak. I would have spoken dierent idiolects if I would have been born in a dierent place or at a dierent time. Idiolects are a natural companion in innocence for the positivist, because we seem to have special and privileged knowledge of our own idiolect. And the positivist will want to say that we have special and privileged knowledge of our own set of rules. I'll assume that I have special and privileged knowledge of my own idiolect ` m that Mary is happy means-in ` m that Mary is happy: I immediately know what it means-in ` m , in virtue of understanding my idiolect. In making this claim, I am making a claim about the three place relation \means-in": a three place relation between sentences in a context, idiolects, and propositions. 1 It's not a dyadic relation between sentences in a context and propositions. Now this sort of knowledge is not essential for semantic competence with your idiolect. For example, maybe you can speak a idiolect without being in a position to think about semantic concepts like means-in. And some philosophers are very interested in explaining our linguistic knowledge without us thinking about those sort of semantic concepts. 2 I am interested in a dierent question. I am focusing on people who already have the semantic concept means-in. I want to explore how they have special and privileged knowledge of what falls under that concept, in their own idiolect. The positivist will go on to claim that I have the same kind of special and privileged knowledge of whatever set of rules o m I accept to be the moral standard: for example, that it demands keeping some particular promise. This case is like our special and privileged knowledge of our own idiolect, in that both involve knowledge of something very complicated. And the same heuristic is initially useful in both cases. As a rst pass at capturing our privileged knowledge of our own idiolect, we can imagine that we've memorized a book mapping sentences in a context to propositions. Moreover, this heuristic has the same limitations in the linguistic case as in the moral case. Our linguistic knowledge sometimes comes harder than it would on this articial model, and is less transparent than it assumes. However, there is less reason to appeal to this heuristic in the linguistic case, because we have a better sense of what explains our linguistic knowledge. Each of us has tacit knowledge of a recursive phonology, a recursive syntax, and a recursive semantics. That's why we interpret particular sentences as meaning what they do. I'll claim that our tacit knowledge of our own set of rules is very much like our tacit knowledge of a recursive semantics. For my purposes, then, we need to be clear about what's plausibly involved in tacit knowledge of a recursive semantics. 1 Or, depending on your view of the semantics/ pragmatics interface, it might be a three place relation between sentences, idiolects, and something like propositional matrices. 2 Davis Lewis is one example, in his \Languages and Language" (1975). 68 I have tacit knowledge of what salad means in my idiolect: present me with an object, and I can usually tell you if it's in the extension of salad. That is, I have a set of dispositions to apply the word salad to some objects and not to others. There is some important relationship between this set of dispositions and my tacit knowledge of what salad means. Some philosophers will think that set of dispositions is what ordinarily grounds my semantic knowledge. Other philosophers will think that that set of dispositions is just a hallmark of this semantic knowledge: that is, in ordinary cases where I have this semantic knowledge, I'll also have this set of dispositions. The latter sort of philosopher is likely to emphasize that there are cases where you can communicate by using the word salad without having these sorts of dispositions. For example, you can use the word deferentially { maybe you intend to refer to talk about whatever your mom is talking about. I want to set those cases aside. Non-deferential uses of salad is my model for what the positivist will say. Some- one who uses the word salad non-deferentially has a set of dispositions to apply the word to some objects and to refuse its application to others. Those dispositions either ground her semantic knowledge, or reveal what sort of knowledge she has about the meaning of salad in her idiolect. To develop this modal, though, we need a technique that allows us to move from linguistic dispositions, to attitudes towards particular idiolects. I'll focus on one technique to answering this question. I'll propose that Ramsifying over someone's attitudes either grounds or reveals what idiolect they're speaking. David Lewis (1970) popularized the technique of Ramsifying as determining what individuals are thinking about. He focused on scientic theories, which he modeled as a set of open sentences. Someone who believes a scientic theory will accept a range of sentences from that theory. For example, a chemist might think that there is some property hydrogen and property oxygen such that8x 1 , y 1 , z 1 [(Hydrogen(x 1 ) ^ Hydrogen(y 1 )^ Oxygen(z 1 )) and x 1 , y 1 , z 1 compose)É they form a molecule of water]. To gure out what the scientist is thinking about, we replace the relevant theoretical terms (Hydrogen and Oxygen) with free variables: 8x 1 , y 1 , z 1 [(X(x 1 ) ^ X(y 1 )^ Y(z 1 )) and x 1 , y 1 , z 1 compose)É they form a molecule of water]. As a rst pass, then, the scientist is thinking of hydrogen and oxygen because those are the properties that make this open sentence true. (This suggestion is only a rst pass, because the open sentences aren't enough by themselves to determine which exact properties they're thinking about.) Now turn to the case of idiolects. I speak a idiolect ` m where Bob is happy and Mary is sad expresses-in ` m the proposition that Bob is happy and Mary is sad. I may have never considered that sentence before. But the claim is still plausible, because my dispositions either constrain or reveal the meanings of the sentence's constituents. For example, I'm disposed to interpret and as meaning that its two arguments are true. I will model this disposition somewhat articially. I will model it as involving a disposition to assent to a sentence about what and means-in my idiolect ` m . That is, the boxed term \` m " is the theoretical term (like Hydrogen) whose content is constrained by Ramsifying over this sentence: 69 I'm disposed to take: pS 1 and S 2 q to mean-in ` m [p <st> .q <st> .p is true and q is true](JS 1 K)(JS 2 K) And Ramsifying over this sentence constrains ` m to be a idiolect where and means [p <st> .q <st> .p is true and q is true]. It's reasonable to attribute that meaning to and in virtue of Ramsifying over my dispositions. That attribution is the one where I get the most right. Ramsifying can be interpreted in two ways. It can be interpreted as articulating the content of our beliefs about meaning, or it can be interpreted as a metasemantic explanation why our beliefs about meaning have the contents that they do. Use `F' for the conjunction of open sentences we're Ramsifying over. I know about my own idiolect that Mary is happy means in ` m that Mary is happy. The metasemantic interpretation of this knowledge involves de re thought about the idiolect that is F, while the content-level interpretation involves thought by description. (Content-level) I know that [the` x : F(` x )] (Mary is happy means- in ` x that Mary is happy) (Metasemantic) I know that Mary is happy means-in ` m that Mary is happy My attitudes are about` m and not some other idiolect, be- cause ` m is F. I'll adopt the metasemantic interpretation throughout what follows. Thought about my own idiolect is de re thought about some particular object, which we can charac- terize formally. Ramsifying is what reveals or explains which object we're thinking about; it bears the metasemantic question. It doesn't infect the content of our attitudes. 4.1.2 Turning to the moral case Earlier chapters have claimed that the only kind of knowledge required for moral knowledge is knowledge de re about your own set of rules. This claim follows from a more general picture, where moral attitudes involve de re thought about your own set of rules. The positivist will explain those de re attitudes by extending the Ramsifying strategy from the previous section. String together all the particular moral judgments that you're disposed to make at some particular moment in time. As a rst pass proposal, take the judgments to be about the set of rules that results from Ramsifying over those judgments, and think of Ramsifying as answering the metasemantic question. To spell out this idea, we need to explain what we're Ramsifying over. According to the positivist, moral judgments are a hybrid of two dierent kinds of attitudes: belief plus acceptance, say. Let's focus on the beliefs that a particular agent Jane is disposed to have about promise-keeping. I characterize my dispositions about and with open sentences, and then Ramied over them. I'll do the same thing here. 70 Belief: if you promised to meet your friend for lunch but can save a life if you don't, x requires saving the life. Acceptance: x is the moral standard Belief: In every other case where you promised to meet your friend for lunch, x requires you to keep the promise. Acceptance: x is the moral standard These dispositions either reveal or ground which set of rules you're thinking about. They play the same role that dispositions about the application conditions of salad play in either revealing or grounding what idiolect I speak. That's the idea that this section develops: take the resources introduced to capture some facts about idiolects, and use them on the positivist's behalf. I'll begin with a rst pass at a general recipe for doing what the positivist needs. That general recipe needs to take an arbitrary set of dispositions to believe, and tell us what set of rules they're thinking about. Take this recipe. Thetask: constructing an set of rules o i that models some individual's beliefs. If the agent believes that A is required to in C, then for any world w 1 where A does not in C, o i ranks some world where A does in C has higher than w 1 . If the agent judges A is forbidden to in C, then for any world w 1 where A does in C, o i ranks some world where A does not in C as higher than w 1 . What's required is modeled with higher-ranked worlds and what's forbidden is mod- eled with lower ranked worlds. (We don't need to add in a separate clause for what's permissible, at this point, as long as actions that are neither forbidden nor required are permitted.) The rest of this chapter focuses on turning this rst pass into something that's fully adequate. 4.2 Rening the rst pass There are several reasons that the recipe I just described is only a rst pass. I'll slowly introduce those reasons one by one, both in this chapter and in the next. (It's too overwhelming to do all at once.) 4.2.1 Sets of rules and orderings One important reason that this recipe is just a rst pass is that doesn't actually deliver what I promised. It doesn't tell us what set of rules the agent is thinking about, because it doesn't deliver quite enough structure to discriminate closely related sets of rules. 71 To appreciate this point, let's ll in some background. Fix on a class of relations R on pairs of possible worlds with two formal features. In order to be a member of this class, R i needs to be transitive { that is,8w 1 ,w 2 ,w 3 , if w 1 R i w 2 and w 2 R i w 3 , then w 1 R i w 3 . Second, it also needs to be re exive { that is,8w 1 , w 1 R i w 1 . I'll call relations that satisfy these two constraints Kratzerian orderings on worlds. Someone might try to use these Kratzerian orderings to give a semantics for modals. That person would hold that pMust(p)q is true relative to an ordering R i i all the worlds that R i ranks highest are worlds where p is true. That is, it's true i [8w 1 : [:9w 2 ] (w 2 R i w 1 ^w 1 6= w 2 )] (p is true at w 1 ). And pMay(p)q is true relative to R i i some world that R i ranks highest are worlds where p is true. That is, it's true i [9w 1 : [:9w 2 ] (w 2 R i w 1 ^w 1 6=w 2 )] (p is true at w 1 ). My rst-pass recipe only determines a Kratzerian ordering. And such orderings are dierent from sets of rules, because they have less structure. Think of a set of rules containing only one rule: a rule that forbids causing two painful headaches in all and only those case where you could only cause one. Compare that set with another set again containing only one rue: a rule that forbids causing an even number of headaches in all and only those case where you could only cause one. The two sets of rules determine the same Kratzerian ordering. Any world where the rst set forbids an action is a world where the second set also forbids it, and any world where the rst set permits an action is a world where the second set permits it too. Since Kratzerian orderings just are rankings on possible worlds, the two sets of rules determine the very same Kratzerian ordering. But it's natural to say that they're dierent sets of rules. My rst-pass suggestion thus doesn't deliver a set of rules that the agent is thinking about. It delivers something coarser-grained. One reaction to this point would be to go looking for another recipe that does better { a recipe that delivers sets of rules. But another response is to rene the positivist's idea in one way. Rather than taking moral discourse to be about sets of rules, take it to be about Kratzerian orderings. I introduced the positivist's view by characterizing it as a view about sets of rules. But I did that, because it's easier to what a set of rules is than it is to understand what a Kratzerian ordering is. By the positivist's lights, though, the choice between the two options is a theory-internal choice. And either option preserves the philosophical virtues that the positivist claims. The positivist's basic idea is that moral knowledge comes a lot easier than you might expect, because it doesn't require knowledge of the realist's fundamental property. In order for this idea to be plausible, though, the positivist needs to say what kind of knowledge moral knowledge does require. After all, I know that killing is wrong communicates that I know some proposition { the positivist's idea is just that the proposition known is not about the realist's fundamental property. The proposition known might be about a set of rules. Or it might be about a Kratzerian ordering. The task of this chapter and the next is to show that it is comparatively easy for ordinary agents to know about either one. I think the positivist should stick to the simplest idea that she can make work, and should prefer orderings to sets of rules until something forces her to opt for sets of rules instead. My basic recipe starts with an agent's convictions about 72 what's true, and reverse-engineers an ordering from those convictions. As we've seen, though, sets of rules can draw ner distinctions than would be re ected in an agent's convictions about what's true. Now suppose we nd something that forces the positivist to opt for sets of rules rather than orderings. If that happens, we've thereby found a new source of data to use in reverse-engineering what the agent is thinking about. That's not an objection. It's the answer to a problem for a set- of-rules positivist. I myself am skeptical that we can nd a source of data to solve this problem for the set-of-rules positivist. 3 So I happen to prefer being a simple Kratzerian-ordering positivist. But in necessariis unitas, in dubiis libertas { and reasonable positivists can disagree among ourselves about set-of-rules positivism and Kratzerian-ordering positivism. 4.2.2 Forms and limits of positivist realism I've just made an ecumenical point. There are several dierent options for develop- ing the basic positivist idea, and the positivist's basic philosophical ambitions are compatible with several of them. That is, the positivist is ecumenical about the arguments to modal vocabulary. She allows that the arguments can be orderings, or sets of rules. This section notes one important limitation on this ecumenicism. In particular, the positivist has to reject one standard feature of the orthodox treat- ment that Kratzer (2012) gives of modals. Kratzer thinks that the arguments to modal vocabulary are even richer than I've considered so far. She thinks they're functions from worlds to something like sets of rules { what she calls ordering sources. She uses these functions to capture examples like (8). (8) If the NC state legislature had been less transphobic, it'd be legally permissible for everyone to use a bathroom that matches their gender identity. Kratzer models (8) with a function from a world to the laws of NC at that world. Since the laws would be dierent if the legislature had been less transphobic, the function takes you to a dierent set of propositions at that world than at the actual world. It is hard for the positivist to make sense of these functions. In order for the positivist to appeal to that function, she would have to take our beliefs at the actual world to determine a function from an arbitrary world to a set of propositions. And it's not particularly plausible that our psychology determines that sort of function. Now if you disagree, I'm perfectly happy to concede the point. Someone who disagrees with me takes the positivist to have one fewer distinctive claim to defend. But trying to make sense of Kratzer's functions within the positivist's framework seems too dicult to me. 3 Section 5, below, discusses some semantic arguments for something like the set-of- rules view. 73 It's more natural for the positivist to explain examples like (8) by appeal to conditional presuppositions. She will associate (8) with these two contents: At-issue: if the NC state legislature had been less transphobic, it'd be permissible o4 for everyone to use a bathroom that matches their gender identity. Not-at-issue: the NC state legislature had been less transphobic É o 4 is the laws of NC Because (8) carries a conditional presupposition, we interpret it as being about a dierent ordering source (or set of rules) than a bare utterance of the consequent is about. Nothing in this suggestion requires Kratzer's functions from worlds to ordering sources. It just requires the resources the positivist already acknowledges: thought about bare sets of rules, or bare ordering sources. We've just seen that the positivist can't be perfectly ecumenical. There is one semantic resource (functions from worlds to ordering sources) that looks natural on traditional approaches but that looks highly unnatural on the positivist's approach. Is this failure to be perfectly ecumenical the exception, or is it the rule? I think it's the exception. In fact, it's the only exception that I've been able to nd. I want to close this section by noting two other important approaches to the semantics of modals, and noting that they are perfectly compatible with the positivist's picture. Orderings, ordering sources, and sets of rules all determine rankings of worlds: this world is higher ranked than that one, and so on. Some philosophers are skeptical that the semantics for modals uniformly rests on a ordering on worlds. They think it instead incorporates a orderings on actions, at least for some cases. (Mark Schroeder (2011) and Fabrizio Cariani (2013) have both made suggestions with this structure.) The positivist can also incorporate their suggestion. To do that, she would change what her ordering is a ordering over: it would be a ordering over actions, rather than over worlds. Nothing else would change. An agent's dispositions to believe would still determine which ordering she is thinking about. It's just that the ordering has a dierent structure than I've assumed up to now. The same point holds for other proposals about the semantics of modals. If you think that the semantics is contrastive, for example, then the positivist will appeal to a formal representation of contrast-sensitivity. Other theorists have suggested that the semantics should be sensitive to some- thing ner-grained than orderings on worlds or ordering sources, or even orderings on actions. Je Horty (2014) has suggested that the semantics should be sensitive to interactions among reasons. He argues that that semantics can capture facts about conditionals that the more traditional semantics cannot capture. It's straightfor- ward to extend the positivist's proposal to incorporate his suggestion. Horty (2012) develops one formal model of interactions among reasons, a model that builds from default logic. The positivist can then claim that our dispositions determine a set of reasons and their weights that we each take to be the moral standard, where those sets of reasons have more internal structure than orderings on worlds and so can represent ner distinctions. 74 Any of these changes to the positivist's approach would be a change of detail, not of the core of the view. The core of the view is that moral knowledge is easy, because it's easy to know about our own ordering/ordering source/set of reasons, and that's the only kind of knowledge that moral knowledge depends on. However, I will continue to characterize the positivist's view as a view about sets of rules. This characterization allows me to preserve uniform terminology throughout the whole dissertation, rather than talking in one way in the rst three chapters and then switching to another partway through Chapter 4. 4.3 Minor renements I've given a rst-pass recipe for guring out what an agent is thinking about. That recipe takes an arbitrary set of dispositions to believe. The task: constructing an object o i that models some individual's beliefs. If the agent believes that A is required to in C, then for any world w 1 where A does not in C, o i ranks some world where A does in C has higher than w 1 . If the agent judges A is forbidden to in C, then for any world w 1 where A does in C, o i ranks some world where A does not in C as higher than w 1 . What's required is modeled with higher-ranked worlds and what's forbidden is mod- eled with lower ranked worlds. We're in the process of rening this rst pass, trying to make it more and more adequate. This section describes some minor renements. 4.3.1 What do we Ramsify over? One signicant choice-point for the positivist is about the dispositions that she Ramsies over. She might Ramsify over our dispositions to believe generalizations: for example, that promises in general ought to be kept. Or she might Ramsify over our very particular dispositions to have beliefs about the moral status of complete naturalistic descriptions of an actions. That is, the dispositions are dispositions about how the agent would classify actions if she were in a position to grasp the complete naturalistic description of them. (None of us nite agents are in a position to grasp those complete naturalistic descriptions, of course. But we can still ask about how we would classify them if we could grasp those descriptions. 4 ) This second conception of Ramsication makes the idea compatible with the sort of particularism that Jonathan Dancy (2004) defends, where generalizations 4 What if we wouldn't be disposed to classify them in any particular way if we could grasp those complete naturalistic descriptions? Then we would be modeled with partial sets of rules, of the sort introduced later. 75 about our moral obligations always have exceptions. But I'll bracket those complete naturalistic descriptions in what follows. My examples will be examples about our beliefs about generalizations, to make the discussion simpler enough to follow. I also acknowledge that our moral attitudes might depend on more than just our beliefs. A range of emotional reactions also seem to be important in the way that we form and revise our moral attitudes. 5 It's no problem if they do. The range of states that we're Ramsifying over would then be correspondingly diverse; we'd then include anger, resentment, fear, and so on. Those states can determine what set of rules we're thinking about, and in that way ground our moral knowledge. But the positivist does not need these states to ground our moral knowledge. If emotional reactions are largely irrelevant, the positivist can just Ramsify over our beliefs. What matters is the mental states that we're disposed to have, not the particular type. I'll continue to focus on beliefs, but only for simplicity. 4.3.2 Transparency's revenge The dissertation started with a naive picture of what the positivist is doing. On that naive picture, the positivist takes moral knowledge to be like the knowledge we would have if we memorized a book mapping actions to how the book classies them. There were two important reasons that that naive picture was unsatisfying. It made moral knowledge too easy, and it made moral belief too transparent. It was too transparent because you couldn't have beliefs about what you ought to do without being aware of the considerations that determined what you ought to do. And that looks implausible. You can have moral beliefs without being aware of what explains why each belief is true. (That possibility is why normative ethics is hard, and why normative ethics is important.) But you might wonder if we've really avoided the transparency problem. After all, orderings or sets of rules threaten to be transparent in the same sort of way. Someone who is thinking about a particular set of rules 6 is in a position to know immediately what considerations it takes to matter. To have a full answer to the transparency challenge, then, we need to say more. In particular, we need to draw on one more formal tool. The further tool is the distinction between unarticulated thought and articulated thought. 7 Think of the dierence between (2a) and (2b). (2a) Bob believes that Russell defended logicism. (2b) Bob believes that Russell defended the proposition that math reduces to logic. 5 Antonio Damasio (1994) describes some of this data; others have emphasized it as well. 6 Remember my terminological choice: though I happen to favor Kratzerian orderings over sets of rules, I'll continue to talk as though it's sets of rules that matter. 7 This distinction is due to Mark Richard (1993). 76 It's possible for (2a) to be true even if (2b) is false. Suppose that Bob believes that Russell defended a view called \logicism" { if asked for defenders of logicism on a test, he'd give Russell as one of the answers. But he doesn't have any opinions about what view logicism is. He doesn't have an opinion about whether it's about the reduction of math to logic. It seems to be possible for Bob to nd himself in that state. And if it is possible, (2b) can be true even if (2a) is false. One way to explain this dierence is to suppose that the complement of (2b) contributes something with more internal structure than (2a)'s complement does. You need to be thinking about at least ve constituents to entertain the proposition that (2b) expresses: Russell, the defends relation, the reduces relation, math, and logic. You need to be thinking about fewer constituents to entertain the proposition that (2a) expresses. You only have to be thinking about the three constituents: Russell, the defends relation, and the proposition that math reduces to logic. And you can think about that proposition without thinking of the reduces relation. That is the dierence between (2a) and (2b): in order for (2b) to be true, Bob needs to be thinking of the constituents of the proposition that math reduces to logic. But he doesn't need to be thinking of those constituents in order for (2a) to be true. And that's why it's possible for (2a) to be true even if (2b) is false { Bob needn't know what logicism is to know that Russell defended it. I will mark this dierence by saying that the proposition that math reduces to logic is an unarticulated constituent of (2a), but an articulated constituent of (2b). I haven't given you any theory of what this dierence comes to { that's a task for a full paper. 8 I've just labeled the phenomenon. Other explanations of the phenomenon are also possible, including some which are compatible with a coarse-grained conception of propositions. The answer to the transparency challenge is that this phenomenon happens in the moral case, too. Ordinary moral thought is unarticulated de re thought about particular sets of rules. (3) Jane believes that she ought to keep this promise. (b) Jane believes that o 1 requires keeping this promise, and accepts that o 1 is the moral standard You don't need to recognize the considerations that o 1 takes to matter in order to think about o 1 , any more than you need to think about the constituents of logicism to think about logicism. The same thing is true in the linguistic case. You can think about your own idiolect, without the syntax or semantics of the idiolect being transparent to you. In general, I'll assume throughout what follows that ordinary de re thought about sets of rules is always thought about an unarticulated constituent. That's why it's not transparent. 8 Scott Soames (2015) gives one such theory, drawing on his cognitive event-type con- ception of propositions. 77 4.4 Contrast with \moral functionalism" So far, we only have an incomplete picture of the positivist's approach. There are lots of other questions to explore. And later sections will explore them. Before doing that, though, it's worth situating the positivist's commitments in the broader metaethical landscape. I'll do that by contrasting positivist realism with a super- cially similar proposal: the \moral functionalism" pioneered by Frank Jackson and Philip Pettitt. 9 The positivist, like Jackson and Pettitt, is Ramsifying over our ordinary beliefs. But the positivist has very dierent and more modest goals. She needs Ramsifying to accomplish much less than they do. Because the positivist's goals are more modest than theirs, it's much more plausible that she will achieve her goals. Jackson and Pettitt are interested in giving a reductive proposal about moral properties. Their reductive account starts with a set of platitudes about morality { things like \if an act is an intentional killing, then normally it is wrong" (Jackson 1998a, 130). They collect the platitudes about morality. Then they replace each occurrence of a moral term (like wrong) with a free variable. That process produces a string of open sentences, of the form ...X 1 ...X 2 ...X 3 ...X 1 ... This string grounds a functional characterization of each property. To get that functional characterization of some particular property X 1 , start by existentially closing all the free variables in the string except X 1 . Once you've done that, use that string as the restrictor in a denite description, and you've got a term that refers to whatever realizes the functional role that X 1 plays. Since natural properties are what realize moral properties, they take themselves to have oered a reduction of the moral to the natural. Now Jackson and Pettitt think of themselves are realists, and not as relativists. When two people are disagreeing about what is right, they're disagreeing about the same property. In order make good on their self-conception as realists, though, they need to explain how we're all talking about the same moral properties. And it is initially hard for them to do that. They can't just conjoin all the moral beliefs that anyone has ever had. There are lots of moral disagreements, so conjoining all those beliefs would result in inconsistency. To get around this problem, they propose that it's the platitudes about morality that matter for Ramsifying. Here is Jackson making this point: \genuine moral disagreement, as opposed to mere talking past one another, requires a background of shared moral opinion to x a common, or near enough common, set of meanings for our moral terms. We can think of the rather general principles that we share as the commonplaces or platitudes or constitutive principles that make up the core we need to share in order to count as speaking a common moral idiolect" (Jackson 1998a, 132). They Ramsify over platitudes: a background of shared moral opinion. The positivist's ambitions lead to a very dierent kind of picture than the moral functionalist has. Moral functionalists retreat to platitudes of folk morality because the body of folk morality is inconsistent, when considered as a whole. Now 9 See Frank Jackson (1998a), and Frank Jackson and Philip Pettit (1995). 78 individuals can have inconsistent moral beliefs, just as much as groups do. But inconsistency in individual belief does not lead the positivist to retreat towards platitudes. The positivist instead introduces a new dierent formal tool. She posits partial sets of rules { that is, sets of rules that altogether fail to classify some actions. Those actions aren't permissible according to the partial set of rules, nor are they forbidden, nor are they required. The set of rules is just silent about them. The positivist models someone with inconsistent beliefs by appeal to a partial set of rules that are silent on just the points where the agent's beliefs are inconsistent. Now there are signicant semantic challenges about the intelligibility of appealing to these partial sets of rules. The Appendix to this chapter develops on approach to those semantic challenges. (These partial sets of rules also play other important roles for the positivist. They're the reason why the positivist does not need to go into some subtle problems about Ramsifying, like what happens when no object satises all the open sentences. There is always some set of rules that satises all the open sentences the positivist uses, because the positivist allows that sets of rules can be partially dened.) The positivist countenances these partial sets of rules, because her ambitions are totally dierent from the functionalist's ambitions. Functionalists are Ramsifying to nd properties to explain moral disagreement. If you countenance partial sets of rules, you've abandoned that ambition. Partial sets of rules denitely don't capture all cases of moral disagreement. That's just what it is to be partial! The positivist's crucial idea is that moral disagreement doesn't have to involve disagreement about the same set of rules in order to be genuine. For her, then, Ramsifying determines something that represents each individual agent's moral beliefs (at a time), not something that represents the moral truth. One notable consequence of this fact is that changes in moral belief can change what set of rules we're thinking about. It's a substantive question for the positivist when that happens. It's not plausible such changes always change what set of rules we're thinking about. But it takes some serious readjusting of the present approach to explain why. We'll take up this question in the next section,x3. Despite some similarities, then, moral functionalists are attempting something much more ambitious than the positivist attempts. We should thus expect that there are serious problems for the moral functionalist that are not also problems for the positivist. And indeed this expectation is born out. One problem for the moral functionalist arises because the proposal is intended to give the current meaning of our moral terms by appeal to the platitudes of mature folk morality. Jackson does not intend it as a proposal about current folk morality. He conceives of his proposal in this way because he takes folk morality to be unsatisfying on certain rst-order questions. For example, he denies that the platitudes of current folk morality about the doing/ allowing distinction is adequate (Jackson 1998a, 133-4). Folk morality has to be developed, he thinks, before its platitudes can be Ramsied. But as Stephen Yablo (2000) has noted, mature is a normative term itself. It can't be replaced with a sociological phrase (like at the end of inquiry). If it were, the account would be vulnerable to empirical hostage. Imagine that there's no moment 79 where everyone accepts the same platitudes, because certain anti-social attitudes last until the end of our species. We would then need to pick out a distinguished subset of the future population. And we probably need normative notions to do that. It then looks like Jackson is abandoning his ambitions to be a reductive realist in appealing to the platitudes of mature folk morality. This kind of problem does not arise for the positivist, because she's trying to do so much less than moral functionalists are trying. She's not Ramsifying to give a reductive account of moral properties. She's instead appealing to Ramsifying to fry psychological sh: she uses it to explain how agents are thinking about particular sets of rules. By contrast, the moral functionalist appeals to Ramsifying to fry metaphysical sh, at least in part: she uses it to explain what moral properties are, and how we refer to them. And the positivist has an important license for optimism that she will be able to fry the psychological sh she needs to fry. Introducing that license for optimism was the whole point ofx1. We are able to think about what sentences mean in our own idiolects. And our ability to do that is either grounded or revealed in our linguistic dispositions. That fact shows that we can think about very complicated objects (in this case, idiolects). And that's all the positivist needs. She needs it to be possible for agents to think about very complicated sets of rules in virtue of their dispositions. She Ramsies to explain how it's possible for them to do that. But Ramsifying isn't essential. What matters is that there is some explanation of why unsophisticated agents can do that. There is then one central commonality between the positivist realist and the moral functionalist, and three central dierences. They agree that Ramsifying helps determine the content of moral judgment. The positivist takes each person's beliefs to be about the set of rules that results from Ramsifying over all the beliefs (or other relevant attitudes) that that person has. The moral functionalist takes each persons beliefs to be about the properties that result from Ramsifying over the platitudes of mature folk morality Theydisagree about whether Ramsifying results in everyone thinking about the same objects/ properties. The positivist expects everyone to be thinking about dierent sets of rules. The moral functionalist takes everyone to be thinking about the same properties. They disagree about which beliefs to Ramsify over. The positivist Ramsies over all of a single person's moral beliefs (at a time). 80 The moral functionalist Ramsies over a distinguished subset, the platitudes of mature folk morality. (This dierence comes from their dierent ambitions: the moral func- tionalist is Ramsifying to nd properties to explain moral disagreement, while the positivist is not.) Ramsifying is inessential to the positivist but essential to the moral functionalist The positivist needs to explain how cognitively unsophisticated agents can think about sophisticated objects (like sets of rules, or like idiolects like English), and Ramsifying is her going hypothe- sis. The moral functionalist's account of how we refer to moral prop- erties depends essentially on Ramsifying over the platitudes of mature folk morality. In general, then, there are some important continuities between what the positivist is doing, and what the moral functionalist is doing. But the positivist needs Ram- sifying to do so much less work. That's the reason why her proposal is so much more plausible than moral functionalism. 4.5 Wrapping up This chapter has explained how agents are able to think about particular sets of rules. The next chapter will turn to knowledge: it will explain how we can know about those sets of rules. This chapter has taken on one substantial and distinctive commitment that I haven't yet defended. I've assumed that it's legitimate to appeal to partial sets of rules { rules that fail to classify some actions in any particular way. In particular, I've assumed that you can appeal to those sets of rules without taking on any problematic commitments about the semantics of modals. I've included an appendix arguing for this assumption. If you're happy to grant that assumption, though, you can skip directly to Chapter 5. 81 4.6 Appendix: Semantics for partial orderings This appendix digs into some subtle semantic issues. So I'm going to switch back { just for this appendix { to talk about Kratzerian orderings, rather than sets of rules. Talking about orderings makes my background assumptions a little easier to follow. The goal here is to articulate an adequate semantics for acknowledging that we can think about partial orderings. 10 4.6.1 Supervaluationist semantics for partial orderings My account of partial orderings is modeled on an example of a partially dened term from Scott Soames (1999). He imagines introducing a new term (smidget) by giving sucient conditions for its applying and sucient conditions for its not applying. If you're shorter than four feet, you're a smidget. If you're taller than ve feet, you're not a smidget. This denition leaves open whether someone between four and ve feet is a smidget. It's partially dened in that sense. It's dened only for some objects. I will give a non-standard semantics for smidget: the semantics will be supervaluationist, and quantify over precisications of the predicate. The positivist does need the supervaluationist semantics, for philosophical reasons the next section details. And there are tricky issues about developing my supervaluationist approach. This section focuses on the technical issues, and the next takes up the philosophical ones. I start with an important distinction among supervaluationist theories, a dif- ference about the truth predicate. On the rst sort of theory, a sentence is true i it is supertrue { that is, if it is true on every precisication. The second sort of theory, by contrast, is disquotational. The truth predicate is characterized by the (non-paradoxical) instances of the T-schema. Here are the schemata for each. (Truth as Supertruth) ` `S' is true' is true i `S' is true on every salient precisica- tion ` `S' is false' is true i `S' is false on every salient precisi- cation (Truth as Disquotational) ` `S' is true' is true i S ` `S' is false' is true i:S The two approaches are very dierent, and incompatible. The disquotationalist is taking the truth predicate to itself be vague, and to admit of dierent precisications. Crucially, though, she also takes the (non- paradoxical) instances of the T-schema to constrain precisications of the truth 10 I'm grateful to John Hawthorne for fantastically helpful conversations about this material. 82 predicate. The semantics for true (and for false) only quanties over those precisi- cations that verify those instances. (Kit Fine (1975) calls this sort of constraint a penumbral connection.) As a result the disquotationalist quanties over another precisication that the supertruth theorist does not: she quanties over precisi- cations of true as well as precisications of `S'. Both the disquotationalist and the supertruth theorist agree that every instance of the law of excluded middle (LEM) is true. But they disagree about whether every instance of (Bivalence) is true. Every instance of (Bivalence) is true, if the disquotationalist is right. But some instances of (Bivalence) are false, if the supertruth theorist is right. (LEM) Either S or:S (Bivalence) Either `S' is true or `:S' is true Both the supertruth theorist and the disquotationalist will vindicate every instance of LEM. Each precisication veries S, or it veries:S. And both theorists ac- cept disjunction introduction, so each precisication veries S or:S. Importantly, though, some instances of Bivalence are false if for the supertruth theorist. For her, Bivalence is true only if S is either supertrue or superfalse. And it's possible for S to be neither. The disquotationalist, by contrast, does vindicate Bivalence, as well as LEM. Start by xing on the simple case where S does not contain any occurrences of true or of false. Fix on some precisication `true i ' of the truth predicate in the relevant instance of Bivalence. The disquotationalist requires this precisication to verify the T-schema. And the disquotationalist already thinks that each precisication veries S, or veries:S. Since she accepts the relevant instances of the T-schema, she also thinks that each precisication veries `S' is true i , or it veries `:S' is true i . Since the disquotationalist accepts disjunction introduction, she vindicates Bivalence. Now turn to the more complicated case, where S itself contains the predicate true or the predicate false. The Truth-Teller is one example. (Truth-Teller) The Truth-Teller is true. (Bivalence) Either Truth-Teller is true or:Truth-Teller is true This instance of Bivalence is true if the disquotationalist is right. It's true because the disquotationalist associates the embedded occurrences of true with a dierent truth predicate (true 1 ) than the unembedded occurrences (true 2 ). The relevant instance of Bivalence will thus be: either `The Truth-Teller is true 1 ; is true 2 or `The Truth-Teller is not true 1 ' is true 2 . And this instance of Bivalence will be true. There are two precisications of the embedded truth predicate true 1 . On the rst, the Truth-Teller is true 1 , and on the second it is not true 1 . These are the only two precisications, because precisications are entirely classical. If something isn't in the extension of is P, it's in the extension of is not P. Every admissible precisication is then a precisication where one disjunct of Bivalence is true 2 . 83 First Precisication of true 1 : The Truth-Teller is true 1 . Then: `The Truth-Teller is true 1 ' is true 2 { by the penumbral connection about true 2 given by the dis- quotational schema: ` `S' is true 2 ' is true 2 i S Second Precisication of true 1 : TheTruth-Teller is not true 1 . Then: `The Truth-Teller is true 1 ' is false 2 { by the penumbral connection about false 2 given by the dis- quotational schema: ` `S' is false 2 ' is true 2 i:S The Truth-Teller is not true 1 { by the penumbral connection about false 2 given by the dis- quotational schema: ` `S' is false 2 ' is true 2 i:S `The Truth-Teller is not true 1 ' is true 2 { by the penumbral connection about true 2 given by the dis- quotational schema: ` `S' is true 2 ' is true 2 i S In general, then, one disjunct of Bivalence is true on every admissible precisi- cation of true 1 . And the background logic is entirely classical. So the relevant instance of Bivalence is true, even for examples containing the truth predicate. I will adopt a disquotationalist conception of the truth predicate, and set the supertruth theory aside. Why? The supertruth theory doesn't let negation and is false play nicely together. We would like all (non-paradoxical) instances of the Interchange schema to be true. (Interchange) `S' is false i:S But the supertruth theorist holds that some instances of Interchange are not true. Suppose that some precisications verify S and others verify:S. The righthand side of Interchange is true on any precisication that veries:S. Unfortunately, though, the lefthand side is false on all precisications. Falsity is superfalsity, and the existence of precisications that verify:S guarantees that `S' is not superfalse. But then Interchange is not true on all precisications, so it's not true, given a supertruth conception of truth. 11 With this background in hand, we can give a supervaluationist semantics for partially dened terms. The semantics all take a set of precisications as given: the set P for smidget, and the set O for modals. 11 See Cian Dorr (2003) and Vann McGee and Brian McLaughlin (1995) for further discussion of these and related issues; Dorr defends the disquotational conception in more detail. 84 Bob is a smidget is true i (8x: x2 P) [Bob2 x] pMust(p)q is true i (8o i : o i 2 O) [(8w::9w 0 (w 0 > oi w) [p is true at w]] pMay(p)q is true i (8o i : o i 2 O) [(9w: :9w 0 (w 0 > oi w) [p is true at w]] Go back to the way Soames introduced smidget. If you're shorter than four feet, is a smidget applies to you. And if you're taller than ve feet, is not a smidget applies to you. My supervaluationist semantics models these facts by taking the precisications P for is a smidget to agree among themselves about people shorter than four feet or people taller than ve, but to disagree among themselves about the other cases. And we generalize this strategy to the semantics of modals. We model partial orderings with a set of fully precise orderings, which all disagree among themselves in those cases where we want the ordering to be partial. In testing this proposal, we should make sure that it vindicates plausible equiv- alences like the Duality schema. (Duality) `May(S)' is true i `:Must(:S)' is true Kratzer's semantics is attractive in part because it vindicates Duality. It'd be terrible if my proposal lost this advantage! Fortunately, though, my proposal does preserve it. Start by noting the object idiolect schema corresponding to Duality is supertrue. (Object Idiolect Duality) May(S) i:Must(:S) Why is (Object Idiolect Duality) true? Well, in developing a supervaluation- ist semantics for modals, I am supervaluing over traditional Kratzerian orderings that are complete: that classify every action as permitted, forbidden, or required. And Kratzer's semantics predicts that Object Idiolect Duality holds for ev- ery traditional Kratzerian ordering. It guarantees, in other words, that Object Idiolect Duality is supertrue. Now the disquotationalist also takes the (non- paradoxical) instances of the T-schema to be supertrue, including the following instance schemata. (T-Schema 1) `May(S)' is true i May(S) (T-Schema 2) `:Must(:S)' is true i:Must(:S) And the background logic is fully classical. So Duality follows straightforwardly. (Duality) `May(S)' is true i `:Must(:S)' is true Things have worked out very elegantly. I don't have to posit an independent \penumbral connection" between may and must to explain whyDuality is true. It rather follows from Kratzer's semantics, plus the relevant instances of the T-schema. 85 4.6.2 Why supervaluationist semantics for partial orderings? That's the end of the purely technical issues. We won't leave technical issues be- hind; the philosophical issues are entangled other technical issues. At this point, though, we're changing gears. I'm going to explain why I chose the supervaluationist semantics for partial orderings that I did. Soames does not give a supervaluationist account of his smidget example. He denes is a smidget directly, without supervaluation over dierent extensions. Is a smidget has a denite extension, a set of objects that fall under it. And it has a denite anti-extension, a set of objects that do not fall under it. But the two fail to be jointly exhaustive. Some objects are neither in the extension nor in the anti-extension. This approach has a lot going for it. It's a lot simpler than my supervaluationist alternative. In specifying sucient conditions for being a smidget, it seems like we're specifying the extension. And in specifying sucient conditions for not being a smidget, it seems like we're specifying the anti-extension. There's no need to appeal to the extra machinery that my supervaluationist proposal does. 12 A Soames-style conception of partial orderings has very dierent philosophi- cal upshots than a supervaluationist conception. Note that the supervaluationist predicts that (4) is always true. (4) Every action is either required, permitted, or forbidden To test to this point, take someone who has no beliefs about some act of promise- keeping, or whose beliefs about the act of promise-keeping are incoherent. The positivist will model this person's belief with a partial ordering that is silent on the act of promise-keeping. And my supervaluationist semantics models that partial ordering with a set of complete orderings that agree about the other acts of promise- keeping, but disagree about that one act of promise-keeping. The dierent orderings are complete, in that each one will classify the act of promise-keeping as either required, permitted, or forbidden. This act of promise-keeping is either required, permitted, or forbidden will then be true on each ordering, and so supertrue. This reasoning generalizes. So (4) will always be be true. On a Soamesian conception of partial orderings, by contrast, (4) won't be true. And this fact turns out to make it much harder for the positivist to draw on a Soamesian conception of partial orderings. But rst: what would a Soamesian conception of a partial ordering o 1 even look like? An ordering would be partial if the extension of predicates about o 1 are themselves partial. It'd be partial if is required by o 1 or is permitted by o 1 are partial. For example, o 1 might not classify some act of promise-keeping in any particular way. It's not in the extension or anti-extension of is required by o 1 . No problem so far. Kratzer's semantics can be modied to allow for this possibility. 13 12 Vann McGee (1990) and Jaime Tappenden (1993) both develop accounts of vagueness that are similar to what Soames suggests about smidget. 13 To see how, adopt the following abbreviations. DetHigher =f<x1;x2 >:x1 is determinately higher ranked byo1 thanx2g 86 Now go back to someone who has no beliefs about some act of promise-keeping, or whose beliefs about the act of promise-keeping are incoherent. The Soamesian partial ordering o S that models this person's beliefs will be partially dened, in that it won't classify that act of promise-keeping as required by o S , forbidden by o S , or permitted by o S . Now it doesn't follow from this point that that act is promise-keeping is not required by o S , or that it's not forbidden by o S , or that it's not permitted by o S . It's as much as mistake to think that that act is required by o S as it is to think that it's not required by o S . In each case, the belief is neither true nor false. As a result, (4) will be neither true nor false. (4) Everything is required, permitted, or forbidden (4) can't be true simpliciter. The act of promise-keeping isn't in the extension of required-by-o S , or in the extension of permitted-by-o S , or in the extension of forbidden-by-o S . And it would have to be in one of those extensions in order for (4) to be true. Similar reasoning about the anti-extension of those predicates prevents (4) from being false. So (4) will be neither true nor false. And problems come up for the positivist if (4) is neither true nor false. Suppose that it is rational to suspend judgment about propositions that are neither true nor false. Then the positivist predicts that (5) is true of anyone who is morally unopinionated about some act of promise-keeping. (5) Such a person is rational in suspending judgment about whether that act of promise-keeping is neither required, permitted, nor forbid- den. Remember that the positivist associates (4) with a pair of attitudes, about the relevant partial ordering. (5lf) Such a person is rational in suspending judgment about whether act of promise-keeping is neither required by o 1 , permitted by o 1 , nor forbidden by o 1 (while accepting that o 1 is the moral standard) DetLower =f <x1;x2 >:x1 is determinately lower ranked by o1 than x2g HighestWorlds(X) =f x1 :9x2 :<x1;x2 >2X^:9x3 :<x3;x1 >2Xg Then we can give the following semantics. pMay(p)q is determinately true i9w(w2 HighestWorlds(DetHigher)\ M ^ p is true at w) pMay(p)q is determinately false i [the w's: w2 HighestWorlds(DetLower) \ M] (p is true at w) pMust(p)q is determinately true i [the w's: w2 HighestWorlds(DetHigher) \ M] (p is true at w) pMust(p)q is determinately false i9w(w2 HighestWorlds(DetLower)\ M ^ p is true at w) 87 On a Soamesian conception of partial orderings, the at-issue content associated with (5)'s complement is neither true nor false. And the positivist's core idea is that epistemic evaluations (like about what's rational) only evaluate the epistemic status of the at-issue content. So as long as it's rational to suspend judgment about propositions that are neither true nor false, (5) has to be true. And this result is implausible. (5) ascribes an attitude about the scope of the demands of morality. It claims that it's rational for our agent to take this act of promise-keeping to fall outside the scope of morality. And on the picture in the last paragraph, ignorance of the moral questions justies suspending judgment about this question. After all, your ignorance is why you're thinking about a partial ordering that doesn't classify the action. And it's implausible that (5) is true for this reason. Ignorance about your obligations doesn't justify this sort of metaphysical conclusion, about the scope of morality! That's the basic reason why a Soamesian conception of partial orderings ts poorly with the positivist's aspirations. Now in developing this problem, I assumed that it's rational to suspend judg- ment about propositions that are neither true nor false. But the problem is robust, even across other assumptions. Maybe you should just refuse to assert propositions that are neither true nor false, and also refuse to assert their negation. Then the positivist predicts that it's rational to refuse to assert that that act of promise- keeping is either permitted, forbidden, or required. Even that prediction may look implausible. The claim that you're refusing to assert is a metaphysical claim, about the structure of morality. It's the claim that any action is either forbidden, required, or permitted. As before, ignorance about your obligations doesn't seem to make it rational to refuse to assert that sort of metaphysical claim. This problem does not arise for my supervaluationist conception of partial or- derings. Sentences like (4) are always true given the supervaluationist conception. So it won't be rational to suspend judgment towards (5)'s complement, even if you're not opinionated about that act of promise-keeping. That's the basic reason I led with the supervaluationist conception of partial orderings. It is the one that best ts the positivist's aspirations. It looks like the positivist needs a particular formal approach to partial or- derings. A rival theory of partial orderings just won't work for her. Does this observation show that something is wrong with her approach? Not necessarily. The positivist needs the supervaluationist approach because of the expressive limitations of the other theory. In particular, that rival theory doesn't discriminate between partial orderings that verify everything is neither required, permitted, or forbidden and those that don't. The supervaluationist approach, by contrast, is expressively rich enough to draw this distinction. We've already seen how to model partial or- derings that don't verify that sentence. But the supervaluationist can also model orderings that do verify it. For example, she might use a set of orderings that are all are partially dened in the way Soames describes, or she might develop a more complicated supervaluationist semantics that has the same eect. Moreover, the supervaluationist can also model orderings that are silent on that sentence. To do that, she would use a set of orderings that include both complete orderings and Soames-style partial orderings. One way to legitimate a formal tool is to show that 88 it expands our expressive resources, allowing us to express more than we otherwise could. That's the basic reason we should recognize the legitimacy of a supervalua- tionist account of partial orderings, over a purely Soamesian approach. The use of supervaluationist tools raises another question. Some philosophers draw on those tools in their accounts of vagueness. Others appeal to the sort of partiality that Soames appeals to. And others draw on yet other tools. Does this account of partiality commit the positivist to a supervaluationist account of vagueness? Not at all. Vagueness and partiality are just dierent phenomena. We can see why even in the present case. It is vague which partial ordering models my beliefs. Particular partial orderings are themselves perfectly precise and have sharp boundaries. Imagine someone who is only opinionated about a sequence of cases about promise-keeping. She otherwise has no moral opinions. In the rst case, I can keep some particular promise without doing anyone any harm. In the last case, I can keep that very same promise, but only by killing someone. Our agent takes me to be required to keep the promise in the rst case, and forbidden to keep it in the last. We do need partial orderings to capture this sequence. But it's vague which partial orderings do capture it; partial orderings are themselves perfectly precise. The positivist thus needs to draw on further resources in theorizing about vague- ness. But she can draw on any theory of vagueness that she pleases. She can be an epistemicist. 14 She'd then hold that each of us is thinking about some particular fully precise partial ordering, without being in a position to know which one. Or she can be a supervaluationist about vagueness. 15 She'd then model agents with sets of sets of orderings. Each set of sets would represent one admissible resolution of the vagueness about which partial ordering the agent is thinking about. And so on through the dierent models of vagueness. In general, then, the positivist doesn't incur any particular commitments about vagueness. She does incur par- ticular commitments about partially dened terms. But those are just dierent topics. 14 See Roy Sorensen (1988) and Timothy Williamson (1994). 15 See Kit Fine (1975) and David Lewis (1982). 89 Chapter 5: De re knowledge We're still in Part I of the dissertation. And the goal of Part I is to show that it would be nice for non-skeptical realists if a particular linguistic claim about presupposition is true. Positivist realism follows from that linguistic claim, and positivist realism has a range of nice upshots for non-skeptical realists. The goal of Chapters 4 and 5 is to show that the linguistic claim really is the crucial pivot. Now positivist realism does have other commitments, beyond that linguistic claim. But Chapters 4 and 5 try to show that those other commitments are plausible enough that they're not a reason for rejecting the view. The view rises and falls with the core linguistic claim. This chapter takes up the positivist's commitments in general epistemology. It shows how the positivist can have the resources she needs to vindicate the de re knowledge of sets of rules that she takes moral knowledge to require. It's ambitions are continuous with the ambitions of the last chapter. In the way the last chapter focused on the positivist's commitments in the philosophy of mind, the chapter focuses on her commitments in general epistemology. Its substantive assumptions are also continuous with the substantive assumptions of that last chapter. The last chapter focused on agent's dispositions to believe in explaining how agents can have de re beliefs about sets of rules. This chapter uses those dispositions to explain how agents can have de re knowledge. Now the last chapter presented an initial rst pass account of how the agent's dispositions determine what set of rules she's thinking about. I rened that rst pass in several ways. This chapter will add one further renement. It modies the rst pass to allow for performance errors { where the agent's dispositions misre in some particular way. This modication has important upshots for the positivist's account of our moral knowledge. 5.1 Performance errors, how I've given a rst-pass recipe for guring out what an agent is thinking about. That recipe takes an arbitrary set of dispositions to believe. The task: constructing an object o i that models some individual's beliefs. If the agent believes that A is required to in C, then for any world w 1 where A does not in C, o i ranks some world where A 90 does in C has higher than w 1 . If the agent judges A is forbidden to in C, then for any world w 1 where A does in C, o i ranks some world where A does not in C as higher than w 1 . What's required is modeled with higher-ranked worlds and what's forbidden is mod- eled with lower ranked worlds. This rst-pass recipe rules out the possibility of performance errors. I'll un- derstand performance errors as cases where an agent really is thinking about one particular set of rules, but makes a mistake about what that set of rules demands. The rst-pass recipe takes our dispositions to believe to immediately determine what set of rules the agent is thinking about. There just isn't space in this recipe for performance errors. Any apparent example of a performance error is interpreted as a case where the agent switches which set of rules they're thinking about. And if they later stop make this performance error, they've switched which set of rules they're thinking about, all over again. And that's implausible. In general, the positivist has to acknowledge that changing your moral beliefs can sometimes change what set of rules you're thinking about. And she has very good reason to acknowledge that point. Your moral beliefs are intimately tied to what set of rules you're thinking about. They either ground or reveal those rules. So changes to those dispositions had better have the potential to change which set of rules you're thinking about! But this basic conviction doesn't require her to deny that there are any performance errors. Linguistic judgments can also involve performance errors. For example, I can miss one of the readings of Mary left her book at the bank, say. Those sorts of performance errors don't show that I'm speaking a dierent idiolect now than I was earlier. The positivist will say something similar. But getting clear on just what to say raises some very important issues. It's possible to take this naive approach to modeling linguistic behavior, too. In that case, we'd appeal to the formal object that perfectly ts our actual use. In the linguistic case, though, it's much more plausible that other considerations matter too. To introduce this point, note that the basic idea of Ramsifying needs to be supplemented with some additional preferences. My dispositions to use some term might underdetermine what it means. For example, maybe my dispositions are compatible with some very weird semantics for and. p <st> :q <st> : 8 < : p is false and q is false if p = that Bob is happy p is true and q is true otherwise (Maybe I've never thought about Bob.) To exclude this sort of candidate for and, we should balance faithfulness to use with a preference for simpler interpretations of the idiolect, perhaps also with a preference for interpretations involving comparatively natural properties. 91 An attractive way to explain performance errors is as taking those further pref- erences to trump faithfulness to use in some cases. Suppose that on Tuesday, I missed one of the readings of Mary left her book at the bank. Every other day of the week, though, I do hear both readings. We might reasonably prefer an interpreta- tion of my linguistic behavior that takes me to speak the same idiolect this whole week, as the simplest interpretation. In interpreting me in that way, we're treating my not hearing of the readings as a performance error, rather than revealing that I switched which idiolect I spoke. Our preference for a simpler theory explains how we might reasonably classify some behavior as a performance error, rather than taking it to reveal that the person has switched which idiolect they speak. Go back to the moral case. My rst-pass recipe described the very simplest way to construct sets of rules, which moves directly from the agent's dispositions to the set of rules that she's thinking about. But the positivist can also draw on the same resources that we use in the linguistic case. She might privilege sets of rules that are simpler, or sets of rules that are sensitive to more natural properties. There are lots of other interesting questions to explore here, and lots of room for the positivist to make her view more plausible by toggling particular features of her view. In toggling those features, though, it's important not to lose sight of the posi- tivist's overall ambitions. She wants to explain why moral knowledge is so easy. And the more our sets of rules swing free from our dispositions, the harder it is to explain that point. To introduce this point, note I speak a idiolect where woodchucks are groundhogs expresses a truth. But I don't have privileged and immediate knowledge of this fact, because my own individual dispositions don't determine whether it's true. It's true because the most natural referent for woodchuck is the same as the most natural referent for groundhog. In this case, the linguistic example is patterning very dierently from the moral case, because the positivist takes the moral case to involve systematic privileged and immediate knowledge of sets of rules. In order to vindicate our moral knowledge, the positivist needs our individual dispositions determine sets of rules more directly than our individual dispositions determine what idiolect we speak. There is still some room for appeal to simplicity or naturalness in the moral case, but less than in the linguistic case. As a result, the positivist will take our sets of rules to be more tightly tied to our moral dispositions than our idiolects are tied to our linguistic dispositions. Much of the rest of this chapter focuses on getting clear on this point. But before moving on, we should pause brie y over another question. Our dispositions leave some moral questions open, because we're sometimes unsure about what we ought to do. What sets of rules model that sort of uncertainty? One option is to hold that simplicity and naturalness then do play a signicant role in determining what set of rules we're thinking about. It's just that we have privileged and immediate knowledge only when our dispositions are the thing that determines what set of rules we're thinking of. The second approach takes our dispositions to the main thing that determines what set of rules we're thinking about. Simplicity and naturalness play only a minor role. Suppose I'm almost always disposed to make some kind of judgment but rarely make the opposite judgment. Maybe in that case, I'm thinking about a simple set of rules that does not vindicate that 92 opposite judgment. But it's pretty unusual that our dispositions get overruled like that. I'll adopt the second approach, because it's simpler to talk about. It allows the positivist to claim that it's almost always easy to know about own set of rules. Be- liefs about our sets of rules are almost always grounded (or revealed) in dispositions to believe. When we have the right dispositions, those beliefs are also pieces of knowledge. (The next chapter will discuss this claim in more detail.) Things would look similar on the rst approach. The positivist would be claiming that it's almost always easy to know about those parts of our own sets of rules that our dispositions determine that we're thinking about. But it's cumbersome to talk in that way, and I don't see any philosophical payos. So I'll bracket the more cumbersome option. The choice between the rst and second approaches does have important down- stream consequences. In particular, there will be a lot more indeterminacy in the sets of rules on my favored approach than on the alternative. After all, our current dispositions don't determine an answer to all moral questions. Just think about the sorts of considerations that can get you out of a promise. Saving a life will, preventing someone from losing their legs probably will, but preventing some minor harm won't. There will be some cases in between where we're not sure. These dispositions might determine complete sets of rules in conjunction with a strong preference for simplicity and naturalness in attributing sets of rules. That is, they might determine complete sets of rules if we took the rst option. But when we take the second option, we should expect the sets of rules to be much more partial. 5.2 Moral knowledge This section explains how the positivist vindicates a range of de re knowledge about sets of rules. I begin with one important piece of terminology. I will say that an agent's dispositions to make moral judgments collectively constitute a mental representation of the demands of some particular set of rules o 1 . A mental representation of o 1 = df the set of dispositions that o 1 models. The core thesis of this chapter is that an agent has privileged and immediate knowl- edge of a set of rules when her mental representation by itself determines what set of rules she's thinking about. She has less privileged and immediate knowledge if other factors also determine what set of rules she's thinking about. The more we weigh simplicity or naturalness and the less we weigh the mental representation, the less we have privileged and immediate knowledge. 93 5.2.1 The good cases: when mental representations support knowl- edge This section describes some good cases where someone's mental representations do support knowledge. I'll do that by focusing on someone whose mental representation totally determines what set of rules she's thinking about, and where she's free of performance errors. In supposing that she's free of performance errors, I'm assuming only that her moral dispositions are perfectly consistent with teach other { that she's not disposed to make inconsistent judgments. Call this person Meredith: Meredith's mental representation by itself determine what set of rules she's thinking about. Meredith is an idealization, in two important ways. This section will describe how moral knowledge works in this idealization, and the next section will relax these idealizations. I claim that Meredith has knowledge about that set of rules, and not mere true belief. To make good on this claim, I need to show that her belief has whatever addi- tional features distinguishes knowledge from mere true belief. And there's immense controversy about just what features those are, of course. So I intend the following discussion less as articulating a denitive statement of the positivist's view, and more as a heuristic to the sort of account that the positivist will oer. My goal will be to put you in a position to tell when a mental representative satises whatever stricture you favor on knowledge. And I'll pursue that goal by discussing some representative accounts of what distinguishes knowledge. In order for a belief to be a piece of knowledge, it has to be justied. And some philosophers are reliabilists about justication. A belief is justied if it's formed by a reliable method. Meredith's beliefs about her set of rules o M will satisfy this requirement, because her mental representation is a reliable way of forming beliefs about the set of rules. This claim should look initially plausible. If it wasn't, it wouldn't be a mental representation about that set of rules. It would be a mental representation of another set of rules. Remember that Meredith's beliefs are about whatever set of rules ts her moral dispositions most tightly. And the details bear this point out. In order for Meredith's set of dispositions to be an unreliable guide to o M , they would have to lead Meredith to have some false beliefs about o M . Consider a concrete case. Suppose that o M always re- quires promise-keeping, come what may. And suppose that Meredith is disposed to acknowledge that it's permissible to break promises in some cases. For example, imagine that she has the following dispositions. Belief: if you promised to meet your friend for lunch but can save a life if you don't, x requires saving the life. Acceptance: x is the moral standard Belief: In every other case where you promised to meet your friend for lunch, x requires you to keep the promise. Acceptance: x is the moral standard In this case, there will be another set of rules o M 0 that ts Meredith's dispositions 94 even more closely than o M . o M 0 would require saving a life, rather than keeping the promise, as o M requires. Since o M 0 ts her dispositions even better, her beliefs have to be about o M 0 , rather than o M . (Remember that we're supposing for simplicity that Meredith's mental representation is the only thing that determines what set of rules she's thinking about, and we supposed that her dispositions are perfectly consistent with each other.) We were wrong in supposing that Meredith's beliefs were about o M rather than o M 0 . Meredith's mental representation have to be a highly reliable guide to whatever set of rules they represent. If they weren't, they would be a representation of a dierent set of rules, and would be a highly reliable guide to that other set of rules. I'll call this approach a Mentalist explanation: it appeals centrally to the agent's mental representation of a set of rules. The positivist will give a similar Mentalist explanation of the other features that distinguish knowledge from mere true belief. Before working through the details, though, it's helpful to compare this explanation with another, more familiar kind of account. In particular, it's helpful to note that this account has has the same structure as an inferential-role explanation of logical knowledge. Suppose that you are competent with the logic of conjunction. From your belief that p and your belief that q, you're disposed to infer p^ q, and from your belief that p^ q, you're disposed to infer p and also to infer q. These dispositions are reliable. They won't lead you astray, to infer something false from something true. They are a highly reliable guide to the facts about conjunction. After all, those dispositions seem to be constitutive of competence with conjunction { you're competent with conjunction just in case you have these dispositions. The explanation of the reliability of these dispositions is thus constitutive, from the nature of competence with the concept. This constitutive explanation of our knowledge has the same structure as the Mentalist explanation of our moral knowledge. Let's spell the constitutive expla- nation out more carefully. Suppose that some agent started out with the conjunction dispositions with a term `*', and only those dispositions about that term. She's dis- posed to infer that pp * qq is true from her belief that p and her belief that q, and to infer p from her belief that pp * qq is true. Suppose then that she changes her dispositions about `*', to the dispositions constitutive of competence with disjunc- tion. 1 Then the content of her thoughts have shifted { `*' now refers to disjunction rather than conjunction. The positivist's mental representations are similar. If the mental representation shifts, it's simply a representation of a dierent set of rules. Both the positivist and this idea about conjunction take the mental representation/ dispositions to determine what kinds of contents the agent is entertaining. It's thus no accident that the representation/ dispositions are reliable ways of forming beliefs about those contents. 2 I've noted that the positivist takes the agent's mental representation to be a 1 She becomes disposed to infer pp * qq from her belief that p alone, and is disposed to infer p from pp * qq only when she believes:q, etc. 2 Now the positivist does not endorse any particular thesis about competence with these representations. That's a central dierence between the two kinds of explanations. But the continuities are more illuminating. 95 reliable way of forming beliefs about her own set of rules. The inferential-role vindication of logical knowledge will predict that the agent's dispositions about a term are also a reliable way of forming beliefs about the content expressed. And it will make that prediction for the same reason that the Mentalist explanation does. If the dispositions weren't a reliable way of forming belief about that content, they wouldn't be beliefs about that content. They'd be beliefs about some other contents. In general, you can see the inferential-role account as providing a recipe for predicting what the positivist's Mentalism will predict. If you want to know how the Mentalist account satises some requirement on knowledge, gure out how the inferential-role account meets that requirement, and then generalize. What about more internalistic conceptions of justication? Suppose, for exam- ple, that justication depends on what's internally accessible: what you have access to through introspection or re ection. That is, you can't be justied in believing that p unless your grounds for believing that p are in some way accessible to you. Start with the inferential-role vindication of logical knowledge. The thing that you can access is something like understanding of the relevant term: you understand that, from your belief that p^ q, you can infer that p and you can infer that q. Importantly, this understanding is constituted by the set of dispositions that you have for the salient term. The positivist's Mentalism supports the same account. The thing that you can access is something like understanding of the relevant set of rules. That under- standing is constituted by your set of dispositions. Now you might have a number of worries about this account. But note that objections to this account are also objections to the inferential-role vindication of logical knowledge. Both appeal to understanding in the same way, and both give the same constitutive account of what it is to understand a term/ set of rules. But this sort of inferential-role vin- dication of logical knowledge is paradigmatically available to an internalist about justication. Since the positivist's Mentalism has the same structure, we should also see it as paradigmatically available to those internalists. We can be more brisk with other requirements on knowledge. Some philosophers think that safety is part of what distinguishes knowledge. Williamson's formulation is characteristic: \if one knows, one could not easily have been wrong in a similar case" (Williamson 2000, 147). 3 The positivist's Mentalism is well positioned to satisfy this constraint. In particular, the positivist will satisfy it whenever an agent avoids performance errors { that is, whenever an agent is consistent in the way that she takes natural facts to matter morally. When would an agent lack safe beliefs about some set of rules s i ? She would have to believe something false about s i at some nearby world. For example, she might believe that s i forbids -ing at the world, when it doesn't. But then there is a set of rules s i 0 that is an even better t for her dispositions: one that is just like s i except that it classies-ing as forbidden according to s i 0 . And the positivist will hold that her beliefs are about s i 0 , rather 3 This formulation needs to be taken with some care. It's a characterization of the structure of our reasoning about what is known, not something that that takes us from more basic primitives to conclusions about what is known. 96 than s i . (Bracketing performance errors is important here. Bracketing those errors is what guarantees that there is a set of rules that ts her dispositions even better.) The positivist'sMentalism does crisply predict that the agent's beliefs will be safe. Other philosophers take knowledge to require the exclusion of accident. Kiera Setiya, for example, argues for the following principle. K: When S knows that p, she knows it by a reliable method, and it is no accident that her method is reliable. (Setiya 2012, 96) Setiya takes inferential-role accounts of logical knowledge to satisfy this requirement. It's no accident that our methods for forming beliefs about `p^ q' is reliable. If they weren't reliable, our mental states wouldn't be about conjunction. The positivist's mental representations do satisfy this no-accident requirement. And they satisfy it for the same reason the inferential-role accounts satisfy it. Our dispositions determine what set of rules we're thinking about. So it's not an accident that they're a reliable guide to that set of rules. 5.2.2 The bad cases: where mental representations don't support knowledge In the last section, we saw that basic structure of the positivist's epistemology { how her mental representations have the right structure to support knowledge in a wide range of cases. But we can gain a better appreciation of the positivist's idea by considering the sorts of considerations that the positivist will take to undermine knowledge. That's what this section will do. I'll focus on two kinds of cases. The rst kind of case involves performance errors, where the agent's mental representation doesn't operate in a systematic enough way. In the second kind of case, the agent's mental representation doesn't by itself determine what set of rules the agent is thinking about. The character I considered in the previous section was idealized to exclude these two sorts of cases; this section describes what happens when the idealizations are relaxed. Let's start with performance errors. I might correctly classify some action in the actual world, but misclassify it at nearby worlds. Imagine, for example, that I accepted a set of rules that only treated three considerations as bearing on promise- keeping. saving a life counts in favor of breaking a promise but that consideration is undercut if the promise was itself to save a life preventing a disaster (hundreds of deaths) counts in favor of breaking a promise Imagine further that I promised to save A's life rather than B's life. I recognize that my set of rules does require me to save A's life. But at some nearby worlds, I forget about the second element of my list. I would think that promises are morally 97 irrelevant once there are any lives to be saved. Even though I do have a mental representation of this set of rules, my mental representation still misres in some cases. In this case, I wouldn't know that I ought to save A's life. My belief isn't safe, so it isn't knowledge. The previous section argued that these cases are only possible when there is a performance error. If there isn't a performance error, there will always be another set of rules that ts the dispositions even better. But you might wonder why performance errors make a dierence. After all, it's possible to construct a set of rules that takes the performance error into account: one that is partial and undened for the cases when there are performance errors. (In our example, the relevant partial set of rules would be undened about breaking a promise to save one life in favor of saving another life.) And the last section argued that my moral beliefs are about whatever set of rules best t my mental representation. So you might expect that what I was calling performance errors only reveal that the agent's set of rules is more partial than we initially thought. In considering this point, it's helpful to return to the comparison with idiolects. Imagine that I almost always acknowledge two readings of Mary left her money in the bank, but occasionally only acknowledge the nancial reading. It's a mistake to interpret me as speaking an idiolect where that sentence only expresses the nancial reading. That interpretation is mistaken, because it misses a systematic pattern about the way that I use my language { that I almost always acknowledge both readings. That systematic pattern is what makes cases where I only acknowledge one interpretation a performance error. The same point holds in the moral case. My dispositions can display systematic patterns without displaying perfect regularities. If we interpret divergences from perfect regularities with partial sets of rules, we'll miss those systematic patterns. To capture those systematic patterns, we should interpret my attitudes as about comparatively complete sets of rules. Performance errors happen when there are systematic patterns that fall short of perfect regularities. And performance errors can undercut knowledge. They can lead to false beliefs. Or they can prevent our beliefs from being safe { they can lead to mistakes in saliently similar cases. Or, for those who favor a more internal constraint on justication, they can partially undercut the agent's understanding of the set of rules, and undercut knowledge in some cases. That's the rst class of cases where an agent's belief falls short of knowledge. I now turn to the second class of cases. The last section idealized in two ways. It idealized away from performance errors. And we've just seen how to relax that idealization. The last section also idealized by supposing that the agent's mental representation totally determined what set of rules she is thinking about. But the positivist acknowledges that this supposition may also be an idealization { that other factors might matter as well. Comparatively simple sets of rules may be better interpretations. And sets of rules that incorporate comparatively natural properties may be better interpretations as well. The more these factors matter, the less knowledge the agent's mental repre- sentation will aord her. Think of the linguistic case. Maybe I have special and 98 privileged knowledge that red objects are colored is true because of my linguistic dispositions, about what red and colored refer to. But I don't have the same sort of special and privileged knowledge that woodchucks are groundhogs is true. The reason is that my linguistic dispositions don't settle what woodchuck and groundhog refer to. In this case, the two terms refer to the same thing because of facts external to the agent { that the most natural referent for the one is the same as the most natural referent for the other. So a preference for simpler interpretations and a preference for more natural interpretations are each another way for an agent's beliefs to fall short of knowledge Imagine a case where a preference for comparatively natural properties has that eect. Maybe that the property being-a-Homo-sapiens is much more natural than being-Caucasian. And suppose that Jeerson is disposed to think it's wrong to enslave Caucasians but permissible to enslave non-Caucasians, but he's much more condent in the former than in the latter. Now dispositions to make the latter judgment might be trumped by a preference for interpreting him as accepting a set of rules that only includes highly natural properties. Then the best interpretation of him might be that he accepts a set of rules that forbids enslaving Homo sapiens as the moral standard. Then Jeerson wouldn't be justied in believing that it's permissible to enslave non-Caucasians. His dispositions are not a reliable guide to his own set of rules, on this question. In fact, his dispositions about enslaving non-Caucasians is a systematically bad guide to his own set of rules on this question. An internalist about justication could say something similar. Jeerson's dispositions don't aord him understanding of his set of rules on this question, because they don't put him in contact with the considerations that matter for that set of rules. This sort of possibility illustrates the second way that an agent's beliefs can fall short of knowledge. A preference for naturalness or for simplicity can trump her dispositions. This possibility is quite dierent from the possibility of performance errors. Performance errors have to involve occasional departures from a systematic pattern. By contrast, this second kind of case can involve a systematic pattern in the agent's mental representation, but a pattern that departs systematically from the agent's set of rules. In either sort of case, though, the agent's mental representation fails to deliver justication, and so fails to deliver knowledge. This section has been very tentative. I've tried to illuminate the sorts of consid- erations that the positivist takes to matter for knowledge, by describing the good cases where an agent has moral knowledge, and the bad cases where she doesn't. But there are many dierent ways to work out the details. You might accept positivist realism, while also taking a preference for simplicity and naturalness to dramati- cally constrain which set of rules the agent is thinking about. In doing that, you're paring back how much moral knowledge ordinary agents have. It makes sense to do that if you're worried about the positivist being excessively liberal in attributing knowledge. If, by contrast, you're not worried about that point, it would make sense to develop a version of positivist realism where we're thinking about whatever sets of rules hew closest to our dispositions. Importantly, though, both of these options would retain the central advantage 99 that the positivist claims. Those advantages come because the positivist rejects what I've called the Standard Assumption. Standard Assumption If moral realism is true, moral knowledge requires knowledge of a property that is independent of and more fun- damental than our individual evaluative attitudes. Knowledge of a set of rules is not knowledge of that sort of property. So the Standard Assumption is false if it's comparatively hard to know about sets of rules. It's also false if it's comparatively easy to know about sets of rules. In other words, the questions explored in this chapter are important for appreciating the details of the positivist's view. But they don't change the philosophical signicance of the view. 5.3 Explaining the agent's beliefs I will close by discharging one nal assumption that I've been making. I've been as- suming that agents somehow end up with dispositions about something they accept to be the moral standard. Belief: if you promised to meet your friend for lunch but can save a life if you don't, x requires saving the life. Acceptance: x is the moral standard Belief: In every other case where you promised to meet your friend for lunch, x requires you to keep the promise. Acceptance: x is the moral standard But I haven't yet explained how agents end up with these dispositions in the rst place. For the positivist, this explanation is task for other disciplines. It's a task for disciplines who are concerned to explain the development of moral judgment { disciples like social psychology and social-cognitive neuroscience. This section will brie y show how to integrate the dierent approaches to explaining our moral dis- positions with positivist realism. I will focus on three basic tools for explaining how we developed these dispositions: those that appeal to our innate psychology, those that appeal to cultural evolution, and those that appeal to idiosyncratic modication of previous beliefs. Start with approaches that appeal our innate psychology. Certain norms of reciprocity are plausible examples. Maybe I'm disposed to believe that I should help you repair your house if you've helped me repair mine. On the approach developed in the previous section, that means that I'm disposed to have this pair of states. Belief: if you've helped me repair my house, x requires me to help you repair yours (at least in a certain range of cases). 100 Acceptance: x is the moral standard And why do we have this disposition? The answer is that certain minimal norms of reciprocity seem to innate to human psychology. 4 But an important question arises about integrating these sorts of explanations with the positivist's proposal. According to the positivist, moral beliefs are a pair of attitudes, which includes the attitude of accepting x as the moral standard. And it's not particularly plausible that having that second attitude is an innate feature of our psychology. For one thing, our primate ancestors likely had the same dispositions of reciprocity. And it's not very plausible that they were disposed to think about what they were morally obligated to do. Fortunately, though, the positivist doesn't need our early primate ancestors to be disposed to be thinking about moral obligation. She can think of the development of moral belief as coming in two stages. In the rst stage, our innate psychology gives us dispositions to act in certain ways { to reciprocate when other people act in a way that benets us, say. In the second stage, we come to integrate those dispositions with our other moral attitudes. Maybe we take certain promises to trump our dispositions to reciprocate in some but not all cases. It's only when we've integrate our reciprocity dispositions with our other moral judgments that the disposition to act becomes a disposition to have moral beliefs. One last point about innate dispositions. There are two importantly dierent ways that our innate psychology can shape our moral beliefs. I've just given an example where the dispositions directly determine some of our moral beliefs. We're immediately disposed to accept certain norms of reciprocity, say, and they shape our moral beliefs as we integrate other moral considerations with the norms of reci- procity. In other cases, though, our innate dispositions might indirectly determine the moral beliefs that we're disposed to have. We might have an innate disposition to form certain dispositions given one environment but to form dierent disposi- tions in another environment. Either kind of explanation will work for the positivist. The positivist just needs some explanation of how we develop our beliefs, and the development of those beliefs can be shot through with contingency. You might think that our innate dispositions totally determine the moral beliefs that we're disposed to have. If you think that, you can move on to the next section: you've already seen how the positivist will explain why we're disposed to have the moral beliefs we're disposed to have. But most people are not in that camp. They doubt that our innate dispositions are the whole story. To satisfy them, the positivist needs to point to other mechanisms that complete the story. I'll assume that cultural evolution is another important mechanism. Selection pressures have determined our innate dispositions; we have those dispositions because of our evolutionary history. In appealing to cultural evolution, I am supposing that selection pressures have also shaped the culture that we've developed. And there are important controversies about the way that selection pressures have done that. Think of the evolution of a disposition to help others. It's initially natural to think that this disposition evolved because it beneted the whole group. 4 [References to relevant literature] 101 Groups with more helpers survived better than groups with more freeloaders, so selection pressures favored groups with more helpers. But this natural thought is mistaken. The gene that disposes you to be a helper will be selected against, at least in some cases. The costs of helping accrue to those that help, while the benets of helping will accrue to the freeloaders. This fact can sometimes create a selection pressure for genes that dispose you to freeload, and against those those dispose you to help. There is, however, a case where there is selection pressure in favor of genes that dispose you to be a helper. The selection pressure exists if your having the gene increases the tness of others with the very same gene. Selection operates on genes, not on groups. 5 There is, in other words, a tight constraint on biological evolution: it has to be explicable in terms of selection on genes. Genes have two central features that make them explanatorily fundamental: they are discrete and faithfully replicated. It is controversial whether cultural evolution also runs through something gene- like. Some theorists insist that it does. Dawkins introduced the word meme for whatever things play the same explanatorily role in cultural evolution that genes play in biological evolution. Like genes, memes are supposed to be discrete and faithfully replicated. Susan Blackmore (1999), Richard Dawkins (1982), and Daniel Dennett (1995) all insist that memes must be what explains cultural evolution. Other theorists, by contrast, doubt that cultural evolution needs to operate on any- thing gene-like to be genuine. They think that pieces of culture can be the objects of selection even if they are neither discreet nor faithfully replicated. Peter Richardson and Robert Boyd helpfully articulate some processes that allow culture evolution to operate even in that dicult case (Richardson and Boyd 81). Cultural evolution is somewhat controversial because some doubt that culture consists of anything meme- like, and hold that evolutionary explanations must appeal to something gene-like. (Dan Sperber (1996) and Christopher Hallpike (1986) have both pushed these kinds of doubts.) Richardson and Boyd powerfully argue that evolutionary explanations need not depend essentially on memes. In any case, either conception of cultural evolution does what the positivist needs it to do. Suppose that cultural evolution happens through selection on memes. Then agents acquire many of their moral beliefs by internalizing memes, and those memes are what explains why they're disposed to believe what they're disposed to believe. For example, they internalize memes about the practice of promise-keeping from their society, and that internalization explains why they're disposed to have the beliefs about promise-keeping that they do. But the same point holds even if cultural evolution operates on something less gene-like. Agents would still acquire many beliefs by picking up the (non-meme) cultural variants that they encounter. They adopt those cultural variants for whatever reason they pick up on other pieces of culture { maybe prestigious members of the community espouse the cultural variant, or maybe the cultural variant increases the tness of those who adopt it, 5 It turns out that selection on genes can be precisely modeled as selection on groups, with the Price covariation equation. The important point is that group level explanations are legitimate only when they're isomorphic to gene level explanations. 102 or something else. So far, I've introduced two important mechanisms that explain why we'd be disposed to have the beliefs we in fact have: our innate dispositions, and cultural evolution. It's worth separating out another mechanism, too. Individuals sometimes innovate in their moral beliefs, even in no one else in their society has made the same innovation. Maybe they innovate to improve the consistency of their moral beliefs { maybe, say, everyone else thinks that it's morally permissible to enslave members of some races but not others, and the agent innovates in making the prohibition independent of race. That innovation changes what set of rules she's thinking about. (The innovator has changed the beliefs that she is disposed to have, and the positivist takes those dispositions to determine what set of rules she's thinking about.) There's plenty of room for such innovation within the positivist's framework, and such innovation can be perfectly rational. 6 This section has indicated the mechanisms that explain the development of our moral beliefs. The positivist relies on those mechanisms, because one of her starting assumptions is that agents are disposed to have moral beliefs. She doesn't herself explain those beliefs; she takes them as given. The mechanisms in this section explain why it's legitimate for her to do that. I've indicated three mechanisms that help explain the development of our moral beliefs. Dierent theorists might reasonably disagree about the amount of work that each mechanism does: whether work is evenly distributed between our innate dispositions, cultural evolution, and individual innovation, or if one of those does more than the others. The positivist is happy with any answer to that question. And she's also happy if some theorist wants to posit further mechanisms to explain the development of moral belief. Let's close this section with another perspective on the contribution the posi- tivist is making. The positivist is in part giving us an account of the relationship between scientic accounts of our moral belief and rst-order normative theorizing. Scientic accounts of moral belief, whether neurological, evolutionary, or whatever, are explaining how we ended up thinking about the set of rules that we in fact ended up thinking about. But the positivist holds that there's more to moral belief than just belief about the demands of some set of rules. She thinks moral belief also involves accepting the salient set of rules as the moral standard. So she can interpret normative ethicists as exploring the nature of that fundamental normative property. That's the reason why normative ethics compliments and doesn't compete with scientic accounts of moral belief. Each addresses questions the other leaves open. Recognizing this point gives us another perspective on the positivist's answer to the debunking challenge, back in chapter 2. Much of the intuitive force of the debunking challenge comes from the concern that scientic explanation of moral belief doesn't seem to leave any room for the sorts of questions the normative ethicists are asking. 6 This sort of innovation in moral belief is an important part of cultural evolution; Richardson and Boyd call it guided variation (115). But it's worth noting separately, because so much constructive work in rst-order moral philosophy involves these sorts of innovations. 103 The positivist allows us to see how there is plenty of room for those questions. The scientists are explaining how we developed the sets of rules we developed. But the sets of rules we've developed don't settle which set of rules is in fact the moral standard { that's an independent question that normative ethicists are investigating. Scientic explanations of moral belief compliment normative ethics. They don't compete. 5.4 New horizons One test for a theory is whether it opens up further questions that are fruitful to pursue. The approach taken in this chapter denitely does. Some philosophers note that there is something odd about moral knowledge formed by testimony { it seems to be missing something. If you understand why some moral claim is true, you seem to be in a better position than someone who knows that it's true because someone told them that it's true. It turns out to be interestingly dicult to explain this dierence. After all, both testimony and understanding seem to be enough for moral knowledge. So we can't say that relying on testimony is worse because it doesn't give you knowledge. But the positivist already allows that there are two dierent ways of believing some moral claim { by deference, or in the special privileged way that this chapter has been articulating. It would be interesting to explore whether this dierence allows the positivist to explain the intuitive dierence between knowledge by understanding and knowledge by testimony. At the very least, she has interesting new resources to bring to bear, and those resources seem to have the right structure to illuminate this question. Other philosophers insist that emotions can be a route to moral knowledge. Your anger at Mike's-ing can ground moral knowledge. The positivist has an ele- gant explanation of how that's possible, if the anger has something like the following structure. Anger: Mike's -ing violated x Acceptance: x is the moral standard. x4 gave the positivist's constitutive account of our moral knowledge. Our individual dispositions to believe determine which set of rules we're thinking about, so beliefs about that set of rules are pieces of knowledge. It's reasonable to expect, though, that the same kind of explanation will generalize to emotions. Our emotions { dispositional or occurrent { help determine which set of rules we're thinking of, and can give us knowledge about that set of rules for that reason. 7 7 Sentimentalist theories of moral knowledge that give an important place to the emo- tions face an important reference problem: they need to explain how our emotions come to be about moral properties. (Mark Schroeder (2008b) and Karl Schafer (2013) both note this problem.) The positivist answers this problem, in a way that can be cleanly integrated into an explanation of how emotions can ground moral knowledge. 104 Part II Part II: Arguing for positivist realism 105 A roadmap for what follows We've just nished Part I. I've argued that it would be nice for non-skeptical moral realists if a linguistic claim about presupposition is true. That linguistic claim leads to a view that I've called positivist realism. Chapters 1 { 3 have argued that positivist realism does have nice upshots for non-skeptical moral realists. I've ac- knowledged that positivist realism has further commitments, beyond the linguistic claim that is at its core. But Chapters 4 { 5 have argued that the linguistic claim re- ally is the crucial pivot for evaluating the view. The positivist's other commitments are all highly defensible. Parts II-III argue for the core linguistic claim at the heart of positivist realism. But the arguments that follow are complicated. And it's easy to lose track of how all the pieces t together. So I want to oer a quick roadmap. The rst goal in Part will be to get clear on one basic fact about presupposition: the fact that knowledge ascriptions don't require knowledge of the presupposition. This basic fact is the foundation for the rest of this Part. But this basic fact has gone unappreciated in the broader literature on presupposition. It has gone unappreciated because linguists and philosophers of language normally share an orthodox ideology about presupposition. And the shared ideology obscures that basic fact; it's very dicult to capture that basic fact given that shared ideology. As a result, these philosophers tend to doubt that the generalization is true. Problem 1: linguists and philosophers of language tend to accept an ortho- doxy ideology that conceals a basic fact about presupposition, a basic fact that is centrally important for the positivist. The very rst task in Part II will be to solve Problem 1. It will show that this basic fact really is a fact, and can bear the weight that the positivist puts on it. Chapter 6 carries out that task. And the second task in Part II will be to show that the positivist's core linguistic thesis is true. Moral discourse does carry just the presupposition that the positivist supposes. This task is important, because no one has evaluated this claim with that care it deserves. Problem 2: no one has investigated whether the realist's fundamental prop- erty is part of the presupposed, not-at-issue commitments of moral discourse. Chapters 7-8 will argue that the positivist's core linguistic thesis is true. These arguments will build from an assumption about presupposition that Chapter 6 has defended. 106 Part II is positive: it argues for a positive thesis about how presupposition works generally, and for a positive thesis about the presupposition of moral discourse. The next Part in the dissertation, by contrast, will be defensive. It takes up a central problem for the positivist. It's reasonable to doubt that the positivist does have an adequate account of moral disagreement. Imagine Immanuel and Jeremy disagreeing about some case of promise-keeping. They may well be thinking about dierent sets of rules. If they are, it's hard for the positivist to explain their disagreement. And if the positivist can't explain what they're disagreeing about, it's less clear that she has preserved the features of moral realism that make it philosophically interesting. Problem 3: it doesn't look like the positivist has an adequate account of moral disagreement. Part III will defend the positivist against this charge. 107 Chapter 6: Presuppositions are not entailments Abstract: This paper focuses on the way that presupposition trig- gers embed under attitude verbs. It has three main goals. It rst shows that we can't explain those embeddings just by drawing on the stan- dard tools that philosophers and linguists have traditionally used. The second goal is to introduce and develop a non-standard tool that does capture these embeddings. The central point of this discussion is that presupposition triggers behave very dierently under attitude ascrip- tions than most other kinds of linguistic constructions, in a way that philosophers and linguists have missed. And the third goal is to unpack the philosophical payos of this discovery. Several central debates in epistemology look very dierent. Utterances of (1a) and (1b) are standard examples of utterances that carry presuppositions. (1a) John has stopped wearing plaid. (1b) John hasn't stopped wearing plaid. Make either utterance, and your hearer will take you to assume that John used to wear plaid. That is, uses of stop trigger a presupposition about the past. A wide range of philosophically interesting constructions trigger presuppositions. Ascrip- tions of knowledge presuppose that the complement is true; the use of a denite description presupposes that something satises the description. Other cases are more contested. 1 But the class of philosophically interesting terms that are presup- position triggers is pretty big, whatever its exact contours. Presupposition triggers raise a range of dicult questions. This paper focuses on one range of questions, about the way that presupposition triggers embed under attitude verbs. It has three main goals. The rst goal is negative: I show that 1 Several philosophers have argued that presupposition is important for understanding normativity (Alex Silk (2016, 2017b) and Stephen Finlay (2014)), epistemology (Daniel L opez de Sa (2008, 2015) and Matthew ?), and as well as more linguistic topics, like complex demonstratives (Glanzberg and Siegel (2006). 108 we can't explain those embeddings just by drawing on the standard tools that philosophers and linguists have traditionally used. That's what I show in Part I. Part I also argues, more strongly, there's no way to modify those standard tools to do what's needed. After making this negative point, I switch to my second goal, which is to introduce and develop a non-standard tool that does capture these embeddings. That's what I'll do in Part II. The central point of Parts I and II is that presupposition triggers behave very dierently under attitude ascriptions than most other kinds of linguistic construc- tions, in a way that philosophers and linguists have missed. Part III describes the philosophical payos of this discovery. Several central debates in epistemology look very dierent. And even before we see the details, we should expect this discovery to have signicant upshots in epistemology. Knows is an attitude verb. And I'm arguing that philosophers have badly misunderstood the way that attitude verbs work when the complement contains a presupposition trigger. Part I: Our toolkit needs expanded 6.1 Background about presupposition This paper is about the class of presupposition triggers. The best way to introduce this class is as a class of linguistic constructions that behave in a distinctive way when embedded. In order for a construction to be a presupposition trigger, it needs to be associ- ated with a commitment that ordinarily projects when embedded. A commitment projects when two conditions are satised. First, a sentence S that contains the construction needs to be associated with a commitment p. That is, normal, sincere utterances of S need to be understood as committing the speaker to p. Second, a range of canonical embeddings of S also need to be associated with the very same commitment p. That is, normal, sincere utterances of pnot-Sq, pperhaps Sq, and pif S, then Tq need to ordinarily be understood as committing the speaker to p. For example, John stopped wearing plaid 2 is associated with the commitment that John used to wear plaid. And that commitment is also associated with the relevant embeddings: sentences like John [didn't stop/perhaps stopped] wearing plaid and If John stopped wearing plaid, his dog is happier. 3 The rst hallmark of presupposition triggers is that they are associated with a commitment that projects in this way. As it happens, though, this hallmark is not enough by itself to single out the class that interests us. Other linguistic con- structions exhibit projection behavior, but are not best thought of as presupposition triggers. Non-restrictive relatives are one example. Mary's mother, who likes Mary, lives near her is associated with the commitment that Mary's mother likes her. But 2 I throughout italicize terms and sentences that I intend to mention. 3 Gennaro Chierchia and Sally McConnell-Ginet (2000) helpfully discuss this sort of test for projection behavior in more detail. 109 this commitment also projects through these embeddings. For example, it's also a commitment of Mary's mother doesn't live near Mary, who likes her. So the class of presupposition triggers is a proper part of the class of projective content. It diers from the broader class in part because the projection behavior is defeasible. Even though the commitment projects in ordinary contexts, there are special contexts where it doesn't. Imagine me saying: Tom never wore plaid, so he didn't stop wearing plaid. Even though stop triggers a commitment that usually projects through not, it doesn't in this case. (The discourse is perfectly felicitous, and it wouldn't be if the commitment projected.) Non-restrictive relatives pattern dierently. Imagine me saying: Mary doesn't like her mother, so Mary's mother doesn't live near Mary, who likes her. That discourse is infelicitous. It's infelicitous because the commitment that Mary likes her mother still projects from the second sentence, and that commitment is inconsistent with the rst sentence. The second hallmark of presupposition triggers is that their projection behavior is fairly defeasible, in the way that we just saw { much more defeasible than the projection behavior of non-restrictive relatives. 4 To understand presupposition triggers, we have to understand why they project in this way. It's important to do that, because it's not just an idiosyncratic fact about some presupposition triggers that they project. Rather, any term that car- ries the same commitments as a presupposition trigger always exhibits the same projection behavior. Take stop. Sentences containing stop are associated with two commitments: a commitment that something was happening in the past, and a commitment that that thing isn't happening in the present. It turns out that any term associated with those two commitments exhibits the same projection behav- ior. That is, when that term is embedded under a negation operator, the rst commitment projects, and the second one is interpreted under the negation opera- tor. This point holds across natural languages, even typographically distinct ones. 5 This robust generalization calls out for explanation. 6.1.1 Pragmatic explanations of presupposition Robert Stalnaker has developed a powerful and in uential explanation of presuppo- sition triggers that predicts that cross-linguistic uniformity of presupposition pro- jection. 6 He starts with a simple conjecture about the semantics of presupposition triggers like stop. (1a) John has stopped wearing plaid. For him, (1a) expresses a conjunctive proposition, the proposition that John used 4 Chris Potts (2005) helpfully works through some of those details (12 { 36). In pointing to these two hallmarks, I'm just giving uncontroversial empirical hallmarks of the class. It's denitely not the deepest theoretical description of what ties the class together. 5 One of the central contributions of Tonhauser et al. (2013) is to show that this generalization does hold generally, even across typographically distinct languages. 6 See Stalnaker (1973, 1974, 2002, 1999). 110 to wear plaid, and isn't anymore. 7 So what does (1b) express? (1b) John hasn't stopped wearing plaid. Stalnaker's answer is simple. (1b) expresses the proposition formed by composing the semantic contribution of not with the semantic contribution of (1a). For him, then, the semantics of presupposition triggers works in exactly the same way as the semantics for other linguistic constructions. Why does not compose with presup- position triggers in the same way across languages? For the same reason that not composes with an arbitrary sentence in the same way across languages. Let's turn now to another important feature of Stalnaker's account. For him, presupposition is fundamentally something that speakers do, rather than a funda- mental property of sentences or utterances. And for him, a speaker presupposes a proposition if she takes it to be part of the common ground { that is, part of what parties to the conversation accept, and take others to accept, and so on. The notion of utterances presupposing a proposition is less fundamental. An utterance of a sentence S presupposes p if and because speakers who use S tend to presuppose p. The task of explaining the cross-linguistic uniformity of presupposition projec- tion then reduces to a simpler task. We need to explain why we tend to interpret someone who asserts the proposition that (1b) expresses by using (1b) as presup- posing something about the past. Stalnaker executes this task by drawing on general facts about rational commu- nication. In fact, he explores several options for doing so. One option focuses on questions. He asks about the conditions under which it's interesting to investigate whether (1a) is true or (1b) is true. Maybe it's normally only interesting to ask whether John used to wear plaid and isn't anymore if you already assume that John used to wear plaid, and want to gure out whether his clothing habits have changed. As Stalnaker puts the idea, \the propositions that P and that Q may be related to each other, and to common beliefs and intentions, in such a way that it is hard to think of a reason that anyone would raise the question whether P, or care about its answer, unless he already believed that Q" (Stalnaker 1974, 205). This kind of theory is usually called a pragmatic theory of presupposition, be- cause it builds from Gricean assumptions about rational communication. 8 A central advantage of pragmatic theories of presupposition is that they explain the cross- linguistic uniformity of presupposition projection. The explanation just sketched 7 In this paper, I'll simplify by assuming that sentences like (1a) express propositions, rather than something more complicated, like propositional matrices. I'm making this simplication just for terminological reasons. If you prefer, read occurrences of \the propo- sition that (1a) expresses" as \the propositional matrix that (1a) expresses", or whatever your preferred ideology is. 8 Proponents of this approach include at least Jay David Atlas (2005), Steven Boer and William Lycan (5976), Stephen Levinson (1983), Ruth Kempson (1975), Philippe Schlenker (2010), Mandy Simons (2001), Robert Stalnaker (1973, 1974, 2002, 1999), Judith Tonhauser et al. (2013), and Deirdre Wilson (1975). 111 also explains why (utterances of) translations of (1b) in other languages presup- pose the same thing. Translations of (1b) express the same proposition that (1b) expresses. And the Stalnakerian explanation is an explanation of why the use of a sentence that expresses something about the past and present both will tend to presuppose the proposition about the past. 6.1.2 Generalizing to attitude reports The rest of this paper will assume that some pragmatic theory of presupposition is right. In addition, I'll also assume Stalnaker's conception of speaker presupposition { that a speaker presupposes a proposition when she takes it to be common ground. This conception of speaker presupposition is the most familiar one, so it's the best conception to assume as the baseline in describing the innovation that I want to make. 9 Our central question will be about presupposition triggers under attitude re- ports, like (2). (2) I'm glad that Tom has stopped wearing plaid. We'll explore what uses of (2) communicate. The obvious answer is that uses of (2) ordinarily communicate the proposition that the sentence semantically expresses. Given Stalnaker's simple semantics for presupposition triggers, for example, (2) would express that I'm glad that Tom used to wear plaid and isn't wearing plaid now. We should denitely start by supposing that that proposition is what's communicated. It's the simplest and most plausible answer. But the use of a presupposition trigger may throw a wrench into this simple answer. And indeed that's what I will suggest. Consider a comparison. If I sar- castically say Bob is a ne friend, I'm not communicating the proposition that the sentence semantically expresses. In fact, I'm communicating the opposite. Now there is immense controversy about the relationship between what's semantically expressed and what's communicated. Some theorists think that the two are very closely related, and that sarcasm illustrates an uncommon phenomenon. Other theorists allow for more systematic gaps between the the two. 10 I'll suggest that the use of a presupposition trigger in (2) is often like sarcastic intonation { that it shifts the proposition communicated, so that the proposition communicated isn't the proposition semantically expressed. Now allowing for sys- tematic gaps between the proposition semantically expressed and the proposition 9 As it happens, I don't myself think that this conception of speaker-presupposition is the best one. I myself favor a view that builds from the framework that Craige Roberts (2012) has developed, as sketched in, for example, Simons et al. (2011). 10 Views where there's a systematic gap between the two include at least Kent Bach (1994), Anne Bezuidenhout (1997), Robyn Carston (2002), Fran cois Recanati (2004), and Scott Soames (2009b), as well as the tradition descending from Dan Sperber and Deirdre Wilson (5986).. Jason Stanley (2007) argues for a more traditional picture. 112 communicated expands our explanatory toolkit. We can explain some readings of (2) as the readings where we hear the proposition semantically expressed, and other readings as the readings where the proposition communicated isn't the proposition semantically expressed. To show that our explanatory toolkit does need to be ex- panded in this way, we need to show that the proposition semantically expressed isn't plausibly what's communicated in a central range of cases. I'll argue that we do need to expand our toolkit in just this way. In fact, I'll argue for a particular way of expanding the toolkit, which I'll call the Fission theory. according to the Fission theory, I can use (2) without thereby communicating that I'm glad about the past and present both. Instead, I can use it to communicate that I'm glad just about the present. In those cases, the only proposition communicated about my gladness is that I'm glad that Tom is not wearing plaid now. The more traditional theory is that we don't need to expand our toolkit in this way. I'll call this theory the Atom theory. according to the Atom theory, normal, sincere uses of (2) only communicate the proposition semantically expressed. 11 I call the second theory the Atom theory because it treats the semantic value of the complement as an atom. If it's right, competent speaker-hearers don't \break up" that proposition in interpreting uses of (2). The rst theory does take the semantic value to get broken up in interpreting uses of (2). That's why I call it the Fission theory. 12 6.2 Why the Atom theory is false The Atom theory is more ambitious than the Fission theory. It attempts to explain the data by using a narrow toolkit { without supposing that the semantic value of the complement is \broken up" in interpreting some uses of (2). This section shows that this narrow toolkit can't work. We need to broaden our toolkit in the way that Fission theory does. 11 The Atom theory is the wholly orthodox view among theorists who write about presupposition; it's tacitly assumed by at least Paul Elbourne (2005), Bart Geurts (1998), Irene Heim (1992), and Judith Tonhauser, David Beaver, Craige Roberts, and Mandy Simons (2013). 12 This disagreement is not a disagreement about presupposition projection; it's fully compatible with any theory of presupposition projection that's compatible with a prag- matic theory of presupposition. If a pragmatic theorist can make good on the predic- tions of the constraint-like conception from theorists like Irene Heim (1982, 1983), David Beaver (2001), or Daniel Rothschild (2011), the Fission theorist can too { and ditto for the anaphoric conception from theorists like Rob van der Sandt (1992) and Bart Geurts (1999). 113 6.2.1 A puzzle Being glad that Tom stopped wearing plaid doesn't require being glad that Tom used to wear plaid. The Atom theorist will try to explain this important fact in several dierent ways. This section shows that none of these explanations can be correct. Each explanation forces the Atom theorist to make incorrect predictions about another important fact. The Atom theorist can explain the rst fact. She won't take it to be a fact about stop, but rather a more general fact about gladness. After all, you don't need to be glad about all the entailments of S to be glad that S. (3a) I'm sad that you're buying a car. But I'm still glad that you're buying a hybrid car. That minimizes the environmental damage of your choice. If you're buying a hybrid car, you're buying a car. But I can be glad about the former even though I'm not glad about the latter. There has to be some explanation of why you can be glad that S without being glad about everything that S entails. The Atom theorist will try to use that explanation to capture the behavior of presupposition triggers under attitude reports. I claim that this strategy can't work. There is an important dierence between the way that presupposition triggers behave under attitude verbs and the way that other kinds of constructions behave { a second fact about presupposition triggers that doesn't hold of other kinds of constructions. The second important fact about presupposition triggers is that they are associated with a privileged commitment: a commitment that you have to be glad about in order for the gladness report to be appropriate. The privileged commitment associated with Tom has stopped wearing plaid is that Tom doesn't wear plaid. You can't be glad that Tom has stopped wearing plaid unless you're glad about that privileged commitment. As evidence for this fact, contrast (4a) and (4b): (4a) James is sad that Tom used to wear plaid. He's still glad that Tom has stopped wearing plaid. (4b) James is sad that Tom doesn't wear plaid. * He's still glad that Tom has stopped wearing plaid. This fact is important, because other kinds of linguistic constructions behave dierently. They do not display the same privileged commitment. Imagine that you were torn between buying a hybrid car and a hybrid Vespa, and I'm relieved that you opted for the former. (I think that the former is more normal, and makes you less of a hippie.) Then (3b) is true. (3b) I'm sad that you're buying something hybrid. But I'm still glad that you're buying a hybrid car. That's more normal. In other words: (3a) is true if the speaker wants you to minimize damage to the 114 environment, and (3b) is true if the speaker wants you to maximize damage. So you're buying a hybrid car doesn't carry any privileged commitment that I need to be glad about in order to be glad that you're buying a hybrid car. The presupposition trigger strop behaves in very dierent way than we see in (3a) and (3b). The goal of this section is to show that the Atom theorist can't explain this dierence. 6.2.2 Attitude ascriptions without privileged commitments The Atom theorist will try to explain the way that stop embeds under attitude verbs by appeal to general facts about attitude verbs, rather than by appeal to facts about the presupposition trigger. So to evaluate this claim, we have to understand how attitude verbs behave. We've seen that you can be glad that S without being about everything that S entails. I'll describe this fact as the fact that is glad isn't closed under (logical) consequence. This section describes three representative accounts of this fact. It then identies a baseline feature that these accounts share, and it explains why any plausible account of is glad has to share that baseline feature. The next subsection will show that that baseline feature prevents any of those accounts from explaining the way that presupposition triggers embed under attitude verbs. In other words: I'll be arguing that the two pairs in the previous section illustrate two distinct phenomena. You can't use the same tool to capture both. It's not just ascriptions of gladness that fail to be closed under consequence. (5) I want you to buy a hybrid car. (5) can be true even if I don't want you to buy a car, but believe that you will anyway. And wants is in several ways simpler than is glad. 13 So I'll use it to introduce three representative accounts of failure of consequence. The rst account starts from a Hintikka-style semantics for attitudes ascriptions. pA wants that Sq is true i all the worlds compatible with what A wants are worlds where the proposition that S is true. But this rst account goes beyond the basic Hintikka-style semantics, by further restricting the worlds it quanties over. One version of this Hintikka-style approach further restricts to worlds compat- ible with what the matrix subject believes. As a rst pass, take pA wants that Sq to be true i all the worlds compatible with what A believes and also what A wants are worlds where the proposition that S is true. To illustrate, go back to (5). Model my wants with a set of worlds that includes some worlds where you don't buy a car. An unmodied Hintikka-style semantics then predicts that (5) is false. There are worlds compatible with my wants where the complement is false. The modied Hintikka semantics, by contrast, correctly predicts that (5) is true. We no longer quantify over worlds where you don't buy a car, because we only quantify 13 For example, it's not factive. 115 over worlds compatible with my beliefs. And my beliefs entail that you do buy a car. 14 I'll call this rst account a modied-Hintikka account. The second account assumes that propositions have more internal structure than mere sets of worlds. Propositions are either built up out of objects, properties, or relations, or built up out of concepts for those things. So even if S entails T, there is no interesting and general relationship between the truth of pA wants that Sq and the truth of pA wants that Tq. 15 And there is a very general reason why there isn't any interesting relationship between the two: entailments between the complements are just irrelevant. For one thing, every sentence entails that arith- metic is incomplete, trivially. And proponents of structured propositions already deny that there is any interesting relationship between having an attitude towards an arbitrary proposition (say, that it's raining now), and having that attitude to the proposition that arithmetic is incomplete. I'll call this second explanation a structured-proposition account. A third explanation builds from the sort of alternative semantics that Mats ? has developed. The context combines with (5)'s complement to supply a set of alternatives: in our case, maybe the setfyou buy a hybrid car, you buy a non- hybrid carg. (5) is true i you prefer the proposition expressed by the complement to the alternatives. So (5) can be true even while I want you to buy a car is false. The set of alternatives for the later could befyou buy a car, you don't buy a carg. And you can prefer the former to the later even while preferring a hybrid car to a non-hybrid car. I'll call this an alternative-semantics explanation of failures of closure. 16 All three of these accounts have signicant promise. The rest of this paper will assume that one of them is correct. Moreover, all three accounts extend straight- forwardly to ascriptions of gladness. Importantly, these explanations all share one baseline feature. They all pre- dict symmetries among entailments. Wanting/ being glad that S doesn't require wanting/ being glad about any privileged entailment of S. (Symmetries among Entailments) If S asymmetrically entails T (that is, if S entails T and T does not entail S), it's possible to [want/be glad] that S, without [wanting/being glad] that T. It's easy to see that the structured-propositions account leads immediately to the Symmetry claim. Of course it's possible to V that S without V-ing that T { they're just dierent attitudes! 14 Irene Heim (1992) develops a sophisticated version of this account; the next page will dig into some further important features of her account. 15 The rst approach, by contrast, still predicts interesting relationships between the two. Suppose that A doesn't believe that T. Then pA wants that Sq is true only if pA wants that Tq is true. 16 Elizabeth ? develops this kind of proposal; it's used to good eect in Pranav and Hacquard (2013), and it's related to some ideas in Simons et al. (forthcoming). 116 The alternative-semantics account also leads to the Symmetry claim. If S asym- metrically entails T, there's some further sentence U such that pU^ Tq is true i S is true. It's possible to prefer (U^ T) to (:U^ T), while also preferring:T to T. In that case, pI want that Sq can be true (if you prefer (U^ T) to (:U^ T)) { even while pI want that Tq is false (because you prefer:T to T). The modied-Hintikka account will also vindicate this Symmetry claim. Fail- ures of closure happen when the matrix subject already believes some entailment. In those cases, the subject doesn't need any positive attitude towards what she believes. We would expect this account to predict Symmetries among Entail- ments: just x on a subject who already believes some entailment. The details are actually somewhat complicated, but this kind of account will also end up vindicating Symmetries among Entailments. 17 As a result, basic structural features of these three accounts force them to vindicate Symmetries among Entailments. Symmetries among Entailments If S asymmetrically entails T, it's possible to [want/be glad] that S, without [wanting/being glad] that 17 The rst pass version of the modied-Hintikka account actually vindicates the Sym- metries claim. But the reason why it vindicates that Symmetries claim is the reason why it can only be a rst-pass version of the account. If you already believe that T, then the rst pass account takes you to want that T. After all, there just aren't any worlds compatible with what you believe, so all the worlds compatible with your beliefs and your wants are (trivially!) worlds where T is true. This result shows that the rst-pass version can only be the rst pass. Even if I believe you'll buy a car, I can want you to not buy it. We thus need to rene the rst-pass presentation of the modied-Hintikka account. To do that, we should hold that pA wants that Sq is true i A prefers each similar world where S is true to any similar world where:S is true, and add that worlds compatible with what she believes are more similar than worlds that aren't compatible with what she believes. Then belief that you'll buy a car isn't enough for wanting you to buy a car. We compare worlds where you don't get a car with worlds where you do. And I can prefer a world where you don't get a car. Moreover, I want you to get a hybrid car can still be true, even if I don't want you to get a car. We compare worlds where you don't get a hybrid car with worlds where you do. Crucially, though, there are some worlds compatible with what I believe where you don't get a hybrid car, where you get a gas-guzzler instead. This sentence is true because I prefer worlds where you get a hybrid to those worlds. (Irene Heim (1992) develops this kind of proposal, inspired in part by some ideas from Stalnaker (1984).) The modied-Hintikka account does predict Symmetries among Entailments. If S asymmetrically entails T, there's some further sentence U such that pU^ Tq is true i S is true. Now suppose that my preferences are: (U^:T) (:U^:T) (U^ T) (:U ^ T). Suppose further that I believe that T. The sentence pI want that Sq is then true. There are worlds compatible what I believe where S is true (worlds where U is true), and worlds where it is false (where U is false). So we only compare worlds that are compatible with what I believe, and all those worlds are worlds where S is true. Crucially, though, the sentence pI want that Tq is false. We have to look outside the worlds compatible with what I believe to nd a world where:T is true. And I prefer any of those worlds to a world where T is true. 117 T. But I come to praise these explanations, not bury them. Any empirically plausible theory of attitude verbs will vindicate Symmetries among Entailments. Go back to the earlier pair. (3a) I'm sad that you're buying a car. But I'm still glad that you're buying a hybrid car. That minimizes environmental damage. (3b) I'm sad that you're buying something hybrid. But I'm still glad that you're buying a hybrid car. That's more normal. In the rst sentence, I'm glad about the option that does the least environment damage. In the second one, I'm glad about the option that does the most. It's thus an example of the readings that Symmetries among Entailments would predict. In general, then, a theory of attitude verbs is empirically plausible only if it vindicates Symmetries among Entailments. It's a virtue of all three of our accounts that they do. 6.2.3 Presuppositions are dierent Let's recap. The overall goal of this section (x2) is to argue against the Atom theory of presupposition triggers. I've already noted that you can be glad that Tom stopped wearing plaid without being glad that he used to. We've seen how the Atom theorist can capture this fact, by drawing on the tools just introduced. This section shows that, in doing that, the Atom theorist is forced to make mistaken predications about another important fact. I'll conclude that there's no way to make the Atom theory work. Go back to this pair: (4a) James is sad that Tom used to wear plaid. He's glad that Tom has stopped wearing plaid. (4b) James is sad that Tom doesn't wear plaid. * He's glad that Tom has stopped wearing plaid. The minimal pair illustrates a crucial asymmetry among the commitments triggered from the complement. Let S 1 = Tom has stopped wearing plaid (Core Asymmetry) S 1 carries some commitment T 1 such that: it's impossible to be glad that S 1 unless you're glad that T 1 . The commitment T 1 is that Tom doesn't wear plaid. (4b) is inappropriate because you can't be glad that S 1 unless you're glad that T 1 . And the rst conjunct in (4b) rules out your being glad that T 1 . 118 I deny that Atom theorists can explain this Asymmetry. According to the Atom theorist, being glad that Tom stopped wearing plaid involves being glad about a conjunctive proposition, about Tom's past and present both. (4a)'s felicity shows that you can have that attitude without being glad about Tom's past. The most natural way for the Atom theorist to capture this fact is to trace it back to the facts about attitude verbs described inx2.2 { that you don't need to be glad about all of S's entailments to be glad that S. But in oering this explanation, the Atom theorist seems to predict that (4b) is also felicitous. That prediction is wrong. The simplest version of the Atom theory doesn't explain the Core Asymmetry. Now Atom theorists do have a standard way to extend the tools fromx2.2 to explain the Core Asymmetry. They start with the plausible observation that we interpret the matrix subject of an attitude ascription as believing presuppositions of the complement. That's why pA V-es that Sq doesn't require A to V S's presup- positions: we interpret A as already believing S's presuppositions. This strategy can capture some instances of the Core Asymmetry. An utterance of pA hopes that Tom has stopped wearing plaidq is only appropriate if A already believes that Tom used to wear plaid. And you can't hope what you believe. So that utterance is appropriate only if A hopes that Tom doesn't wear plaid. 18 But this strategy simply cannot capture the Core Asymmetry in full generality. Stipulate that James knows that Tom used to wear plaid, and also knows that he doesn't wear plaid now. Even given that stipulation, the Core Asymmetry still holds { that is, (4a) is still felicitous and (4b) is infelicitous. (4a) James is sad that Tom used to wear plaid. He's glad that Tom has stopped wearing plaid. (4b) James is sad that Tom doesn't wear plaid. * He's glad that Tom has stopped wearing plaid. Sadness that S and gladness that S are both compatible with knowledge that S. As a result, the Atom theorist simply has to fall back on the tools fromx2.2. And those tools just can't explain the Core Asymmetry. Philosophers and linguists have missed this point in part because their diet of examples is too narrow. Matrix belief in the presupposition plus the tools from x2.1 might capture the Core Asymmetry for attitudes that are incompatible with belief. But it just can't work in full generality. It can't capture emotive factives, because they're perfectly compatible with belief, and indeed perfectly compatible with knowledge. The examples in (4a) and (4b), with presupposition triggers, thus illustrate a dierent phenomenon than the examples in (3a) and (3b), without presupposition triggers. (3a) I'm sad that you're buying a car. But I'm still glad that you're buying a hybrid car. That minimizes environmental damage. 18 Paul Elbourne (2005), Bart Geurts (1998), and Irene Heim (1992) all adopt this kind of idea; Heim is particularly explicit in doing so. 119 (3b) I'm sad that you're buying something hybrid. But I'm still glad that you're buying a hybrid car. That's more normal. And there are highly general reasons the two pairs illustrate distinct phenomena. Any viable account of (3a) and (3b) will predict Symmetries among Entail- ments. Symmetries among Entailments If S asymmetrically entails T, it's possible to [want/be glad] that S, without [wanting/being glad] that T. Symmetries among Entailments is just a description of the data about the rst minimal pair. But an account that predicts Symmetries among Entailments can't explain the Core Asymmetry. I conclude that the Atom theory can't capture the Core Asymmetry, at least in fully generality. So the Atom theory has to be wrong. Before moving on, though, I turn to one objection on behalf of the Atom theorist. The objection is that I've gotten the empirical facts wrong. Imagine that James is sad that Tom is neglecting his parent's wishes for him to wear plaid, so he's sad that Tom doesn't wear plaid. James is also glad that Tom is listening to his partner, who doesn't want Tom to wear plaid { so James is also glad that Tom has stopped wearing plaid. In that case, both sentences in (4b) are appropriate. So the Atom theory is correct to allow (4b) to be felicitous. This response is too glib, and we can reformulate the Core Asymmetry to avoid it. The imagined situation is also a situation where James is sad that Tom doesn't wear plaid and also glad that Tom doesn't wear plaid. That is, it's a situation where you're sad and glad about the very same proposition. That combination of attitudes is possible when you're sad for one reason and glad for a dierent reason. James is sad that Tom doesn't wear plaid because of what Tom's parents want, and he's glad that Tom doesn't wear plaid because of what Tom's partner wants. Crucially, though, James can't be sad that Tom doesn't wear plaid and glad for the very same reason that he doesn't wear plaid. So we can just reformulate our minimal pair, and observe the same asymmetry. (4a) James is sad that Tom used to wear plaid { and he's glad for the very same reason that Tom has stopped wearing plaid. (4b) James is sad that Tom doesn't wear plaid { * and he's glad for the very same reason that Tom has stopped wearing plaid. The Atom theorist can't explain this data. I conclude that the Atom theory just can't be made to work. In the next section, I'll show that we can capture the Core Asymmetry by abandoning the Atom theory. Before doing that, though, I want to note that this section is describing a genuine and distinctive phenomenon. Start by noting that this phenomenon happens very distinctively and systematically. First, it happens systematically with emotive factives. 120 (7a) James is sad that Tom used to wear plaid. But still he's [happy/relieved/pleased] that Tom has stopped wearing plaid. (7b) James is sad that Tom doesn't wear plaid. * But he's still [happy/relieved/pleased] that Tom has stopped wearing plaid. (7c) James is glad that Tom used to wear plaid. But still he [is sad/is angry/is upset/regrets] that Tom has stopped wearing plaid. (7d) James is glad that Tom doesn't wear plaid. * But he's still [is sad/is angry/is upset/regrets]that Tom has stopped wearing plaid. Second, it happens systematically with other kinds of attitudes. (7e) James doesn't [desire/expect/conjecture/hope/fear] that Tom used to wear plaid. But he does [desire/expect/conjecture/ hope/fear] that Tom has stopped wearing plaid. (7f) James doesn't [desire/expect/conjecture/hope/fear] that Tom doesn't wear plaid. * But he does [desire/expect/conjecture/ hope/fear] that Tom has stopped wearing plaid. Third: this same pattern happens systematically, with every plausible candidate for a presupposition trigger. Start: James is sad that Tom hasn't worn plaid before. But he's still glad that Tom has started wearing plaid. James is sad that Tom does wear plaid. * But he's still glad that Tom has started wearing plaid. Also: James is sad that Bill had dinner in New York. But he's still glad that Tom also had dinner in New York. James is glad that Bill had dinner in New York. He's sad that Tom had dinner in New York. * But he's still glad that Tom also had dinner in New York. Knows: James is sad that Tom's mom died. But he's still glad that Tom knows that his mom died. James is sad that Tom knows that his mom died, if she did. * But he's still glad that Tom knows that his mom died. 121 Go on constructing similar examples for whatever other presupposition triggers you like. You'll nd the same contrast. When we've appreciated this kind of point, we can see it popping up in other ways, too. For example, (6) is always true, even when asserted discourse-initially. (6) If I'm glad that Tom doesn't wear plaid, then I'm glad at the very least that Tom has stopped wearing plaid. (Maybe I'm also glad that he never wore plaid.) This fact contrasts very sharply with the same sort of example without a presup- position trigger. (8a) and (8b) aren't always true. (8a) If I'm glad that you're buying a car, then I'm glad at the very least that you're buying a hybrid car. (Maybe I'm also glad that you're buying something hybrid.) (8b) If I'm glad that you're buying something hybrid, then I'm glad at the very least that you're buying a hybrid car. (Maybe I'm also glad that you're buying a car.) Imagine the speaker who wants to maximize damage to the environment: (8a) isn't true in her mouth. Then imagine the speaker who wants to minimize damage: (8b) isn't true in her mouth. The contrast between (6) and (8a)/ (8b) is another illustration of the central phenomenon I'm describing. (6) is appropriate because you can only hope that Tom has stopped wearing plaid if you hope that he doesn't wear plaid. It doesn't matter if you hope that he used to. There is, in other words, an asymmetry among the entailments of the complement. And the examples in (8a) and (8b) do not involve the same asymmetry. Again, the Atom theorist can't capture this asymmetry. 6.3 Fission explanations of the Core Asymmetry Let's recap.x1 contrasted two dierent theories about presupposition triggers under attitude reports, as in (2). (2) I'm glad that Tom has stopped wearing plaid. theFission theory holds that I can use (2) without thereby com- municating that I'm glad about the past and present both. In- stead, I can use it to communicate that I'm glad just about the present. In those cases, the only proposition communicated about my gladness is that I'm glad that Tom is not wearing plaid now. The more traditional theory is that we don't need to expand our toolkit in this way. I'll call this theory the Atom theory. 122 the Atom theory holds that the cases that the Fission theory is about are still cases where uses of (2) only communicate the proposition semantically expressed. x2 argued that the Atom theory can't work, because it can't explain the Core Asymmetry. Let S 1 = Tom has stopped wearing plaid (Core Asymmetry) S 1 carries some commitment T 1 such that: it's impossible to be glad that S 1 unless you're glad that T 1 . This section outlines how Fission theorists can explain this Asymmetry. Even before we get into the details, we should expect Fission theorists to explain this Asymmetry. Uses of S 1 are associated with two commitments: one commitment about Tom's past activity, and one commitment about his present inactivity. If the Fission theory is true, being glad that S 1 doesn't require being glad about both commitments. It only requires being glad about Tom's present inactivity. So of course one of S 1 's commitments is privileged! We don't even interpret the other one as something that the person is glad about. So if the Fission theory is true, it smoothly predicts the central pieces of data that we've set out to explain. As a result, the Fission theorist has very dierent burdens that the Atom the- orist. The central task for the Fission theorist is to motivate her theory in the rst place. After all, the Atom theory has a simple and standard picture of what's communicate by attitude ascriptions that embed a presupposition trigger. They communicate what a standard compositional theory takes the attitude ascription to semantically express. Since the Fission theorist doesn't agree, she owes us her own systematic account of what is communicated. This section outlines two ap- proaches that a Fission theory might take. 6.3.1 A purely pragmatic Fission theory Here's the rst approach. It starts with a fully traditional semantics for presup- position triggers, where sentences containing a presupposition trigger semantically express a conjunctive proposition. (2) I'm glad that Tom has stopped wearing plaid. (Conjunctive) I'm glad that Tom used to wear plaid and isn't anymore. There are at least two ways forConjunctive to be true. It can be true if Present is true, and it can be true if Past is true. (Present) I'm glad that Tom isn't wearing plaid. (Past) I'm glad that Tom used to wear plaid. 123 Either proposition is sucient for the truth of Conjunctive, because is glad isn't closed under consequence. Importantly, though, communicating Past or Present is more informative than communicatingConjunctive would be. For one thing,Conjunctive is true in a wider range of worlds than Past is true in, and a wider range than Present is true in. (It's true at least in the union of the worlds where Past with the worlds where Present is true.) Since it's true in a wider range of worlds, it is less informative; the hearer learns less about her location in logical space if she learns that it's true than if she learns that one of the other propositions is true. As a result, it's more cooperative to use (2) to communicate Past or to communicate Present. The rst Fission theory draws heavily on this observation. But this observation only goes so far. It tells us that Past and Present are tied as interpretations of uses of (2). Without something to break this tie, it would still be best to interpret uses of (2) as communicating Conjunctive. The rst Fission theory breaks this tie by appeal to a pragmatic theory of presupposition. A pragmatic theory of presupposition explains why utterances of Tom stopped wearing plaid presuppose that Tom used to wear plaid. And part of what it is for an utterance to presuppose p is for competent speaker-hearers to understand that p is not part of the main point of the utterance. Now a speaker who uses (2) ordinarily presupposes that Tom used to wear plaid. That is, she's treating the proposition about Tom's past activities as not the main point of her utterance. It's then less likely that she intends to communicate about her gladness about Tom's past activity. That's why competent speaker-hearers tend to understand utterances of (2) as communicating Present, rather than any of the alternatives. This explanation also generalizes to third-person attitude ascriptions, given one further claim. Stipulate for a moment that attitude ascriptions somehow commu- nicate that the matrix subject is disposed to presuppose presuppositions triggered from the complement. With (6), for example, utterances of (6) somehow communi- cate that James is disposed to presuppose that Tom used to wear plaid. (6) James is glad that Tom has stopped wearing plaid. And in presupposing this proposition, James is disposed to treat it as not the main point of the utterance. Now, as before, there are three interpretations of what uses of (6) communicate: Conjunctive, Present, and Past. Since James is disposed to treat propositions about Tom's past as not the main point, Past is a less good interpretation of (6) than Present. And Present and Past are both more informative than Conjunctive. So Present is the best interpretation of what ordinary uses of (6) communicate. That's basically all the detail I'll give about this rst Fission theory. (I'll return to the stipulation in a moment.) So I haven't really introduced an explanation of the Fission theory. For one thing, I haven't given enough detail to tell what sort of empirical predictions this explanation would make. I've rather indicated some resources that you might use in explaining that theory. 124 My overarching goal in this paper is to shift the working hypothesis about presupposition triggers under attitude reports. The working hypothesis has tradi- tionally been that the Atom theory is true. I think that the working hypothesis should rather be that the Fission theory is true. But it's hard for a single paper to shift the working hypothesis while also zeroing in on the very best version of the working hypothesis. So my goal here is modest. I want to show that there are several attractive ways of explaining the Fission theory. In doing that, I'm trying to convince you that there will be some adequate explanation of the Fission theory, without convincing you of its exact contours. Moreover, the Fission theory faces some important problems. And it's easier to appreciate those problems as they arise for dierent explanations of the Fission theory. For example, I just introduced an explanation that stipulates something important: that we normally interpret the matrix subject as being disposed to presuppose presuppositions of the complement. And this stipulation looks like a liability of this rst explanation. After all, the Atom theorist might claim to explain the facts that I've had to stipulate. For an Atom theorist, being glad that Tom stopped wearing plaid is being glad about a conjunctive proposition about the past and present both. And we should agree that being glad that p and q requires believing p and q. And believing that Tom used to wear plaid is a great reason for being disposed to presuppose that he used to. Advantage Atom theory: it explains an important fact that the Fission theory has to stipulate. This apparent advantage is illusory. The Atom explanation of this pattern doesn't work, at least with the full generality that it would need to be plausible. The pattern is highly general. (General Pattern) In order to V that S, you normally need to be disposed to presuppose S's presuppositions. For example, hoping that Tom stopped wearing plaid also seems to require believing that Tom used to wear plaid. And the Atom account does not capture this general pattern. For one thing, it's not plausible that hoping that p and q requires believing p and q. (You can hope that it's cloudy and rainy, without believing that it's cloudy and rainy.) Extant Atom theorists have recognized this point, tacitly stipulating that the General Pattern holds, without explaining why. 19 In other words: Atom theorists need an explanation of the General Pattern as much as Fission theorists do. So we can't use the General Pattern to discriminate between the Atom theory and the Fission theory. 6.3.2 A hybridized pragmatic explanation This section introduces another explanation of the Fission theory. This explanation changes an odd feature of the previous explanation. That previous explanation in eect imagines the matrix subject as the speaker; it asks 19 Paul Elbourne (2005), Bart Geurts (1998), and Irene Heim (1992) are some examples. 125 what she would treat as not the main point if she were talking about Tom's activities. But this approach is oddly indirect. It's not the matrix subject who's using a construction that triggers the relevant presuppositions. It's the speaker who's doing that. Why not focus more on what the speaker is doing? The second explanation of the Fission theory aims at doing that. It begins with a dierent semantics for presupposition triggers. This semantics is multi-dimensional. It associates the presupposition trigger with a pair of propositions, rather than a single conjunctive proposition. Multi-dimensional Semantics JstopK g = h <et> .x <e> .t <s> . Content 1: x isn't h-ing at t Content 2: x was not h-ing before t This semantics does not mark either of the two propositions in any special way. It doesn't mark the rst one as the main point of the utterance, and it doesn't mark the second one as presupposed. (The labels { \Content 1" and \Content 2" { are only part of the meta-language that I use to talk about this proposal.) As a result, this semantics does not stipulate any facts that a pragmatic theorist is concerned to explain. It makes only one change from the traditional account. Conjunctions are eliminated; propositions that were previously conjoined are instead each associated with the presupposition trigger. 20 Verbs like stop embed under other constructions. They embed under attitude reports, for one thing. So in adopting a multi-dimensional semantics, we need a principled account of these embeddings. Do both contents embed? If not, what determines which content embeds? The second Fission explanation uses a pragmatic theory of presupposition to answer these questions. Its basic claim is that embeddings \ignore" presupposed content. Since utterances of Tom stopped wearing plaid presuppose that he used to, embeddings ignore the proposition about his past. So utterances of James is glad that Tom stopped wearing plaid communicate only that James is glad about Tom's present activity: that James is glad that Tom isn't wearing plaid. This approach diers from thex3.1 approach in being exclusively speaker focused. The speaker's choice of a presupposition trigger is the only thing that matters in determining what embeds. I'll call this approach a hybridized explanation of the Fission theory. It draws on a multi-dimensional semantics for presupposition triggers, and semantic theories of 20 Is this multi-dimensional semantics itself o limits for a pragmatic theorist? No. Take the conjecture that natural language never encodes conjunctions, but always encodes con- tent that is traditionally taken to be conjunctive by associating the term with two distinct propositions. This conjecture is as simple and principled as the traditional conjecture, that the semantics is always conjunctive. It doesn't stipulate anything that the traditional conjecture doesn't also stipulate. 126 presupposition characteristically draw on that tool. But it also draws on a pragmatic theory of presupposition. It hybridizes the two kinds of resources together. The hybridized approach holds that semantic composition is most basically something that agents do. 21 Why do embeddings \ignore" presupposed content? The answer is that the speaker's presuppositions aect which of the two propositions are the predication target of the embedding. If the speaker is presupposing p, p is normally not the predication target of embeddings. On this picture, then, semantic composition is importantly like other, non-semantic processes. If I say this apple is red, I'm normally predicating the predicate is red of a proper subpart of the whole apple. I'm just talking about the surface. I'm not saying that the apple is red all the way through. The hybridized approach takes semantic composition to work similarly. In the apple case, there are several predication targets: you can be predicating red of the surface, or of the whole thing. In the stop case, there are at least two predication targets: the proposition about the past, and the proposition about the present. Now the hybridized theorist needs to explain why the speaker's presupposi- tions would aect semantic composition. But we shouldn't be pessimistic about her prospects. A common thread running through many dierent pragmatic the- ories of presupposition is that the attitude of presupposition involves treating the presupposition as not the main point. It then makes sense that competent hearers would not interpret the presupposition as the predication target. They understand that the speaker is treating the presupposition as not the main point { so they'll interpret that presupposition as not being the predication target. To illustrate, take an utterance of James is glad that Tom stopped wearing plaid. For the hybridized theorist, the speaker uses this sentence to predicate the property being-something-that-James-is-glad-about of some proposition. The hearer's task is to gure out which one. Since she usually knows that the presupposed proposition is not part of the main point, she tends to infer that that proposition is not the predication target. Moreover, the speaker is tacitly expecting the hearer to make this inference. Competent speakers and competent hearers will both expect an utterance of this sentence to be interpreted as asserting that James is glad that Tom isn't wearing plaid. And that's just what we should hope for. There might be other ways of developing this hybridized theory. One version would dispense with the multi-dimensional semantics altogether, instead taking stop to express just the conjunctive proposition that it's traditionally taken to express. However, it would deny that this conjunctive proposition is the usual predication target. The usual predication target is just the conjunct about Tom's present activity. It takes the analogy with red apples even more seriously. Filling out any version of this story would take a great deal of work. It would re- quire a more concrete picture of the attitude of speaker presupposition, and a much more detailed account of the inferences that competent speakers and competent hearers expect each other to make. But this paper isn't trying to develop any such 21 It thus builds from the recent work developing \cognitive act" conceptions of propo- sitions, as in Peter ?, Bjorn ?, Frederika ?, and Scott Soames (2015). 127 account. My goal is rather to shift our working hypothesis about presuppositions under attitude reports { to the Fission theory, and away from the Atom theory. The hybridized explanation shares at least one liability with the purely prag- matic theory. It doesn't explain why we tend to infer that the matrix subject presupposes presuppositions triggered from the complement. However, as we saw earlier, the Atom theorist doesn't explain this fact either { at least in the full gen- erality that it needs to be explained. One important further question about the two explanations of the Fission theory is whether they can be extended to explain this fact. I think that the hybridized explanation can, and I doubt that the purely pragmatic explanation can. So I suspect that the hybridized explanation is more promising. But making good on this suspicion would take another paper. 6.4 A problem: local accommodation Let's recap. Part I aimed at eliminating the Atom theory, showing that it can't be correct. Part II sketches how the Fission theory can capture the data that the Atom theory can't. Now I haven't developed any version of the Fission theory in enough detail to evaluate. I've rather been trying to highlight the range of resources that the Fission theorist can use in developing her theory. The goal has been to convince you to adopt the Fission theory as our working hypothesis about presupposition triggers under attitude reports. This section considers a central argument against the Fission theory, and for the Atom theory. The argument tries to show that no version of the Fission the- ory can capture an important range of data, about attitude ascriptions where the presupposition is locally accommodated. Imagine that there's a prize for former plaid-wearers. You get the prize i you once wore plaid but no longer do. If James just learned about this prize, and if he likes Tom, (6) might be true. (6) James is glad that Tom has stopped wearing plaid. In this case, though, (6) isn't true because James is glad only about Tom's present activities. Rather, (6) is true because James is glad about the past and present both. That is just the reading that the Atom theory predicts, and the reading that the Fission theory seems to deny. It thus seems like good evidence for the Atom theory. In fact, though, the Fission theory is compatible with these readings, at least in principle. The Fission theory makes a claim about a possible use of presupposition triggers. It is silent about whether this possible use is the only possible use of presupposition triggers. theFission theory holds that I can use (6) without thereby com- municating that I'm glad about the past and present both. In- stead, I can use it to communicate that I'm glad just about the present. In those cases, the only proposition communicated about my gladness is that I'm glad that Tom is not wearing plaid now. 128 the Atom theory holds that the cases that the Fission theory is about are still cases where uses of (6) only communicate the proposition semantically expressed. The basic question in this paper is whether we need to expand the tools that the Atom theorist uses, to include the sort of tool that the Fission theorist uses. And I've been arguing that we do. As a result, the basic Fission theory doesn't say anything about local accommodation, one way or another. In general, the Fission theory is less ambitious than the Atom theory. The Atom theory tries to explain presupposition triggers by drawing on a comparatively narrow toolkit. So if the Atom theory ts the data, it's much more theoretically satisfying. The Fission theory, by contrast, holds that we need to expand our toolkit { that we need to use new tools, in addition to the tools the Atom theorist uses. As a result, it's less theoretically satisfying. But because the Atom theory is more ambitious, it's easier to counterexample. That's why the examples inx2 are so important. They show that there's no way for the Atom theorist to make good on her theoretical ambitions. But the Fission theorist does not have the same kind of theoretical ambitions. Even if her view is right, there still could be contexts where the data patterns in the way the Atom theorist predicts. So we shouldn't expect to nd an example that the Atom theorist gets right and the Fission theorist gets wrong. But the devil is in the details. The rest of this section will consider two particular ways that local accommodation might make trouble for the Fission theory. 6.4.1 Explanations of the Fission theory Go back to the hybridized explanation fromx3.2. That explanation built from a multi-dimensional semantics for presupposition triggers. Multi-dimensional Semantics JstopK g = h <et> .x <e> .t <s> . Content 1: x isn't h-ing at t Content 2: x was not h-ing before t This semantics seems to force a choice: we can interpret embedding operators (like James is glad...) as applying to one proposition or the other, but not both. And cases of local accommodation involve both propositions being interpreted under that embedding operator. So this semantics seems to predict, incorrectly, that local accommodation is impossible. That's one way that local accommodation could make trouble for the Fission theory: by eliminating some ways of explaining the theory. And that may be the upshot for the hybridized explanation { that it's empirically implausible, because it doesn't allow for the readings that are in fact available. 129 Importantly, though, local accommodation just can't pose the same problem for every explanation of the Fission theory. Go back to the purely pragmatic theory inx3.1. Given that theory, attitude ascriptions that embed presupposition triggers semantically express just what the Atom theory says they do: an attitude to a conjunctive proposition. But competent speaker-hearers usually interpret those at- titude ascriptions as communicating something more informative. They are usually interpreted as communicating that the matrix subject has the attitude towards a proper subpart of what the complement expresses. They are interpreted as commu- nicating that more informative proposition for pragmatic reasons, from reasoning about what it is cooperative for the speaker to say. As a result, the purely pragmatic theorist can capture cases of local accommo- dation. To do that, she needs only to show that the normal pragmatic reasoning will be disrupted for some reason. And we should be optimistic that she will have prin- cipled way of showing just that. The pragmatic reasoning depends crucially on the assumption that the matrix subject is disposed to presuppose the presupposition { that is, to treat it as not part of the main point. But this assumption is false in the case described. Tom's past activities are very much part of the main point. He has to have worn plaid to be eligible for the prize! The normal pragmatic reasoning thus doesn't get o the ground. And that pragmatic reasoning is why we don't interpret the proposition semantically expressed as what's communicated. Since it doesn't get o the ground, we do hear the proposition semantically expressed { which is just the reading where the presupposition is locally accommodated. More generally, then, it is possible to develop the Fission theory to allow cases of local accommodation. And that's just what we should expect. The Fission theory expands our toolkit, rather than contracting it. It'd be surprising if expanding our toolkit made it impossible to explain some things we could explain before. Now go back to the hybridized explanation of the Fission theory. I conceded that local accommodation posed a problem for the hybridized theory. But I'm not sure it really does. To capture cases of local accommodation, the hybridized theorist needs a conjunctive proposition to be the predication target of certain embeddings. And I think that the hybridized theorist can allow that that happens sometimes. In order for it to happen, agents would have to do one more thing than they usually do. They would rst conjoin the two propositions associated with stop. Compare the introduction of conjunctions in event semantics. 22 Suppose that adjuncts like in Paris just express properties of events: e.In(e, Paris). Sentences like John danced in Paris still communicate conjunctive propositions:9e(Dancing(e) ^ Agent(e, John)^ In(e, Paris). The event semanticist with our simple view of ad- juncts thinks that conjunctions are introduced, even though they're not part of the semantic values. The hybridized theorist just does the same thing. Then her story continues as before. Moreover, this suggestion is plausible given the basic hybridized approach. Cases of local accommodation are cases where the presupposition is part of the main point. This observation is important for the purely pragmatic theory, because 22 As pioneered in ?. 130 it means that the matrix subject usually treats the presupposition as part of the main point. But it can also be important for the hybridized theory, because it means that the speaker will also take the presupposition to be part of the main point. Both propositions that are associated with stop are then part of the main point. That's the reason why the predication target is a conjunctive proposition. 6.4.2 Back to the Core Asymmetry Local accommodation could also make trouble for the Fission theory in another way. It could undermine the data fromx2 that I used to eliminate the Atom theory. Consider the following case. There's an evil demon who intends to kill everybody who has never worn plaid, and reward everybody who currently wears plaid. James knows these facts, so his preferences are: First Best: Tom is currently wearing plaid. Second Best: Tom used to wear plaid, but isn't currently. Worst: Tom hasn't ever worn plaid. Given these preferences, and James' knowledge that Tom used to but isn't currently wearing plaid, (6) seems appropriate. (6) James is glad that Tom has stopped wearing plaid. Can the Fission theorist explain why (6) is appropriate? Yes. (6) is appropriate because the presupposition is locally accommodated. In this context, using (6) communicates that James is glad that Tom used to wear plaid and isn't anymore. And that proposition is true, given the preferences that James has. Now the Atom theorist might try to leverage this point to undermine the data back inx2. I there suggested that sentences like (4b) are always inappropriate. (4b) James is sad that Tom doesn't wear plaid { * and he's glad that Tom has stopped wearing plaid. And I claimed that the Atom theorist can't explain this fact. But the Atom theorist might think that we've just found a case where (4b) is appropriate. Both conjuncts are true. And even the Fission theorist concedes that both conjuncts are true! So I've gotten the empirical facts wrong. There's no objection to the Atom theorist. This objection makes a more sophisticated version of an earlier mistake. I concede that (4b) can be appropriate given dierent reasons for being sad and for being glad. The most pointed challenge to the Atom theory controls for this fact. It changes (4b) to be about sadness and gladness for the same reason. (4b) James is sad that Tom doesn't wear plaid { * and, for the same reason { he's glad that Tom has stopped wearing plaid. 131 And this discourse is uncontroversially defective, even in the case imagined. The empirical facts still hold. This challenge is forceful because (4a), unlike (4b), can still be true. (4a) James is sad that Tom used to wear plaid { and, for the same reason { he's glad that Tom has stopped wearing plaid. (4b) James is sad that Tom doesn't wear plaid { and, for the same reason { * he's glad that Tom has stopped wearing plaid. Now (4a) isn't true in the case described in this section. But the challenge to the Atom theorist is to explain why it's ever possible for (4a) to be true, even while it's impossible for (4b) to be true. In this section, I've considered an argument it is possible for (4b) to be true. This argument fails. We can rest condent in the data that I've been leveraging against the Atom theorist. Remember that the Atom theorist has dierent theoretical ambitions than the Fission theorist. The Atom theorist thinks that all the data can be explained by drawing on a narrow toolkit. And the Fission theorist thinks that we need to broaden our toolkit. In order to show that the Atom theorist is wrong, we just need to nd one context where the data patterns in a way that she can't explain. By contrast, the Fission theorist can allow that there are contexts where the data patterns in the way that Atom theorist expects. The Fission theorist can allow that there's a great deal of heterogeneity in the data. The Atom theorist can't. We've now reached the end of Part II. The goal of Part II was to show how the Fission theory might capture the data that the Atom theory cannot. That is, its goal is to shift our working hypothesis, from the Atom theory to the Fission theory. In pursuing this goal, I've been comparing families of views, rather than detailed versions of the dierent views. I hope that it's clear by now why it's fruitful to do that. Each family of views shares lots of problems, and lots of advantages. Part III: Philosophical upshots The choice between the Fission theory and the Atom theory also has important philosophical upshots. As before, the philosophical upshots are shared among mem- bers of the Fission theory { they follow from structural features of the basic Fission view. The goal of Part III is to introduce two of the most important upshots. 6.5 Singular thought and thought by description The Fission theory is incompatible with a very natural picture of our cognitive limitations, a natural picture that goes back at least to Russell. 23 That picture acknowledges only two ways of thinking about objects in the world. We can think 23 Russell (1905); ?. 132 about objects by description, as when we think about whoever is uniquely king of France. Alternatively, we can think about objects singularly, without any de- scription mediating our thought. And this neo-Russellian picture imposes some constraints on singular thought. That is, purely general grounds aren't enough for thinking singularly about some object. 24 In the most familiar version, you have to be acquainted with the object in order to think singularly about it. Importantly, then, views that are maximally liberal about singular thought { like that of John Hawthorne and David Manley (2012) { reject this neo-Russellian picture. The Fission theory is incompatible with the neo-Russellian picture. It's in- compatible with it, because it's incompatible with a Russellian account of denite descriptions. For our purposes, the key fact about a Russellian account is that a sentence containing a denite description entails that something satises the re- strictor of the description. pthe F is Gq is true i [9x: (Fx^8y (Fy! x = y))] (Gx). So pthe F is Gq entails that [9x] (Fx). Importantly, though, the proposition that something satises the restrictor is also a presupposition of the use of a denite. For one thing, it projects through embeddings. Utterances of the king of Sweden isn't home tend to presuppose that Sweden has a king. So the Russellian account is a conjunctive account of the proposition expressed { just the sort of conjunctive account that Atom theorists accept. However, we also see the same pattern with denite descriptions that we see with other presupposition triggers. (10a) I'm sad that there's a spy. But I'm glad that the spy was caught. (10b). I'm sad that, if there is a spy, he's been caught. * But I'm glad that the spy was caught. We see the now-familiar asymmetry that presupposition triggers display. Being glad about the presupposition (that there is a spy) isn't enough for being glad that the spy was caught. The Fission theorist has a standard explanation of this asymmetry. Utterances of I'm glad that the spy was caught communicate that the speaker is glad about a proposition that does not entail the presupposition. And they also somehow commu- nicate that the speaker presupposes something that does entail the presupposition. The Fission gloss has to have the following structure: (2) I'm glad that the spy was caught. I'm glad that i 1 was caught. I presuppose that i 1 is a spy. 24 Robin ? discusses standard versions of Russellian picture in illuminating detail, in- cluding some of its costs. Philosophers who accept some version of this picture include at least Kent ?, Steven Boer and William Lycan (6986), Keith ?, Gareth ?, Robin Jeshion (2010), John ??, Francois ?, Marga ?, Nathan ?, and Scott ??. 133 What is it to be glad that i 1 was caught? One answer is that that attitude in- volves thinking singularly about some individual, and being glad that that indi- vidual was caught. This answer assumes that purely general grounds suce for singular thought. After all, I can have purely general grounds for being glad that the spy was caught. (Maybe I know that the CIA has been running badly, and I infer that there's a spy. When the CIA starts running more smoothly, I can be glad that the spy is caught { all without having anything but general grounds for my attitudes.) So this answer rejects any neo-Russellian picture that imposes some constraints on singular thought. It requires the sort of liberality about singular thought that Hawthorne and Manley (2012) endorse. There are other ways to make sense of the Fission gloss on gladness reports. One approach is to expand the neo-Russellian picture by positing a new way that our thought and language can be about particular individuals. Grant that we can think quanticationally, about all spies, say. Grant that we can also think singularly about particular individuals, and insist that there are some constraints on singular thought. This approach introduces a sui generis third way for thinking about objects. I will call this way thought with unanchored discourse referents. Thought with unanchored discourse referents is like singular thought, in that it does not require the thinker to think of the object as falling under any particular description. But it's unlike singular thought, in that there need be no unique worldly individual that you're thinking about. 25 As a result, another way to develop the Fission theory is take uses of (2) to communicate a proposition about an unanchored discourse referent. If I'm glad that the spy was caught, I'm glad about an unanchored discourse referent i 1 { glad that i 1 was caught, while presupposing that i 1 is a spy. 26 In general, the Fission theory is incompatible with any neo-Russellian account of the limits of thought. (2) can be true even if I have nothing but general grounds for my attitudes. So the neo-Russellian can't take (2) to communicate that I'm glad about singular propositions. She has to take me to be glad about a proposition that entails that something satises the restrictor property. In doing that, her view is incompatible with the Fission theory. The Fission theory denies that I'm glad about a proposition that entails that something satises the restrictor property. As a result, the Fission theory forces us to introduce some tool that the neo-Russellian does not countenance. And this neo-Russellian picture has been in uential across a wide range of debates, while also facing its share of critical attention. 27 But inasmuch as this 25 Irene Heim (1982); ? and Hans Kamp (1981) both in uentially developed these ideas, around the same time. The distinction between anchored and unanchored discourse referents is due to Kamp (ms).. 26 Linguists have long recognized that a Fission approach to presupposition triggers requires some non-standard machinery. This basic observation goes back to Karttunen and Peters (6979). Linguists who favor this sort of appeal to unanchored discourse referents include Paul Dekker (2008), Robert van Rooij (2005, 2010), David Oshima (2006), and Yasutada Sudo (2012). 27 Robin ? gives a helpful description of the lay of the land. 134 paper has shown that the Fission theory is true and the Atom theory false, it has shown that the neo-Russellian picture is false. I've emphasized that the Fission theory is less ambitious than the Atom theory. It denies that we can explain everything by the linguistic tools that the Atom theo- rist uses. The choice between the Fission theory and the Atom theory corresponds to an important linguistic joint. All versions of the Atom theory face similar prob- lems and have similar advantages, and all versions of the Fission theory do too. In this section, we've seen that this linguistic joint also corresponds to an important philosophical joint. If you accept any version of the Fission theory, you have to re- ject the neo-Russellian picture. You have to acknowledge examples that just can't be explained in a neo-Russellian way. 6.6 Applications in epistemology This section describes another philosophical upshot of the Fission theory. If the Fission theory is true, knowledge attributions don't require knowledge of the pre- supposition. Knows is an attitude verb, like wants or is glad. And the whole point of the Fission theory is that attitude attributions normally don't require the at- titude towards presuppositions of the complement. As a result, certain kinds of knowledge will be easier to acquire if the Fission theory is true than if the Atom theory is true. But we have to work through a subtle point before fully appreciating this upshot of the Fission theory. Attitude ascriptions do require that the matrix subject is disposed to presuppose presuppositions of the complement. And one reason why you might be disposed to presuppose those presuppositions is that you know them. So knowledge of a presupposition is sucient for meeting one of the requirements of a knowledge attribution. My claim is that it's not necessary if the Fission theory is true. There are other reasons you might be disposed to presuppose presuppositions of the complement. Maybe you're disposed to treat the presupposition as true in the present context { that is, you're disposed to accept them. Then you'll also meet this requirement. 28 In general, then, the Fission theory predicts that knowledge reports don't re- quire knowledge of the presupposition, though they are associated with a require- ment that can be satised by knowledge of the presupposition. Do the empirical facts t this prediction? Consider clefts. An utterance of it wasn't Tom who took out the trash presupposes that someone took it out. (11a) I don't know whether Tom took out the trash. But I'll accept for now that he didn't. * I denitely know that it wasn't Tom who took it out. The trash is empty, so someone took out out. (11b) I don't know whether someone took out the trash. But I'll accept for now that someone did. I denitely know that it wasn't Tom who 28 See Stalnaker (2002) for an illuminating discussion of the reasons why it's intelligible to presuppose p. 135 took it out. I've been watching him all day, and he went nowhere near it. The second sentence { about what I accept { is important to satisfy the requirement that the matrix subject of the knowledge report does accept presuppositions of the complement. And this pattern is just what the Fission theory predicts. Knowing that it wasn't Tom who took out the trash doesn't require knowing that someone took it out, but it does require knowing that Tom didn't take it out. Interestingly, though, this pattern is somewhat more subtle with other presup- position triggers, including the paradigmatic example in this paper. (12a) I don't know whether Tom is wearing plaid now. But I'll accept for now that he is. * I denitely know that he has stopped wearing plaid, because he just told me that he's never worn plaid in the past. (12b) I don't know whether Tom used to wear plaid. But I'll accept for now that he did. I denitely know that he has stopped wearing plaid, because he just told me that he's never wearing plaid from today forward. (12b) is a bit hard to process. Despite that diculty, it is appropriate, in a way that (11a) just isn't. Knowledge that Tom has stopped wearing plaid doesn't require knowledge about his past, though it does require knowledge about the present. I've just given two pairs that pattern in the way that the Fission theory pre- dicts. Unfortunately, though, both of them are somewhat articial { not the sorts of utterance you'd hear on the Clapham omnibus. They're important here as cor- roborating evidence for the claims that the Fission theory makes. They help us rest secure in the Fission theory: that it does plausibly generalize even to knowledge attributions. This discovery about knowledge attributions has important upshots in episte- mology. Certain kinds of knowledge come easier if the Fission theory is true than if the Atom theory is true. Suppose that utterances of S are associated with two propositions p 1 and p 2 , with p 1 being a presupposition of utterances of S. Knowing that S requires less if the Fission theory is true than if the Atom theory is true. Knowing that S requires knowing two propositions given the Atom theory: p 1 and p 2 . On the Fission theory, by contrast, knowing that S only requires knowing one proposition: it requires knowing p 2 . One concrete example of this upshot is in metaethics. Moral realists take moral properties to be independent of and more fundamental than our individual eval- uative attitudes. Alex Silk (2016, 2017b) has argued that commitment to those properties is a presupposition of moral discourse if it's present. He has not appreci- ated the important upshots that this argument would have for moral epistemology. If it's right, moral knowledge would come much more easily than moral realists have traditionally assumed. It wouldn't require knowledge about the fundamental realist properties. It would only require agents to accept propositions about those properties. There isn't space here to work out these consequences in detail, but we should expect there to be dramatic consequences. 136 In general, presupposition is philosophically more important if the Fission the- ory is true than if the Atom theory is true. If the Fission theory is true, certain kinds of knowledge come much easier than philosophers have appreciated. 29 6.7 Wrapping up This paper has both an expository agenda and an argumentative agenda. The ex- pository agenda is to introduce the Fission theory as an alternative to the orthodox Atom theory, and to articulate the important dierences between the two theories. The Atom theory and the Fission theory are both broad tents; they include lots of dierent concrete theories. But the choice between them is a deep and fundamental one. If you adopt any version of the Atom theory, you will automatically secure a particular range of advantages, and incur a particular range of costs. Ditto for the Fission theory { adopting any version of it automatically secures a particular range of advantages, and incurs a particular range of costs. I've tried to describe the choice between the two theories at a high enough level to bring out the basic advantages and the basic costs. The Atom theory is theoretically more ambitious. It tries to use a comparatively narrow toolkit to explain the full range of data. Its basic problem is empirical. You can't explain the full range of data just by appeal to that narrow toolkit. The Fission theory is theoretically less ambitious; it uses a comparatively broad toolkit. Its problem is theoretical. It needs to articulate the principles that determine what parts of the toolkit bear on which empirical phenomenon. My argumentative agenda has been to eliminate the Atom theory. I've argued that it's impossible to make good on its theoretical ambitions. To recall the key contrast: (4a) James is sad that Tom used to wear plaid. He's still glad that Tom has stopped wearing plaid. (4b) James is sad that Tom doesn't wear plaid. * He's still glad that Tom has stopped wearing plaid. The Atom theorist has to explain (4a)'s felicity by appeal to the behavior of the attitude verb, rather than something special about the presupposition trigger. (Her view is that there isn't anything special about the presupposition trigger. So of 29 This second philosophical upshot is more complicated than the rst. Any version of the Fission theory is incompatible with the neo-Russellian picture. But there are ver- sions of the Fission theory where presupposition is philosophically less interesting. The Fission theorist will agree that presuppositions can in principle be locally accommodated in knowledge attributions. Dierent versions of the Fission theory will take that sort of local accommodation to happen with more or less frequency. Some theorists think that local accommodation is basic and normal (Nirit ? and Craige? are some examples) while others think that it's marked and exceptional (Irene Heim (1983) and David ? are some examples). The interest of this upshot for epistemology depends to some extent on which group is right. 137 course she has to appeal to the behavior of the attitude verb!) And I've shown that it's impossible to predict the asymmetry between (4a) and (4b) by appealing to the behavior of the attitude verb. We should all adopt the Fission theory as our working hypothesis. The alterna- tive just can't work. Appreciating this point has important philosophical upshots. It forces us to abandon a neo-Russellian picture of our cognitive limitation, in favor a more liberal view. And it reveals that certain kinds of knowledge are much easier to acquire than we've realized before now. Getting clear on presupposition triggers in attitude ascription bears important fruit, both philosophical and linguistic. 138 Chapter 7: Four arguments for positivist realism At this point in the dissertation, we're in a position to start evaluating the core linguistic thesis that lies at the heart of positivist realism. The goal of this Part of the dissertation is positive: I want to put together the evidence for that thesis. However, the core linguistic thesis is in one way more complicated than I've acknowledged up to now. The core linguistic thesis rests on two basic building blocks, rather than just one. The rst important building block is the positivist's hybrid theory of moral belief and moral knowledge, where knowledge reports receive the following analysis. (1a) Jane knows that she ought to keep this promise. At-issue: Jane knows that s 1 requires keeping this promise, and accepts that s 1 is the moral standard Not-at-issue: s 1 is the moral standard On this hybrid theory, Jane only needs to know one proposition, about what the set of rules s 1 demands. This hybrid theory follows from the claim that the fundamental realist property is part of the presupposed, not-at-issue content associated with a moral utterance. (2) Jane ought to keep this promise. At-issue: s 1 requires Jane's keeping this promise Not-at-issue: s 1 is the moral standard This simple conjecture is what leads to the hybrid theory, because of the facts about presupposition that the previous chapter emphasized. I'll begin by arguing that the fundamental realist property is in fact part of the presupposed, not-at-issue content associated with a moral utterance. However, the hybrid theory is not enough by itself to secure the full range of philosophical advantages that the positivist claims. One further building block is necessary. This chapter will also describe and motivate that further building block. 139 7.1 The Asymmetry Argument for the hybrid theory Chapter 6 developed a new perspective on presupposition triggers. That new per- spective supports a simple argument for the hybrid theory. Presupposition triggers embed very dierently than other kinds of terms. In particular, they embed very dierently under attitude ascriptions, like about what the subject wants or is glad about. Sentences free of presupposition triggers exhibit what I called Symmetries among Entailments. (Symmetries among Entailments) If S asymmetrically entails T (that is, if S entails T and T does not entail S), it's possible to [want/be glad] that S, without [wanting/being glad] that T. Take the sentence I'm glad that he bought a hybrid car. That sentence can be true even if I'm sad that he bought a car { if, for example, I know that he's going to buy a car, and I think that it's better to buy a hybrid than a gas-guzzler. That sentence can also be true even if I'm sad that he bought something hybrid { if, for example, I know that he's going to buy a car, and I think he'd be less of a damn hippie if he buys a car rather than a hybrid motorcycle. Presupposition triggers behave dierently. They do not display these sorts of symmetries. They rather display what I called the Core Asymmetry. (Core Asymmetry) S 1 carries some commitment T 1 such that: it's impossible to be glad that S 1 unless you're glad that T 1 . To be glad that Bill stopped wearing plaid, say, you have to be glad that he isn't wearing plaid. It's not enough if you're glad that he used to wear plaid. This construction thus exhibits the Core Asymmetry: in this case, the commitment is that Bill isn't wearing plaid. And it does not display Symmetries among Entailments. Moral discourse exhibits the Core Asymmetry. Imagine someone who has spent all her career arguing for the error theory, but changes her mind late in life. She can be sad about this discovery. At the same time, she can be glad about some substantive truths. In that case, (5a) is perfectly assertable. (5a) I'm sad that there objective, mind-independent moral facts. Given that fact, though, I'm glad that killing is usually wrong. I've always ar- gued that if there were objective, mind-independent moral facts, killing would usually be wrong. And I'm glad I wasn't wrong about every- thing. Now switch the case, to a normative ethicist who's come to believe that killing isn't wrong, and regrets that conclusion. At the same time, though, she's glad that she was right all along about there being objective, mind-independent truths. 140 (5b) I'm sad that morality would forbid killing if it forbids anything. * Given that fact, though, I'm glad that killing is usually wrong. I've always argued that there are objective, mind-independent moral facts. And I'm glad I wasn't wrong about everything. This pattern of behavior is a hallmark of presupposition triggers. The best way to explain it is to adopt the linguistic conjecture that the positivist favors: (*) Killing is usually wrong. At-issue:o 1 usually forbids killing Not-at-issue: o 1 is the moral standard This linguistic conjecture predicts that (+) is associated with these two propositions. (+) I'm glad that killing is usually wrong. At-issue: I'm glad that o 1 forbids killing and accept that o 1 is the moral standard. Not-at-issue: o 1 is the moral standard (+) doesn't require you to have any positive attitude towards the property being- the-moral-standard. That's why (5a) is assertable: (+) can be true even if you have nothing but regret about that property. But (+) does require you to be glad that killing is forbidden by the set of rules that you accept to be the moral standard. That's why (5b) isn't assertable. Its rst sentence is incompatible with being glad about that proposition. So if you accept the main argument of Chapter 6, you should accept the hybrid theory; the hybrid theory is the right way to explain this instance of the Core Asymmetry. I think that the Asymmetry argument decisively establishes the hybrid theory. So I think that every moral realist whatsoever should accept the hybrid theory. It should be entirely orthodox. As we'll see later on, though, the Asymmetry argument won't move you all the way to positivist realism. We need one further piece. But before introducing that further piece, and arguing for it, I want to articulate a more complete picture of arguments for the basic hybrid theory. 7.1.1 Unidimensional accounts of attitude asymmetries Let's start with an important complication. Chapter 6 countenanced two expla- nations of the Core Asymmetry. One of them was the hybrid theory of attitude ascriptions, of the sort just illustrated. But it also acknowledged a dynamic ex- planation of that Asymmetry. That dynamic explanation made crucial use of a backgrounding operator `': attitude ascriptions \ignore" material under that op- erator. This dynamic approach will gloss (+) in a signicantly dierent way. The property being-the-moral-standard is interpreted under that operator. 141 (+) I'm glad that killing is usually wrong. I'm glad that9x: (x is the moral standard)^ x usually forbids killing (5a) is assertable, because (+) doesn't require me to have any positive attitude towards the property being-the-moral-standard, since that property is interpreted under the operator. And (5b) isn't assertable, because (+) does require me to have a positive attitude towards something which is interpreted as being the moral standard. This dynamic suggestion is not the hybrid theory that is at the heart of positivist realism. But this dynamic suggestion has exactly the same philosophical upshots as the hybrid theory. Attitude ascriptions do not require the matrix attitude towards the realist's fundamental property. That property is interpreted under the opera- tor, and attitude ascriptions are supposed to ignore material under that operator. 1 I will take the Core Asymmetry to require the hybrid theory. But because the dynamic theory has the same upshots as the hybrid theory, you could equally well take the Core Asymmetry to require the dynamic theory. You might expect me to consider an alternative to the hybrid theory and the dynamic suggestion: the suggestion that the realist's property is part of the restric- tor of a Russellian denite description. This suggestion associates (+) with these two propositions: (+) I'm glad that killing usually wrong. I'm glad that [the x: x is the moral standard] (x usually forbids killing) This suggestion draws only on familiar tools. The semantics for the is straightfor- wardly quanticational: the embedded proposition is true i [9 x: x is the moral standard^ [8 y] (y is the moral standardÉ x = y)] (x forbids killing]. This sugges- tion has a good claim to be orthodox conception of moral semantics, at least among realists who acknowledge an ordering semantics for modal terms. A dirty little secret of Chapter 6 is that its conclusion is incompatible with a Russellian account of denite descriptions. Denite descriptions are presuppositions triggers. And the Russellian account is a paradigmatic example of a conjunctive semantics for a presupposition trigger that does without the dynamic operator. And the whole point of Chapter 6 was to argue that no account with that structure can explain the Core Asymmetry that characterizes presupposition triggers. By my lights, then, the orthodox realist conception of moral semantics is o the table before we even started this chapter. We have to accept the hybrid theory or the dynamic suggestion once we have the right understanding of presupposition triggers. 1 What does it mean to \ignore" that material? I'm not sure. That's why I prefer the hybrid theory. 142 7.2 The argument from local accommodation This section presents another argument for the basic hybrid theory. This second argument should increase our condence in the hybrid theory. It's a distinct line of evidence that conrms the same point. I'm defending the linguistic conjecture that the property being-the-moral-standard is part of a backgrounded presupposition of moral discourse. This linguistic conjec- ture makes important predictions about the way that we interpret embeddings of moral vocabulary, under operators like not. Backgrounded content embeds dier- ently than foregrounded content. Predicates like married and unmarried illustrate the way that foregrounded content embeds. Consider (4a). (4a) * Bob is not married, and he's not unmarried either. Either kind of belief would involve the same mistake. (4a) normally seems defective. 2 It seems defective because (5a) is a conceptual truth. (5a) If someone isn't married, they're unmarried. If (5a) is true, everything is either in the extension of is married or in the extension of is unmarried. And (4a) communicates that Bob isn't in either. But moral discourse does not pattern in the same way. A sentence like (4b) is interpretable. (4b) Killing is not morally impermissible, and it's not morally permis- sible either. Either kind of belief would involve the same mistake. For example, if I told my doctor that I was teaching (4b) in my classes, the doctor might well think that I'm corrupting the youth. This point is a point about the doctor's linguistic competence: that she can conclude certain things about what I'm teaching. 3 At the same time, though, (5b) expresses something like a conceptual truth. (5b) If an act isn't morally impermissible, it's morally permissible. 2 There is, however, one exception, which I take up below. 3 Now some philosophers (like Ronald Dworkin (1996)) doubt that utterances like (4b) are in the end intelligible. But their arguments are linguistically naive. It's a fact about our linguistic competence that we can understand utterances like (4b) in a way that we normally can't understand utterances like (4a). If a linguistic theory can capture this dierence, it's more plausible than one that can't. Dworkin's argument is plausible only if we can't nd a linguistic theory that does capture this dierence. And I'm claiming that we can. So it doesn't matter if some error theorists wouldn't assert (4b). I'm using it to illustrate a fact about our linguistic competence, not as a canonical formulation of the error theory. 143 There is a striking (if not totally perfect) contrast between the (a)-pair and the (b) pair. My presuppositional conjecture is necessary to capture this contrast. To ap- preciate why, though, we need some further background. Presuppositions usually project from embeddings under negation. If an utterance of S presupposes p, an utterance of p:Sq does too. 4 But there are important exceptions to this gener- alization. Presuppositions can be locally accommodated under negation, so that the presupposition is understood to be part of what's rejected. For example, John stopped smoking presupposes that John used to smoke. But we can coherently in- terpret someone who says John didn't stop smoking, because he never started. On the coherent interpretation, the negation targets the not-at-issue content as well as the at-issue content. When not-at-issue content is interpreted under an operator like not, linguists will say that the not-at-issue content is locally accommodated. My presuppositional conjecture is the best way to explain the contrast between the moral case and the marriage case. Given that conjecture, there are two ways of interpreting moral utterances containing a negation operator like not. The property being-the-moral-standard can either project globally, or be accommodated locally. The proposition about the moral standard projects globally when we interpret (5b), leading us to interpret it as communicating an at-issue content that's conceptually true. At the same time, we interpret the presupposition in (4b) as locally accom- modated, leading us to interpret it as a substantive content that's not a conceptual truth. (5b) If an act isn't morally impermissible, it's morally permissible. At-issue: If it's not the case that o 1 permits killing, o 1 does not permit killing Not-at-issue: o 1 is the moral standard (4b) Killing is not morally impermissible, and it's not morally permis- sible either. Either kind of belief would involve the same mistake. At-issue: It's not the case that o 1 is the moral standard and o 1 does not permit killing, and it's not the case that o 1 is the moral standard and o 1 does permit killing. Either kind of belief would involve the same mistake. Not-at-issue: [nothing] In order to endorse this explanation, you have to allow that the property being- the-moral-standard is part of a content that is sometimes interpreted globally, and sometimes locally accommodated. And in doing that, you're taking that property to be part of the presupposed, not-at-issue content. Local accommodation is a 4 For a generalization of this point to the sub-sentential case, see Philippe Schlenker (2010) and the references cited there. 144 genuine hallmark of presupposes, not-at-issue content. No other kind of content exhibits the same behavior. 5 Now I've claimed that sentences like (4a) aren't interpretable. (4a) * Bob is not married, and he's not unmarried either. And I suggested that (5a) expresses a conceptual truth. (5a) If someone isn't married, they're unmarried. However, there is an important exception to my rst claim. (4a) is perfectly inter- pretable if Bob isn't a person at all { if, for example, Bob is a chair. But this exception bolsters my presuppositional conjecture, rather than under- cutting it. It is highly plausible that sortally restricted predicates (like is married) trigger a presupposition about the relevant sortal (for is married, that the subject is a person). 6 And (4a) is interpretable if that presupposition is locally accom- modated { if it communicates that Bob is not both a person and married, and that Bob is not both a person and unmarried. Moreover, this explanation of the interpretability of (4a) also explains why we hear (5a) as expressing a conceptual truth. Someone triggers the presupposition that the salient object is a person, and that presupposition guarantees that the presupposition triggered from unmarried is satised. In general: when can we interpret two sentences with the form `A is not F and A is not un-F' and `if an X is not F, then it is un-F'? Only when F triggers a presupposition. Then the rst sentence is interpretable because we hear that presupposition is locally accommodated, and the second sentence is interpretable because we hear the presupposition as globally accommodated. Since we've seen the moral discourse exhibits this pattern, we have to see moral discourse as involving the sort of presuppositions I've suggested. Knowledge ascriptions furnishes another, related pattern of evidence for the pre- suppositional conjecture. We can understand someone who self-attributes knowl- edge of (4b), in a way that we can't understand someone who self-attributes knowl- edge of (4a) (while taking Bob to be a person). (6a) *? I know that Bob is not married and I know that he's not unmarried either. (6c) I know that if someone isn't married, they're unmarried. (6b) I know that killing is not morally permissible, and I know that it's not morally impermissible either. (6d) I know that if an act isn't morally impermissible, it's morally permissible. 5 Christopher Potts argues for this point in illuminating detail (Potts 2005, 35-6). 6 For discussion, see Thomason (1972). 145 We can't make good sense of (6a). (Or if we can make sense of it, we interpret the speaker as thinking that Bob is not a person.) By contrast, we can make good sense of (6b). We can interpret it as a cocky expression of the error theory. (Think of the dierence between telling your doctor (6a) and telling her (6c). The latter would elicit moral disapproval, but the former only confusion.) The presuppositional conjecture crisply captures this dierence. (6b) I know that killing is not morally impermissible, and I know that it's not morally permissible either. At-issue: I know that it's not the case that o 1 is the moral standard and o 1 permits killing, and it's not the case that o 1 is the moral standard and o 1 does not permit killing. Not-t-issue: [nothing] (6d) I know that if an act isn't morally impermissible, it's morally permissible. At-issue: I know that if it's not the case that o 1 permits killing, o 1 does not permit killing, and I accept that o 1 is the moral standard Not-at-issue: o 1 is the moral standard The presupposition conjecture predicts that are two coherent ways of interpreting the predicate is morally permissible. That predicate triggers a presupposition that can project globally, or be accommodated locally. That's why we can understand both (6b) and (6d). The behavior of the predicate is morally permissible under negation is another powerful source of evidence for the thesis that being-the-moral-standard is part of the presupposition. It's another source of evidence that the hybrid theory should be the orthodox theory of the moral use of modals. 7.3 Monism and Pluralism I now turn a central complication. I've said that a linguistic thesis about presuppo- sition is the central pivot for evaluating positivist realism. Unfortunately, though, the linguistic thesis that leads to positivist realism is more complicated than I've ac- knowledged so far. It rests on two fundamental building blocks. One of the central building blocks is the hybrid theory that I just argued for. But another building block is also important for the positivist. I'll introduce that further building block slowly. Moral realists use their fun- damental property being-the-moral-standard to play several important roles. One important role is as part of what distinguishes the moral use of modals from other uses. Knowledge ascriptions like (1a) can be used to assert something about Jane's knowledge about morality. But they can also be used in other ways: to talk about 146 her knowledge about what prudence requires, say. The realist takes the moral use of (1b) to communicate something about the property being-the-moral-standard: that Jane has some attitude about that property. And this hybrid theory is perfectly general, extending to other moral attitudes, like belief. (1b) Jane believes that she ought to keep this promise. At-issue: Jane believes that s 1 requires keeping this promise, and accepts that s 1 is the moral standard Not-at-issue: [nothing] (believes is a presupposition plug) But this hybrid theory by itself leaves another important question open. Take a systematic moral disagreement. Aquinas and I disagree about charging interest on loans. He thought it was always wrong, and I think it's sometimes ne. Are we both thinking about the same set of rules, and accepting it as the moral standard? Or are we thinking about dierent sets of rules, and each accepting our own set of rules as the moral standard? If you insist that we're always thinking of the same set of rules, you have a Monist conception of the hybrid theory. If you allow that we're thinking of dierent sets of rules, you have a Pluralist conception of the hybrid theory. My discussion in Chapters 4 and 5 took the positivist to be a Pluralist. And there is good reason for making this assumption; some of the positivist's metaeth- ical advantages depend on a Pluralist conception of the hybrid theory. So a full understanding of the positivist's metaethical advantages thus requires a full under- standing of the relationship between the general hybrid theory and the Pluralist conception of it. 7.3.1 Why the hybrid theory leaves this choice open I've noted that the hybrid theory follows from the conjecture that the fundamental moral property being-the-moral-standard is part of the presupposed, not-at-issue content associated with a moral utterance. On the Pluralist conception of the hybrid theory, the relevant agent's background assumptions have priority in determining what's asserted: the assertions are about whatever set of rules the relevant agent accepts to have that property. On a Monist conception, by contrast, the truth about that proposition has priority in determining what's asserted: the assertions are about whatever set of rules actually does have that property. The bare linguistic conjecture about presupposition may not be enough by itself to settle the choice between these two conceptions of the hybrid theory. It may not be enough, because it's somewhat unclear how presuppositions af- fect linguistic interpretation. It's unclear if they always behave as the Pluralist takes them to behave, or if they sometimes behave as the Pluralist supposes and sometimes as the Monist supposes. It should be uncontroversial that the speaker's background assumptions some- times have priority in determining what's asserted. Gender marking on pronouns 147 is is plausibly presuppositional. Think of an utterance of (3). (3) He came to town last night. An utterance of (3) plausibly presupposes that the salient individual is male. 7 Sup- pose, though, that the speaker has false beliefs about someone's gender. The speaker believes that Taylor is male and Alex is female, though you and I know that Taylor is female and Alex male. The speaker says Taylor and Alex were in the state this week, and he came to town last night. We'll interpret the second sentence as com- municating something about the individual that the speaker accepts to be male, not about the individual that we know to be male. Theorists who take gender marking to contribute a presupposition are assuming that we sometimes interpret the presupposition so that the speaker is accepting something true. If all presuppositions patterned with this case, the Pluralist conception of the hybrid theory would be the only viable option. And I would be perfectly happy if that turned out to be the case. The Pluralist conception is the one that secures all metaethical advantages that I claimed for positivist realism. But it may be contro- versial whether all presuppositions do pattern with the example in (3). In order for an example to pattern dierently, the speaker's background assumptions would have to be irrelevant { that is, these kinds of presuppositions have an automatic eect on the proposition communicated, without help from the speaker's assumptions. It's helpful to have an example where some theorists might take the speaker's background assumptions to be totally irrelevant. I'll illustrate this possibility within a Discourse Representation Theory framework. That framework makes it natural to associate The F is G with the at-issue content that i is G and the not-at-issue content that i is F. 8 Now consider (4). (4) The president of the US came to town last night. On some uses of (4), it's just irrelevant who the speaker accepts to be the president. She's talking about Obama, whatever her background beliefs. (Maybe there are some other contexts where uses of (4) communicate something else. The important point is that (4) has some uses where the speaker's background assumptions are totally irrelevant.) In this case, the speaker's attitudes about who satises the presupposition are irrelevant. 9 In general, then, presuppositions might aect the interpretation of an utterance in two ways. Sometimes we interpret the utterance as about whatever object the relevant agent accepts to satisfy the presupposition. There may also be cases where we interpret the utterance as about whatever object actually does satisfy the pre- supposition. Now go back to the broader questions we're exploring. We're focusing 7 See Robin Cooper (1983), Philippe Schlenker (2003), and Yasutada Sudo (2012) for discussion of this point. 8 I'm here simplifying very complicated issues in the research traditions running from Hans Kamp (1981) and Irene Heim (1982). 9 Hans Kamp (ms) discusses these points in helpful detail. 148 on moral realists, who posit a property being-the-moral-standard as the property that distinguishes the moral use of modals from other uses, like the legal use. At this point, we're restricted our attention to the view that that property is part of the presupposed, not-at-issue content associated with a moral utterance. The view is compatible with the claim that uses of modals express propositions about the set of rules that in fact is the moral standard. It's also uses of modals can express propositions about the set of rules that the relevant agent accepts to be the moral standard. We're now in a position to state the contrast between Monist and Pluralist views more precisely. The Monist holds that just talking about morality determines what set of rules is incorporated into the proposition expressed. The agent's background assumptions are irrelevant. The Pluralist disagrees. Talking about morality doesn't by itself determine what set of rules is incorporated into the proposition expressed. The agent's attitudes matter too. This Pluralist idea has striking and controversial commitments that the Monist does not. The Pluralist needs to articulate the additional facts that determine what set of rules agents are thinking about, in addition to their intention to think about morality. Chapter 4 suggests, for example, that the agent's disposition to accept substantive moral judgments determine which set of rules she's thinking about. Dierent agents would then be disposed to think of dierent sets of rules inasmuch as they're disposed to make dierent substantive judgments. Let's illustrate. Go back to a belief ascription like (1b). (1b) Jane believes that she ought to keep this promise. You can use (1b) to communicate something about Jane's beliefs about morality, or her beliefs about prudence. Monists and Pluralists have very dierent accounts of the propositions communicated. To describe this dierence, suppose that Jane has false beliefs both about what morality demands, and also about what prudence demands for her. We can model this supposition in the following, highly idealized way. Stipulate for distinct M, P, A, and B: M is the set of rules that is the moral standard P is the set of rules that is the standard of prudence for Jane A is the set of rules that best ts Jane's judgments about her moral obligations B is the set of rules that best ts Jane's judgments about her prudential obligations Monism: the moral use of modals express propositions about whatever set of rules is the moral standard, even when that set of rules is not the best t for the agent's judgments about her moral obligations. 149 The moral use of (1b) communicates that Jane believes that ought M (You keep that promise), while accepting M as the moral standard. The legal use of (1b) communicates that Jane believes that ought P (You keep that promise), while accepting P as the standard of prudence for Jane. Pluralism: the moral use of modals express propositions about what- ever set of rules the agent accepts to be the moral standard. The moral use of (1b) communicates that Jane believes that ought A (You keep that promise), while accepting A as the moral standard. The legal use of (1b) communicates that Jane believes that ought B (You keep that promise), while accepting B as the standard of prudence for Jane. These two approaches are very dierent. 7.3.2 Why Pluralism is so important Many of the important advantages of positivist realism turn on a Pluralist con- ception of the hybrid theory. Contrast what it takes to have safe moral beliefs on the Pluralist's conception, and what it takes to have safe beliefs on the Monist's conception. For the Pluralist, it's pretty easy our moral beliefs to be safe. Our dispositional moral beliefs determine what set of rules we're thinking about. Now in evaluating whether our beliefs are safe, we only consider worlds where we form beliefs by relying on the same methods that we actually rely on. Since those meth- ods determine which set of rules we're thinking about, our moral beliefs are usually true across those nearby worlds. In other words: the Pluralist's conception of moral belief guarantees that moral beliefs are usually safe. The Monist's conception does not issue the same guarantee. Whatever methods we use to form our moral beliefs, those beliefs are always about the same set of rules. The positivist has been assuming a Pluralist conception of the hybrid theory. Consider the positivist's response to evolutionary debunking arguments. The de- bunker takes evolutionary explanations of moral belief to undercut ordinary moral knowledge. Evolutionary in uences on moral belief are supposed to be like the in u- ence of a hypnotist who is indierent to the truth: both undermine knowledge. The positivist responds that the only kind of knowledge necessary for ordinary moral knowledge ascriptions like (1a) is knowledge about a particular set of rules s 1 . (1a) Jane knows that she ought to keep this promise. At-issue: Jane knows that s 1 requires keeping this promise, and accepts that s 1 is the moral standard Not-at-issue: s 1 is the moral standard 150 Jane doesn't also have to know propositions about the realist's fundamental prop- erty being-the-moral-standard. The positivist holds that it's comparatively easy to know propositions about s 1 , because you can know about them in virtue of your dispositions to have moral beliefs. To see the role the Pluralist conception plays, consider a way of sharpening the debunker's challenge. Insist that there are (conceptually possible) worlds where we continue to form beliefs in the same way that we do, but where the moral standard is very dierent than it actually is. (Sharon Street insists on this kind of point: \as a purely conceptual matter, these independent normative truths might be anything" (Street 2008b, 208). 10 ) You can take these conceptually possible worlds to show that our beliefs aren't safe, and conclude that we don't know. The positivist responds by denying that our ordinary moral knowledge depends on a modally robust guarantee that the set of rules we accept to be the moral standard is in fact the moral standard. Even at the conceptually possible worlds that we've imagined, Jane still has de re knowledge about s 1 . Now s 1 isn't the moral standard at that world, even though it is at the actual world. But that doesn't matter, for the positivist, because moral knowledge only depends on knowledge of Jane's particular set of rules s 1 . Crucially, though, this response works only for a Pluralist. The Monist takes moral belief and moral knowledge to be about whatever set of rules is the moral standard. Beliefs at the conceptually possible world are then about a dierent set of rules than they are at the actual world. Let s @ be the moral standard at the actual world, and let s c be the moral standard at the conceptually possible world. s @ requires keeping this promise; s c doesn't. Monism At the actual world, Jane that believes s @ requires keeping this promise, and accepts that s @ is the moral standard At the conceptually possible world, Jane believes that s c requires keeping this promise, and accepts that s c is the moral standard Jane's belief at the conceptually possible world is false So Jane's belief at the actual world is not safe, and can't amount to knowledge. Pluralism At the actual world, Jane that believes s @ requires keeping this promise, and accepts that s @ is the moral standard At the conceptually possible world, Jane believes that s @ requires keeping this promise, and accepts that s @ is the moral standard 10 Justin Clarke-Doane (2012) has a helpful discussion of the importance of this point in debunking arguments { see pp. 320. 151 Jane's belief at the conceptually possible world is true So Jane's belief at the actual world is safe, and can amount to knowledge. The positivist's vindication of the safety of our moral beliefs requires our disposi- tional moral beliefs to x what set of rules what we're thinking about. That's the reason why we're still thinking about s @ at the conceptually possible world. That world is a world where the moral standard is dierent, but our dispositional beliefs are still the same. And this explanation is crucially Pluralist. Our dispositional moral beliefs are totally irrelevant for the Monist. Positivist realism then builds from two crucial building blocks. The rst build- ing block is the hybrid theory of moral judgment, and the second is the Pluralist conception of the hybrid theory. Earlier sections (x7.1 andx7.2) have focused on giving some initial evidence for the hybrid theory, without mentioning the choice between Monist and Pluralist construals of it. So I've been idealizing away from an important complication. We've just seen that this idealization is only an idealization. Positivist realism needs two building blocks { both the hybrid theory of moral judgment, and also a Pluralist conception of the hybrid theory. In the rest of this chapter, I'll develop two arguments for a Pluralist construal of the hybrid theory. The next chapter will develop a third such argument. 7.4 A hermeneutical argument for Pluralism Throughout the rest of this chapter, I'll just assume the hybrid theory. Think of this assumption as assumption for conditional proof. The goal here is to show that you should adopt the Pluralist conception of the hybrid theory if you accept the hybrid theory at all. The rst argument is simple. Moral knowledge is easy in the way that the Pluralist predicts. It only requires knowledge of the considerations that bear on the moral status of the action. It doesn't require whatever additional features that the Monist will require it to have. Think, for example, of someone who comes to believe that the fact that an action is painful is a reason not to do it. Moreover, that belief is stable. She won't give it up. The Pluralist predicts that that belief is also a piece of knowledge. I think that that prediction is correct. In that case, we would take the agent to know what she beliefs. So this prediction is a piece of evidence for Pluralism. The only way to disarm this argument is to develop a Monist view where the bar for moral knowledge is as low as the Pluralist predicts. And it is very hard to do that. The strictures that normally matter for general epistemology are just irrelevant if the Pluralist is right. It's not at all clear how you could mimic the Pluralist's predictions, while trying to satisfy those strictures. This simple argument is particularly forceful for some kinds of philosophers. Chapter 3 argued that moral naturalists are better positioned to vindicate some 152 central hallmarks of moral knowledge given positivist realism. But this argument depended essentially on a Pluralist conception of the hybrid theory. Moral natural- ists should see this argument as a hermeneutical argument for Pluralism. Chapter 13 will make a similar argument about rule consequentialists. They need a Plu- ralist conception of the hybrid theory in order to intelligibly draw on the method of re ective equilibrium. In fact, many chapters in the dissertation can be seen as providing related evidence for the Pluralist conception. Isn't there something methodologically odd about what I've just done? I've argued for a linguistic thesis (Pluralism), from some facts about moral epistemology. And it doesn't seem like moral epistemology is the sort of thing that should constrain linguistic work! I don't agree. Monism and Pluralism make the same predictions outside of intensional contexts. A sentence that's free of intensional vocabulary is true for the Monist i it's true for the Pluralist. The only way to discriminate the two views is by appeal to our intuitions about sentences containing intensional vocabulary like knows. And you can't articulate the truth-conditions for I know that killing is usually wrong without doing moral epistemology. 11 And the choice between Monism and Pluralism crucially depends on the truth-conditions of sentences containing knows. So of course we'll be drawn into moral epistemology in adjudicating the present linguistic question. Let's put together the arguments that I've given so far. I've given two arguments for the basic hybrid theory, and one argument for Pluralism. Favors the hybrid theory? Favors Pluralism? x7.1: Attitude asymmetries Yes! No x7.2: Local accommodation Yes! No x7.4: Knows ascriptions No Yes! x7.5: Chapter 8: I've been using the label \positivist realism" for a realist who accepts a Pluralist conception of the hybrid theory. We need arguments in both of my columns here to defend positivist realism. I'll go on to give more evidence for positivist realism. The positivist is pushing a big change in moral semantics, and the broader the range of evidence she can bring to bear, the better. At the same time, though, I think that enough of the heavy lifting has been done at this point. Every moral realist should accept the hybrid theory of the moral use of modals, or the dynamic alternative. (And both of those options have the same philosophical upshots.) Moral realists haven't realized that 11 Non-homophonic truth-conditions, of course. 153 they are forced to choose one of those two options { but only for bad reasons, because they've been operating with a mistaken conception of presupposition. And once we accept the hybrid theory, there is simple evidence for the Pluralist conception of it over the Monist conception. 7.5 The anthropological use of moral language This section introduces a piece of evidence that favors both the hybrid theory and the Pluralist conception of it. The argument is about the anthropological or \in- verted commas" use of moral language { the use of moral language to describe a set of demands that the speaker herself doesn't accept. Now the argument here is evidence both for the hybrid theory and for the Pluralist conception of it. So I'll start out by targeting traditional realists, who take the fundamental normative property to be part of an at-issue proposition that moral utterances communicate. That is, I'll start by targeting philosophers who reject the hybrid theory. Imagine an anthropologist asserting (5a), about the moral demands of some particular community. (5a) Children are morally required to house their aging parents. Some traditional realists might think of normal utterances of (5a) as asserting the at-issue proposition that [the x: x is the moral standard] (children are required x to house their aging parents). Other traditional realists might think of it dierently. What unies them all is that the realist property is part of the at-issue proposition communicated. But the anthropological uses of (5a) pose an important problem for these re- alists, because they don't seem to communicate the proposition it ordinarily does. It's appropriate for the anthropologist to use (5a), even if she doesn't herself accept those demands as the demands of morality. Since the anthropological use doesn't commit the speaker to believing the traditional realist's proposition, it has to be explained in a dierent way. It's standard to call this an `inverted commas' use of moral language. I think that the only way to explain this use is to accept the hybrid theory. In fact, I think that I can force you to accept the hybrid theory with a series of pointed questions. The rst question is whether the moral predicate in (5a) is used or just men- tioned. If you think that the predicate is just mentioned, you're making what I'll call the `metalinguistic suggestion'. This suggestion is highly implausible. For one thing, it's not even clear what receives a metalinguistic interpretation in (5a). Sup- pose, for example, that the predicate are morally required is mentioned and not used. Then it's unclear what proposition (5a) is supposed to communicate. In general, cognizing a proposition involves predicating a property or relation of some- thing. And it's unclear what property or relation we're supposed to predicate of something. It's only the use of a predicate that contributes a property or a rela- tion. The mention of a predicate contributes an object, an object made up of letters strung together. 154 Setting aside this problem, though, we can also test the proposal more directly. Let's start with a detour through negative polarity items (NPIs), like ever or at all. NPIs are systematically appropriate under negation, and systematically inappropri- ate in atomic sentences. I'm at all hungry and he's ever doing what he should are inappropriate, while I'm not at all hungry and he isn't ever doing what he should are both appropriate. The best extant explanation of the licensing of NPIs is se- mantic: that they are licensed in environments with a certain semantic property, a property which negation has. 12 NPIs are important evidence for our purposes, because they are only licensed by object-level negation, and not by metalinguistic negation. (6a) That car was built ten days ago. It's not old! (6b) That car was built ten days ago. It's not at all old! (6c) That car is antique. It's not old! (6d) That car is antique. ? It's not at all old! (7a) That car has always been blue. It has not been red! (7b) That car has always been blue. It has not ever been red! (7a) That car has always been maroon. It has not been red! (7b) That car has always been maroon. ? It has not ever been red! Nor is surprising that we would nd this sort of pattern. Robyn Carston has suggested that metalinguistic cases involve negation scoping over material that is echoically used. She follows Dan Sperber and Deirdre Wilson (1b986) in suggesting that \a representation is used echoically when it reports what someone else has said or thought and expresses an attitude to it" (Carston 1996, 320). You can echo a thought or an utterance but reject it, and thereby indicate that there is some- thing wrong with the thought or utterance. But no one would make the utterances that scope under the negation in (8b), if it were metalinguistic. Those utterances would be utterances of the infelicitous sentence that car is at all old. That is the reason metalinguistic negation does not license NPIs Carston emphasizes this point (Carston 1996, 321-2), as one of the central arguments for her approach. Let's return back to the main thread, about moral utterances. (5b) Children aren't at all permitted to turn away their aging parents! (5c) Children aren't ever permitted to turn away their aging parents! If the negation of (5b) were metalinguistic, NPIs like at all or ever would be marked in the way that they are marked in (8b). But that is not what we nd. Now we've already seen that the metalinguistic interpretation was not plausible about examples 12 In particular, they seem to be licensed in the scope of downward entailing operators. Nothing about this formal point matters in what follows, except that negation is downward entailing. 155 like (5a). But it's especially implausible as an account of (5b). So we should set aside the metalinguistic account of inverted commas uses. We should insist that utterances like (5a) { (5c) involve using and not just men- tioning the moral predicate. The next question is whether the use of the predicate ends up communicating a dierent relation in the anthropological use than in the sincere use. For example, maybe the use ends up communicating something about the relation is-believed-to be-permitted-by-the-moral-standard, rather than just the relation is-permitted-by-the-moral-standard. (Remember that we're considering tra- ditional realists, who take the moral property to be part of the at-issue content associated with a moral utterance.) The belief-centric idea is unpromising. The anthropologist may be drawing out subtle consequences of the community's moral beliefs. And the community might fail to recognize those consequences. The anthro- pologist can still use sentences like (5a) to express the subtle consequences of her investigation. But it won't be plausible that the community believes those subtle consequences. Another option would be to posit another avor of modality: the anthro-moral avor, which you use to talk about the moral beliefs of a community that's not your own. Sentences like (5b) express propositions about that other avor, rather than about the fundamental property being-the-moral-standard. This suggestion is also implausible. Suppose that Jake is a member of the community under study. An anthropologist could also say (5d). (5d) Jake knows that children aren't ever morally permitted to turn away their aging parents. (5d) doesn't communicate a proposition about the anthro-moral avor. (Jake is himself a morally committed agent; his beliefs are beliefs about the property being- the-moral-standard.) But the anthropologist can still use (5d). So the anthropolog- ical use can't communicate propositions about a special anthro-moral avor. More generally, (5d) is a powerful argument against a range of approaches to the anthropological use. Consider \content ctionalist" accounts of inverted com- mas uses { that is, accounts where the anthropological use of (5a) communicates a dierent proposition than the sincere use would. Maybe, for example, the anthro- pological use communicates that children aren't permitted to turn away their aging parents, according to some relevant ction. This content ctionalist account has the same problem as appealing to the anthro-moral avor does. Ordinary agents in that society like Jake take themselves to be thinking and knowing just the same propositions as the anthropologist does in her sincere moments. They don't take themselves to be thinking propositions about a ction, for example. (For one thing, Jake might disagree with the anthropologist about what morality really demands, in virtue of the knowledge that he ascribes to himself in (5d).) The content c- tionalist then predicts that (5d) is, unlike (5a), in being inappropriate. And that's implausible. In general, (5d) refutes any approach to the anthropological use that takes the predicate are morally permitted to communicate a relation other than the relation 156 that sincere uses of moral language communicate. After all, Jake's beliefs are about the relation that the sincere use communicates! Moreover, knows is factive. The speaker has to agree with the proposition that Jake believes. (So (5d) is another kind of evidence against \metalinguistic" accounts of the anthropological use. Jake might not even know English. So it's not plausible that he believes a proposition about the English predicate aren't ever morally permitted.) So the only possible explanation of the anthropological use of (5d) takes it to communicate the same proposition(s) as the sincere use. The anthropological use is dierent in the attitude required towards that property. The standard name for this kind of view is \force ctionalism". The idea is that the sincere use is associated with the same proposition as the anthropological use, but the proposition is put forward with a dierent \force". We should see positivist realism as the best kind of force ctionalist account. The positivist already thinks that a range of dierent attitudes can constitute ac- ceptance. You can accept a proposition by believing it. But you can also accept a proposition to draw out its consequences. And that's what she can say about these cases. The anthropologist is temporarily accepting the other society's set of rules as the moral standard, in order to draw out its consequences. It's not that the anthropologist is doing something metalinguistic, or communicating propositions about what that society believes. The propositions communicated are just ordinary propositions about an set of rules. It's just that the anthropologist has a dierent attitude towards the proposition that the relevant set of rules is the moral standard. This explanation is very simple and elegant. However, positivist realism is not the only kind of force ctionalism. In fact, it's not even the simplest kind of force ctionalism. The positivist predicts that the anthropological use of (5d) involves one proposition asserted with ordinary force, and one proposition that's just accepted for anthropological reasons. In other words, the anthropological use of (5d) involves: asserting that Jake knows that o j doesn't ever permit turning away aging parents and accepts that o j is the moral standard accepting that o j is the moral standard for anthropological reasons A simpler kind of force ctionalism would take the anthropological use to just involve accepting one proposition. accepting that Jake knows that [the x: x is the moral standard] (o x doesn't ever permit turning away aging parents) for anthropological reasons. This simpler alternative doesn't involve asserting any propositions with their ordi- nary force. There is powerful evidence for the more complicated positivist alternative. (5e) can be consistent in a way that (5f) can't. (5e) Speaking as an anthropologist, Jake denitely knows that chil- 157 dren aren't ever morally permitted to turn away their aging parents. Though, taking o my anthropologist's hat, I don't myself know if Jake's conception of morality is really right. (5f) Speaking as an anthropologist, Jake denitely knows that children aren't ever morally permitted to turn away their aging parents. * Though, taking o my anthropologist's hat, I don't myself know if Jake's conception of morality really does forbid children turning away their aging parents. (5e) is somewhat hard to process. But (5f) is just inconsistent. The positivist predicts this dierence. (5e) is consistent because asserting (5d) just involves ac- cepting the proposition that o j is the moral standard. When we process the second sentence in (5e), we understand that the speaker is switching what propositions she accepts. (This switch is why (5e) is a bit hard to process.) By contrast, (5f) is inconsistent. In asserting (5d), we're straightforwardly asserting that Jake knows that Jake knows that o j doesn't ever permit turning away aging parents. So (5f) is Moore-paradoxical: it has the form `Jake knows that p but I don't know if p'. (Ut- terances with this form are Moore-paradoxical because knows is factive. Utterances with that form entail something with the form `p but I don't know if p', which is paradigmatically Moore-paradoxical. ) Simple force ctionalism cannot capture this dierence. Given that simple view, all of (5d)'s commitments are on a par: they're all just accepted for anthropological purposes. None of them are straightforwardly asserted. Moreover, those commit- ments are inconsistent both with the second sentence in (5e) and the second sentence in (5f). So simple force ctionalist just can't account for the contrast between (5e) and (5f). If it predicts that (5f) is inconsistent, it predicts that (5e) is inconsistent as well. (In order to predict that (5f) is inconsistent, it has forbid changes in what propositions the speaker predicts. And if such changes are forbidden, (5e) will be inconsistent.) Similarly if it predicts that (5e) is consistent { then (5f) should be consistent as well. The only way to explain this contrast is to adopt a hybrid picture where an assertive utterance of (5d) requires merely accepting one proposition and asserting another. Note that this explanation does require a Pluralist conception of the hybrid theory. It works only if Jake's set of rules is dierent from the ones that are in fact the moral standard. The best explanation of the anthropological use is the hybrid one. 158 Favors the hybrid theory? Favors Pluralism? x7.1: Attitude asymmetries Yes! No x3: Local accomodation Yes! No x4: Knows ascriptions No Yes! x5: Anthropological uses Yes! Yes! Chapter 8: 7.6 Wrapping up I'm claiming that moral philosophers have made a systematic mistake about modal semantics, and that this mistake has made them think that a range of questions in moral epistemology are much more dicult than they actually are. This chapter has developed several strands of argument for that claim. But it should be sociologically unsurprising that moral philosophers have made this mistake. The questions we've been exploring are very subtle. It takes a lot of work to develop and adjudicate arguments that illuminate those questions. Moreover, the hybrid theory doesn't by itself have the interesting philosophical upshots that I've been concerned to detail. It's only the Pluralist conception of the hybrid theory that has those upshots. If the Monist conception is right, all the tra- ditional debates in moral epistemology are forceful as they are on totally traditional conceptions of moral semantics. The hybrid theory should be uncontroversial. The linguistic evidence that discriminates it from the alternatives breaks decisively in its favor. And the hybrid theory by itself doesn't have striking and revisionary upshots for moral epistemology. In other words: the interesting question is if the Pluralist or the Monist is right. Everyone should accept the hybrid theory. And the right way to settle the choice between Monist and Pluralism is holistic: we evaluate all the strands of evidence that bear on the choice. In eect, that's what the whole dissertation is doing. 159 Chapter 8: Shifty contextualism about epistemics Let's review where we are in the dissertation. We're in Part II. And the goal of Part II is to defend the linguistic claim about presupposition that is at the heart of the positivist's picture. As we got into the details, though, we saw that the linguistic claim actually had two crucial pieces. One piece was the conjecture that the realist's property is a presupposition of the use of modals. And the other piece is a particular Pluralist picture of just how that presupposition aects the interpretation of moral utterances. The last chapter gave what I regard as decisive evidence for the rst piece of the picture: that the realist's property is a presupposition. The last chapter also gave good evidence for Pluralism. But the evidence is less decisive than the evidence for the basic claim about presupposition. So it's worth exploring other sources of evidence for Pluralism. That's what this chapter will do. It will consider the epistemic use of modals { that is, the use of words like might or must to talk about our own evidence. I'll argue that the best explanation of the epistemic use is a Pluralist one. And I'll close by suggesting that the correct account of the epistemic use should illuminate the correct account of the moral use. I will show that Pluralism about epistemic modals picture solves two important challenges. The rst of these challenges is very familiar, and the second less so, but equally important. Imagine Ignorant Ian and Well-Informed Wendy looking for some keys. Ian: The keys might be in the drawer. Wendy: What you said is not true { I've already checked. The two challenges to the contextualist come from two central facts about this example. The rst central fact is that it can be appropriate for Wendy to disagree with Ian, to say that what he said is not true, or to say that he was wrong, or that he made a mistake. (Imagine that she knows that the keys aren't in the drawer.) The second central fact is that Wendy can acknowledge that Ian is reasonable even while insisting that he's wrong. After all, Ian might not know whether the keys are in the drawer. If he doesn't, it's appropriate and reasonable for him to think that the keys might be in the drawer. It is hard for contextualists to explain both features of the example. They can explain why the two agents are reasonable in making their utterances. The two agents are simply describing dierent bodies of information. Wendy is describing one that represents the contents of the drawer, and Ian is describing one that doesn't. 160 But this suggestion seems to get disagreement wrong. If the two parties are only describing dierent bodies of information, they wouldn't be disagreeing. If you said I'm hungry, and I said I'm not hungry, we wouldn't be disagreeing. We're just talking about dierent people. Something similar for the contextualist: the two parties would just be talking about dierent bodies of information. It's then hard for the contextualist to explain why Wendy can take herself to disagree with Ian's reasonable belief. 1 I will argue that these problems arise because we haven't thought about the relationship between modals and attitude reports in a systematic enough way. We can use the lessons from the previous two chapters to do better. 8.1 The new theory This section extends the positivist's account of the moral use to the epistemic use. But we need one new resource to do that. That new resource is a new property which I will epistemic correctness: the property of containing all truths. A modal base is epistemically correct if it contains all truths. And one modal base is more epistemically correct than another if the rst contains more truths. I propose that judgments about what might be involve accepting a modal base as epistemically correct. For example, asserting an epistemic is a hybrid state, consisting of two attitudes. It both consists in asserting something about what's compatible or incompatible with a particular modal base, and also consists in ac- cepting that modal base as epistemically correct. Here's an example. In asserting that the keys might be in the drawer, I'm asserting that some modal base m 1 is compatible with the keys being in the drawer, and 1 There has been an explosion of work on this challenge { articulations of important theoretical approaches and arguments for and against. There isn't always perfect clarity about the relations between these dierent approaches, as Mark Schroeder (2015) de- scribes in detail. But broadly expressivist approaches (or approaches that draw heavily on expressivist resources) include at least Ernest Adams (1975), Joshua Knobe and Seth Yalcin (2014), Ben Lennertz (2013, 2014), Jonathan Bennett (2003), Dorothy Edgington (1995), Martin Montminy (2012), Sarah Moss (2013, 2015), Benjamin Schneider (2010), Eric Swanson (2006), and Seth Yalcin (2007, 2011). Broadly relativist approaches include at least Andy Egan (2007), Egan, Hawthorne, and Weatherson (2011), John MacFarlane (2007, 2011, 2014), and Tamina Stephenson (2007). Broadly contextualist approaches in- clude at least Pranav Anand and Valentine Hacquard (2013), Kent Bach (2011),Gunnar Bj ornsson and Alexander Alm er (2010), Gunnar Bj ornsson and Stephen Finlay (2010), David Braun (2012), Herman Cappelen and John Hawthorne (2009), Ezra Cook (2013), Janice Dowell (2011, 2013), Angelika Kratzer (1977, 1981, 2012), Kai von Fintel and An- thony Gillies (2008, 2011), John Hawthorne (2007), Tornn Huvenes (2015), Justin Khoo (2015), and Crispin Wright (2007). Broadly dynamic views include at least Frank Veltman (1996) and Malte Willer (2013, 2014). 161 accepting that m 1 is epistemically correct. Other attitudes are also a hybrid of two attitudes. Belief and knowledge, for exam- ple, both are. If you believe that the keys might be in the drawer, you believe that m 1 is compatible with the keys being in the drawer, and accept that m 1 is epistemically correct. If you know that the keys might be in the drawer, you know that m 1 is compatible with the keys being in the drawer, and accept that m 1 is epistemically correct. 8.1.1 How Pluralism explains both disagreement and reasonable- ness Pluralism explains both of the central desiderata that we want to explain. It explains how two agents can disagree while both being reasonable. Say that Ignorant Ian and Well-Informed Wendy are looking for some keys. Ian: The keys might be in the drawer. Wendy: No, they can't be there. (I've already checked.) Modal bases are, let's say, sets of propositions. Ian is talking about m 1 , and Wendy is talking about m 2 . m 1 =fthe keys are not in the garage, the keys are not in the kitcheng m 2 =fthe keys are not in the garage, the keys are not in the kitchen, the keys are not in the drawerg Pluralism then makes these two predictions. Ian asserted that m 1 is compatible with the keys being in the drawer, while accepting that m 1 is epistemically correct. Wendy asserted that m 2 rules out the keys being in the drawer, while accepting that m 2 is epistemically correct. Then Wendy and Ian disagree in what they accept. They are both accepting proposi- tions that can't both be true. The proposition that m 1 is correct and the proposition that m 2 is correct can't both be true, since it can't be that both contain all truths. 162 That's the simple account of disagreement. How can the two agents nonetheless be reasonable? Reasonable belief is also a hybrid state, consisting of acceptance of some modal base as epistemically correct, plus reasonable belief about that modal base. It crucially does not require reasonable belief about what's epistemically correct. So Ian is reasonable because he has reasonable beliefs about what is com- patible with his own modal base, the one that leaves open whether the keys are in the drawer. Now the Pluralist is expanding the kind of disagreements that the contextual- ist can explain. She is suggesting that our two agents can be disagreeing about what's epistemically correct. Importantly, though, she still allows for the sort of disagreement that more traditional contextualists can explain. Suppose that two mathematicians are disagreeing about what follows from a set of axioms. One might assert that p must be true, while the other asserts that p might be false. In this case, they would be disagreeing about what's compatible with some modal base: the set of axioms. Moreover, we can evaluate the reasonableness of the dierent beliefs about what's compatible with the axioms. We might deny that a mathematician is reasonable in believing that p must be true, if we don't think that the mathemati- cian has a proof of p. Traditional contextualists have an elegant explanation of this kind of case, and Pluralism retains that elegant explanation. The Pluralist account of disagreement generalizes straightforwardly to all cases of disagreement, including eavesdroppers. Imagine that Ian is talking with Wendy. Eavesdropping Evelyn isn't part of the conversation, but she knows where they are. Ian: The keys might be in the drawer. ... Eavesdropping Evelyn [to a friend]: No, they can't be there. (I've already checked.) Many contextualists struggle to explain this exchange, but the Pluralist explains it crisply. The Pluralist takes disagreement to be disagreement about what's epistem- ically correct. But she thinks of epistemic correctness in a striking and distinctive way. It's a comparative notion: some modal base is more correct than some group of bases. Which ones? Any modal base whatsoever. Let's illustrate with Ian. He's thinking about some modal base m 1 . In accepting that m 1 is epistemically correct, he's taking it to be more epistemically correct than any other modal base what- soever. Now m 1 is incomplete. It doesn't say where they keys are. Imagine that does know where the keys are: say, that they're in the drawer. Then she accepts a modal base m 2 that includes that proposition, as epistemically correct. That's why they disagree. Evelyn accepts that m 2 is more correct than m 1 , and Ian accepts that m 1 is more correct than m 2 . This proposal works only because it takes Ian to be doing something very odd. He's accepting a proposition that he's in a position to know is false. Ian knows that either the keys are in the drawer or they're not. He also knows that if the keys are in the drawer, m 1 is not epistemically correct, because m 1 plus the proposition about the keys is more correct. And he knows that some other modal base if more 163 correct if they're not in the drawer. So Ian is in a position to know that it's false that m 1 is epistemically correct. In general, though, it can perfectly appropriate to accept propositions that you can know to be false. Here are two examples. First: you can accept that Hamiltonian mechanics is true and relativist mechanics is false to simplify your calculations in an engineering context. In doing that, you're accepting a proposition that you know to be false. Second: some religious people will insist that God loves his children, while also insisting that God isn't male. Since his presupposes that the salient individual is male, they're accepting a proposition that they don't believe. In this case, they're accepting it because they regard it as the least unsatisfactory option. Other possessive pronouns (in English) presuppose that the individual is female, or that the group has several members. The central feature of my account of epistemics is that epistemics pattern with the use of Hamiltonian mechanics, or gendered language for God. In both cases, your practical interests can make it perfectly appropriate to accept propositions you know to be false. The practical interest in the epistemic case is to reason from the largest amount of information that you have, without relying on anything that your information leaves open. The latter goal is why ordinary agents only accept an incomplete modal base as epistemically correct. 8.1.2 Contrasting the Pluralist with other contextualists The Pluralist strictly improves on the approaches that other contextualists have developed. To see why, imagine that Wendy told us (1). (1) Ian is heading towards that drawer, because he reasonably believes that the keys might be there. But I know that they can't be there; I've already checked. So it would be a mistake to follow Ian in believing that the keys might be there. On the hybrid suggestion, uses of (1) communicate these three propositions. Ian is heading towards that drawer, because he (reasonably be- lieves that m 1 is compatible with the keys being there and accepts that m 1 is epistemically correct). But I, Wendy, know that m 2 is incompatible with the keys being there and accept that m 2 is epistemically correct). So Ian is wrong in (believing that m 1 is compatible with the keys being there and accepting that m 1 is epistemically correct). Dierent modal bases are incorporated into the complements of Ian's knowledge reports than are incorporated into the complements of Wendy's knowledge reports. Why? Because the two agents take dierent modal bases to be epistemically correct. This suggestion contrasts with an important contextualist suggestion from Jan- ice Dowell (2011, 2013). She holds that, when one person assertively utters an 164 epistemic and her interlocutor assertively utters its negation, the two parties are disagreeing because they're expressing propositions about the very same modal base. Dowell gives a powerful defense of this proposal, suggesting that speaker intentions determine the modal base incorporated into the propositions expressed, and that speakers can intend to talk about the same modal base even if they don't know exactly what modal base they're talking about. We can simplify the compar- ison with Dowell by pretending that speakers intend to talk about whatever modal base is epistemically correct in the present context. Dowell thinks that they're talking about whatever base is in fact epistemically correct, while Pluralism takes them to be talking about dierent particular modal bases which they accept to be epistemically correct. Does Dowell's view dier substantively from Pluralism? Yes. The two make dif- ferent predictions about the reasonableness of judgments about what might be. The Pluralist predicts that it's really easy to have reasonable beliefs about what might be. Reasonable belief depends only on reasonable belief about what's compatible with a set of propositions that you're already thinking about. Dowell does not make it as easy to have reasonable beliefs about what might be. Since Ian and Wendy disagree, the two attitude reports should express propositions about the very same modal base. But it's hard to see how Ian could have reasonable beliefs about that modal base, unless he is sensitive to the information that Wendy takes to bear on the question. And it's unclear how he could be sensitive to that information. He isn't aware of what it is! We can appreciate another distinctive feature of the hybrid picture by comparing it with a proposal by Kai von Fintel and Anthony Gillies (2011). Properties like epistemic correctness play no role in their proposal. They instead propose that utterances of epistemic modals communicate a cloud of propositions, propositions about several dierent bodies of information. One of those propositions describes only the speaker's body of information; another describes the body of information of the entire group. They suggest that it's appropriate to assert an epistemic modal when you accept one of those propositions { typically, the one about your body of information. But your assertion puts the other propositions in play. So other speakers can react to your utterance by reacting to some of the other propositions put in play. This proposal explains why it's easy for a speaker to appropriately assert an epistemic. It's appropriate as long as her own information is compatible with the prejacent. But it also explains why her hearers can object. They're objecting to the other propositions put in play. Their proposal doesn't generalize to the attitude reports under discussion. (1) Ian is heading towards that drawer, because he reasonably believes that the keys might be there. But I know that they can't be there; I've already checked. So it would be a mistake to follow Ian in believing that the keys might be there. What proposition does Ian believe? It seems like he only believes the proposition about his own body of information. (Ian may have no beliefs about any other salient 165 bodies of information; he may suspend judgment on them.) But then it's unclear what Ian is mistaken in believing. He is not mistaken in believing the proposition about his own body of information. That proposition is true. But that proposition may be the only relevant proposition he believes. Since pIan is wrong in believing Sq entails that Ian believes S, it's hard to see how von Fintel and Gillies can explain this case. 2 Pluralism does much better. 3 The new contextualist account makes it much easier to see how someone can have reasonable but mistaken beliefs about what might be. However, there is a danger that the account makes it too easy to have mistaken beliefs about what might be. Virtually all `might' beliefs involve accepting an incomplete modal base as epistemically correct. (Ordinary agents aren't omniscient, and mostly know that they aren't.) But we saw in the last section that no incomplete modal base could be epistemically correct. So the account seems to predict that all ordinary `might' beliefs are mistaken. And this prediction would be very costly. For one thing, I don't regard my own `might' beliefs as mistaken, whatever this account might predict. 4 Fortunately, though, the proered account does not make this prediction. Dis- tinguish two pictures of a mistake in acceptance. On a belief-centric picture, you can appropriately say pit would be a mistake to accept pq if you yourself believe:p. On an acceptance-centric picture, you can appropriately say pit would be a mistake to accept pq if you yourself accept:p. The acceptance-centric picture turns out to disarm the problem in the previ- ous paragraph, while also being signicantly more plausible than the belief-centric alternative. I'll describe the dierence between the two using gendered religious language. Suppose that Bob and Carl both use gendered pronouns for God, be- cause they regard those pronouns as the least inadequate option. But they believe that God is neither male nor female. Now since his presupposes that the salient individual is male, an utterance of (2) communicates the proposition (+). (2) It would be a mistake to follow Carl in believing that God infuses grace in his children. (+) It would be a mistake to follow Carl in (believing that God infuses grace in God's children and accepting that God is male) The belief-centric picture predicts that Bob can appropriately say (2) even if Bob agrees that God infuses grace in God's children. After all, Bob doesn't believe that 2 Ben Lennertz (2013) has emphasized this kind of problem, in his chapter 2,x5. 3 It's also important to compare Pluralism with other similar contextualist approaches. Several contextualists supplement their approach an expressivist pragmatics: examples include Stephen Finlay and Gunnar Bjornsson (2010), Ben Lennertz (2013), and Martin Montminy (2012). The hybrid approach is importantly dierent from those approaches, but it's more useful to contrast the two approaches late in the paper, inx4.2. 4 I'm grateful to an anonymous Ergo referee for posing this objection in a particularly forceful way. 166 God is male. (He believes that God is neither male nor female.) Since he doesn't believe the proposition Carl accepts, the belief-central account predicts that (2) is appropriate. And that prediction is incorrect. Bob is ex hypothesi perfectly happy to use gendered pronouns for God { he doesn't disagree with Carl about his use of pronouns. In order for (2) to be appropriate, Bob and Carl have to disagree about the infusion of grace. The acceptance-centric picture, by contrast, avoids this prediction. That picture grounds mistakes in acceptance in disagreement in ac- ceptance. Since Bob and Carl both accept that God is male, Bob can appropriately say (2) only if he disagrees with Carl about whether God infuses grace in God's children. And our initial problem about ordinary `might' beliefs does not arise on the acceptance-centric conception. I grant that we should believe that no incomplete modal bases is epistemically correct. (The last page gave a compelling argument for that point.) But that belief doesn't require us to accept that no incomplete modal base is epistemically correct. On the contrary, we have very good reason to accept incomplete modal bases as epistemically correct. And in diagnosing mistakes in acceptance, I reason from the propositions that I myself accept. That's what explains why someone who's less informed than me is mistaken what modal base they accept to be epistemically correct. Given the propositions that I myself accept, one of the propositions that they accept is false. Reasoning from what I accept also explains why I won't regard myself as mistaken about what I accept. Given the propositions that I myself accept, all the propositions that I accept are true. 5 Now there are lots of fruitful and interesting questions to explore about the acceptance-centric conception, and we'll take up some of them as the paper goes on. The important point for now is that that conception has the right structure for a Goldilocks solution to the contextualist's problem: not too few mistaken beliefs about what might be, and not too many, either. 8.2 A systematic picture of attitudes towards epistemics The broader goal of this paper is to connect the contextualist's problems about disagreement to more general questions about attitudes about what might be. The last section showed that it's possible for contextualists to explain epistemic disagree- ment, while also explaining reasonable epistemic belief. The explanation draws on 5 This observation explains why it's Moore-paradoxical to assertively utter the keys might be there but I am wrong in so believing. In asserting the rst conjunct, you're asserting that some modal base o1 is compatible with the keys being there and accepting that o1 is epistemically correct. And in asserting the second conjunct, I'm asserting that I'm wrong in (believing that o1 is compatible with the keys being there and accepting that o1 is epistemically correct). When I'm evaluating whether it's a mistake to accept thato1 is epistemically correct, I reason from the propositions I accept, which includes the proposition that o1 is epistemically correct. Since that proposition trivially entails itself, it's not a mistake to accept thato1 is epistemically correct. But then the only way for the second conjunct to be assertable is for me to not believe that o1 is compatible with the keys being there. So the assertion has Moore-paradoxical assertability-conditions. 167 two distinctive tools: the property being-epistemically-correct, and the attitude of acceptance. But it's reasonable to worry that these tools are both ad hoc { that is, unmotivated additions to the basic contextualist idea. This section introduces a systematic account of attitudes about what might be. It shows how those two tools are essential parts of that systematic account. That is, it shows how thinking more systematically about attitudes about what might be illuminates the questions in the last section. After doing this, I'll explain how the account introduced in x1 bears a range of other exciting fruit, include an explanation of the \epistemic contradictions" that Seth Yalcin (2007) has highlighted. 8.2.1 A central but underappreciated question My systematic account answers a central but underappreciated question about the epistemic use of modals. This question is genuinely underappreciated. So I will begin by arguing for the importance of this question, and arguing for it in some depth. This central question is about attitudes towards epistemics. It's an instance of a more general question about attitudes towards modals, like (3). (3) Jill believes that she must keep this promise. The positivist has a distinctive picture of what utterances of (3) communicate. But let's set that distinctive picture aside for a page, and think more systematically about (3). Everyone should agree (3) can be used to talk about the demands of morality, the demands of prudence, or the demands of some particular set of laws. It's standard to explain this point by supposing that she must keep this promise has one more argument place than its surface syntax suggests. The moral use of (3) communicates a proposition with that argument place saturated with something that represents the demands of morality, the prudential use with something that represents the demands of prudence, and so on. The positivist is exploring a central question arises after we explain what all the uses of (3) have in common. She's interested in asking what distinguishes the dierent uses of (3). What distinguishes the proposition the moral use communi- cates from the proposition the legal use communicates? The positivist rejects a Naive Thought that many other philosophers have tacitly accepted. That Naive Thought holds that they're distinguished because the propositions believed have dierent truth-conditions. NaiveThought: The moral use of (3) communicates that Jill believes a proposition with dierent truth-conditions than the proposition that the legal use of (3) takes her to believe. Though I've been assuming that this Naive Thought is wrong, it's worth noting that it is pretty natural, at least at rst. Jill's moral belief is dierent from her belief about the laws, because the two propositions have dierent modal proles. The moral proposition is true because of what the moral facts are, and the legal 168 proposition is true because of what the legal facts are. And the moral and legal facts don't always line up. Unfortunately, though, the Naive Thought doesn't work in full generality. It is possible for (i) the laws of some society to coincide exactly with the laws of morality, across all of modal space, even while (ii) someone gets confused, and thinks that the laws don't coincide. Imagine that some set of laws does coincide with morality in just that way. It's still possible for Jill to think that she's morally required to keep this promise, without thinking that she's legally required to. In that case, the moral use of (3) asserts something true, and the legal use assert something false. But the Naive Thought can't explain this fact. In this case, the moral proposition is true at a world w i the legal proposition is true at w. There isn't any dierence in modal prole for the Naive Thought to latch onto to explain how the moral but not the legal use of (3) asserts something true. Let's make this point more concrete. We might model the demands of morality with orderings on worlds: w 1 is ranked higher than w 2 , which is tied with w 3 , which are both above... And I'm imagining a case where the very same ordering that models the demands of morality also models the laws of some society. (The society designed its laws so that they correspond with morality throughout all of modal space.) Call that ordering o 1 . Given the Naive Thought, the moral and legal uses of (3) both communicate that Jill believes that o 1 requires her keeping this promise. And that won't work, as it stands. The moral use asserts something true and the legal use asserts something false. As a result, we need to draw on richer resources than the Naive Thought uses. We need to explain how Jane can be thinking of o 1 in dierent ways, as the demands of morality, or the demands of positive law. I will call the dierent ways we think about the orderings modes of presentation. You might think of these modes of presentation as properties that we predicate of orderings. For example, I've appealed to the property being-the-moral-standard in developing the positivist's idea. I've been tacitly assuming that that property is something like a mode of presentation for the salient ordering. This assumption is one way of developing an important realist conviction: that the moral use of (3) communicates that the subject has some attitude towards the realist's fundamental property. Now there are also other ways of developing this important realist conviction. You might think of modes of present as the concepts that we use to cognize the orderings. Or you might think of them as grist for Stalnakerian diagonalization. 6 But however we work out the details, we can say that the moral use of (3) asserts that Jane believes that o 1 (thought of as the demands of morality) requires her keeping this promise, which can be true even if she doesn't think that o 1 (thought of as the demands of the law) requires her keeping this promise. 6 In setting aside the Naive Thought, I am not taking a stand on the nature of propositions { whether they're individuated more nely than sets of worlds. I'm just pointing out that you need to draw on whatever tools you use to capture hyper-intensional dierences in your account of modals. 169 8.2.2 Back to the epistemic case Let's return to epistemic modals. I've just given an argument that it's necessary to posit modes of presentation to distinguish the moral use of modals from the legal use. Once we're familiar with this kind of argument, we're in a position to appreciate how to generalize it for the epistemic use. The argument requires two dierent uses of modals: in the initial case, moral and legal uses. So I need a use of modals that pairs with the epistemic use, to power my argument. I will use what I'll call the circumstantial use of modals. We can say the tree can grow to be ten feet tall to indicate something about its nature: that it's compatible with its nature that it grows to be that tall. Given this use, we can construct the same sort of case constructed earlier. It's possible for someone's information to coincide exactly with the the facts that are relevant for the circumstantial use. Say that John nds himself in that situation. There is still a dierence between epistemic belief ascriptions and circumstantial belief ascriptions. (4a) can be appropriate even if (4b) isn't. (4a) John believes that this tree might grow to be ten feet tall. (4b) John believes that this tree can grow to be ten feet tall. John is aware of the information that he has. He knows that that information is compatible with the tree growing to be ten feet tall. (4a) is then perfectly appropriate. But he doesn't realize that that information coincides with the facts about the nature of the tree. In fact, he suspends judgment about the nature of the tree. Then (4b) isn't appropriate. The best way to explain this contrast is to posit a mode of presentation for the circumstantial use, and a mode of presentation for the epistemic use. (4a) and (4b) both communicate propositions about the same body of information { the same modal base, to use a phrase from Kratzer. 7 But (4b) is inappropriate even though (4a) is appropriate, because appropriate use of (4b) requires John to think of that body of information as being about the nature of the tree. The explanation is exactly like the explanation of other instances of Frege's puzzle. John thinks that Hesperus is bright can be appropriately assertable even if John thinks that Phosphorus is bright isn't. Why? The use of Phosphorus suggests a dierent mode of presentation than the use Hesperus does. So we should posit a mode of presentation to distinguish the epistemic use of modals. 8 7 This assumption is very important for this style of argument. And I haven't defended it here. A full defense of it would need to show that there are only two unpronounced arguments-places: one for something like a modal base, and one for something like an ordering or an ordering source. But I take Angelika Kratzer (1977, 1981, 2012) to have given powerful considerations in favor of that claim, and for those considerations to have been borne out in the subsequent literature. 8 Other accounts of the dierence between (4a) and (4b) might tempt you. You might note the syntactic dierences that Valentine Hacquard (2006) notes between epistemic and root modality, and take those syntactic dierences to make some kind of semantic dierence. I'm happy to allow for this point. Even granting it, though, why need some 170 Appreciating this point puts us in a better position to appreciate the Pluralist view fromx1. That theory was using a particular account of the mode of pre- sentation that distinguishes the epistemic use. It appealed a modal base being epistemically correct { that is, as containing all truths. It uses that property as the mode of presentation for the epistemic use. (4a) is appropriate because the modal base that John accepts to be epistemically correct is compatible with the tree growing that tall. Now remember that the overarching goal of this section is to show that the Pluralist picture of epistemic modals is highly principled, and not ad hoc. I'll pursue that goal by showing that that picture is a simple and natural answer to what I will call the Central Question. Central Question: what mode of presentation is associated with the epistemic use of modals, and how is it associated with that use? Features of the hybrid account that initially look ad hoc do not look the same way when we see them as answering this question. The appeal to epistemic correctness is a natural idea about the mode of presen- tation that distinguishes the epistemic use. And we should be willing to entertain this idea, once we see that we do need some such mode. (The conclusion will brie y consider possible modes of presentation that distinguish the epistemic use.) Now the positivist has a general and distinctive account of how modes of presen- tation gure in what attitude reports communicate. Applying that account to the epistemic use associates the epistemic us of (6) with the following pair of attitudes. (6) Jane thinks that the keys might be in the kitchen. Jane thinks that m 1 is compatible with the keys being in the kitchen, and Jane accepts that m 1 is epistemically correct. Is the appeal to epistemic correctness ad hoc? Not at all Epistemic correctness is introduced as the mode of presentation that distinguishes beliefs about what epistemically might be, from other kinds of modal beliefs. Pluralism is not ad hoc. Instead, it is one the simplest and most conservative ways of extending a basically Kratzerian understanding of modals, to answer a question that everyone needs to answer. 8.2.3 Consequences of this explanation of Pluralism This presuppositional explanation of Pluralism has an important advantage. To appreciate this advantage, let's recall one reason the Pluralism initially looked im- plausible. That theory takes ordinary talk about what might be to involve accep- account of why (4a) can be appropriate even while (4b) isn't. And I'm claiming that the best account of that dierence appeals to modes of presentation, which may be interestingly informed by syntactic considerations. 171 tance of a proposition that the speakers can know to be false. Asserting that the keys might be in the drawer means asserting that your own body of information m 1 is compatible with the keys being in the drawer, while also accepting that m 1 is epistemically correct. In accepting that m 1 is epistemically correct, you're ac- cepting that it's more correct than any modal base whatsoever. Now you're not omniscient. So you're in a position to know that the proposition that you accept is false. Why? Let m 1 be the modal base that you accept to be epistemically correct. Since you're not omniscient, m 1 is incomplete. So there is some way of extending it to be more complete: there's some p where adding either p or:p would make it more complete. If p is true, m 1 plus p is more epistemically correct. And if:p is true, m 1 plus:p is more epistemically correct. m 1 can't be epistemically correct; in order for it to be, it would have to be more correct than every other modal base. This feature of the hybrid account cleanly explains eavesdroppers. It's always reasonable for a better-informed eavesdropper to disagree with you, because they reject the proposition you accept: that m 1 is epistemically correct. But this feature of the hybrid account might look implausible. Ordinary discourse about what might be just doesn't involve this sort of commitment! I appealed to two analogies to make the idea more plausible. One was about religious language: religious people who use a gendered pronoun to talk about God, even though they don't believe that God is male. I suggested that they're accepting that God is male, for conversational purposes. Now I take pronouns like he and him to trigger the presupposition that the salient individual is male. JheK = f <et> At-issue: x is f Not-at-issue: x is male An utterance of God loves his children presupposes that God is male. 9 It's nonethe- less intelligible for a religious person to make that utterance, even if they don't believe this presupposition. They only need to accept it for their conversational purposes. The Pluralist takes the epistemic use of modals to involve accepting a propo- sition you don't believe. And the analogy with gendered pronouns is quite tight. The two phenomena have exactly the same linguistic explanation: they are both due to the fact that you only need to accept presuppositions of your utterance. Though this picture of the epistemic use is initially surprising, it ends up having a lot going for it. The epistemic use involves temporarily treating a body of information as the only truths. That's why any eavesdropper will disagree with you. They don't treat that body of information as the only truths. But you yourself have great practical reasons for treating that body of information as the only truths. You have practical reason for not relying on any more complete body of information: any such body of information will contain a proposition that you're not sure about. 9 See Robin Barwise and Cooper (1981), Philippe Schlenker (2003), and Yasutada Sudo (2012) for a defense of the claim. 172 My presuppositional explanation of Pluralism makes the theory signicantly more plausible. It makes the theory much more simple and elegant. And it shows that natural language already has the resources that the Pluralist uses in explaining the data: the analogy with gendered pronouns is exact. But the presuppositional explanation gives rise to a further question. I've emphasized that Pluralism can explain disagreement about what might be. And given the presuppositional expla- nation of Pluralism, disagreement about what might be is usually disagreement in presupposition. We're disagreeing about which modal base is epistemically correct. But you might wonder if disagreement in presupposition licenses the full range of disagreement markers that are in fact licensed. After all, it's very odd to use certain locutions if you want to reject the presuppositions of an utterance. (7) It wasn't Bill who stole the tarts. (8) ?? [No/ that's wrong], because the tarts weren't stolen. No and that's wrong seem appropriate only as objections to what (7) asserts (that Bill didn't steal the tarts), not as objections to the presupposition. Epistemics pattern quite dierently. No and that's wrong are both perfectly appropriate. (7 0 ) The keys might be in the drawer. (8 0 ) [No/ that's wrong.] They can't be there. This problem is a perfectly general problem about the positivist's approach. And it is a very important problem. All of Part III is devoted to exploring it. So I want to bracket it here, until we take up the perfectly general problem in the next chapter. 8.3 Yalcin's \epistemic contradictions" This account bears exciting fruit for other questions about the epistemic use. One example is about a puzzle from Seth Yalcin. He has noted that (15) and (16) both seem defective. (15) # Suppose it is raining and it might not be raining. (16) # If it is raining and it might not be raining, then... (Yalcin 2007, 985) And he has argued that traditional contextualists can't explain this data. But my hybrid theory can. It associates the utterances with the following logical forms: (15lf) At-Issue: suppose it is raining and m 1 is compatible with it not raining Not-at-Issue: m 1 is epistemically correct 173 (16lf) At-Issue: if it is raining and m 1 is compatible with it not raining, ... Not-at-Issue: m 1 is epistemically correct The content supposed and the content in the antecedent are each inconsistent with the not-at-issue content. If it is raining and m 1 is compatible with it not raining, then there is another modal base that is more epistemically correct than m 1 : m 1 plus the proposition that it is raining. In general, this sort of inconsistency leads to signicant infelicity. (17a) # Suppose that it's not raining and John knows that it's raining. At-Issue: suppose that it's not raining and John knows that it's raining Not-at-Issue: it's not raining (17b) # If John knows that it's raining and it's not raining, ... At-Issue: if it's not raining and John knows that it's rain- ing, ... Not-at-Issue: it's not raining In both of these cases, the content supposed and the content in the antecedent are each inconsistent with the not-at-issue content. In fact, presuppositions in general behave like this example illustrates. 10 For my purposes, it doesn't matter why they behave like this { only that they do. 11 Since presuppositions in general behave like 10 Here are three more examples: (17c) # Suppose it never has rained and it just stopped raining. (17d) # Suppose that no one stole the tarts and it was John who stole the tarts. (17e) # Suppose that no one stole the tarts and the tart thief was English. 11 You might think that this example has nothing at all to do with presupposition. The at-issue content is itself contradictory { so it's no surprise that we hear them as inconsistent! In making this suggestion, you are assuming that presuppositions are entailments of their triggers. That is, you're assuming that if an utterance of a sentence S presupposes p, then p is one of S's entailments. Someone who thinks this needs to explain why attitudes towards presupposition triggers don't require the same attitude towards the presupposition. (You can hope that John knows that his mother died, without hoping that his mother died.) Once you explain this point, I will adopt your explanation to my hybrid theory. I will say that pMight pq entails that the salient modal base is epistemically correct, and draw on your account of why attributions of reasonable belief don't require reasonable belief about that entailment. 174 this, Pluralism explains Yalcin's data, without any further assumptions. 12 Pluralism also explains some further facts about Yalcin's data. Cian Dorr and John Hawthorne have noted that there are special cases where constructions like Yalcin's are assertable. (18) If there is no spider in the closet and I am having a fatal heart attack because there might be, then I am about to die an absurd death. (Hawthorne and Dorr 2013, 877) Similar constructions are possible with presupposition triggers, like the Krugman- esque (19). (19) If Obama hasn't greatly increased government spending, and the budget is under discussion because everyone knows that Obama has greatly increased government spending, the discussion of the budget will be a mess. It is possible (if dicult) for a presupposition to be somehow cancelled, in the way that (19) illustrates. 13 Given this point, Pluralism predicts that there would be examples like (18), where the presupposition about epistemic correctness is some- how cancelled. And if that presupposition is somehow cancelled, (18) is perfectly consistent. Moreover, it's plausible that the presupposition would be cancelled in this context. A use of (18) doesn't seem to presuppose that the salient modal base is epistemically correct! But Pluralism also predicts that this sort of exam- ple is possible only in special cases, because presuppositions can be canceled only in special contexts. Pluralism has exactly the right structure to capture Yalcin's phenomenon. 14 12 Alex Silk (forthcoming) has suggested a similar account of Yalcin's data, an account which also draws on a presupposition. But the distinctive contribution of this paper is its account of reasonable belief and disagreement, not its account of Yalcin's data. I'm mentioning Yalcin's data only to illustrate how views like mine can capture it. 13 Robert Stalnaker (1974) in uentially unied a range of otherwise dicult data by supposing that this is possible, and Scott Soames (2009a) notes important further points about this phenomenon. Dorit Abusch (2010) has an especially helpful discussion of these points. 14 Silk (forthcoming) discussed this point in much more detail: see hisx3.4. There are some structural similarities between this presupposition explanation and the one from Hawthorne and Dorr (2013). The explanations are still quite dierent, as Silk helpfully brings out. Their explanation appeals to a conversational implicature, not to a presup- position. And my explanation follows from a simple and highly conservative conjecture about the epistemic use of modals. 175 8.4 Wrapping up It's helpful to close by contrasting my approach to the epistemic work with some supercially similar work by Stephen Finlay and Gunnar Bjornsson (2010). 15 They start by noting our practical reasons for using the most complete body of informa- tion available. The fuller body is less likely to lead us astray. So when Wendy says that Ian is wrong to think that the keys might be in the kitchen, she is recommend- ing against relying on the information that Ian has. This account of disagreement is dierent than any account from previous sections. It's not disagreement about what's compatible with a modal base. But neither is it disagreement about which modal base has some further property, like being epistemically correct. Both sorts of disagreements are disagreements about which propositions are true. By contrast, Finlay and Bjornsson are thinking of the disagreement as practical disagreement: disagreement about what to do, rather than what to believe or what to accept. Now Finlay and Bjornsson are interested in epistemic expressions in part be- cause they accept a broadly relativist conception of moral belief. Finlay, for ex- ample, defends an end-relational semantics for moral sentences, where they express propositions about our ends. 16 He allows that dierent agents may have dierent ends even when they're talking about what they morally ought to do. When I say you ought to keep this promise and you disagree, we may well be talking about dierent ends. And we may be making true claims about our dierent ends. But we'll continue to disagree morally even when we agree about what would promote various ends. So he needs some additional account of moral disagreement that goes beyond the propositions that are communicated. He lls that need with a broadly expressivist account that parallels his treatment of epistemics. Finlay and Bjornsson explicitly take their account of epistemic expressions as a license for optimism in defense of their relativism. They write: \since contextualism invokes the very same sort of pragmatic considerations (in particular an appeal to privileged contexts) to account for both sorts of context-insensitive assessments, no signicant extra cost is added to defend standard relativity" (Finlay and Bjornsson 2010, 36). The Pluralist will under undercut this license for optimism. You don't need expressivist resources to defend the basic contextualist idea against problems about disagreement. Or so I claim. But I haven't made good on this promise, yet. The whole of the next section will be devoted to making good on it. If I can succeed, though, the epistemic use won't be any evidence that the expressivist's fully general problems can be solved. For example, it's not evidence that the Frege-Geach problem can be solved. So a fortiori there's no guarantee that Finlay or Bjornsson's metaethical relativism is defensible. In fact, the comparison with Finlay and Bjornsson gives us a more complete picture of the hybrid account this paper has introduced. Finlay and Bjornsson im- plicitly adopt the same account of the epistemic use of modals as of the moral use. 15 Ben Lennertz (2013) and Martin Montminy (2012) have made similar suggestions. 16 Finlay (2014) 176 In both cases, the propositions communicated aren't enough to explain all cases of disagreement, so they also draw on expressivist tools. But this uniform treatment of the epistemic use and the moral use is somewhat unusual. It's more common for contextualists to treat the moral uses of modals quite dierently than they treat the epistemic uses. To explain the moral use, they introduce a mode of presen- tation being-the-moral-standard, over and above the ordering or ordering source. This mode of presentation explains moral disagreement. Moral disagreement is dis- agreement about what is the moral standard. But extant contextualists tend not to use the same strategy for epistemics. They don't rely on an additional property over and above the modal base. The whole situation is very odd. Contextualism is appealing as a way of unifying all the dierent uses with a common semantic core. But extant contextualists tend to use that common semantic core in a very dierent way in the epistemic case than in the moral case. The Pluralist is distinctive, because she uses the same strategy in both cases. In that way, her approach is like Finlay and Bjornsson's. Unlike them, though, she is tacitly assuming an altogether traditional picture of moral disagreement. It's al- ways disagreement about what is the moral standard. The Pluralist generalizes this traditional picture from the moral case to the epistemic case. Epistemic disagree- ment is also disagreement about what's epistemically correct. There is, however, a good reason why most contextualists haven't extended the traditional strategy from the moral case to the epistemic case. It's hard to identify a property like being-the- moral-standard for the epistemic use that (2) covers all cases of disagreement, and (b) is something that ordinary agents can have reasonable beliefs about. Some philosophers have tried. Dan Lopez de Sa (2008), for example, has also defended contextualism by positing a new presupposition: the presupposition that the modal base represents what's common information to parties to the conversation. Alex Silk (forthcoming) has suggested in a somewhat similar vein that a use of an epis- temic modal presupposes that the salient modal base is endorsed in the context. 17 Both of these suggestions are closer to satisfying the second desideratum, that it be something that ordinary agents can have reasonable beliefs about. Unfortunately, though, it's hard to develop these ideas to satisfy the rst re- quirement, that it cover all cases of disagreement. As John MacFarlane (2007) emphasizes, Lopez de Sa's proposal struggles to explain eavesdroppers. Ignorant Ian asserts that the keys might be in the drawer. Someone much better informed hears what Ian said, but has decisive evidence that the keys aren't there. Our Eaves- dropping Evelyn could intelligibly react by saying that Ian is wrong, that the keys can't be there. It's very hard for Lopez de Sa to explain what they're disagreeing about. They're talking about dierent modal bases, so they aren't disagreeing about what's compatible with any one modal base. Moreover, Evelyn knows that they're not party to the same conversation. So it's not plausible that she is disagreeing with Ian about what's common information to parties in the conversation. 18 17 And he intends this claim to dissolve arguments from disagreement, as it clearer in other work, like Silk (2016) and Silk (2017b). 18 And it's very unclear if Silk's account can capture the full range of disagreement that 177 Pluralism, by contrast, does have an adequate account of eavesdroppers. But it has that adequate account because it abandons desideratum (b): that the property that covers all cases of disagreement is something that ordinary agents can have reasonable beliefs about. Eavesdroppers are disagreeing about what is epistemically correct. It's always intelligible for a better-informed eavesdropper to do that, given Pluralism. Someone who asserts an epistemic is accepting a proposition that they can know to be false: the proposition that their own incomplete modal base is epistemically correct. 19 Pluralism is like a three legged stool. Take away any leg, and the whole thing falls apart. First: disagreement about what might be is disagreement about what is more epistemically correct than anything whatsoever. Take away that leg, and you end up with problems in explaining all cases of disagreement. Second: reasonable belief about what might be is not reasonable belief about what is epistemically correct. Take away that leg, and you end up with problems in explaining how easy it is to have reasonable belief about what might be. Third: modals presuppose that the salient modal base is epistemically correct. Take away that third leg, and it's hard to see how the resulting view is coherent. (If you can disagree with a claim about what might be because the modal base isn't correct, why doesn't reasonable belief about what might be require reasonable belief about epistemic correctness?) Together, though, the three legs support a powerful explanation of the full range of data. You don't need to know the presupposition of the complement in order to know the complement. So the speaker's knowledge about what might be doesn't depend on her knowledge about what's epistemically correct. This theory is a nal piece of evidence in favor of the positivist's general ap- proach. It's just the account of the epistemic use that we would expect if the positivist is right. my account captures. It would take a full paper to gure out if it can. By contrast, there isn't any question that my account does capture the full range of disagreement. 19 I introduced epistemic correctness as the mode of presentation that distinguishes atti- tudes about epistemic modal bases from attitudes about circumstantial modal bases. But this way of motivating the mode of presentation does not lead directly to epistemic correct- ness. There are lots of dierent ways of thinking about the relevant mode of presentation. Eavesdroppers are relevant to this choice: it would be good to identify a mode of presen- tation that they could disagree about. That's why I've opted for epistemic correctness, but other choices might be possible. 178 Favors Pluralism? Favors Pluralism? x7.1: Attitude asymmetries Yes! No x3: Local accomodation Yes! No x4: Knows ascriptions No Yes! x5: Anthropological uses Yes! Yes! Chapter 8: Yes! Yes! The epistemic use is evidence for the Pluralist version of my hybrid conjecture, I think, because it patterns just how the Pluralist would expect it to pattern. Now you might wonder why the Pluralist's commitment about the moral use makes any predictions about other uses of modals. I take up this question in the Appendix to this chapter. I conceive of that challenge as a technical challenge about the way that the positivist thinks about context-sensitivity. If you're not worried about that technical challenge, you can skip the Appendix. We're now at the end of Part II. I've explained why the positivist's linguistic claim is highly plausible. It illuminates a range of dierent uses. But there is an important problem about the positivist's account of disagreement: it's unclear if she can really explain the full range. Part III will show that this problem can be solved. 179 8.5 Appendix: The account of context-sensitivity Let's start with the reasons why contextualist approaches like Kratzer's are attrac- tive in the rst place. We use our modal language in strikingly diverse ways - to describe the demands of morality, of prudence, or to describe our evidence. In fact, most (perhaps all) natural languages have terms that are used in this range of ways. We should give an explanation that captures this uniformity, without positing am- biguity in the lexicon. (The data seem too systematic to be explained in that way.) Kratzer's proposal unies all these dierent uses as involving an invariant semantic content. The dierent uses of modal language are explained as dierent ways of saturating that invariant semantic content. The moral use saturates the invariant semantic content with something that represents the demands of morality, and the epistemic use saturates with something that represents our evidence. The task of this Appendix is to show that the there is no lexical ambiguity { that the Pluralist accepts Kratzer's semantics. This task is unfortunately complicated. There are many dierent ways that semantic values might combine with contexts to determine the propositions expressed. And those dierences go very deep; basic technical and interpretative choices divide the dierent approaches. The dierences go deep enough that the Pluralist can't give a concrete description of her proposal without making some technical choices that some theorists reject. So I'll assume a particular framework in describing the Pluralist's proposal. But it should be possible to translate from this particular framework into another framework. 20 The particular framework I assume is due to Scott Soames. The framework involves claims about two dierent topics. It involves a claim about the meanings of sentences independent of context, plus a proposal about the way those context-invariant meanings interact with contexts to determine the propositions expressed. the meaning M of a sentence S is (or imposes) a set of constraints on what normal, literal uses of S (without conversational implicatures that force reinterpretation) assert, or express. When S contains indexicals, or is otherwise semantically incomplete (e.g., lacks any syntactic repre- sentation of some semantically required element), M will not determine a complete proposition by itself, and so must be pragmatically supple- mented in order for a use of S to express a complete thought, or result in a truth-evaluable assertion. (Soames 2010, 163) Possessives (like Mary's cat) are one illustration. On this approach, the meaning of the sentence I saw Mary's cat is not a single proposition. It is rather a set of constraints on what normal, literal uses of that sentence express. The constraints require the context to supply a relation that is predicated of Mary and the cat: the cat that Mary owns, the one that she likes, the one that she has just been talking about. 20 For example, it should be possible to translate this proposal into the framework that David Kaplan (1989) introduced, perhaps as developed by Nathan Salmon (2002). 180 I'll conceive of the meaning of sentences containing possessives as propositional matrices - that is, as incomplete contents that need to be saturated with something further to be complete, truth-evaluable propositions. I will indicate the places that need further saturation with ellipses and long underscores; `... 1 ', `... 2 ', ` 1 ', ` 2 ', and so on. The meaning of I saw Mary's cat combines with something to express a complete proposition. So the meaning is the propositional matrix that I saw the cat that Mary ... 1 . So the meaning constrains that sentence to express one of these propositions: I saw the cat that Mary owns, I saw that cat that Mary likes, I saw the cat that Mary has been talking about, and so on. On this approach, we characterize the meaning of modals by characterizing the way they constrain what is expressed by normal, literal uses of sentences containing them. Kratzer's distinctive semantics takes the context to supply two arguments. One argument is a modal base, and the other is something that determines an ordering. Now the Pluralist has a distinctive picture of those arguments. She thinks that they are associated with presuppositions about the relevant mode of presentation. For each argument, then, she thinks the context needs to do two things. First, it needs to ll in the presupposition with the relevant mode of presentation (as about morality, or some set of laws, or whatever). Second, it needs to saturate with the relevant ordering or modal base. Since Kratzer relies on two arguments, the Pluralist thinks that context will saturate four values. Pluralism JmayK = At-issue: p <st> .p is true at one of the worlds in 1 that 2 ranks highest 21 Not-at-issue: 1 is the modal base that ... 1 Not-at-issue: 2 is the ordering of possible worlds that ... 2 JmustK = At-issue: p <st> .p is true at all worlds in 1 that 2 ranks highest Not-at-issue: 1 is the modal base that ... 1 Not-at-issue: 2 is the ordering of possible worlds that ... 2 The not-at-issue contents are presuppositions. And the at-issue content is the tra- ditional semantic value - for example, the one that embeds under negation. So the Pluralist takes the context to have a two-fold eect. First, the context supplies the mode of presentation that is relevant in the present context. There are 21 I will use `p<st>' as a variable over objects of type<st> - that is, propositions. Even though I assume that propositions have internal structure, they still have that familiar type, since they at least determine a function from worlds to truth values. 181 lots of interesting and dicult questions about the way that that happens, of course; I'm just assuming that it happens in some way. Those mode of presentations settle the kind of presupposition we're talking about - whether it's moral, prudential, or whatever. To illustrate the rst eect of the context, suppose that we're in a context where the modal base is constrained to be what's epistemically correct, and the ordering is constrained to be what's morally correct. Epistemic-cum-moral uses: JmayK = At-issue: p <st> .p is true at one of the worlds in 1 that 2 ranks highest Not-at-issue: 1 is the modal base that is epistemically correct Not-at-issue: 2 is the ordering of possible worlds that is morally correct JmustK = At-issue: p <st> .p is true at all the worlds in 1 that 2 ranks highest Not-at-issue: 1 is the modal base that is epistemically correct Not-at-issue: 2 is the ordering of possible worlds that is morally correct Importantly, though, saturating with the descriptive avors supplied by the context still leaves two argument-places unsaturated. The context has another central eect at this point. Because the two not-at- issue contents are presuppositions, we interpret these utterances as incorporating whatever ordering and modal base the relevant agent accepts to be epistemically correct/ morally correct. 22 The incomplete contents that we've just described are then saturated with those arguments. Epistemic-cum-moral uses: JmayK = At-issue: p <st> .p is true at one the worlds in m 1 that o 1 ranks highest Not-at-issue: m 1 is the modal base that is epistemically correct 22 Who is the relevant agent? Usually the speaker. The exception is embeddings under non-factive attitude verbs. In those cases, the matrix subject is the relevant agent. 182 Not-at-issue: o 1 is the ordering of possible worlds that is morally correct We then get a fully saturated content. And that content feeds into the processes that determine what propositions are asserted by particular utterances. The Pluralist proposal outlined in this paper retains this central advantage of Kratzer's semantics. There is no lexical ambiguity. 183 Part III Part III: Disagreement 184 A brief review of the roadmap Positivist realism is built around a linguistic conjecture. The goal of Part I was to show that that positivist realism has important upshots for metaethics, if it's true. And the goal of Part II was to defend the linguistic conjecture. In the course of defending it, though, we found that the core linguistic conjecture actually has two importantly dierent parts. One part is the hybrid theory of attitude ascriptions, which follows from the conjecture that the realist's property is part of the presupposition of the moral use. The other part is a Pluralist conception of the hybrid theory { a conception where one speaker's moral utterance can communicate propositions about a dierent set of rules than some other person's moral utterance would. Both these pieces are important to the full philosophical advantages that the positivist claims. I think that there is decisive evidence for the rst part: the hybrid theory of attitudes. And there is good evidence for the Pluralist conception. However, there is a powerful argument against the Pluralist conception of the hybrid theory, and in favor of the Monist alternative, or an account of moral dis- course that is even more traditional. The argument purports to show that the Pluralist does not have an adequate account of moral disagreement. Part III sys- tematically disarms this argument. The Pluralist has a fully adequate account of moral disagreement. 185 Chapter 9: Ascriptions of agreement and disagree- ment I'm calling someone who accepts the Pluralist conception of the hybrid theory a positivist. I won't call someone who accepts the Monist conception of the hybrid theory a positivist. Monism has dierent enough philosophical upshots that we should have a distinctive label for someone who accepts both the hybrid theory and the Pluralist conception of it. The positivist occupies an intermediate position in the broader metaethical landscape. Some metaethicists are absolutists, who suppose that there is some fundamental moral property being-the-moral-standard that our moral discourse de- scribes, a property that is independent of and more fundamental than our particular mental states. (Paradigmatic absolutists include Aristotle and Mill, Scanlon, Foot, and Railton.) These absolutists contrast with relativists, like Harman, Finlay, and Street, who suppose that our moral discourse is about something idiosyncratic to particular agents, and that varies between them. Classical versions of this kind of view take moral discourse to be about our individual sentiments, or our individual goals, or something similar. Each approach has certain advantages. The absolutists face metaphysical and epistemic challenges. They need to explain how we know about what we ought to do. They also need to answer certain metaphysical ques- tions, like why the moral supervenes on the non-moral. Relativists have an easier time answering those epistemic and metaphysical questions. But many philosophers take them to face problems in explaining moral disagreement and agreement. If my use of moral language is about something dierent than your use, it seems like we're talking past each other when we disagree. Absolutists, by contrast, have an elegant explanation of moral disagreement. Disagreement is disagreement about what's the moral standard. One helpful way to think about the positivist realist is that she is trying to claim the epistemic advantages that the relativist claims. Crucially, though, the positivist is an absolutist. She does think that there is some fundamental property that our moral language describes. She just denies that our moral knowledge depends on knowledge about that fundamental property. So the positivist is trying to build a halfway house between absolutism and relativism. She is trying to claim the relativist's advantages about our knowledge, while retaining the core absolutist ideas, and in particular the absolutist's account of disagreement. Part I has defended one side of the halfway house, showing that the positivist can claim important 186 epistemic advantages over more traditional absolutists. This Part, Part III, shows that the positivist does retain the absolutist's account of disagreement. Now dierent philosophers might have dierent impressions about the impor- tance of this Part. Some philosophers doubt that that relativists do have an ad- equate account of moral disagreement. This chapter is important if you're one of those philosophers. It is showing that this problem for relativists is not a problem for positivist realism: that you should take positivist realism to pattern with other kinds of realism. But other philosophers are convinced that relativists do have an adequate account of moral disagreement. This chapter is much less interesting for these philosophers. Showing that the positivist retains the absolutist's account of moral disagreement just isn't that important if you think that absolutists and relativists both have an adequate account of moral disagreement. Because the im- portance of this chapter depends on there being an interesting dierence between absolutist accounts of disagreement and relativist's account, I will talk in a way that assumes that there is. 9.1 The positivist's basic advantage A relativist might take moral discourse to describe something idiosyncratic to each of us. This position poses an initial problem about moral disagreement. Jeremy: You may tell that lie. Immanuel: You may not tell that lie. So what Jeremy said is false. Given some kind of relativism, Jeremy and Immanuel would both assert a proposi- tion about something idiosyncratic to them. But then Immanuel's response looks inappropriate. Each of them is talking about something idiosyncratic to them. Compare the case where Jeremy says: I'm hungry. Immanuel can respond: I'm not hungry. But it's very odd for Immanuel to continue with So what Jeremy said is false. In the hunger case, Immanuel's initial claim doesn't support his conclusion that what Jeremy said was false. (They're talking about dierent people!) It seems like the relativist should make the same prediction about the moral case. They're talking about something idiosyncratic to them, so they're talking past each other. And like this one then look like evidence against the relativist. The central question in this chapter is if there is the same kind of evidence against the positivist. The positivist will associate Jeremy and Immanuel's utter- ances with the following pairs of propositions. Jeremy: You may tell that lie. at-issue: o J permits telling that lie not-at-issue: o J is the moral standard Immanuel: You may not tell that lie. 187 at-issue: o I does not permit telling that lie not-at-issue: o I is the moral standard Suppose that Immanuel's propositions are both true. Then one of Jeremy's propo- sitions is false. If o J and o I are the same set of rules, it's false that o J permits telling that lie. If they're dierent sets of rules, it's false that o J is the moral stan- dard. That's the reason why Immanuel's utterance supports the conclusion that what Jeremy said was false. One of Jeremy's two propositions has to be false if Immanuel is right. This observation is the raw material that the positivist can use to explain this discourse. The two agents disagree because they've committed themselves to propo- sitions that can't all be true. It turns out, interestingly, that there are several ways of assembling the raw material. 1 And surprisingly subtle issues come up in evalu- ating the dierent options. Most of this chapter and the next will focus on those issues. But the issues we'll explore are dierent in kind from the issues that aict relativists. The positivist already has all the raw material necessary to explain the disagreement between Immanuel and Jeremy. The relativist does not. She has to supplement her basic idea with further claims about the attitudes expressed, or something along those lines. The positivist is thus in a much better position. The interesting questions for her are about the way natural languages implement the raw material she provides, not about whether she has the raw material to explain the disagreement. In fact, the positivist will always have the raw materials to explain disagreement. The introduction suggested that traditional absolutists have a powerful explanation of disagreement. Moral disagreement always involves disagreement about the same property, being-the-moral-standard. Jeremy asserted that the moral standard allows telling this lie, and Immanuel asserts that it doesn't. That's the reason they're not talking past each other { they're disagreeing about the extension of the very same property. The positivist has a license for optimism she'll always have the raw materials to explain disagreement. The license for optimism is that the Absolutist proposi- tions are obvious a priori consequences of the positivist's propositions. Take some particular moral judgment - that you ought to keep that promise. For the Abso- lutist, that judgment is the judgment that the moral standard requires keeping that promise. For the positivist, by contrast, that judgment is the judgment that o 1 requires keeping that promise, and the judgment that o 1 is the moral standard. We can generalize. License for Optimism: The Absolutist's proposition is an obvious a priori consequence of the positivist's propositions. (The propositions that S expresses if the other Absolutists are right is an obvious a priori consequences of the propositions that S expresses if the positivist is right.) 1 Does what Jeremy said sometimes refer to a not-at-issue proposition? Or does it refer to a lambda abstract of the at-issue proposition? 188 This license for optimism guarantees that the positivist has the raw materials to explain disagreement. The right propositions are always laying around. But there are important limits on this license for optimism, because the license itself doesn't guarantee that the materials come together in a plausible way. There are two challenges to this license for optimism. The rst challenge is whether there are plausible linguistic means for us to disagree or agree about the full range of propositions that the positivist associates with moral utterances. Embeddings under attitude verbs are one important example. Even if the positivist has an account of disagreement in discourse, she might not have an adequate account of sentences like Jeremy and Immanuel [disagree/agree] that... The second challenge comes from the positivist's appeal to acceptance. The positivist takes basic moral judgments to consist in a pair of attitudes, including accepting a proposition as the moral standard. For traditional absolutists, by con- trast, basic moral judgment consists in only one attitude; belief is nothing more than belief in the absolutist proposition. And it's possible that the attitude puts the traditional absolutist's account of disagreement o-limits. After all, some atti- tudes don't support disagreement. You can wonder whether p without disagreeing with someone who wonders whether:p. Other attitudes, by contrast, do support disagreement. If you believe p, you disagree with someone who believes:p. That is, belief is inconsistency-transmitting, while wondering is not. 2 The second challenge is whether acceptance to be inconsistency-transmitting in the right range of cases, in order to explain disagreement. This chapter and the next take up these two challenges in order. As a result, I will just assume in this chapter that acceptance is inconsistency-transmitting in the right range of cases, and wait to the next chapter to defend that assumption. I do that because the second challenge turns out to be downstream from the rst chal- lenge. That is, the tools introduced in this chapter are also important in answering the second challenge. So we need to work through the rst challenge before we can turn to the second. 9.2 Ascriptions of disagreement and agreement Start with an ascription of disagreement. Imagine that Immanuel and Jeremy have very dierent moral codes for lying. Immanuel thinks that it's always wrong, and Jeremy thinks that it's permissible as long as the benets outweigh the harms. (0) is then true of a lie whose benets outweigh the harms. (0) Immanuel and Jeremy disagree about whether this lie is permissible. Can the positivist explain why (0) is true? It doesn't seem like she can. Her basic commitments force her to associate (0) with one of the following logical forms. 2 This vocabulary is from Mark Schroeder (2008a). 189 (0a) Immanuel and Jeremy (disagree about whether o I permits this lie, and accept that o I is the moral standard) (0b) Immanuel and Jeremy (disagree about whether o J permits this lie, and accept that o J is the moral standard) (0) is problematic for the positivist, because neither (0a) nor (0b) is true. Jeremy doesn't accept Immanuel's set of rules o I as the moral standard, and Immanuel doesn't accept Jeremy's set of rules o J as the moral standard. This is a deep problem. Now there is a glib suggestion about this problem that we should set aside. The glib suggestion is that the dis in disagrees to somehow contribute a negation operator that scopes wide. (0) Immanuel and Jeremy disagree about whether this lie is permissible. (0c) Immanuel and Jeremy do not (agree that o I permits this lie, and accept that o I is the moral standard) (0d) Immanuel and Jeremy do not (agree thato J permits this lie, and accept that o J is the moral standard) On this suggestion, (0c) and (0d) would both be true, because there isn't any set of rules that they both accept to be the moral standard. Now this suggestion should look somewhat ad hoc. It's not what you would expect if you mechanically applied the basic positivist's idea. Moreover, this suggestion won't work in full generality. The positivist has the same kind of problem about ascriptions of agreement. And the same trick won't help with this problem. Take (1). (1) Immanuel and Jeremy agree that you can't torture kittens for fun and prot. Imagine that Immanuel and Jeremy agree, but for dierent reasons. Jeremy thinks that it's wrong because it's an in iction of pain without corresponding benet. Immanuel thinks that it's wrong because the torture twists the rational soul of the torturer. The positivist models these judgments with dierent sets of rules. Jeremy accepts a set of rules o J that encodes the harm as counting against the torture but doesn't encode the twisting of the rational soul as mattering. Immanuel accepts a set of rules o I that encodes the signicance of the twisting of the rational soul but not the signicance of the harm. Given these two sets of rules, it looks like there only two propositions that (1) might communicate. (2a) Immanuel and Jeremy both (agree that o I forbids torturing kittens for fun and prot, and accept that o I is the moral standard) (2b) Immanuel and Jeremy both (agree that o J forbids torturing kit- tens for fun and prot, and accept that o J is the moral standard) 190 And the same problem comes up. Neither of these propositions are true. Jeremy doesn't accept Immanuel's rules as the moral standard, and Immanuel doesn't ac- cept Jeremy's rules as the moral standard. Ascriptions of agreement show just how deep the problem really goes. So the bulk of this chapter will focus on ascriptions of agreement, and show that the positivist does have an elegant and principled account of those ascriptions. The chapter will close with a discussion of disagreement, and show that the account of agreement smoothly supports an account of disagreement, as well. Before turning to my own account, though, I want to explain why I'll ignore another possible account of this phenomenon. Earlier chapters have suggested that moral utterances can express propositions about partial sets of rules. For example, that's how I modeled agents with inconsistent beliefs. You might wonder about using the same idea here. Maybe the positivist should hold that (2) is true because Jeremy and Immanuel both agree about an extremely simple set of rules: the set of rules that is partial except in forbidding torturing kittens for fun and prot. This idea doesn't help the positivist. It collapses the dierence between moral knowledge and true moral belief. If ascriptions of agreement can express proposi- tions about maximally partial sets of rules, ascriptions of knowledge should as well. And it seems trivial to know about that sort of rule; it's very natural to think that you know about it as soon as you have true beliefs about it. The positivist does not and should not accept this account of agreement. 9.3 A companion in innocence Fortunately, though, the positivist has a companion in innocence, a companion in innocence that suggests that thex2 problem can be solved. x2 made several tacit assumptions in developing those problems. It's possible to show that those assumptions can't all be right. Consider a construction like (3). (3) Bill and Jane are sad that a kitten killed that mouse. They hope that it didn't torture the mouse. Constructions like (3) are sometimes called Geach discourses, after Peter Geach (1967), who drew contemporary attention to them. This section considers the dierent possible accounts of Geach discourses, and shows that each account is incompatible with some tacit assumption from the last section. Then it shows how each account of Geach discourses crisply solves the problem that the positivist encounters. (3) is very puzzling. It's not just that Bill and Jane agree about the existential closures of the complements. They denitely are sad that there's some mouse-killing kitten. And they denitely hope that there's no kitten that tortured the mouse. But that's not all. Their thoughts are all linked together: they're thinking about the very same kitten in each case. Their hope wouldn't be satised if there were a pacist kitten that refrained from both killing and torture. So their hope is more than just 191 that [9x: kitten (x)] (x didn't torture the mouse). In other words: attitudes towards the existential closures of the individual complements aren't enough to explain this discourse. Our central question about (3) is: what links the two attitudes together, so that they are about the very same kitten? In tackling this question, I'll focus only on articulating a plausible set of truth-conditions for (3). I won't fuss about how or whether those truth-conditions can be derived compositionally, or what mix of syntactic, semantic, and pragmatic processes associate (3) with those truth- conditions. The rst option for explaining (3) takes the hope to be about a proposition enriched with descriptive content from the rst complement. That enrichment is what links the two thoughts together. (3e) Bill and Jane hope that [9x: kitten (x) and x killed that mouse] (x didn't torture that kitten rst). Unfortunately, though, this suggestion looks implausible. It seems to require that Bill and Jane both hope that there's a kitten that killed that mouse. And the whole point of the rst sentence in (3) is that they denitely don't have that hope! Now there might be some way to solve this problem. I'll consider some options later. But those options are quite complex. So I want to set this approach aside, for a few pages, and pursue two other avenues before returning back to it. 9.3.1 Liberalized singular thought The second approach to (3) takes an existential quantier to scope wide. (3) com- municates the same proposition as (4) (4) [9x (Bill and Jane are sad that x killed that mouse, and they hope that x didn't torture the mouse rst)] This approach takes on some striking and controversial commitments. To see the basic reason why, assume that Miley or Mikey are the only cats at the world. Then (4) is true only if (5) is true. (5) [Bill and Jane are sad that Mikey killed that mouse, and they hope that Mikey didn't torture the mouse rst] OR [they are sad that Miley killed that mouse, and they hope that Miley didn't torture the mouse rst] And these truth-conditions for (3) are controversial. They require Bill and Jane to have singular thoughts about Miley, or to have singular thoughts about Mikey. These truth-conditions are controversial, because they are only viable if it's possible to think singularly about some object in virtue of purely general grounds. After all, (3) could be true even if the agents had purely general grounds for their attitudes. Imagine that they both think that only kittens kill mice. They've heard 192 that a mouse died. They infer (with sadness) that a kitten killed that mouse. And they have the empathetic hope that the kitten gave the mouse a quiet death. Moreover, many philosophers deny that purely general grounds are sucient for singular thought. Robin Jeshion is one example. She is quite permissive about the grounds for singular thought. For example, she doesn't take singular thought to require any causal connection with the object of thought. She holds that an agent can think singularly about an individual only when that agent has initiated a mental le on that individual. And she proposes that such initiation is possible only when that individual is signicant to that agent: Signicance Condition: a mental le is initiated on an individual only if that individual is signicant to the agent with respect to her plans, projects, aective states, motivations. (Jeshion 2010, 136) This suggestion is more permissive than most. But even it is not enough to support singular truth-conditions for (3). Jeshion still holds that some agents can't think singularly about some objects: whenever the Signicance Condition isn't satised. Imagine that Bill and Jane have only general grounds for their attitudes in (3). The object that satises those general grounds { Mikey, say { might not be signicant for them, in the way that matters for Jeshion. So singular truth-conditions will be implausible for (3), if Jeshion is right. This point generalizes systematically. Suppose that there is some substantive constraint on singular thought, that some but not all agents satisfy. Suppose fur- ther that Bill and Jane don't satisfy that substantive constraint towards whatever satises their general grounds for thinking about kittens. Then singular truth- conditions will be implausible for (3). Now some views do abjure constraints on singular thought. David Kaplan suggests, for example, that \what allows us to take various propositional attitudes toward singular propositions is not the form of our acquaintance with the object but is rather our ability to manipulate the conceptual apparatus of direct reference" (Kaplan 1989, 536). This sort of liberality about singular thought can make sense of singular truth-conditions for (3). 3 And indeed it's the only way of doing so. Let's turn back to the positivists's problem fromx2. If you explain (3) by abjur- ing any constraints on singular thought, the positivist will use your commitments to solve the problem fromx2. She would then take the agreement in (1) to be agreement about whatever is in fact the correct moral standard. Call that set of rules o c . (1) Immanuel and Jeremy agree that you can't torture kittens for fun and prot. (2c) Immanuel and Jeremy both (agree that o c forbids torturing kittens for fun and prot, and accept that o c is the moral standard) 3 David Kaplan (1989) and David Manley and John Hawthorne (2012) all favor this kind of liberalism about singular thought. For critical discussion, see (Jeshion 2010, 125-9). 193 (1) was initially problematic, because there was no one set of rules that they both accept to be the moral standard. They don't accept that Immanuel's set of rules o I is the moral standard, and neither do they accept that Jeremy's set of rules o J is the moral standard. On the present idea, though, there is one set of rules that they both accept to be the moral standard. It's the one that they're thinking singularly about. Now there is an important tension between this idea and the positivist's gen- eral ambitions. The positivist takes Immanuel and Jeremy to be thinking of two particular sets of rules, which are dierent. But I just suggested that (1) is true because they're thinking singularly about the same set of rules, o c . So the present suggestion is taking them to think about the same set of rules, in virtue of the fact that they are thinking about dierent sets of rules. And this suggestion is very odd. In general, though, it is very odd to explain Geach discourses by appeal to singular thought. The oddity comes from this strategy for explaining Geach dis- courses, not the positivist's distinctive commitments. To see this point, consider another kind of example. Suppose that Bill is sad that Mikey killed that mouse and hopes that he didn't torture it. Jane, by contrast, thinks that Chester killed that mouse. She is sad that Chester killed it, and hopes that he didn't torture it. We can still use (3) to describe this their attitudes. (3) Bill and Jane are sad that a kitten killed that mouse. And they hope that it didn't torture it rst. In this case, the singular explanation of Geach discourses is very odd. If (3) at- tributes singular thought about one worldly kitten, Bill and Jane are thinking about the same worldly kitten in virtue of thinking about dierent worldly kittens. So this example suggests a quite general point. Individuals can share linked thoughts in virtue of thinking about particular objects, even if the particular objects are dierent. So whatever explanation we give of linked thoughts had better explain how that's possible. I myself dislike this approach to Geach discourses. I'm skeptical that agents think singularly about one world kitten in virtue of thinking singularly about dif- ferently worldly kittens. But if you can stomach that consequence, the positivist has a crisp account of her problem. 9.3.2 Unanchored discourse referents This subsection introduces the third approach to Geach discourses, which is the approach that I take to be most promising. This approach posits a new way that our thought and language can be about particular individuals. Grant that we can think descriptively about particular indi- viduals, as when we think or talk about the current president of the US. Grant that we can also think singularly about particular individuals, and insist that there are some constraints on singular thought. (General grounds aren't enough to think sin- gularly about some individual.) The third approach introduces a sui generis third 194 way for thinking about objects. I will call this third way thought with unanchored discourse referents. Thought with unanchored discourse referents is like singular thought, in that it does not require the thinker to think of the object as falling under any particular description. But it's unlike singular thought, in that there need be no unique worldly individual that you're thinking about. I'll introduce unanchored discourse referents by introducing their truth-conditions. I'll start with an extensional example, and build up to an example about attitude reports. Let subscripted `i's be terms for unanchored discourse referents. (3a) Something killed that mouse. It tortured the mouse rst. i 1 killed that mouse. i 1 tortured the mouse rst. The discourse in (3a) is true at a world-state w i w contains some object that veries the constraints on the unanchored discourse referent. In other words: it's true i if the world contains some object that both killed that mouse, and also tortured the mouse rst. That's the account of truth at a maximally specic world- state. Importantly, though, the denition generalizes smoothly to truth at non- maximal situations. Quantify over all the ways of making the non-maximal situation maximal, and make sure that those maximal ways all verify the constraints on the unanchored discourse referent. So the discourse in (3a) is true at a non-maximal situation s i there is a maximal way of extending s that veries the constraints on the discourse referents, and in fact all maximal ways do. Unanchored discourse referents model the accumulation of information about objects. We can be accumulating information about several such objects at once, as in (3b). (3b) A cat killed a mouse. It tortured the mouse rst. i 1 is a cat. i 2 is a mouse i 1 killed i 2 i 1 tortured i 2 rst. And the discourse in (3b) is true if there is some pair of objects <x, y> where x is a cat, y a mouse, x killed y, and x tortured y rst. And we can draw a range of inference from these truth-conditions { that (3b) entails that there is a cat, that there is a torturer, and so on. And we can see the conditions where the agent can appropriately infer (3b): if they think that there is a mouse-torturing cat that also killed its victim. In general, the functional role of unanchored discourse referents is as denoting concepts that play a similar role to Fregean individual concepts. They allow in- dividuals to think and talk about individuals without being in a position to think singularly about those individuals. And there is a crucial dierence between these 195 unanchored discourse referents and Russellian denite descriptions. Russellian def- inites require uniqueness: there has to be some one individual that satises the relevant descriptive information. There is no similar requirement on unanchored discourse referents. Discourses containing them are true if the discourse is true of every individual that satises the relevant descriptive information. It doesn't matter if there is a unique satiser, or if several individuals satisfy it. The truth-conditions for attitude ascriptions builds from the account of truth at non-maximal situations. (4) Jane believes that something killed that mouse. She also believes that it tortured the mouse rst. Jane believes that i 1 killed that mouse. She also believes that i 1 tortured the mouse rst. i 1 killed that mouse i 1 tortured the mouse rst. (4) depends on what's compatible with Jane's beliefs. It's true i there is some maximal extension of her beliefs that veries the items in the bullet-pointed list, and indeed every maximal extension does. To nd the maximal extensions, start with the propositions that Jane believes. A maximal extension is the result of adding propositions to what Jane already believes until you get a complete, consistent description of the world. A maximal extension veries the items in the bullet- pointed list if the objectual existential closure of that list is true at that maximal extension. Maybe one maximal extension of Jane's belief's includes the propositions: <Mikey killed that mouse> and <Mikey tortured that mouse rst>. That maximal exten- sion veries the bullet-pointed list, because the objectual existential closure of that list is true at that maximal extension. Another maximal extension might include the propositions that Fido did both things. That maximal extension also veries the bullet-pointed list. Importantly, though, Jane need not be thinking about Mikey or Fido in order for these maximal extensions to be relevant. The maximal extensions are what are left open by her beliefs. In some cases, then, it's precisely because she's never thought about Mikey or Fido that they count as among the maximal extensions of her beliefs. If she's never thought about them, she's never formed any beliefs about them that would eliminate them as candidates for the killer and torturer. (This fact about unanchored discourse referents is that basic reason why unanchored discourse referents are unlike unconstrained singular thought.) (4) would be false if there is some maximal extension of Jane's beliefs where dierent things did the killing and torturing. For example, it would be false if there's a maximal extension of her beliefs that includes the propositions: <only Mikey killed that mouse> and<only Fido tortured that mouse rst>. The objectual existential closure of the bullet-pointed list is not true at that extension. But that's exactly the prediction we want to make. If Jane's beliefs allow that dierent things did the killing and the torturing, then she doesn't have the beliefs that (4) describes her 196 as having. In order to have those beliefs, you've got to think that the same thing did the killing and the torturing. Universal quantication over maximal extension captures exactly the truth-conditions we want. 4 That's the basic idea. Moreover, this basic idea generalizes smoothly to ascriptions of sadness and hope, like in (3). (3) Bill is sad that a kitten killed that mouse. And he hopes that it didn't torture it rst. (3) is true if there is some maximal extension of the propositions that Bill hopes in and the propositions that Bill is sad about where there is some individual that killed the mouse and didn't torture it rst. However, subtle issues come up in generalizing this proposal to ascriptions of agreement. (5a) Bill and Jane agree that something killed that mouse. They also agree that it tortured the mouse rst. Bill and Jane agree that i 1 killed that mouse. They also agree that i 1 tortured the mouse rst. i 1 killed that mouse i 1 tortured the mouse rst I claim that the correct semantics for agrees checks the maximal extensions of each individual's beliefs, rather than the maximal extensions of the group's beliefs. (5a) is true i every maximal extension of Bill's beliefs veries the bullet-pointed list and every maximal extension of Jane's beliefs veries the bullet-pointed list. We have to appeal to maximal extension of each individual's beliefs. We get the wrong result if we rst pool everything they believe, and then consider maximal extensions of that pooled information. Suppose that Bill thinks that the killer has spots, while Jane thinks that the killer does not. Then there is no object x such that<x killed the mouse> is in the maximal extension of the group's beliefs. Either x has spots or it doesn't. (Remember that we're considering maximal extensions, which specify everything about an object, including whether it has spots.) If x does have spots, <x killed the mouse> is not in any maximal extension of the group's beliefs, because Jane thinks that the killer does not have spots. And if x does not have spots, <x killed the mouse> is not in any maximal extension of the group's beliefs, because Bill thinks that the killer does have spots. So the group-centered approach predicts that (5a) is false if Bill and Jane disagree about whether the killer had spots. This prediction is empirically mistaken. If it were correct, the discourse in (5b) would be inconsistent. 4 Irene Heim (1982) and Hans Kamp (1981) both in uentially developed these ideas, around the same time. The distinction between anchored and unanchored discourse refer- ents is due to Kamp (ms). 197 (5c) Bill and Jane agree that something killed that mouse. They also agree that it tortured the mouse rst. But they disagree about whether it had spots. But this discourse is consistent. So the group-centered approach is unpromising. Let's go back to the to the problem fromx2, which motivated the considera- tion of Geach discourses. Unanchored discourse referents solve that problem. The problem was to give a plausible account of (1). (1) Immanuel and Jeremy agree that you can't torture kittens for fun and prot. (1) was initially problematic, because there was no one set of rules that they both accept to be the moral standard. They don't accept that Immanuel's set of rules o I is the moral standard, and neither do they accept that Jeremy's set of rules o J is the moral standard. If we posit unanchored discourse referents, though, we can avoid this problem. (2d) Immanuel and Jeremy both (agree that i 1 forbids torturing kittens for fun and prot, and accept that i 1 is the moral standard) And (2d) is true if there is some maximal extension of Immanuel's attitudes that veries all the constraints on i 1 and there is some maximal extension of Jeremy's attitudes that veries all the constraints on i 1 . And such maximal extensions indeed exist! The maximal extension that matters for Immanuel is about his set of rules o I : it includes the propositions that o I forbids torturing kittens for fun and prot, and that o I is the moral standard. And the maximal extension that matters for Jeremy is about his set of rules o J : it includes the propositions that o J forbids torturing kittens for fun and prot, and that o J is the moral standard. Unanchored discourse referents crisply solve the problems for the positivist. If they are the right way to capture Geach discourses, the positivist is home free. 9.4 The positivist's problem is a problem about intensional anaphora The last section showed how two natural accounts of Geach discourses also give the positivist the resources to solve her problem. In fact, though, something stronger is true. Any adequate account of Geach discourses will also give the positivist the resources to solve her problem. Go back to the core conjecture that drives positivist realism: that uses of modals carry presuppositions. The positivist's distinctive moral epistemology follows from this basic conjecture, because you don't need to V presuppositions of the comple- ment in order to V the complement. Take (6), for example. (6) Jim hopes that Tom has stopped beating his dog. 198 (6) can be true even if Jim doesn't hope that Tom used to beat his dog. It can be true if Jim accepts that Tom used to beat his dog and hopes that Tom isn't beating it anymore. Given the positivist's conjecture that modals carry presuppositions, moral knowledge would be like this. It would only require acceptance that the salient set of rules is the moral standard, plus knowledge about what that set of rules demands. In developing the positivist's idea, I have drawn on a distinctive explanation of this fact about presupposition. I have assumed a multi-dimensional semantics for presupposition triggers, where presupposition triggers like stop are associated with two distinct propositions. For example, Tom has stopped beating his dog is associ- ated with the proposition that Tom used to beat his dog and with the proposition that Tom isn't beating his dog now. The former is interpreted as not-at-issue, and the latter interpreted as at-issue. I explain the distinctive behavior of presupposi- tions under attitude reports by claiming that only the at-issue proposition embeds. pA V-es that Sq semantically expresses that A V-es S's at-issue content. (6), for example, semantically expresses that Jim hopes that Tom isn't beating his dog. Further pragmatic processes usually enrich the semantic content with the further content that A accepts S's not-at-issue content. Given a multi-dimensional semantics for presupposition triggers, many attitude ascriptions that embed presupposition triggers are themselves Geach discourses. Go back to the initial example of a Geach discourse. (3) Bill is sad that a kitten killed that mouse. And he hopes that it didn't torture it rst. (3) ascribes dierent attitudes to two dierent propositions, but the two propositions are \linked", in that they're both about the same individuals. (7) is an example of the same phenomenon, if the multi-dimensional semantics is right. (7) Bill and Jane agree that a kitten stopped torturing a mouse. There are two propositions, p 1 and p 2 where an assertive utterance of (7) communicates that: Bill and Jane agree about p 1 Bill and Jane both accept p 2 Moreover, p 1 and p 2 need to be linked, to be about the same individual. If they're not linked, the truth-conditions for (7) would be too weak. One way for the propositions to be unlinked is if they're just propositions with bare existential quantication. Then (7) would communicate that Bill and Jane agree that [9x] (x isn't torturing a mouse now) and accept that [9x] (x wasn't torturing a mouse before now). But suppose that they think that some kittens are always torturing a mouse, and also that some kittens never tortured a mouse. Then they accept that there's some kitten that used to torture a mouse, and a kitten that isn't torturing a mouse now. These truth-conditions are then satised. But (7) isn't 199 assertable in that case. It requires that the very same kitten that used to torture that mouse not be torturing it anymore. So we should expect that any adequate account of Geach discourses will solve the positivist's problem. 5 These ascriptions of agreement are themselves Geach discourses! Or, more carefully, they're Geach discourses, given a multi-dimensional semantics for presupposition triggers. 9.5 The positivist's problem in a unidimensional semantics The last section argued for a conditional: if the right semantics for presupposition triggers is multi-dimensional, the positivist's problem has a solution. Given a multi- dimensional semantics, many attitude ascriptions are themselves Geach discourses. In particular, the attitude ascriptions that create problems for the positivist are Geach discourses. And since they are Geach discourses, adequate accounts of Geach discourses will solve the positivist's problems. This section shows that the positivist's problem will also have a solution, given a unidimensional semantics for presupposition triggers. Recall a Central Empirical Fact about presupposition. (Central Empirical Fact) You don't need to V presuppositions of the complement in order to V the complement. If the explanation of this Fact isn't a multi-dimensional semantics for presupposition triggers, it's a unidimensional, conjunctive semantics. Tom stopped beating his dog expresses the proposition that Tom used to beat his dog and Tom isn't beating his dog anymore. To explain the Central Empirical Fact, we have to suppose that the presupposi- tional part of the content is somehow interpreted as presupposed. I'll follow David Beaver in writing that marking with a `'. Tom stopped beating his dog would then somehow come to be associated with the proposition that (Tom used to beat his dog) and Tom isn't beating his dog anymore. In that proposition, the conjunct about Tom's past activity is somehow interpreted as presupposed. We can then try to explain the Central Fact by holding that attitude ascriptions ignore material under the material. Because attitude ascriptions ignore that material, you can hope that Tom stopped beating his dog as long as you hope that he's not beating his dog now. Now a full development of this theory needs to explain why attitude ascriptions \ignore" material under the operator. 6 5 It was noted quite early on that multi-dimensional semantics for presupposition trig- gers seems to make mistaken predictions about examples like 7; Lauri Karttunen and Stanley Peters (1979) noted the problem in their development of their multi-dimensional approach. For more contemporary discussion, see Paul Dekker (2008), Robert van Rooij (2005, 2010), David Oshima (2006), and Yasutada Sudo (2012). 6 I'm skeptical that such an explanation is possible. That's one of the reasons I've been working with a multi-dimensional semantics rather than this option. 200 The operator solves the problems for the positivist. We want to explain why (1) can be true even if Jeremy and Immanuel accept dierent sets of rules to be the moral standard. (1) Immanuel and Jeremy agree that you can't torture kittens for fun and prot. (1) can then communicate an existential content. (1lf) Immanuel and Jeremy agree that [9x: x is the moral standard] (x forbids torturing kittens for fun and prot) These occurrences of `x' are bound inside the attitude ascription. So they do not require Immanuel or Jeremy to think singularly about any particular set of rules. Crucially, though, the material under the operator is somehow backgrounded. And that's just what the positivist needs in order for her account to be plausible. Strictly speaking, then, we don't need to consider Geach discourses in the dis- cussion of unidimensional semantics for presupposition triggers. It's enough to note that the proponent of a unidimensional semantics has to acknowledge tools that allow the positivist to solve her problem. But the discussion of unidimensional se- mantics does give us a better perspective on an earlier point about Geach discourses. My discussion of Geach discourses started by considering a very conservative expla- nation of the data. (3) was my example of a Geach discourse. (3) Bill and Jane are sad that a kitten killed that mouse. They hope that it didn't torture the mouse rst. A conservative explanation of this data takes the second sentence communicates the same proposition as (3e). (3e) Bill and Jane hope that [9x: kitten (x) and x killed that mouse] (x didn't torture that kitten rst). I earlier dismissed this suggestion as implausible. It seems to require that Bill and Jane both hope that there's a kitten that killed that mouse. But this suggestion does become plausible if we supplement it with the op- erator described in this section. (3) communicates that the agents hope that [9x: kitten (x) and x killed that mouse] (x didn't torture that kitten rst). And hope in that content doesn't require hope in the material under the operator. To evaluate this idea, though, we need a better understanding of what the operator is doing. John Hawthorne and David Manley (2012) have outlined an idea that may answer this question, at least as it arises in the present context. They take sentences like (3) to involve a special kind of cover domain restriction. I can say I parked the car by the house, even though the world contains more than one car and more than one house. In interpreting that utterance, we somehow restrict our attention so that we're only not attending to the other cars or the other houses. Maybe we're restricting attention to the car that I own, and the house that I see. 201 Hawthorne and Manley propose that, in cases like (3), the speaker is \pre- supposing that there is exactly one object that her use of the indenite is about" (Hawthorne and Manley 2012, 126). The speaker is presupposing, in other words, that covert domain restriction singles out one object. How, then, should we inter- pret \[9x: kitten (x) and x killed that mouse]..."? Well, the material under the operator signals that the same covert domain restriction is still in place, and we're still in a domain where only one kitten is available to talk about. Crucially, then, you can hope that [9x: kitten (x) and x killed that mouse] ( x didn't torture that kitten rst), without hoping that a kitten killed that mouse. So in order to defend this highly conservative account of Geach discourses, we have to introduce a backgrounding operator. And that backgrounding operator is enough to solve the positivist's problems about agreement. Even in this case, then, Geach discourses motivate the introduction of tools that allow the positivist to defend her proposal. 9.6 A master argument We can now introduce a master argument that the positivist will have an adequate account of ascriptions of agreement. The argument proceeds by cases. (1) Immanuel and Jeremy agree that you can't torture kit- tens for fun and prot. Master Argument Is the correct semantics for presupposition triggers multi-dimensional, or unidimensional? If it's multi-dimensional, then ascriptions of agreement or dis- agreement like (1) are themselves Geach discourses. { The correct account of Geach discourses will capture exam- ples like (1). If the correct semantics is unidimensional, then presupposed ma- terial has to be interpreted under a backgrounding operator. { Then the property being-the-moral-standard will be inter- preted under the backgrounding operator, and the posi- tivist can take (1) to involve agreement about an existential content about that property. So the positivist will have an adequate account of ascriptions of agreement, like (1). Now this paper opened by considering ascriptions of disagreement. (0) Immanuel and Jeremy disagree about whether this lie is permissible. 202 But I quickly switched to ascriptions of agreement, like (1). (1) Immanuel and Jeremy agree that you can't torture kittens for fun and prot. I claimed that ascriptions of agreement end up raising more dicult questions than ascriptions of disagreement. And I promised that the account of agreement will generalize smoothly to explain disagreement. I can now make good on this promise. In making good on this promise, I will again proceed by cases. Suppose rst that the right semantics for presupposition triggers is unidimensional. Then utterances of (0) would communicate disagreement about an existential proposition about the moral standard. (a) Immanuel and Jeremy disagree about whether [9x: x is the moral standard] (x permits this lie) This logical form does capture their disagreement. They disagree about whether there is something that is both the moral standard and also permits telling this lie. Now turn to a multi-dimensional semantics for presupposition triggers. Then ascriptions of disagreement like (0) are Geach discourses, just like (5). (5) Bill and Jane agree that a kitten killed that mouse. They disagree about whether it tortured the mouse rst. (5lf) Bill and Jane agree thati 1 killedi 2 ^i 1 is a kitten^i 2 is a mouse. They disagree about whether i 1 tortured i 2 rst. And they disagree i one of them thinks that i 1 tortured i 2 rst, and one of them thinks thati 1 didn't torturei 2 rst. In other words: one of them needs to satisfy the constraints in Column 1, and the other needs to satisfy the constraints in Column 2. Column 1 i 1 killed i 2 i 1 is a kitten i 2 is a mouse i 1 did torture i 2 Column 2 i 1 killed i 2 i 1 is a kitten i 2 is a mouse i 1 did not torture i 2 (5) is true i every maximal extension of one of their beliefs satises Column 1, and every maximal extension of the other's beliefs satises Column 2. The same point will hold for (0). It would communicate (b). (0) Immanuel and Jeremy disagree about whether this lie is permissible. (b) Immanuel and Jeremy disagree about whether i 1 permits this lie and accept that i 1 is the moral standard. 203 Column 1 i 1 permits this lie. i 1 is the moral standard Column 2 i 1 does not permit this lie i 1 is the moral standard (0) is true i every maximal extension of one of their beliefs satises Column 1, and every maximal extension of the other's beliefs satises Column 2. So (0) will be true. Immanuel and Jeremy accept dierent sets of rules as the moral standard, and their sets of rules disagree about this lie. Immanuel accepts that his own set of rules o I is the moral standard. And that set of rules does not permit the lie. So every maximal extension of Immanuel's attitudes will satisfy Column 2. Jeremy accepts his set of rules o J as the moral standard, and that set of rules does permit the lie. So every maximal extension of Jeremy's attitudes will satisfy Column 1. The account of agreement does generalize smoothly to disagreement. This paper opened by noting a license for optimism that the positivist will have a satisfying account of agreement and disagreement. It noted that traditional absolutists have a powerful explanation of disagreement. Moral disagreement always involves disagreement about the same property, being-the-moral-standard. (0) is true if Jeremy asserted that the moral standard allows telling this lie, and Immanuel asserts that it doesn't. That's the reason they're not talking past each other { they're disagreeing about the extension of the very same property. This observation gives the positivist a license for optimism she'll always have the raw materials to explain disagreement. The license for optimism is that the Absolutist propositions are obvious a priori consequences of the positivist's propositions. Take some particular moral judgment - that you ought to keep that promise. For the Absolutist, that judgment is the judgment that the moral standard requires keeping that promise. For the positivist, by contrast, that judgment is the judgment that o 1 requires keeping that promise, and the judgment that o 1 is the moral standard. We can generalize. License for Optimism: The Absolutist's proposition is an obvious a priori consequence of the positivist's propositions. (The propositions that S expresses if the other Absolutists are right is an obvious a priori consequences of the propositions that S expresses if the positivist is right.) This license for optimism guarantees that the positivist has the raw materials to explain disagreement. The right propositions are always laying around. This chapter has vindicated this license for optimism, against one way that it might fail. It has shown that this license for optimism generalizes to attitude ascriptions. The next chapters will take up further challenges to this license for optimism. 204 Chapter 10: A theory of acceptance We're currently in Part III of the dissertation. And the goal of this Part is to explore whether the positivist has an adequate account of moral disagreement. I am argue that the positivist does { that her account of moral disagreement is adequate inasmuch as traditional moral realists have an adequate account of disagreement. The last chapter defended this claim by considering ascriptions of agreement and disagreement, like (0). (0) Immanuel and Jeremy disagree about whether this lie is permissible. I showed how the positivist does have an adequate account of these constructions. She just needs to draw on the best account of intensional anaphora. And any viable account of intensional anaphora will solve the positivist's problem. As emphasized before, the positivist has a fully general license for optimism that she has an adequate account of disagreement. License for Optimism: The traditional moral realist's proposition is an obvious a priori consequence of the positivist's propositions. (The propositions that S expresses if the traditional moral realists are right is an obvious a priori consequences of the propositions that S expresses if the positivist is right.) This license for optimism guarantees that the positivist has the raw materials to explain disagreement. The right propositions are always laying around. But we saw in the last chapter that there might be limits on this license for optimism. The license doesn't itself guarantee that the materials come together in a plausible way. The last chapter emphasized two ways that the license for optimism might fail. First, there there might be no plausible linguistic means for us to disagree (or agree) about the full range of propositions that the positivist associates with moral utterances. That was the problem that the last chapter tackled. It argued that there will always be plausible linguistic means to disagree or agree about the full range of propositions. The second way that the license for optimism might fail comes from the attitude of acceptance. Remember that the positivist takes basic moral judgments to consist in a pair of attitudes. Belief, for example, is belief in one proposition plus acceptance of another. For traditional absolutists, by contrast, basic moral judgment consists in only one attitude; belief is nothing more than belief in the absolutist proposition. 205 The positivist's additional attitude might somehow break the license for optimism. Acceptance might not support disagreement in the right range of cases. This chapter takes up this second problem. In tackling it, though, we rst need a more complete picture of what it is to accept a proposition in the sense that the positivist cares about. This chapter opens by developing a more complete picture. Then it shows that this more complete picture does support the positivist's license for optimism. 10.1 Acceptance and disagreement Some attitudes don't support disagreement. You can wonder whether p without disagreeing with someone who wonders whether:p. Other attitudes, by contrast, do support disagreement. If you believe p, you disagree with someone who believes :p. I'll describe this dierence by saying that belief is inconsistency-transmitting, while wondering is not. 1 One important question is: is acceptance uniformly inconsistency-transmitting? If it is, the positivist would be in the same position as the traditional absolutist. Agent A V-es that you may tell that lie. Traditional Absolutist: A V-es that the moral stan- dard permits telling that lie Positivist: A V-es that o 1 permits telling that lie and accepts that o 1 is the moral standard. Agent B V-es that you may not tell that lie. Traditional Absolutist: B V-es that the moral stan- dard forbids your telling that lie. Positivist: B V-es that o 2 forbids your telling that lie and accepts that o 2 is the moral standard. Traditional absolutists predict that A and B disagree if V-ing is inconsistency- transmitting. The two agents V inconsistent propositions { so of course they dis- agree. The positivist makes the same prediction whenever acceptance is inconsistency- transmitting. Either o 2 and o 1 are identical or they're distinct. If they're identical, the two agents V inconsistent propositions about the demands of the same order- ing. If they're distinct, the two agents accept inconsistent propositions about which ordering is the moral standard. So they're guaranteed to disagree whenever V-ing and acceptance are both inconsistency-transmitting. But the positivist can't just stipulate that acceptance is inconsistency-transmitting. The positivist is making a conjecture about moral discourse that she intends to be 1 This vocabulary is from Mark Schroeder (2008a). 206 simple and highly principled. The conjecture is that the realist's fundamental nor- mative property is part of the presupposition of moral discourse. This conjecture leads directly to the positivist's hybrid conception of moral attitudes, where accep- tance of a set of rules as the moral standard plays a central role. It leads directly to that conjecture, because acceptance is the attitude appropriate for the presupposi- tions of an attitude. In other words: the attitude of acceptance plays a important theoretical role, totally independently of the positivist's particular use of it. Since it plays that important theoretical role, the positivist can't just stipulate facts about it. If she wants to hold that that attitude is always inconsistency-transmitting, she has to show that whatever plays that theoretical role has to be inconsistency- transmitting. And it's highly implausible that whatever attitude plays this theoretical role is always inconsistency-transmitting. For example, presupposition triggers are per- fectly appropriate during the course of reasoning for reductio. In the course of proving that there is no largest prime, say, you might use phrases like the largest prime. For example, you might say multiply the largest prime by all the smaller primes, and add one. In uttering that sentence, you are accepting that there is a largest prime. (Denite descriptions presuppose the existence of something satisfy- ing the restrictor.) But in doing that, you don't really disagree with someone who accepts that there isn't a largest prime because they believe that there isn't a largest prime. For one thing, you might reason by reductio when you already believe that there isn't a largest prime. Maybe you're teaching a class, and use this example to introduce reasoning by reductio. It's just not plausible that acceptance is always inconsistency-transmitting. Our overall question in this part of the dissertation is if the positivist is in a worse position than traditional moral realists in explaining moral disagreement. This chapter is focusing on one aspect of that question: whether the positivist's appeal to acceptance puts her in a worse position. We've just seen that this question is harder than the positivist might have hoped. The positivist can't insist that acceptance is always inconsistency-transmitting. But the game's not over yet. It might still be possible for the positivist to retain the traditional realist's account of disagreement. She could retain it if the cases where acceptance fails to transmit inconsistency are irrelevant for the traditional realist's account of disagreement. And that is just what I'll suggest in this chapter. 10.2 The basic account of acceptance We've just seen that there are very general constraints on the positivist's account of disagreement. Her account makes crucial appeal to what agents accept. And there are very general constraints on when acceptance is inconsistency-transmitting. The goal of this section is to introduce one general picture of acceptance, which makes good on those constraints. Later sections will show that this account of acceptance allows the positivist to retain to traditional realist's account of disagreement. I want to start with a picture of acceptance that Robert Stalnaker develops. 207 He glosses acceptance as \a category of propositional attitudes and methodological stances toward a proposition, a category that includes belief, but also some atti- tudes (presumption, assumption, acceptance for the purposes of an argument or an inquiry) that contrast with belief, and with each other. To accept a proposition is to treat it as true for some reason. One ignores, at least temporarily, and perhaps in a limited context, the possibility that it is false" (Stalnaker 2002, 716). He's assuming that there is an important class of attitudes that are all tied together in an interesting way. They all involve treating propositions as true. I'll assume that he's right that there is some such class. I want to highlight one particularly important feature of this class of attitudes. For each attitude A i in that class, rationality requires probabilistic coherence among the propositions that you A i . For example, rationality requires that your beliefs be probabilistically coherent. Rationality also requires that the propositions that you're assuming be probabilistically coherent. And so on. As some evidence that rationality does require probabilistic coherence within these attitudes, note that Dutch book arguments generalize to each member of the category. Start with the classic example of belief. If your beliefs are probabilistically incoherent, then you'll be disposed to take a series of bets that guarantees that you'll take a loss. 2 And you aren't rational if you're disposed to do that. So you're not rational if your beliefs are probabilistically incoherent. This Dutch book argument extends to each attitude A i that is a kind of ac- ceptance. In A i -ing a proposition, you are pursuing some goal. But if your A i -ing is probabilistically incoherent, you're going to be worse at pursuing that goal than you otherwise would be. Take whatever goods constitute that goal. If your A i -ing is probabilistically incoherent, someone can write you a Dutch book that leaves you with less of those goods than you started out with. And the existence of such books suggest that your A i -ing is irrational. We can generalize: probabilistic incoherency within any kind of acceptance is irrational. This generalization makes acceptance just the right sort of thing to explain dis- agreement. If there's a rational prohibition on being probabilistically incoherent in what you A i , it's not surprising if dierences in what we A i can support disagree- ment. The generalization about the inter-personal case looks very natural when you accept the generalization about the intra-personal case. There is, however, an important problem with this natural thought. In the last section, we saw that acceptance sometimes fails to be inconsistency-transmitting. How should we explain this fact, within this Stalnakerian picture? I'll propose that acceptance is question-relative: that it's perfectly intelligible and perfectly rational to accept dierent propositions relative to dierent questions. Now there has been a general turn recently towards this sort of question- relativity. Seth Yalcin (2016) has argued, for example, that belief is question- sensitive in a closely related way. But the idea that acceptance is question-relative is signicantly more modest than Yalcin's conjecture about belief. Acceptance is paradigmatically something that you do in conversation. And conversations can 2 Frank Ramsey (1926) introduced this style of argument. 208 have dierent goals. So it can make sense to do dierent things in dierent conversa- tions, given your dierent goals. Questions are a natural way of regimenting the idea that conversations have dierent goals. We individuate the goals of a conversation by appeal to the questions under exploration in that conversation. In other words: it should be uncontroversial that we need something to play the role that questions play in the following picture. The motivations for Yalcin's question-sensitivity, by contrast, are tied to more controversial commitments about the nature of proposi- tions. So our guiding idea is that dierent conversations have dierent goals. I'm proposing regimenting this guiding idea with a comparatively rich conception of conversations, which I take directly from Craige Roberts' work. 3 The rst part of the picture is a particular question that all parties to the conversation take to be under discussion. These questions are questions that a particular conversation is aimed at answering. We might ask why James ate lunch at 11:30 today, or why it's raining now. That is, we might tacitly understand the conversation as aimed at answering that question. And I'll use those questions to regiment the idea that dierent conversation have dierent goals. I'll reify those questions: I'll say that conversations are associated with Questions under Discussion, which I'll abbreviate QUD. Conversation A has dierent goals than Conversation B if the question under discussion in A is independent of the Question under Discussion in B. Agents can disagree about what propositions to treat as true when they're investigating a question. If I think Zeus controls the rain and you think he doesn't exist, we're going to treat dierent propositions as true when exploring why it's raining now. 'll treat propositions about Zeus' actions as true. And you'll treat them as false. So we'll disagree, because we treat dierent propositions as true in exploring this question. In making this point, I'm appealing to the triadic relation x treats y as true relative to z. This triadic relation is a relation between agents, propositions, and questions. How should we understand this notion? I'll appeal to the agent's dis- positions. This relation holds only if the agent is disposed to use the proposition as a premise in reasoning about the question under discussion. I treat propositions about Zeus' actions as true because I'm disposed to use them as premises in further reasoning. And you treat propositions about Zeus' non-existence as true because you're disposed to use them as premises in your further reasoning. I'll treat this triadic relation between agents, propositions, and questions as a primitive. I'll build an edice on top of this primitive; it's at the heart of my account of disagreement in acceptance. I've just given you a heuristic to tell when this relation holds. It holds when the agent is disposed to use the proposition as a premise when reasoning about the question. Now it doesn't matter for me if this relation ultimately holds because a sui generis, question-directed mental state holds, of if the relation is grounded in yet more primitive mental states, which may not be question-directed. For our purposes, we can treat this relation as a primitive, whatever the deep metaphysical facts. 3 See Roberts (2012). 209 We can treat this triadic relation as a primitive because our purpose is to develop generalizations about agents who take propositions as true. The central general- ization that I want to introduce here is a generalization about when two agents disagree. My idea is that agents disagree if they accept incompatible propositions. Because acceptance is question-relative, my ocial account is: Disagreement in Acceptance: pA and B disagree about pq is true as uttered in a context where Q is the question under discussion if, relative to Q, A accepts some proposition that is inconsistent with a proposition that B accepts. I've glossed acceptance as treating a proposition as true relative to a question. So a more revealing gloss on the picture uses that vocabulary. Disagreement in Acceptance: pA and B disagree about pq is true as uttered in a context where Q is the question under discussion if there is some proposition that A is disposed to treat as true relative to Q that is incompatible with a proposition that treats as true relative to Q. And I've given a heuristic for telling when an agent treats a proposition as true relative to a question. She treats it as true if she's disposed to use it as a premise in reasoning about that question. It's worth emphasizing at the outset that this account is fully classical, in that it grounds disagreement in logical relations between the objects of the attitudes. Moreover, the background logic can be fully classical. The picture of disagreement in acceptance patterns exactly with disagreement in belief, and licenses all the same inferences that disagreement in belief does. Take the corresponding claim about belief. Disagreement in Belief: pA and B disagree about pq is true if there is some proposition that A believes that is incompatible with a proposition that B believes. Disagreements in belief are grounded in logical relations between the propositions believed. There isn't any disagreement if the propositions believed are all consis- tent. My account of disagreement in acceptance doesn't draw on any odd resources other than the triadic treating-as-true relation. All the other resources are just the resources we would use to explain disagreement in belief. My account in that way highly conservative. 10.3 Connecting the basic account to the target phenomenon The last section introduced my basic account of disagreement in acceptance. I introduced a range of notions (questions under discussion, treating a proposition as true...). The task of this section is to show that my formal account does link up 210 with an ordinary notion of disagreement. 4 The last section proposed the following sucient condition for disagreement in acceptance. Disagreement in Acceptance: pA and B disagree about pq is true as uttered in a context where Q is the question under discussion if, relative to Q, A accepts some proposition that is inconsistent with a proposition that B accepts. This condition is only a sucient condition. It doesn't say that all ordinary cases of disagreement will t this condition, but only that the ordinary notion of disagree- ment applies when this condition is met. This modesty is independently motivated. The phenomenon I'm describing has to do with conversations: I'm trying to explain when people would disagree with each other when talking with each other. There is no guarantee that ordinary notions of disagreement only include disagreement in conversation. And it doesn't matter for me if it does. I'll discuss ve examples where we'll ascribe disagreement. The rst case only an illustration of how the account works in an ordinary case involving only the agent's beliefs. The other cases involve more recharch e examples. Imagine that we're discussing why the Dodgers won the game last night, and only drawing on our beliefs about what happened { the strategies that the dier- ent teams used, what happened when they tried to execute those strategies, and so on. The question under discussion is then why did the Dodgers win last night. The propositions that we each accept are about what happened in the game, and about whether the strategies used were eective. In this case, the dispositions to accept propositions are explained by our beliefs: we're disposed to accept those propositions because we believe them. Disagreement in Acceptance is plausi- ble in this case. Suppose that one of accepts that there was a run scored in the fth inning, and the other doesn't. Ordinary speakers would use the sentence they disagree about whether a run was scored in the fth inning to describe that possibil- ity. Similarly, suppose that one of accepts that the strategy used in the fth inning was eective, and the other doesn't. Ordinary speakers would use the sentence they disagree about whether the strategy used in the fth inning was eective to describe that possibility. This rst case is just an ordinary case of disagreement in belief. It's helpful as an illustration of the way the account works. But it doesn't have the broader philosophical payos that matter here. Remember that it's really important for the positivist's broader purposes that there are cases of disagreement in acceptance that aren't also cases of disagreement in belief. Remember that the positivist denies that moral knowledge requires belief about the fundamental realist property: she thinks that it involves a much less demanding attitude. Now the positivist also claims that this more minimal attitude also supports disagreement. But it's not possible to defend this claim just by appeal to the sort of case described in the last paragraph. That case doesn't show anything about attitudes other than belief. 4 I'm grateful to Steve Finlay for getting me to appreciate the importance of this task. 211 What we need are cases where agents are treating a proposition as true, but for some reason other than that they believe it. I'll discuss four possible examples of this phenomenon, in hopes of convincing you that it is genuine. Here's the rst example. You and agree that three possibilities are exhaustive and mutually exclusive: that the answer is in book A, that it's in book B, and that it's in book C. There are time constraints on where we look to nd the answer. We can only look in one book. So we have to proceed as if one of the three propositions is true. Now I think that the rst possibility is .4 likely, and the other two are .3 likely. You, by contrast, think that the last possibility is .4 likely, and the other two are .3 likely. In that case, (3) is true. (3) We disagree about whether the answer is in book A or in book C. This case is important, because it is a case of disagreement that is not grounded in disagreement in belief. It's not plausible that I believe that the answer is in book A. After all, my credence in that possibility is less than .5 { I'm more condent that that possibility doesn't hold than that it does. It's very hard to understand how I could genuinely believe a proposition while thinking that it's more likely to be false than to be true. Moreover, it's not plausible that you believe that the answer is in book C, for the same reason. We instead disagree because we're accepting incompatible propositions. Because of the time constraint, I have to treat one of the propositions as true. And because of our dierent credences, it makes sense to treat dierent propositions as true. It seems merely accepting incompatible propositions is enough to support ascriptions of disagreement, like (3). This case is a perfectly clean example of disagreement in acceptance that's not also disagreement in belief. If it convinces you, you can just skip the rest of this section. I'll just be giving more examples of the same phenomenon. But these other examples do add something. I've suggested that there is a class of attitudes that are tied together in an interesting way: that all the attitudes in this class involve treating a proposition as true for some reason. Belief falls into this class. And we've just found another member of this class: treating a proposition as true because it's the proposition that's most likely to be true. Moreover, there are philosophical payos to exploring just how broad this class of attitudes is. Suppose we nd that a broad class of attitudes patterns in a similar way. This discovery should increase our condence that there are interesting generalizations about this class of attitudes. That's why I'll explore more examples in the rest of this section. I want to convince you that there is a broad class of attitudes that are tied together by an interesting generalization. The next example is about gendered language in a religious context. Some religious people will insist that God loves his children. They'll also insist that God isn't male. Since his presupposes that the salient individual is male, they're accepting a proposition that they don't believe. In this case, they're accepting it because they regard it as the least unsatisfactory option. Other possessive pronouns (in English) presuppose that the individual is female, or that the group has several 212 members. 5 There is also disagreement about gendered language for God. Some people think that it does more harm than good, and avoid it. We can describe them as disagreeing. (4a) They disagree about whether God is a he. This observation is just what Disagreement in Acceptance predicts: that dis- agreement in what the agents accepts is sucient for ascriptions of disagreement like (4a). If you agree that (4a) is appropriate, you recognize a case where someone is treating a proposition as true for some reason than that they believe it. Now there are interesting further complications about gendered religious lan- guage. Other apparently similar constructions are more controversial. (4b) They disagree about whether God is male. You won't accept (4b) if you think that God transcends gender but still use the masculine pronoun. This observation looks like trouble for Disagreement in Ac- ceptance. It suggests that disagreement about whether to accept that God is male is not sucient for ascriptions of disagreement. However, someone who avoids the masculine pronoun is likely to accept (4b): they think that it accurately cap- tures the disagreement. This observation suggests that there are several notions of disagreement oating around: disagreement in belief as well as disagreement in acceptance. The two parties agree in belief, but disagree in acceptance. The per- son who rejects (4b) interprets it as describing disagreement in belief, and rejects it for that reason. And the person who accepts (4b) interprets it as describing disagreement in acceptance, and accepts it for that reason. Gendered language reveals an important fact about the grounds of acceptance. It's initially natural to think that an agent's cognitive states determine what she accepts. She accepts p if she believes p, or if she believes p if q, if she's in a con- versation that assumes q. But gendered language suggests that an agent's practical states can also help determine what she accepts. Religious believers who use gen- dered pronouns may use those pronouns for practical reasons { because they're the least bad option that lets them talk about God as a person. And an agent's practical states help determine what she accepts in other cases, too. Suppose we're lost in the woods, and come to a branch in the road. We're in a group together, and we're all going to head the same way. We decided to ip a coin to guide our decision. Unfortunately, we don't coordinate. You and ip two dierent coins and the coins come up dierently. Before I know what you did, I'd say that the way out is to the left, and you'd say the opposite. It's then true that: (5) We disagree about whether the way out is to the left. 5 See Robin Barwise and Cooper (1981) and Philippe Schlenker (2003) for arguments that gender marking triggers presuppositions. 213 This disagreement doesn't go very deep. The group would rerun the coin-clip, or decide that one of the two coin- ips is the ocial one. But we would still use (5) to describe it, before it's resolved. Now it is easy to switch the case so that the impression of disagreement disap- pears. Imagine, that we're not part of the same group and we're facing the same decision on dierent days. Even if we ipped a coin and the coins came up dier- ently, we wouldn't use (5) to describe the disagreement. But this case can also ts the general framework. In the modied case, dierent questions are under discus- sion. In group 1 on Tuesday, they're asking whether the way out for group 1 on Tuesday is to the left. And in group 2 on Wednesday, they're asking whether the way out for group 2 on Wednesday is to the left. 6 I've described four examples of disagreement in acceptance, where the theo- retical tools that I've described line up with intuitive judgments about when two people disagree. Unfortunately, though, there are other examples that don't t my theoretical tools as well. You and I know both know that relativistic mechanics is correct and Hamiltonian mechanics is not. Imagine that we're interested in some engineering problem { how to design some device. I think Hamiltonian mechanics is a close enough approximation, and so much easier to use. In drawing up my de- sign for the device, I'm accepting Hamiltonian mechanics. I'm disposed to use the theory in further reasoning. You, by contrast, accept relativistic mechanics even in the engineering context that we nd ourselves in. You do the laborious calculations each time, despite the time constraints on our nishing the project. You're disposed to use relativistic mechanics as a premise in further reasoning. It's natural to describe this case as a case where you and I accept dierent propositions. I accept that Hamiltonian mechanics is correct, and you do not. If this gloss is right, Disagreement in Acceptance predicts that (6) should be perfectly assertable. (6) We disagree about whether Hamiltonian mechanics is correct. And (6) does not seem to be appropriate. We agree about the correctness of Hamil- tonian mechanics. This observation suggests that it's wrong to think of this case as involving acceptance of inconsistent propositions. We should think of it as a case of Stevensonian disagreement in action: whether to do the calculations in one way rather than in another way. Then my theoretical apparatus simply does not apply. It applies only when agents are accepting dierent propositions, and we're no longer understanding this case as involving disagreement about any propositions. The framework fromx2.1 gives us a better perspective on this point. When would the two parties disagree about whether Hamiltonian mechanics is correct? 6 Aren't they really discussing the same question: whether the way out is to the left? An answer to that question entails an answer to the question they're discussing. So they should be discussing the general question if they're discussing the more particular question. Not at all { questions under discussion don't work like that. If they did, one of the questions under discussion would always be: what is the way the world is? And that result would undermine the theoretical interest of questions under discussion. 214 Well, we would have accept inconsistent propositions relative to the same question under discussion. And it's hard to nd a question under discussion that would do. The question could either be factual, or prudential. factual: something like: is Hamiltonian mechanics correct? prudential: something like: given our practical goals, ought we use Hamil- tonian mechanics in our calculations? Neither would license disagreement about whether Hamiltonian mechanics is cor- rect. We agree on purely factual questions. So there won't be any disagreement if the question under discussion is just about the facts. And the truth about Hamilto- nian mechanics isn't really relevant, if we're exploring the prudential question about what to use. Let's recap. I've argued that disagreement in acceptance is a genuine phe- nomenon. I've given four genuine cases of disagreement in acceptance: (i) quo- tidian disagreement about why the Dodgers won, (ii) disagreement grounded in dierent credences, (iii) disagreement about gendered religious language, and (iv) disagreement about the path out of the woods. In each of these cases, it is natu- ral to describe the parties as disagreeing. Crucially, though, only the rst kind of case involves disagreement in belief. Cases (ii) through (iv) are all possible even if the relevant parties all have the same beliefs. They should lead us to acknowledge disagreement in acceptance as a genuine phenomenon. There are complications. The discussion of Hamiltonian mechanics suggested that it's sometimes dicult to tell when we've found a case of disagreement in acceptance. Even in that case, though, the framework fromx2.1 helps explain why that case isn't a case of disagreement in acceptance. There doesn't seem to be a question under discussion that the two parties disagree about. That framework looks very fruitful. 10.4 Returning to positivist realism Here's where we are. At a high level, we're exploring whether the positivist has an adequate account of moral disagreement. It's important to explore this point, be- cause the positivist is trying to build a halfway house between traditional metaeth- ical absolutism and traditional metaethical relativism. She is trying to claim the relativist's explanation of moral knowledge, while retaining the absolutist's account of disagreement. This chapter is exploring one facet of this high-level question. We're asking whether the positivist's appeal to acceptance prevents her from re- taining the absolutist's account of moral disagreement. I've just introduced my ocial framework for theorizing about acceptance. This section and the next takes up our question within my ocial framework. Things turn out just how the posi- tivist would hope. Let's start by reviewing the problem that I'm concerned to answer. Remember that some attitudes don't support disagreement. You can wonder whether p without 215 disagreeing with someone who wonders whether:p. Other attitudes, by contrast, do support disagreement. If you believe p, you disagree with someone who believes :p. I'll describe this dierence by saying that belief is inconsistency-transmitting, while wondering is not. 7 If acceptance were uniformly inconsistency-transmitting, the positivist would be in the same position as the traditional absolutist. Agent A V-es that you may tell that lie. Traditional Absolutist: A V-es that the moral standard permits telling that lie Positivist: A V-es that o 1 permits telling that lie and accepts that o 1 is the moral standard. Agent B V-es that you may not tell that lie. Traditional Absolutist: B V-es that the moral standard forbids your telling that lie. Positivist: B V-es that o 2 forbids your telling that lie and accepts that o 2 is the moral standard. Traditional absolutists predict that A and B disagree if V-ing is inconsistency- transmitting. The two agents V inconsistent propositions. So they disagree. The positivist makes the same prediction whenever acceptance is inconsistency-transmitting. o 2 and o 1 are identical or they're distinct. If they're identical, the two agents V inconsistent propositions about the demands of the same ordering. If they're dis- tinct, the two agents accept inconsistent propositions about which ordering is the moral standard. So they're guaranteed to disagree whenever V-ing and acceptance are both inconsistency-transmitting. The positivist's account would coincide with traditional realism if acceptance were always inconsistency transmitting. But we saw earlier that acceptance isn't always inconsistency-transmitting. You can accept a proposition in the course of reasoning by reductio. For example, you can accept that there is a largest prime, in the course of arguing that no such prime exists. And you can be consistent in doing that even while you believe that there is no largest prime. But belief also supports acceptance. You can be disposed to accept p because you believe p. As a result, you can be consistent in accepting-for-reductio that there is a largest prime while also be disposed to accept-in-a-belief-grounded-way that there is no largest prime. Now those two attitudes are both instances of acceptance. So acceptance isn't always inconsistency-transmitting. 7 This vocabulary is from Mark Schroeder (2008a). 216 The last section gave a range of examples where acceptance is inconsistency- transmitting. But we've just seen an example where it isn't. Does my framework have the right structure to distinguish these two kinds of examples? Yes. That framework is built around my Disagreement in Acceptance claim. Disagreement in Acceptance: pA and B disagree about pq is true as uttered in a context where Q is the question under discussion if, relative to Q, A accepts some proposition that is inconsistent with a proposition that B accepts. Now Disagreement in Acceptance states a sucient condition for disagree- ment. In order for this framework to distinguish our two types of examples, it has to hold that the sucient condition is met in one class of examples, and not met in the other. In particular, it would have to hold that the sucient condition is not met when we're comparing accepting-for-reductio and accepting-in-a-belief- grounded-way. And the only way for that sucient condition to not be met is for the agent to be accepting dierent propositions relative to dierent questions. And it is highly plausible that the agent is exploring dierent questions in the two contexts. In the context where beliefs matter, the agent is exploring the nature of mathematical reality. The Question Under Discussion might be: QUD1: At-issue: What are the facts about prime numbers? Not-at-issue: [nothing] And the question is quite dierent when the agent is reasoning by reductio. One option to characterize the QUD is with the sort of unanchored discourse referents that the last chapter described. Where `i' is a term for the unanchored discourse referent, the QUD might be: QUD2: At-issue: What properties does i have? Not-at-issue: i is the largest prime number We might explore QUD2 to try to show that there is no consistent set of prop- erties that i can have. Now I can intelligibly accept dierent propositions when investigating the rst question than when investigating the second question. For example, I can intelligibly accept that there is no largest prime number when ex- ploring QUD1, while not accepting that proposition while exploring QUD2. That's why acceptance isn't inconsistency-transmitting across the two kinds of contexts. The sucient condition that usually makes acceptance inconsistency-transmitting is not met. This result is exactly what I hoped for; I relativized acceptance to questions to capture just this point. I observed that dierent conversations have dierent goals. I regimented this observation by appeal to questions. Two conversations 217 have dierent goals when they are exploring dierent questions. It is intuitive that a conversation where I'm reasoning by reductio has dierent goals than a conversation we're I'm reasoning just from the propositions that I believe. My framework makes good on this intuitive conviction by regimenting the dierent goals with dierent questions. Now you might not like appealing to unanchored discourse referents { you might think that there is something illegitimate about doing that. There are other ways of capturing the data that I use unanchored discourse referents to capture, as described in the previous chapter. However, I don't want to work through those other options here. Once you see how the explanation goes with unanchored discourse referents, you can reconstruct it within whatever approach you prefer. The important point is that there is some QUD that distinguishes contexts where we're reasoning by reductio from those contexts where we're reasoning just from what we believe. Accepting p relative to QUD1 isn't inconsistency-transmitting with accepting p relative to QUD2 because QUD1 is suciently dierent from QUD2. 10.5 How the moral case works Let's take up the moral case { nally! I'm claiming that the positivist has a license for optimism that she will have the account of disagreement that traditional realist has. Now one way for her to have the traditional realist's account is for acceptance to be inconsistency-transmitting across every context. But we've seen that acceptance isn't plausibly inconsistency- transmitting in that way. But another way for the positivist to have the tradi- tional account of disagreement is for contexts where acceptance isn't inconsistency- transmitting to be irrelevant for the traditional realist { that is, they're contexts where the traditional realist doesn't intend her account to apply. And that is just what I'll suggest. To illustrate how some contexts could be irrelevant for the traditional realist's account of disagreement, think about anthropological or inverted commas use of moral language. I can use ordinary moral sentences when speaking as an anthro- pologist. I can describe the beliefs of a community by saying children are morally required to care for their aging parents. And I can do that even while disagreeing with those beliefs. The traditional moral realist doesn't take this case to be a case where I'm disagreeing with myself. Rather, she thinks that my commitments in the anthropological context are somehow irrelevant to my commitments in a context where I'm investigating what morality itself demands. What the traditional realist really cares about, I think, are the contexts where we're investigating what morality itself demands. She intends her account of dis- agreement to be an account that applies in those contexts. Now I individuate contexts by appeal to the question under discussion. So the traditional realist is interested in contexts where she takes the QUD to be: Traditionalist QUD: what does the moral standard demand? 218 The anthropological context is not a context where we're investigating this QUD. And that's the reason I don't disagree with myself when I assertively utter one sentence in an anthropological context and assertively utter its negation in a context where we're investigating this QUD. I claim that the positivist has the realist's account of disagreement in those con- texts where the traditional realist takes us to be investigating this QUD. If this claim is true, the positivist is home free. The traditional realist account of disagreement is only plausible for those contexts where we're investigating the Traditionalist QUD. In making this claim, then, I'm claiming that the positivist's account of dis- agreement coincides with the traditional realist's account in those cases where the traditional realist's account is plausible. Fix on the contexts where the traditionalist takes us to be investigating the Traditionalist QUD. Now the positivist will have a slightly dierent theory about the QUD under discussion in those contexts. The positivist diers from the traditional moral realist in taking propositions about the moral standard to be presupposed and not-at-issue. So she will take the QUD in these cases to be about something that is presupposed to be the moral standard. As in the earlier prime number case, we might use an unanchored discourse referent in characterizing this QUD. The QUD might then be: Positivist QUD At-issue: What does i demand? Not-at-issue: i is the moral standard Any context where the traditional realist takes the Traditionalist QUD to be under discussion is a context where the positivist takes the Positivist QUD to be under discussion. And { crucially! { acceptance is inconsistency-transmitting across all contexts where the Positivist QUD is under discussion. Given my ocial framework, acceptance is inconsistency-transmitting whenever the same question is under dis- cussion. To recall: Disagreement in Acceptance: pA and B disagree about pq is true as uttered in a context where Q is the question under discussion if, relative to Q, A accepts some proposition that is inconsistent with a proposition that B accepts. I've been noting that there are cases where acceptance fails to be inconsistency- transmitting. My framework captures those cases by holding that dierent questions are under discussion in each of the two cases. That's what I did in the discussion of the two cases about the largest prime number. So as soon as the positivist identies a single QUD as what's being investigated in two dierent contexts, she predicts that acceptance is inconsistency-transmitting across those contexts. And that's just what I've done. I've identied one QUD (the Positivist QUD) as being investigated across the range of contexts where the 219 traditionalist takes her Traditionalist QUD to be investigated. I've thus shown that acceptance is inconsistency-transmitting across the range of contexts where the traditionalist's account of disagreement is plausible. And that's just what I set out to show. Let's wrap up. I've argued that the positivist has a license for optimism that she has the same account of disagreement as the traditional moral realist. License for Optimism: The traditional moral realist's proposition is an obvious a priori consequence of the positivist's propositions. (The propositions that S expresses if the traditional moral realists are right is an obvious a priori consequences of the propositions that S expresses if the positivist is right.) This chapter and the last have considered the two ways that this license might fail. First, there there might be no plausible linguistic means for us to disagree (or agree) about the full range of propositions that the positivist associates with moral utterances. Second, acceptance might not be inconsistency-transmitting in the right range of cases. At this point, we know enough to see that neither possibility will obtain. We can be condent that the positivist's license for optimism is genuine. 220 Chapter 11: Disagreements about knowledge I've argued in Part I that positivist realism has important upshots for metaethics. And I've been trying to show in Part III that the positivist really does have an adequate account of moral disagreement. Is there tension between these two goals? I've had to introduce some striking new resources to explain moral disagreement. And you might reasonably worry that those striking new resources undercut the philosophical payos of Part I. After all, I haven't said anything at all about how my striking new resources interact with knowledge ascriptions. This chapter remedies this lacuna. It describes how the striking new resources would interact with knowledge ascriptions. In the course of doing that, I'm show how positivist realism still has the same important upshots for metaethics, even if it draws on those striking new resources. 11.1 Knowledge and unanchored discourse referents Early chapters of the dissertation introduced positivist realism as having important epistemic consequences. A knowledge ascription like (1) is associated with two propositions. (1) I know that slavery is wrong. At-issue: I know that o 1 forbids slavery, and I accept that o 1 is the moral standard Not-at-issue: o 1 is the moral standard I've claimed that I can know about o 1 in virtue of my mental states { in virtue of my mental representation of o 1 . And I've claimed that that knowledge is comparatively easy, and does not require the realist to take on controversial commitments in epistemology. It is natural to think that the positivist needs to change her account of what utterances of (1) communicate when she accepts the lessons of the previous two chapters. Aquinas and I disagree about a range of questions. For one thing, he thinks it's morally wrong to charge interest on loans, and I think it's morally per- missible. But we will still take each other to have a signicant amount of knowledge in other cases. We would both accept (2). 221 (2) Aquinas and I both know that it's sometimes permissible to kill in self-defense. Suppose we glossed (2) as involving knowledge of sets of rules. It might be about Aquinas' set of rules, or about mine. As about Aquinas' set of rules At-issue: I know that o A forbids slavery, and I accept that o A is the moral standard Not-at-issue: o A is the moral standard As about my set of rules At-issue: I know that o M forbids slavery, and I accept that o M is the moral standard Not-at-issue: o M is the moral standard But then we've found another instance of the problem explored in Chapter 9. Nei- ther of the at-issue propositions are true. Aquinas doesn't accept my set of rules as the moral standard, and I don't accept his as the moral standard. The positivist might react to this problem in several dierent ways. But it seems to be the very same kind of problem that we encountered in Chapter 9. It would be reasonable to hope that the solution to that problem also solves this problem. In fact, it might well count against the positivist if she has to draw on some a dierent kind of resource to explain how agents with some dierent moral beliefs could still share some pieces of moral knowledge. Her view would be comparatively complex if she had to do that. So we should hope to explain (2) by appeal to whatever tool we favored from Chapter 9. I favor unanchored discourse referents. So I want to gloss (2) as communicating a proposition about an unanchored discourse referent. (2) Aquinas and I both know that it's sometimes permissible to kill in self-defense. At-issue: Aquinas and I both know that i forbids slavery, and both accept that i is the moral standard Not-at-issue: i is the moral standard But to interpret this gloss, we need to understand what it is to know a proposition with an unanchored discourse referent. I claim that knowledge about an unanchored discourse referent can be grounded in knowledge of a singular, anchored proposition. If this claim is true, the the shift to unanchored discourse referents leaves all the positivist's philosophical advantages intact. The only kind of knowledge required for (2) is knowledge about an unanchored discourse referent. 222 Here's a concrete illustration. I know about my own set of rules: I know that o M forbids slavery. (See Chapter 5 for a defense of this conclusion.) And Aquinas also knows about his own set of rules: he knows that o A forbids slavery. (Again, see Chapter 5.) That is, we know singular, anchored propositions about particular sets of rules. If that anchored knowledge also grounds unanchored knowledge about i, we both know that i forbids slavery. And anchored knowledge does ground unanchored knowledge. Consider (8). (8) The linguists know that Chester killed the mouse. They also know that Chester tortured the mouse rst. And the philosophers know that a cat killed the mouse. And they know that it tortured the mouse rst. But they don't know which cat it was. If (8) is true, (9a) is also true. (9a) The linguists and philosophers all know that a cat killed the mouse. They also know that it tortured the mouse rst. This example shows that anchored knowledge about an individual grounds unan- chored knowledge, that is, knowledge of an unanchored proposition. ((9a) has to ascribe knowledge of an unanchored proposition. The philosophers don't know any anchored propositions about mice-killing cats. We'd hear (9a) as false if it did as- cribe knowledge of anchored propositions.) Since (9a) follows from (8), the linguists' anchored knowledge grounds unanchored knowledge. Once we recognize this point, all the positivist's advantages reappear. Earlier chapters have explained how we can know anchored propositions about our indi- vidual sets of rules. And knowledge of anchored propositions grounds knowledge of the corresponding unanchored propositions. The same point holds of justication. (10) The linguists are justied in believing that Chester is a cat who killed the mouse. They also are also justied in believing that Chester tortured the mouse rst. (11) The philosophers are justied in believing that Mikey is a cat who killed the mouse. They also are also justied in believing that Mikey tortured the mouse rst. Justication in believing the anchored propositions again grounds justication in believing the unanchored propositions. (9b) is true if (10) and (11) are true. (9b) The linguists and philosophers are justied in believing that a cat killed the mouse. They also are also justied in believing that it tortured the mouse rst. Justication and knowledge both transmit from anchored thought to unanchored thought. Now there are still dierences between the two. Agents can be justied in believing inconsistent propositions, but they can't know inconsistent propositions. 223 The positivist smoothly explains this point. She's already explained why agents can't know inconsistent anchored propositions. And that explanation also explains why agents never know inconsistent unanchored propositions in this way: we're re- stricting our attention to explanations of our knowledge of unanchored propositions that grounds it in knowledge of anchored propositions. So appeal to unanchored discourse referents does not threaten the positivist's philosophical payos. The payos come because the positivist thinks that moral knowledge comes much easier than philosophers have appreciated. Now we've just seen that knowledge of anchored propositions is sucient for knowledge of unan- chored propositions. The positivist has given us an account of a sucient condition for knowing an anchored proposition about a set of rules. (That's what Chap- ter 5 did.) Since anchored knowledge is sucient for unanchored knowledge, the positivist has thereby given us a sucient condition for knowing an unanchored proposition about a set of rules. The positivist's philosophical advantages survive intact. It's reasonable to think that a similar point will hold whatever account you favor of intensional anaphora. The inference from (8) to (9a) shows that knowing a singular proposition is sucient for knowing the kind of content (whatever it is) that's involved in intensional anaphora. The positivist's philosophical advantages should survive any theory of intensional anaphora. 11.2 Is moral knowledge too easy? The rest of this chapter will explore some technical questions about knowledge of unanchored discourse referents, and their relation to knowledge of anchored dis- course referents. Feel free to skip the rest of this chapter if you buy the story I just gave inx11.1. Part III of the dissertation makes it easier for speakers to agree or disagree about what they ought to do. They don't need to accept the same set of rules as the moral standard. They only need to accept sets of rules that coincide in their demands in this case. Now it takes more for two people to know what they ought to do than for them to agree about what they ought to do. In making it easier for two people to agree, it's important that we haven't made it too easy for them to both know. And there is a danger that we might have made it too easy for them to know what they ought to do. Suppose that Jeremy's set of rules is dierent from Im- manuel's, but the two coincide in some case. They both forbid this particular lie. (It's both a lie and something that has bad long-term consequences.) But Immanuel doesn't think that Jeremy knows what he ought to do. He gets too many similar cases wrong. Can the positivist vindicate Immanuel's judgment? It's helpful to formulate this challenge as an inference. Suppose that Jeremy truly believes that this lie is wrong. Suppose further that he's justied in believing that his own set of rules forbids that lie, and indeed that his justied belief has whatever features prevent the belief from being Gettier-ed. (Maybe, for example, the belief about his 224 own set of rules is also safe.) Immanuel should still be able to deny that Jeremy knows that this lie is wrong. Can the positivist explain why? The answer is somewhat complicated. There are two interpretations of this in- ference, depending on whether we're interpreting Jeremy's beliefs as about anchored propositions or about unanchored propositions. The inference is invalid on the an- chored interpretation, and valid but unsound on the unanchored interpretation. Let's start with the anchored interpretation: Immanuel's set of rules is o K and Jeremy's o J . (a) Jeremy justiably believes that this lie is impermissible. at-issue: Jeremy (justiably believes that o J forbids that this lie) and accepts that o J is the moral standard. not-at-issue: [nothing] (b) And this lie is impermissible. at-issue: o K forbids that this lie not-at-issue: o K is the moral standard (c) And Jeremy's belief has whatever feature distinguishes justied true belief from knowledge. at-issue: Jeremy's belief that o J forbids that this lie has whatever feature distinguishes justied true belief from knowl- edge. not-at-issue: [nothing] (d) So Jeremy knows that this lie is impermissible. at-issue: Jeremy knows that o K forbids that this lie and accepts that o K is the moral standard. not-at-issue: o K is the moral standard This inference is invalid. Nothing in these premises guarantees that Jeremy accepts Immanuel's set of rules o K as the moral standard. The anchored interpretation won't make Jeremy's knowledge too easy. Isn't there another problem here? The inference from these premises to this conclusion is intuitively valid, and does not seem to have an invalid reading. But the positivist predicts that there is an invalid reading. Fortunately, though, the positivist has a companion in innocence. Consider this inference. (a) Beth is tall. (b) And Bill is tall. (c) So there are at least two people who are tall. 225 We interpret this inference as valid. It's very hard to hear one reading of this inference where it's valid, and another inference where it's invalid. But there is actually a reading of this inference where it's invalid. Interpret the rst and third occurrences of tall as taller than 6ft, and the second occurrence as taller than 5ft. The fact that Bill is taller than 5ft doesn't guarantee that Bill is taller than 6ft. On this reading, then, (c) doesn't follow. This problem is structurally parallel to the problem that arises for the positivist. In both cases, the semantics for the construction will allow for more readings than we in fact hear. It's unclear what to say about this question. It suggests that there is a strong preference for interpreting context-sensitive terms uniformly throughout an inference like this one. We somehow don't consider the readings of the dierent sentences that make the inference fail. So the positivist predicts that the natural language inference from (a) to (d) has a reading that's invalid. I've noted that we don't hear this inference. But it's not surprising that we don't hear the invalid inference. The invalid inference involves interpreting context sensitive items inconsistently, as sometimes referring to o J and sometimes to o K . But the tall case suggests that those interpretations are dispreferred. Inferences like the one from (a) to (d) are generally interpreted with context-sensitive items interpreted uniformly. It's a general fact that we don't hear the invalid readings where context-sensitive items are interpreted inconsistently. As a result, the only reading of this inference that we hear is the unanchored reading, where Jeremy's beliefs are all about an unanchored discourse referent. This inference is valid but unsound. (a) Jeremy justiably believes that this lie is impermissible. at-issue: Jeremy (justiably believes that i 1 forbids that this lie) and accepts that i 1 is the moral standard. not-at-issue: [nothing] (b) And this lie is impermissible. at-issue: i 1 forbids that this lie not-at-issue: i 1 is the moral standard (c) And Jeremy's belief has whatever feature distinguishes justied true belief from knowledge. at-issue: Jeremy's belief that i 1 forbids that this lie has whatever feature distinguishes justied true belief from knowl- edge. not-at-issue: [nothing] (d) So Jeremy knows that this lie is impermissible. 226 at-issue: Jeremy knows that i 1 forbids that this lie and ac- cepts that i 1 is the moral standard. not-at-issue: i 1 is the moral standard To vindicate Immanuel, the positivist has to deny that (a), (b), or (c) is true. But (b) is denitely true. That was the whole payo of Chapter 9 { I want to explain how Immanuel and Jeremy could regard themselves as agreeing, and so attribute true beliefs to each other. And the positivist has to take (a) as true. She thinks that familiarity with our own set of rules is enough to give you justied beliefs about what you ought to do, at least in most cases. (Remember that you're justied in some unanchored belief if you're justied in the corresponding anchored belief.) And it seems like (c) should also be true. (c) requires Jeremy to have whatever feature distinguishes justied true belief from knowledge. Maybe it requires safety, or the exclusion of accident, or something like that. And surely Jeremy also has safe beliefs about the demands of his own set of rules! One of the core positivist claims is that it's fairly easy to satisfy safety requirements. You can satisfy it entirely in virtue of your own attitudes about your set of rules. I deny that (c) is true. The present challenge is pointed only when (b) and (c) are both true { that is, the conjunction in (f) is true: (f) And this lie is impermissible. And Jeremy's belief has whatever feature distinguishes justied true belief from knowledge. at-issue: i 1 forbids that this lie, and Jeremy's belief that i 1 forbids this lie has whatever feature distinguishes justied true belief from knowledge. not-at-issue: i 1 is the moral standard And in this case, the conjunction in (f) is not true. It requires Jeremy to have safe beliefs about the set of rules (whatever it is) that bears on the permissibility of telling the lie. And he denitely doesn't! Or at least he doesn't if Immanuel is right. According to Immanuel, the consequences don't matter for the permissibility of the lie. Since Jeremy thinks that they do matter, his beliefs about the permissibility of the lie aren't safe. Just change the consequences of telling the lie! A similar point holds if knowledge require the exclusion of accident. It's only an accident that Jeremy got this case right. He was lucky enough not to be in a case where the (morally irrelevant) consequences happened to line up with the obligations that he in fact has. That's the intuitive gloss on why (f) isn't true. Unfortunately, though, the details take some care. (f) constrains the maximal extensions of the speaker's beliefs: those extensions have to turn out in a certain way for (f) to be true. But what are the exact constraints? A natural rst pass turns out to be inadequate. First Pass i 1 forbids this lie 227 i 1 is the moral standard all the nearby worlds where Jeremy believes that i 1 forbids this lie while accepting that i 1 is the moral standard are worlds where i 1 does forbid this lie. This First Pass is inadequate, because it requires objectual quantication into Jeremy's beliefs. The speaker accepts these three unanchored propositions i the objectual existential closure of these three constraints is true at all the maximal extensions of her beliefs. Unfortunately, though, there is no one set of rules that the speaker and Jeremy both accept to be the moral standard, either at the actual world or at nearby worlds. So the objectual existential closure of these three constraints is false at all the maximal extensions of the speaker's beliefs. To avoid this problem, we have to distinguish the speaker's unanchored discourse referent (which I'll label `i 1 ') from Jeremy's (which I'll label `i 2 '). Then we need to require that the two line up in the cases we care about. Final Pass i 1 forbids this lie i 1 is the moral standard all the nearby worlds where Jeremy believes that i 2 forbidss this lie while accepting that i 2 is the moral standard are worlds where i 1 does forbid this lie. The objectual quantier does not bind i 2 . Jeremy knows that this lie is impermissible is true i all maximal extensions of the speaker's beliefs verify that 9x [x forbids this lie, and x is the moral standard, and all the nearby worlds where Jeremy believes that i 2 forbids this lie while accepting that i 2 is the moral standard are worlds where x does forbid this lie]. This gloss preserves what the positivist needs to preserve. Immanuel will deny that Jeremy knows that this lie is impermissible, because Jeremy wrongly thinks that the consequences determine whether it's permissible. Jeremy doesn't know even if he gets lucky and ends up with the true belief that this lie is impermissible. That true belief isn't knowledge, because there are nearby worlds where Jeremy thinks that the act has dierent consequences; at those worlds, his beliefs don't line up with what Immanuel takes to be the moral standard. The positivist can distinguish knowledge from mere true belief, and also from justied true belief. 228 11.3 Other kinds of knowledge attributions The suggestion from the last section also answers one more challenge about knowl- edge ascriptions. Suppose that Jeremy's belief that this lie is impermissible falls short of knowledge. Then (19) is true. (19) Jeremy' belief that this lie is impermissible isn't also a piece of knowledge. How can the positivist explain this point? Jeremy's belief here is about the set of rules that Jeremy accepts to be the moral standard, or so it seems. Belief attribu- tions aren't factive, so they aren't constrained to be about whatever set of rules the speaker accepts to be the moral standard. But Jeremy does know the proposition that his own set of rules. (It's pretty easy to know about your own set of rules, at least if the positivist is right.) It then seems like the positivist should predict that (19) is false { Jeremy's belief that this lie is impermissible is in fact a piece of knowledge. Note in the rst place that this problem does not arise if the belief and knowledge are both about unanchored discourse referents. (20) Jeremy' belief that i 1 forbids this lie and acceptance that i 1 is the moral standard isn't also a piece of knowledge that i 1 forbids this lie and acceptance that i 1 is the moral standard, where i 1 is the moral standard. And we've already explained why knowledge about unanchored discourse referents is more demanding than knowledge about your own set of rules plus coincidence of that set of rules with the moral standard in the particular case under discussion. If (19) has (20) as its logical form, we can explain why it's true. But (19) should have another use, where it ascribes an anchored belief about Jeremy's own set of rules to Jeremy. Is that a reading we can hear? First, let's get clearer on the syntax of (19). I assume that (19) contains material that has been deleted or moved, so that it's as in (19lf). (19lf) Jeremy' belief that this lie is impermissible isn't also a piece of knowledge that this lie is impermissible. And context-sensitive items in deleted or moved material are resolved in the same way. Fix on a deictic use of (21). (21) Jeremy' belief that he's hungry also a piece of knowledge. (21lfI) Jeremy' belief that he's hungry also a piece of knowledge that he's hungry. On the deictic use, both occurrences of he are interpreted as being about the same individual. (Suppose that the rst occurrence of he refers to Billy. (21) doesn't get 229 to be true just because Jeremy also knows that some other guy John is hungry.) So the sets of rules in (19lf) have to be the same. (19lf) Jeremy' belief that o 1 forbids this lie and acceptance that o 1 is the moral standard isn't also a piece of knowledge that o 1 forbids this lie and acceptance that o 1 is the moral standard, where o 1 is the moral standard. The problematic use of (19) was in a context where we knew that Jeremy and the speaker accept dierent sets of rules to be the moral standard. When we hit the rst occurrence of `o 1 ', we interpret it as being about Jeremy's set of rules. But we can't sustain this interpretation by the end, since we're required to interpret `o 1 ' as referring to whatever set of rules the speaker accepts to be correct. And no particular set of rules satises both those requirements. Fortunately, though, there is another interpretation of this discourse, where it's about unanchored discourse referents. That's the reason why that's the only one we hear in this context. It's the only reading that satises all the constraints on interpreting the utterance. 11.4 Wrapping up Part III has been exploring if the positivist have a stable intermediary position be- tween traditional absolutists (Foot, Scanlon, Railton, ...) and traditional relativists (Harman, Street, Finlay, ...). Now traditional absolutists take moral judgment to involve an attitude to only one proposition. The positivist, by contrast, takes moral judgment to involve two attitudes to two dierent propositions. This chapter and the last has tried to show that the positivist also has the absolutists' advantages in explaining disagreement and agreement. The last chapter opened with a general license for optimism on this point. (License for optimism) The absolutist proposition is an obvious a priori consequence of the positivist's propositions. So if two agents make opposing moral judgments, each of them is committed to the corresponding absolutist proposition. And those absolutist propositions are sucient to explain moral disagreement. The positivist can mimic the absolutist's account of disagreement, if the license for optimism is correct. I've been making good on this license for optimism. I've been showing how the positivist conceives of a range of dierent moral disagreements, and giving the evidence for her approach. 230 Part IV Part IV: Normative upshots 231 Chapter 12: A positivist vindication of the method of re ective equilibrium There is a range of powerful evidence for positivist realism, as we saw in Part II. There is also the central philosophical challenge to the positivist: whether she can explain moral disagreement. But we've seen in Part III that she does have a solution. It's reasonable to conclude that positivist realism is true. This conclusion has important upshots for normative ethics. The goal of Part IV is to introduce some of them. I'll focus on two upshots in particular. Positivist realism makes it easier to answer the sort of cluelessness objection that James Lenman (2000) has leveled against consequentialists. It also vindicates the sort of evolutionary debunking arguments that philosophers like Peter Singer (2005) has leveled against rival views { it explains how those arguments can be legitimate. This chapter lays the groundwork for those later upshots. Normative ethicists standardly rely on the method of re ective equilibrium. They defend their theory by showing that it best ties our particular moral convictions into a coherent whole. The positivist agrees that this method is the right method to use in normative ethics. But her explanation of the method is very distinctive. It is that distinctive explanation that grounds the upshots of her view. My goal in this chapter is to introduce the positivist's distinctive explanation of the method. 12.1 Introduction Let's review. The positivist has a distinctive hybrid conception of moral judgment, where moral judgment is a hybrid of judgment de re about some set of rules, plus acceptance that that set of rules is the moral standard. Knowledge, for example, is a hybrid of knowledge de re about some set of rules, plus acceptance of that set of rules as the moral standard. (1) I know that killing is usually wrong. At-issue: I know that o 1 usually forbids killing, and accept that o 1 is the moral standard Not-at-issue: o 1 is the moral standard The picture follows from the discovery that the fundamental realist property is a 232 presupposition of moral utterances. It follows from that discovery because accep- tance is the attitude appropriate for presuppositions. Now remember that there are several dierent species of acceptance. You accept a proposition if you treat it as true. So belief or knowledge are themselves perfectly intelligible grounds for accepting a proposition. But there are other grounds as well. You can accept a proposition to humor your fellow philosophers, or in the course of reasoning for conditional proof. Finally { and most importantly for our purposes here { your credences can ground acceptance. Suppose that three possibilities are jointly exhaustive and mu- tually exclusive: that the answer is in book A, that it's in book B, and that it's in book C. There are time constraints on where we look to nd the answer. We can only look in one book. I think that the rst possibility is .4 likely, and the other two are .3 likely. I can then accept that the answer is in book A, because of my credences. I'll call this credally-grounded acceptance. The positivist takes sincere, committal use of moral language to require credally- grounded acceptance. Fix on a case where you're using moral language to describe what you really think { the use without \inverted commas". In that case, the positivist takes you to be accepting that some particular set of rules o 1 is the moral standard. Your accepting that proposition needs grounded in your credences. You need to be more condent that that particular set of rules is the moral standard than that any other particular set of rules is the moral standard. Now you might still have a very low credence in the proposition that you accept. What matters is that you have an even lower credence in any alternative. Parts II and III may have obscured this important fact. The positivist acknowl- edges uses of moral language that do not involve credally-grounded acceptance. The anthropological use of moral language is one example. In fact, the positivist takes that use of moral language to be important evidence for her proposal. But it doesn't constrain what she says about the sincere, committal use. She can and does insist that that use requires credally-grounded acceptance. The positivist thus assumes that ordinary agents already have credences about what is the moral standard. And those credences ground what she accepts to be the moral standard. Importantly, though, the positivist just takes those credences as given. It's not her job to explain how we developed those credences. It's the job of the sciences concerned with human behavior { social-cognitive neuroscience, for example. 1 What the positivist contributes is an explanation of how those credences can support moral knowledge. 12.2 Normative theorizing: its nature and method Normative ethicists have generated an enormous body of work. They argue back and forth about surgeons and trolleys, about killing and letting die, about deonto- logical theories versus consequentialist theories, and on and on. The positivist has a distinctive picture of what normative ethicists are doing in those debates. 1 See Chapter 5 for a fuller discussion of this point. 233 In fact, there are in principle two ways of conceiving of work in normative ethics, given positivist realism. On the rst, the fundamental aim is to describe the nature of the property being-the-moral-standard. On the second, the fundamental aim is to describe the set of rules that the speaker herself accepts. The rst option is far more plausible. The positivist realist is a realist. She thinks that there are truths about what we ought to do that are independent of and more fundamental than our individual evaluative attitudes. Since moral truths are mind-independent, and normative ethics is an investigation of moral truth, normative ethics is an investigation of mind-independent facts. Only the rst conception of normative ethics takes it to be investigation of mind-independent facts. On the second conception, normative ethics is about a psychological question { about what I myself accept to be the moral standard. And that's not how a realist will think about it. The only way for the positivist realist to think about normative ethicist is as investigating the property that is part of the presupposed, not-at-issue proposition: the property being-the-moral-standard. Theses in normative ethics are then theses about that fundamental property. For example, rule consequentialism is a thesis about that property. (Rule Consequentialism) 8x [x is the moral standard i x is the set of rules whose embedding has the best consequences] And Scanlonian contractualism would also be a claim about this property: that x is the moral standard if no one could reasonably reject x. Now almost all rule consequentialists take themselves to disagree with Scanlonian contractualists. The positivist conceives this disagreement as disagreement about the fundamental prop- erty being-the-moral-standard. That's how the positivist thinks of the goal of normative theorizing. But nor- mative ethicists tend to pursue this goal in a distinctive way. They use the method of re ective equilibrium. 2 That is, they try to show that their normative theses best tie together our moral intuitions into a more coherent whole. For example, they might appeal to our intuition that it's permissible to switch the trolley from the track where it kills ve to the track where it kills one, and also to our intuition that it's impermissible for a surgeon to cut up one person and use their organs to save ve people. Of course, not all normative ethicists rely on concrete intuitions about trolleys and surgeons. Some appeal to highly abstract and general intuitions, like intuitions about goodness and maximization, rather than more concrete intu- itions. 3 I'll set those sorts of normative ethicists aside for now, and return to them in Chapter 14. 2 In uential discussions of this method include at least: Norman Daniels (1979), Je McMahan (2000), John Rawls (1971), and T. M. Scanlon (2002). 3 Henry Sidgwick (1907) is one familiar example. In fact, consequentialists sometimes argue explicitly against the method of re ective equilibrium { those reacting just to Rawls (1971) include Peter Singer (1974) and R. M. Hare (1973). 234 The positivist's distinctive hybrid theory extends to talk about intuitions as well. My intuition is just an intuition about my own set of rules. (3) I have the intuition that cutting up one to save ve is impermis- sible. At-issue: I have the intuition that o 1 forbids cutting up one to save ve, and accept that o 1 is the moral standard Not-at-issue: [nothing] This prediction follows from the positivist's general proposal. Have the intuition is an attitude verb, like believes, or knows, or hopes. So the positivist has to give the very same gloss on (3) that she gives on other attitude ascriptions. We're focusing on normative ethicists who use their intuitions as a source of evidence. Unfortunately, though, it's not at all clear that it makes sense to do that, if the positivist is right. The evidence just doesn't match the goal. The goal is not to describe our own set of rules. The goal is to describe the fundamental moral property being-the-moral-standard. But there's no guarantee that our set of rules does line up with that property. So it's not clear if the method of re ective equilibrium does make sense, within the positivist's framework. 12.3 In ationary versus de ationary vindications of the method Now the method of re ective equilibrium would make sense if something guaranteed that our set of rules did mostly line up with the moral standard. So one way to defend the method is to try and nd some such guarantee. For example, you could defend the method by arguing that we do have a reliable way of forming beliefs about the fundamental realist property. And in fact, a range of philosophers do defend the method in just this way. Je McMahan (2000) is one example: he argues that reasoning from our intuitions is a reliable way of forming beliefs about the moral standard. I'll call this approach an in ationary vindication of the method of re ective equilibrium. It in ates the normative ethicist's commitments, to include an expla- nation our reliability. But the positivist is trying to avoid these sorts of commitments. Maybe we each do have a reliable way of forming beliefs about the moral standard, maybe we don't. Ordinary moral knowledge survives either possibility. That was what Chapters 1-3 showed: it just doesn't matter if something guarantees that our set of rules do line up with the moral standard. So it would be unfortunate for the positivist if she needed these sorts of commitments to make sense of the method of re ective equilibrium. Fortunately, though, the positivist can avoid in ationary commitments. Take my intuition that it's wrong for the surgeon to cut up one to save ve. That intuition involves the intuition that my own set of rules o m forbids cutting up one to save ve. In order to use that intuition to do normative ethics, I have to treat my own 235 set of rules as evidence about the fundamental moral standard. An in ationist like McMahan thinks that it makes sense to do that only if my own set of rules are a reliable source of evidence about the moral standard. That's why he takes on his in ationary commitments. But the in ationist is wrong. It makes sense to treat my own rules as a source of evidence whether or not I am reliable. My having the intuition also involves me accepting that o m is the moral standard. And, trivially, if o m is the moral standard, it's a reliable guide to the moral standard. So my having the intuition requires me accepting that o m is a reliable guide to the moral standard. We don't need any in ationary defense of the method of re ective equilibrium. There is a trivial inference from one of the constituents of a moral intuition to the conclusion that my own set of rules is a reliable guide to the moral standard. Here's another way to make this central point. If you don't accept your own rules to be the moral standard, you just don't have moral intuitions. Someone who nds herself with a moral intuition is tacitly accepting that her own set of rules is the moral standard. Now maybe it's a mistake to have any moral intuitions at all. The next section will take up that possibility. But virtually none of us are actually in that position. We all (or almost all) have moral intuitions. The positivist accepts the following Hermeneutical Picture. (Hermeneutical Picture) Normative ethicists use their own set of rules as evidence about the moral standard. The positivist needs to explain how it makes sense to use your own set of rules as evidence about the moral standard. She's a realist; she acknowledges that the two can be dierent. I've contrasted two defenses of the Hermeneutical Picture. Why does it make sense to use your own set of rules as evidence about the moral standard? (The Inflationary Defense) because you are reliable in forming beliefs about the moral standard. (The Deflationary Defense) because of the commitments that fol- low from your having an intuition. In having a moral intuition, you're accepting your own set of rules as the moral standard. If what you accept is true, then (trivially) your set of rules are a reliable guide to the moral standard. I favor the de ationary defense. The rest of Part IV will focuses on positivists who do accept it. I'll call them de ationary positivists. Importantly, the de ationary positivist is adopting an explicitly psychologistic conception of normative ethics. My own work in normative ethics will aim at articulating the structure that lays behind the set of rules I accept to be the moral standard. Nothing more, and nothing less. This psychologistic conception of normative ethics is unusual. But it is worth noting that it is continuous with some suggestive methodological remarks from Rawls. 236 Now one may think of moral theory at rst (and I stress the provisional nature of this view) as the attempt to describe our moral capacity; or, in the present case, one may regard a theory of justice as describing our sense of justice. ... A conception of justice characterizes our moral sensibility when the everyday judgments we do make are in accordance with its principles. These principles can serve as part of the premises of an argument which arrives at the matching judgments. We do not understand our sense of justice until we know in some systematic way covering a wide range of cases what these principles are. (Rawls 1971, 41,x9) There is a close relationship between Rawls' talk about our \moral sensibility" and the positivist's talk about our individual set of rules. Now even though Rawls endorses this sort of psychologism, it's highly contro- versial. One central problem is that it struggles to explain moral disagreement. If the task of normative ethics is to unpack our individual moral sensibility, and if individual moral sensibilities vary from individual to individual, normative ethicists may well be talking past each other when they disagree. But normative ethicists aren't talking past each other. 4 But the positivist has an adequate account of moral disagreement. Establishing that fact was the whole point of Part III! So the de ationary positivist is perfectly entitled to her psychologism. She's already done all the hard work to disarm the traditional problems. 12.4 Is a de ationist attitude psychologically possible? You may nd de ationary positivism deeply unsatisfying. You might argue that there are substantive philosophical questions that the de ationary positivist is re- fusing to answer. This chapter closes by arguing that no such argument can be found. You might argue that the de ationary positivist is violating some rational norm, somewhere. Maybe the de ationary positivist is violating a rational norm in believ- ing propositions about her own set of rules? Chapter 5 addressed this challenge. It showed that you can be perfectly rational in believing propositions about your own set of rules. Maybe the de ationary positivist is violating a rational norm in ac- cepting some ordering as the moral standard? Chapter 10 addressed this challenge. There is only one general norm on acceptance: the propositions that you accept have to be probabilistically coherent. And the de ationary positivist can be perfectly probabilistically coherent. Imag- ine that I'm certain that the moral standard is one of nine sets of rules. (For any other set of rules o i where i > 9, my credence that o i is the moral standard is 0.) I can be .2 condent that o 1 is the moral standard, and .1 condent that any of o 2 through o 9 is. That set of attitudes is probabilistically coherent, and so satises 4 See C. L. Stevenson (1937) for one early statement of this kind of point. 237 the requirements that rationality imposes on acceptance. It's just not possible to use rational constraints to argue against the de ationary positivist. The only remaining challenge is a psychological one. The de ationist does need to accept some particular set of rules as the moral standard. And her acceptance has to be credally-grounded: she has to be more condent that that particular set of rules is the moral standard than that any other set of rules is the moral standard. And the in ationist might try to show that it's not psychologically possible for a de ationist to distribute her credences in that way. After all, the de ationary positivist is happy to allow for many dierent scientic explanations of why we accept what we accept to be the moral standard. The in ationist might argue that it's not psychologically possible for a de ationist to accept any particular set of rules as the moral standard when you recognize that sort of scientic explanation. It's psychologically possible to do that only if you think that we are somehow reliable in forming beliefs about what's the moral standard. This challenge is very, very ambitious. It purports to establish that something is psychologically impossible for creatures like us. It should look pretty implausible, precisely because it is so ambitious. But it's the only remaining challenge to make, if you nd de ationary positivism unsatisfying. And, even though this challenge looks implausible at rst, it raises issues that will be important in later chapters. So I want to give my ocial answer to it here. I'll rely on this answer later on. The core answer is that a range of moral beliefs are an ineliminable part of hu- man life. For example, they're important for our reactive attitudes, like resentment or indignation { attitudes involved in holding someone responsible. I follow P. F. Strawson (1974) in holding that the reactive attitudes are an ineliminable part of human life. He writes: the human commitment to participation in ordinary inter-personal re- lationships is, I think, too thoroughgoing and deeply rooted for us to take seriously the thought that a general theoretical conviction might so change our world that, in it, there were no longer any such things as inter-personal relationships as we normally understand them; and be- ing involved in inter-personal relationships as we normally understand them precisely is being exposed to the range of reactive attitudes and feelings that is in question. (Strawson 1974, 14) I'll use this Strawsonian part in explaining why moral beliefs are psychologically ineliminable. The rationality of our reactive attitudes plausibly depends on our moral beliefs. Importantly, it's their rationality, and not their existence, that depends on our moral beliefs. We should reject the Existence-on-Moral thesis, and accept the Rational-on-Moral thesis. (Existence-on-Moral): I [resent/am indignant/feel guilty...] to- wards x for -ing only if I believe it was morally wrong for x to 238 (Rationality-on-Moral): I'm rational in [resenting/being indig- nant/feeling guilty...] towards x for-ing only if I believe it was morally wrong for x to The Rationality thesis is weaker than the Existence thesis. If the Existence thesis is true, the Rationality thesis is too, but not vice versa. (If you can't have the attitude without having the relevant belief, you can't rationally have the attitude without having the belief.) Because theRationality thesis is weaker, it is more plausible. Imagine some- one who was raised Catholic, and raised to accept prohibitions on recreational sex. They have come to reject those prohibitions. But they still experience something phenomenally like guilt about recreational sex. If the existence of guilt depends on moral belief, these experiences are not experiences of guilt. So if you think that they are genuine experiences of guilt, you reject the Existence-on-Moral thesis. If, by contrast, you think that that guilt is genuine but irrational, then you accept Rationality-on-Moral. 5 And there is good reason to accept theRationality-on-Moral thesis. There does seem to be something irrational about feeling guilty about recreational sex, if you think that there's nothing wrong with it. Rationality-on-Moral explains why. It's irrational given Rationality-on-Moral, because you don't have the corresponding moral belief. Let's return to the psychological challenge. The challenge is that it's not psy- chologically possible for a de ationist to accept some particular set of rules as the moral standard. Now the positivist takes moral belief to require accepting a partic- ular set of rules as the moral standard. So the challenge succeeds only by showing that it's not psychologically possible for a de ationist to have moral beliefs. Sup- pose that I'm such a de ationist, and imagine that I was just punched in the face for no reason. The psychological challenge purports to show that I can't believe that that punch was wrong. That is, it purports to show that (4) is false. (4) I believe that the unprovoked punch is wrong. Why does the psychological challenge entail that (4) is false? (4) requires me to accept that some particular set of rules is the moral standard. (4) isn't true if I'm equally condent that the moral standard permits the punch as that it allows it. I'm going to resent an unprovoked punch. Now givenRationality-on-Moral, I'm rational in resenting the punch only if I believe the punch is wrong { that is, if (4) is true. Moreover, I can regard myself as rational in resenting the punch even while fully attending to the Rationality-on-Moral thesis. Why? The best explanation is that I do believe that the unprovoked punch is wrong. And in giving that explanation, we do not appeal to any of the facts that the in ationist appeals to. I don't need to regard myself as having a reliable way of forming beliefs about 5 Joseph Butler (1729) and John Rawls (1971) both accept the strongerExistence-on- Moral, and so accept Rationality-on-Moral. R. Jay Wallace (1998) argues against the Existence claim, but accepts the Rationality claim. 239 the moral standard. I just nd myself resenting the punch, and come to believe that it was wrong. That's all I need to do. And if it's all I need to do, the in ationist's challenge misres. The core idea really is very modest. It's that our moral beliefs are an inelim- inable part of human life. The core idea can be so modest because the psychological challenge has to be so ambitious. It purports to establish a point about what's psy- chologically impossible for creatures like us. I'll wrap up by noting another way this core idea really is modest. Its conclu- sion is perfectly compatible with some kinds of error theory about morality. Now error theorists disagree among themselves about the rational response to their error theory. Some of them think that the rational response is to stop making moral judgments and having moral beliefs. Proponents of this view (like Ian Hinckfuss (1987)) are sometimes called abolitionists. They contrast with error theorists like Richard Joyce (2001) who think that the rational response to the error theory is to maintain the pretense that moral beliefs are true. If my argument is right, abo- litionism is wrong. Creatures like us just can't give our moral beliefs up, at least if we continue to regard our reactive attitudes as reasonable. If we become error theorists, we should adopt a pretense theory of the sort that Joyce recommends. I'm giving a Strawsonian argument for doing that, but the upshot is the same. There are other ways of making the same modest point. Joyce has given a very dierent kind of argument for having something like moral beliefs. He argues that it's instrumentally useful to do so. However, if [an agent] is thinking that-ing simply must be performed, then the possibilities for rationalization diminish. Someone armed with moral beliefs thinks of cooperation as categorically required, and the distinctive value of categorical imperatives is that they silence calcula- tion. (Joyce 2001, 213) Joyce's argument has the same upshots for the de ationary positivist as my Straw- sonian argument. If either one succeeds, it's psychologically possible for the de- ationary positivist to accept substantive propositions about the moral standard. Crucially, though, that psychological possibility involves treating her own set of rules as a good guide to the moral standard. It thus involves just what the Hermeneu- tical Picture takes normative theorizing to involve. Here's a heuristic for evaluating what I've tried to accomplish in this chap- ter. Imagine that you somehow came to accept an error theory about morality. Would you still feel a psychological pull towards believing that unprovoked punches are wrong? Would Strawsonian observations about the reactive attitudes pull you there? Or would Joyce's argument? If either would, then you're on board with my conclusion in this chapter. After all, believing that unprovoked punches are morally wrong involves accepting that your own set of rules is a guide to the moral standard. And normative theorizing only involves accepting your own set of rules as a guide to the moral standard. Normative theorizing would make perfect sense. And it would still make perfect sense even given the positivist's de ationary take 240 on central questions in moral epistemology. 12.5 Looking forward The positivist accepts the following Hermeneutical Picture. (Hermeneutical Picture) Normative ethicists use their own set of rules as evidence about the moral standard. The positivist needs to explain how it makes sense to use your own set of rules as evidence about the moral standard. She's a realist; she acknowledges that the two can be dierent. I've contrasted two defenses of the Hermeneutical Picture. Why does it make sense to use your own set of rules as evidence about the moral standard? (The Inflationary Defense) because you are reliable in forming beliefs about the moral standard. (The Deflationary Defense) because of the commitments that fol- low from your having an intuition. In having a moral intuition, you're accepting your own set of rules as the moral standard. If what you accept is true, then (trivially) your set of rules are a reliable guide to the moral standard. I favor the de ationary defense. I've been concerned to show that it is psychologi- cally stable { that it is compatible with our moral beliefs. Note an important point about the de ationary suggestion. Normative the- orizing doesn't by itself deliver knowledge about the moral standard. Normative theorizing starts with you accepting that your set of rules is a good guide to the moral standard. And it involves you accepting further conclusions about the moral standard based on your own set of rules. Acceptance in, acceptance out. But normative theorizing can still extend our moral knowledge, even though it doesn't extend our knowledge about the moral standard. To illustrate, take someone who once thought that interracial marriages were morally wrong. And imagine such a person coming to believe otherwise, by the method of re ective equilibrium. Imagine her reasoning from her beliefs about the moral signicance of marriage, and her beliefs about the moral signicance of race in most contexts, and coming to the conclusion that interracial marriages are perfectly ne. Does she then know that conclusion? Yes. (4) She knows that interracial marriages are perfectly ne. At-issue: she knows that o 1 permits interracial marriages and accepts that o 1 is the moral standard. Not-at-issue: o 1 is the moral standard 241 Her normative theorizing involved her coming to accept that the new set of rules o 1 is the moral standard, by reasoning from what she accepts. But her accepting that o 1 is the moral standard doesn't need to have any positive epistemic status in order for (4) to be true. As a result, it doesn't matter that normative theorizing only delivers more states of acceptance. It can still extend moral knowledge. The positivist's vindication of re ective equilibrium is very striking and distinc- tive. It's very dierent from a coherentist vindication of the method, a la John Rawls (1975). Those coherentists argue that our moral intuitions gain their epis- temic credentials by their relations to each other. It's also very dierent from a traditional intuitionist vindication of the method, of the sort that Je McMahan (2000) has developed. He does take up central issues in moral epistemology, in the course of trying to vindicate the method of re ective equilibrium. The de ationary positivist has a de ationary attitudes towards debates about coherentism and debates about intuitionism. Maybe some kind of coherentism is true, and we know propositions about the moral standard because of the relations between those propositions. And maybe some kind of intuitionism is true, and we know propositions about the moral standard because we apprehend those proposi- tions in the right way. But maybe the coherentists and intuitionists are both wrong. Even then, normative theorizing still delivers moral knowledge. The following chapters will show that this de ationary attitude has important methodological upshots. In particular, it illuminates what sorts of considerations might be defeaters for a moral intuition { that is, considerations that make it a mistake to rely on that intuition. It denitely seems like some considerations are defeaters. For example, discovering that my intuition that p is the result of my socio-economic position defeats that intuition. But it's unclear what range of considerations pattern with this example. Consider this list. 1. My socio-economic position explains why I have the intuition that p. 2. I'm ignorant of the non-normative facts that bear on whether p. 3. My evolutionary history explains why I have the intuition that p. It's very unclear if the nal two considerations are defeaters for our intuitions, and when they are. De ationary positivism illuminates these questions. In brief: igno- rance of the non-normative facts is a much less signicant defeater if de ationary positivism is true than if it's false. And facts about my evolutionary history are a much more signicant defeater if de ationary positivism is true than if it's false. Chapter 13 establishes the rst point, and Chapter 14 the second. 242 Chapter 13: Rule consequentialists should be posi- tivists The last chapter described the distinctive way that the de ationary positivist vin- dicates the method of re ective equilibrium. And I promised that this vindication has distinctive upshots for the methodology of ethics. In particular, I promised that it would illuminate the conditions under which the following range of considerations are defeaters for an intuition that p. 1. My socio-economic position explains why I have the intuition that p. 2. I'm ignorant of the non-normative facts that bear on whether p. 3. My evolutionary history explains why I have the intuition that p. The rst consideration denitely seems like a defeater. But it's very unclear if the second two are, or under what conditions they are. This chapter focuses on the second kind of consideration: ignorance of relevant non-normative facts. There has been bizarrely little discussion of when such ignorance is a defeater. This chapter thus has two parts. The rst part establishes a baseline, about the considerations under which such ignorance looks like a defeater. The second part shows how positivist realism would dramatically shift that baseline. Now it's easier to establish a baseline by focusing on a particular normative view. I've picked rule consequentialism. If ignorance of the non-normative facts ever defeats an important range of moral intuitions, it would defeat an important range of intuitions if rule consequentialism is true. (After all, ordinary agents like you and me tend to be ignorant of the sorts of consequences that would matter if rule consequentialism is true.) At the same time, though, rule consequentialists do tend to draw on the method of re ective equilibrium. Our question thus involves holding them to a methodological standard they themselves accept. xx1{2 will develop a baseline generalization about the importance of ignorance about non- normative facts. If that baseline generalization is right, it's not possible to defend rule consequentialism as the outcome of re ective equilibrium. Positivist realism dramatically shifts the baseline. Ignorance of non-normative facts is much less likely to be a defeater if the positivist is right than if she's wrong. In particular, it does make sense to defend rule consequentialism as the outcome of re ective equilibrium, if positivist realism is true. In fact, something stronger is true. You need to accept a view like positivism realism if you want to defend 243 rule consequentialism as the outcome of re ective equilibrium. Some view like that is necessary to disarm the argument developed inx1. This point will be the rst illustration of the important normative upshots of positivist realism. 13.1 Rule consequentialism in re ective equilibrium? Not on standard assumptions! Rule consequentialists take an act to be wrong just in case it is forbidden by the best set of rules. And they understand what set of rules would be \best" in terms of the overall consequences of those rules being embedded in some sense in the community { widely or universally followed, on some views, or accepted, or internalized, by some or all or most of the community. 1 In defending this thesis, they tend to draw on the method of re ective equilibrium. They try to show that rule consequentialism is the best way of tying our more particular moral convictions into a coherent whole. This paper will argue that, if rule consequentialism is true, it's not possible to defend rule consequentialism in that way. The problem will arise because ordinary agents tend to be ignorant of the sort of consequences that would matter if rule consequentialism is true. I develop the problem in two stages. The rst stage shows that this sort of ignorance threatens the justication of rule consequentialism as the outcome of re ective equilibrium. And the second stage develops some diagnostic tools to reveal when this sort of ignorance does occur. Those tools pick up some of the points about cluelessness that James Lenman (2000) has made. They constitute the baseline generalization that I'm concerned to establish. Let's start with two representative rule consequentialists. Brad Hooker holds \an act is wrong if and only if it is forbidden by the code of rules whose internalization by the overwhelming majority of everyone everywhere in each new generation has maximum expected value in terms of well-being (with some priority for the worst o)" (Hooker 2000, 32) Derek Part holds that \everyone ought to follow principles whose universal acceptance would make things go best" (Part 2011, 375). Both Hooker and Part use the method of re ective equilibrium to defend their rule consequentialism. Hooker is explicit about this approach: \the best argument for rule-consequentialism is that it does a better job than its rivals of matching and tying together our moral convictions, as well as oering us help with our moral disagreements and uncertainties" (Hooker 2000, 104). 2 Part is similarly reliant on 1 Richard Brandt (1963), John Harsanyi (1977), Brad Hooker (2000), Derek Part (2011a), Holly Smith (2010), and Michael Ridge (2006b) are representative of some of the dierent options. 2 The second advantage - oering us help with moral disagreements - is downstream from the rst. It's only because the consequentialist theory matches and ties together our moral convictions that it's trustworthy enough to help us with out moral disagreements. 244 our particular moral judgments. For example, he thinks that Kantian strictures on treating agents as mere means are mistaken. And he thinks that we can see the mistake by what they predict in particular cases. Third Earthquake: You and your child are trapped in slowly col- lapsing wreckage, which threatens both your lives. (Part 2011a, 222). ... In Third Earthquake, you cannot save your child's life except by crushing Black's toe, without Black's consent. This act, I believe, would be justied (Part 2011a, 231) Part use our beliefs about this case as evidence against Kantian strictures. The overarching goal of this section and the next is to argue that it's self- defeating to defend rule consequentialism in this way. That is, they aim at showing that, if rule consequentialism is true, it's a mistake to rely on these sorts of beliefs in normative ethics. Now this charge is tricky to develop. Suppose that the later part of the paper shows that I don't know some moral belief, because I'm ignorant of the consequences that matter. Even if I establish this point, it might still be legitimate for the rule consequentialist to continue relying on that belief in arguing for her theory. It's controversial whether re ective equilibrium requires beliefs with any special epistemic standing, like being known. For example, it's not clear that John Rawls' in uential discussion of re ective equilibrium incorporates any such requirement. He instead seems to take our moral beliefs to gain epistemic standing in virtue of their relations to each other. Several of the most prominent accounts of re ective equilibrium have a similar structure. 3 Maybe the consequentialist is already thinking of re ective equilibrium in this way. It's not that we start out with justied beliefs about what we ought to do. Rather, we systematically re- ect on our more particular judgments, and then become justied in believing the consequentialist theory. I don't want to contest this point. But there is still an important problem for the rule consequentialist, even if we think of re ective equilibrium in this undemanding way. Ignorance of the conse- quences also shows that the belief does have a special epistemic standing. It shows that it is defeated. Our ignorance about long-term consequences is the defeater. And this observation creates a problem for the rule consequentialist. Work on re- ective equilibrium uniformly allows that initial judgments can be defeated, and insists that it's a mistake to rely on defeated judgments. (See, for example, TJx9, p, 42.) We've already seen one reason for allowing initial judgments to be defeated. We might come to recognize that some particular judgment is best explained by our socio-economic position. That fact would be a defeater for the particular judgment. A conception of re ective equilibrium that disallows this possibility is implausible. 3 John Rawls (1971, 2005), T. M. Scanlon (2002, 2014), and Norman Daniels (1979) are all examples. Pp. 114 - 6 of Rawls (2005) is one of the loci classici. 245 13.1.1 Ignorance of the consequences The rest of this section will argue that too many moral beliefs would be defeated if rule consequentialism is true. Now this argument doesn't need to show that all our ordinary beliefs are defeated to establish the broader point. The method of re ective equilibrium is contrastive. You defend a theory by showing that it ts our undefeated beliefs better than the alternatives. And the danger in taking too many beliefs to be defeated is that there won't be enough beliefs left to distinguish rule consequentialism. Take the belief that it's wrong to kill at random. Several normative theories will vindicate that belief { Scanlonian contractualist as much as rule consequentialists. So that belief doesn't discriminate between the theories. I'm thus happy to concede that a range of moral knowledge would be undefeated, even if rule consequentialism is true. My problem is about the beliefs that you use to discriminate rule consequentialism from the alternatives. I'll suggest that those beliefs are defeated if rule consequentialism is true. This section will introduce one diagnostic tool for telling when our moral beliefs would be defeated. Start with a problem that James Lenman (2000) has described for act consequentialists. Many of our actions are identity-aecting { they aect which people happen to exist. Identity-aecting actions have signicant eects. The person who otherwise wouldn't have existed could do a great deal of good, or a great deal of harm. Since the good or bad done is part of the total consequences of our identity-aecting actions, those consequences are part of what determines whether the action was right or wrong. Identity-aecting actions are a particularly vivid illustration of an important general point. Our actions ramify in ways that nite creatures like us aren't in a position to appreciate. And we're usually unable to discriminate these cases. Now to make this point precise, we need to situate it in a picture about knowledge in general. A natural picture holds that knowledge requires safe belief. Williamson's formulation is characteristic: \if one knows, one could not easily have been wrong in a similar case" (Williamson 2000, 147). 4 The idea is to compare similar cases where you still form your beliefs in the way that you actually form them. A belief is safe if all those similar cases are cases where your belief is still true. Safety illuminates the structure of Lenman's reasoning. Take an act of scape- goating. Ordinary agents will continue to have the same beliefs about the wrongness of scapegoating whether this act of scapegoating leads to the birth of some mass murder or prevents that birth. If act consequentialism is true, ordinary agents could easily been incorrect about this act of scapegoating in similar cases. For example, ordinary agents would still think that this act of scapegoating is wrong even in a case where the scapegoating prevents the birth of a terrible mass murderer. Now what justies our treating this case as a similar case? Well, Lenman does seem right that act consequentialists can't vindicate ordinary moral knowledge in these cases { that is, the corresponding belief is defeated. A safety theorist could capture 4 This formulation is a characterization of the structure of our reasoning about what is known, not something that that takes us from more basic primitives to conclusions about what is known. 246 this intuitive judgment by taking this kind of case as a similar case. Talk about safety allows us to see the structure the lies behind the intuitions that Lenman is eliciting. It's not a way of convincing someone who disagrees. Rule consequentialists have a similar problem, though their problem initially looks less severe. It is hard to have safe beliefs about the consequences of embedding dierent rules. Dierent rules would have dierent consequences if something cata- clysmic happens { if, for example, the globe warms past a certain point. (Theorists in the social contract tradition are presumably right to assume moderate scarcity of resources in their theorizing, and to take this assumption to be important in the rules that would be chosen.) Now if a cataclysmic event actually happens, it could aect the truth of our moral beliefs, and so their amounting to knowledge. Any rule consequentialist has to acknowledge that fact. But focus on cataclysmic events that don't happen, but that could easily hap- pen, given what we know right now. Supposing that the earth actually warms n degrees, focus on worlds where it warms 2n or 5n degrees. Rules are likely to have dierent consequences at those worlds. And that means that what we ought to do will be quite dierent at those words. Moreover, any moral belief that would be false at those worlds is no longer safe. After all, we're imagining that those worlds count as similar cases. And a belief is safe only if it's true throughout all similar cases. Ordinary moral beliefs aren't much safer if rule consequentialism is than if act consequentialism is true. Let's get more concrete, by going back to Part's Third Earthquake case. Third Earthquake: You and your child are trapped in slowly col- lapsing wreckage, which threatens both your lives. (Part 2011a, 222). ... In Third Earthquake, you cannot save your child's life except by crushing Black's toe, without Black's consent. This act, I believe, would be justied (Part 2011a, 231) I claim that ordinary agents are not in a position to know about the consequences that matter, even in this case. Given only moderate scarcity of resources, rules allowing the in iction of harm may well have the best consequences. But it's less clear if resources are very scarce. If resources are scarce enough, the loss of a toe would be a death sentence. (You'd couldn't feed yourself anymore.) In that case, it's much less clear whether rules that allow for toe-crushing have the best consequences. They might instead be much more stringent about the in iction of harm. Inasmuch as we're counting worlds where the globe warms signicantly more than it actually does, the rule consequentialist does not vindicate ordinary moral knowledge even in this case, despite initial appearances. The important point here is modest. I concede that some possible agents could be in a position to know about the consequences that matter. I'm just suggesting that ordinary agents, like you or me, or Hooker or Part, are not in that position. Our ignorance is a defeater for us. So it's illegitimate for us to rely on these beliefs when evaluating rule consequentialism. There might be some other people who could rely on their beliefs about Third Earthquake to defend rule consequen- 247 tialism. But we're not them. We cam sharpen this problem with one more plausible assumption. We can as- sume that basic, non-testimonial moral knowledge requires sensitivity to the morally signicant features of the action. It's not enough if you happen to classify the action correctly, as permitted, or whatever { you have to be sensitive to the reasons why it's permitted. 5 Now in Part's Third Earthquake case, it may matter that the life saved is your child, and it might not. If it does matter, moral knowledge requires sensitivity to the fact it's your child. If it doesn't matter, moral knowledge doesn't require that sensitivity. In the rst case, the rules with the best consequences allow you to in ict toe-level harm, but only if the life saved is your child's; in the second case, you may always trade toe-level harm for a life. Ordinary agents are not in a position to track which of those two rules has the best consequences. Imagine that I rob the rich to save the poor. (I in ict a toe-crush level nancial loss on someone rich, and save someone with the stolen money.) The second rule permits this theft. But it is hard to know whether rules permitting this action have the best con- sequences. For one thing, they make society somewhat unstable { it's much harder to plan if it's permissible for others to in ict these kinds of harms. Maybe the lives saved really do outweigh the costly instability of the rules. And maybe they don't. The point is that ordinary agents like you and I aren't in a position to know which. Now take the other rule, which permits the in iction of harm only when the life saved is your child's. To know that that rule is best, you need to know that it is better than the rule that permits robbing the rich to save the poor. But we don't know which rule is better. 13.1.2 A more general diagnostic tool The last section introduced safety as a diagnostic tool for telling when some moral belief is defeated. This section introduces a more general diagnostic tool. Knowledge can build on knowledge. Further pieces of knowledge can put you in a position to know something further. For example, it seems like (S2) is usually true if (S1) is. (S1) Anne knows that this glass contains water. (S2) If Anne gained testimonial knowledge that that water is H 2 O, she would be in a position to know that this glass contains H 2 O. There is a connection between the two in those cases where Anne's knowledge that this glass contains water survives her learning about the nature of water. If it does, the two pieces of knowledge put her in a position to know that this glass contains H 2 O. And there is a connection between (S1) and (S2) because of general facts about knowledge, not because of particular facts about water and H 2 O. 5 Set aside derivative moral knowledge, like from testimony. We're interested in worlds with only ordinary moral agents. If none of them know what they ought to do in a basic way, nobody else can acquire knowledge from anybody else. 248 Now consider this inference. (D1) Anne knows that she may sacrice someone's toe to save her child's life. (D2) If Anne gained testimonial knowledge that rule consequentialism is true, she is in a position to know that the rules that allow you to sacri- ce a toe to save anyone's life have the best consequences. If this example is like the scientic example, (D2) is usually true if (D1) is true. The only exceptions are those mentioned earlier, where gaining the testimonial knowledge in (D2)'s antecedent destroys the knowledge in (D1). Setting aside that exception, though, (D2) seems to be true whenever (D1) is. It's an instance of the same schema as (S1) and (S2). And that schema seems to hold in virtue of general facts about knowledge. (D2) is important in this paper, because it is plausible only for those agents who are in a position to know about the long-term consequences of the rules. Af- ter all, testimonial knowledge of rule consequentialism isn't by itself enough to support knowledge about the long-term consequences. Imagine that you're totally ignorant about what you ought to do in particular cases, and also ignorant of the consequences of dierent rules. If a reliable source of testimony told you that rule consequentialism is true, you still don't know what you ought to do. You also need to know about the consequences of dierent rules. As a gloss, then, we might say that (D2) is true only for those agents who are in a rough and ready position to know about the consequences { not that they're already propositionally justied in knowing about the consequences, but rather that the thing that would put them in a position to know about the consequences is not itself knowledge about the consequences. It is implausible that ordinary agents are in a rough and ready position to know about the consequences. Go back to Hooker's formulation, to illustrate. An act is wrong if and only if it is forbidden by the code of rules whose internalization by the overwhelming majority of everyone everywhere in each new generation has maximum expected value in terms of well- being (with some priority for the worst o). (Hooker 2000, 32) My grandparents are not in a rough and ready position to know about these sorts of facts. Now maybe there are some exceptions. Maybe they're in a rough and ready position to know that rules that tend to prohibit killing have Hooker's property more than rules that don't. But those exceptions tend to be the sort of beliefs that normative theories agree about. So they don't help us discriminate among them. The earlier point about safety is a special case of this general point. If you're in a rough and ready position to know about about the consequences, then the corre- sponding beliefs are safe. And much ordinary moral knowledge looks unsafe if rule consequentialism is true. (Think of global warming far more severe than actually occurs.) But the important point is more general. A rule consequentialist who at- 249 tributes a lot of moral knowledge to ordinary people has to think, implausibly, that they're in a rough and ready position to know about the consequences. Dierent epistemologists might cash this point out dierently, safety-theoretically or not. Now I've acknowledged that there is an exception to the generalization that (D2) is true if (D1) is. (D1) Anne knows that she may sacrice someone's toe to save her child's life. (D2) If Anne gained testimonial knowledge that rule consequentialism is true, she is in a position to know that the rules that allow you to sacri- ce a toe to save anyone's life have the best consequences. The exceptions are if the knowledge in (D2) would destroy the knowledge in (D1): if Anne's knowledge in (D1) depends on her not knowing that rule consequentialism is true. Can the rule consequentialist appeal to this exception? Not in the course of defending her view as the outcome of re ective equilibrium. If she did appeal to this exception, she could never come to know her normative theory by building from her knowledge of ordinary judgments. After all, appealing to this exception concedes that knowledge of the former would undermine knowledge of the latter! 13.2 More general questions 13.2.1 Background epistemic assumptions Let's connect this problem with broader issues in general epistemology. Take this inference. 1. I have a hand. 2. If I have a hand, I'm not a brain in a vat. 3. So I'm not a brain in a vat. Does your justication in believing the premises depend on your antecedently being justied in believing the conclusion? Several philosophers think not; James Pryor (2004) and Christopher Tucker (2010) are some examples. They think that you can be justied in believing 1. before you're justied in believing 3. You can then rely on your justication for believing 1. to acquire justication for believing 3. This sort of approach is available to a wide range of epistemologists. Maybe the rule consequentialist can suggest something similar. It's not that we antecedently know about the consequences. It's rather that we come to be in a position to know about the consequences as soon as we know what we ought to do. This response concedes the thesis that this paper is arguing for. I'm claiming that an agent's moral knowledge would require her to be in a position to know about the consequences that matter (if rule consequentialism is true). Distinguish Correlation and Dependence versions of this claim. 250 Correlation If an agent knows p, she's in a position to know about the consequences that matter. Dependence An agent's moral knowledge depends on whether she's in a position to know about the consequences that matter. The Pryor/Tucker-inspired response denies Dependence but insists onCorrela- tion. And the present problem for the rule consequentialist arises as long as Cor- relation is true. Ordinary agents plausibly don't have rough and ready knowledge about the consequences. That fact combines withCorrelation to entail that ordi- nary agents don't know. This ignorance is a defeater for their moral beliefs, making it illegitimate to rely on the defeated beliefs in moral theorizing. This problem arises because there is independent evidence, emphasized inx13.1.2, for denying that ordinary agents do know about the consequences. The Moorean example is dierent. There isn't usually independent evidence against my knowing if I'm envatted. Moreover, when there is independent evidence, we will also deny knowledge of 1. Imagine a demon who's ipping a coin, and will envat you if it lands heads. Then you don't know that you're not envatted { not even if the coin lands tails. But neither do you know that you have hands. I've been arguing that much ordinary moral knowledge is like this last case, if rule consequentialism is true. Many theorists may have failed to recognize this point. But this failure isn't surprising. Similar confusion is possible in the Moorean case. An undergraduate might insist that I do know that I have hands even if there's an envatting demon around. It can then be helpful to make the sort of argument I've been making in this paper. First show that, if the agent knows p, the agent is in a position to know q. Then show that the agent isn't in a position to know q. You've shown that the agent doesn't know p, and that her belief is defeated. 13.2.2 The scope of the problem The method of re ective equilibrium is the rule consequentialist's fundamental claim to do better than act consequentialists. Ordinary moral beliefs are supposed to be the evidence on their behalf. Act consequentialists don't need that kind of evidence. They can instead appeal to abstract intuitions about goodness and maximization. 6 And the rule consequentialist can't plausibly appeal to those intuitions. 7 But be- cause the act consequentialist can appeal to those abstract intuitions, cluelessness about the consequences matters less for her. She only needs to explain how our or- dinary moral beliefs are an adequate decision procedure for maximizing our chances of doing what we ought to do. 8 The rule consequentialist needs more. She needs 6 Henry Sidgwick (1907) is one familiar example. In fact, act consequentialists some- times argue explicitly against the method of re ective equilibrium { those reacting just to Rawls (1971) include Peter Singer (1974) and R. M. Hare (1973). 7 For discussion, see Richard Arneson (2005), Robert Card (2007), and Edmund Wall (2009) 8 See, e.g., G. E. Moore (1903), J. J. C. Smart (1973), Alastair Norcross (1990), Elinor Mason (2004), and Peter Railton (1984). 251 our ordinary beliefs to be undefeated evidence for her theory. As a result, our cluelessness is much more important for rule consequentialists than for act conse- quentialists, in a way that has gone unappreciated. The rule consequentialist who recognizes her own cluelessness can't defend her own theory. Now a similar problem will arise for other theorists. For example, a similar problem will arise for Scanlon's contractualism. Suppose we're considering a de- tailed principle about the moral basis of promise-keeping, like a version of Scanlon's Principle M with additional detail lled in. 9 In particular, we're considering what sort of exceptions are permitted by the most adequate version of the principle. Scanlon allows that you can reasonably reject some more concrete principle for its social eects. But ordinary nite agents like us aren't in a position to know about the social eects. So we're not in a position to know which particular principles are reasonably rejectable. But this result looks like the problem we found for the con- sequentialist. Ordinary nite creatures like us don't know about the considerations that determine what we ought to do. However, contractualists like Scanlon can concede the points in this section much more easily than rule consequentialists can. Scanlon has the same kind of problem in one place. But ordinary agents are usually in a better position to know about the sorts of considerations that matter for him. So the range of intuitions that he can take to be knowledge is much broader than the range of intuitions the rule consequentialist can take to be knowledge. It's thus much more important for rule consequentialists to nd some mistake in the diagnostic tools I've suggested in this paper. However, other kinds of normative ethicists may also have a stake in nding some mistake. This section and the last (xx1{2) have argued that ordinary agents don't have rough and ready knowledge about the consequences that the rule consequentialist takes to matter. They have also argued that that lack of rough and ready knowledge is a defeater. That is, they've aimed at establishing this Baseline Generalization. (Baseline Generalization) Lacking rough and ready knowledge of the relevant non-normative facts is a defeater. The arguments have been complicated. But they have been complicated because I've been concerned to argue for a highly general thesis, by articulating the weakest set of assumptions that lead that thesis. The comparison with knowledge of water was my way of doing that. 9 Scanlon's Principle M: \in the absence of special justication, it is not permissible for one person, A, in order to get another person, B, to do some act, X (which A wants B to do and which B is morally free to do or not to do but would otherwise not do), to lead B to expect that if he or she does X then A will do Y (which B wants but believes that A will not otherwise do) when in fact A has no intention of doing Y if B does, and A can reasonably foresee that B will suer signicant loss if he or she does X and A does not reciprocate by doing Y" (Scanlon 1998, 298, italics mine). 252 13.3 The baseline is very dierent given positivist realism The Baseline Generalization is false if positivist realism is true. It is much easier for a rule consequentialist to defend her theory as the outcome of re ective equilibrium if the positivist is right. To appreciate this point, let's start by reviewing the positivist's conception of normative theorizing. She takes it to aim at establishing propositions about the moral standard. For example, rule consequentialism is a claim about the moral standard. (Rule Consequentialism) [8x [x is the moral standard i x is the set of rules whose embedding has the best consequences] Rule consequentialists are arguing for this thesis. Scanlonian contractualism is also a claim about this property: that x is the moral standard if no one could reason- ably reject x. Now almost all rule consequentialists take themselves to disagree with Scanlonian contractualists. The positivist conceives this disagreement as disagree- ment about the fundamental property being-the-moral-standard. And normative ethicists use their intuitions about their own set of rules as evidence in arguing for this thesis. (Hermeneutical Picture) Normative ethicists use their own set of rules as evidence about the moral standard. And the positivist's core idea is that moral knowledge does not require knowledge about the moral standard, but acceptance of propositions about it. (1) Jane knows that she ought to keep this promise. At-issue: Jane knows that o 1 requires keeping this promise, and accepts that o 1 is the moral standard Not-at-issue: o 1 is the moral standard I've throughout used an articial example to illustrate the kind of knowledge. Sup- pose that someone wrote out a book describing what o 1 demands, and that Jane memorized that set of rules. After memorizing it, she comes to accept that it ar- ticulates the demands of morality. Because she has memorized that set of rules, she knows what o 1 demands. On the positivist's picture, that's the only kind of knowledge that's necessary for moral knowledge. Jane doesn't also need to know any proposition containing the fundamental property being-the-moral-standard. She only needs to accept that her set of rules o 1 is the moral standard. The baseline describe inxx1{2 is then mistaken. (Baseline Generalization) Lacking rough and ready knowledge of the relevant non-normative facts is a defeater. 253 It doesn't matter whether we're ignorant about the non-normative facts. And the reason that it doesn't matter is that moral knowledge doesn't depend on knowledge about the fundamental property being-the-moral-standard. In particular, it doesn't matter if we're in a position to know about the consequences that determine which set of rules are best Before working through this point in more detail, though, I want to note an important caveat. The positivist takes non-normative facts to matter in two impor- tantly dierent ways. A non-normative facts can matter for the application of the best rules, or it can matter for the determination of which rules are best. Suppose that o 1 forbids all unprovoked punches. Knowing that o 1 forbids some unprovoked punch requires knowing that that punch is a punch. 10 That non-normative fact matters for the application of o 1 . If it hadn't been a punch, o 1 wouldn't have forbidden it. Then the Baseline Generalization is ambiguous between two claims. (ApplicationGeneralization) Lacking rough and ready knowledge of the application facts is a defeater. (Determination Generalization) Lacking rough and ready knowl- edge of the determination facts is a defeater. The positivist accepts the Application Generalization, and rejects the Determina- tion Generalization. The rest of this chapter will show that the arguments inxx1{2 require the Determination Generalization to work. Let's make this abstract point less abstract. One way of having rough and ready knowledge is having safe beliefs. So there are closely related generalizations about safe belief. (Safe Application) Lacking safe belief about the application facts is a defeater. (Safe Determination) Lacking safe belief about the determination facts is a defeater. x1 argued that ordinary moral beliefs aren't safe, if rule consequentialism is true. Then I relied on Safe Determination to show that rule consequentialists can't defend their theory as the outcome of re ective equilibrium. The positivist denies that Safe Determination is true. Start by supposing that o 1 is the moral standard at the actual world, and that Jane accepts that it is. Suppose further that Jane has memorized the way that o 1 works. She has memo- rized what considerations it takes to matter, and how it takes those considerations to matter. That memorization gives her knowledge of what o 1 demands. Nowx1 imagined worlds where the globe warms two or three times as much as it actually warms. Dierent rules have better consequences at those worlds. In this paper's terms, another set of rules o 2 would be the moral standard at those worlds. But we 10 Set aside derivative knowledge, like knowledge by testimony. I'm here focusing on non-derivative, direct knowledge of what an set of rules like o1 forbids. 254 might well continue forming our moral beliefs in the same way, and accepting that o 1 . It doesn't seem like we would have safe beliefs about the moral standard. The positivist can concede this point, because she doesn't take our moral knowl- edge to depend on safe beliefs about the moral standard. Moral knowledge only depends on safe beliefs de re, about the set of rules that are the actual moral stan- dard. And { crucially! { those beliefs are safe. Go to the counterfactual world where the globe warms much more than it actually does. Now Jane is still forming her beliefs about o 1 in the same way at that world, by relying on what she memorized. When we're asking whether some belief is safe, we're only comparing worlds where it's formed in the same way it actually is. Now memorization is a reliable (if not infallible) guide to o 1 . Ordinary cases where Jane relies on what she memorizes are cases where she has true beliefs about o 1 . Those beliefs are no longer about the set of rules that are the moral standard, but they're still true beliefs about o 1 . Since those beliefs are true throughout these similar worlds, they're safe. So Jane still knows. It's thus important to distinguish two ingredients in the positivist's conception of moral knowledge: knowledge de re of some set of rules, plus acceptance of what's in fact the moral standard as the moral standard. But the two ingredients are modally independent of each other. It's possible to have the rst without having the second. Nearby worlds where dierent sets of rules are the moral standard might be important for testing for the rst ingredient. In asking whether Jane continues to have true beliefs about o 1 across those counterfactual worlds, we're exploring whether she in fact has the de re knowledge that is the rst ingredient of moral knowledge. I've been explaining why her beliefs about o 1 are safe across that range of worlds, to defend the positivist's claim that she does have de re knowledge about o 1 . But those counterfactual worlds are irrelevant for the second ingredient, acceptance of what's in fact the moral standard as the moral standard. Moral knowledge at the actual world depends only on her having the second ingredient of moral knowledge at the actual world. It's totally irrelevant whether she also has that ingredient at counterfactual worlds. The positivist proposes a radical shift in our theorizing about moral knowledge, as noted in earlier chapters. And this radical shift has the right structure to solve the problem that arises for the rule consequentialist. 13.4 The nature of the positivist's proposal The last section started to explain why the positivist rejects the Baseline General- ization that powered the arguments inxx1{2. But I skated over some very subtle points. I'll now work through them. 13.4.1 Rough and ready knowledge x2 developed a heuristic, to distinguish the cases where ordinary moral knowledge requires rough and ready knowledge of the consequences. It compared an inference 255 about scientic knowledge with an inference about moral knowledge. (S1) Anne knows that this glass contains water. (S2) If Anne gained testimonial knowledge that that water is H 2 O, she would be in a position to know that this glass contains H 2 O. (D1) Anne knows that she may sacrice someone's toe to save her child's life. (D2) If Anne gained testimonial knowledge that rule consequentialism is true, she is in a position to know that the rules that allow you to sacri- ce a toe to save anyone's life have the best consequences. (S1) usually supports (S2): (S2) is usually true if and because (S1) is true. And the challenge from xwas if (D1) supports (D2) in the same kind of way. If it does, ordinary moral knowledge depends on rough and ready knowledge of the consequences. The positivist predicts that (D1) never supports (D2). She associates (D1) with two propositions: At-issue: Jane knows that o 1 permits sacricing some- one's toe to save her child's life, and accepts that o 1 is the moral standard Not-at-issue: o 1 is the moral standard The only thing that follows from these two propositions plus the supposition that Anne knows that rule consequentialism is true is a conclusion about what she would accept. (D0) If Anne gained testimonial knowledge that rule consequentialism is true, she would accept that the rules that allow you to sacrice a toe to save anyone's life have the best consequences. Knowledge plus acceptance in, only acceptance out. And the objectionable condi- tional (D2) does not follow from this benign conditional. Accepting some propo- sition doesn't requiring being in a position to know it. Moreover, the conditional (D0) is plausible. If you're accepting that you can sacrice the toe, and you know that rule consequentialism is correct, you had better accept that sacricing the toe is ne by rule consequentialist standards! Moral knowledge has a very dierent prole than scientic knowledge, because moral knowledge has a distinctive hybrid structure that physical knowledge does not. (D1) Anne knows that she may sacrice someone's toe to save her child's life. 256 Anne knows that o 1 allows sacricing someone's toe. Anne accepts that o 1 is the moral standard. (S1) Anne knows that this glass contains water. Anne knows that this glass contains water. Ordinary physical knowledge involves knowledge about the property that the the scientist is trying to describe. Ordinary moral knowledge, by contrast, does not involve knowledge about the property that the rule consequentialist is trying to describe. Chapter 11 was concerned to defend a de ationary vindication of normative theorizing. Normative theorizing still makes sense even if you don't have any reliable way of coming to accept truths about the moral standard. This chapter is exhibiting one virtue of that de ationary vindication of normative theorizing. It makes it possible for rule consequentialists to defend their normative claims as the outcome of re ective equilibrium, without supposing that the ordinary folk have rough and ready knowledge about the consequences. 13.4.2 What is the aim of normative inquiry, again? In order for this response to work, though, the positivist has to think of the aims of normative theorizing in a very distinctive way. And she has to reject a very natural picture of normative theorizing. It's initially natural to expect normative theorizing to aim at establishing bicon- ditionals connecting moral facts to the facts ground the moral facts. For example, that picture would take the rule consequentialist to be trying to show that8 [I ought to i -ing is demanded by the rules with the best consequences]. The positivist has to reject this picture. If it were right, the positivist would still pre- dict that ordinary moral knowledge depends on rough and ready knowledge of the consequences. Here's why. Suppose that you know that rule consequentialism is true and you know that you ought to keep this promise. Then you would be in a position to know that keeping this promise is demanded by the rules with the best consequences. The inference is straightforward. I know that (8 [I ought to i-ing is demanded by the rules with the best consequences]). I know that I may sacrice someone's toe to save my child's life.. Modus ponens allows me to infer that sacricing the toe is permitted by the rules with the best consequences. If I make that inference, the resulting belief would normally be a piece of knowledge. And the positivist would predict that ordinary knowledge that (D1*) puts me in a position to know that sacricing the toe is permitted by the rules with the best consequences. 257 (D1*) I know that I may sacrice someone's toe to save my child's life. (D2) If I gained testimonial knowledge that rule consequentialism is true, I would be in a position to know that the rules that allow me to sacrice a toe to save anyone's life have the best consequences. Then (D1*) depends on rough and ready knowledge about the consequences. The same problem arises all over again. The positivist does not accept the picture of normative theorizing that powers this objection. Instead, she understands normative theorizing as aimed at estab- lishing theses about the fundamental realist property. (Rule Consequentialism) [8x [x is the moral standard i x is the set of rules whose embedding has the best consequences] Testimonial knowledge of Rule Consequentialism does not involve knowledge of the earlier biconditional. If I know that I ought to keep this promise, I know that o 1 requires keeping this promise, and accept that o 1 is the moral standard. Even if I know Rule Consequentialism, nothing follows about my knowledge of the long-term consequences of keeping the promise. Knowledge plus acceptance in, only acceptance out. Now if I know Consequentialism, I will accept the earlier biconditional. I will accept that I ought to i -ing is the action demanded by the best rules. But any rule consequentialist worth her salt had better accept that biconditional anyway, and be disposed to change her views about what she ought to do if she comes to accept that the rules with the best consequences don't require that action. So the positivist does need to think of the aim of normative theorizing in a distinctive way. But Chapter 12 has already emphasized that she does think of normative ethics in a distinctive way. There are two possible ways of thinking about normative theses like rule consequentialism within the positivist's framework. 8 [I ought to i -ing is demanded by the rules with the best con- sequences]. Set-of-Rules Conception (Rule Consequentialism) At-issue: [8x [o 1 requires-ing i-ing demanded by the rules with the best consequences] Not-at-issue: o 1 is the moral standard. Property Conception (Rule Consequentialism) [8x [x is the moral standard i x is the set of rules whose embedding has the best consequences] 258 The rst conception (the Set-of-Rules Conception) takes normative ethics to aim at unpacking the speaker's idiosyncratic set of rules. And that conception of normative theorizing is a poor t for the way that normative ethicists tend to think of themselves. They think they're doing more than just describing the idiosyncratic set of moral beliefs that they happen to have acquired throughout their life. That's sociology, not normative ethics. It's the second conception (the Property Conception) that ts their self-conception. 11 13.5 Rule consequentialists should be positivists This chapter opened by mentioning three kinds of considerations that might be defeaters for an intuition that p. 1. My socio-economic position explains why I have the intuition that p. 2. I'm ignorant of the non-normative facts that bear on whether p. 3. My evolutionary history explains why I have the intuition that p. The rst consideration is denitely a defeater. But it's unclear if the second two considerations are defeaters, and it's unclear when they are. It's worth getting clear on whether the second kind of consideration is a defeater. It's very hard to defend rule consequentialism if it is. If it is, the rule consequentialist wouldn't have enough undefeated intuitions to defend her theory as the most adequate theory. Positivist realism makes a big dierence. Traditional conceptions of re ective equilibrium have very dierent upshots for this question. Ignorance of the non- normative facts is a defeater given those traditional conceptions. And it is not a defeater given the positivist's conception. In the rst half of this chapter, I argued that traditional conceptions should take ignorance of the non-normative facts to be a defeater. Then I've explained why that sort of ignorance won't be a defeater if the positivist is right. The upshot is that the positivist makes it possible to defend rule consequentialism as the outcome of re ective equilibrium, and that traditional conceptions of re ective equilibrium don't. In fact, though, I think that something stronger is true. Rule consequentialists need to be positivists in order to defend their view as the outcome of re ective equilibrium. Go back to the pair of inferences discussed earlier. (S1) Anne knows that this glass contains water. (S2) If Anne gained testimonial knowledge that that water is H 2 O, she would be in a position to know that this glass contains H 2 O. 11 Seex1 of chapter 10 for a more detailed discussion of this claim. 259 (D1) Anne knows that she may sacrice someone's toe to save her child's life. (D2) If Anne gained testimonial knowledge that rule consequentialism is true, she is in a position to know that the rules that allow you to sacri- ce a toe to save anyone's life have the best consequences. Let's recap, one last time. (S1) usually supports (S2): (S2) can usually be true because (S1) is true. Does (D1) also support (D2)? If it does, ordinary moral knowledge depends on being in a rough and ready position to know about the consequences. But ordinary agents don't have rough and ready knowledge about the consequences. So many of our moral beliefs would be defeated if (D1) supports (D2). In order to defend rule consequentialism as the outcome of re ective equilib- rium, then, you have to hold that (D1) does not support (D2). So you have to nd some dierence between the (D1)/(D2) pair and the (S1)/(S2) pair. Now it's easy to get confused about this point, because of a complicated feature of the re- lationship between (S1) and (S2). (S1) doesn't always support (S2). Suppose that Anne's knowing that this glass contains water is somehow incompatible with her knowing that water is H 2 O. In that very unusual case, (S1) wouldn't support (S2). A confused rule consequentialist might deny that (D1) supports (D2) for the same reason: knowing that rule consequentialism is incompatible with ordinary moral knowledge. But that response itself prevents the consequentialist from defending her theory as the outcome of re ective equilibrium, as noted in more detail earlier. To use this method, then, the rule consequentialist has to somehow deny that (D1) supports (D2). And her options for doing so are very tightly constrained. Suppose that (S1) supports (S2) because of general facts about the transmission of knowledge. Weak Transmission If you know that a is F, and that knowledge is compatible with knowing that all Fs are Gs, then you're in a position to know that a is G if you come to know that all Fs are Gs. GivenWeak Transmission, (D1) has to support (D2) as long as moral knowledge require knowledge about the realist's fundamental normative property. Given that requirement, (D1) involves knowledge of the fundamental realist property. In par- ticular, it involves knowing that an action a has that property F. (a = sacricing a toe, and F = being-the-moral-standard.) And rule consequentialism is a thesis about that fundamental property: that all Fs are Gs. If (D1) is true, Anne is then in a position to know that a is G if she comes to know that all Fs are Gs. And that's just the objectionable conclusion (D2). GivenWeak Transmission, the rule consequentialist can't defend her view as the outcome of re ective equilibrium unless moral knowledge swings free of knowl- edge of the realist's fundamental property. And the only way for a realist to allow moral knowledge to swing free of that kind of knowledge is by adopting the pos- itivist's hybrid account of knowledge. So the rule consequentialist who accepts 260 Weak Transmission needs to be a positivist. Moreover, it's hard to see a princi- pled account of the transmission of knowledge where (S1) usually supports (S2) but (D1) does not support (D2). That is, it's hard to see a principled account where Weak Transmission is false. It's hard enough that we should see positing such an account as obscurantism about the transmission of knowledge. In general, the rule consequentialist has only three options. Does the (S1) to (S2) inference hold generally? Do (S1) and (D1) have the same structure? Obscurantism about the transmission of knowledge (D2) is true if (D1) is true Positivist realism YES no yes NO The leftmost option makes ordinary moral knowledge depend on rough and ready knowledge of the consequences. xx1{2 have argued that it's impossible to defend rule consequentialism as the outcome of re ective equilibrium given that sort of de- pendence. And the rightmost option makes rule consequentialist hostage to debates in general epistemology, about the transmission of the knowledge. The best way for a rule consequentialists to defend their view as the outcome of re ective equilibrium is to accept positivist realism. 261 Chapter 14: The normative signicance of sociobi- ology In this nal portion of the dissertation, we're exploring some upshots that positivist realism would have for normative theorizing. We're focusing in particular on what sorts of considerations might be defeaters for a moral intuition { that is, consid- erations that make it a mistake to rely on that intuition. It denitely seems like some considerations are defeaters. For example, discovering that my intuition that p is the result of my socio-economic position defeats that intuition. But it's unclear what range of considerations pattern with this example. Consider this list. 1. My socio-economic position explains why I have the intuition that p. 2. I'm ignorant of the non-normative facts that bear on whether p. 3. My evolutionary history explains why I have the intuition that p. It's very unclear if the nal two considerations are defeaters for our intuitions, and when they are. De ationary positivism illuminates these questions. We saw in the last chapter that ignorance of the non-normative facts is a much less signicant defeater if de ationary positivism is true than if it's false. This chapter focuses on the third kind of possible defeater: facts about my evolutionary history. This sort of defeater is much more signicant for de ationary positivists than for her more traditionalist rivals. Positivist realism has very dier- ent upshots for type-2 defeater than for type-3 defeaters. Type-2 defeaters aren't genuine, while type-3 defeaters are. It shouldn't be surprising that facts about our evolutionary history are defeaters, at least in some cases. Some people are disposed to believe that a range of sex are acts morally wrong, because they are \impure", in some sense. There are powerful evolutionary explanations of these dispositions. 1 I think that these evolutionary explanations are defeaters for those beliefs. Given those evolutionary explanations, it's a mistake to rely on these beliefs as evidence in normative ethics. 2 We can 1 For relevant discussion, see Rozin et al. (1993), and the literature discussed there. 2 Jesse Prinz gives an example of the sort of argument that I have in mind: \by labeling certain sexual acts as unnatural, Church leaders may have been able to tap into parishioners' biological dispositions to disgust. Disgust obeys a logic of contamination. If sex is construed as dirty, then sex with multiple partners can be seen as a way to spread 262 reasonably expect some explanation of why it would be a mistake to take these beliefs as a source of evidence. The positivist has a distinctive way of vindicating this sort of argument. And her vindication illuminates other important questions about this sort of argument. One important question is about the scope of these arguments. It's not just philosophers with a liberal view about sex who appeal to this sort of argument. Consequentialists also appeal to this sort of argument, in trying to disarm some central objections. (They might try to debunk intuitive beliefs about the signicance between doing and allowing, or about the scope of benecence.) Within the positivist's framework, this sort of argument is just as legitimate as the liberal's argument. In other words: the positivist has a very expansive account of the scope of this sort of argument { just the sort of view that amplies their signicance. 14.1 A central problem A range of normative ethicists have appealed debunking arguments. An act con- sequentialists like Peter Singer might appeal to evolutionary explanations to un- dermine our intuitive beliefs about the moral signicance of the dierence between doing and allowing. Roger Crisp (2006) has given a debunking explanation of why we value accomplishment independently of pleasure. And Tim Mulgan (2006) has oered a debunking explanation of beliefs about the experience machine. Recent debunking arguments are particularly likely to appeal to evolutionary accounts. But scientic explanations in general raise the same kinds of issues. For example, Marx also oered debunking explanations of certain moral beliefs. 3 Though this explanation does not appeal to an evolutionary explanation, it raises many of the same issues that evolutionary debunking does. I'll mostly focus on arguments from our evolutionary history, but the discussion will generalize. Arguments from our evolutionary history are highly controversial in normative ethics. And they are highly controversial for very good reason. It is hard to make those arguments without thereby committing yourself to global debunking argu- ments { arguments that purport to show that none of our moral beliefs amount to knowledge. 4 But it's self-defeating for a normative theorist to make an argument that leads to this sort of global skepticism. Normative ethicists are trying to extend our moral beliefs. And global skepticism tends to destabilize belief. The task for impurity. By cultivating squeamish attitudes toward sexuality, Church leaders could have more easily transmitted norms against polygamy, divorce, and remarriage" (Prinz 2009, 233). 3 Interpreting these explanations as attempts at normative debunking requires inter- preting him as a moralizer, as Allen Wood (1972) does, and not as only a social scientist. 4 See Street (2006, 2008a, 2008b, ms) and Joyce (2006a,b). Street argues for the global conclusion only on a realist construal of the domain. She herself favors a constructivist construal of the domain, and so thinks that we do have moral knowledge. For other versions of this general kind of challenge, see Joshua Greene (2008), Phillip Kitcher (2007), Michael Ruse (1998), and Michael Ruse and E. O. Willson (1986). For discussion of the dierences between these challenges, see Russ Shafer-Landau (2012) and Katie Vavova (2014a). 263 proponents of local arguments from our evolutionary history is to prevent those arguments from motivating global arguments. Let's start with an example. I'm disposed to believe that it's wrong for a surgeon to cut up one to save ve. I'm also disposed to believe that believe that it's permissible to switch a trolley on its way to kill ve people to only kill one instead. This pair of beliefs seems like evidence against act consequentialism. If both are correct, it doesn't seem like the consequences determine what we ought to do. Peter Singer draws on work by Joshua Greene to give a debunking explanation of these beliefs. He suggests that we've evolved a strong aversion to immediately in icted violence, which is triggered when we imagine the surgeon cutting up the one. But we haven't evolved a strong aversion to more industrial killings, of the sort the trolley problem illustrates. Singer then complains: \what is the moral salience of the fact that I have killed someone in a way that was possible a million years ago, rather than in a way that became possible only two hundred years ago? I would answer: none" (Singer 2005, 348). He concludes that we shouldn't rely on this pair of beliefs as evidence. Singer is here making a debunking argument. He's trying to show that it's a mistake to rely on these intuitions in normative ethics. Now his argument may not show that the relevant intuitions are false. It's possible that our evolutionary history happens to line up with a signicant fault line in normative reality. But they do suggest that those intuitions have some epistemic defect that make them fall short of being knowledge. Maybe the evolutionary explanation shows that they're not justied, for example, or that they're unsafe, or something similar. Unfortunately, though, there is a slippery slope from local debunking to global debunking. Local normative debunking takes the evolutionary explanation of our moral beliefs to show that those beliefs are unjustied, or are unsafe, or lack some other feature that distinguishes knowledge from mere true belief. But all our moral beliefs will admit of evolutionary explanation, or at least admit of some sort of sci- entic explanation. If it's the fact that our beliefs have an evolutionary explanation that makes them unjustied (or unsafe, or ...), all our moral beliefs look unjustied (or unsafe, or...). And that's the claim that distinguishes the global debunker. To avoid this slippery slope, the local debunker needs to explain what distinguishes the moral beliefs she accepts as justied from the moral beliefs that she takes to be unjustied. Guy Kahane puts this point sharply. Until the [global evolutionary debunking argument (EDA)] argument has been defused, [philosophers like Singer] cannot deploy local EDAs in normative ethics with a clear conscience. And even if the global argument can be resisted, it might be resisted in a way that leaves no space for EDAs of any kind (Kahane 2011, 117) Kahane's point holds even if he hasn't show that local debunking must collapse into global debunking. The burden of proof seems to be on the local, normative debunker, to distinguish what she's doing from what the global debunker is doing. 264 The local debunker appears to have only one option for distinguishing what she's doing from what the global debunker is doing. She has to draw what I'll call invidious distinctions among dierent evolutionary explanations of our moral beliefs. In drawing those invidious distinctions, she would be identifying some evolutionary explanations as undercutting the epistemic credentials of the beliefs, while identifying other evolutionary explanations as not undercutting the epistemic credentials of the beliefs. Peter Singer agrees. He admits that his argument requires distinguishing a special class of moral beliefs (those with a \rational basis"), which are formed in a dierent way than we form the beliefs that are problematic for his act consequentialism. He concedes that \even to specify in what sense a moral belief can have a rational basis is not easy. Nevertheless, it seems to me worth attempting, for it is the only way to avoid moral skepticism" (Singer 2005, 351). Then it seems like the legitimacy of local debunking depends on invidious dis- tinctions between dierent kinds of evolutionary explanations. If you're skeptical that such distinctions can be found, you should be skeptical of local debunking ar- guments. Moreover, it seems like proponents of local debunking arguments owe us these sorts of invidious distinctions. We should apportion our condence in their arguments to match our condence that the invidious distinctions do exist { so we shouldn't be at all condent in a local debunker unless we're condent in the relevant invidious distinction. 14.2 Evolutionary arguments in normative ethics, without invid- ious distinctions I think evolutionary explanations of our moral beliefs can have the upshots that Singer cares about. Moreover, we don't need to draw invidious distinctions among evolutionary explanations in order for them to have those upshots. That is, I'll defend the legitimacy of what Singer is doing, but in a dierent and less committal way. The debunking arguments mentioned in the last section purport to show that some moral belief is epistemically awed and falls short of knowledge in some way. I'll reserve the label debunking arguments for arguments that try to establish this sort of conclusion. But philosophers who make local debunking arguments really care about showing that those beliefs are no evidence against some normative view. I'll call an argument for the latter conclusion a no evidence argument. Debunking arguments are one kind of no evidence argument. But I'll suggest that they're not the only kind. Let's start with a concrete example. John Finnis (1980) holds that there is a good internal to heterosexual marriage that isn't available outside of it. (1) Finnis believes that that there is a good internal to heterosexual marriage that isn't available outside of it. And he thinks that he apprehends this good in an epistemically basic way, so that he is justied in so believing. So he thinks that (2) is true. 265 (2) Finnis is justied in believing that that there is a good internal to heterosexual marriage that isn't available outside of it. He relies on this claim in arguments about sexual morality. For example, he argues that homosexual sex is immoral, because they are \against" this good. Finns' belief is just the sort of belief that is ripe for evolutionary explanation. That belief may well be an adaptively advantageous. Communities that believe that there is a good internal to heterosexual marriage that's unavailable outside of it may well be more likely to reproduce than communities that lack that belief. 5 After all, moral judgments often have practical upshots. So a community that takes heterosexual marriage to be a good that's otherwise unavailable may well include more people in heterosexual unions than there otherwise would be. Another kind of evolutionary explanation might appeal to subconscious factors. Maybe the best explanation of Finnis' belief is that he subconsciously thinks that a range of sex are acts are \impure", in some sense. There are powerful evolutionary explanations of these dispositions. 6 It seems like these evolutionary explanations bear on Finnis' claims. Why? Debunkers have one answer. They try to show that he is not justied in believing as he does { that the evolutionary explanation somehow defeats his justication. In doing that, they will claim that (2) is false.It is this attempt that leads the problem in the last section. They need to explain how any of us are justied in believing as we do. This chapter will introduce a quite dierent picture. This picture grants that Finnis is justied in believing as he does. That is, it grants that (2) is true. But it insists that Finnis' justied belief is no evidence for the theses he accepts in normative ethics. The picture is thus very weird. It grants that Finnis can have justied beliefs about a domain, while in the very same breath denying that he can legitimately rely on those beliefs when theorizing about the domain. This initially weird picture has one crucial advantage over debunking arguments. It does not need to answer Kahane's challenge. It does not take evolutionary ar- guments to undermine Finnis' justication in believing as he does. So it doesn't threat the justication of any moral beliefs. That's the reason this picture doesn't lead to global debunking; it's just not a species of debunking at all. 7 Despite its 5 If you're worried that this belief is too intellectual to be widespread, then you're worried about Finnis' view. He does think that ordinary agents tacitly apprehend the good of heterosexual marriage. 6 For relevant discussion, see Rozin et al. (1993), and the literature discussed there. Jesse Prinz gives an example of the sort of argument that I have in mind: \by labeling certain sexual acts as unnatural, Church leaders may have been able to tap into parish- ioners' biological dispositions to disgust. Disgust obeys a logic of contamination. If sex is construed as dirty, then sex with multiple partners can be seen as a way to spread impurity. By cultivating squeamish attitudes toward sexuality, Church leaders could have more easily transmitted norms against polygamy, divorce, and remarriage" (Prinz 2009, 233). 7 In fact, this picture can even concede that Finnis would know what he believes if what he believes had been true. 266 weirdness, then, this kind of argument would be dramatically less costly than de- bunking arguments than debunking arguments. It would work without drawing any invidious distinctions among evolutionary explanations. That's the structure I want: a picture that explains why Finnis' belief is no evidence for his normative theory, even though it's justied. And it is possible to develop a theory with this structure. Compare the role of observations in scientic explanation. If an observation o is more likely on theory T 1 than T 2 , then o is some evidence for T 1 and some evidence against T 2 . But if o is equally likely on T 1 as on T 2 , then o does not favor one theory over the other. One way to defend a theory against a recalcitrant observation is to show that that recalcitrant observation is in fact as likely on that theory as on rival theories. Take a scientist { Sue { who observes that p by using some method M. Un- fortunately, though, Sue's own theory predicts that:p. Even worse, her enemy's theory predicts that p. Her observation then looks like evidence against her theory and evidence for her opponent. But she can disarm this evidence by explaining why that observation is just as likely on her theory as on her opponent's theory. One way to do that that would be for her to argue that method M will misre if her theory is true. This argument is not a way of explaining the content of the recalcitrant observation. It's rather a way of explaining why the fact that we've made that observation is unsurprising. Sue's approach is the model for the rest of the paper. Go back to Finnis. He thinks that there's a good internal to heterosexual marriage that's unavailable outside it. Someone with a liberal view about sex will disagree. That liberal should model her answer to Finnis on what Sue says. Even if the liberal's view about sex is right, we would expect people to have the beliefs that Finnis has. We're assuming in this chapter that Finnis' belief does have some evolutionary explanation. So the liberal can help herself to that evolutionary explanation. She should conjoin her normative view about sex with that evolutionary explanation. It as likely that Finnis would have his belief on that conjunction as it is if Finnis' own normative view is correct. Then the liberal is in exactly the same position that Sue is in. Finnis' belief is no evidence against the liberal, no more than the recalcitrant observation is evidence against Sue. I will call this kind of argument a no-evidence argument. It's a way for a normative ethicist to show that some belief is no evidence against her theory. The rest of this paper will develop and defend this kind of argument. Before leaving Sue behind, though, we can learn one more thing from her. My no-evidence argument allows that Finnis is justied in his belief, while insisting that that belief is no evidence for his normative view. I noted that that possibility looks very weird. If you're justied in believing p, surely p can be some evidence for your theory! However, Sue's case seems to involve the same sort of weirdness. It's very natural to think that her opponent is justied in believing p. It's what his theory predicts, and it's what has been observed. Maybe he even knows his theory! If he does, and if he knows that his theory entails p, he's plausibly in a position to know that p. As an outsider to this dispute, then, I can take him to know that p. And I can take him to know even while acknowledging that Sue is right that the 267 observation that p is no evidence for his theory over her theory. In other words: Sue's opponent can have a justied belief that is no evidence for his theory. Sue's case involves just the weirdness that my no-evidence argument involves. You might nd this possibility particularly weird if you get confused about a piece of terminology. Williamson (2000) holds that something is a piece of evidence i it's a piece of knowledge. This biconditional is very strong. It's weaker and more plausible to hold that, if something is a piece of knowledge, it's a piece of evidence. This paper is not using the word evidence in the way that Williamson is using it, or in the way that the weaker conditional uses the word. James Joyce (2004) helpfully distinguishes between what he \calls a theory of evidential relevance, which seeks to determine whether and how much \e speaks in favour of h", and a theory of evidential status, which species the sort of \creditable standing" that e must have in order to count as evidence for anything" (Joyce 2004, 296) 8 This paper is about the theory of evidential relevance. It is not about the theory of evidential status, at all. Sue's claim is that the observation that p does not speak in favor of her opponent's theory. That is, her claim is a claim about evidential relevance. We can grant this claim. And we can do that even while we acknowledge that her opponent has a justied belief and indeed knowledge that p. If we accept a Williamsonian theory about evidential status, we would then grant that the opponent's has a perfect epistemic status. The opponent knows p, so p is a piece of evidence that he has. My no-evidence argument is modeled on the Sue example. So it attempts to establish a conclusion about evidential relevance, not about evidential status. The conclusion is that Finnis' belief in the integral good of heterosexual marriage has no evidential relevance to work in normative ethics. I'll continue to use the word evidence to talk about evidential relevance. There are other ways to use the word, too, but I'm setting those aside. 14.3 Wait { what are the data for normative ethics? Let's recap. I've outlined a very dierent way of thinking about the relevance of evolutionary explanations for normative ethics. Those explanations don't threaten the justication of the targeted beliefs. They rather threaten the evidential rele- vance of those beliefs. But note this outline doesn't t the promises I made at the beginning of the paper. I promised to show how positivist realism supports a new way of thinking about the normative relevance of evolutionary explanations. And my no-evidence argument never even mentioned positivist realism! The devil is in the details. There is an orthodox assumption about norma- tive ethics that is widely accepted. And my no-evidence argument fails if that assumption is true. This assumption is, I think, the central reason that philoso- phers haven't explored my no-evidence argument. Now positivist realism cleanly explains why that orthodox assumption is false. So it's important in this paper because it disarms the central reason for discounting the no-evidence argument. 8 The quotes embedded in this passage are from p. 186 of Williamson (2000) . 268 In other words, I can introduce my no-evidence argument without mentioning the positivist. Positivism comes in when we work out the details. The orthodox assumption is about the data in normative ethics. The last section talked as if the data are beliefs. But talk about \belief" is ambitious. It's ambiguous between talk about the state of belief and talk about the object of belief. Take the phrase Finnis' belief that there's a good internal to heterosexual marriage that's unavailable outside it. That phrase can refer either to the proposition believed (that there's a good internal...), or to a mental state (the state of believing that proposition). I've been assuming that the data to be explained are about my mental states { that we need to explain why I have that belief. However, the orthodox assumption in normative ethics is that the data are the propositions believed. The best normative theory is the one that best ts those propositions into a more coherent whole. Je McMahan is characteristic of a wide range of normative ethicists on this point. He characterizes himself as accepting a picture that takes \certain substantive moral beliefs that we already have as reliable starting points for moral inquiry. It presupposes that at least some of our moral intuitions have a certain prima facie normative authority" (McMahan 2000, 93). The locus of normative authority is, for him, the propositional objects of our intuitions. My no-evidence argument fails on this orthodox conception of normative ethics. Finnis believes that there's a good internal to heterosexual marriage that's unavail- able outside it. Our evolutionary history is just irrelevant for whether that good exists. The existence of that good depends on facts about what's practically rea- sonable, on his theory. And our evolutionary history just has no bearing on what is practically reasonable. The same things will always be practically reasonable for creatures like us, whatever our evolutionary backstory. As a result, facts about our evolutionary history are just irrelevant. So if the observation to be explained is the proposition that there's a good internal to heterosexual marriage that's unavailable outside it, evolution doesn't help the liberal at all. It only helps if the observation to be explained is that Finnis believes that proposition. As a result, my no-evidence argument won't work given the orthodox conception of normative ethics. The positivist rejects the orthodox assumption that the data are the proposi- tions believed. She instead conceives of the data in the stative way: the evidence is that we have certain mental states. For the positivist, moral beliefs involve beliefs about our own set of rules. We do regard our moral beliefs as reliable { but the sense in which they're reliable is that they're reliable guides to our own set of rules. But normative ethics doesn't aim at articulating our own idiosyncratic set of rules. That's sociology, not philosophy. Rather, normative ethics aims at articulating the nature of the fundamental property being-the-moral-standard. And there is no guar- antee that we have reliable means of forming beliefs about that property. So the positivist denies that particular moral beliefs are reliable starting points for inquiry. And she denies that they have prima facie normative authority. However, normative ethics still makes sense on this picture. We're accepting our set of rules as the moral standard. Because we do that, we can intelligibly treat our own set of rules as a source of evidence about the moral standard. In 269 fact, something stronger is true. Someone who isn't willing to treat her own set of rules as a source of evidence doesn't really accept her own set of rules as the moral standard. If she did accept it as the moral standard, she would be willing to treat it as a source of evidence. But { crucially! { this picture is a picture where the starting points of normative ethics are our psychological states. It's because we're in certain psychological states that we do treat our own set of rules as evidence. 9 The no-evidence argument makes perfect sense on the positivist's picture. And it doesn't make any sense on the orthodox picture. Positivist realism is an elegant way of making sense of the no-evidence argument. It explains why the data for nor- mative ethics might be facts about our mental states, rather than the propositions believed. Now note that, if the positivist is right, there are two options in explaining a moral belief that doesn't t some normative theory. You can rene the norma- tive theory, readjusting your proposal about the property being-the-moral-standard. This rst option is what almost always goes on in normative ethics. (Frances Kamm (2006) is an extreme example.) The second option is to explain why it's unsurprising that we have that belief, even if your normative theory is true. The two responses are perfectly on a par, for the positivist. The parity between the two responses is how the positivist makes sense of my no-evidence argument. That argument just involves taking the second option, rather than the rst. 14.4 Does the no-evidence argument discriminate between nor- mative theories? You might now feel like you've lost your bearings. I've claimed that there are two equally good options for explaining a recalcitrant moral belief: (i) revise your normative theory, or (ii) explain the belief away. And I take scientic explanations of the belief to be perfectly good ways of doing the second. But there are scientic explanations for all our moral beliefs! Then you might argue that the two options can't really be on par. If they are, any normative theory whatsoever can help itself to my no-evidence argument. And that leaves normative ethics at an utter impasse. Any time a normative theorist faced some problematic belief, she could just appeal to my no-evidence argument. Since we can made progress in normative ethics, the positivist's two options for explaining a moral belief just can't be on a par. I agree that we can make progress in normative ethics. And I agree that there are scientic explanations of all our moral beliefs. But I deny that my no-evidence argument is equally helpful for all normative theories. We should care about several theoretical virtues in our normative theorizing. It does seem like delity to our antecedent moral beliefs is a virtue. But there are other theoretical virtues as well. Simplicity is an example. We should prefer a simpler theory to a more complex theory, even if the two theories are equally faithful to our antecedent moral beliefs. 9 See Chapter 12 for a more detailed discussion of this point. 270 And dierent moral theories prioritize dierent theoretical virtues. Some theo- ries trade delity for simplicity. Act consequentialists are a paradigmatic example. Other theories trade simplicity for delity. Rossian pluralists that posit an irre- ducible plurality of grounds of reasons are a paradigmatic example. It's helpful for my purposes to think of normative theories as grouped by their simplicity. Start with the class of normative theories that give the simplest explanations of our moral beliefs, without regard to plausibility. This class might include objective act con- sequentialism and egoism. And end with the class of normative theories that give the least simple explanations of our moral beliefs, without regard to plausibility. This class might include particularist theories, of the sort that Jonathan Dancy has defended. My no-evidence argument decreases the importance of delity to our antecedent moral beliefs, and increases the importance of other theoretical virtues. After all, the argument explains why it wouldn't be surprising if the correct normative theory lay at a distance at a distance from some of our ordinary convictions. It's then less costly to suppose that it does. In other words, my no-evidence argument does not help in discriminating within these classes of equally simple theories. Crucially, though, it does help discriminate among these classes. In particular, my no-evidence argument makes it less costly to trade delity to ordinary moral belief for simplicity. There might be other dimensions that matter in evaluating normative theories, in addition to the simplicity of the dierent theories. For one thing, we might want to identify a property that creatures like us can intelligibly care about. 10 My no-evidence argument has similar upshots for those dimensions. They are more important given my argument than they would otherwise be. I can now state the full version of my no-evidence argument. Go back to Fin- nis' belief that there's a good internal to heterosexual marriage that's unavailable outside it. He accepts a normative theory where that belief is true. There are lots of more liberal views about sexual morality where that belief is false. Consider, for example, a liberal view where consent is the only thing that matters for sex. Finnis' view is more complicated: it takes consent to matter, but also takes the partner's gender to matter, and also takes the partner's marital status to matter. The liberal view is simpler, and that fact counts in its favor. Finnis thinks that delity to his antecedent belief is more important than simplicity, at least in this case. So he favors his normative theory over a more liberal alternative. The no-evidence argument undercuts the importance of delity to his antecedent belief. A No-Evidence argument against Finnis 1. There is an evolutionary explanation of Finnis' belief. 2. Given that evolutionary explanation, Finnis is as likely to have his belief if the liberal theory is true as if it's false. 3. If some belief is equally likely on either of two theories, then that belief does not favor one theory over the other. 10 This desire plays a role, I think, in the arguments in Scanlon, 1998). 271 4. So Finnis' belief does not favor his normative theory over the liberal theory. 5. Finnis' normative theory doesn't have any theoretical virtues, over the liberal theory. 6. The liberal theory has theoretical virtues that Finnis' normative theory does not; it's simpler. 7. Since we should prefer a theory with more theoretical virtues, we should prefer the liberal theory to Finnis' theory. This no-evidence argument thus has two key parts. It rst uses a scientic expla- nation to eliminate the virtue that Finnis claims for his theory. Then it uses the theoretical virtues of the liberal theory to break the parity between Finnis' theory and the liberal theory in favor of the liberal theory. This argument is still compatible with Finnis being justied in his belief. So (2) is still true, as long as he remains unpersuaded by the argument. (2) Finnis is (propositionally) justied in believing that that there is a good internal to heterosexual marriage that isn't available outside of it. The positivist allows that (2) is true inasmuch as Finnis is justied in his beliefs about his own set of rules: he's justied in thinking that heterosexual marriage is good according to o F . Now the no-evidence argument purports to show that it's a mistake to believe as Finnis does. If he stops making that mistake, he will switch what set of rules he accepts to be the moral standard. And if he switches what set of rules he accepts to be the moral standard, (2) will no longer be true. Call the new set of rules o N . He's not propositionally justied in believing that heterosexual marriage is good according to o N . The dispositions that would allow him to think about o N don't justify him in believing that proposition. By contrast, the dispositions that allow him to think about o F do justify him in believing that heterosexual marriage is good according to o F . The positivist thus predicts that (2) is true now. She makes that prediction even though (2) wouldn't be true if Finnis had true moral beliefs. 14.5 Isn't the simplest normative theory highly implausible? You might still feel like you've lost your bearings. There is a maximally simple normative theory: Permissivism. Permissivism: the moral standard permits everything. Every other coherent normative theory is more complicated. Moreover, it looks like the argument from the last section should favor Permissivism over every rival. It's even simpler than act consequentialism, for example. And the Permissivist 272 can leverage my no-evidence argument against the act consequentialist. The fact that act consequentialism coincides with our judgments about wrongness or about obligation is no evidence in its favor. It's not surprising that we make those judg- ments, even if Permissivism is true. In each case, there will be some evolutionary explanation of the judgment. 11 Permissivism is such an implausible conclusion that it constitutes a reductio of any argument that leads to it. For one thing, it may not even even be possible for creatures like us to accept Permissivism in a stable way. Chapter 12 emphasized an important tie between our reactive attitudes and our judgments about wrongness { that those reactive attitudes seem dependent on our judgments about what's wrong. More carefully, it emphasized an important tie between the rationality of our reactive attitudes and those beliefs. But believing that an attitude is irrational tends to do destabilize the attitude. Coming to accept Permissivism would plausibly destabilize our reactive attitudes. And our reactive attitudes seem like an ineliminable part of our life together. If they are, it wouldn't be possible to stably accept Permissivism. Resentment and indignation would keep leading us to accept propositions about what the moral standard forbids, which is incompatible with Permissivism. And the positivist takes normative theory to aim at delivering propositions about the moral standard that we can accept in a stable and consistent way. I deny that my no-evidence argument supports Permissivism. In fact, I think that the objection in the last two paragraphs gave the basic reason why. It's not possible for creatures like to accept Permissivism in a stable and consistent way. And I claim that should only accept a normative theory that we can accept in a stable way. If that claim is true, we shouldn't accept Permissivism. This claim is a claim about what we should do. But what sense of `should' is it about? It can't be the `should' of moral obligation. For one thing, it's not clear why we would have that sort of moral obligation. For another, it would be dialectically ineective to construe the claim in that way. The Permissivist could just oer a no-evidence argument. Instead, I take the `should' is the `should' of constitution. What it is to accept a thesis in normative ethics is to accept it in a stable and consistent way. That is the only way to think of normative ethics given the positivist's picture. So the positivist is committed to regarding this claim as a constitutive fact about normative ethics. And this constitutive constraint is a lexical constraint on normative ethics. In defending a normative thesis, the rst thing that you need to check is whether ordinary agents can accept the thesis in a stable and consistent way. If we can't, your normative theory can't get o the ground. Lexical Constraint: You shouldn't accept a normative thesis unless you can accept it in a stable and consistent way. Once you've satised the Lexical Constraint, you can start worrying about other constraints, like Fidelity and Simplicity. 11 I'm grateful to Alex Dietz for making this point. 273 Fidelity: You should prefer a normative theory that is faithful to our antecedent moral beliefs. Simplicity: You should prefer a simpler normative theory. Now unlike the Lexical Constraint, these latter two `should's are not the `should's of constitution. They are instead the `should's of epistemic normativity. Fidelity, for example, is grounded in the requirement that our attitudes be consistent. Given the sort of creatures that we are, we will continue to accept propositions about the moral standard. Fidelity requires that those states be consistent with what normative theses we accept. My no-evidence argument is safe after all. It doesn't lead to unacceptable conclusions, like the conclusion that Permissivism is true. 14.6 Yay for consequentialists! I've introduced a way of thinking about the relevance of our evolutionary history for normative ethics. A normative ethicist can appeal to that history to reduce the importance of delity to our antecedent moral beliefs, and increase the importance of theoretical virtues like simplicity. I've called this kind of appeal my no-evidence argument. I suggest construing all appeals to our evolutionary history as oering a no- evidence argument. This construal is important, because some appeals to our evo- lutionary history are more controversial than others. I've been focusing on a liberal debunking argument, that targets some beliefs of John Finnis. It's also possible to give a no-evidence argument on behalf of consequentialists. One standard objection to consequentialists is that their view is implausibly de- manding { that it predicts obligations that are far more stringent than we in fact accept. 12 For example, many consequentialist views seem to predict that we have striking obligations to far-o strangers, inasmuch as resources that we normally use for ourselves could do much more good for them. This prediction is taken to be evidence against the consequentialist theory. My no-evidence argument crisply disarms this source of evidence. There are presumably good scientic explanations of why we would not accept striking obli- gations to far-o strangers. Maybe, for example, it wasn't adaptive for communities to accept that we have such obligations. A No-Evidence argument against Demandingness arguments 1. There is an evolutionary explanation of our belief that morality is comparatively undemanding. 2. Given that evolutionary explanation, we are as likely to have his belief if some consequentialist theory is true as if it's false. 12 For one helpful discussion, see Tim Mulgan (2001). 274 3. If some belief is equally likely on either of two theories, then that belief does not favor one theory over the other. 4. So our belief about morality's demands does not favor a non- consequentialist normative theory over a consequentialist theory. 5. The non-consequentialist doesn't have theoretical virtues, over the consequentialist theory. (The virtue that it claims { closer delity to our antecedent moral beliefs { is no virtue at all.) 6. The consequentialist theory has theoretical virtues that the non- consequentialist theory does not; it's simpler. 7. Since we should prefer a theory with more theoretical virtues, we should prefer the consequentialist theory. As before, the no-evidence argument has two parts. First, it undermines the impor- tance of delity to our antecedent moral beliefs. Second, it highlights the theoretical virtues of one alternative over the other. Remember that my no-evidence argument works without drawing invidious dis- tinctions among scientic explanations. It does not need to single out a class of scientic explanations as the ones that undermine justication, while explaining what distinguishes members of that class from scientic explanations that do not undermine justication. It's a big advantage of the no-evidence argument that it does without invidious distinctions. There's no need to worry that the no-evidence arguments will support global debunking arguments. Global debunking arguments try to show that moral beliefs lack some feature that distinguishes knowledge from mere true belief. And the no-evidence argument simply does not threaten those features. So it doesn't need to draw invidious distinctions, and mark some scientic explanations as threatening those features, and others as not. Go back to Finnis. Suppose that you're convinced that it's a mistake to use his antecedent beliefs about sexual morality as evidence in normative ethics, and you're convinced that it's a mistake because of the evolutionary history of Finnis' beliefs. I've developed an argument that makes sense of your conviction. Crucially, thought, this argument is equally available to consequentialists, in disarming objections to their view. In acknowledging the force of this argument against Finnis, you're committed to acknowledging its force against non-consequentialists. The shoe is then on the other foot. It is the non-consequentialist who owes us invidious distinctions among dierent kinds of evolutionary explanations. Or, more carefully, it is the non-consequentialist who wants to oer a debunking argument against Finnis who owes us those invidious distinctions. My no-evidence construal of the argument against Finnis is equally available to the consequentialist. If you accept that construal of the argument, you are pushed into the consequentialist's welcoming arms. Resisting that push means nding invidious distinctions among the dierent scientic explanations of our moral beliefs. And we should be skeptical that such distinctions can be found, absent some positive evidence for them. There are two coherent positions about scientic explanations of moral beliefs. One position is that they're never relevant for normative theorizing. They don't 275 threaten Finnis, and they don't threaten the non-consequentialist. The other is that they are always relevant. They threaten Finnis, and they threaten the non- consequentialist. The positivist takes the later position. If she's right, someone who takes scientic explanations to threaten Finnis should also take them to threaten the non-consequentialist. 14.7 Against trolleyology The positivist has a striking and highly controversial conception of the methods of normative ethics. It's a mistake to focus on particular intuitions about surgeons and trolleys, or about our intuitions about the limits of our obligations of benecence. If your normative theory can't account for those intuitions, you should just oer a no-evidence argument against them. Normative ethicists should focus instead on articulating the role of moral judg- ments in our life together, and try to gure out the most plausible candidate for what plays that role. I've noted that there is an important connection between our moral judgments and our reactive attitudes. And I've used that connection to eliminate a normative view: Permissivism. Let's review. Why can't the no-evidence argument undercut the connection between moral judgment and reactive attitudes? Answer: creatures like us can't reject that con- nection in a stable way. The no-evidence argument succeeds only when it is psy- chologically possible for us to accept the contrary view. For example, that's why the no-evidence argument works against Finnis. We can accept a liberal view about sex in a stable way. The positivist acknowledges two sources of evidence in normative ethics. Facts about the role of moral judgment in our lives together are the rst source of evi- dence. I've illustrated this kind of evidence with the reactive attitudes. But moral judgments play a range of roles in our life together. Any of those roles is an in- telligible source of evidence in normative ethics. Theoretical virtues like simplicity are another source of evidence. The fact that a theory is simpler is evidence in its favor. That's it. I am optimistic that a consequentialist view is the best way of making sense of these two kinds of evidence. In general the positivist does not acknowledge the evidential force of particular intuitions about surgeons, or about trolleys. She rejects trolleyology. I want to close this dissertation with some autobiography. I've had an overt agenda in writing the dissertation. I'm trying to convince you that a particular lin- guistic claim about presupposition is a central pivot in metaethics. If that linguistic claim is true, a range of problems in moral epistemology are pseudo-problems. And those epistemic problems have important upshots for debates in moral metaphysics and normative ethics. Those debates look very dierent given that particular claim about presupposition. I also have a covert agenda. I buy into a very non-standard conception of normative ethics. I think that normative ethics involves, as Allen Wood puts it, a method, which I nd not only in Kant but also in utilitarians such as 276 Bentham and Mill, that would draw the fundamental moral principle from very general and fundamental considerations about the nature of rational desire and action, and would then attempt to reconcile these principles with common moral opinions only insofar as those opinions can be seen as applications of the principles. (Wood 2011, 59) I nd a consequentialist account of the fundamental moral property more attractive than Wood does. But I agree with his methodological stance. I think it's a mistake to Chisholm away at the account of the fundamental moral principle to align it with common moral opinions. But this methodological stance faces a powerful challenge. It treats certain \very general and fundamental considerations" as genuine evidence, and does not treat common moral opinions as a source of evidence. It needs to justify that dierential treatment. One option is to take the general and fundamental considerations to be apprehended in a dierent way than common moral opinions. Maybe, for example, we know those general and fundamental considerations apriori, in some special and foundational way. I regard that option as extremely costly. I don't want my normative theorizing to hang on baroque theses in moral epistemology. The positivist has a good shot at vindicating this methodological stance, with- out baroque theses in moral epistemology. It frees us to develop ambitious and highly general work in normative ethics, without holding that work hostage to the vicissitudes of trolleyology. Unfortunately, I haven't made as much progress in de- veloping this methodological stance as I had hoped. This chapter is the rst to outline how positivist realism might have this sort of methodological upshots. Cru- cial details still need to be lled in. But at this point, it still looks like positivist realism is a way of vindicating this methodological stance. 277 Bibliography Abrus an, M arta. 2011. \Predicting the presuppositions of soft triggers." Linguistics and Philosophy 34:491{535. Abusch, Dorit. 2010. \Presupposition Triggering from Alternatives." Journal of Semantics 27 (1):37{80. Adams, Ernest. 1975. The Logic of Conditionals. Reidel, Dordrecht. Alm, David. 2000. \Moral Conditionals, Non-Cognitivism, and Meaning." Southern Journal of Philosophy 38:355{377. Anand Pranav and Hacquard, Valentine. 2013. \Epistemics and Attitudes." Se- mantics & Pragmatics 6 (8):1{59. Arneson, Richard. 2005. \Sophisticated Rule Consequentialism: Some Simple Ob- jections." Philosophical Issues 15:235{251. Atlas, Jay David. 2005. Logic, Meaning, and Conversation: Semantical Underde- terminacy, Implicature, and Their Interface. Oxford: Oxford University Press. Bach, Kent. 1994. \Conversational impliciture." Mind and Language 9:124{162. |. 1999. \The Myth of Conventional Implicature." Linguistics and Philosophy 22:327{366. |. 2011. \Perspectives on Possibilities: Contextualism, Relativism, or What?" In Andy Egan and Brian Weatherson (eds.), Epistemic Modality, 19{59. Oxford: Oxford University Press. Bar-On, Dorit and Chrisman, Matthew. 2009. \Ethical Neo-Expressivism." In Russ Shafer-Landau (ed.), Oxford Studies in Metaethics, V. 4, 133{166. Oxford: Oxford University Press. Barker, Stephen. 2000. \Is Value Content a Component of Conventional Implica- ture?" Analysis 60:268{279. Barkhausen, Max. 2016. \Reductionist Moral Realism and the Contingency of Moral Evolution*." Ethics 126:662{689. ISSN 0014-1704. 278 Barwise, Jon and Cooper, Robin. 1981. \Generalized Quantiers and Natural Lan- guage." Linguistics and Philosophy 4 (2):159{219. Beaver, David. 2001. Presupposition and Assertion in Dymanic Semantics. Stan- ford, CA: CSLI Publications. Bedke, Matt. 2009. \Intuitive non-naturalism meets cosmic coincidence." Pacic Philosophical Quarterly 90:188{209. |. 2014. \No Coincidence." In Russ Shafer-Landau (ed.), Oxford Studies in Metaethics, Volume 9, 102{125. Oxford: Oxford University Press. Bennett, Jonathan. 2003. A Philosophical Guide to Conditionals. Oxford University Press: Oxford. Bezuidenhout, Anne. 1997. \Pragmatics, semantic undetermination and the refer- ential/attributive distinction." Mind 106:375{409. Bj ornsson, Gunnar and Finlay, Stephen. 2010. \Metaethical Contextualism De- fended." Ethics 121 (1):7{36. Bj ornsson Gunnar and In, Alexander Almer. 2010. .. 2010. \The Pragmatics of Insensitive Assessments: Understanding the Relativity of Assessments of Judg- ments of Personal Taste, Epistemic Modals, and More." In The Baltic Interna- tional Yearbook of Cognition, Logic, and Communication, volume 6, 1{45. New Prairie Press: Manhattan, KS. Blackmore, Susan. 1999. The meme machine. Oxford University Press. Boer, Steven and Lycan, William. 1976. The Myth of Semantic Presupposition. Indiana University Linguistics Club. Boisvert, Daniel. 2008. \Expressive-Assertivism." Pacic Philosophical Quarterly 89:169{203. Boyd, Richard. 1988. \How to be a Moral Realist." In Georey Sayre-McCord (ed.), Essays on Moral Realism, 181{228. Cornell University Press. Brandt, Richard. 1963. \Toward a Credible Form of Utilitarianism." In H.-N. Casta~ neda and G. Nakhnikian (eds.), Morality and the Language of Conduct, 107{43. Detroit: Wayne State University. |. 1979. A Theory of the Good and the Right. Oxford: Oxford University Press. Braun, David. 2012. \An Invariantist Theory of Might Might be Right." Linguistics and Philosophy 35:461 { 89. Brink, David. 1989. Moral Realism and the Foundations of Ethics. Cambridge: Cambridge University Press. 279 |. 2001. \Realism, naturalism, and moral semantics." ocial Philosophy and Policy 18:154{76. Brown, Campbell. 2011. \A New and Improved Supervenience Argument for Ethical Descriptivism." In Russ Shafer-Landau (ed.), Oxford Studies in Metaethics 6, 205{18. Oxford: Oxford University Press. Butler, Joseph. 1729. \Sermons." In L A Selby-Bigge (ed.), British Moralists: Being Selections from Writers Principally of the Eighteenth Century, 181{244. New York: The Bobbs-Merrill Company. Cappelen, Herman and Hawthorne, John. 2009. Relativism and Monadic Truth. New York: Oxford University Press. Card, Robert. 2007. \Inconsistency and the Theoretical Commitments of Hooker's Rule-Consequentialism." Utilitas 19(02):243{258. Cariani, Fabrizio. 2013. \Ought and Resolution Semantics." No^ us 47(3):534{558. Carston, Robyn. 1996. \Metalinguistic Negation and Echoic Use." Journal of Pragmatics 25 (3):309{330. |. 2002. Thoughts and Utterances. Oxford: Blackwell. Chierchia, Gennaro and McConnell-Ginet, Sally. 2000. Meaning and Grammar: An Introduction to Semantics. Cambridge, MA: The MIT Press, second edition. Christensen, David. 2007. \Epistemology of Disagreement: The Good News." The Philosophical Review 116(2):187{217. Clarke-Doane, Justin. 2012. \Morality and Mathematics: The Evolutionary Chal- lenge." Ethics 122 (2):313{340. Cohen, Stewart. 1999. \Contextualism, Skepticism, and The Structure of Reasons." Philosophical Perspectives 13:57{89. Cook, Ezra. 2013. \Epistemic Modals and Common Ground." Inquiry 56:179{209. Cooper, Robin. 1983. Quantication and Syntactic Theory. Dordrecht: Reidel. Copp, David. 2000. \Milk, Honey, and the Good Life on Moral Twin Earth." Synthese 124:113{137. |. 2001. \Realist-Expressivism: A Neglected Option for Moral Realism." Social Philosophy and Policy 18:1{43. |. 2007. Morality in a Natural World: Selected Essays in Metaethics. New York: Cambridge University Press. 280 Copp, David. 2008. \Darwinian skepticism about moral realism." Philosophical Issues 18:186{206. Crisp, Roger. 2006. Reasons and the Good. Oxford: Oxford University Press. Cuneo, Terrence and Shafer-Landau, Russ. 2014. \The Moral Fixed Points: New Directions for Moral Nonnaturalism." Philosophical Studies . Damasio, Antonio. 1994. Descartes' Error: Emotion, Reason, and the Human Brain. G.P. Putnam's Sons. Dancy, Jonathan. 2004. Ethics Without Principles. Oxford University Press. Daniels, Norman. 1979. \Wide Re ective Equilibrium and Theory Acceptance in Ethics." The Journal of Philosophy 76 (5):256{282. Das, Ramon. 2016. \Evolutionary debunking of morality: epistemological or meta- physical?" Philosophical Studies 173 (2):417{35. Dawkins, Richard. 1982. The extended phenotype: The gene as the unit of selection. San Francisco: Freeman. Dekker, Paul. 2008. \A multi-dimensional treatment of quantication in extraordi- nary English." Linguistics and Philosophy 31:101{127. Dennett, Daniel. 1995. Darwin's dangerous idea: Evolution and the meanings of life. New York: Simon & Schuster. DeRose, Keith. 2002. \Assertion, Knowledge, and Context." The Philosophical Review 111 (2):167{203. Dorr, Cian. 2003. \Vagueness without ignorance." Philosophical Perspectives 17:83{ 114. Dorr, Cian and Hawthorne, John. 2013. \Embedding epistemic modals." Mind 122:867{913. ISSN 00264423. doi:10.1093/mind/fzt091. Dowell, Janice. 2011. \A Flexibly Contextualist Account of Epistemic Modals." Philosophers Imprint 1{25. |. 2013. \Flexible Contextualism about Deontic Modals: A Puzzle about Information-sensitivity." Inquiry 56, ( 2-3):149{178. |. 2016. \The Metaethical Insignicance of Moral Twin Earth." Oxford Studies in Metaethics 11:1{26. Dreier, James. 1992. \The Supervenience Argument against Moral Realism." South- ern Journal of Philosophy 30(3):13{38. 281 Dunaway, Billy. 2015. \Supervenience Arguments and Normative Non-naturalism." Philosophy and Phenomenological Research 91 (3):627{655. Dunaway, Billy and McPherson, Tristram. 2016. \Reference Magnetism as a Solu- tion to the Moral Twin Earth Problem." Ergo 3. Dworkin, Ronald. 1996. \Objectivity and truth: you'd better believe it." Philosophy and Public Aairs 25:87{139. Edgington, Dorothy. 1995. \On Conditionals." Mind 104:235{329. Egan, Andy. 2007. \Epistemic modals, relativism, and assertion." Philosophical Studies 133 (1):1{22. Egan, Andy, Hawthorne, John, and Weatherson, Brian. 2005. \Epistemic Modals in Context." In Gerhard Preyer and Georg Peter (eds.), Contextualism in Phi- losophy. Oxford University Press. Elbourne, Paul. 2005. Situations and Individuals. The MIT Pres: Cambridge, MA. Elga, Adam. 2007. \Re ection and Disagreement." No^ us 41 (3):478{502. Enoch, David. 2010. \The epistemological challenge to metanormative realism: how best to understand it, and how to cope with it." Philosophical Studies 148:413{ 438. |. 2011. Taking Morality Seriously: A Defense of Robust Realism. Oxford: Oxford University Press. Feldman, Richard. 2007. \Reasonable Religious Disagreements." In L. Antony (ed.), Philosophers without Gods, 194{214. Oxford: Oxford University Press. Fine, Kit. 1975. \Vagueness, Truth and Logic." Synthese 30:265{300. |. 1994. \Essence and Modality." In James Tomberlin (ed.), Philosophical Per- spectives 8: Logic and Language, 1{16. Atascadero: Ridgeview. Finlay, Stephen. 2004. \The Conversational Practicality of Value Judgement." The Journal of Ethics 8 (3):205{223. |. 2014. Confusion of Tongues: A Theory of Normative Language. Oxford Uni- versity Press. Finnis, John. 1980. Natural Law and Natural Rights. Oxford: Oxford University Press. FitzPatrick, William. 2014. \Debunking evolutionary debunking of ethical realism." Philosophical Studies 172. 282 Foot, Philippa. 1972. \Morality as a System of Hypothetical Imperatives." Philo- sophical Review 81(3):305{316. |. 2001. Natural Goodness. Oxford: Oxford University Press. Fraser, Benjamin James. 2014. \Evolutionary debunking arguments and the relia- bility of moral cognition." Philosophical Studies 168 (2):457{473. Geach, Peter. 1967. \Intentional Identity." Journal of Philosophy 64(20):627{632. Geurts, Bart. 1998. \Presuppositions and Anaphors in Attitude Contexts." Lin- guistics and Philosophy 21:545. ISSN 0165-0157. |. 1999. Presuppositions and Pronouns. Oxford: Elsevier. Gibbard, Allan. 2003. Thinking How to Live. Cambridge: Harvard University Press. Greene, Joshua. 2008. \The Secret Joke of Kant's Soul." In Walter Sinnott- Armstrong (ed.), Moral Psychology, Vol. 3: The Neuroscience of Morality, Emo- tions, Brain Disorders and Development., 35{80. Cambridge, MA: MIT Press. Hacquard, Valentine. 2006. Aspects of Modality. Ph.D. thesis, MIT. Hallpike, C. R. 1986. The principles of social evolution. New York: Oxford Univer- sity Press. Hampton, Jean. 1998. The Authority of Reason. Cambridge: Cambridge University Press. Hare, R M. 1973. \Rawls Theory of Justice." The Philosophical Quarterly 23:144{ 155. Harman, Gilbert. 1975. \Moral Relativism Defended." Philosophical Review 84:3{ 22. |. 1977. The Nature of Morality. Oxford University Press. |. 1996. Moral Relativism and Moral Objectivity. Oxford: Blackwell. Harsanyi, John. 1977. \Rule Utilitarianism and Decision Theory." Erkenntnis 11:25{53. Hawthorne, John. 2007. \Eavesdroppers and epistemic modals." Philosophical Issues 17:92{101. Hawthorne, John, Rothschild, Daniel, and Spectre, Levi. 2016. \Belief is weak." Philosophical Studies 173:1393{1404. Hawthorne, John & Manley, David. 2012. The Reference Book. New York: Oxford University Press. 283 Hazlett, Allan. 2010. \The Myth of Factive Verbs." Philosophy and Phenomeno- logical Research 80(3):497{522. Heim, Irene. 1982. The Semantics of Denite and Indenite Noun Phrases. Ph.D. thesis, University of Massachusetts. |. 1983. \On the Projection Problem for Presuppositions." In Daniel Flickinger Michael Barlow and Michael Westcoat (eds.), Second Annual West Coast Con- ference on Formal Linguistics, 114{126. Stanford. |. 1992. \Presupposition Projection and the Semantics of Attitude Verbs." Journal of Semantics 9:183{221. Hinckfuss, Ian. 1987. \The Moral Society: Its Structure and Eects." Hooker, Brad. 2000. Ideal Code, Real World: A Rule-consequentialist Theory of Morality. Oxford University Press. Horty, John. 2012. Reasons as Defaults. Oxford University Press. |. 2014. \Deontic Modals: Why Abandon the Classical Semantics?" Pacic Philosophical Quarterly 95:424{460. ISSN 02790750. doi:10.1111/papq.12044. Hume, David. 1739. A Treatise of Human Nature. Oxford: Clarendon, 2nd edition. Hursthouse, Rosalind. 1999. On Virtue Ethics. Oxford: Clarendon Press. Huvenes, Tornn. 2015. \Epistemic Modals and Credal Disagreement." Philosoph- ical Studies 172:111{987. Jackson, Frank. 1998a. From Metaphysics to Ethics: A Defence of Conceptual Analysi. Oxford: Clarendon Press. |. 1998b. From Metaphysics to Ethics: A Defence of Conceptual Analysis. Oxford: Oxford University Press. |. 1999. \Non-Cognitivism, Validity, and Conditionals." In Dale Jamieson (ed.), Singer and His Critics, 18{37. Oxford: Blackwell. Jackson, Frank and Pettit, Philip. 1995. \Moral functionalism and moral motiva- tion." Philosophical Quarterly 45 (178):20 { 40. Jeshion, Robin. 2010. \Singular Thought: Acquaintance, Semantic Instrumental- ism, and Cognitivism." In Robin Jeshion (ed.), New Essays on Singular Thought, 105{140. Oxford: Oxford University Press. Joyce, James. 2004. \Williamson on Knowledge and Evidence." Philosophical Books 45:296{305. Joyce, Richard. 2001. The Myth of Morality. Cambridge University Press. 284 |. 2006a. \Metaethics and the empirical sciences." Philosophical Explorations 9 (1):133{148. |. 2006b. The evolution of morality. Cambridge: MIT Press. |. 2013. \The many moral nativisms." In Kim Sterelny and Richard Joyce (eds.), Cooperation and its Evolution. Kahane, Guy. 2011. \Evolutionary debunking arguments." No^ us 45:103{125. Kahneman, Daniel and Tversky, Amos. 1974. \Judgment under Uncertainty: Heuristics and Biases." Science 185 (4157):1124{1131. Kamm, Frances. 2006. Intricate Ethics. Oxford: Oxford University Press. Kamp, Hans. ???? \Proper Names from a communication-theoretic Perspective." |. 1981. \A Theory of Truth and Semantic Representation." In Groenendijk Janssen and Stokhof (eds.), Formal Methods in the Study of Language. Mathe- matical Centre, Amsterdam. Kaplan, David. 1989. \Demonstratives." In Joseph Almog, John Perry, and Howard Wettstein (eds.), Themes from Kaplan. New York: Oxford University Press. Karttunen, Lauri. 1974. \Presupposition and Linguistic Context." Theoretical Linguistics 1:181 { 94. Karttunen, Lauri and Peters, Stanley. 1979. \Conventional implicature." In Choon- Kyu Oh and David Dinneen (eds.), Syntax and semantics vol. 11: Presupposition, 1{56. New York: Academic Press. Kempson, Ruth. 1975. Presupposition and the Delimitation of Semantics. Cam- bridge: Cambridge University Press. Khoo, Justin. 2015. \Modal Disagreements." Inquiry 58 (5). Kitcher, Phillip. 2007. \Biology and Ethics." In David Copp (ed.), The Oxford Handbook of Ethical Theory, 163{85. Oxford: Oxford University Press. Knobe, Joshua and Yalcin, Seth. 2014. \Epistemic modals and context: Experi- mental data." Semantics & Pragmatics 10:1{21. Kolodny, Niko and MacFarlane, John. 2010. \Ifs and Oughts." Journal of Philoso- phy 107 (3):115{143. Kratzer, Angelika. 1977. \What `Must' and `Can' Must and Can Mean." Linguistics and Philosophy 1:337{355. |. 1981. \The Notional Category of Modality." In H J Eikmeyer and H Rieser (eds.), Words, Worlds, and Contexts, 38{74. New York: de Gruyter. 285 |. 2012. Modals and Conditionals. New and Revised Perspectives. Oxford: Oxford University Press. Kripke, Saul. 2009. \Presupposition and Anaphora: Remarks on the Formulation of the Projection Problem." Linguistic Inquiry 40 (3):367{386. Laskowski, Nicholas and Finlay, Stephen. 2017. \Conceptual Analysis in Metaethics." In Tristam McPherson and David Plunkett (eds.), Routledge Hand- book of Metaethics. Routledge. Lenman, James. 2000. \Consequentialism and Cluelessness." Philosophy and Public Aairs 29 (4):342{370. Lennertz, Ben. 2013. Reasoning with Uncertainty and Epistemic Modals. Ph.D. thesis, USC. |. 2014. \Simple Contextualism about Epistemic Modals is Incorrect." Thought: A Journal of Philosophy 3(4):252{262. Levinson, Stephen. 1983. Pragmatics. Cambridge: Cambridge University Press. Lewis, David. 1970. \How to Dene Theoretical Terms." Journal of Philosophy 67:427{446. |. 1975. \Languages and Language." In Keith Gunderson (ed.), Language, Mind and Knowledge, volume 7 of Minnesota Studies in the Philosophy of Science, 3{35. Minneapolis: University of Minnesota Press. |. 1982. \Logic for Equivocators." Nous 16:431{441. |. 1983. \New Work for a Theory of Universals." Australasian Journal of Philos- ophy 61:343{377. L opez de Sa, Dan. 2008. \Presuppositions of Commonality: An Indexical Rela- tivist Account of Disagreement." In Garca-Carpintero and Kolbel (eds.), Relative Truth, 297{310. Oxford University Press. MacFarlane, John. 2007. \Relativism and Disagreement." Philosophical Studies 132:17{31. |. 2011. \Epistemic Modals are Assessment Sensitive." In Andy Egan and Brian Weatherson (eds.), Epistemic Modality, 144{178. New York: Oxford: Oxford University P. |. 2014. Assessment Sensitivity: Relative Truth and its Applications. Oxford: Oxford University Press. Mackie, J L. 1977. Ethics: Inventing Right and Wrong. London: Penguin. 286 Maguire, Barry. 2014. \T. M. Scanlon Being Realistic About Reasons." Times Literary Supplement 26. Mason, Elinor. 2004. \Consequentialism and the Principle of Indierence." Utilitas 16 (3):316{321. McGee, Vann. 1990. Truth, Vagueness and Paradox: An Essay on the Logic of Truth,. Indianapolis: Hackett Publishing. McGee, Vann and McLaughlin, Brian. 1995. \Distinctions Without A Dierence." Southern Journal of Philosophy 33:203{251. McMahan, Je. 2000. \Moral Intuition." In Hugh LaFollette (ed.), The Blackwell Guide to Ethical Theory, 92{110. Blackwell. McPherson, Tristram. 2015. \What is at Stake in Debates among Normative Real- ists?" Nous 49 (1). Meinong, Alexius. 1904. Untersuchungen zur Gegenstandstheorie und Psychologie. Leipzig: Barth. Montminy, Martin. 2012. \Epistemic Modals and Indirect Weak Suggestives." di- alectica 66:583{606. Moore, G E. 1903. Principia Ethica. Cambridge: Cambridge University Press. Moss, Sarah. 2013. \Epistemology Formalized." Philosophical Review 122 (1):1{43. |. 2015. \On the semantics and pragmatics of epistemic vocabulary." Semantics and Pragmatics 8:1 { 81. Mulgan, Tim. 2001. The Demands of Consequentialism. New York: Oxford Uni- versity Press. |. 2006. Future People: A Moderate Consequentialist Account of our Obligations to Future Generations. Oxford: Oxford University Press. Nagel, Thomas. 1986. The View from Nowhere. Oxford: Oxford University Press. Norcross, Alastair. 1990. \Consequentialism and the Unforeseeable Future." Anal- ysis 50 (4):253{256. Oshima, David. 2006. Perspectives in Reported Discourse. Ph.D. thesis, Stanford University. Part, Derek. 2011a. On What Matters, volume 1. Oxford: Oxford University Press. |. 2011b. On What Matters, volume 2. Oxford: Oxford University Press. 287 Pigden, Charles. 2012. \Identifying Goodness." Australasian Journal of Philosophy 90(1):93{109. Potts, Christopher. 2005. The Logic of Conventional Implicatures. Oxford: Oxford University Press. Prinz, Jesse. 2009. The Emotional Construction of Morals. Oxford: Oxford Uni- versity Press. Pryor, James. 2004. \What's Wrong with Moore's Argument." Philosophical Issues 14:349{77. Railton, Peter. 1984. \Alienation, Consequentialism, and the Demands of Morality." Philosophy and Public Aairs 13 (2):134{171. |. 1986. \Moral Realism." The Philosophical Review 95 (2):163{207. Ramsey, Frank P. 1926. \Truth and Probability." In Henry E. Kyburg and Howard E.K Smokler (eds.), Studies in Subjective Probability. Huntington, NY: Robert E. Kreiger Publishing Co. Rawls, John. 1971. A Theory of Justice. Cambridge, MA: Harvard University Press. |. 1975. \The Independence of Moral Theory." Proceedings and Addresses of the American Philosophical Association 48:5{22. |. 2005. Political Liberalism. New York: Columbia University Press. Recanati, Fran cois. 2004. Literal Meaning. Cambridge University Press. Richard, Mark. 1993. \Articulated Terms." Philosophical Perspectives 7:207{230. Richardson, Peter and Boyd, Robert. 2005. Not by genes alone. University of Chicago Press. Ridge, Michael. 2006a. \Ecumenical Expressivism: Finessing Frege." Ethics 116:36{ 302. |. 2006b. \Introducing Variable-Rate Rule-Utilitariansim." Philosophical Quar- terly 56:242{53. |. 2007. \Anti-Reductionism and Supervenience." Journal of Moral Philosophy 4:330{48. |. 2014. Impassioned Belief. Oxford: Oxford University Press. Roberts, Craige. 2012. \Information Structure: Towards an integrated formal the- ory of pragmatics." Semantics and Pragmatics 6:1{69. 288 Roberts, Craige, Simons, Mandy, Beaver, David, and Tonhauser, Judith. 2009. \Presupposition, Conventional Implicature, and Beyond: A unied account of projection." In Nathan Klinedist and Daniel Rothschild (eds.), Proceedings of Workshop on New Directions in the Theory of Presupposition, ESSLI 2009. Rosen, Gideon. 2010. \Metaphysical Dependence: Grounding and Reduction." In Bob Hale and Aviv Homan (eds.), Modality, 109{136. Oxford: Oxford University Press. Ross, W D. 1930. The Right and the Good. Oxford: Oxford University Press. Rothschild, Daniel. 2011. \Explaining presupposition projection." Semantics & Pragmatics 4:1{42. Rozin, Paul, Haidt, Jonathan, and McCauley, Clark. 1993. \Disgust." In L. F. Lewis, M.; Haviland-Jones, J. M.; Barrett (ed.), Handbook of Emotions, 757{ 776. New York: Guilford Press. Ruse, Michael. 1998. Taking Darwin Seriously. Oxford: Blackwell Publishing. Ruse, Michael and Wilson, E. O. 1986. \Moral Philosophy as Applied Science." Philosophy 61:173{92. Russell, Bertrand. 1905. \On Denoting." Mind 479{493. Salmon, Nathan. 2002. \Demonstrating and Necessity." The Philosophical Review 111 (4):497{537. Scanlon, T M. ???? \Contractualism and Utilitarianism." In Steven Cahn and Joram Haber (eds.), 20th Century Ethical Theory, 647{665. Prentice Hall. |. 1998. What We Owe to Each Other. Cambridge, MA: Harvard University Press. |. 2002. \Rawls on Justication." In Samuel Freeman (ed.), The Cambridge Companion to Rawls, 139{167. Cambridge University Press. |. 2014. Being Realistic about Reasons. New York: Oxford University Press. Schafer, Karl. 2013. \Perception and the Rational Force of Desire." Journal of Philosophy 110 (5):258{281. Schlenker, Philippe. 2003. \A Plea for Monsters." Linguistics and Philosophy 26:29{120. ISSN 0165-0157. doi:10.1023/A:1022225203544. |. 2010. \Presuppositions and Local Contexts." Mind 119:377{391. Schneider, Benjamin. 2010. \Expressivism Concerning Epistemic Modals." Philo- sophical Quarterly 60 (240):601{615. 289 Schroeder, Mark. 2005. \Realism and Reduction: The Quest for Robustness." The Philosophers' Imprint 5 (1). |. 2007. Slaves of the Passions. Oxford: Oxford University Press. |. 2008a. Being For: Evalauting the Semantic Program of Expressivism. Oxford: Oxford University Press. |. 2008b. \How Does the Good Appear to Us?" Social Theory and Practice 34(1):119{ 130. |. 2009. \Hybrid Expressivism: Virtues and Vices." Ethics 119 (2):257{309. |. 2011. \ntextitfOughtsg, Agents, and Actions." Philosophical Review 120 (1):1{ 41. |. 2015. \Is Semantics Formal?" In Expressing our Attitudes, 209 { 224. Oxford University Press. Setiya, Kieran. 2012. Knowing Right From Wrong. Oxford University Press. Shafer-Landau, Russ. 2003. Moral Realism: A Defence. Oxford: Oxford University Press. |. 2012. \Evolutionary Debunking, Moral Realism and Moral Knowledge." Journal of Ethics and Social Philosophy 7 (1):1{37. Sider, Ted. 2014. Writing the Book of the World. Oxford. Sidgwick, Henry. 1907. The Methods of Ethics. Hackett. Silk, Alex. 2016. Discourse Contextualism: A Framework for Contextualist Seman- tics and Pragmatics. New York: Oxford University Press. |. 2017a. \How to Embed an Epistemic Modal: Attitude Problems and Other Defects of Character." Philosophical Studies . |. 2017b. \Normative Language in Context." In Russ Shafer-Landau (ed.), Oxford Studies in Metaethics, Volume 12. Oxford University Press. Simons, Mandy. 2000. Issues in the Semantics and Pragmatics of Disjunction. Routledge. |. 2001. \On the Conversational Basis of Some Presuppositions." In Semantics and Linguistic Theory 11. Singer, Peter. 1974. \Sidgwick and Re ective Equilibrium." The Monist 58:490{517. |. 2005. \Ethics and Intuitions." Journal of Ethics 9:331{352. 290 Sinnott-Armstrong, Walter. 2006. Moral skepticisms. Oxford: Oxford University Press. Skarsaune, Knut Olav. 2011. \Darwin and moral realism: survival of the iest." Philosophical Studies 152:229{243. Smart, J J C. 1973. Utilitarianism: For and Against. New York: Cambridge University Press. Smith, Holly. 2010. \Measuring the Consequences of Rules." Utilitas 22:413{33. Soames, Scott. 1999. Understanding Truth. Oxford: Oxford University Press. |. 2009a. \Presupposition." In Philosophical Essays, Volume 1: Natural Language: What It Means and How We Use It, 73{130. Princeton, NJ: Princeton University Press. |. 2009b. \The Gap between Meaning and Assertion: Why What We Literally Say Often Diers from What Our Words Literally Mean." In Philosophical Essays, Volume 1: Natural Language: What It Means and How We Use It, 278{297. Oxford: Oxford University Press. |. 2010. Philosophy of Language. Princeton, NJ: Princeton University Press. |. 2015. Rethinking Language, Mind, and Meaning. Princeton, NJ: Princeton University Press. Sorensen, Roy. 1988. Blindspots. Oxford: Oxford University Press. Sperber, Dan. 1996. Explaining culture: A naturalistic approach. Oxford: Blackwell. Sperber, Dan and Wilson, Deirdre. 1986. Relevance: Communication and Cognition. Cambridge, MA: Harvard University Press. Stalnaker, Robert. 1973. \Presuppositions." Journal of Philosophical Logic 2:447{ 457. |. 1974. \Pragmatic Presuppositions." In Milton K Munitz and Peter K Unger (eds.), Semantics and Philosophy. New York: New York University Press. |. 1984. Inquiry. Cambridge, MA: MIT Press. |. 1999. Context and Content: Essays on Intentionality in Speech and Thought. Oxford: Oxford University Press. |. 2002. \Common Ground." Linguistics and Philosophy 25:701{721. Stanley, Jason. 2007. Language in Context: Selected Essays. Oxford: Oxford Uni- versity Press. 291 Stephenson, Tamina. 2007. \Judge Dependence, Epistemic Modals, and Predicates of Personal Taste." Linguistics and Philosophy 30 (4):487{525. Stevenson, Charles Leslie. 1937. \The Emotive Meaning of Ethical Terms." Mind 46 (181):14{31. Strandberg, Caj. 2012. \A Dual Aspect Account of Moral Language." Philosophy and Phenomenological Research 84 (1):87{122. |. 2015. \Options for Hybrid Expressivism." Ethical Theory and Moral Practice 18 (1):91{111. Strawson, P F. 1974. \Freedom and Resentment." In Freedom and Resentment and other essays. Methuen and Co, Ltd. Street, Sharon. ???? \Objectivity and Truth: You'd Better Rethink It." |. 2006. \A Darwinian Dilemma for Realist Theories of Value." Philosophical Studies 127 (1):109{166. |. 2008a. \Constructivism about reasons." Oxford Studies in Metaethics 3:207{ 245. |. 2008b. \Reply to Copp: Naturalism, Normativity, and the Varieties of Realism Worth Worrying about." Philosophical Issues 18:207{228. Sturgeon, Nicholas. 1988. \Moral Explanations." In Georey Sayre-McCord (ed.), Essays on Moral Realism. Ithaca: Cornell University Press. |. 2006. \Ethical Naturalism." In David Copp (ed.), The Oxford Handbook of Ethical Theory, 91{121. New York: Oxford University Press. Sudo, Yasutada. 2012. On the Semantics of Phi Features on Pronouns. Ph.D. thesis, Massachusetts Institute of Technology. Swanson, Eric. 2006. Interactions with Context. Ph.D. thesis, Department of Lin- guistics and Philosophy, MIT. Tappenden, Jamie. 1993. \The Liar and Sorites Paradoxes: Toward a Unied Treatment." Journal of Philosophy 60:551{77. Thomason, Richmond. 1972. \A semantic theory of sortal incorrectness." Journal of Philosophical Logic 209-258. Thompson, Michael. 2012. Life and Action. Cambridge, MA: Harvard University Press. Thomson, Judith Jarvis. 1997. \The Right and the Good." Journal of Philosophy 94(6):273{298. 292 Tonhauser, Judith, Beaver, David, Roberts, Craige, and Simons, Mandy. 2013. \Towards a Taxonomy of Projective Content." Language 89 (1):66{109. Tucker, Christopher. 2010. \When Transmission Fails." The Philosophical Review 119 (4):497{529. van der Sandt, Rob. 1992. \Presupposition projection as anaphora resolution." Journal of Semantics 9:333{377. van Fraassen, Bas. 1980. The Scientic Image. Oxford: Clarendon. van Rooij, Robert. 2005. \A modal analysis of presupposition and modal subordi- nation." Journal of Semantics 22:281{305. ISSN 0167-5133. |. 2010. \Presuppsoition: An (un)common attitude?" In Rainer Bauerle, Uwe Reyle, and Thomas Zimmermann (eds.), Presuppositions and Discourse: Essays Oered to Hans Kamp. Emerald Group Publishing Limited. Vavova, Katia. 2014a. \Debunking Evolutionary Debunking." In R Shafer-Landau (ed.), Oxford Studies in Metaethics, Volume 9, 76{101. Oxford: Oxford Univer- sity Press. |. 2014b. \Moral Disagreement and Moral Skepticism." Philosophical Perspectives 28(1):302{333. |. 2015. \Evolutionary debunking of moral realism." Philosophy Compass 10:104{ 116. Veltman, Frank. 1996. \Defaults in Update Semantics." Journal of Philosophical Logic 25:221{261. von Fintel, Kai and Gillies, Anthony. 2008. \CIA Leaks." Philosophical Review 117 (1):77{98. |. 2011. \Might Made Right." In Andy Egan and Brian Weatherson (eds.), Epis- temic Modality, 108{130. Oxford University Press. Wall, Edmund. 2009. \Hooker's Consequentialism and the Depth of Moral Experi- ence." Dialogue 48 (2):337{351. Wallace, R. Jay. 1998. Responsibility and the Moral Sentiments. Cambridge, MA: Harvard University Press. Wedgwood, Ralph. 2007. The Nature of Normativity. Oxford: Oxford University Press. |. 2016. \Review: Being Realistic about Reasons." The Philosophical 66 (262):213{ 217. 293 White, Roger. 2010. \You just believe that because." Philosophical Perspectives 24:573{615. Wielenberg, Erik. 2010. \On the evolutionary debunking of morality." Ethics 120:441{464. Willer, Malte. 2013. \Dynamics of Epistemic Modality." Philosophical Review 122(1):45{92. |. 2014. \Dynamic Thoughts on Ifs and Oughts." Philosophers' Imprint 14 (28):1{ 30. Williamson, Timothy. 1994. Vagueness. London: Routledge. |. 2000. Knowledge and its Limits. Oxford: Oxford University Press. Wilson, Deirdre. 1975. Presupposition and Non-Truth Conditional Semantics. New York: Academic Press. Wood, Allen. 1972. \The Marxian Critique of Justice." Philosophy and Public Aairs 1:244{82. |. 2011. \Humanity as an End in Itself." In On What Matters, Volume 2, 58{82. New York: Oxford University Press. Wright, Crispin. 2007. \New Age Relativism and Epistemic Possibility: The Ques- tion of Evidence." Philosophical Issues 17:262{283. Yablo, Stephen. 2000. \Red, Bitter, Best." Philosophical Books 41:13{23. Yalcin, Seth. 2007. \Epistemic Modals." Mind 116:983{1026. |. 2011. \Nonfactualism about epistemic modality." In Andy Egan and Brian Weatherson (eds.), Epistemic modality, 295{332. New York: Oxford University Press. |. 2016. \Belief as Question-Sensitive." Philosophy and Phenomenological Research . 294

Abstract (if available)

Abstract This dissertation introduces a new picture of moral language and thought. ❧ This picture has significant upshots for moral epistemology, for moral metaphysics, and for normative ethics. I will call my new picture positivist realism. I develop it in four Parts. Part I introduces the view, and explains why it would be nice for non-skeptical moral realists if it is true. The goal in this part won’t be to argue that the view is actually true. Rather, I’ll try to show that it would be nice if it were. Part II will argue that positivist realism is true, by marshaling a range of linguistic evidence. Part III will answer a central problem for the positivist. And then Part IV will introduce some upshots that positivist realism has for normative ethics. ❧ The positivist holds that a range of central problems in moral epistemology are pseudo-problems. They seem forceful only because we’ve misunderstood the way our language works. In particular, they seem forceful only because we’ve failed to appreciate that moral discourse carries substantial presuppositions. ❧ The positivist makes non-skeptical moral realism easier to defend, in two different ways. First, it’s much easier to answer moral skeptics if the positivist right. It’s just not possible to formulate the skeptics’ arguments in this picture. Second, the positivist increases the range of realist views that are worth taking seriously. An important class of metaethical arguments start from our conviction that we do have moral knowledge. Arguments in this class leverage that conviction as evidence against some realists. Sharon Street (2006) is one example. She argues that her non-realist conception of moral belief is the only way to vindicate that conviction. In doing that, she’s taking that conviction as evidence for her view and as evidence against her realist rivals. This style of argument also figures in intramural debates among realists. Non-naturalists like T. M. Scanlon (2014) are one example. They use our conviction as evidence against their naturalist rivals. The positivist systematically disarms this kind of argument. ❧ The positivist thinks that metaethicists have made a very important mistake about the semantics and pragmatics of moral language. This mistake has led them to try to address some problems that are in fact only pseudo-problems. Moreover, attempting to answer these pseudo-problems has had bad effects in other areas of moral philosophy. In particular, it has had bad effects in moral metaphysics and in normative ethics.

Linked assets