Vagueness, legal content, and legal interpretation

PDF

Download

Open document

Transcript (if available)

Content VAGUENESS, LEGAL CONTENT, AND LEGAL INTERPRETATION by Hrafn Asgeirsson _______________________________________________________________________ A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (PHILOSOPHY) May 2012 Copyright 2012 Hrafn Asgeirsson ii DEDICATION To my wife, Anna iii ACKNOWLEDGMENTS I would like to thank my primary committee members – Andrei Marmor and Gideon Yaffe – for all their help during the writing of this dissertation. Both have been immensely generous with their time and advice. It has been a privilege to have Andrei as an advisor – his expertise and patience have helped improve all aspects of the material. I consider myself particularly lucky to have been able to work with him on issues at the intersection of philosophy of law and philosophy of language during the past several years. Gideon provided me with excellent advice and instruction on countless occasions. Many of the ideas in this dissertation are there in large part because of him. I would also like to like to thank Scott Altman for taking the time to be on my committee and Gary Watson for several helpful discussions over the course of writing the dissertation. Several of their comments have helped me frame the issues in this dissertation better, particularly in the last chapter. During my time at USC, I have also benefited particularly from advice and instruction from Scott Soames and Mark Schroeder. I also owe a great debt to Kory Declark. I am incredibly lucky to have him as a friend and philosophical advisor. Finally, my greatest debt is to my wonderful wife, Anna Thorvaldsdottir. I’d like to thank her, well, for everything, and dedicate this dissertation to her. iv TABLE OF CONTENTS Dedication ii Acknowledgements iii Abstract v Introduction 1 Chapter One: Endicott on the Value of Vague Law 9 Chapter Two: Sorensen on Vagueness and Power-Delegation 59 Chapter Three: Marmor on Non-Literal Legislative Speech 98 Chapter Four: Indeterminacy about Utterance Content 152 Chapter Five: Textualism, Content, and Interpretation 172 Chapter Six: Resolving Cases of Vagueness 215 Bibliography 257 Appendix: List of Legal Cases 264 v ABSTRACT Lawmaking is – paradigmatically – a type of speech act: people make law by saying things. It is natural to think, therefore, that the content of the law is determined by what lawmakers communicate. However, it is sometimes vague what content they communicate, and even when it is clear, the content itself is sometimes vague. In my dissertation, I examine the nature and consequences of these two linguistic sources of indeterminacy in the law. I argue that vagueness in the law is sometimes a good thing, although its value should not be overestimated. I also propose a strategy for resolving borderline cases, which – if correct – sheds significant light on the debate about legal interpretation. 1 INTRODUCTION Lawmaking is – paradigmatically – a type of speech act: people make law by saying things. It is natural to think, therefore, that the content of the law is determined by what lawmakers communicate. However, it is sometimes vague what content they communicate, and even when it is clear, the content itself is sometimes vague. As a result, the law is to some extent vague. In my dissertation, I examine the nature and consequences of these two linguistic sources of indeterminacy in the law. The aim is to give plausible answers to three related questions: In virtue of what is the law vague? What might be good about vague law? How should courts resolve cases of vagueness? Many laws not only contain vague expressions, but contain expressions that are likely to generate deep disagreements over how to appropriately apply the relevant norms. These expressions – such as ‘neglect’, ‘reasonable’, and ‘with all deliberate speed’ – are typically associated with multiple attributes that cannot be measured in common units, which entails that for many cases there will be no answer to the question whether the relevant law applies to them. As a result, the content of these laws is to a considerable extent vague. The second type of vagueness in the law is primarily due to the fact that lawmakers and their audience rarely share much information regarding legislative intent. As a result, interpreters are often not in a position to determine whether or not the legislature intended the relevant law to be construed literally. Consequently, it is 2 sometimes vague whether the content of a law is its literal content or some non-literal pragmatic enrichment of that content. Legal indeterminacy of these two types, I argue, is sometimes a good thing. First, lawmakers sometimes do better to enact law with vague content, and to let interpreters handle borderline cases if and when they arise, than to try to work out a more determinate alternative. Second, lawmakers sometimes do better to make it vague what the content of a law is than to not legislate at all. However, we also have at least three good reasons not to overestimate the value of vague law. First, some of the value associated with the use of vague terms in the law is wrongly attributed to vagueness. Second, due to the highly contingent nature of behavioral incentives, we should not overestimate the extent to which vague law incentivizes people to take desirable courses of action. Third, because there is reason to doubt that courts generally have any special expertise with respect to the underlying purposes of either contract law or negligence law, we should not overestimate the extent to which they are able to maximize fidelity to law in deciding borderline cases. Nevertheless, I argue, courts can often appropriately decide borderline cases by consulting reliable statements about legislative purpose made by lawmakers at key points in the legislative process. In particular, if lawmakers can be expected to possess relative expertise with respect to the relevant domain of conduct or if there is a sufficient need to respect the legislative bargains that facilitated the relevant legislation, then judges should resolve vagueness in the law by deference to such statements of intent. 3 Traditionally, people have argued that vague law can be a good thing, either because it appropriately delegates limited lawmaking power to officials or because it elicits certain kinds of desired behavior on the part of subjects. Timothy Endicott has argued, for example, that vague law allows the courts to work out the details of the law, which may be preferable in domains where the courts satisfy certain competence conditions, and that vague law can give people incentives to avoid creating risk to others or to contract out of liabilities in cost-effective ways. 1 Some forms of conduct, he says, are better regulated by vague norms than by precise norms. Although I agree with certain aspects of Endicott’s view, I argue in chapter one that due attention to issues like the highly contingent nature of behavioral incentive and the limits of judicial competence shows that the value of vagueness in the law should not be overestimated. Endicott has also argued, however, that vagueness in the law is sometimes valuable because some forms of human conduct simply cannot be regulated by precise norms. 2 In these cases, Endicott thinks, vagueness is valuable in virtue of being a necessary means to valuable legislative ends. I argue that vagueness is actually not a means to the relevant legislative ends. Rather, vague law is a necessary consequence of an even more important feature of the law, which I call incommensurate multidimensionality. Value, however, only “transmits” from ends to means, but not to necessary consequences of those means. This critique also applies to many versions of 1 Endicott (2011). 2 Endicott (2011). 4 the popular view that vagueness is valuable because it is necessary for securing valuable flexibility in the law. In chapter two, I turn to a thoroughly skeptical view regarding the value of vagueness in the law. Roy Sorensen has argued that the traditional arguments for the value of vagueness are mistaken – vagueness in the law neither has power-delegating value nor does it give people the behavioral incentives required in order for the traditional arguments to work. 3 I will be concerned primarily with challenging Sorensen’s claim that vagueness does not have a valuable power-delegating function, since his argument threatens to take away what is traditionally thought to be the main reason for thinking that vagueness in the law can be a good thing. I argue that lawmakers sometimes do better to enact vague law – and thus to delegate limited lawmaking power to agencies and courts – than to set simple bright-line rules or to work out complex precise alternatives. I also briefly address Sorensen’s argument against the claim that vagueness is valuable because uncertainty about liability gives people a valuable behavioral incentive, since his argument raises important questions about another major reason that vagueness has been taken to be valuable to law, especially by scholars influenced by law and economics. Chapters one and two are about the value of using vague terms in legislation. We can call this kind of indeterminacy vagueness of content. In chapter three, I turn to another – somewhat neglected – form of vagueness: vagueness about content. As I aim to show, not only is the content of a legislative utterance often vague, it is often vague what the content of such an utterance is. 3 Sorensen (2001a). 5 The starting point for this discussion is the simple observation that in ordinary conversational settings it frequently happens that speakers assert something other than what they literally say. Given the ubiquity of this phenomenon, it makes sense to ask whether it also frequently happens that legislatures assert something other than what they literally say. Andrei Marmor has argued that non-literal legislative speech is rare. 4 According to Marmor, a speaker succeeds in asserting something other than what she literally says only if it is obvious that she cannot be intending to assert the literal content of her remark. And that rarely happens in law, he says. I argue that the epistemic constraint on non- literal assertion is often lower than Marmor holds – although its exact “height” may (like standards of proof) vary between legal domains. I also argue, however, that Marmor’s argument can be successfully revised: the content of a statutory directive is rarely determinately different from its literal content, due to the fact that legislative contexts generally contain little unequivocal information about legislative intent. The fact that the conversational background in the legislative speech context generally isn’t rich enough to warrant a reinterpretation, however, does not entail that the content of the law is – by default – its literal content. In chapter four, I argue that once we analyze all the relevant pragmatic aspects of legislative utterances, we see that there is plenty room for indeterminacy about utterance content on the framework I adopt. The basic idea is that in certain cases in which the audience is warranted in withholding belief regarding the speaker’s communicative intention, the primary content of the relevant 4 Marmor (2008). 6 utterance is indeterminate between the literal content of the sentence uttered and some non-literal enrichment thereof. This has significant consequences for the analysis of a number of important but controversial legal cases, which I discuss in some detail. Given what I say in chapters three and four, my view may seem to imply a version of textualism. And, as I discuss in chapter five, in one sense it does, at least insofar as textualism is taken to be a theory of legal content rather than legal interpretation (on the understanding that legal interpretation is the act of developing the law in the face of indeterminate/inconsistent legal content or a particularly problematic result). Textualism, however, is more than just a theory about legal content – it is primarily a normative theory of adjudication of which such a theory is a part. And as I explain, my view differs significantly from full-fledged textualism. Part of the reason is that I think that legal content is often less determinate than textualist rhetoric suggests. Mainly, however, the difference is that I don’t think that the standard textualist considerations support their limited view of the judge’s role in all cases in which interpretation is called for. In chapter six, I conclude the dissertation by arguing that interpreters can use similar strategies to resolve the two types of vagueness that I have discussed: sometimes, the law’s background purpose(s) will dictate a decision. However, I also argue that there are significant restrictions on the sort of background rationale that courts can legitimately appeal to in settling borderline cases. I propose that – insofar as the aim is to maximize fidelity to law – such appeal is normally restricted to what I call the operative rationale 7 of the law: roughly, rationale that lawmakers have offered under sincerity-inducing conditions during the legislative process. It is one thing, however, to say that borderline cases can be settled by reference to legislative purposes and another to say that they should, and so we need to consider what reasons judges might have to resolve cases of vagueness by deference to the relevant statute’s background justification. I think that there are at least two types of reasons for such deference. First, judges can have authority-based reasons to defer to legislative intentions, if the legislature can be expected to possess expertise regarding the relevant domain of conduct. Second, however, judges may also have non-expertise-based reasons to defer to the legislature’s intentions. I argue that one of the strongest reasons in favor of resolving borderline cases by reference to legislative rationale is that doing so respects legislative bargaining and thereby helps preserve a fundamental feature of the legislative institution. Thus, although my dissertation is primarily about specific questions regarding the nature and value of vagueness in the law, my discussion also sheds significant light on other important issues in jurisprudence. It will emerge from chapters five and six that two of the leading theories of legal interpretation in a significant sense often complement – rather than compete with – each other. If what I say in these chapters is correct, a linguistically sophisticated version of textualism is the best theory of legal content – the content of the law is determined by what the legislature can reasonably be taken to have said – while a constrained version of intentionalism is often the best theory of legal interpretation in borderline cases – in cases of vagueness, the law should sometimes be 8 precisified in accordance with the rationale that lawmakers have offered for it under sincerity-inducing conditions during the legislative process. 9 CHAPTER ONE: ENDICOTT ON THE VALUE OF VAGUE LAW In “The Value of Vagueness,” Timothy Endicott argues that vagueness can be valuable to law, contrary to the impression that many have of the normative nature of law. 5 If the law aims to regulate conduct and vague norms entail that there may be cases in which it is justified neither to say that the norm applies nor that it does not apply, it can seem like a plainly bad idea to introduce vagueness into the law. Endicott seeks to resolve this apparent tension, in part by emphasizing that the law aims not just to guide action, but to guide action for a purpose. When we evaluate whether vagueness and precision are valuable features of a given scheme of regulation, we must always make sure to evaluate those features with respect to law’s various purposes, both general and specific. According to Endicott, the value of vagueness in the law has two main sources: (i) some forms of human conduct are better regulated by vague norms than by precise norms and (ii) some forms of human conduct cannot be regulated by precise norms. When both precise and vague schemes of regulation are available to lawmakers, they must determine whether, in a given scheme of regulation, the consequences of precision are worse than the consequences of vagueness. I call this scenario the Lawmakers’ Challenge, and we can call the associated value of vagueness its comparative value. However, when precise schemes of regulation are impossible, as Endicott claims is very commonly the case, lawmakers have only one “alternative”, if they wish to regulate. In such circumstances, the only means for regulating the relevant forms of human conduct is to issue vague 5 See Endicott (2011). 10 norms. Thus, lawmakers sometimes must use vague regulation. We can call the value of vagueness associated with this scenario its instrumental necessity value. I will argue that Endicott’s argument for the instrumental necessity value of vagueness for law is unsound, but that he is basically right about the comparative value of vagueness. First, I will provide a brief overview of Endicott’s view regarding the varieties and values of vagueness in the law. Since his view is quite complex, it will help us understand and evaluate his arguments if we first get an idea about how he thinks about vagueness in the law and about how that connects to the claims that he makes about its value. I will then go on to examine his argument for the claim that vagueness has instrumental necessity value, as well as his argument for the claim that it has comparative value. 1.1 Endicott’s framework and its relationship to his arguments Endicott seems to distinguish between (at least) three categories of vagueness, each of which has certain explanatory significance. Their exact nature is not made explicit in Endicott’s text, but they are motivated via examples. Two of the labels are Endicott’s – trivial vagueness and extravagant vagueness – and one is invented – evident vagueness. 6 The categories all have core examples, but their boundaries cannot be expected to be sharp. Trivial vagueness includes vagueness associated with terms that are otherwise as precise as expressions get – terms like ‘year’ etc. – and with hedging expressions like 6 Thanks to Andrei Marmor for the label. 11 ‘approximately’ and ‘more or less’, which are typically “pegged” to something precise (consider ‘approximately 2 feet tall’). 7 The vagueness associated with the former is generally unavoidable – for any speaker of a natural language – but hedging expressions are easy to avoid in law. Although Endicott is not clear about what it is that makes these sorts of vagueness trivial, it seems sensible to say that in the case of hedging it is the fact that such expressions simply aren’t used in law and that in the case of relatively precise expressions it is the fact that, relative to clear cases, borderline cases are rare, and if they do come up, they are not likely to be practically consequential. Thus, this type of vagueness is very unlikely to generate what Endicott calls “serious and deep” disputes over the relevant norm’s principles of application and it therefore seems quite appropriate to call this category of vagueness trivial vagueness. 8 Evident vagueness includes vagueness associated with paradigmatic vague terms like ‘heavy’, ‘old’, ‘slow’, ‘tall’, ‘bald’, ‘heap’, ‘child’, ‘crowd’, etc. Some of these terms – like ‘heavy’, ‘old’, ‘slow’, and ‘tall’ – are unidimensional, i.e. associated only with one attribute – weight, age, speed, and height, respectively – and some – like ‘bald’, ‘heap’, ‘child’, and ‘crowd’ – are multidimensional, i.e. associated with more than one attribute – for ‘crowd’, e.g., the number of people and the density of the relevant gathering. 9 Whether a term is evidently vague, I will take it, depends on the salience of a totally 7 See Endicott (2000), 92. 8 This does not mean that vagueness in measurement terms is never significant. We can readily imagine, for example, that instances of nanotechnology will produce practically consequential borderline cases with respect to existing – and more coarse-grained – regulation. 9 See Endicott (2000), 45. 12 ordered dimension – like weight or height. In the unidimensional cases mentioned above, the totally ordered dimension is the only dimension associated with the respective term and so these will be the most evidently vague terms. Some multidimensional terms also have certain relatively salient totally ordered dimensions and so will also count as (more or less) evidently vague – consider e.g. ‘child’ and age, as well as ‘crowd’ and the quantity of people. Now, according to Endicott, if a term is associated with multiple dimensions and two or more of these dimensions are incommensurate, i.e. if they correspond to “attributes that cannot be measured in common units,” 10 this entails that the term is vague. The reason, he thinks, is that incommensurability entails partially defined comparatives (‘is more/less F than’). That is, if the attributes associated with a term cannot be measured in common units, then for any (actual or possible) x, there will be some (actual or possible) y, such that it is indeterminate whether x is more/less F than y. If so, then it is not possible to identify any x such that x is less F, or just as F, than any other y:Fy and more F than any z:~Fz. In other words, it is not possible to identify a sharp boundary between the Fs and the non-Fs. And if this is not possible, then F is a vague 10 Endicott (2000), 42. Two attributes can fail to be measurable in a common unit in at least two ways: (i) the attributes may each have a natural numerical ordering but there is no common unit to which the relevant scales can non-arbitrarily be reduced, and (ii) the attributes may not have any natural numerical ordering. 13 predicate. 11 Incommensurability (across associated dimensions), then, entails vagueness, in Endicott’s view. 12 Having briefly introduced the notion of incommensurability and its purported relation to vagueness, it is time to introduce the category that Endicott appears to judge most important to law: extravagant vagueness. Extravagantly vague expressions include terms such as ‘reasonable’, ‘fair’, and ‘good’. Such expressions typically have a multiplicity of incommensurate dimensions, which entails that for any (actual or possible) x, there will be significantly many (actual or possible) “items” y, such that it is indeterminate whether x is more/less F than y. That is, the indeterminate cases will start to crowd out the determinate ones. Such extravagant vagueness is, on Endicott’s view, primarily associated with normative and evaluative terms associated with multiple attributes, although nothing in his text indicates that it is restricted to those types of expressions. Borderline cases that arise due to the occurrence of extravagantly vague terms are very likely to be practically consequential. Thus, this type of vagueness is also very likely to generate serious and deep disputes over the relevant norm’s principles of application. This is true for example of terms like ‘neglect’ and ‘reasonable’, both of which are associated with multiple incommensurate dimensions and both of which are common in 11 See Endicott (2000), 36 (especially footn. 6). 12 Others that have argued for the claim that incommensurability entails vagueness include Griffin (1986) and Broome (1997), although Griffin takes vagueness to be rather unimportant. Raz (1986) argues that incommensurability does not entail vagueness, although he thinks it is compatible with it. 14 tort, contract, and administrative law (and the latter even in commercial law). I imagine that phrases like ‘substantial and unjustifiable risk’, which forms part of the definitions of recklessness and negligence in the Model Penal Code (MPC), belong in this category as well. 13 From what Endicott says, it is not entirely clear whether artifact terms – like ‘trade’, ‘vehicle’, and ‘dwelling’ – and terms that are used to refer to human actions – like ‘commit’, ‘carry’, and ‘take’ (all of which figure in criminal law statutes) – belong in the category of extravagant vagueness. I take it that some may – like ‘vehicle’, ‘chair’, ‘game’, and ‘carry’ – and some may not – like ‘crowd’ (assuming the only relevant dimensions are the number of people and the density of the gathering). And some, of course, may be borderline. 1.2 The values of non-trivial varieties of vagueness When lawmakers deliberately use vague language, Endicott says, they typically use extravagantly vague terms – especially normative terms – and they generally avoid both evidently vague terms – like ‘child’ – and certain trivial vagueness – like hedging expressions. 14 Endicott is not as explicit about his view regarding lawmakers’ use of non- 13 M.P.C. §2.02(c) and (d). It is interesting to note that in an attempt to make the MPC’s requirements of culpability “precise”, the Alaskan legislature defined substantial and unjustifiable risk partly in terms of what a reasonable person would do; see A.S. §28.35.400(a), concerning reckless driving. 14 It is reasonable to wonder whether hedging expressions really give rise only to trivial vagueness. Imagine, for example, that the law required people to drive more or less on the right side of the road. Surely, such vagueness would not be trivial. Nevertheless, it is possible to make sense of Endicott’s claim. Firstly, as he notes, such expressions are routinely avoided in legal texts. Since their occurrence is rare, the frequency borderline cases will be relatively low. 15 normative vague terms that could either count as extravagantly vague or borderline so – like ‘trade’ and ‘vehicle’ – but he seems to think that it varies whether such terms are used, depending on the costs and benefits of vague versus precise schemes of regulation. When it comes to building regulations or tax law, for example, he claims that lawmakers have a choice between “more or less complicated and detailed” schemes of regulation. 15 I take it that he is implying that, at least in some respects, vague schemes in which such terms occur could be replaced by relatively precise ones, although this may not necessarily be desirable. It may be tempting, based on Endicott’s claims about the ways in which lawmakers use vague language, to infer that – typically – evident and non-normative extravagant vagueness can be avoided but extravagant vagueness associated with normative terms cannot. This would certainly explain the common occurrences of vague normative terms in law. 16 And to further infer that the comparative value of vagueness concerns only vagueness associated with non-normative terms and its instrumental necessity value concerns extravagant vagueness associated with normative terms. But this Secondly, and more importantly, when hedging expressions are used, they usually introduce only a “fuzzy edge” – compare ‘at 2pm’ and ‘at around 2pm’; see Endicott (2000), 92. Although, the extent of this fuzzy edge may vary depending on the conversational or practical context, hedging expressions are typically used with relatively precise expressions, in which case the resulting penumbra is relatively small. Thus, we can reasonably expect that hedging expressions are very unlikely to generate “serious and deep disputes” over the relevant norm’s principles of application and so would appropriately count as trivially vague. 15 Endicott (2011), 25. 16 I say ‘occurrences’ because the issue is not that a large set of vague terms is used, but rather that the use of certain such terms is widespread (‘reasonable’ is a good example). 16 would be too hasty. Endicott’s view regarding the value of vagueness in law is more complicated than that. We see that this is so from the fact that Endicott thinks that there are no precise alternatives to some vague schemes of regulation in criminal law and that there are more precise alternatives to extravagantly vague schemes of regulation in contract law. He argues, for example, that it is in general not possible to regulate the use of violence with precise schemes, but that such schemes would suffice for the regulation of contracts (although the precise alternative might not be better). If he is right, then there are legislative scenarios in which vagueness associated with non-normative terms has instrumental necessity value and extravagant vagueness associated with normative terms has comparative value. There is a further complexity to Endicott’s view – a distinction between individual norms and schemes of regulation, which turns out to play an important role in explaining cases in which vagueness has both instrumental necessity value and comparative value. Schemes of regulation are sets of norms intended to regulate particular forms of human conduct – such as conduct that is specially damaging to children, or is particularly likely to be. In some domains – like the domain of child care law – lawmakers have no choice but to use a vague scheme of regulation, i.e. a set of norms, some members of which contain occurrences of vague terms. However, lawmakers do have some choices to make regarding the exact composition of the relevant set of norms – in other words, the scheme of regulation can be vague to a greater or lesser extent. It must contain some vague norms, but it may be possible to incorporate 17 certain precise norms as well, in particular when there is some sufficiently relevant feature that can be quantitatively measured – such as age (as a proxy for maturity) or speed (as a proxy for safe vehicular conduct). We can therefore say that in case it is better to refrain from incorporating a precise norm in a given vague scheme of regulation, vagueness has comparative value in addition to instrumental necessity value. If I understand Endicott correctly, what I said in the previous paragraph applies also to the value of vagueness in negligence law. There is no way, he thinks, in which we could devise precise schemes to regulate risk-creating forms of conduct. So vagueness has instrumental necessity value. However, it is surely possible to incorporate some precise norms into the relevant set of norms. Nevertheless, Endicott argues, it is better – at least in some cases – to refrain from doing so. In such cases, vagueness also has comparative value. Now that we have a preliminary picture of how Endicott thinks about vagueness in the law and some idea about how it connects to the claims that he makes about its value, it is time to examine in more detail the support he gives for those claims based on the framework I have described. 2.1 Endicott’s argument from instrumental necessity and the transmission of value In his argument for the instrumental necessity value of vagueness in law, Endicott is primarily motivated by the following claim: Vagueness is a necessary means for achieving general regulation of a widely varying range of conduct. That is, in order to 18 regulate forms of conduct that vary widely, the law must use vague general terms. In particular, lawmakers must sometimes issue extravagantly vague norms, norms that are very likely to generate actual borderlines cases that are practically consequential. In order to adequately prevent, and punish, conduct that is specially damaging to children, for example, the law must use terms like ‘neglect’ and ‘abandoned’ – or at least use approximate, and usually non-exhaustive, definitions thereof, which are themselves extravagantly vague. Assuming – as we certainly can in this case – that it is valuable to prevent such conduct (and valuable to punish it when it occurs), it is valuable to use vague terms, Endicott argues. Let us say that the law adequately regulates undesirable conduct if it prevents an adequate number of instances of it and adequately punishes the instances that do occur. We can then analyze Endicott’s argument in the following way: 17 P1. It is valuable that: The law adequately regulates conduct that is specially damaging to children. P2. That the law uses vague terms is a necessary means for adequately regulating conduct that is specially damaging to children. C. It is valuable that: The law uses vague terms. The argument is valid, and I certainly do not wish to dispute P1. What I have to show, then, is that P2 is false – vagueness, as it turns out, is not a necessary means for regulating the relevant form of conduct. 17 Thanks to Olav Gjelsvik for helpful discussion about Endicott’s argument. 19 The following discussion will focus on the term ‘neglect’ as it used in child care law, since that is one of Endicott’s favorite examples of vague terms that the law must use. I will begin by explaining very briefly the main characteristics of the type of vagueness with which Endicott is concerned – what he calls extravagant vagueness. Then, I argue that although extravagant vagueness is indeed necessary for regulating the relevant form of conduct, it is not a necessary means for doing so. Rather, it is a necessary consequence of the semantic property that actually facilitates reference to the relevant forms of conduct – what I call incommensurate multidimensionality. And value, I argue, only “transmits” from ends to means, but not to necessary consequence of those means. 18 2.2 Incommensurate multidimensionality as a means According to Endicott, the immediately relevant issue is that terms like ‘reasonable care’ and ‘neglect’ cannot possibly be replaced by precise phrases that are sufficiently similar in extension, or by any disjunctions thereof. In general, only terms that have one or more totally ordered dimensions, or that are otherwise significantly correlated with such an attribute, can be so replaced. ‘Citizens are required to drive safely’, for example, can be replaced with ‘Citizens are required to drive at no more than n mph’ in tandem with a host of other more or less precise rules, because speed is such a weighty factor in safe driving, and the resulting scheme of laws can still be said to have the purpose of 18 I should note that my discussion owes a great deal both to Soames (2011) and to Kolodny (unpublished). 20 preventing automobile-related accidents/injuries/damage. Similar with ‘Only mature citizens may vote’ and ‘Only citizens older than n years of age may vote’ (even though age is only a significantly correlated attribute, since age is not something in virtue of which one counts as mature). Vague terms or phrases that are not associated with any attribute that can be quantitatively measured, on the other hand, do not generally admit of replacement by relatively precise alternatives, nor do terms that are associated with a relatively insignificant proportion of such dimensions, since – in most cases – the resulting regulation could not properly be said to be regulation of the conduct in question (more on this below, in the section “On the necessity of incommensurate multidimensionality and – hence – vagueness”). Arguably, ‘neglect’ falls in the latter category, since, although time (spent with or away from one’s children) and age (of, say, baby-sitters) are certainly relevant attributes, these dimensions – along with other totally ordered dimensions – do not suffice to make for an adequate replacement. In these cases, vagueness is indeed necessary for the relevant regulatory ends. But that, I hope to show, does not suffice for vagueness to have value, even if the ends in question are valuable. Let me begin by saying that what lawmakers should do in order to regulate a wide variety of conduct with a general standard is not – in the first instance – to use terms that are vague. Rather, it is to use broad general terms associated with multiple incommensurate dimensions, or perhaps to use approximate – and usually non-exhaustive – definitions thereof which are themselves highly general and multidimensional. (It is important to note that non-exhaustive definitions do not replace their definienda; rather, 21 they are meant to be guidelines for application.) Endicott even indicates as much in at least two places. He claims for example that the relevant issue regarding negligence law is the “sheer, mind-boggling variety of ways in which people can create more or less unreasonable risks to other people,” 19 and that the relevant issue regarding child care law is the “daunting variety of things that a child needs from its parents [which] corresponds to a wide variety of ways in which a parent may more or less neglect a child.” 20 Now, I think that Endicott is right that extravagantly vague laws generally do not have any relatively precise replacements, since terms tend to be extravagantly vague due to the fact that they have multiple dimensions that cannot be measured in a common metric (and some are bound to be impossible to measure in any reasonable way, at least cardinally). But I think he is wrong in maintaining that this is the issue that is relevant to determining whether or not vagueness has value in virtue of being a necessary means. The reason, I will argue, is that vagueness is not the semantic feature that helps bring about this “general regulation of a widely varying range of conduct.” 21 To see better what semantic feature it is that does help bring this about, let us take a look at what seems to me to be generally true of the relevant lawmaking scenarios: 1. Lawmakers need/want to regulate a set, {T 1 , …, T n }, of multiple related types of behavior. 19 Endicott (2011), 25. 20 Endicott (2011), 24. 21 Endicott (2011), 28. 22 2. There exists a type T, denoted by an existing term t, such that each member of {T 1 , …, T n } is a subtype of T. 22 3. It is not possible for lawmakers to compile a set – precise or not – of less general terms, {t 1 , …, t n }, such that there is a one-to-one mapping between {t 1 , …, t n } and {T 1 , …, T n }. In Endicott’s favored example, lawmakers want to prohibit certain forms of conduct in order to prevent certain types of harm to children (and to punish instances that do occur), namely (something like) those types that constitute unreasonable failure on behalf of parents to pay attention to the needs that their children have in virtue of their parent-child relationship. And indeed there exists a term that covers the relevant forms of behavior – namely ‘neglect’. Due to the nature of such conduct, however, it is not possible for lawmakers to devise a list of less general terms – precise or not – that disjunctively covers the relevant behavior (i.e. a necessarily coextensive list of action-type terms). As Soames (2011) notes: “one who undertook the task of more precisely delineating either [the class of clear neglect or the class of clear non-neglect] using only more highly- focused language designating specific behavior – about the regularity and content of meals, frequency of trips to doctors, time with parents, age of baby-sitters, and the like – would … find it stupefying at best, and practically impossible, at worst.” 23 22 Note that worries about family resemblances are premature, since it suffices for this characterization that, for example, ‘game’ denotes a property that is informative and necessary of all instances of every sub-type of game – such as chess, solitaire, and golf. It is only in relation to the exact nature of this property that issues about family resemblances come up, since no one would deny that, necessarily, chess, solitaire, and golf are all games. 23 Soames (2011), 40. 23 If lawmakers were to try to use any such attempted list, they would likely get things gravely wrong; they would probably end up failing to prevent (and punish) a significant amount of harmful conduct as well as prevent (and punish) a significant amount of non-harmful conduct. 24 That is, any attempt to come up with a list would result in excessively over- and underinclusive regulation. Lawmakers, then, should not attempt to specify the set of action-types meant to be prohibited. Given that they want to adequately prevent (and punish) harm that is specially damaging to children, they should just use the term ‘neglect’. That, for example, is exactly what UK lawmakers did when they wrote the Child and Young Persons Act (1933); as Endicott notes: “The variety of ways in which children may be left more or less alone led to the vague standard of ‘neglect’ in the [Act].” 25 2.3 The impossibility of specification There are, of course, a variety of ways in which something can be impossible, and so far I haven’t said anything about in what sense it is impossible for lawmakers to devise an adequate, more specific, list of terms. Is it metaphysically impossible, or epistemically impossible, or perhaps simply practically impossible? It may be tempting to think that the impossibility of compiling an informative list of subtypes of neglect has to do with the context-sensitive nature of evaluation. And 24 To be sure, the law has some tools for handling errors like these – such as the necessity defense; more on this below. 25 Endicott (2011), 24. 24 probably it does have something to do with it. However, the immediate issue of unspecifiability has nothing to do with evaluation, as such. To see this, let us say that someone who utters ‘The Joneses have been neglectful towards their child’ succeeds in communicating the proposition that the Joneses have unreasonably failed to pay attention to the needs of their child. 26 We can – at least for practical purposes – separate the evaluative aspect of this (rough, but sufficient) analysis from the descriptive aspect, and focus just on the latter (i.e. on the proposition that the Joneses failed to pay attention to the needs of their child). With respect to unspecifiability, ‘neglect’ is in an important sense no different from many purely descriptive terms: just as there is a vast variety of ways, for example, in which people can walk to work or use a firearm, so there is a vast variety of ways in which people can, as parents, fail to pay attention to the needs of their children (in part due to the diversity of a child’s needs). In general, there is nothing distinctively evaluative about the inability to list exhaustively the ways in which one can realize a given action-type, even if the relevant term is properly characterized as evaluative; we could repeat this line of reasoning e.g. with ‘cruel’, ‘lewd’ ‘unjust’, and so forth. Now, this inability could merely be practical. Perhaps there is a finite number of ways in which parents can fail to pay attention to the needs of their children but for practical reasons we just can’t list them. Or there is a countably infinite number of ways 26 I say ‘communicates’ because I wish to remain neutral about the “linguistic location” of the evaluate aspect of such an utterance. That is, nothing turns on if the evaluative aspect is part of the semantics of the term or if it gets attached pragmatically (by way of assertion, implicature, or presupposition). 25 and so it is impossible to know what a complete list looks like. If so, then it will of course also be impossible to construct one. Or perhaps the number is uncountably infinite, in which case any sort of complete list is impossible (try e.g. to imagine counting possible worlds), let alone knowable or humanly constructible. Soames, for one, seems to endorse the claim that it is at least practically impossible, as evidenced by the quote above. It is unclear, however, whether he would be willing to go any further. To be sure, when it comes to normative terms, there is a further – potentially significant – complication: we are only interested in those instantiations of the relevant types that have the further property of being unjustified (whatever property that turns out to be). If parents, for example, for justifying reasons do not feed their children regular nutritious meals – say, during a widespread famine – then presumably we wouldn’t complain that they are being neglectful. 27 So, what we are ultimately after is a list of action-types in (types of) circumstances: A1 in C1; A2 in C2; and so on. Each item on this more complicated list, then, is what is correctly said to be less general than ‘neglect’, even if the relation isn’t straightforwardly taxonomic (like it is with ‘tiger’ and ‘mammal’). Thus, even if we had a reasonable list of action-types that corresponds to the ways in which parents can fail to pay attention to their children’s needs, we would need to revise that list so as to include only unreasonable instantiations of those types. And here, we are faced with a problem similar to the immediate one described above, in that we 27 Or consider the morality-based motivation for allowing battered mothers to seek a defense of duress or of diminished capacity against prosecution for Failure to Protect. 26 need a list of exceptions, determined by a theory of right and wrong action. Depending on one’s metaethical view, however, such a list may be finite, countably infinite, or uncountably infinite. Only if the list of exceptions is finite and practically manageable can we expect an adequate specification of ‘neglect’. This is a tall order, and probably not to be expected. As Väyrynen (2009) points out, outside ethics it has not proved promising to provide finite lists of exceptions to Ceteris Paribus generalizations, and there is no a priori reason to think that ethics is special in that respect. 28 If what I have said in this section and in the section before it is true, then we have good reason to believe that using terms that are associated with multiple incommensurate dimensions is a necessary means for adequately regulating, for example, conduct that is specially harmful to children. This, of course, does not suffice to show decisively that P2 is false, since A’s being a necessary means for C does not preclude B from also being a necessary means for C, but it does very strongly suggest that vagueness is not what is really doing the work, vis-à-vis the regulation of the relevant forms of conduct. In the next section, I want to tackle an issue that I have left unaddressed so far. I have been taking it pretty much for granted that using a certain type of terms is indeed a necessary means for regulating conduct that is specially damaging to children and that the primary issue is what semantic aspect of those terms is doing the work. But, as I hope to show below, it is not obvious that this is true. In many cases, it is better to regulate conduct with terms or phrases that are associated with relatively few dimensions – and which are therefore relatively precise – than with vague multidimensional terms. The 28 Citing e.g. Fodor (1991) and Pietroski and Rey (1995). 27 question is why this is not also the case when it comes to regulating conduct that is specially damaging to children. Why exactly are there supposed to be no good proxies for terms like ‘neglect’, as it is used in child care law? 2.4 On the necessity of incommensurate multidimensionality and – hence – vagueness Let us suppose, based on what I said in the previous section, that it is in fact – at a minimum – practically impossible to compile a list of action-types (in circumstances) that is necessarily coextensive with ‘neglect’. It is natural to wonder if this is any different from other forms of conduct that lawmakers have an interest in regulating. Surely, in many cases, lawmakers do the right thing by trying to replace general terms with something more specific – and thus often better action-guiding – even if it is does result in a specification that is likely to be both underinclusive and overinclusive. Sometimes, it is simply better to settle for some sort of rough approximation. When it comes to drinking and voting age, for example, it would make bad sense to formulate the law in terms of maturity. Given the extent to which people would be at the mercy of officials and the cost of evaluating each case, it is better to avoid that mess and make age the legally relevant factor, as was done in the National Minimum Drinking Age Act (1984) and in the 26 th Amendment of the US Constitution. 29 Why shouldn’t we do something similar with ‘neglect’? Couldn’t we for example either (i) construct an admittedly underinclusive but manageable list of relevant action- types or (ii) find some quantifiable aspect – such as the minimum age of babysitters, 29 23 U.S.C. §158 and U.S. Const., Am. 26, respectively. 28 minimum number of visits to the doctor pr. month, minimum number and nutritional value of meals pr. day etc. – and rely on available legal tools, such as legal defenses, to mitigate any over- and underinclusiveness? In the case of neglect, I believe the answer is no. And I think we can see why if we attend to the fact that the relevant notion of being an adequate proxy has two dimensions. For our purposes, we can say that a term t 1 is an adequate proxy for a term t 2 iff it is not the case that: the product of the number of errors and their disvalue resulting from the use of t 1 > the product of the number of errors and their disvalue resulting from the use of t 2 . What makes suggestions (i) and (ii) above inadequate proxies has, I think, primarily to do with the gravity of the resulting errors – i.e. with the gravity of the over- and underinclusiveness and the limits of legal tools to mitigate these. It seems to me that, on any reasonable theory of value, what is at stake in adequately preventing child neglect is considerably weightier than what is at stake in preventing, say, immature people from voting or drinking. By saying this, I just mean that the interest that children have in not being subject to neglect (underinclusion) and the interest that parents have in not being held legally responsible for morally permissible or excused conduct (overinclusion) is weightier or stronger than any comparable interests that may be present in the context of voting and drinking. If so, then it is reasonable to think that instances of over- or underinclusiveness in the attempt to prohibit child neglect will be quite serious. And given that any attempt to provide a fairly specific, but manageable, list of relevant conduct is likely to be widely off the mark, if not entirely so, we can reasonably say that the resulting over- and 29 inclusiveness would be excessive. Given the number of errors and their disvalue, it is too off the mark to count as a means to the lawmakers’ end of adequately preventing and punishing child neglect. This provides, I believe, a reasonable explanation why nothing will make for a relatively precise proxy for ‘neglect’, as opposed for example to ‘mature’. As I mentioned, the law does have some tools for handling errors like these – most notably legal defenses that aim at exculpating those who break the law excusably or justifiably (examples include Insanity and Necessity, respectively). These tools, however, have significant limitations. The necessity – or “lesser evils” – defense, for example, which allows for acquittal in case the defendant’s best course of action really was to break the law, is in general not known to people, and authorities arguably have an interest in keeping it that way. 30 This means that circumstances in which subjects will do what is best – by breaking the law – are likely to be relatively rare. And even if people did know about the defense, they would presumably be very hesitant to take the risk and hope for a successful defense. In fact, it seems that they should be, since – as Yaffe (2009) explains – the calculation of lesser/greater evils is far from consistent between courts, because it is not settled (in any jurisdiction) what standards are to be employed in evaluating the alternatives or what it is exactly that needs to be better than what. 31 30 For a discussion, see Dan-Cohen (1984). 31 See in particular Yaffe (2009), Sec. 2, 372–378. I should note that Yaffe doubts that the necessity defense is best characterized as a justificatory defense. Rather, he thinks, it is a procedural estoppel defense, much like the defense of double jeopardy. If the government were to find the defendant guilty, its conduct would be “inconsistent in some way with some prior government conduct with respect to the defendant,” vis-à-vis its authority over her. 30 In addition to all this, legal defenses can only mitigate the effects of overinclusion. But in the case of neglect, in contrast to most other legislative scenarios, 32 underinclusion is arguably the more serious side of the coin (or at least equally serious), and the only possible way of mitigating underinclusion seems to be the employment of purposive legal interpretation (which can of course also be used to handle cases of overinclusion). But this presupposes that cases of underinclusion reliably reach the courts, and much has to happen for a case that is not covered by the law (but which should be) to get that far in the system. It is probably not an exaggeration to say that most cases of underinclusion will stay out of the courts’ reach and, as a result, judicial interpretation is a rather inefficient tool for dealing with underinclusion. From all this, it seems to me appropriate to conclude that in the case of neglect, using a list of fairly specific action-types is not an option that is available to lawmakers, given the end that they have – at least not as a means to that end. I won’t venture to define the threshold that – no doubt, vaguely – divides courses of action into means and non-means, relative to ends, but I do think we can say that, whatever that threshold is, using a set of fairly precise requirements just cannot be considered a means to adequately regulate conduct that is specially damaging to children. It is just too far off the mark. I should note, though, that even if this is the case, I do not think that lawmakers are rationally required to use the term ‘neglect’ undefined, as the UK lawmakers did in the Children and Young Persons Act. It may well be that although no manageable list will 32 As Husak (2008) notes, there is in law a presumption against overinclusion/overcriminalization (see 154); such a presumption, however, can be justified only if overinclusion is generally worse than underinclusion. 31 count as an adequate proxy, some sort of approximating high-level definition will. If so, then it is not necessarily the case that lawmakers ought to use the term ‘neglect’ undefined, and certainly not the case that they must. In the United States, for example, the federal government provides a non- exhaustive definition of ‘neglect’ that constrains any state legislation: “the term … means, at a minimum, any recent … failure to act on the part of a parent or caretaker, which results in death, serious physical or emotional harm, sexual abuse or exploitation, or … failure to act which presents an imminent risk of serious harm” (my emphasis). 33 To be sure, when non-exhaustive definitions are provided, they do not replace their definienda, and many states (such as California, Connecticut, and Kentucky) have followed the federal government in defining ‘neglect’ in this incomplete way (with some variation of course). 34 However, several states (including New Jersey, Texas, and Utah) do provide what appear to be complete high-level definitions of ‘neglect’, i.e. definitions that do replace the “ordinary meaning” of the term even if they are intended to capture or approximate it. 35 Although it is true that this attempt on behalf of US state lawmakers to be more informative than the 1933 UK lawmakers probably does result in over- and underinclusiveness, it is not so widely off the mark that it is doesn’t merit serious 33 See The Child Abuse Prevention and Treatment Act (CAPTA), 42 U.S.C. §5101 et seq; 42 U.S.C. §5116 et seq. 34 California Welf. & Inst. Code §300; Connecticut Gen. Stat. § 46b-120; Kentucky Rev. Stat. § 600.020. 35 New Jersey Ann. Stat. § 9:6-8.21; Texas Fam. Code § 261.001; Utah Ann. Code § 78A-6-105. 32 consideration. The over- and underinclusiveness is bad, of course, but it is arguably not excessive, and so the use of approximating high-level definitions is a possible alternative to using the term ‘neglect’ either undefined or non-exhaustively defined. Thus, it can’t be the case that, in order to adequately prevent (and punish) harm that is specially damaging to children, lawmakers have to use the term ‘neglect’, although a case could be made that they ought to. Even if it is the best way to do so, it is certainly not the only way. What, then, should we make of the claim that incommensurate multidimensionality – and, hence, vagueness – is necessary in order to adequately regulate conduct that is specially damaging to children? The answer, I think, is this: for every way in which you can reasonably formulate a legal text that aims to adequately regulate conduct that is specially damaging to children, it is the case that, necessarily, if lawmakers enact a norm using a text formulated in that way, then the resulting law contains terms that are associated with multiple incommensurate dimensions and, hence, vague. Or, to put it another way, there are more or less general ways, as well as more or less informative ways, in which lawmakers might try to achieve the relevant purpose, but the law will be multidimensional – and therefore vague – no matter which one of them is chosen. In fact, I believe that in the case of neglect, the law will be extravagantly vague, just as Endicott claims, since any proposal that can be considered a means to the relevant end has to be multidimensional in much the same way as ‘neglect’ is. The US definitions of ‘neglect’ are good examples of this. And multidimensionality of that sort entails extravagant vagueness, as we have already seen. 33 I hope to have made a reasonable case for the claim that expressions that cover classes of widely varying conduct do so not in virtue of the fact that they are vague, but rather in virtue of the fact that they have multiple – and often incommensurate – dimensions. To be sure, the vagueness of the relevant terms is, by necessity, tied to their multidimensionality, but not in a way that makes vagueness central in our explanation of how the law can, or should, achieve regulation of very broad classes of conduct. Rather, vagueness is a necessary consequence of the feature that actually does facilitate reference to these classes. To use vague terms, then, is something lawmakers have to do in order to achieve certain valuable ends, but that does not make their vagueness a means to those ends. 2.5 A possible reply: Vagueness is still necessary It is natural to wonder whether Endicott cannot simply respond to what I have said by claiming that, although the value of vagueness is perhaps not as “immediate” as we might have thought, it nevertheless does have value. After all, vagueness – like multidimensionality – is necessary for achieving certain valuable ends; it is still true that in order to achieve these ends, lawmakers have to use vague terms. Does it not follow, then, that vagueness is – like multidimensionality – valuable? We can represent this argument by simply tweaking Endicott’s P2 a little bit: P1. It is valuable that: The law adequately regulates conduct that is specially damaging to children. 34 P2*. In order to adequately regulate conduct that is specially damaging to children, the law must use vague terms. C. It is valuable that: The law uses vague terms. Amended this way, both premises of the argument are true, and so if I want to show that it doesn’t go through, I will have to show that the following general argument form is invalid (I use the ‘sees to it’ locution, common in the logic of action, to emphasize that we are talking about what someone needs to do in order to do some other thing): It is valuable that: x sees to it that p In order that x sees to it that p, x must see to it that q ∴ It is valuable that: x sees to it that q Call the underlying rule of inference ‘Value Transmission’ – ‘VT’ for short: VT If it is valuable that x sees to it that p, and, necessarily, if x sees to it that p, then x sees to it that q, then it is valuable that x sees to it that q. I will have to show that VT does not hold universally and that the most plausible restriction of it will not suffice to save Endicott’s argument. Since the principle has significant intuitive pull, this is not an easy feat. I will argue that it does not apply to necessary consequences of necessary means, which – if what I say is otherwise correct – suffices to show that the proposed amendment of Endicott’s argument is invalid. 35 Certainly, we can say that, if lawmakers don’t enact vague law, the probability of adequately regulating – at least by law – conduct that is specially damaging to children is zero, or at least sufficiently close thereto. This would seem to suffice to make enacting vague law a practical necessity – it is in fact something that lawmakers must do, if they are to achieve their goals. However, in order that value “transmit” in the required way, it does not suffice that vagueness simply be necessary. It must also be a means. That is, it must in some relevant sense help bring about the relevant legislative ends. Consider the following example, borrowed and adapted from Kolodny (unpublished). Imagine that a boxer aims to land a punch, but that, necessarily, whenever he decides to throw a punch, he signals his decision with a facial cue. There is a clear sense, therefore, in which the boxer has to signal in order to land a punch. Signaling, however, is not a means for landing a punch – it does not help to bring that end about. Consequently, Kolodny thinks that no value transmits from boxer’s aim to the signaling. And I agree. Even if it is valuable for the boxer to see to it that he lands a punch, it is not valuable for him that he sees to it that he signals his decision. In fact, the signaling lowers the probability of achieving the end and so if the signaling has any value given the boxer’s goal, then it seems to be negative value. Signaling is simply a by-product of something that (in some worlds) does help bring the end about, namely deciding to throw a punch. Signaling, therefore, appears to be a practical necessity to which the relevant ends transmits no value. VT thus seems to fail for practical necessities that aren’t means. Now, although vagueness is presumably more closely related to certain legislative ends than signaling is to landing a punch, it is still not sufficiently related to make it a 36 means. Let me borrow and adapt an example from Kim (1974) to substantiate this claim. Say that Meletus sees to it that Socrates dies. By doing so, he also sees to it that Xantippe becomes a widow. But there is no causal relationship between Socrates’s dying and Xantippe’s becoming a widow – the dependence is of another sort. 36 Imagine also, for the sake of argument, that in order to protect the youth of Athens, Meletus has to see to it that Socrates dies. Given the relation between seeing to it that Socrates dies and seeing to it that Xantippe becomes a widow, Meletus also has to see to it that Xantippe becomes a widow in order to achieve his end. Suppose now that it is in fact valuable that Meletus sees to it that the youth of Athens is protected and that – consequently – it is valuable that he sees to it that Socrates dies. It would seem odd to say that, by the same token, it is valuable that Meletus sees to it that Xantippe becomes a widow. Although her fate is something that has to be brought about in order to achieve a valuable end, it is not a means to that end – there is no interesting relation between her being or not being a widow and the youth of Athens. As in Kolodny’s boxer case, her becoming a widow is just a necessary by-product of something that does help bring the relevant end about: Socrates’s dying. Hence, her fate seems to be another practical necessity to which no value transmits. VT, therefore, apparently fails in this case too. 36 This sort of dependency is often called Cambridge dependency; see e.g Kim (1974). For those who don’t like Kim-style events and prefer Davidson-style events instead, I can make the same point by arguing that value-talk and reasons-talk is intensional rather than extensional. This isn’t very hard, since standard deontic logic assumes that only logically equivalent descriptions can be substituted for one another. So even if we have theoretical reasons for saying that the lawmakers’ using an incommensurately multidimensional term is in fact identical with the lawmakers’ using a vague term and the former is instrumentally valuable, it does not follow that the latter is also (since these events are identical without being logically equivalent). 37 Vagueness, I want to claim, is much like Xantippe’s becoming a widow. What is going on in law, on my account, is that by using terms with multiple incommensurate dimensions, lawmakers also use terms that are (extravagantly) vague. And although multidimensionality (of the right sort) is a means for achieving regulation of widely varying forms of conduct, vagueness is not. Hence, the value associated with the relevant legislative ends does not transmit to vagueness. To sum up, the examples borrowed from Kolodny and Kim arguably show, or at least provide good evidence, that VT does not hold for necessary consequences of necessary means. If that is correct, then the proposed amendment of Endicott’s argument is invalid. I conclude, therefore, that Endicott has not shown that “vagueness is valuable as a technique for achieving the general regulation of a widely varying range of conduct,” 37 even if it is indeed the case that it is necessary for such regulation. In the next section, I want to briefly consider a possible reply to my criticism, a reply that highlights the difference that theories of vagueness might conceivably make for conclusions about its value. After that, I also want to explain what I take to be a broader payoff of the foregoing discussion. Lastly, I briefly examine the suggestion that what is really going on in the law has to do neither with vagueness nor with multidimensionality, but with the essential contestability of the relevant concepts. 37 Endicott (2011), 28. 38 2.6 Another possible reply: What does vagueness consist in? My criticism of Endicott’s argument relied crucially on two claims. One is that vagueness is a not a means for achieving the general regulation of a widely varying range of conduct, but rather a necessary consequence of something that is – namely incommensurate multidimensionality. The other is that value does not transmit from ends to necessary consequences of necessary means to those ends. It might be complained, however, that my argument for the first claim assumes that the vagueness of a predicate consists in its being susceptible to having borderline cases. If vagueness of the sort associated with terms like ‘neglect’, on the other hand, could be said to consist in their incommensurate multidimensionality, then it does follow that vagueness has value for law in virtue of being a necessary means. It might be thought, for example, that vagueness just is whatever gives rise to borderline cases – i.e. their source – and incommensurate multidimensionality, as we have seen, does exactly that. Thus, one’s theory about what vagueness consists in will make a difference to what conclusions can be drawn regarding its value. It is hard to see, however, how the vagueness of a predicate could consist in its having multiple incommensurate dimensions, even if we are limiting our discussion to predicates that aren’t straightforwardly Sorites-susceptible. Granted, it is not enough that predicates simply have borderline cases, but that doesn’t mean that we should say that having borderline cases isn’t partly constitutive of vagueness. As many have noted, vague terms must also be tolerant: a vague term must be applicable to any object that is sufficiently similar – for the purposes of the conversation – to some object to which the 39 term determinately applies (we can call it a paradigm if we like). Of course, different theories will have different things to say about the exact nature of each feature, but it is unclear to me what motivation we have for thinking that their combination is not what vagueness consists in. To be sure, we may need explanations of what it is that gives rise to these features, but that doesn’t mean that the explanans is what vagueness really consists in. Some explanations are constitutive, some aren’t. In addition to the above, the reply I am considering here will not be available if we adopt Endicott’s own constitutive explanation of vagueness. “To understand a vague word,” he says, “is to be able to use paradigms, and vagueness is flexibility in their use,” 38 which indicates that he takes tolerance – as described above – to be the defining feature of vague terms. If so, then the question is whether that feature helps lawmakers bring about the relevant regulation. I see no reason to think that it does, although it may of course have other kinds of utility. Next, I want to turn to what I take to be a broader payoff of my critique. It is a payoff, I think, that it applies not just to Endicott’s argument but to a number of related arguments regarding the value of vagueness for law. If I am correct, these arguments also wrongly attribute value to vagueness on the grounds that it is a means to some end that is valuable for law. 38 Endicott (2000), 155. 40 2.7 A broader payoff It is sometimes said that vagueness can be valuable because it enables lawmakers to enact norms that are “flexible”, “open-ended”, “elastic”, or “adaptable”. In a recent paper, for example, Jeremy Waldron says that open-ended standards invoke people’s capacity for practical deliberation in both structured and unstructured ways, because they incentivize them to act according to their own practical thinking in certain circumstances. At the end of his paper, he has the following to say about the value of vagueness: [It] is good to focus on the need for legislative flexibility [in certain areas]… The adaptability of law, secured precisely by what others would call its indeterminacy, is … a valuable legislative resource and a respectful one too, for it works in tandem with the most sophisticated understanding of people’s powers of practical reasoning. 39 From this, it is reasonable to conclude that Waldron takes vagueness to be valuable for law on the grounds that enacting extravagantly vague norms makes the law flexible and open-ended, which in turn invokes a capacity the exercise of which realizes the dignity of human agents. Now, in an important sense, vagueness can allow for flexibility and open- endedness in the law. Sometimes, lawmakers – for various reasons – don’t want to settle beforehand how to classify borderline cases. It is better to deal with them in the courts, if and when they come up, since that allows decision-makers to evaluate, for each case, whether it (and sufficiently similar ones) should be within or outside the reach of the law 39 Waldron (2011), 82. 41 (and for what reasons). This is indeed appropriately characterized as flexibility or open- endedness and vagueness does help bring this about. However, this is not the only way in which it is valuable that the law is flexible or open-ended. Often, the relevant matter is that lawmakers do not want to draw sharp lines that make determinate cases fall on the wrong side of law’s boundary, vis-à-vis the relevant law’s rationale. If there is an indefinite variety of ways, for example, in which one can neglect a child, the worry is not that using more specific language will arbitrarily settle borderline cases of neglect, but rather – as we have seen – that some determinate cases of neglect will not be covered by law and that some determine cases of non-neglect will be. And this matter has nothing to do with vagueness, as such. Now, if we attend to why it is that terms like ‘reasonable’ invoke the deliberative capacity that Waldron values, we see that it is arguably the flexibility and open- endedness associated with multidimensionality, and not with vagueness, that matters to him: The legislature does not want [certain] norms pinned down to a precise and exact meaning that will govern all future cases; that would detract from the very elasticity that it is aiming at and it would detract from the sort of active consideration by citizens that it is seeking authoritatively to elicit. … If someone asks about the meaning of ‘reasonable’ or ‘appropriate’, all we can do is indicate that these are flexible all-purpose predicates of evaluation that invite us to consider a number of possible factors in an open-ended way. 40 40 Waldron (2011), 74; my emphasis. 42 Thus, I believe that Waldron – and whoever argues in a similar vein – wrongly attributes instrumental value to vagueness, at least in part. It is wrongly attributed, I think, because the two notions of flexibility and open-endedness are conflated. One is properly associated with vagueness and the other with multidimensionality. As we have seen, multidimensional terms are often extravagantly vague, as a matter of necessity, but this does not mean that (in those cases) vagueness has instrumental value if multidimensionality does. 2.8 An alternative explanation: Essentially Contested Concepts? It has been suggested to me, by Andrei Marmor, that what might really be going on – vis- à-vis why lawmakers cannot come up with specific replacements for (at least certain) normative terms in the law – does not have to do with either vagueness or multidimensionality, but rather with the fact that the associated concepts are essentially contested, i.e. concepts whose nature makes it the case that they don’t have stable criteria of application, although there will be paradigms associated with each such concept. The notion of essentially contested concepts was introduced by Gallie (1956) and is distinguished primarily by the following five necessary conditions: 1. The concept must be appraisive; that is, it must signify some sort of valued (or disvalued) achievement; 2. The concept must be internally complex; the value (or disvalue) associated with it applies to the concept as a whole; 43 3. The value (or disvalue) associated with the concept must somehow be explained by the way in which the parts contribute to the whole; 4. The achievement that is valued (or disvalued) must be sensitive to context in a way that cannot be predicted in advance – i.e., the concept must be “open”; 5. Each user of the concept must recognize that her own use of the concept is contested by other users in understandable ways. I take it that Marmor’s suggestion is that lawmakers use vague normative terms not primarily because they are vague or multidimensional, but because lawmakers unavoidably disagree on how to use the relevant concepts, and thus unavoidably disagree on how to specify their content for the purposes of legislation. I think that in the case of ‘neglect’, it is not contestedness that’s at work. To be sure, the concept of neglect satisfies each of 1–4, but that is presumably because these are conditions that apply to all multidimensional normative concepts. As for condition 5, I do not think that the concept of neglect is contested in the same way that art or democracy are (these are among Gallie’s target concepts in his paper). To say that it is would make the notion of contestedness much too thin. Condition 5 is pretty strong, as it places a constraint on what it takes to be competent with a concept. 41 My intuition about neglect is that it does not have such a constraint. Thus, I believe that neglect does not satisfy condition 5, which – according to Gallie – is the defining feature of essentially contested concepts. 41 See Ruben (2010) for a possible critique of this claim. 44 I do think, however, that a good case can be made that contestedness is at work in Endicott’s other favored examples. In negligence law, for example, the relevant concept is that of reasonable care, and in contract law that of fair and reasonable contract terms. As opposed to child care law, there indeed exist competing theories of the underlying purposes of negligence law and contract law, which make for different understandings of the respective concepts. People, for example, who see reasonability requirements in negligence law through the lens of corrective justice are bound to understand it differently from those who see it through the lens of economic analysis, or even from those who see it through the lens of distributive justice. The same applies to reasonableness in contract law, I think. And, arguably, a person who is familiar with these concepts does recognize that others may, and frequently do, contest the way in which she understands them. Still, I think that multidimensionality is part of the full story in these cases, too. The reason is that even if lawmakers did manage to settle on particular understandings of the relevant concepts, those concepts would satisfy conditions 1–4. And it is a consequence of condition 4 that those concepts are “open”, such that it is not predictable what falls under them in future circumstances. But this presents the same specification problem that we encountered in the case of neglect. There are indefinitely many ways in which people can create risk for other for other people, and indefinitely many ways in which one can unjustifiably do so. So there is no way to list the forms of behavior that are meant to be prohibited. Similarly, there are indefinitely many ways in which one can create a contract and indefinitely many ways in which one can unfairly or unreasonably 45 do so. As a result, there is no way to specify in specific terms what can or cannot be included when a contract is made. And this will be true even if lawmakers embrace, say, economic criteria for the application of these terms (since multiple factors, in unpredictable ways, determine the probabilities relevant to each case). Thus, even if unavoidable disagreement is a reason for lawmakers to use vague normative language in these cases, it is nevertheless multidimensionality that provides lawmakers with the means to regulate the relevant forms of conduct. Getting back to Endicott’s argument, it is not, I think, fatal for his overall view that he has not established that vagueness has instrumental necessity value. The reason is that I believe he is in certain respects right about the comparative value of vagueness. 3.1 Endicott’s argument from comparative value In his argument for the comparative value of vagueness in law, Endicott identifies three ways in which precision and vagueness may be (comparatively) valuable. I will start by listing and briefly explaining them in the abstract and then move on to a more detailed discussion. (1) From the perspective of the law’s subjects, the value of precision is that it lets them know their legal rights and obligations and the value of vagueness is that it can encourage desirable forms of behavior. Vague law can, Endicott says, give people incentive to be innovative in avoiding liability or to cost-efficiently 46 contract out of liability. He calls these values guidance value and private ordering value, respectively. (2) With respect to officials, the value of precision is that it lets them know how to apply the relevant norm and the value of vagueness is that it can appropriately delegate to them limited lawmaking power. According to Endicott, vagueness for example allows the courts to work out the details of the law, which may be preferable in domains where the courts satisfy certain competence conditions. He calls these values process value and power allocation value, respectively. (3) With respect to the purpose of the relevant norm, the value of precision is that it constrains the discretion of officials and the value of vagueness is that it allows officials to act in a way that is faithful to that purpose. In certain borderline cases, the background purpose, or purposes, of the law may dictate an outcome; unlike precision, vagueness affords a case-by-case determination of borderline cases, and so may sometimes be better suited to bring the law closer in line with its underlying purposes. Endicott calls these values constraint value and fidelity value, respectively. 42 It is by reference to these three types of values that lawmakers must determine whether the consequences of precision are worse than the consequences of vagueness. That is, these values form the “basis of calculation” in the Lawmakers’ Challenge. To be sure, vagueness is often worse than precision. When it comes to drinking and voting age, for example, it would make bad sense to formulate the law vaguely and have the courts or law enforcement agencies deal with borderline adults, by asking whether or not this or 42 Endicott (2011). 47 that person should count as one. Given the extent to which people would be at the mercy of officials and the cost of evaluating each borderline case, it is better to make a somewhat (but not totally) arbitrary cut-off point by mentioning a particular age – say, twenty-one, as was done in the National Minimum Drinking Age Act (1984), or eighteen, as was done in the 26 th Amendment of the US Constitution. 43 However, the value of precision can sometimes be negligible compared to the value of vagueness, or so Endicott argues. 3.2 Arbitrariness: The cost of precision may be greater than the cost of vagueness Endicott’s primary examples of vagueness comparing favorably to precision come from contract law, negligence law, and child care law. The Unfair Contract Terms Act (1977) in English contract law, for example, uses vague terms like ‘reasonable’ and so on. 44 The purpose, according to Endicott, was to “delegate to courts the power to determine the reasons for which limitation of liability clauses were to be permitted or not.” 45 This delegation is justified, he says, given that judges have special expertise in developing norms of contract law, that the doctrine of precedent is well suited to facilitate such development, and that the courts are in a position to hear the arguments of disputing sides. In other words, the courts are in a better position than lawmakers to develop the 43 23 U.S.C. §158 and U.S. Const., Am. 26, respectively. 44 It is reasonable to assume that what applies in the case of the Unfair Contract Terms Act in English law carries over to the unconscionability clause of the Uniform Commercial Code in US law. 45 Endicott (2011), 26. 48 details of contract law norms. The Act’s low process value is negligible, then, compared to its high power allocation value. Moreover, if the Act were replaced with a more precise alternative, the increased process value would be negligible compared to the low fidelity value – making the law easier to apply would make it significantly less faithful to its underlying purpose. The Act also provides service providers with an incentive to be innovative in avoiding liability. They may decide that it is in their best interest to accept liability or to take out insurance, or they may devise various contract clauses that (at least attempt to) make reasonable exclusions of liability. This sort of desirable creativity in the private domain would not have been encouraged by a precise scheme of regulation. The Act’s low guidance value is negligible, therefore, compared to its high private ordering value, Endicott says. In negligence law, vague schemes of regulation are necessary, according to Endicott, but there is still a choice about whether to try to incorporate some precise norms. It is here that comparative judgments must be made. Endicott argues that the uncertainty arising from relatively vague schemes may encourage private parties to avoid creating risks to others or to cost-effectively contract out of liability, thus assigning responsibility for action where it best belongs. Precise standards, on the other hand, do not encourage this sort of behavior, but instead encourage people to do no more than just what is required. In cases where private persons are in a better position than lawmakers to devise ways of avoiding harm to others, it is therefore better to refrain from including a 49 precise standard. In these cases, the low guidance value is negligible compared to the high private ordering value. Endicott is not explicit about how vagueness compares with precision with respect to officials, but it is plausible to take him to believe that the delegation of power in negligence law is justified for the same reasons as in the case of contract law – the courts are in a better position than lawmakers to develop the details of negligence law norms, due to the fact that judges have special expertise in developing such norms. If that is correct, the low process value is negligible compared to the high power allocation value. In child care law, vague schemes are – as in negligence law – unavoidable, on Endicott’s view. But, as before, it remains to be determined whether to incorporate certain precise norms – like, for example, a norm prescribing the minimum age of babysitters. Endicott argues that refraining from including such a norm may allocate responsibility to the persons who ought to have it, i.e. to the appropriate persons. Refraining from specifying a minimum age for babysitters, for example, appropriately assigns certain responsibilities to parents, which makes it the better option. In this case, and in similar ones, the low guidance value is negligible compared to the high private ordering value. As for the power allocating value of vagueness, it is low with respect to the courts, since they do not have any special expertise in this domain. However, it is presumably high with respect to social services officials (if they function well), although Endicott does not say anything about the latter. I would not go so far as to say that the 50 low process value is negligible compared to the high power allocation value, but aggregating all of the relevant values will presumably still yield that vagueness is preferable to precision. These examples constitute Endicott’s main case for the comparative value of vagueness. I think he is right about a number of things, but would like to voice a couple of worries. One concerns the highly contingent nature of economic incentives; the worry is that vagueness may give subjects incentive both to undercomply and to overcomply, both of which gives us reason not to overestimate the private ordering value of vagueness. The other concerns judicial competence; the worry here is that there is reason to doubt that courts have any special expertise with respect to either contract law or negligence law, which gives us a reason not to overestimate the power allocation value of vagueness. 3.3 Some remarks on Endicott’s framework Before I go on to discuss these worries, however, I would like to make a remark about Endicott’s threefold distinction of value. Value, as we saw, was distinguished by its relation to (i) law’s subjects, (ii) officials, and (iii) legislative purposes. The reason I want to start here is that I think the third category is undermotivated, and to some extent obscures rather than clarifies the Lawmaker’s Challenge. Endicott himself says very little about this third category, and the two associated values – constraint value and fidelity value – seem primarily to have to do with the presence or absence of private or official discretion/responsibility. I think it would be better to recognize just two domains of value 51 – value with respect to private persons and value with respect to officials – and to construe Endicott’s third category as identifying the source of those values. This would also eliminate the problem of associating a source of value with that third category. It seems to me best to say that guidance value, private ordering value, process value, and power allocation value are all determined by the (sometimes conflicting) purposes of the law, both law’s general (contingent and/or necessary) purposes – such as guiding action and assigning responsibility for action where it best belongs, etc. – and specific legislative purposes – such as ensuring that only mature individuals vote, and so on. Vagueness and precision have value, then, insofar as they promote law’s ends. And this seems right. 3.4 Vagueness and undercompliance Recall Endicott’s argument that vagueness in negligence law, under certain circumstances, gives private persons incentive to avoid creating risk to others or to contract out of liabilities in cost-effective ways. On Endicott’s view, then, one of the consequences of enacting vague law is that it gives efficient incentive, in the sense that it motivates agents to take courses of action that are considered desirable from the point of view of the law. In many circumstances, this may be right. Recent work, however, on the ways in which people actually behave in the face of vague law seems to suggest that the relevant uncertainty contributes less to such motivation than Endicott (and others) have predicted. Feldman and Teichman (2008), for example, report a study that they designed in order to test the hypothesis that the type of uncertainty with which a legal subject is 52 faced does not matter vis-à-vis the calculation of probability of liability (conditional on taking a particular action), which is – within the framework of Rational Choice Theory – the probability used to determine incentives given by law. 46 Rational Choice Theory has traditionally treated uncertainty about the law’s content as being on a par with – or fungible with respect to – uncertainty about enforcement. As Feldman and Teichman note, traditional theories assume that “[t]he expected sanction if illegality is certain and the probability of detection is fifty percent is identical to the expected sanction if detection is certain and the probability of illegality is fifty percent.” 47 Endicott is not explicit about whether he adopts this view, but it appears to me to be implicit in his arguments. Although the study and the reported findings should be taken with a grain of salt, it did show a significant positive difference between the percentage of people willing to follow through with a borderline illegal course of action (given certain enforcement) and the percentage of people willing to do so when the probability of enforcement is .5 (given certain liability). 48 Insofar as it is not desirable (from the point of view of the law) that people engage in borderline legal behavior, it is appropriate to call this statistically significant pattern of behavior a tendency to “undercomply” in borderline cases. Feldman and Teichman seem to have shown that we cannot expect uncertainty about the legal status of a given course of action to contribute to the expected sanction to 46 For a description of the study, see Feldman & Teichman (2008), 985–86. 47 Feldman & Teichman, (2008), 985. 48 See, in particular, Feldman & Teichman (2008), 1001–2. 53 the same extent that uncertainty about enforcement does. It is not clear how the uncertainty should be modeled, but we can nevertheless infer that if vague law is – in borderline cases – to able to give the same incentives as other (better understood) types of legal uncertainty, it seems that the sanctions for violating such laws would have to be raised. 49 This, however, is not feasible since most of the cases that reach the courts will presumably be clear ones, and imposing higher sanctions only for borderline behavior is remarkably problematic. Because it is not feasible to raise sanctions, it seems that, in borderline cases, the uncertainty associated with the legal consequences of the relevant course of action will often not give the incentive that Endicott identifies in his arguments. I take the findings reported by Feldman and Teichman to indicate that Endicott may not have taken account of circumstances under which people have the incentives that frustrate law’s purposes. I do not want to suggest that Endicott did not consider undercompliance at all, but it is reasonable to conclude, given the fact that he does not discuss it, that he thinks the relative frequency of undercompliance – prior to the consideration of the circumstances discussed by Feldman and Teichman – is negligible. But if Feldman and Teichman are right, then the relative frequency of such cases is presumably non-negligible. It is an open question to what extent this affects Endicott’s conclusion, but if Endicott is correct in his claim that the value of enacting a precise scheme of negligence regulation is negligible compared to the value of enacting a vague one, then it is 49 Unless people tend to behave irrationally in borderline cases, in which case it is not clear that any remedy is possible. 54 reasonable to expect that, although taking account of the possibility and disutility of undercompliance will reduce the expected value of the vague alternative, vagueness will still be preferable to precision. Nevertheless, if the reduction can plausibly be shown to be sufficient – or further negative consequences need to be taken account of – it will prove hard to argue that vagueness has significant advantage over precision. This, however, is not currently the case, and so I will take the consideration regarding undercompliance not to be fatal to Endicott’s argument, but to give us a definite reason not to overestimate the comparative value of vagueness. 3.5 Vagueness and overcompliance As noted by Calfee and Crasswell (1984), uncertainty about the content of the law may, under certain circumstances, give people incentive to overcomply, relative to standards supplied by economic theory. In the domain of negligence law, people may for example take excessive care (relative to these standards) – a strategy known as ‘safe harbor’ – in order to avoid negligence liability. That is, it can pay for agents to make sure that their behavior falls within the determinate anti-extension of the predicates used in the relevant legal texts, and they may do so in ways that have nothing to do with innovation or efficient contracting-out – they just spend an inefficient amount of resources on taking precaution. I take this to indicate that Endicott may have overestimated the proportion of cases in which people are given sufficient incentive to behave in ways considered to be desirable from the point of view of the law (like innovation and contracting-out). Now, I 55 do not want to suggest that Endicott overlooked safe harbor strategies. Instead, I take Endicott’s silence regarding this type of incentive to indicate that he thinks that the relative frequency of such behavior is negligible. Crasswell and Calfee’s work suggests that it may not be. As in the case of undercompliance, it is an open question to what extent the possibility of overcompliance affects Endicott’s conclusion. My view is that it varies by domain. But, again, if Endicott is correct in his claim that the value of enacting a precise scheme of negligence regulation is negligible compared to the value of enacting a vague one, then – as before – it is reasonable to expect that, although taking account of the possibility and disutility of overcompliance will reduce the expected value of the vague alternative, vagueness will continue to be preferable to precision. In any case, since I expect everyone to agree that undercompliance with law is generally worse than overcompliance (because overcompliance is generally more likely than undercompliance to promote law’s ends), the comparative value of vagueness is less affected by the current argument than by the undercompliance argument. Consequently, I will take the consideration regarding overcompliance not to be fatal to Endicott’s argument, but to give us some reason not to overestimate the comparative value of vagueness. 3.6 Do courts have special expertise? To close my discussion of Endicott’s view, I will explain my own main worry about his view, vis-à-vis the comparative value of vagueness, based on the doubts that I have that courts in fact have the sort of special expertise that is necessary for vagueness to have 56 significant power allocation value. Since Endicott’s claim about judicial competence in negligence and contract law provides much – though not all – of the basis for his claim that vagueness is valuable compared to relative precision in these domains, due to the high power allocation value, worries about judicial competence – relative to law’s purposes – are potentially serious. I will assume that we have an adequate idea about how courts are likely to behave in the face of vague legislation in contract and negligence law. The relevant question, thus, is not what courts are likely to do, but whether what we can reasonably take them to do is faithful to law’s purposes. The practical authority of expertise is, after all, based on the premise that the recommendations of the expert allow people to better act in accordance with the reasons that apply to them. So, in order for the power allocation value of vagueness to be as high in negligence and contract law as Endicott claims, it must be the case that the courts have special expertise with respect to the underlying purposes of those areas of law. I think there are at least two reasons to doubt that courts in general have such expertise. One is that it is not settled what the purposes of these areas of law really are. Is the goal of negligence law to provide a mechanism for corrective justice or for distributive justice, or perhaps to minimize the sum of the cost of accidents and the cost of avoiding them? And what about contract law? Legal practice does not seem to give settled answers to these questions. Thus, claims about special expertise on behalf of judges seem overly optimistic. 57 Another reason, downstream from the first one, is that claims about special judicial expertise are hard to justify given any of these possible fundamental goals. For example, courts are – on average – definitely not experts on principles of economic efficiency. As Craswell (1993) notes, since public utility commissions – whose dedicated job it is to ensure reasonable rates for utility services – have not managed to inspire confidence among economists, we have no reason to think that common law courts would do any better at determining general economic reasonableness. 50 Nor are courts experts on distributive or corrective justice. These are moral-political goals, and if there are indeed any experts on these matters, I see no reason to think that the courts have more expertise than the legislature does. Now, even if I am correct, this does not mean that all power allocation value is lost with respect to these areas of law. As we will see in the next chapter, it is sometimes better to deal with borderline cases when they come up than to try either to consider in advance possible borderline cases and work out sophisticated, more precise, alternatives, or to halt deliberation early on and opt for cheap and simple bright line rules. And, as Endicott rightly points out, the courts are well suited to do just that. But this is true even if judges do not have any special expertise in the relevant area of law. 50 See Craswell (1993), 20, footn. 36. For a discussion about the limited competence of courts to identify economically efficient outcomes in cases concerning incomplete contracts, see Hadfield (1994b). It should be noted, however, that in some legal systems there are courts specialized, for example, in commercial law, it which case the courts may have significantly greater expertise in the relevant matters than the average lawmaker (since the relevant judges will in most cases be well trained in matters economic). These courts, of course, have more expertise than the legislature – as an institution – only if lawmakers do not have reliable access to advice from comparable economic experts (assuming that economic efficiency is the underlying goal of commercial law). Thanks to Burkhard Schafer for bringing this point to my attention. 58 3.7 Summary In this section, I have argued that Endicott’s argument for the comparative value of vagueness is in many respects correct, although we should be aware of certain reasons not to overestimate that value. I argued that none of the issues I raised regarding the variable behavioral incentives that vagueness provides are fatal, but that they give us reason not to overestimate the private ordering value of vagueness for law. I also argued that common law courts do not in general have any privileged access to the underlying purposes of either contract law or negligence law, on any conception of those purposes, which gives us reason not to overestimate the power allocation value of vagueness. 59 CHAPTER TWO: SORENSEN ON VAGUENESS AND POWER-DELEGATION In “Vagueness has no function in law”, 51 Roy Sorensen argues that it is a mistake to think that vagueness has a constructive function in law, such as delegating limited law-making power to officials or eliciting certain kinds of desired behavior on the part of subjects. His argument, therefore, is aimed at exactly the sort of arguments discussed in the last part of the previous chapter (although Sorensen’s paper pre-dates Endicott’s). Sorensen argues that vagueness in the law merely appears to be functional, due to “a cluster of logical and linguistic errors” about its nature. In addition, he thinks that vagueness in the law often generates serious problems, and so cannot have a function in the law, on the understanding that something has a function in a system iff its presence is explained by how it serves a goal of the system. Thus, for Sorensen, the issue is whether vagueness is valuable vis-à-vis law’s aims, i.e. whether the presence of vagueness in the law does or can promote the common good. 52 The main problems he identifies are that (i) vague legislation is often used in a way contrary to the promotion of the common good, 53 and (ii) vague legal language, in genuine hard cases, forces serious judicial insincerity. 54 Sorensen believes, then, that he can explain away the evidence for the claim that 51 Sorensen (2001a). 52 Sorensen (2001a), 398. 53 Sorensen (2001a), 399. 54 Sorensen (2001a), 388–392. 60 vagueness has a function in law, and, further, produce evidence for the contrary claim that it has no such function. One of the main aims of this chapter is to examine one of Sorensen’s claims: that vagueness in the law – properly understood – cannot be justified by appeal to the value of power-delegation. Sorensen appears to think that the delegation of power to officials is justified only if these officials are in a better position to discover the right answer in the relevant cases. Since vagueness proper entails that there is no answer to be discovered, power-delegation will not be justified. Or so Sorensen argues. If he is right, then he will have taken away what is traditionally thought to be the main reason for thinking that vagueness can have a constructive function in law. I will begin by presenting Sorensen’s epistemic account of vagueness, his distinction between absolute and relative borderline cases, and his argument that absolute borderline cases do not have a constructive power-delegating function in law. I should note that the examples I present are mine, except for Brown v. Board of Education. 55 So in case any of them are unpersuasive, Sorensen is not at fault. I will then examine his argument, arguing that it is unsound. More specifically, I will argue that the following claim is false: Delegation of decision making authority, vis- à-vis borderline cases regarding something’s being F, is valuable only if the relevant delegates are in a better position to answer the question whether x is F than those delegating the power. Delegation of decision-making authority can be valuable, I argue, even if the relevant delegates are not in a better position to answer the question whether x 55 347 U.S. 483 (1954). 61 is F. The key to seeing why is to acknowledge that when faced with absolute borderline cases, the courts must engage with a related normative question – whether x ought, relative to the purposes of the law, to count as an F. I will try to show that, under certain circumstances, it is indeed better to let the relevant delegates answer the normative question – i.e. whether x ought, relative to the rationale of the relevant law, to count as an F. If I am successful, then there are situations in which delegation of decision-making authority is valuable even if the relevant delegates are not in a better position to answer the question whether x is F than those delegating the power. That is, if I am correct, then Sorensen’s argument is unsound. 1. Sorensen’s view It is standard to define vagueness with reference to borderline cases, in the sense that most writers on vagueness hold either that a term is vague only if it has possible borderline cases or that a term is vague if, and only if, it has possible borderline cases. 56 Since Sorensen is committed to the latter claim, and nothing in this particular paper hangs on which one of these claims is true, I will assume it here for the sake of accurately representing his argument. The nature of borderline cases is controversial, but most theorists would accept the characterization that these are cases in which there is inherent uncertainty regarding 56 Sorensen himself accepts the stronger claim that a term is vague if, and only if, it has possible borderline cases, but several authors on vagueness take the existence of borderline cases alone to be insufficient for vagueness; see e.g. Soames (1999). It is also important to distinguish between intensional vagueness – the possibility of having borderline cases – and extensional vagueness – actual borderline cases. Vagueness is properly characterized in terms of the possibility of borderline cases. This distinction will also play a role in determining the value of vagueness for law. 62 whether or not the relevant term applies. They will vary, however, in how they think this uncertainty ought to be understood. Some will say that the appropriate explanation is fundamentally epistemic (i.e. concerns what we can know), others that it is linguistic (i.e. concerns the rules of language), and yet others that it is at bottom ontological (i.e. concerns the way the world is). Sorensen thinks that the proper explanation of vagueness is primarily epistemic. A borderline case, unlike a clear case, is a case in which the question regarding whether a predicate ‘F’ applies to an object o has no knowable answer. The reason that the question has no answer is that there is nothing that makes either a positive or negative answer correct. For example, in case <F, o> = p and p is a borderline proposition, p will lack a truth-maker (and, thus, so will ~p). Here, a truth-maker is something “in the world” which makes the relevant truth true; Joe’s being bald, for example, makes the proposition that Joe is bald a true proposition (which in turn makes the sentence ‘Joe is bald’ a true sentence). In general, truth-makers are also the entities via which we come to know the truth-values of propositions. On Sorensen’s view, the borderline proposition p will still have a truth-value – it’s either true or false – but since we cannot, via any truth-maker (o’s being F), access this truth-value we are irremediably in the dark as to what that truth- value is. 57 This is what makes Sorensen’s approach epistemic. Since one cannot, even in principle, justifiably believe that a borderline sentence is true (or false), it follows that 57 Borderline cases, on Sorensen’s view, therefore do not involve truth-value gaps, but truth- maker gaps. 63 one cannot, even in principle, come to know the relevant proposition. In borderline cases, then, one’s lack of knowledge is irremediable. 1.1 Absolute borderline cases, relative borderline cases, and answering resources But, says Sorensen, there are borderline cases and there are borderline cases. He distinguishes between absolute borderline cases – in which the ignorance really is irremediable – and relative borderline cases – in which the ignorance is remediable – and he thinks that confusing the two may mislead one to believe that genuine vagueness is functional in law. 58 A term is genuinely vague iff it has absolute borderline cases, and a term has absolute borderline cases iff it has cases that are borderline given any means of answering “Is x F?” That is, in principle, there exist no resources for us to answer this question with respect to genuine borderline cases. Relative borderline cases, on the other hand, are not evidence of vagueness proper, since the ignorance in such a case is remediable by some means or other (which may or may not be available at a given time). Sorensen thinks that most borderline cases are relative ones. 59 Relative cases include e.g. ignorance as the result of a “measure-by-eye” test for whether an x falls in this or that size-category, which may be resolved by a “measure-by-ruler” test. An x may be categorized as borderline relative to the former test (i.e. the test doesn’t, for the relevant x, return an answer to the question “Is x F?”), while relative to the latter it may 58 See Sorensen (2001a), 392–400. 59 Sorensen (2001b), Ch. 1. 64 be categorized as a member or non-member of the relevant set (i.e. that test may return an answer to the question “Is x F?”). Imagine, for example, a couple of fishermen determining by eye whether a given fish is big or not. From what they can determine by eye, the fish may be borderline big. Yet, if later measured by a ruler, it may turn out that it does fall within the category big fish. If so, then the case was a relative borderline case. To take a legal example, consider the Pollution Prevention Act (1990), under which “pollution that cannot be prevented should be recycled in an environmentally safe manner whenever feasible.” 60 Due to the complexity of environmental matters, it is clear that there are a great many possible cases that lawmakers could not categorize as safe or unsafe, but that experts nevertheless could. That is, relative to the body of information available to the average member of Congress, there may be many cases that are borderline without being borderline relative to the evolving body of information available to an agency like the Environmental Protection Agency. Things are different when it comes to absolute borderline cases. Such cases include e.g. borderline patches of color in between, say, yellow and green (chartreuse). Here, arguably, the ignorance – as to whether a chartreuse patch is green or not – resulting from a “measure-by-eye” test cannot be remedied by appeal to another test. It is not to be expected, for example, that a measurement of the spectrum of reflected light waves – or any other imaginable test – will provide us with an answer to whether a patch of chartreuse really is green or not. Likewise, there will presumably be some absolute borderline cases of ‘environmentally safe recycling.’ In such cases, no body of 60 42 U.S.C. §13101-13109. 65 information – no matter how sophisticated – can settle whether it is safe or not. Absolute cases are resistant to any further inquiry. Measuring devices, bodies of information, sets of inference rules etc. are answering resources – considerations capable of furnishing an answer to a question of the form “Is x F?”. 61 Such a resource may e.g. provide a test for being F (e.g. if condition C is satisfied, then x is F). A test, of course, may be more or less fine-grained. Less fine- grained ones are likely to yield more borderline cases. And a case that is borderline relative to a certain answering resource may not be borderline relative to some other – perhaps more fine-grained – resource, as we have seen. 62 If there is some answering resource relative to which a case is not borderline, then the case is only relatively borderline. Cases in which no answering resource is capable of returning an answer, on the other hand, are absolute borderline cases. 1.2 Borderline cases and the delegation of power Sorensen thinks that legal theorists are really interested in the functionality of relative borderline cases, since the relevant ignorance can be remedied by appeal to an appropriate answering resource. Delegation of power to administrative agencies, e.g., by means of introducing relative borderline cases into the law is valuable, Sorensen claims, only because the relevant agencies have answering resources equipped to answer the 61 Sorensen (2001a), 392. 62 Sorensen says that formal systems in which a statement which is undecidable in one but provable in another show clearly this structural relationship. Thus borderline cases are spawned by incompleteness of the relevant answering system. See Sorensen (2001b), 23. 66 question whether x is F. In these cases, he says, legislatures typically make it the case that certain cases are borderline relative to the answering resources provided by law, i.e. the law will provide no test that will, for every x, provide an answer as to whether x is F. But this doesn’t mean that alternative non-legal answering resources aren’t available. “What is undecidable relative to current law,” he says, “may be decidable with the help of supplementary premises and procedures.” 63 If I understand Sorensen correctly, I take it that, for example, the legislature may in certain cases not have the resources to answer whether or not a recycling process is “environmentally safe” – i.e. the law does not tell us under which conditions such a process is safe – while scientific studies may indicate at least some such conditions. That is, it may be the case that there exists no specific regulation defining what is or is not environmentally safe, in which case an appropriate answer cannot be given by the law, yet it may be had by appealing to “extra-legal” answering resources. The important thing is this: if such resources are in principle available, then the relevant case isn’t an absolute borderline case – it is merely borderline relative to existing law. Sorensen’s own main legal example is Brown v. Board of Education, 64 a case in which the Supreme Court ordered that school desegregation occur “with all deliberate speed.” It did not provide any explanation of what this phrase meant and there was no precedent either to furnish legally clear cases of things happening “with all deliberate 63 Sorensen (2001a), 399. I am not sure why these additional answering resources cannot count as legal, assuming that officials are authorized to consult them, but for the sake of argument I’ll go along and agree that these are somehow “beyond the law”. 64 347 U.S. 483 (1954). 67 speed”. In some cases, of course, a superficial test may suffice, e.g. if it is clear that most everybody would think that desegregation clearly had, or had not, happened with all deliberate speed. Such a case does not call for a finer, more discriminating, test. But it may also happen that it is neither immediately clear that it happened with all deliberate speed nor that it didn’t. In this case, the courts may e.g. appeal to explanations of what it really is for something to happen with such a speed. But in some cases, i.e. genuine borderline cases, no acceptable theory or resource of any other type will furnish an answer to the relevant question. 65 It is these sorts of cases that typically bother philosophers of logic and language. So it seems that the phrase “with all deliberate speed” is genuinely vague. However, the flexibility that this phrase was supposed to facilitate does not, Sorensen says, have to do with genuine vagueness, i.e. it does not have to do with absolute borderline cases, which exist due to the inability to answer the question whether x is F. Rather, the utility of the vague phrase has to do with relative borderline cases and the ability to furnish answers to the relevant questions. In Brown, the flexibility that the phrase allows is relevant to cases in which the people closest to it are in a position to say whether a particular case of desegregation has, or has not, happened with the required speed. That is, the decision of the Supreme Court allows the law to rely on the people who deal with the actual case to furnish the relevant answer, which is arguably better than if the court had come up with some definition of the phrase, or used instead a more 65 Although, as Sorensen notes, Dworkin would deny this, since, on his view, all alleged hard cases concern relative borderline cases. 68 precise one. The complexity and unforeseeability of such matters as school desegregation make it the case that it is reasonable to delegate the task of finding out whether “x is F” to those with special experience and/or who are in close proximity with the actual situation. Such individuals arguably have better discriminatory abilities – that is, they are better equipped to classify the relevant cases – and so delegating certain legal powers to them is in an important respect analogous to utilizing a more fine-grained test for “F-ness”. Cases that may seem borderline relative to a rather limited body of information possessed by the legislature (or the Supreme Court) may turn out not to be borderline relative to a richer body of information available to the relevant officials. So relative borderline cases may well have a valuable power-delegating function in law. Absolute borderline cases, on the other hand, do not have any such power- delegating value, Sorensen says, since no answering resource is capable of providing an answer to whether the relevant xs are F. Hence, no one could ever be in a position to answer the relevant question and the delegation of power to agencies or the courts would be useless (for the purposes of settling the matter). The phrase ‘endangerment to the health of persons’, 66 for example, presumably has absolute borderline cases, but there is no benefit, Sorensen would say, in delegating power vis-à-vis absolute borderline cases of such endangerment since no one can be in a better position than any other with respect to determining whether such a case constitutes endangerment in the relevant sense. Not even the best experts. Same goes for absolute borderline cases of “with all deliberate speed.” 66 See the Federal Water Pollution Control Act (1948), 33 U.S.C. §1364(a). 69 Let me conclude this part of the chapter by summarizing Sorensen’s argument: P1. Delegation of decision-making authority, vis-à-vis borderline cases, is valuable only if the relevant delegates are in a better position to answer the question whether x is F than those delegating the power. P2. When x is an absolute borderline case of ‘F’, it is not the case that the relevant delegates are in a better position to answer the question whether x is F than those delegating the power. C. Therefore, when x is an absolute borderline case of ‘F’, it is not the case that delegation of decision making authority, vis-à-vis x, is valuable. Sorensen’s conclusion from all this, then, is that the notion of power-delegation cannot be used to show that vagueness – proper – can be a valuable feature of law. Sorensen’s distinction between absolute and relative borderline cases is quite interesting and the argument indeed threatens to take away what is traditionally taken to be the main reason for thinking that vagueness can have a constructive function in law. If the benefits of being able to delegate limited law-making power are had by introducing relative borderline cases into the law, rather than absolute ones, then power-delegation won’t be a benefit due to vagueness proper. Moreover, the distinction does not depend on Sorensen’s particular theory of vagueness – rather, it is one that any broad theory of vagueness should take seriously. More particular to our purposes here, it seems to me that P1 and P2 are jointly consistent with any viable theory of vagueness. Thus, we cannot avoid the 70 problem posed by Sorensen’s argument simply by arguing against his account of vagueness. It must be tackled some other way. Sorensen’s P2 seems safe enough. In fact, it is arguably entailed by any respectable theory of vagueness, whether epistemic, supervaluational, psychological, or indexical/contextual. As far as absolute borderline cases go, all theories will claim that for every cognizer x, every cognizer y, and every borderline proposition p, it is never the case that x is epistemically better situated than y vis-à-vis the truth value of p. P2 is simply a limited case of this general claim. Now, I agree with Sorensen that the distinction between absolute and relative borderline cases is often overlooked and that, as a result, value that is appropriately associated with relative borderline cases is erroneously associated with absolute ones. Still, I think that there are cases in which the delegation of power – and the resulting discretion – is justified even if the relevant delegates are not in a better position to find out whether x is F. In certain cases, it may well be better to leave the stipulation as to whether x is F – i.e. the decision whether x ought, for the purposes of the law, to count as an F – up to competent delegates. In the next part of the chapter, I try to explain some conditions under which this is the case. If I succeed, then P1 of Sorensen’s argument will be false. 2. The value of vagueness Sorensen recognizes that absolute borderline cases prompt judicial discretion, and characterizes this discretion as discretion to substitute the question whether x is F for the 71 question whether x should count as an F. In case a court is faced with a borderline case of, say, “business establishment” – as it was in the case of Curran v. Mount Diablo Boy Scouts 67 – the courts may relativize the concept in question to suit the purposes of the law, Sorensen says. 68 In Curran, for example, such relativization led to the Boy Scouts being counted as a business establishment. If I understand Sorensen correctly, he has in mind the practice of the courts to use phrases of the following sort: ‘for the purposes of [such-and-such a statute], x is F’. The Boy Scouts, then, presumably counted as a business establishment for the purposes of California’s Unruh Civil Rights Act. 69 To taken an example from the constitutional realm, many have argued (and the Supreme Court has agreed) that for the purposes of the First Amendment, flag-burning counts as speech (and so is protected). 70 These, of course, are not answers to whether x is F, but to whether x should legally count as an F relative to some particular aims of the law. Thus, in absolute borderline cases, the question is – implicitly or explicitly – changed. Moreover, it is changed to one that in many cases has better prospects of being answered (i.e. a non-zero probability, unlike ‘Is x F?’). The reason I bring this up is that I have a hard time seeing why this question- changing discretion may not be valuable. In fact, I believe that legislators often rely on this pervasive judicial response. When legislatures deliberately use vague language, they 67 17 Cal. 4th 670 (1998). 68 Sorensen (2001a), 414. 69 Cal. Civ. Code §51. 70 See Texas v. Johnson, 491 U.S. 397 (1989), and U.S. v. Eichman, 496 U.S. 310 (1990). 72 are not asking the courts to find out whether absolute borderline Fs really are Fs; rather, they are asking the courts to do exactly what Sorensen describes, to engage with the normative question whether x ought – relative to the purposes of the law – to count as an F. And I fail to see any argument in Sorensen’s paper to the effect that the judicial response to this legislative “request” cannot, under certain circumstances, be valuable. And if it can be shown that discretion to change the question can be valuable, then – since such discretion is due to absolute borderline cases – vagueness proper will have a constructive function in law. One issue which may be in the back of Sorensen’s mind is, I think, the following. It seems that if it is valuable to let the courts figure out whether x should count as an F, then that is because there is an answer to whether x should count as an F. And if there is an answer to whether x should count as an F, then it seems that there is a fact of the matter whether x is a case that ought to be regulated, on the narrow understanding that a case c is regulated by a rule R iff c is included in R’s domain of application. But, and here is the catch, if the immediate aim of the lawmakers is to regulate everything F and x is absolutely borderline F, then there is arguably no fact of the matter whether x ought or ought not to be so included. Thus, combine the relevant immediate regulatory aim with the absolute borderline status of x and you may have a hard time explaining how it could be valuable to let the courts decide whether or not it should be regulated. If there is a 73 truly arbitrary distinction to be made, it seems better simply to make it at the legislative level. 71 I suspect that the answer here lies in the combination of two relatively common aspects of lawmaking and adjudication: first, in the fact that many vague laws don’t wear their justifications “on their sleeves,” to borrow a phrase from Schauer (1991), 72 and, second, in the question-changing element of the judicial response to absolute borderline cases. I do think that Sorensen’s arbitrariness worry has some weight in cases in which the enacted law “embodies” – so to speak – its own primary justification. To borrow and adapt one of Schauer’s examples, compare e.g. a law that prohibits the presence of “all annoying agents” on public buses with a law that prohibits the presence of “all dogs” on such vehicles. Following Schauer, we can say that the former rule is couched in terms of its own justification while the latter is not. A law prohibiting the presence of “all annoying agents” on public buses is obviously vague, mainly as a result of the multiple incommensurable factors that may 71 There is a potentially important issue here regarding the question whether some case x ought to be regulated. When we ask whether x ought to be regulated, we can be asking (at least) two things: (i) ought it be the case that there is some rule R such that x is regulated by R?, or (ii) for some particular rule R and case x, ought it be the case that x is regulated by R? The former question concerns whether x ought to be regulated by law (at all), while the latter concerns whether x ought to be regulated by some particular law. It is possible that, in some cases, this distinction matters – since lawmakers have an unconstrained choice of rules, while the discretion of the courts is limited to precisifications of particular (already enacted) rules – but I will not pursue these matters here. 72 See Schauer (1991), 50, fn. 15. 74 determine whether or not an agent is annoying. 73 And if the rule itself “exhausts” its justification/rationale, then it does appear – as Sorensen claims – that resolving the relevant borderline cases is an arbitrary matter. More generally put: if a rule R, the primary predicate of which is ‘F’, “wears its justification on its sleeve”, then it is Fness that determines whether or not x ought to be regulated by R. In case x is borderline F, therefore, the choice to include or exclude x in R’s domain of application will be arbitrary. And if that is the case, then there is no answer to the question whether or not x ought to be regulated by R. Or so the worry goes. There are at least two things to note here. First, it is rare that laws wear their justifications entirely on their sleeves; for most laws, it will be possible to point to some further aims that they are intended to promote, which may suffice to settle certain cases. In addition, the facts that determine whether or not some x ought to be regulated by some rule R are not exhausted by substantive normative facts pertaining to (the properties of) x – in some cases, there may be procedural reasons for favoring the inclusion or exclusion of x from R’s domain of application. Second, and more importantly, many vague laws don’t at all wear their justifications on their sleeves. The background justification of the well-worn fictional “No vehicles in the park” statute, for example, may suffice to answer the question whether a skateboard ought to count as a vehicle. In case the background justification includes, say, to minimize noise pollution in the park, then that may point in favor of skateboards counting as vehicles “for the purposes of the statute.” If, on the other 73 But also because it may be borderline both whether or not a thing counts as an agent and whether or not an object counts as a public bus. 75 hand, the aim was to reduce exhaust pollution and/or risk of fatal accidents, then that seems to clearly count in favor of not counting them as vehicles. Now, it may of course happen that the relevant background aims do not determine a verdict either way – that is, they may fail to resolve the matter. Consider the normative question once again: Ought x count as an F? Let’s semi-formalize the question in the following way (‘O’ here reads ‘It ought to be the case that’): O(x counts as an F)? We can then construct a complex predicate [λx O(x counts as an F], which may have borderline instances, both relative and absolute – just as F, by hypothesis, does. This, however, is not a theoretical problem, and should in fact be expected. I would be highly skeptical of an account that predicted otherwise. My claim is not that looking to legislative purpose can in all borderline cases guide judicial decision – that is, I am not suggesting a uniform practical decision-procedure for all borderline cases. Nor need I suggest one in order to reply to Sorensen. I am merely pointing out that while it may be arbitrary with respect to Fness whether x ought to count as an F, it may happen – and probably often happens – that it is not arbitrary with respect to the purposes of the law. I should note that the law seems to embrace this question-changing resolution strategy in other varieties of hard cases too, even ones in which the particular behavior is within the determinate anti-extension of the relevant predicate. It seems appropriate to understand both the doctrine of transferred intent and the doctrine of willful blindness as prescriptions to treat particular cases in certain ways, for the purposes of the law. Take for example the case of a person A who intends to harm another person B, but who – by accident – ends up harming yet another person C. In such a case, the law often treats A as 76 having intended to harm C. This is known as transferred intent (or transferred malice, in English law). Consider the case of Bradshaw v. Richey, 74 in which Richey was found guilty of aggravated felony murder on the basis of this doctrine. Richey had intended to kill his ex-girlfriend and her boyfriend but ended up killing a little girl, Cynthia Collins, instead. That is, Richey’s intended violation of the law failed and instead he killed a person he had no intention of killing. Still, he was found guilty of murdering the girl, a felony that requires intent. The doctrine of transferred intent arguably operates on the premise that the behavior to which it applies is just as bad as the intended behavior. Thus, the question facing the courts is a version of the normative question: ought x – for the purposes of the law – to count as an F? The purpose, in this case, is to prevent or punish behavior that is equally bad as intended violations of the law (assuming attempt). In Richey, for example, the question was whether Richey’s unintentional killing of young Cynthia Collins was – given the fact that it was the result of an attempt to murder another person – equally bad as the intended murder. The doctrine of willful blindness operates on a similar premise: the doctrine is supposed to apply if a defendant’s deliberately ignorant violation of the law is just as bad as a knowing violation. In US v. Jewell, 75 for example, Jewell was convicted for knowingly transporting marijuana across the US-Mexico border in his car, despite the fact that he deliberately avoided positive knowledge of what was in the car’s 74 546 U.S. 74 (2005). 75 532 F.2d 697 (1976). 77 compartment. In this case, too, the court opted to “change the question”, from a factual to a normative one: ought Jewell’s violation – relative to the law’s purpose of preventing or punishing behavior that is equally bad as knowing violations of the law – to count as a knowing violation? The reason I am mentioning the doctrines of transferred intent and willful blindness is that they arguably show that the question-changing strategy shows up in different sorts of hard cases. That is, it is not a strategy particular to borderline cases, although the particular context of the normative question is different because – in the typical cases where transferred intent or willful blindness are taken to apply – the relevant behavior is within the determinate anti-extension of the relevant predicates and the background purpose is a quite particular one. In vagueness-related cases, the courts can of course also ask if the relevant behavior is just as bad as behavior that is within the determinate extension of the predicate in question, but I see no reason to think that the resolution of borderline cases is restricted to that particular purpose. There may be all sorts of reasons why x ought – for the purposes of the law – to count as an F. Now that we have relieved the arbitrariness worry, by proposing a strategy grounded both in normative theory and in practice, we can ask whether the normative question is ever better left to the courts or administrative agencies. If it is, then Sorensen’s argument will be unsound. 78 2.1 Evaluating the value of vagueness in terms of “better than” I want to start this section by focusing on what kind of questions it is appropriate to ask when we are considering the value of vagueness in the law. The reason is that Sorensen’s framework for talking about law and vagueness can sometimes make for misleading questions. In particular, Sorensen’s strict notion of function is poorly suited for asking the appropriate questions regarding the value of vagueness in the law. His criterion of function was this: something has a function in a system iff its presence is explained by how it serves a goal of the system. This prompts an unhelpful way to think about the value of vagueness, since we are in effect forced to ask whether the presence of absolute borderline cases is valuable relative to the aims of the law. It is unhelpful since it seems odd from the get-go to even entertain the thought that somehow the presence of a particular absolute borderline case in the law is a good thing. What value does a particular hard case promote? This way of conducting the inquiry makes claims about the value of vagueness seem dubious from the start and so should be rejected and substituted for a more sensible way of asking the relevant questions. If we don’t, we run the risk of not getting to the real issues. I propose that we evaluate the value of vagueness in law using the comparative phrase ‘better than’, which underlines the fact that the value of using terms that have absolute borderline cases depends on the alternative options available to the lawmakers. The appropriate question is, I think, whether it is better to leave a law vague and let the courts deal with borderline cases than to have the legislature work out a more precise alternative. Or, to put it another way: is it ever better to leave it to the courts (or 79 administrative agencies) to answer the question whether x ought to count as an F (assuming x is a genuine borderline case)? As for Sorensen, he must think that an answer can be affirmative only if the relevant delegates are in a better position to answer it than the legislators. This, however, would also be a misleading way of framing the issue. That A is in a better position than B to answer a question does not – as the phrase seems to superficially suggest – entail that A has better knowledge than B. A can also be in a better position if A has better tools than B for finding an answer, or if A’s cost of finding the answer is lower than B’s. This latter notion of answering cost may be particularly helpful, I think, in understanding the benefit of letting delegates answer the question whether a borderline x ought to count as an F. Since the answering cost we are concerned with here has to do with a practical question – whether to count x as an F – we can presumably substitute it for the more familiar notion of cost of deliberation. 76 Now, in some cases it is clearly not better to let delegates decide whether or not particular borderline cases ought to be regulated. When it comes to drinking age, for example, it would make bad sense to formulate the law vaguely (“Only adults are permitted to purchase and publicly possess alcoholic beverages.”) and have the courts or law enforcement agencies deal with borderline adults, by asking whether or not this or that person should count as one. The overall cost of deliberation would be enormous. Better to avoid that mess and make a somewhat (but not totally) arbitrary cut-off point by 76 For the time being, I am including under the heading of ‘cost of deliberation’ the appropriateness of the deliberator. That is, the extent to which the deliberator is inappropriate for the task of deliberating will figure in the cost of deliberation. Ultimately, this cost should be considered separately, but I’ll leave that discussion for a later occasion. 80 mentioning a particular age – say, twenty-one. That may of course be done at least in two ways, either by leaving out any mention of adults in favor of explicit age, as was done in the National Minimum Drinking Age Act (1984), 77 or by defining ‘adult’. (What the courts do in response to vague law amounts, it seems to me, roughly to an incremental version of the latter strategy, as evidenced by the use of phrases like the aforementioned ‘For the purposes of this statute…’) However, in other cases, it may be better to stick with a vague formulation and let the courts change the question to whether x ought to count as an F relative to the purposes of the relevant law. Let me try out a fictional example. The federal code concerning drive-by shooting related to major drug offenses specifies, among other things, that if one fires a weapon into a crowd “[causing] grave risk to any human life”, then one shall receive such-and-such punishment. 78 The phrase ‘grave risk’ here is vague, mainly in virtue of the term ‘grave’. As a borderline case, we can imagine that someone, in relation to a major drug offense, fires rubber bullets into a crowd and that, given the nature of the projectiles and the circumstances of the shooting, it is indeterminate whether the shooter caused grave risk to human life. In other words, the shooting constitutes an absolute borderline case of the behavior prohibited by the statute. Now, it seems to me that it would be a bad move for the legislature – whose aims presumably include the reduction of danger to innocent bystanders and of the public fear 77 23 U.S.C. §158. 78 18 U.S.C. §36(b1). 81 that drive-by shootings induce – to try to work out a precise alternative statute. In addition to the fact that it would be very cumbersome, it is very likely, due to the multiplicity of factors that make for the absence or presence of risk, that any attempt to come up with a formula that is neither excessively over- nor underinclusive is bound to fail. 79 Also, substituting the vague law for a more precise one would mean to exclude in advance some unpredictable borderline cases and include some, thereby eliminating the possibility of determining – when they come up – whether the law has an interest, vis-à- vis the purpose of the code, in regulating them. To put this point more generally, it seems to me a bad bargain, all in all, to have legislators try to deal with intensional vagueness instead of letting the courts deal only with extensional vagueness. No matter how many concrete cases end up in the courts, they will always be far fewer than even the most modest sets of possible absolute borderline cases. That is, the cost of deliberation associated with determining in advance what to do regarding possible borderline cases will in most cases greatly exceed the cost of deliberation associated with determining what to do regarding actual borderline cases when they come up. Although, as we saw above, sometimes the foreseeable actual borderline cases are so many that it is better to draw a simple bright line, as long as this “cheaper” rule is sufficiently acceptable (with respect to the relevant legislative aims). But sometimes, such a strategy is not feasible and the more feasible option is to leave the 79 I say ‘excessively’ since, as Schauer (1991) notes, most laws are arguably both over- and underinclusive to some extent. 82 language vague and let the courts deal with extensional vagueness on a case-by-case basis. If this is correct, then Sorensen’s argument against the claim that vagueness proper can have a power-delegating value is unsound. In particular, it is not the case that the delegation of limited law-making power is justified only if the individuals or entities to which the power is delegated are in a better position to find out whether x is F. Also, it seems that – at least in some cases – we can, using Sorensen own strong notion of function, say that the occurrence of absolute borderline cases in the law is indeed explained by how they contribute to the promotion of the aims of the law. They do so not by delegating to the courts the task to find out whether x is F. Rather, they do so by delegating to them the task to find out whether x ought, relative to the purposes of the relevant law, to count as an F. The above example regarding drive-by shooting is of course very limited. But there are other, more wide-ranging, cases in which similar reasoning applies. Take for example the general requirements of culpability, as defined by the Model Penal Code, section 2.02. In sub-section 2(a), 80 the vague ordinary term ‘purposely’ is defined partially in terms of awareness, belief, and hope. Now, there will be both relative and absolute borderline cases of persons purposely violating the law in the ordinary sense of ‘purposely’. The definition used in the model penal code is arguably a precisification vis- à-vis relative borderline cases. It gives us a coarse-grained test for the relevant quality, i.e. the definition supplies a somewhat helpful answering resource. But I fail to see that 80 M.P.C. §2.02(a). 83 the definition helps eliminate absolute borderline cases in any meaningful sense. At best, it trades one set of cases for another, since ‘awareness’, ‘belief’, and ‘hope’ are at least as vague as ‘purposely’ (I’m not assuming that we have a clear theoretical way of comparing vagueness – I only mean that the following holds: If ‘purposely’ is vague, then so are the other terms). The question I want to ask here is this: Which of the following options seems better? (1) The legislature tries to preempt the occurrence of absolute borderline cases by adding further, non-circular, definitions of ‘awareness’, ‘hope’, and ‘belief’, stipulating meanings in order to eliminate absolute borderline cases of purposely violating the law. (2) The legislature – at some fairly coarse-grained level – leaves the provision vague and gives judges the discretion to ask whether actual absolute borderline cases that reach the courts ought – relative to the purpose of the code – to be regulated and thus to be counted as being purposeful. It seems to me that (2) is the better option. Option (1) is feasible only if the lawmakers are able to evaluate a significant number of possible borderline cases or if it makes sense to halt deliberation and opt for a cheap and simple bright line rule. The former isn’t feasible given the high cost of deliberation and limited cognitive resources of normal human beings. And the latter isn’t feasible given the high likelihood of error – i.e. of drawing an unacceptable boundary, relative to the relevant regulatory aim. Better to opt 84 for (2) and let the courts deal with borderline cases incrementally, by dealing with actual borderline cases of purposeful violations as they reach the courts. If what I have said is correct, then Sorensen’s argument against the power-delegating function of vagueness is unsound. Vagueness may indeed have such a function, and, moreover, this function seems to be significant rather than marginal. I do agree, however, that many power-delegating instances are valuable because there is hope that the delegates will in fact discover whether x is or is not F, and that we should be careful to distinguish the value of relative borderline cases from that of absolute ones. 2.2 Taking stock I began this chapter by introducing Sorensen’s epistemic account of vagueness and his distinction between absolute and relative borderline cases, and explained his argument that absolute borderline cases do not have a constructive power-delegating function in law. The argument was summarized in the following way: P1. Delegation of decision-making authority, vis-à-vis borderline cases, is valuable only if the relevant delegates are in a better position to answer the question whether x is F than those delegating the power. P2. When x is an absolute borderline case of ‘F’, it is not the case that the relevant delegates are in a better position to answer the question whether x is F than those delegating the power. C. Therefore, when x is an absolute borderline case of ‘F’, it is not the case that delegation of decision making authority, vis-à-vis x, is valuable. 85 My reply was to argue that P1 is false. Delegation of decision-making authority can be valuable even if the relevant delegates are not in a better position to answer the question whether x is F than those delegating the power. The key to seeing why, I claimed, is to acknowledge that when faced with absolute borderline cases, the courts must engage with a related normative question – whether x ought, relative to the purposes of the law, to count as an F. I tried to show that, under certain circumstances, it is indeed better to let the relevant delegates answer the normative question – i.e. whether x ought, relative to the purpose of the relevant law, to count as an F. It is better when both of the following options are worse: (i) lawmakers consider in advance possible borderline cases and work out a sophisticated, more precise, alternative; (ii) lawmakers halt deliberation early on and opt for a cheap and simple bright line rule. These options are typically worse if option (i) incurs extravagant cost of deliberation and option (ii) is likely to result in a rule that is unacceptable relative to the relevant regulatory aim. If I am correct that there are situations in which neither of these options is feasible, then there are situations in which delegation of decision-making authority is valuable even if the relevant delegates are not in a better position to answer the question whether x really is F. Before I move on to the next chapter, I would like to relatively briefly address another argument due to Sorensen, one that also connects nicely with the discussion in the previous chapter on Endicott’s arguments for the value of vagueness in law. The foregoing discussion about vagueness and power-delegation concerned the value of 86 vague law in the domain of legal officials – i.e. of legislators, administrative officials, and judges. I want to now turn to the value of vague law in part as it concerns the domain of legal subjects, replying to Sorensen’s argument – also found in Sorensen (2001a) – for the claim that uncertainty due to vagueness-related ignorance has no constructive function in law. 81 My aim in this particular section is not to give a counterargument to Sorensen’s claim, but just to that show it is insufficiently supported, due to an unsound intermediate argument. The reason is that, in chapter one, I already discussed – and, with certain important qualifications endorsed – Endicott’s argument for the claim that the relevant type of uncertainty can – under certain circumstances – be valuable in law, and so I take what I said there to provide an implicit counterargument to Sorensen’s claim. 3. Vagueness, uncertainty, and desirable behavior A number of authors argue that vagueness can be valuable to law since it entails (or consists in) uncertainty. On such views, vagueness can have a function in law because uncertainty can. The basic idea underlying the arguments for these views is that vagueness generates unpredictability that somehow promotes certain important legal ends. Hadfield (1994a), who is Sorensen‘s main target, for example argues that – under certain circumstances – adopting vague rules over precise rules will, over the long run, produce legal standards that are comparatively close to optimal ones, partly in virtue of 81 See Sorensen (2001a), 410–11. 87 the varying behavioral incentives that vague rules give to those subject to the relevant laws. When the law is precise, Hadfield says, subjects can, relative to all possible behavior, assign the probability 0 or 1 to the proposition that they will incur liability. This, of course, is not the same as being held liable, the probability of which may be somewhere between 1 and 0 due to factors like imperfect enforcement etc. – that is, to factors that are the product of official discretion. In the case of vague legal rules, Hadfield says, this uncertainty is amplified, partly because it is not possible – even in principle – for people to assign, relative to all possible behavior, the probability 0 or 1 to the proposition that they will incur liability and partly because official discretion is inevitable in borderline cases. The mainstream view among law and economics scholars working on unclear law and deterrence used to be that the effect of the type of uncertainty described in the previous paragraph is that it leads to undercompliance, but recent work in that field seems to show that – if the uncertainty resulting from the enactment of vague rules is not too great – it may instead, at least under certain conditions (especially if agents are risk averse), lead to overcompliance. 82 In some circumstances, Hadfield says, “[t]he decrease in probability of being held liable gained by overcompliance may be substantial enough to offset the personal cost of overcompliance; it can therefore make sense for an 82 See Calfee & Craswell (1986). 88 individual to overcomply in order to reduce expected damages.” 83 This behavioral strategy is known as ‘safe harbor’. However, Hadfield says, vague rules don’t lead exclusively to undercompliance or overcompliance – the behavioral incentive that they give is more nuanced than that. What vague rules do, and which may be desirable on Hadfield’s view, is to correlate small changes in behavior with small changes in the probability of liability. The cost of compliance along with the probability of liability will, for each subject, determine whether complying is the optimal thing to do. As Sorensen puts it, “[t]he uncertainty of the vague rule [is supposed to create] a spectrum of expected utilities that varies gently with the cost of compliance.” 84 Precise rules, on the other hand, steer people either into the same safe harbor or, if that is too costly, causes people to ignore the rule altogether. The result of such legislation will be “excessive compliance or nil.” 85 In certain domains of law – in particular in domains in which it is likely that, if lawmakers would choose precise rules, they would choose sub-optimal ones – it is better, Hadfield argues, to enact vague law. For example, in cases in which there is “uncertainty among lawmakers and regulators about what constitutes optimal behavior,” enacting a vague law encourages a “small degree of experimentation by individuals acting on the margin,” which “introduces variability into the types of cases heard by courts or 83 Hadfield (1994a), 544. 84 Sorensen (2001a), 410. My emphasis. 85 Hadfield (1994a), 545. 89 regulatory agencies” and thereby “gives these institutions more information about the nature of the activity regulated, thus improving their ability to develop law.” 86 When it comes to choosing, for example, between strict liability – which is a bright-line rule – and a rule of negligence – which is a vague rule – for products liability, there are good reasons to choose the latter, Hadfield (1994b) argues. A strict liability rule to a large extent ignores the various factors that determine the most efficient outcome in particular cases – such as relative costs and benefits of precautionary measures and the potential for the manufacturer to innovate new safety mechanisms – by introducing a “blanket standard”, while a rule of negligence allows sensitivity to those factors by “[directing] the court to explicitly weigh the expected costs and benefits of preventing an accident.” 87 However, it isn’t necessarily that the vague standard – as is – is the optimal standard. Rather, what it allows for - and the precise one doesn’t – is that, in practice, it is eventually “reduced to collections of refined rules when applied to any particular subset of facts.” 88 That is, over time, the vague standard is precisified in a principled manner – a manner which, at least in some domains, eventually makes for a standard (or a collection thereof) that is closer to an optimal standard than any precise standard chosen at the outset would have been. 86 Hadfield (1994a), 549. 87 Hadfield (1994b), 610. 88 Hadfield (1994b), 610. 90 As we saw in the previous chapter, I do have some worries about this line of reasoning. But since I have already discussed these worries in relation to Endicott’s argument for the comparative value of vagueness for law, I will let it suffice here to briefly explain what Sorensen thinks is – in general – wrong with arguments like Hadfield’s and then go on to provide what I take to be a rebuttal of his argument. 3.1 Sorensen on vagueness and uncertainty in law Sorensen’s main objection to views like Hadfield’s – views that take vagueness to be valuable in law because uncertainty is – is that, due to the tendency of the courts to treat like cases alike, whatever unpredictability is generated by vague legislation will (unlike proper randomization) not be sufficient to promote the relevant aim. Judges, he says, are poor randomizers, even by informal standards, both because normal human beings are bad at producing sufficiently random-like results and – more importantly – because judges will attempt to decide problem cases in a principled fashion. Sorensen’s argument is brief and he does not provide any detailed explanation of why vague law fails to generate sufficient unpredictability. However, the argument raises important questions about one of the primary reasons that vagueness has – especially by scholars influenced by law and economics – been taken to be valuable to law, and so I think that it merits careful dissection. If we take care to reconstruct Sorensen’s argument, we will be able both to identify certain problematic assumptions that he makes and to see why – despite the attempt of judges to decide problem cases in a principled fashion – vague law can still often generate uncertainty sufficient for it to be valuable. 91 We can reconstruct Sorensen’s main argument in the following way: P1. Vagueness-related uncertainty is valuable in law only if the resulting unpredictability regarding probability of (criminal or civil) liability gives subjects incentive to engage in desirable behavior. P2. It is not the case that the unpredictability regarding probability of (criminal or civil) liability resulting from vagueness-related uncertainty gives people incentive to engage in desirable behavior. C1. It is not the case that vagueness-related uncertainty is valuable in law. P1 seems fairly safe, and I will treat as such here. I will therefore – like Sorensen – focus on P2 and argue that it is false. Sorensen’s argument for P2 is as follows: P3. The unpredictability regarding probability of (criminal or civil) liability resulting from vagueness-related uncertainty gives people incentive to engage in desirable behavior only if judges/courts behave sufficiently like unbiased random-decision generators. P4. Judges/courts do not behave sufficiently like unbiased random-decision generators. C/P2. It is not the case that the unpredictability regarding probability of (criminal or civil) liability resulting from vagueness-related uncertainty gives people incentive to engage in desirable behavior. 92 On the face of it, it may seem that P3 is dubious and that P4 is plausible. In fact, however, I think it is the other way around. We see why if we examine the arguments for P3 and P4. Let‘s start with P3: P5. The unpredictability regarding probability of (criminal or civil) liability resulting from vagueness-related uncertainty gives people incentive to engage in desirable behavior only if it is sufficiently hard to detect patterns in decision-making in borderline cases. P6. It is sufficiently hard to detect patterns in decision-making in borderline cases only if judges/courts behave sufficiently like unbiased random-decision generators. C/P3. The unpredictability regarding probability of (criminal or civil) liability resulting from vagueness-related uncertainty gives people incentive to engage in desirable behavior only if judges/courts behave sufficiently like unbiased random-decision generators. P5 is the safe premise here. P6 may seem less safe, but once we acknowledge the sufficiently qualification, it turns out to be very plausible. Judges/courts don‘t have to really be unbiased random-decision generators, they simply have to appear to be. That is, as long as decision-making patterns are sufficiently hard to discern for the relevant group of people, the relevant requirement is satisfied. Thus, it suffices that judges/courts be so- called pseudo-random-decision generators relative to the relevant perceiver. In casinos, for example, pseudo-random-number generators are used because – for the purposes of running a gambling facility – they behave sufficiently like genuine unbiased random- 93 number generators: they make it hard enough for the majority of people to detect patterns in the relevant set of outcomes. P3, therefore, is fairly plausible. Turning to P4, I think we start to see where the problem lies with Sorensen‘s argument. It is also here that we get to the issues that are most interesting from a legal- theoretic perspective. Here is a reconstruction of the support that Sorensen provides for P4, i.e. for the claim that judges/courts do not behave sufficiently like unbiased random- decision generators: P7. When judges/courts make an effort to treat like cases alike, they are trying to act in a conscientious, principled fashion. P8. When judges/courts try to act in a conscientious, principled fashion, they do not behave sufficiently like unbiased random-decision generators. C5. When judges/courts make an effort to treat like cases alike, they do not behave sufficiently like unbiased random-decision generators. P9. Judges/courts make an effort to treat like cases alike. C/P4. Judges/courts do not behave sufficiently like unbiased random-decision generators. I think that P8 is where Sorensen’t argument goes wrong (I will treat P7 and P9 as unproblematic). I don’t think that the way in which judges/courts attempt to act in a conscientious, principled fashion entails that they don’t behave like pseudo-random- decision generators relative to subjects – even relative to subjects well-informed about precedent. My main reason for thinking so is twofold. One reason is that even if the courts make an effort to treat like cases alike, the doctrine is – as such – substantially 94 empty. 89 That is, the doctrine cannot be used to predict the decision of a court unless the predictor has good evidence of the criteria of similarity that the court will use in determining whether or not two cases are alike. However, since the criteria of relevance are contestable and often controversial, the doctrine of treating like cases alike alone won’t ensure that courts fail to behave with sufficient unpredictability in borderline cases. Further, although in rare cases, subjects with good knowledge of precedent and of the inclinations of particular courts can perhaps predict what the court will decide, this will not generally be the case – especially since judges are likely to disagree about the relevant substantive criteria for determining the likeness of cases. 90 An additional reason to doubt P8 is that even if we concede that courts in general attempt to treat like cases alike and that there is rough agreement about the relevant criteria, this is by no means the only consideration that courts take into account when deciding cases. In particular, predictability is often significantly undermined by the practice of distingushing – i.e. the practice of not following precedent even though the ratio of the earlier case applies to the facts of the present case. When cases are distinguished, it is claimed that the present case has a property that relevantly distinguishes it from prior precedent-setting cases, warranting a contrary decision. To be sure, the probability that a court will distinguish is often lower than the probability that it 89 See Duxbury (2008), 143. See also Schauer (1987), 596–7. 90 It may of course happen that, in particular areas of law, there is significant general predictability about outcomes in borderline cases. In American criminal law, for example, the Rule of Lenity presumably makes it fairly likely that a borderline case will be decided in favor of the defendant. However, such fairly settled rules are rare outside the domain of criminal law. 95 will follow precedent, but if we add this measure of uncertainty on top of the frequent uncertainty regarding the criteria of similarity, then this common judicial practice helps make it the case that – at least in many cases – there will be significant unpredictability regarding court outcomes. Or, in Sorensen‘s jargon, the fact that courts try to act in principled ways does not guarantee that they do not behave sufficiently like random- decision generators. For the reasons mentioned above, it may be sufficiently hard for many subjects to detect any reliable patterns in the decisions of courts in borderline cases. Thus, P8 is false, which leaves C4/P4 and C2/P2 unsupported, which in turn leaves Sorensen’s main conclusion – that vagueness-related uncertainty is not valuable in law – unsupported. A third reason to treat P8 with skepticism – although not as weighty as the combination of the two preceding ones – is that although it may be true that in general people don’t reliably produce sets ot outcomes that closely resemble genuinely random ones, they don’t reliably detect patterns either. As psychological research has shown, 91 people tend to see patterns where there aren’t any and to fail to see those that are there. That is, in general, the pattern-recognition capabilities of normal human beings generate a great deal of both false positives and false negatives. It is of course controversial whether and to what extent these experimental results generalize to different spheres of human life, 92 but they – at the very least – suggest that even if judges don’t come very close to 91 See e.g. Gilovich et al., (1985), Kahneman & Tversky (1972), Bar-Hillel & Wagenaar (1991), and Hahn & Warren (2009). 92 For a general discussion of the application of general experimental results to judicial cognitive behavior, see Schauer (2010). 96 resembling genuine random-decision generators, the cognitive capacities of normal legal subjects will often guarantee that they aren’t able to reliably predict future outcomes based on their perception of previous judicial behavior. 93 I should emphasize that I am not arguing that vagueness – and hence uncertainty – in the law isn’t diminished by judicial decisions. Of course it is. But all Hadfield’s argument requires, for example, is that, initially, vague law creates incentive for people to experiment “at the margins,” thus providing legal institutions with better information about how to develop the law in the courts. And as the law is precisifed through judicial decision-making, thereby decreasing uncertainty and – if Hadfield is correct – sometimes getting the law closer to optimality, it seems to me that we are simply trading one reasonably good thing for another. In addition, I take the considerations I have mentioned to show that although court outcomes in borderline cases do precisify the law, the increase in predictability gained from each individual outcome is fairly modest. If what I have said above is correct, we have good reason to believe that Sorensen’s argument for the claim that vagueness-related uncertainty is not valuable in law is unsound. What more, what I have said in fact seems to support the opposite conclusion. In particular, it seems to support the falsity of P2, i.e. of the claim that the unpredictability regarding probability of (criminal or civil) liability resulting from vagueness-related uncertainty does not give people incentive to engage in desirable 93 It is interesting to note that one fairly recent study comparing the reliability of model vs. expert panel prediction about US Supreme Court cases showed that the expert panel had only a 59% accuracy rate, which isn’t all too impressive. It should be noted, however, that within the expert panel there were certain identifiable sub-groups that had significant accuracy rates – most notably a group of appellate attorneys. See Ruger et al. (2004), 1152, 1163–71. 97 behavior. Here, however, I will not attempt to provide my own counterargument to Sorensen’s claim. I will instead take it that both Hadfield and Endicott provide us with good reasons to believe that vagueness-related uncertainty can be a valuable feature of the law, although – as I argued in the previous chapter – we should be careful in not overestimating that value. Thus, I counter Sorensen’s argument only implicitly, in my qualified endorsements of both Endicott’s and Hadfield’s arguments. 98 CHAPTER THREE: MARMOR ON NON-LITERAL LEGISLATIVE SPEECH Chapters one and two have been about the value of using vague terms in legislation. We can call this kind of indeterminacy vagueness of content. In chapters three and four, I turn my attention to a second form of indeterminacy, one that appears to have been somewhat neglected in the literature – both in law and in language. As I aim to show in these chapters, not only is the content of a legislative utterance often vague, it is often vague what the content of such an utterance is. We can call the kind of indeterminacy vagueness about content. In the present chapter, I discuss the conditions for successful non-literal speech and address the question whether these conditions are satisfied in the legal context. I will focus on an argument made by Andrei Marmor for the claim that non-literal legislative speech is rare. 94 According to Marmor, a speaker succeeds in asserting something other than what she literally says only if it is obvious that she cannot be intending to assert the literal content of her remark. And that rarely happens in law, he says. In the first part of this chapter, I argue that a simple version of Marmor’s argument is unsound. In the second part, however, I argue that the argument can be successfully revised. In chapter four, I argue that the revised argument has some significant consequences for the extent to which we should take the content of the law to be determinate. The basic idea is that in certain cases in which the audience is warranted in withholding belief regarding the speaker’s communicative intention, the primary content 94 Marmor (2008). 99 of the relevant utterance is indeterminate between the literal content of the sentence uttered and some non-literal enrichment thereof. This has significant consequences for the analysis of a number of important but controversial legal cases, which I discuss in some detail. 1. Marmor on non-literal legislative speech In ordinary conversational settings, it frequently happens that speakers assert something other than what they literally say. When I utter the words ‘I haven’t had breakfast’ to the waiter at the airport café, for example, it is obvious that I don’t intend to assert that I haven’t had breakfast ever. Rather, I am asserting that I haven’t had breakfast today, perhaps in the hope that he tells me they are still serving. Given the ubiquity of this phenomenon, it makes sense to ask whether it also frequently happens that legislatures assert something other than what they literally say. Consider for example the Federal Tort Claims Act, which guarantees the government immunity from “[a]ny claim arising in respect of […] the detention of any goods, merchandize, or other property by any officer of customs or excise or any other law enforcement officer.” 95 On a literal reading of the Act, it guarantees immunity from claims arising in respect of certain actions on behalf of law enforcement officers of any kind. But it is reasonable – given the context of the Act – to wonder whether the legislature intended to communicate something slightly different, namely that immunity is guaranteed in respect of certain actions on behalf of law enforcement officers working 95 28 U.S.C. 2680(c). 100 in customs or related domains. The difference in content is significant. Abdus-Shahid M.S. Ali, for example, wanted to sue the government because prison officers lost property belonging to him, but this would require a non-literal reading of the Act. The Supreme Court held that a literal reading was appropriate and Ali was not allowed to sue. 96 Or consider the US Criminal Code, 18 U.S.C. § 922(g)(1), which makes it “unlawful for any person … who has been convicted in any court, of a crime punishable by imprisonment for a term exceeding one year … to … possess … any firearm.” 97 Read literally, there is no restriction on the location of the conviction – a conviction in any court will do. But – again – it makes sense to ask whether the legislature intended to communicate something slightly different from the literal content of the statute, namely that possession of a firearm is unlawful if one has been convicted in a US court. As before, the difference in content is significant. Gary Small, for example, had been convicted of violating 18 U.S.C. § 922(g)(1), but the Supreme Court reversed the decision on appeal, on the basis of a non-literal reading of the Code. 98 These cases are by no means unique – there is a wealth of cases the outcome of which turns on whether the legislature can reasonably be said to have asserted something other than what it literally said. Thus, questions about the existence and extent of non- 96 Ali v. Federal Bureau of Prisons; 552 U.S. 214 (2008). I discuss this case in more detail below. 97 18 U.S.C. § 922(g)(1). 98 Small v. United States; 544 U.S. 385 (2005). I also discuss this case in more detail below. 101 literal assertion in the law have considerable practical, in addition to theoretical, significance. 1.1 Marmor’s skepticism about non-literal legislative assertion In “The Pragmatics of Legal Language,” Andrei Marmor argues that – contrary to ordinary conversation – it rarely happens that the law asserts something other than what it literally says. 99 The reason, he thinks, is that the pragmatic aspects of legal language are unique, in that they involve considerations that are not typically present in ordinary conversational settings. In ordinary conversation, for example, the context – i.e. the background against which utterances are evaluated – is typically rather rich, which according to Marmor makes it possible for a speaker to assert something other than what she literally says. In some cases, for example, a rich conversational background makes it obvious to the hearer that the literal content of the sentence uttered is false, which prompts her to infer that the speaker intends to communicate something different. And, in other cases, a clear sense of specific conversational purpose makes it clear – on pain of irrelevance – that the speaker intended to assert something other that what she literally says. But in law, Marmor says, the conversational background is usually rather thin; the legislature and its audience do not share much common ground and it is often unclear what exactly the specific context of a particular legislative utterance is (although it is of 99 Marmor (2008). For an interesting – and relatively early – discussion of non-literal legislative speech and epistemic constraints, see Goldsworthy (1994), esp. 168–70. 102 course clear enough that the accepted general purpose of the “conversation” is to enact a law). As a result, Marmor is skeptical about the idea that the legislature can assert something other than what it literally says, where this is taken to mean that (i) the content of the legislature’s assertion is not identical to the content of the sentence(s) uttered nor (ii) is the content of the sentence(s) uttered among the content asserted. This notion of asserting something other than what one literally said will be further sharpened below. Marmor substantiates this skepticism by first contrasting successful cases of pragmatic enrichment with unsuccessful ones, in an effort to identify the feature that the former cases have and the latter cases lack, and by then showing that – typically – this feature is absent in the context of legislative utterances. The distinguishing feature, Marmor thinks, is obviousness: typically, in cases in which a speaker successful asserts something other than she literally says, it is – due to a sufficiently rich conversational background – just obvious that she couldn’t be intending to assert the literal content of her remark. It is this feature that Marmor thinks is – for the most part – absent in the context of legislation. There are two ways, then, in which one can respond to Marmor’s argument. One is to focus on the claim that asserting something other than what one literally says requires it to be obvious that one doesn’t intend to assert what one says. The other is to focus on the claim that it is rarely obvious that the legislature couldn’t be intending to assert just what it literally said. I think both claims need some revision, but here I will focus only on the first claim. I will argue that we can find a robust set of counterexamples to what I will call the obviousness criterion and that we should instead opt for a weaker, 103 more flexible, requirement, one that can accommodate both the examples that prima facie motivate the obviousness criterion and the examples that I take to constitute counterexamples to it. The basic idea is that a speaker succeeds in asserting something other than what she literally says only if it is significantly more likely than not that she is intending to do so. Since the relevant significance is determined by what is at stake in a particular context, the requirement is to some extent context-sensitive. After explaining and criticizing Marmor’s obviousness criterion, and after outlining my own account, I will consider a possible reply to my argument against the obviousness requirement. Although I do not take it to be successful, it is still both prima facie plausible and raises issues that are worth addressing. First, however, I will turn to explaining the motivation for positing a strict requirement of the sort endorsed by Marmor. 1.2 What is required for asserting something other than what one says? Marmor notes that in ordinary conversation, we often comes across examples like the following, in which the assertive content clearly goes beyond what is literally said: 100 100 Note that the following examples and their analyses assume that the literal content of the relevant utterances are complete propositions. This assumption is controversial, and some would be inclined to argue that the literal content is incomplete (i.e. not truth-evaluable). For a discussion, see e.g. Soames (2010), 155–156. Here, however, I will – for the sake of argument – assume that the literal content of the utterances in examples (1) and (2) are complete propositions. 104 (1) A doctor examining a gunshot wound tells the patient: “Don’t worry, you are not going to die.” Clearly, the doctor did not assert that the patient is never going to die, but only that he is not going to die from this particular wound. (2) When I get home in the evening and my wife asks me “Have you eaten?” it is quite clear that she is asking me if I have already eaten dinner tonight, not whether I have ever engaged in the activity of eating. 101 What makes these cases obvious is in large part that certain things are taken for granted by the participants in the conversation, including the purpose of the exchange, and partly that other interpretations would make the respective utterances less than fully cooperative (given the purpose of the relevant conversation). In (1), for example, what makes it obvious that the speaker asserted something other than what she literally said is the contradiction between the literal content and the presupposition that everyone dies, plus the fact that what is relevant to the hearer is his current state of health. 102 And in (2), it is both taken for granted that the hearer has at some point engaged in the activity of eating and that – in circumstances like these – people are normally interested in whether or not their spouses have had adequate nutrition over the course of the day (or perhaps just whether or not they are hungry). 103 101 Marmor (2008), 426. As Marmor notes, these examples are slight variations of examples discussed by Bach (1994) and Soames (2008). 102 In Gricean terms, this case involves an apparent flouting of the First Maxim of Quality: Do not say what you believe to be false. See Grice (1989), 27. 103 In Gricean terms, this case is arguably best analyzed as involving an apparent flouting of the Maxim of Relevance: Be relevant. See Grice (1989), 27. 105 For our purposes, the general lesson to take away from these examples is the following, according to Marmor: “A speaker would normally succeed in conveying assertive content that differs from what he says, when it would be obvious to the hearer, in the particular context of the conversation, that it just cannot be the case that the speaker asserts exactly what he says.” 104 Call the italicized part of the preceding quote, which is restricted to normal circumstances, the ‘obviousness criterion’. Now, as stated, the obviousness criterion is a sufficient condition (on the assumption that certain other background conditions are satisfied), and is supported for example by the fact that – in cases like (1) and (2) – a rational hearer doesn’t even begin to entertain alternative interpretations. If it is obvious, given that a speaker S has uttered some sentence ‘s’ in a context C, to a rational hearer, knowing the relevant conversational background and context, that S doesn’t intend to assert what she literally said, then S asserts something different from what she literally says. It seems clear, however, that Marmor thinks that, in normal circumstances, the obviousness criterion is a necessary condition as well. It is quite possible, to borrow an example from him, that in uttering ‘Have you had blackberries?’ in a particular circumstances, a speaker can be meaning to ask whether or not the hearer has ever had blackberries or can be meaning to ask whether or not the hearer has had blackberries that day. If the contextual background is not sufficiently rich to make it obvious which question is intended, Marmor says, and it would consequently make sense for the hearer 104 Marmor (2008), 428; my emphasis. 106 to ask for clarification, then “this would be a clear indication that [the speaker has] not succeeded in asserting some content that differs from what [she] said.” 105 Given what Marmor says about the contrast between successful and unsuccessful cases of asserting something different from what one literally says, then, I think it is reasonable to restate the obviousness criterion as a biconditional (note that in this discussion, I’m assuming that q is a complete, truth-evaluable proposition): OC On the assumption that, in some normal context C, something has been successfully asserted: In uttering a sentence ‘s’ with semantic content q in C, a speaker S does not assert q iff, in C, it is obvious to a rational hearer H, knowing the relevant conversational background and context, that S just cannot be intending to assert q. Stated this way, the obviousness criterion is supposed to explain our intuitions regarding successful cases of asserting something other than what is literally said – like (1) and (2) above – as well as our intuitions about unsuccessful cases – like the “blackberry scenario”. I think that Marmor is quite right that – assuming certain background conditions – obviousness is a sufficient condition for successfully asserting something other than what one literally said, so my focus here will be on obviousness as a requirement on such success: 105 Marmor (2008), 428. 107 OR On the assumption that, in some normal context C, something has been successfully asserted: In uttering a sentence ‘s’ with semantic content q in C, a speaker S does not assert q only if, in C, it is obvious to a rational hearer H, knowing the relevant conversational background and context, that S just cannot be intending to assert q. Further, given that obviousness of the relevant sort depends – according to Marmor – on how rich the contextual background is, we have a further requirement: RB In a normal context C, it is obvious to a rational hearer H, knowing the relevant conversational background and context, that, in uttering a sentence ‘s’ with semantic content q in C, a speaker S just cannot be intending to assert q only if, in C, S and H share a great deal of (relatively) specific background assumptions. Together, OR and RB produce a second necessary condition on successfully asserting something other than what one literally says – we can call it the richness requirement: RR On the assumption that, in some normal context C, something has been successfully asserted: In uttering a sentence ‘s’ with semantic content q in C, a speaker S does not assert q only if, in C, S and H share a great deal of (relatively) specific background assumptions. In this part of the chapter, however, I will concentrate on OR, leaving the discussion of RR for the next one. I will argue that we can find a robust set convincing counterexamples to OR. 108 For our present concerns, the significance of these three generalizations is that Marmor thinks that when it comes to legislative utterances the relevant principles are rarely satisfied. Very little can jointly be taken for granted by the legislature and its audience(s), which generally makes it less than obvious that it just cannot be the case that the legislature intended to assert what it literally said. Given OR, it follows from this that – in general – if the legislature intends to assert something other than what is literally says, it does not succeed in doing so. In order to illustrate this general claim, Marmor discusses the famous case of Church of the Holy Trinity v. US, 106 in which the relevant question was whether the Alien Contract Labor Act (ACLA) of 1885, 107 which prohibited the importation of “labor or service of any kind” except “professional actors, artists, lecturers or singers” and “persons employed strictly as personal or domestic servants,” prohibited the importation of people intended to serve as rectors or clergymen. The circuit court for the southern district of New York held that it did, while the Supreme Court held that it did not. As Justice Brewer put it, in delivering the opinion of the court: “It is a familiar rule that a thing may be within the letter of the statute and yet not within the statute, because not within its spirit nor within the intention of its makers.” Marmor asks whether it is plausible to argue – in the spirit of Justice Brewer’s remark – that in enacting the ACLA, the legislature asserted something different from 106 143 U.S. 457 (1892). 107 23 Stat. 332 (codified as amended at 29 U.S.C. §2164 (1901)), superceded by Act of Mar. 3, 1903, Pub. L. No. 162, 32 Stat. 1213. 109 what it said. That is, can we argue that although the importation of rectors and clergymen was prohibited according to what the law literally said, the assertive content of the legislative utterance was different, in that it did not apply to such conduct? Marmor thinks that for a couple of reasons it would be a stretch to argue this, and I think he is quite right. Given Marmor’s OR, we can say that the primary reason for this is that the context in which the ACLA was enacted has a number of features in virtue of which it makes sense to ask whether Congress could have had in mind to prohibit the importation of foreign clergymen. And if that is so, then it cannot be obvious that it did not mean to assert what it literally said. So in case Congress really did intend to assert something different from what it literally said, it did not succeed. Or so Marmor argues. Now, as Marmor recognizes, it is perhaps not entirely fair to pick Holy Trinity as the only case study. And I think that is correct. He believes, however, that we get the same results with simpler examples (assuming they are sufficiently realistic, of course 108 ). For our purposes, it is not important how these examples go. Marmor’s explanation for why the law – in these cases – does not succeed in asserting something different from what it literally says is the same: “the context is not rich enough to make it obvious and transparent that the legislature could not have meant/asserted what it said.” 109 108 It is quite possible, I think, to invent cases in which lawmakers and their audience(s) share a wealth of beliefs/knowledge and that – consequently – the law has no problem asserting something other than what it says. But this does not affect Marmor’s argument, since he is only committed to the claim that, given the way the law tends to work, such cases will be relatively rare. 109 Marmor (2008), 429. 110 1.3 The obviousness requirement I fleshed out Marmor’s obviousness requirement for asserting something other than what one literally says in the following way: OR On the assumption that, in some normal context C, something has been successfully asserted: In uttering a sentence ‘s’ with semantic content q in C, a speaker S does not assert q only if, in C, it is obvious to a rational hearer H, knowing the relevant conversational background and context, that S just cannot be intending to assert q. I should note right away that Marmor is not alone in positing comparably strict requirements. We have something rather similar, for example, in Grice’s own pragmatic framework. Assuming that Grice would have classified the examples we are concerned with as examples of conversational implicature (rather than assertion), it follows from Grice’s (1975) definition thereof that the relevant speech acts are successful only if the hearer must reinterpret the speaker’s remark in order to make it consistent with the Cooperative Principle. 110 And, as Sperber and Wilson (1995) observe: When... [ordinary hearers] are asked to explain how they know which interpretation is correct, they generally offer something that looks like a truncated logical argument: the speaker must have intended this interpretation rather than that, because this is the only interpretation that is true; or the only one that gives the required information; or the only one that makes sense. 111 110 See Grice (1975), 30–31. 111 Sperber and Wilson (1995), 13; my emphasis. Cited by Recanati (2002). Note that Sperber and Wilson do not themselves endorse such a strict requirement. 111 The underlying thought – shared by Marmor and Grice, and reflected in certain intuitions of ordinary speakers – is, I take it, that in order for a speaker to successfully assert something other than what she literally said, something must effectively prompt the hearer to reinterpret her remark accordingly. On the picture under scrutiny, then, reinterpreting a speaker S to be asserting something, p, other than what she literally said, q, is warranted only if a rational and informed hearer is – conditional on the assumption that S is being cooperative – in a position to assign a probability marginally greater than 0 to the proposition that S intended to assert q. 112 To be sure, this requirement seems to be satisfied in many cases – most notably in standard cases in the literature that involve the flouting of a conversational maxim. On the assumption, for example, that the doctor from Marmor’s example (1) is observing the maxim Don’t say what you believe to be false, any rational and informed hearer will assign probability fantastically close to 0 to the proposition that the doctor intends to assert that the patient won’t die ever, assuming that there is no reason for her to believe that the doctor is not being cooperative. The question, however, is whether it is true of all normal cases in which a speaker asserts something other than what she literally says that it is virtually impossible – relative to the body of information available to the hearer – that she could be intending to assert what she says. It seems to me that it isn’t. I think that there is a distinctive set of 112 I say marginally greater than 0 because if the hearer assigned probability 0 to the proposition that S intended to assert q, then no amount of evidence would suffice to change her mind (assuming she is – by Bayesian standards – rational). 112 cases in which a speaker successfully asserts something other than what she literally says in which there is indeed a reasonable possibility that she intends to assert what she literally says. These are cases in which it is still significantly more likely than not that she does not intend to assert what she literally says. These cases, I will argue, give us good reason to reject OR in favor of a weaker, more flexible, requirement. 1.4 Assertion, warrant for belief, and the obviousness requirement In order to get at what I take to be the heart of the matter, we need to dissect the obviousness requirement a little further. First, Marmor and I both accept an account of assertion on which the following informal statement is true: Assertion A speaker asserts p only if a rational hearer, knowing the relevant conversational background and context, is warranted in taking her to be intending to assert p. 113 On this account, the epistemic position of the hearer plays a part in determining the content of the speaker’s assertion. 114 That is, it is partly in virtue of facts about epistemic justification that a speaker counts as having asserted a particular proposition. 115 For our 113 See e.g. Marmor (forthcoming), section 3. 114 Several authors – many of whom accept the basic tenets of Grice’s framework – embrace this partly objective account of assertive/communicative content, but some devout Griceans favor a subjective account, on which a speaker means, says, asserts, etc. what he intends to mean, say, assert, etc; see e.g. Neale (2000). 115 Note that, although this is one sense in which the relevant account of assertion is properly said to be normative, this is different from the normativity associated with the so-called norms of 113 purposes here, we can flesh this principle out in the following way and call it the warrant requirement: WR In a normal context C, a speaker S, in uttering a sentence ‘s’, succeeds in asserting p only if, in C, a rational hearer H, knowing the relevant conversational background and context, is warranted in taking S to be intending to assert p. Further, when p is not the semantic content q of the sentence uttered, H is warranted in taking S to be intending to assert p only if H is warranted in taking S not to be intending to assert q. 116 Thus, for our purposes here, we can say that, restricted to contexts in which a speaker asserts something other than what she literally says, the warrant requirement takes the following form: WR* On the assumption that, in some normal context C, something has been successfully asserted: A speaker S, in uttering a sentence ‘s’ with semantic assertion. While the norms of assertions are partly constitutive of an utterance’s being an assertion, the normative considerations I am discussing in this section concern what it takes for an utterance not just to be an assertion but to be an assertion of p. That is, it concerns the content of a given utterance. 116 It might be argued that descriptively enriched assertions that involve sentences whose semantic contents are singular propositions falsify this claim, because the hearer’s inference to the content asserted is not prompted by her recognition that the speaker cannot be intending to assert the singular proposition. Rather, descriptive enrichment appears to be a brute fact about how we use sentences that contain occurrences of proper names: as a rule, rational, reasonably well-informed hearers will add salient descriptive information associated with the corresponding names to the semantic contents of the relevant sentences. However, I don’t this “brute fact” actually falsifies my claim. The reason is that if, in a normal context C, a hearer H is warranted in taking a speaker S to be intending to assert an enriched proposition p and the singular proposition q doesn’t follow from p and C, then H is warranted in taking S not to be intending to assert q. 114 content q in C, succeeds in not asserting q only if, in C, a rational hearer H, knowing the relevant conversational background and context, is warranted in taking S not to be intending to assert q. In plain English, WR* reads something like the following: A speaker succeeds in asserting something other than what she literally says only if a rational hearer, knowing the relevant conversational background and context, is warranted in taking her to be intending to do so. Marmor’s obviousness requirement is produced when we combine WR* with the principle that, in normal contexts, a rational hearer is warranted in taking a speaker to be intending to assert something other than what she literally says only if it obvious that she just cannot be intending to assert what she literally says: Warrant On the assumption that, in some normal context C, something has been successfully asserted: A rational hearer H, knowing the relevant conversational background and context, is, in C, warranted in taking a speaker S to be, in uttering in uttering a sentence ‘s’ with semantic content q in C, not intending to assert q only if, in C, it is obvious to H, knowing the relevant conversational background and context, that S just cannot be intending to assert q. It is this principle which brings out what I believe is at issue between Marmor and me. Contra Marmor, I think that there are successful instances of speakers asserting something other than what they literally say in which it is less than obvious to a rational and informed hearer that the speaker could not be intending to assert the literal content of 115 her remark. If I am able to find counterexamples to Warrant, I will by the same token have found counterexamples to OR. I turn to this task next. 1.4.1 Loose talk scenarios as systematic counterexamples to Warrant? I should begin by explaining what sort of cases I believe make for the most plausible counterexamples to the principle Warrant, and consequently to Marmor’s obviousness requirement. In the first instance, these have to be cases in which it is not obvious that the speaker just cannot be intending to assert the semantic content of the sentence she utters, yet in which nothing appears to be amiss, pragmatically speaking. That is, they must be cases in which the communicative act can reasonably be said to have gone well. A case like this goes reasonably well – I submit – if (but not necessarily only if) (i) the speaker intends to assert something other than what she literally says, (ii) it is, relative to the epistemic position of a rational hearer, knowing the relevant conversational background and context, significantly more likely than not that the speaker is so intending, and (iii) the hearer in fact takes the speaker to be so intending. In many cases of this sort – as I will illustrate below – the hearer may not have any significant reason to ask for clarification, although it may perhaps not be outright silly to ask for it either. In such cases, it would seem that things have gone well and so I see no reason to say that the speaker did not succeed in asserting something other than what she literally said. It seems to me that the speaker succeeded in doing what she intended to do. I should note that if I am right about these cases, then Marmor’s “clarification- test” for successful non-literal assertion is too strict: a speaker successfully asserts 116 something other than what she literally said only if it would be silly of the hearer to ask for clarification about what the speaker meant. Sometimes, as I hope to show, non-literal assertion is successful despite the fact that it would not be outright silly to ask for clarification. These are cases in which the hearer has a reason to ask the speaker for clarification only if that reason comes from the hearer’s own desire to know for certain what the speaker intended to communicate Now that I have explained the general features that I think successful counterexamples must have, let us fix the relevant ideas by trying out two examples in which the speaker is speaking “loosely”. Imagine first that my friend Kory tells me that he can only stay for 20 minutes or so for the premiere of my wife’s new orchestra piece and that he asks me how long the piece is. Imagine also that in response I utter the following: ‘It’s 12 minutes long.’ In this context, it is surely possible that I am intending to speak literally, i.e. it is not obvious that I just cannot be intending to communicate that the piece is exactly 12 minutes long. Yet, it is – given what is “required by the accepted purpose of the conversation” – significantly more likely than not that I am intending to communicate that the work is approximately/about/roughly 12 minutes long. Thus, it is true both that (i) in uttering the relevant sentence in the relevant context, I have given Kory significantly more reason than not to believe that I intended to communicate that the piece is approximately 12 minutes long and that (ii) it doesn’t matter much, for the purposes of the conversation, if I in fact intended to communicate that the piece is exactly 12 minutes long; since Kory is able to stay for the entire piece, nothing hangs on it. 117 It seems to me that there is no reason here to claim that Kory is not warranted in taking the speaker (i.e. me) to be intending to assert something other than the literal content of the sentence uttered. In case Kory infers that I intended to communicate that the piece is approximately 12 minutes long, nothing appears to be amiss – and nothing appears to be amiss because nothing is amiss. Further, if Kory has any reason to ask me to clarify what I meant to communicate, this reason will be grounded exclusively in his desire to know for certain what I meant. If all this is correct, then Warrant is false – the bar for successful assertion is sometimes lower than is postulated by OR. For a second example, imagine that someone asks an artist how her show went and that in response she utters the following: ‘Nobody came to my show!’ In most situations, a rational hearer would infer that the artist intended to communicate that almost nobody came to her show. But it’s not outright impossible that the artist is intending to assert that zero people showed up. However, given that it is rare that absolutely nobody shows up for a show, the artist has given her interlocutor good reason to believe that she is intending to communicate that almost nobody came to the show – it is significantly more likely than not that she is so intending. Moreover, it does not matter – vis-à-vis this particular context – which proposition the artist in fact intended to communicate; either way, the artist has conveyed that the show didn’t go well, which suffices for the purposes of the conversation. As before, it may not be outright silly to ask the artist whether she meant that absolutely nobody or almost nobody showed up, but in this stipulated scenario there is – given what is “required by the accepted purpose of the conversation” – no reason to ask for clarification that is independent of the interlocutor’s 118 desire to know for certain. This further indicates that Warrant is false, and that successful non-literal assertion does not require obviousness – at least not across the board. The scenarios I have just described are not out of the ordinary – they are normal cases of successful non-literal assertion. The same reasoning, I maintain, can be applied to many, if not all, low-stakes loose-talk scenarios – and such situations appear to occur quite frequently, at least frequently enough to make them non-exceptional. It seems, therefore, that we have a robust and systematic set of counterexamples to OR. I want to discuss a possible reply to what I have said, but before I go on to do so, I believe I also need to provide an analysis of Marmor’s blackberry case, an analysis that is consistent with a weaker epistemic constraint on non-literal assertion. The reason is that in that particular case something did indeed seem to be amiss, evidenced by the fact that it seemed that the wife indeed had a desire-independent reason to ask for clarification. I propose to tackle the case in the following way. As the case is described, there is no mention of what the likelihood is that the speaker intends to assert something other than what he literally says or of what the likelihood is that he intends to assert just what he literally says. Now, on the reasonable assumption that uncertainty of this sort translates to equal likelihood, we have an explanation – consistent with my view – of why it makes sense for the hearer to ask for clarification. Although my account is weaker than Marmor’s, it does require that it be significantly more likely than not than the speaker intends to assert something other than what she literally says, where the required 119 significance is determined by what is at stake in the relevant scenario. This requirement is not satisfied on my analysis of the blackberry scenario. It is possible, of course, to construct examples about which it seems reasonable to say that if it isn’t obvious that the speaker can’t be intending to assert what she literally says, then the right thing to do is to ask for clarification. However, these seem to make up only a subset of those cases in which a speaker succeeds in asserting something other than what she literally says. And I should note that when I say that the stakes in a case are high, I don’t necessarily mean that something morally significant depends on the hearer’s ascertaining what the speaker means. It can of course be something of that sort – as it often is in law – but it can also be something much more mundane. The sincere utterance of a question, for example, presupposes that the speaker has an interest in getting information from the hearer. Hence, in normal interrogative contexts, it is to be presumed that something sufficiently significant is at stake and thus it makes sense – if possible – to ask for clarification if there are competing interpretations of the speaker’s question. Next, I will consider a possible reply to what I have said – one that seeks to show that, in my loose-talk scenarios, the hearer is warranted only in believing to a certain degree that the speaker intends to assert something other than what she literally says, but not in believing it outright. If that’s true, then these scenarios are not counterexamples to Warrant, on the assumption that this allows the objector to deny that there has been successful assertion in my examples. 120 1.5 Possible reply: Degree of belief vs. outright belief In the first loose-talk example, I reasoned in the following way about the hearer’s justified belief about speaker intention. Given that the relevant sentence was uttered in the relevant context, it is significantly more likely than not – relative to Kory’s epistemic position – that I am intending to assert that my wife’s new piece is approximately 12 minutes long (rather than intending to assert that it is exactly 12 minutes long). That is, I provided Kory with good evidence of my non-literal communicative intention – good enough so that he was warranted in taking me not to be intending to assert what I literally say on the basis of that evidence. This line of reasoning, however, invites the following question. If the evidence supports the conclusion that it is significantly more likely than not that I am intending to assert something other than what I literally say, why do I say that Kory is warranted in outright believing that I so intend? That is, why is Kory’s epistemic warrant in this case not confined just to the appropriate degree of belief? The problem for me is that if it is, then my loose-talk example isn’t a genuine counterexample to Warrant, because it would no longer be safe to assume – as the principle does – that something has been successfully asserted. There are at least two attractive ways to respond to this objection – both of which originate in theories that attempt to explain what has come to be known in epistemology as pragmatic encroachment. Both theories attempt to explain how pragmatic factors – what is at stake in a context – can affect an agent’s justification for outright belief without 121 affecting her justification for degrees of belief. 117 (I should note right away that, on both accounts, the justification of degrees of belief is sensitive only to evidence.) One theory – Pragmatic Credal Reductivism – attempts to explain this directly by arguing that the degree of confidence in a proposition that constitutes believing it can vary between contexts, while the other – the Reasoning Disposition Account – tries to do so indirectly by arguing that, because of the function that outright belief serves for human beings, such belief is justified as long as the relevant proposition is significantly more probable than its negation and the cost of being wrong is not too great. The theories are motivated by the intuitive normative difference between low- stakes cases and high-stakes cases of the following sort, as described by Ross and Schroeder (forthcoming): Low: Five minutes ago, Hannah made three sandwiches and placed them in the refrigerator. She told Sarah that she placed the peanut butter sandwich on the left, the tuna sandwich in the middle, and the almond butter sandwich on the right. Hannah then departed just as Sarah’s friend Almira arrived for lunch. Sarah knows that Almira has no allergies. Almira says: “I’d love an almond butter sandwich.” And so Sarah opens the refrigerator door, points to the sandwich on the right, and says: “The sandwich on the right is an almond butter sandwich. You can have it.” High: This case is just like Low, except here it is Sarah’s nephew Algernon who is visiting for lunch, and he has a severe peanut allergy. He asks Sarah for a sandwich. Sarah knows that the peanut butter sandwich would be fatal to Algernon, but that the almond butter sandwich would be harmless. She also knows that he would slightly prefer the almond butter sandwich to the tuna sandwich. When Sarah goes to the fridge, she can tell, by visual inspection, which is the tuna sandwich, but she cannot tell, by visual inspection, which is 117 See Ross and Schroeder (forthcoming), 1. 122 the peanut butter sandwich and which is the almond butter sandwich. So she gives him the tuna sandwich. As Ross and Schroeder note, the natural explanation seems to be that in Low Sarah knows that the sandwich on the right is the almond butter sandwich, while in High she doesn’t. The explanation for this difference in knowledge is taken to be that in Low Sarah is justified in believing that the sandwich on the right is the almond butter sandwich, while in High she is not. The two types of theories differ in how they explain why Sarah is justified in believing that the sandwich on the right is the almond butter sandwich in Low but not in High. Without going to into the details of the debate, Pragmatic Credal Reductivism explains this by saying that in Low, but not in High, the justified degree of belief is high enough to constitute justified outright belief, while the Reasoning Disposition Account explains it by claiming that in both scenarios the evidence justifies a dispositional belief, but that only in Low does it also justify an occurrent belief – in High, the justification for outright belief is overridden because the cost of being wrong is particularly salient. 118 Coming back to my loose-talk scenario, let’s say that r is the proposition that that I intended to assert something other than what I literally said. I can argue, then, either that Kory is warranted in believing r because he is, given the evidence available in the context, warranted in believing r to degree d and believing r to degree d in this context constitutes believing r outright, or I can argue that Kory – as a rational hearer – has, given the evidence available in the context (and the low stakes), a justified non- 118 See Ross and Schroeder (forthcoming), section 0. 123 overridden disposition to believe r, from which it follows that he is justified in occurrently believing r. Here, I will assume that it doesn’t matter much which of these two theories I adopt, but I do acknowledge that Ross and Schroeder are persuasive in their argument that only their account – the Reasoning Disposition Account – can really be squared with a seemingly truistic principle that they call Sufficient Evidence: It is rational to believe a proposition p only if one’s evidence significantly favors p over its negation. 119 Simply for that reason I will favor their account as my defense against the objection that, in my loose-talk scenario, Kory (i.e. the hearer) is warranted only in believing to some degree d<1 that I intended to assert something other than what I literally said. On the Reasoning Disposition Account, beliefs have a distinct functional role, due to the competing ends that they serve for human beings. Belief, on this account, serves two competing ends: to allow the believer to arrive at good deliberative conclusions and to prevent the cognitive overload that would result from reasoning in an ideal Bayesian manner on the basis of credences alone. 120 The optimal balance seems to be struck by something like the following principle: an agent is permitted to treat a proposition p as true in her reasoning if the evidence significantly favors p over its negation and the cost of being wrong isn’t too great. Treating these uncertain propositions as true in our reasoning (i.e. believing them), then, is a sort of optimal heuristic, justified by its benefit over the long run. And this can explain why, in my loose-talk scenarios, the hearer is 119 Ross and Schroeder (forthcoming), section 2.4. 120 Ross and Schroeder (forthcoming), section 2.1. 124 warranted in taking the speaker to be intending to assert something other than what she literally said. What I have sketched above constitutes my primary reply to the objection that, in my examples, the hearer is warranted only in believing that it is significantly more likely than not that the speaker intended to assert something other than what she literally said. But there is another reason to reject the view underlying the objection that is worth a brief mention. There is certainly a plausible ring to the claim that one only has as much reason to believe that p as the evidence for p allows, but if we take outright belief in p to be justified only if the probability of p is 1 (or fantastically close thereto), then we would rarely be warranted in believing any empirical claim outright. The reason is that it presumably rarely happens that the probability of an empirical proposition, conditional on some body of evidence, is 1 (or fantastically close thereto). The view underlying the objection discussed in this section, then, has very counterintuive consequences regarding justified outright belief, which I take to count significantly against it. 121 And once we lower the bar, I hope that it becomes relatively easy to see the motivation for my view. 122 121 The counterintuitive nature of this consequence becomes fairly evident when stated normatively: Believe that p only if you know that p. 122 One might wonder why Marmor couldn’t just reply that while what is at stake in a context does affect a person’s epistemic situation, what varies between the contexts relevant to our discussion is simply the height of the bar for obviousness. If the level of certainty required for obviousness partly depends on practical facts, then it may perhaps be argued that OR holds after all. Although I do think that practical considerations affect what counts as obvious, I don’t this fact will suffice to save the obviousness requirement. The main reason is that obviousness is 125 1.6 A further problem for the obviousness requirement: speaker competence There is another – indirect – reason to doubt the obviousness criterion, one that stems from the implications that the view has for our ideas about speaker competence. My worry is that Marmor’s view predicts that a competent speaker will form an intention to assert something other than what she literally says only if she believes that it will be obvious to the hearer that she intends to do so. It seems to me that competent speakers frequently – indeed systematically – behave differently, and so the OR seems implausibly strict. The argument I have in mind goes like this (note that the notion of belief that I have in mind is a rather loose one – it does not require anything like conscious entertainment of the relevant propositions): P1. A competent speaker will form an intention to assert something other than what she literally says only if she believes that she will succeed in asserting something other than what she literally says. P2. A competent speaker believes that she will succeed in asserting something other than what she literally says only if she believes that it will be obvious to the hearer that she intends to do so. still a pretty strict constraint, so strict that the flexibility allowed by variation with practical context would be quite limited. It would be a stretch, for example, to argue that in my loose-talk example, it is in fact obvious that I couldn’t be intending to assert what I literally said. In fact, as I describe the case, it is – by stipulation – non-obvious. In order to accommodate my example, then, it would have to be argued that for all propositions p, if p is significantly more probably that its negation, then p is obvious. And that seems implausible. 126 C. A competent speaker will form an intention to assert something other than what she literally says only if she believes that it will be obvious to the hearer that she intends to do so. Now, P1 is just a limited application of a more general principle of rational action, endorsed e.g. by Grice (1971): that a rational agent will, at a time t, form an intention to ϕ only if, at t, she believes she will ϕ. 123 This is not an uncontroversial principle, but it is still fairly plausible. 124 And if Marmor is right, P2 is pretty plausible too; it just states that a competent speaker believes the content of OR. And together, P1 and P2 appear to have some significant implications – given certain further assumption that I address below – about the rational formation of communicative intentions. Granted, belief is not closed under necessary consequence, but it is closed under necessary consequence that is both obvious and relevant. And, I think, it is not implausible to argue that, to competent speakers, the consequence in question ought to be both obvious and relevant, which means that if they in fact don’t behave in accordance 123 Others who have endorsed this view include Audi (1973), Harman (1976), Davis (1984), and Ross (2009). 124 In case P1 is thought to be too strong, I think we can get a sufficiently similar result by running the argument using a weaker principle – the conclusion still does not seem to match how actual competent speakers behave: P1. A competent speaker will form an intention to assert something other than what she literally says only if she doesn’t believe that she won’t succeed in asserting something other than what she literally says. P2. A competent speaker doesn’t believe that she won’t succeed in asserting something other than what she literally says only if she doesn’t believe that it won’t be obvious to the hearer that she intends to do so. C. A competent speaker will form an intention to assert something other than what she literally says only if she doesn’t believe that it won’t be obvious to the hearer that she intends to do so. 127 with C, then we have reason to believe that one or both of the above premises is false. Of course, I suspect that P2 is the culprit, but ultimately a good argument for P1 is required in order to effectively single out P2 as the problematic premise. Here is not the time to provide such an argument, but I hope that I have at least provided another reason to be skeptical of the obviousness criterion. If, however, we weaken the constraint on what it takes to assert something other than what one says – as I did above – then the problematic argument does not go through, which I take to count in favor of my account. 1.7 Taking stock I began the first part of this chapter by introducing Marmor’s obviousness requirement on asserting something other than what one literally says and by explaining its role in his argument that – contrary to ordinary conversation – in rarely happens that the law succeeds in asserting something other than what it literally says. Next, I argued that a speaker may successfully assert something other than what she literally says even if it isn’t obvious that she cannot be intending to assert what she literally says, and I proposed to replace Marmor’s obviousness requirement with a requirement according to which a speaker succeeds in asserting something other than what she literally says only if it is significantly more likely than not that she so intends. I then defended my counterexamples against the objection that, in my examples, I allow the hearer to go beyond what the evidence suggests. I argued that once we recognize that practical considerations are relevant to epistemic justification, it is easy enough to motivate a view on which an agent is permitted to treat a proposition p as true 128 in her reasoning if the evidence significantly favors p over its negation and the cost of being wrong isn’t too great. Finally, I explained what I took to be another problematic consequence of taking the obviousness requirement to apply across the board. The problem was that the view implausibly predicts that a competent speaker will form an intention to assert something other than what she literally says only if she believes that it will be obvious to the hearer that she intends to do so. If what I have said so far is correct, Marmor’s argument that non-literal legislative speech is rare is unsound. In the next part of this chapter, I examine how his argument can best be revised. 2. Revising Marmor’s argument: Restrict, reconstruct, or both? There are, I believe, two rather straightforward ways in which to respond to my objection to Marmor’s argument for the claim that the law rarely asserts something other than what it literally says. One is to restrict the argument to the legal context and claim that, in law, the stakes are high enough so that obviousness is indeed required for successful non- literal assertion, and the other is to reconstruct the argument using what I called the richness requirement instead of the obviousness requirement. The former is probably the most natural way to reply to what I have said, but both are worth examining in some detail. 129 I will argue that neither revision suffices on its own, but that conjoining them will provide a plausible revision of Marmor’s original argument. The best way to revise the argument, I think, is both to reconstruct it using the richness requirement and to restrict that requirement to the legislative context. It is quite plausible that, in law, a speaker succeeds in asserting something other than what she literally says only if she shares a great deal of relevant common ground with her intended audience. And that – I agree – is rarely the case in law. 2.1 Restricting Marmor’s argument: Is obviousness required in law? In the first part of this chapter, I spent a great deal of effort casting doubt on Marmor’s claim that in order for a speaker to assert something other than what she literally says, it has to be obvious to a rational hearer, knowing the relevant conversational background and context, that she could not be intending to assert the literal content of her remark. I proposed to replace Marmor’s obviousness requirement with a somewhat weaker, more flexible requirement. The basic idea was that a speaker succeeds in asserting something other than what she literally says only if it is significantly more likely than not that she is intending to do so. What counts as significantly more likely than not, I argued, depends on the practical context – on what is at stake. If nothing much is at stake, the epistemic constraint is relatively low, but if enough is at stake, then obviousness may indeed be required for successful non-literal assertion. 130 It makes sense, then, to ask whether, in law, the practical context is such that obviousness is required. If it is, then it is possible to simply restrict the linguistic premise of Marmor’s argument to the legal context, and the argument will be sound: P1. In law, speakers normally assert something other than what they literally say only if it is obvious to a rational hearer, knowing the relevant conversational background and context, that they cannot be intending to assert the literal content of their remark. P2. In law, it is rarely obvious to a rational hearer, knowing the relevant conversational background and context, that a speaker cannot be intending to assert the literal content of her remark. C. In law, speakers rarely assert something other than what they literally say. On the face of it, this amendment is quite plausible. In fact, it seems to be in the nature of law that it deals only with practically significant matters. At the very least, it seems safe to say that there is a (defeasible) presumption both that legislative utterances are made for good normative reasons and that something significant is at stake in all cases in which legal interpretation is called for. Despite its strong prima facie plausibility, however, I think we have reason to believe that P1 is false. Moreover, I think so for reasons having to do primarily with facts about law itself. The thought underlying the rest of this section is twofold. First, the facts that determine the stakes to which epistemic constraints in the legal context are sensitive are either exhaustively or primarily facts about the consequences of outcomes in legal cases pertaining to the parties involved, rather than for example facts about the societal 131 consequences of particular verdicts. 125 Second, these facts are not just “simple” facts about the utility and disutility appropriately assigned to winning and losing on behalf of the individual parties; rather, they are comparative facts about the extent to which the parties should to be taken to have equal normative status – i.e. they are facts about the symmetry, or asymmetry, between the respective utility assignments. Consider the varying standards of proof present in US law. In most civil law cases, for example, the standard of proof is proof “on the balance of probabilities” – the party bearing the burden of proof must prove that her case is more likely than not. In certain serious civil law cases, however, the standard of proof is higher – the evidence has to be “clear and convincing”, meaning that the party bearing the burden of proof must prove that her case is substantially more likely than not. 126 And in criminal law cases, the standard of proof is proof “beyond reasonable doubt” – it has to be practically certain that the defendant is guilty. Now, I do not want to suggest that the varying standards of proof are identical to the epistemic constraints on legislative speech. In fact, I will argue in the next chapter that these two epistemic standards cannot be the same. However, I do take the varying standards of proof to provide us with a good indirect reason to believe, first, that the epistemic constraint on non-literal assertion varies between legal contexts and, second, 125 Thanks to Andrei Marmor, Gideon Yaffe, and Anders Nes for pressing me to clarify this thought better. 126 Note that I am not using the phrases ‘substantially more likely’ and ‘significantly more likely’ interchangeably. The former pertains to evidence and probabilities only, while the latter pertains to practical considerations also. 132 that we can learn a great deal about this constraint by attending to the considerations responsible for the variation in requirements on legal proof. The main reason for this is that although legislative pronouncements have very much in common with ordinary utterances, they are speech acts performed within a particular institutional setting and so it is natural to think that the institutional factors that influence standards of proof may also influence the epistemic constraints on non-literal legislative assertion. 127 Although there are many competing explanations/justifications for the varying standards of proof, I will briefly explain just one set of justifications, found in Redmayne (1999), in order to give an example of how the different practical and institutional concerns in different legal contexts may affect epistemic constraints. The civil law “preponderance-of-the-evidence” standard of proof is often justified by claiming that a >.5 standard minimizes the expected numbers of errors in civil cases (i.e. errors made by the fact-finder). On the assumption that the practical considerations pertaining to the plaintiff are equal in weight to the practical considerations pertaining to the defendant, it can be derived that a >.5 standard is appropriate. To assume equality in normative status is just to assume that the utility of winning if you deserve it is the same for both parties and that the same goes for the disutility of losing if you deserve to win. Using this framework, it is also possible to explain why the height of the epistemic bar increases in serious civil law cases and in criminal cases. In serious civil 127 For an interesting discussion of the similarities and differences between “ordinary conversation”, legal interpretation, literary interpretation, and scientific theorizing, see Shaer (unpublished). In particular, Shaer argues that while ordinary utterance interpretation primarily involves a trade-off between cognitive effort and effect, legal interpretation also crucially involves trade-offs between institutional effort and effect. 133 cases, for example, the defendant may be facing both a pretty serious outcome – such as deportation – and an opponent with substantially greater resources – like the state. In such cases, then, the inequality between the practical situations of plaintiff and defendant gives rise to a considerably higher standard than mere “preponderance-of-the-evidence” – the evidence must instead be “clear and convincing”. And in criminal law cases, the outcomes are typically very serious and the defendant almost always faces an opponent with far superior resources. Thus, a very high standard of proof – proof “beyond reasonable doubt” – is in order. As I said before, we should not take the standards of legal proof to constitute the varying epistemic constraints on non-literal legislative assertion. Instead, we should see them as a clear indication that the epistemic constraint on non-literal assertion would – under many circumstances – be lower than argued by Marmor. And although what I have said here only provides an indirect argument for the claim that this particular epistemic constraint varies between legal domains, I think that – given the strong reason to expect such variation – the burden to proof is on those who claim that the constraint is invariant (either as a general matter or – at a minimum – in law). If that is correct, then we have good reason to conclude that that simply restricting the linguistic premise of Marmor’s argument does not make for a sufficient revision. Although Marmor’s argument cannot be successfully revised by simply restricting the main premise to the legal context, it is worth examining whether or not it can be successfully reconstructed using what I called the richness requirement. I will argue that 134 the richness requirement is false if it is taken to apply to human communication generally, but that it is quite plausible if we restrict it to law. The reason is that – typically – the kind of pragmatic enrichment we would be interested in in the legislative context would arguably require that lawmakers and their audience share a great deal of relevant information. Thus, reconstructing Marmor’s argument using the richness requirement and restricting it to the legal context gives us a very plausible revision of the argument. Or so I will argue. But first I need to show why it doesn’t suffice to just use the richness requirement unrestricted. 2.2 Reconstructing Marmor’s argument: The richness requirement In the first part of this chapter, I explained how the obviousness requirement produces a further requirement on the success of non-literal assertion when combined with the principle that it is obvious to a competent hearer that a speaker is intending to assert something other than what she literally says only if the speaker and hearer share a great deal of (relatively) specific background assumptions: RR On the assumption that, in some normal context C, something has been successfully asserted: In uttering a sentence ‘s’ with semantic content q in C, a speaker S does not assert q only if, in C, S and H share a great deal of (relatively) specific background assumptions. But if OR is false, then RR is not derivable in the way proposed. 135 One way, however, of replying to my critique of Marmor’s argument is to say that although RR cannot be derived from OR in the intended way, it is still independently plausible. It might even be argued that whatever the epistemic constraint on non-literal assertion turns out to be, that constraint is satisfied only if the conversational background is relatively rich. If that is correct, then Marmor’s argument can easily be reconstructed using RR instead of OR (stated informally): 128 P1. Normally, speakers assert something other than what they literally say only if they share a great deal of (relatively) specific background assumptions with their audience. P2. Legislatures rarely share a great deal of (relatively) specific background assumptions with their audience. C. Legislatures rarely assert something other than what they literally say. For the time being, I will accept P2 as true, although it may vary across the different legal contexts how much information is shared between speaker and hearer. Presumably, the 128 I should note that – strictly speaking – this is not a valid argument. It is really an inductive argument, given that P2 is a frequency claim. The correct form of the argument is the following (semi-formally stated): P1. Pr(q⏐p) = 1 P2. Pr(q) = x C. Pr(p) = x≥y The correct conclusion is that the probability of the antecedent is not greater than the probability of the consequent. Whether it is lower – and if so, to what extent – depends on further factors not considered in the argument under consideration here. For our purposes, however, there is no harm in stating the argument informally. 136 legislature shares more common ground, for example, with administrative agencies and courts than it does with ordinary citizens. That, then, leaves P1, which I believe is too strong, at least applied across the board. I will make my case in two steps, which is intended to highlight the two analytically distinct roles that shared information plays in non-literal assertion: the role it plays in determining that the speaker intends to assert something other than what she says and the role it plays in determining what the speaker intends to assert (instead). 129 Let us start with the first role. In order for it to be significantly more likely than not that a speaker isn’t intending to assert what she literally says, all the speaker has to do – assuming certain background conditions obtain – is to utter a sentence the semantic content of which is incompatible in some way with some part of the conversational background (things presupposed, purpose of the conversation etc.). Strictly speaking, then, all that is required – vis-à-vis the common ground – is that the speaker and hearer knowingly share one bit of information. They may, of course, normally share all sorts of information – both knowingly and unknowingly – but what is required for the conversational background to play this first role in non-literal assertion is very minimal. The most basic examples presumably involve quantification – both universal and existential. Consider, for example, universal quantification restricted – on the semantic level – only to humans, e.g. of the form Everyone is F. There are very few predicates that apply to absolutely everyone and so if a speaker utters the words, say, ‘Everyone is 129 Note that I think it is an open question whether either role is analytically prior to the other. I suspect that neither is prior to the other, or that – if so – it varies between types of speech acts. 137 coming to the party!’ it will be obvious to any competent hearer that she does not intend to assert the literal content of her remark (on the assumption that the speaker is observing the conversational maxims). And in order for this to be obvious, all that is required is that the speaker and hearer knowingly share the information that no party is such that everyone is coming to it. Consider also attributive uses of definite descriptions. Routinely, speakers use sentences containing such descriptions to say something true despite the fact that the descriptions don’t pick out any object uniquely. Say, for example, that a speaker utters the words ‘I parked the car in Lot B’. The literal content of the speaker’s remark is false, and this will be obvious to any competent hearer (again, assuming that she is observing the maxims). And all that is required here is that it be part of the common ground that the world contains more than one car. These examples – and many others like them – show that it is not generally true that the conversational background has to be fairly rich in order for it to play its first role in non-literal assertion. In many normal cases, only a minimal common ground is required for determining that the speaker intends to assert something other than what she says. Next, let us turn to the second role, the role that the conversational background plays in determining what the speaker intends to assert. Consider again the case in which a speaker utters the words ‘Everyone is coming to the party!’ Assuming it is significantly more likely than not that the speaker does not intend to assert the literal content of her remark, we can proceed to ask how much it takes – vis-à-vis shared information – to 138 determine what the speaker intended to assert instead. Now, depending on what the speaker in fact intends to assert, it presumably varies hugely how much common ground is required in order for the hearer to recognize the speaker’s communicative intention. In many cases, a rich conversational background is certainly necessary, but in other – perfectly normal – cases, the speaker and hearer have to share very little information. Say, for example, that – in addition to knowingly share the information that no party is such that everyone is coming to it – the speaker and hearer share the information that the speaker has recently sent out party invitations. In tandem with the fact that the speaker uttered the words ‘Everyone is coming to the party!’ these two bits of shared information give the hearer good reason to believe that the speaker intended to assert that everyone who was invited is coming to the party. Nothing more is strictly required. No doubt, in real-world situations similar to the one described, people frequently do share a great deal of related information (what kind of party, date and time, etc.), but the fact that people often assert something other than what they literally say in circumstances in which the conversational background is relatively rich does not entail that such a rich background is required for such assertion. I should also note that in many real-world cases in which the conversational background is relatively rich, the hearer’s warrant for concluding that the speaker is intending to assert that p is overdetermined, meaning that the context provides the hearer with multiple sufficient reasons to draw the relevant conclusion. Only one such reason, of course, is strictly required. 139 2.3 The conversational background as a constraint on non-literal assertion Although a rich conversational background is not – as a general matter – required for successful non-literal assertion, the conversational background will still limit the possibility of such assertion, by limiting the ways in which the conversational maxims can be violated and by limiting the range of plausible candidates for intended assertive content. This suggests that a rich conversational background is presumably required for rich pragmatic enrichment; in other words, the content of an assertion can only be as rich as the conversational background allows. We should note, however, that there are at least two types of pragmatic enrichment for which a rich background is needed. On the one hand, there are cases in which the material that gets added to the semantic content is relatively rich. Say, for example, that a speaker utters the words ‘Tom is too old’, intending to assert that Tom is too old to first run the Badwater Ultramarathon in Vibram FiveFingers shoes and then climb Mt. Whitney right away using the East Face route. This is a lot of content to add by way of pragmatic enrichment, and – presumably – the conversational background has to be very rich in order for the speaker to succeed. At a minimum, the speaker and hearer have to knowingly share the information that Tom is intending to first run the Badwater Ultramarathon in Vibram FiveFingers shoes and then climb Mt. Whitney right away using the East Face route. This bit of information is pretty rich in content, at least compared to the information, for example, that no party is such that everyone is coming to it or that the world contains more than one car. 140 On the other hand, there are cases in which material gets added to relatively rich semantic content. Say, for example, that a speaker utters the words ‘Tom is too old to first run the Badwater Ultramarathon in Vibram FiveFingers shoes and then climb Mt. Whitney right away’, intending to assert that Tom is too old to first run the Badwater Ultramarathon in Vibram FiveFingers shoes and then climb Mt. Whitney right away using the East Face route. This isn’t a lot of material to add, but in order for it to be significantly more like than not that the speaker is intending to assert something other than what she literally say, it must – in the first instance – be part of the common ground that it is not the case that Tom is too old to first run the Badwater Ultramarathon in Vibram FiveFingers shoes and then to climb Mt. Whitney right away using some route or other. Assuming that it is part of the common ground that there is some way that Tom could pull this off – e.g. by taking the Mt. Whitney trail – the hearer is prompted to look for evidence of what enriched proposition the speaker intended to communicate. And if it is for example common ground that the East Face route is the only truly grueling route up Mt. Whitney, and perhaps that Tom is likely to want to take that route, then the hearer will be in a position to recognize the speaker’s communicative intention. But it should be clear that – in this case – successful non-literal assertion requires a great deal of shared information, at least compared to the cases I described above. In the next section, I will argue that in the legislative context, questions about pragmatic enrichment are typically questions about enrichment of rich semantic content. If that is correct, then that will provide us with a natural way of revising Marmor’s 141 argument: we reconstruct it using the richness requirement but we also restrict that requirement to the legislative context. 2.4 Pragmatic enrichment in the legislative context Legislative texts are usually pretty complex, and even the least complex ones don’t come anywhere close to the brevity of sentences like ‘Everyone is coming to the party’ and ‘I parked the car in Lot B’. Consider for example the English Parliament (Qualification of Women) Act of 1918, which is probably among the shortest ones ever enacted: A woman shall not be disqualified by sex or marriage for being elected to or sitting or voting as a Member of the Commons House of Parliament. Most legislative texts carry considerably more information than the Qualification of Women Act and so we should, as a general matter, expect that pragmatic enrichment in legislation – if it occurs – consists in the enrichment of already rich semantic content. If that is correct, then pragmatic enrichment in the context of legislation will require a relatively rich conversational background. The cases of Holy Trinity and Ali, which I described earlier in the chapter, provide good cases in point. 130 In Holy Trinity, as we recall, the question was whether the Alien Contract Labor Act (ACLA) of 1885, 131 which prohibited the importation of “labor or 130 143 U.S. 457 (1892) and 552 U.S. 214 (2008), respectively. 131 23 Stat. 332 (codified as amended at 29 U.S.C. §2164 (1901)), superceded by Act of Mar. 3, 1903, Pub. L. No. 162, 32 Stat. 1213. 142 service of any kind” except “professional actors, artists, lecturers or singers” and “persons employed strictly as personal or domestic servants,” also prohibited the importation of people intended to serve as rectors or clergymen. According to many commentators, the Act only prohibited the importation of manual labor or service, and it is therefore natural to ask whether an argument can be made that the legislature asserted something other than what it literally said. (It is important to note that, at this point, I make no claims about whether or not the proposed requirements are met in the examples under discussion.) Here is the full text of the relevant section of the Act (which consists of six sections): [I]t shall be unlawful for any person, company, partnership, or corporation, in any manner whatsoever to prepay the transportation, or in any way assist or encourage the importation or migration, of any alien or aliens, any foreigner or foreigners, into the United States, its Territories, or the District of Columbia, under contract or agreement, parol or special, express or implied, made previous to the importation or migration of such alien or aliens, foreigner or foreigners, to perform labor or service of any kind in the United States, its Territories, or the District of Columbia. 132 Although the issue concerns the addition of very limited material – i.e. the content of the term ‘manual’ – the addition was allegedly made to content that is already very rich. As a result, the common ground has to be very rich in order for it to be significantly more likely than not that the legislature did not intend to assert the literal content of the text. 132 23 Stat. 332 (codified as amended at 29 U.S.C. §2164 (1901)), superceded by Act of Mar. 3, 1903, Pub. L. No. 162, 32 Stat. 1213. 143 And it has to be even richer in order for it to be significantly more likely than not that the legislature intended to assert that the prohibition concerned only manual labor. Recall also Ali, in which the Supreme Court ruled that Abdus-Shahid M.S. Ali could not sue the government for having lost property belonging to him. The reason, the court said, was that the Federal Tort Claims Act preserves immunity for all federal law enforcement officers, since it guarantees immunity for “[a]ny claim arising in respect of […] the detention of any goods, merchandize, or other property by any officer of customs or excise or any other law enforcement officer.” 133 Ali, of course, argued, and the dissent agreed, that the clause did not guarantee immunity for the government in relation to the actions of prison officers, but only of officers working in customs or related domains. Here is the full text of the relevant subsection of the Federal Tort Claims Act: Any claim arising in respect of the assessment or collection of any tax or customs duty, or the detention of any goods, merchandise, or other property by any officer of customs or excise or any other law enforcement officer, except that the provisions of this chapter and section 1346(b) of this title apply to any claim based on injury or loss of goods, merchandise, or other property, while in the possession of any officer of customs or excise or any other law enforcement officer, if— (1) the property was seized for the purpose of forfeiture under any provision of Federal law providing for the forfeiture of property other than as a sentence imposed upon conviction of a criminal offense; (2) the interest of the claimant was not forfeited; (3) the interest of the claimant was not remitted or mitigated (if the property was subject to forfeiture); and (4) the claimant was not convicted of a crime for which the interest of the claimant in the property was subject to forfeiture under a Federal criminal forfeiture law. 134 133 28 U.S.C. 2680(c); my emphasis. 134 28 U.S.C. 2680(c). 144 This section of the Act is – like the ACLA – very rich in content. As a result, if the proposed non-literal assertion is to be considered successful, the common ground must be rich as well, in order for the hearer both to determine that it is significantly more likely that not that the legislature was not intending to assert what it literally said and to determine what it was that it intended to assert instead. On the assumption that neither the ACLA nor the Federal Tort Claims Act are exceptional in any way, it seems safe to say that non-literal assertion in the legislative context requires that the legislature and its audience share a great deal of relevant information. In other words, if we restrict our attention to legislative utterances, we have a very plausible version of Marmor’s richness requirement: RR-Leg On the assumption that, in some normal legislative context C, something has been successfully asserted: In uttering a sentence ‘s’ with semantic content q in C, a speaker S does not assert q only if, in C, S and H share a great deal of (relatively) specific background assumptions. Or in plainer English: The legislature asserts something other than what it literally says only if it shares a great deal of common ground with its audience. The restricted richness requirement interacts with the flexible epistemic constraint in various ways to produce different requirements for the different areas of law. In criminal law, for example, the common ground has to be substantial, since the epistemic constraint is strict. In civil law, however, the common ground does not have to be quite as rich, since the constraint is weaker. Given these different requirements, it would be 145 natural to expect that the conditions for non-literal assertion are more favorable in civil law than in criminal law. When it comes to the US legal system, however, this expectation must be significantly tempered. Only a limited amount of civil law, for example, comes from legislative enactment. This means that in the area of law in which the relevant constraints on non-literal legislative assertion are weakest, there is – comparatively speaking – a very limited amount of legislative speech. And in criminal law, where most, if not all, law is statute-based, the constraints are very strict. In addition, most of criminal law is arguably addressed to the general public, with whom the legislature shares relatively little common ground (vis-à-vis lawmaking). I should note, though, that there is an important part of US civil law that is largely statute-based – namely administrative law. In administrative law, therefore, the epistemic constraint is weaker than in criminal law and there is arguably a considerable amount of common ground shared by speaker and hearer, which makes it all the more likely that the epistemic constraint is satisfied. Thus, we have good reason to expect that if the legislature ever successfully asserts something other than what it literally says, it will primarily be in the area of administrative law, and perhaps in the area of criminal law addressed to the officials. However, even if the lawmakers do share substantial common ground with both judges and administrative officials, there is reason to believe that the relevant background is often not quite rich enough to determinately satisfy the relevant requirements. As I will argue in the next section, there is a relevant difference in what the common ground must be like when we compare non-literal (descriptive) assertion and non-literal directives. 146 And although there is no structural difference between the two types of context as far as the pragmatic reasoning goes, it is still the case that, as a matter of fact, people tend to have less access to the relevant facts when it comes to non-literal directives – including, and especially, legal directives. Thus, although it is indeed natural to think that the conditions for non-literal assertion are most favorable in administrative law and least favorable in criminal law addressed to the general public, with criminal law addressed to courts somewhere in between, we should nevertheless considerably temper our expectations concerning the existence of actual, determinate cases of such assertion. In the next section, I will explain in some detail why that is. 2.5 The nature of legal directives Let us, following Bach and Harnish (1979), define directives as utterances that “express the speaker’s attitude toward some prospective action by the hearer and his intention that his utterance, or the attitude it expresses, be taken as a reason for the hearer’s action.” 135 Familiar Gricean reasoning applies in the case of directives, too, and so we need some plausible reformulations of at least some of the maxims for that particular context. In order to get at some such reformulations, let us look at a few examples. Say that a speaker instructs a hearer to ‘invite everyone to the party’ and that the attitude expressed is that of desire. Clearly, it would be obvious to any competent hearer that the speaker doesn’t desire that absolutely everyone be invited to the party. Or, to borrow and adapt an example from Manning (2003) (who borrowed and adapted it from 135 Bach & Harnish (1979), 41. 147 MacCallum (1966)), say that your boss tells you to ‘gather all the ashtrays you can find and bring them to the conference room’. It is obvious to you that your boss does not want you to bring all the ashtrays you can possibly find. Rather, she wants you to bring all the ashtrays you can find in the building, perhaps also assuming they belong to the company, that they are not bolted down, etc. To take a legal example, consider the case of Small v. United States, 136 in which the majority argued that the relevant legal directive could not – if interpreted literally – reasonably be regarded as the legislature’s view of how the addressees to behave. In Small, the question was whether petitioner Gary Small, who had been convicted for smuggling in a Japanese court in 1994, had – by owning a firearm – violated the US Criminal Code, 18 U.S.C. § 922(g)(1), which makes it “unlawful for any person … who has been convicted in any court, of a crime punishable by imprisonment for a term exceeding one year … to … possess … any firearm.” 137 The question before the court was whether the phrase ‘convicted in any court’ applied “only to convictions entered in any domestic court or to foreign convictions as well.” 138 The Court held that it applied only to domestic convictions and reversed the prior court’s judgment. The majority claimed that – as a matter of common sense – “Congress generally legislates with domestic concerns in mind.” 139 For our purposes here, we can characterize 136 544 U.S. 385 (2005). 137 18 U.S.C. § 922(g)(1). 138 18 U.S.C. § 922(g)(1). 139 Here, the Court is referring to Smith v. United States, 507 U.S. 197, 204, n. 5 (1993). 148 this “legal presumption” as a fact about the concerns of the legislature that is – via its commonsensical nature – part of the common ground between the legislature and officials (and perhaps even between the legislature and the general public, although that claim is more difficult to defend). Read literally, then, the relevant part of the Code clashes with the common ground, prompting a non-literal interpretation. On the assumption that Congress generally legislates with domestic concerns in mind, a reasonable hearer cannot attribute to Congress the view that officials ought to convict a person for owning a firearm if she has been sentenced in a court somewhere. Thus, read literally, the directive cannot be authoritatively binding. Or so the reasoning goes. In the examples above, the hearer is prompted to reinterpret the speaker’s directive because she cannot reasonably attribute to the speaker the attitude associated with the literal interpretation of the utterance. Thus, in the context of directives, the first Gricean maxim of Quality – which is normally the operative maxim in non-literal assertion – is perhaps best taken to be something like the following: Do not direct a person to do that which you do not desire/intend/etc. In the context of non-literal directives, then, the common ground has to contain information about the speaker’s relevant attitudes – i.e. about her desires/ends/values/preferences. However, in the legislative context, it is rarely the case that such facts are part of the common ground, even among lawmakers and officials – at least not determinately so. Thus, even if the general common ground – measured simply by the total information knowingly shared – is in some sense not particularly thin in either administrative law or in criminal law addressed to officials, it rarely determinately 149 contains the facts that are required for successfully issuing non-literal directives. We can say that in the case of legislation, the common ground relevant to generating such directives is not sufficiently rich. Here, I will not go into the familiar problems regarding epistemic access to the desires/wishes/preferences of other people. I have nothing to add to the discussion of these problems, but I will say that humans seem to be quite good at recognizing the preferences of others. It is one of the skills that allow us to adequately navigate in our interactions with other people. I know full well, for example, that my wife does not like pu-erh tea (and she knows that I know, etc.) and so when I ask her if I should make some tea, what I’m really asking is if I should make some of the tea she likes. In the case of legislation, however, matters regarding epistemic access to the relevant desires/values/aims are particularly problematic for at least two reasons. First, relatively few of law’s ends can truly be said to part of what is taken for granted in the way required for pragmatic reasoning. No doubt, some very general ends do have good claim to be part of the relevant common ground. Law’s aim to protect freedom of speech might be one. But relatively few are like that. Second, and more importantly, law’s ends compete with each other in extremely complex ways – so complex that the structure of “law’s preferences”, i.e. the relationship that holds between law’s various ends, is – at least in general – too complex to be part of what is taken for granted in legislative speech contexts. In order, however, for a non-literal legal directive to succeed, it is not sufficient simply that a literal reading be at odds with just some end that is taken for granted in the context. As Marmor (forthcoming) notes, it is “not necessarily irrational to express a wish 150 or intention that is in conflict with a hope or desire.” 140 The same applies in pragmatic reasoning about the linguistic content of a directive. It is not necessarily irrational for a speaker to express a wish or intention that is in conflict with a hope or a desire that she has, even if knowledge of that hope or desire is part of the common ground between speaker and hearer. Rather, the wish associated with the literal content of the directive has to be at odds with the proper balance of the relevant hopes or desires. In the case of a legal directive, then, the desire associated with its literal content has to be at odds with the proper balance of those ends of the law that can be said to be part of the common ground. 141 And very few facts about such balances can be taken for granted. Consequently, the common ground between the legislature and its audience rarely determinately contains facts of the kind required for (successfully) issuing non-literal directives. The above considerations suggest a reconstruction of Marmor’s second premise, which together with RR-Leg (Marmor’s richness requirement restricted to legislation) makes for what I believe is a sound revision of his argument: P1*. The primary communicative content of a statutory directive is different from its literal content only if the legislature and its audience share a great deal of (relatively) specific background assumptions regarding its desires/ends/values/preferences. 140 Marmor (forthcoming), 17. 141 Of course, there may be some obvious facts about “the law’s preferences” that are part of the common ground, at least between the legislature and officials. In some legal systems, for example, Lex Posterior, Lex Superior, and Lex Specialis each embody such a preference. 151 P2*. The legislature and its audience rarely share a great deal of (relatively) specific background assumptions regarding its desires/ends/values/preferences. C. The primary communicative content of a statutory directive is rarely different from its literal content. Thus, although it does not suffice simply to restrict Marmor’s original argument to the legislative context or to reconstruct it using his original richness requirement, we can combine these strategies to produce a plausible revision of Marmor’s first premise. The bottom line is that despite the fact that the conversational background between lawmakers and officials often contains quite a bit of information, it rarely contains very much of the kind of information required for lawmakers to issue non-literal directives. 152 CHAPTER FOUR: INDETERMINACY ABOUT UTTERANCE CONTENT After the discussion in the previous chapter, it is natural to ask what exactly happens – vis-à-vis the content of a legislative utterance – when the conversational background isn’t rich enough to warrant a reinterpretation. And it may seem equally natural to infer that in such cases, the primary content of the relevant statutory text just is its literal content. If that is correct and if the conversational background is generally not rich enough to warrant a reinterpretation, then the primary content of a statutory text normally just is its literal content. In this chapter, I aim to show that this would be a significant overstatement. To be sure, the way I set up my revision of Marmor’s argument did make it seem like there are very few cases in which it is indeterminate what the content of the law is: If the richness requirement isn’t satisfied, then the content of the law is – by default – its literal content. I hope to show, however, that once we analyze, on the one hand, the relationship between the common ground and the epistemic constraint on non-literal assertion and, on the other hand, the relationship between the epistemic constraint and a mutual expectation of literalness, we see that there is plenty room for indeterminacy about utterance content on the framework I have adopted. Although the details of this discussion can get somewhat complex, the underlying thought is rather simple: That one is not warranted in inferring that p does not entail that one is warranted in inferring that not-p – sometimes, one should simply withhold belief. When this happens in communication (with respect to the speaker’s communicative intention), then the primary 153 content of the relevant utterance will be indeterminate between the literal content of the sentence uttered and some non-literal enrichment thereof. In addition to explaining how – as a general matter – utterance content can be indeterminate in this way, I also hope to show how the framework allows us to identify such indeterminacy as a significant source of indeterminacy in the law. Finally, I will try to show why it is appropriate to label the relevant sort of indeterminacy as vagueness. 1. Revisiting the epistemic constraint on non-literal assertion In order to explain what happens when the richness requirement isn’t satisfied, it will be necessary to return to the discussion about the epistemic constraint on non-literal assertion. The reason is that – although for the purposes of revising Marmor’s argument it was possible to discuss several aspects of the richness requirement independently of epistemic requirements – the question whether the primary content of an utterance is (i) its literal content, (ii) some non-literal content, or (iii) indeterminate between the two is settled by how the conversational context interacts with the epistemic constraint. Basically, the epistemic constraint determines how likely it has to be that the speaker isn’t intending to assert what she literally says and the context determines how likely it actually is. Thus, what happens – in the first instance – when the richness requirement isn’t satisfied is that the epistemic constraint isn’t satisfied, and so it will pay off to continue the explanation of how indeterminacy about content arises with a focus on the (dis)satisfaction of the epistemic constraint. 154 The crucial thing to note is that epistemic constraints on non-literal assertion don’t exist in a vacuum; rather, they are part of a pragmatics framework, the task of which is to explain and predict how linguistic communication succeeds (and fails). So, if there exists such an epistemic constraint – as I assumed in the previous chapter – then something must explain why it exists, and why it has the weight/strength that it is taken to have. On the framework embraced by both Marmor and myself, the epistemic constraint on non- literal assertion is determined by what Bach and Harnish (1979) call a presumption of literalness: on their strong view, the mutual belief that if a speaker could (under the circumstances) be speaking literally, then she is speaking literally. 142 Thus, they say, if it is evident to the hearer that the speaker could not be speaking literally, the hearer appropriately reasons that she must be speaking non-literally and tries to infer what non- literal content she is intending to communicate. 143 Bach and Harnish formulate the literalness presumption strictly – normally, people just assume that speakers will intend to speak literally. This is the literalness presumption in its strongest form, since it stipulates that people assign a probability of close to 1 to the proposition that speakers intend to assert what they literally say. 144 On this invariant view, there isn’t really a need to talk about the literalness presumption being or not being of a particular strength – it suffices to just talk about the presence or 142 Bach & Harnish (1979), 12; see also 61. 143 Bach & Harnish (1979), 12. 144 Note that it can’t be the case that people assign a full probability of 1 to the proposition that speakers intend to assert what they literally say, because then – by the rules of Bayesian reasoning – no amount of evidence would suffice to change their minds. 155 absence of a literalness presumption (in a context). My own view is – unsurprisingly – that although there is indeed a presumption of literalness in linguistic communication, it is often considerably weaker than is postulated by the strong invariant view and its strength can vary by what is at stake in the relevant context. That is, while I think it is sometimes quite useful to distinguish only between contexts in which the literalness presumption is present – such as high-stakes context – and contexts in which it is absent – such as in a game of Bullshit – I also think that a less simplified picture is often appropriate. Even within contexts in which the presumption is present, it may vary to what degree people are expected to intend to speak literally, and this variation is often due to a variation of what is at stake in the relevant context. If that is correct, then – at least sometimes – it matters not just to identify the relevant type of context and what assumptions can be taken for granted in them, but also what weight these assumptions should be assigned. Be that as it may. What is clear is that Marmor, Bach and Harnish, and I are all committed to the view that (i) in normal speech contexts, there is some presumption of literalness, and (ii) the strength of the literalness presumption dictates the height of the epistemic bar. Thus, how much evidence is needed for non-literal assertion depends on the prior probability that the speaker is intending to speak literally. If the epistemic bar is high – i.e. if very strong evidence is required in order for the hearer to be warranted in inferring that the speaker is intending to communicate something other that what she literally said – then that is because there is a strong presumption that speakers intend to 156 speak literally. And if the epistemic bar is low, then that is because that presumption is weak. Although authors neither agree on how strong the literalness presumption is nor on whether or not it can vary by context, there still appears to be a widespread commitment to the claim that there is a presumption of literalness in the legislative context. That is, there appears to be widespread agreement that interpreters of statutory texts are warranted in having some expectation that the legislature to be intending to speak literally, 145 an expectation that is outweighed only by sufficient evidence to the contrary. So we can at least say this much: Anyone who is committed to the literalness presumption being of a certain strength is also committed to the epistemic bar having a particular height, and vice versa. This applies to invariant views and stake-sensitive views alike. Next, therefore, I want to analyze a bit further the relationship between the literalness presumption and the epistemic constraint. As I indicated above, I will analyze the literalness presumption in terms of the expectation that a speaker is intending to speak literally. If we then characterize expectations in terms of probabilities, it will be easy enough to see how the literalness presumption affects the epistemic constraint. It will also allow us to see clearly how what a speaker asserts may be indeterminate, something that my above revision of Marmor’s argument somewhat manages to obscure. 145 Consider for example, Justice Brewer’s acknowledgment in Holy Trinity that the claim that the plain meaning was intended had “great force”; or the dissent’s view in Weber that the language used provided powerful evidence of a corresponding communicative intention. In my view, such claims about the considerable weight of the language used as evidence of corresponding communicative intentions are vindicated by the existence of a literalness presumption, analyzed in terms of the prior probability that the speaker intends to communicate the literal content of her remark. 157 2. The relationship between the literalness presumption and the epistemic constraint on non-literal assertion Let us, for the sake of illustration, assume that the presumption of literalness is invariably strong: in any given context, the (prior) probability that the speaker is intending to assert what she literally says is, say, .9. Then, in order for it to be significantly more likely than not that the speaker isn’t intending to assert what she literally said – conditional on the fact that the she said what she said in the relevant context (i.e. taking everything into account) – the evidence available to the hearer in that context (excluding the fact that she said what she said) has to make it approximately ≤.05 probable that the speaker is intending to assert what she literally said. 146 Thus, we see how a strong literalness presumption spawns a very high epistemic bar on non-literal assertion: In this case, the assertive content of a speaker’s utterance is the literal content of the sentence uttered unless there is approximately a ≤.05 probability that the speaker is intending to assert what she literally said. Getting back to Marmor’s view, we can therefore see that if the epistemic constraint is supposed to be invariantly close to 0 (i.e. if it has to be close to impossible that the speaker is intending to assert the literal content of her remark), then the literalness expectation has to be somewhere around .97 – just marginally less than 1. This seems to me implausibly strict, even in criminal law where the stakes are extremely high. 146 Here is the Bayesian “equation” underlying the calculation of whether or not there is – all things considered – sufficient evidence that the speaker did not intend to assert what she literally said: It is – all things considered – significantly more likely than not that a speaker S is not intending to assert the literal content, p, of the sentence ‘s’ uttered (in a context C) iff pr(S is intending to assert p (in C)) ⎥ S uttered ‘s’ (in C)) is approximately equal to or greater than .6. 158 Thus, although a high bar view need not be implausible in itself – particularly if it is confined to contexts in which much is at stake – such a view would presumably have to settle for a lesser constraint than obviousness. It is of course very difficult to determine with confidence how strong the literalness presumption is in the various contexts, and since the relevant probabilistic concepts are vague it will of course be impossible to map them onto precise probabilities. Most likely, the best we can do is to use precise probabilities to explain the relationship between the literalness presumption and the epistemic constraint and to say that we should analyze the strength of the presumption in terms of a non-sharp set of probabilities that varies somewhat between legal domains. I should note, though, that we do seem to have a reasonable idea about the upper and lower bounds of the probabilities. It seems that anything over (approximately) .9 seems implausibly strong. As I said above, I see no reason to believe that – as a general matter – people enter into conversation with a near certain expectation that speakers will be intending to speak literally, not even if a lot is at stake. That’s just not how language works – as competent language users we know that even in high stakes scenarios people sometimes intend to communicate something different from what they literally say. So we seem to have an approximate upper bound on the literalness presumption: around .9. And unless there is flat out an expectation that speakers are not speaking literally, the epistemic constraint also has an approximate lower bound. Even assuming indifference about literalness – i.e. a .5 expectation that speakers intend to speak literally in the relevant context – will produce an epistemic constraint of around 0.6, or else it will 159 not be significantly more likely than not that the speaker is not intending to assert what she literally said (conditional on the fact that the she said what she said in the relevant context). Thus, the plausible range for the literalness expectation appears to be between .5 and around .9, and that – consequently – the corresponding range for the epistemic constraint is – roughly – between .6 and .95. One of the upshots of establishing this is that it resolves an important issue that was left unresolved by the discussion in the previous chapter – it shows that the varying standards of legal proof cannot constitute the epistemic constraints on non-literal assertion, because no such epistemic constraint can be equivalent to the preponderance- of-the-evidence (>.5) standard of proof. Thus, we should rather take the varying standards of proof to be good indication that the height of the epistemic bar and the strength of the literalness presumption vary between legal domains. 3. Pragmatics and indeterminacy about assertive content Now that I have explained the relationship between the strength of the literalness presumption and the weight of the relevant evidence available in a given context, we are in a better position to understand how borderline cases arise within the framework. Again, any use of precise probabilities is for ease of explanation only. For the purposes of illustration, let us say that the strength of the literalness presumption is – say – around .9 and that – consequently – the epistemic constraint is around .95. If, then, the available evidence (excluding the fact that the speaker said what she said) makes it anywhere between approximately .05 and .15 probable that the speaker 160 is intending to assert what she literally said, then it is – all things considered – neither significantly more nor significantly less likely than not that the speaker is intending to assert what she literally said. And if it is – all things considered – neither significantly more nor significantly less likely than not that the speaker is intending to assert what she literally said, then the hearer is justified only in withholding belief about the speaker’s communicative intention. Consequently, since assertive content is partly determined by the inferences that the hearer is warranted in making about the speaker’s communicative intention, the assertive content in such cases is indeterminate. Thus, the relevant region of “evidential indifference” makes for a corresponding region of indeterminacy about utterance content. The same reasoning applies if we lower the epistemic constraint (or heighten it for that matter). If, for example, we take the strength of the literal presumption to be .8, then the relevant range is approximately between .15 and .25. In case the available evidence makes it .2 probable that the speaker is not intending to assert what she literally said, we are faced with a borderline case. So, in general, no matter the strength of the literalness presumption, for any (approximate) .10 range below the epistemic bar, it will be indeterminate whether the content of the utterance is the literal content of the remark or some pragmatic enrichment thereof. 161 4. Indeterminacy and levels of pragmatic reasoning At this point, it is important to note that indeterminacy can enter into the picture at two different levels, given the two roles that the common ground plays in pragmatic reasoning. It can either be indeterminate that the speaker asserted something other than what she literally said, as explained immediately above, or it can be indeterminate what pragmatically enriched proposition the speaker asserted. As discussed in chapter three, these notions are – although closely related – analytically distinct. Let’s first revisit two significant legal cases, also discussed in the previous chapter, that I believe are best analyzed as cases in which it is indeterminate that the legislature asserted something other than what it literally said: Church of the Holy Trinity v. US and Ali v. Federal Bureau of Prisons. In the context of Holy Trinity and the Alien Contract Labor Act, it was presumably pretty clear to all parties that what Congress wished to accomplish was to prevent the importation of cheap unskilled labor. Further, as Vermeule (1998) notes, the Court later argued that “in no event could [the Act] be construed as applying to a contract for the services of a rector or a pastor of a religious corporation,” 147 suggesting that it took it to be common knowledge that Congress did not wish to “prohibit the immigration under contract of a Christian minister.” 148 This is all significant evidence that Congress did not intend to enact the literal content of the Act. Still, the Act did include a number of specific exceptions that cannot count as exceptions on a pragmatically enriched reading of the Act. An exception for professional artists, for 147 US v. Laws, 163 U.S. 258, 264. Cited by Vermeule (1998), 1842, footn. 41. 148 Vermeule (1998), 1842. 162 example, cannot be an exception to a law that prohibits the importation of manual labor or service. This significantly undercuts the evidence mentioned so far (at this point, I wish to remain silent about whether or not the explicit exceptions provide evidence that Congress did intend the relevant part of the Act to be taken literally). 149 Given the few facts that can appropriately be said to have been part of the common ground between Congress and the Court (in the sense of being taken for granted), it seems to me reasonable to say that, in Holy Trinity, the common ground was neither rich enough to make it significantly more likely than not that Congress did not intend to prohibit the importation of all types of labor or service, including those of rectors and clergymen, nor rich enough to make it significantly more likely than not that they positively did intend to do so. Taking everything into account, the evidence seems to favor neither claim over the other. If that is correct, the content of the Alien Contract Labor Act was to some extent indeterminate. There is simply no answer to the question ‘Does the Act apply to people intended to serve as pastors?’ Similar reasoning applies in Ali, I believe. It seems safe to say that it was part of what was (reasonably) taken for granted both by Congress and the Supreme Court that, in enacting the Federal Tort Claims Act, Congress wished – as Justice Kennedy of the dissent put it – to “[allow] those who allege injury from governmental actions over a vast sphere to seek damages for tortious conduct.” 150 Citing the case of Kosak v. United States, 149 As it turns out, there was also significant debate about the issue in Congress, but since the legislative history was not part of the common ground, it did not – on my view – play a part in determining the content of the law. 150 552 U.S. 214 (2008). 163 Justice Kennedy also pointed out – rightly, I think – that “unduly generous interpretations of the exceptions [established by the FTCA] run the risk of defeating the central purpose of the statute.” 151 The majority, however, also pointed out that since the exception relevant to Ali concerned the detention of property, it was far from obvious that interpreting the exception to apply to absolutely all law enforcement officers would go against the wishes of Congress. Thus, given the fact that Congress uttered the relevant text in the relevant context, it is arguably neither significantly more likely than not that Congress intended federal immunity from the relevant type of tort claims to extend only to officers working in customs or related domains nor that it intended such immunity to extend to officers of all kinds. If that is correct, then the Federal Torts Claim Act is indeterminate in that respect. There were of course a number of other issues that were relevant to deciding the case, but since questions about non-literal statutory content turn just on the language used and the common ground, these issues concerned the proper interpretation of the Act, given that it was indeterminate what the exact content of it was. In addition to cases in which it is indeterminate whether or not the speaker asserts something other than what she literally says, there are also cases in which it is significantly more likely than not that the speaker did not intend to assert what she literally said but indeterminate what enriched proposition she asserted instead. Consider for example a case in which a speaker utters the following: ‘Everyone is coming to the party!’ And let us say that the hearer is in charge of supplying chips and salsa for said party. Now, let us further stipulate that – given the context – it is indeterminate whether 151 465 U.S. 848, 854 (1984). 164 the speaker intended to assert that everyone in the department is coming to the party or that everyone in the group of students is coming to the party. In this case, what the speaker asserted is indeterminate, but it is still determinate that she did not assert the literal content of her remark. To take an analogous legal example, consider the case of United States v. Gonzales. 152 In Gonzales, the question was whether or not the Court could order that a federal prison sentence could run concurrently with a state imposed sentence, despite the fact that a section of 18 U. S. C. §924 states that a sentence imposed under it “shall [not] . . . run concurrently with any other term of imprisonment.” The majority held that it could, but Justice Stevens of the dissent argued that the statute could not “reasonably be interpreted as containing any command to state sentencing judges or as requiring the suspension of any federal sentences when concurrent state sentences are later imposed.” 153 “[C]ommon sense,” he said, “requires us to reject a purely literal reading of the text.” 154 I take this to indicate that Justice Stevens thought that it was obvious that – taken literally – the code has certain legal consequences and that it was obvious that these consequences clashed with Congress’s wishes. If – for the sake of argument – we concede that Justice Stevens is correct, then it is indeed significantly more likely than not that Congress did not intend the statute to be understood literally. However, as he rightly notes, it still remains to be determined what 152 520 U.S. 1 (1997). 153 520 U.S. 1 (1997). 154 520 U.S. 1 (1997). 165 non-literal content was intended, and this may well make a difference to the outcome of the case. According to Justice Stevens, there were two plausible candidates. Either Congress intended to use the phrase ‘any other term of imprisonment’ to mean any other term of imprisonment that has already been imposed or to mean any other federal term of imprisonment. Only the latter was in favor of the respondent. Justice Stevens cited three reasons why he thought that it was more likely than not that Congress intended to use the phrase to mean any other federal term of imprisonment. However, he did not go so far as to claim that it was obvious that Congress could not have intended to use the phrase in the former way. (Two of these reasons actually had to do with legislative history, so only one of the reasons had a genuine bearing on question of legal content.) Since Gonzales was a criminal case, and the epistemic constraint is therefore high, this arguably means that it was not significantly more likely than not that Congress intended to say that a sentence imposed under the relevant code “shall [not] . . . run concurrently with any other [federal] term of imprisonment.” But it certainly does not mean that it was significantly more likely than not that Congress intended to say that a sentence imposed under the relevant code “shall [not] . . . run concurrently with any other federal term of imprisonment [that has already been imposed].” As a result, it is determinate that the content of the statute is not its literal content but indeterminate what exactly the content of the statute is (on the assumption that Justice Stevens is right). 166 5. Indeterminacy about content and the notion of a well-functioning legal system There is a prima facie worry here. Neither the Alien Contract Labor Act, the Federal Torts Claim Act, nor 18 U. S. C. §924 seem like unique – or exceptional – cases. If that is correct, then indeterminacy about the content of the law appears to be rampant. If such indeterminacy is seen as a failure – e.g. as a failure on behalf of lawmakers to communicate to subjects what their legal rights and obligations are – then this would indicate a widespread failure in the legal system. Hasn’t something gone seriously wrong if it is the normal state of affairs that the persons to whom the law is addressed are not able to fully determine what the lawmakers intended to communicate? The answer, I think, is no. Scenarios like the ones we have been considering do not necessarily indicate a general pragmatic failure – neither in ordinary conversation nor in law. In the “salsa case”, for example, the reason for the speaker’s assertion – the knowledge of which we stipulate to be shared by speaker and hearer – is that she wants the hearer to be able to bring enough chips and salsa to the party. In case the difference between the two sets of people is immaterial, vis-à-vis the relevant purpose, nothing has really gone wrong. And although the speaker presumably had in mind one of the two candidate propositions, the student-to-faculty ratio in the department may be high enough so that it does not matter that the hearer has insufficient evidence about which enriched proposition the speaker intended to communicate. 155 155 I’m assuming in this case – safely, I think – that it has no non-instrumental value for the speaker to have her communicative intention fully recognized. 167 Similarly, despite being indeterminate to some extent, the Alien Contract Labor Act still determinately applies to manual labor. And the immunity established by the Federal Torts Claim Act still determinately extends to officers working in customs or related domains. Finally, the section of the 18 U.S.C. §924 that states that a sentence imposed under it “shall [not] . . . run concurrently with any other term of imprisonment” still determinate prohibits that a sentence imposed under it run concurrently with any other federal term of imprisonment that has already been imposed. The general principle is that even if it may be indeterminate what the content of a particular statute is, the statute still determinately applies to those cases to which it applies on any reasonable enrichment. Ultimately, then, it is the practical context that determines if, and to what extent, something has gone wrong, and cases like the salsa case appear to be a regular – and relatively unproblematic – feature of successful human communication. Similarly, indeterminate legal directives like the ACLA, the FTCA, and 18 U.S.C. §924 are regular features of any legal system. Insofar as the number of actual borderline cases resulting from such directives is manageable, such indeterminacy is also relatively unproblematic. Although the law is often to some extent indeterminate, due to limited common ground, this fact will not necessarily prevent it from functioning reasonably well. As a general principle applying both to the case of ordinary conversation and to the legislative context, we can perhaps say that as long as a rational hearer, knowing the conversational background and context, is able to extract sufficient information from the speaker’s 168 utterance, relative to the purpose(s) of the conversation, the relevant speech act counts as reasonably successful. 6. Indeterminacy about speech act content as vagueness In this chapter, I have been discussing what I have called indeterminacy about the content of a speech act. In this section, I want to explain why I take such indeterminacy to be a case of vagueness. I believe this needs to explained because – on the face of it – indeterminacy about the content of a speech act seems rather removed from familiar cases of vagueness (such as borderline cases of ‘blue’, ‘crowd’, ‘vehicle’, etc.). However, although indeterminacy about content is indeed in some respects special, I think it can be shown that it is really a form of vagueness of content – namely, vagueness of speech act terms like ‘say’, ‘assert’, ‘command’, etc. I should emphasize that this claim applies only to speech act terms that are associated with epistemic constraints. As with any type of indeterminacy, there are certain questions regarding indeterminacy about content to which there are no answers. If it is indeterminate, for example, whether S asserted p or p*, then there will be no answer to the questions ‘Did S assert p?’/’Did S assert p*?’ Correspondingly, the propositions that S asserted that p and that S asserted that p* have no truth-value – or in more technical parlance: the phrase ‘S asserted that’ is undefined for the value-pairs <C, p> and <C, p*> (‘C’ denotes the relevant context). However, in order for this type of indeterminacy to count as vagueness proper, it doesn’t suffice that there are contents for which the relevant speech act phrases (e.g. ‘S 169 asserted that’) are undefined. The terms must also be “tolerant”, in the sense that certain minute changes in the facts cannot make a difference to their proper application. In the case of speech act terms/phrases associated with epistemic constraints – such as ‘S asserted that’ – tolerance is due primarily to vagueness in the intuitive probabilistic concepts involved. Recall that assertion was partly defined in terms of epistemic warrant: Assertion A speaker asserts p only if a rational hearer, knowing the conversational background and context, is warranted in taking her to be intending to assert p. And epistemic warrant, in this context, was defined partly in terms of a vague probabilistic concept: EW A rational hearer, knowing the conversational background and context, is warranted in taking a speaker to be intending to assert p only if it is significantly more likely than not that the speaker is intending to assert p. The phrase ‘ It is significantly more likely than not that’ is obviously vague. To be sure, there are very many cases for which the predicate is defined, but there are also many borderline cases. Say for example that, given the evidence in a context, it is 80% likely that S intended to assert p and 20% likely that S intended to assert something else. It is 170 clearly significantly more likely than not that S intended to assert p. Imagine, then, that there is a small change in the context and that the result is that it is now 79% that S intended to assert p. Such a small change doesn’t change anything with respect to whether or not it is appropriate to say that it is significantly more likely than not that S intended to assert p. But if we introduce enough small changes into the context, we can end up with a likelihood, say, of 57%, in which case it may – depending on what is at stake – be indeterminate whether or not it is significantly more likely than not that S intended to assert that p. And if we keep on going, we get a full-fledged Sorites-series. Holding constant what is at stake, it is for example clearly not significantly more likely than not that S intended to assert p if the likelihood of it is an even 50 percent. It seems, then, that there is no sharp cut-off point between the determinate extension and the determinate anti-extension of the phrase ‘It is significantly more likely than not that’. If that is correct, then indeterminacy about content is ultimately a form of vagueness – namely vagueness in the relevant speech act terms (i.e. those terms associated with an epistemic constraint). And this vagueness is due primarily to the vagueness of the probabilistic notions used to define the relevant epistemic constraints. I would like to conclude this brief chapter by explaining why I have chosen to distinguish between vagueness of content and vagueness about content rather than semantic and pragmatic vagueness, respectively. The reason is quite simple: the former distinction tracks what I am after in my explanation of the relevant types of vagueness, whereas the latter does not. First, semantic vagueness in a sense becomes pragmatic vagueness once the relevant terms are used in conversation. The content of an utterance 171 of the sentence ‘Roses are red’, for example, is vague because the word ‘red’ is semantically vague. So the distinction between semantic and pragmatic vagueness is not really mutually exclusive. Second, and more importantly, pragmatic vagueness is not confined to vagueness about content. Pragmatic vagueness can for example include pragmatic enrichment of a “vaguefying” propositional constituent, such as when the sentence ‘We should meet at 5 o’clock’ is used to communicate that we should meet at approximately 5 o’clock. Thus, for my purposes in this dissertation, the distinction between vagueness of content and vagueness about content is – albeit a bit unwieldy – the appropriate distinction to use. 172 CHAPTER FIVE: TEXTUALISM, CONTENT, AND INTERPRETATION So far, I have been assuming that the legal content of a statute – i.e. the content that its enactment contributes to the law – is, or directly corresponds to, the content that the legislature communicates to its intended audience by enacting the relevant text. And in the previous two chapters, I hope to have showed that – typically – the primary communicative content of a statutory enactment is either the literal content of the relevant text or it is indeterminate between that content and some non-literal enrichment. My view may therefore seem to imply a version of textualism. And in one sense it does, at least insofar as textualism is taken to be a theory of legal content rather than legal interpretation (on the understanding that legal interpretation is the act of developing the law in the face of indeterminate/inconsistent legal content or a particularly problematic result). But textualism is more than just a theory about legal content – it is primarily a normative theory of adjudication of which such a theory is a part. And as I will try to make clear, my view differs significantly from full-fledged textualism. Part of the reason is that I think that legal content is often less determinate than textualist rhetoric suggests. Mainly, however, the difference is that I don’t think that the standard textualist considerations support their limited view of the judge’s role in all cases in which interpretation is called for. I begin the chapter by explaining what textualists think that judges shouldn’t do and why. This will give us a sense of what textualists think is at stake in the debate. I then 173 explain what I call the textualist thesis of legal content: the thesis that the legal content of a statute is the linguistic content that a reasonable member of the relevant audience would, knowing the context and conversational background, associate with the enactment. I note how little argument has been given for this thesis and proceed to provide one that I find to be fairly straightforward, given certain assumptions about law and language. But while I think the thesis itself is correct, I also note that the willingness of contemporary textualists to recognize pragmatic factors as contributors to legal content appears to get them into a certain amount of trouble. In particular, the way in which they conceive of the context and conversational background of legislative utterances indicates that – despite their rhetoric – they do not care only about the actual communicative content of statutory enactments. Next, I apply the pragmatic framework underlying the thesis to two longstanding issues in the debate about textualism. First, I argue that the framework allows us to show that there is – contra critics – a principled distinction between the different textualist reasonings in the cases of Green v. Bock Laundry Machine Co. and United States v. Marshall. 156 Second, I argue that the framework confers a non-negligible amount of plausibility on what Perry (2011) calls conception-textualism – a view that he has pronounced “confused, implausible, and unworkable.” 157 Finally, I argue that when it comes to judicial decision-making in cases of vagueness, the standard textualist considerations against judicial lawmaking do not apply. 156 490 U.S. 504 (1989) and 908 F.2d 1312 (7 th Cir. 1990), respectively. 157 Perry (2011), 106. 174 As a result, judges will have to look elsewhere for guidance about how to resolve such cases. 1. Textualism and legislative intentions Textualism, as a normative theory of adjudication, has both a critical aspect and a constructive aspect. The critical aspect of the theory tells us what judges shouldn’t do and why, while the constructive aspect tells us what judges should do and why. As for the critical aspect, textualism holds that there are primarily two things that judges shouldn’t do: they should neither (i) attempt to attribute purpose and meaning to a statute by putting themselves “in imagination in the position of the legislature which enacted the measure,” 158 nor (ii) consult legislative history in the hope of finding evidence of the lawmakers’ actual communicative (i.e. illocutionary) intensions or actual further (i.e. perlocutionary) intentions in enacting the statute. There are several reasons why textualists maintain that judges should not attempt to attribute purpose and meaning to a statute by putting themselves in the shoes of the enacting legislature, but the weightiest one is presumably that they think that the more judges engage in the “imaginative reconstruction” of legislative intent, the more they are assuming authority that they do not possess, insofar as their speculations lead to binding judicial decisions. That is, in doing so – and to the extent that they do it – they are making assumptions that are not theirs to make, since they are neither democratically 158 Hart & Sacks (1994), 1414. 175 elected representatives of the people nor authorized by Article 1, Section 7 of the Constitution to make law. Thus, says the textualist, both respect for democracy and for legislative supremacy requires that judges refrain from engaging in imaginative reconstruction of legislative intent. The argument has also been made that judges shouldn’t allow such reconstruction to influence their judicial decisions because that would diminish the separation of powers (legislative and judicial). Finally, some have argued that it lessens predictability, vis-à-vis application of the law in question, and thereby significantly frustrates the ideal of the Rule of Law. According to the textualist, then, judges should – for important political reasons – not attempt to attribute purpose and meaning to a statute by putting themselves in the shoes of the enacting legislature. But why are judges not permitted to look for concrete evidence of actual legislative intentions by consulting legislative history, given the fact that inference on the basis of concrete evidence seems like a much more constrained activity than imaginative reconstruction? There are two primary reasons why textualists don’t think judges should consult legislative history in the hope of finding evidence of the legislature’s intentions (whether illocutionary or perlocutionary). First, they think that there are – due in significant part to the adversarial nature of legislative institutions in modern democratic societies – very few intentions that can in good faith be attributed to the legislature as a collective. Legislative history is, according to the textualist, not a window into the coherent, collective mind of Congress; on the contrary, it usually contains evidence of several competing viewpoints and the danger is that judges will pick and choose those viewpoints that support the legal 176 outcome they prefer. Thus, judges tend (consciously or unconsciously) to select evidence from legislative history according to the own preferences, and in doing so – the textualist thinks – their actions frustrate the aforementioned ideals: democratic legitimacy, legislative supremacy, separation of powers, and the Rule of law. Second, legislative history not only contains evidence of several competing viewpoints open to cherry-picking, but also contains strategically placed statements intended to influence the decisions of judges who choose to consult legislative history. Politicians have become very good at manipulating legislative history in their favor and it therefore provides, according to the textualist, too little reliable evidence of the legislature’s intentions to be worth consulting. 2. Textualism, legal content, and utterance content As for the constructive aspect of the theory, textualists hold that judges should – in the first instance – determine the content of the relevant statute, which in their view amounts to determining the meaning of the statutory text. And the meaning of a statutory text, according to contemporary textualists, is the linguistic content that a rational and competent speaker of the English language would, knowing the relevant context and conversational background, have associated with the relevant text at the time of enactment. 159 That is, they think that legal content is determined by the objective content of the legislative utterance, which – in turn – is partly determined by contextual factors. 159 See e.g. Easterbrook (1990), 443; see also Manning (2005), 434. 177 As Scalia (1997) puts it, “the import of language depends upon its context, which includes the occasion for, and hence the evident purpose of, its utterance.” 160 As it turns out, however, there is a remarkable dearth of direct argument for the textualist claim that the legal content of a statute is the objective content of the respective utterance (i.e. the enactment). And although some rather obscure hints at arguments can be found in the literature, 161 many textualists seem content to focus primarily on arguing against the alternatives (e.g. that legal content is determined by the intentions of the lawmakers) and then more or less simply propose that judges should instead focus on determining the content that a rational hearer would, knowing the relevant conversational background and context, associate with the statutory text. 162 Nevertheless, I think that the textualist thesis of legal content is basically correct and that a fairly straightforward argument can be provided. The argument is in many respects quite simple, especially if we are permitted to take for granted that enactments are communicative acts: P1. The legal content of a statute S is p iff, in enacting S, the legislature has communicated that p. P2. In enacting a statute S, the legislature communicates that p iff a rational hearer, knowing the relevant conversational background and context, is warranted in taking the legislature to have intended to communicate that p (in enacting S). 160 Scalia (1997), 144. 161 See e.g. Manning (2006), 79. 162 See e.g. Easterbrook (1988). 178 C. The legal content of a statute S is p iff a rational hearer, knowing the relevant conversational background and context, is warranted in taking the legislature to have intended to communicate that p (in enacting S). Following Greenberg (2011), call this the communicative content theory of law. The first premise claims that the legal content of a statute is – or directly corresponds to – the communicative content of the enactment – whatever the best definition of such content turns out to be. I take the truth of P1 to follow from the fact that, necessarily, lawmaking is a communicative speech act. Many authors take this to be uncontroversial, although I should note that there are a few dissenters – Greenberg probably being the most forceful one. 163 The second premise follows from one of the leading contemporary theories of communicative content – a theory that embraces what we can call an objective notion of communicative content. According to this theory – which I adopted in chapter three – the content that a speaker counts as having communicated is determined by the inferences that a rational hearer, knowing the context and conversational background, is warranted in making about the intended content of the utterance. It is interesting to note that in adopting the objective theory of communicative content, we get the textualist thesis – i.e. C – almost “for free” (assuming P1 holds). What I mean by this is just that although P2 generates some deep commitments in the philosophy of language, it is – unlike P1 – not a jurisprudential claim; it is neither a claim about the nature of law/lawmaking nor a normative claim about why the content that a 163 Followed closely by Hurd (1990). 179 rational hearer would associate with the legislation is the content that matters for the purposes of the law. P2 simply follows from what many take to be the best available theory of communicative content. Now, although I – along with many others 164 – think that the objective theory of communicative content is the best available theory, some authors do favor subjective theories, i.e. theories according to which the communicative content of a speaker’s remark is simply the content that she intended to communicate. 165 A speaker – in uttering a sentence – means, says, asserts, etc. what she intends to means, say, assert, etc. Often, the content of a speaker’s communicative intention will (sufficiently) match the content that a rational hearer will associate with the speaker’s remark, in which case it does not really matter with what the content of the law is identified. But sometimes there is a (significant) mismatch, in which case it does matter. And in order to successfully argue that – from the point of view of the law – the content that matters is the content that a rational hearer would have associated with the legislative utterance, normative argument must presumably be offered. This is one way, then, in which relatively subtle choices between linguistic theories can have significant consequences for theorizing in philosophy of law. As it is beyond the scope of this dissertation, I won’t venture to argue for the objective theory or to argue against the subjective one. Rather, I will continue to assume 164 Such as Soames (2011) and Marmor (forthcoming). 165 See e.g. Neale (2005). 180 that the former is the better one and proceed to explore how far the resulting theory of legal content can take us/the textualist. 3. Textualism and pragmatic enrichment So far, I haven’t said much about what sort of content can, according to the textualist, form part of the objective content of a legislative utterance. Contemporary textualists, however, are very explicit about the fact that they believe that such content can go beyond the literal content of the sentence uttered. Quoting Scalia (1997), for example, “the good textualist is not a literalist.” 166 If that is true, then textualists must allow that some types of pragmatic content (sometimes) form part of the content of the law. And as I will try to show, standard textualist reasoning in prominent legal cases suggests that they are at least willing to concede that insofar as such content forms part of the primary content of the relevant utterance, pragmatic content can form part of the content of the law. In other words, pragmatic enrichment (of literal content) can, according to the contemporary textualist, contribute to the content of the law. 167 Let us first consider Scalia’s famous dissent in the case of Smith v. United States. 168 During a drug trafficking crime, Mr. Smith had exchanged a firearm for drugs and the question was whether or not he had thereby violated 18 U.S.C. § 924(c)(1) of the Federal Code, which mandates certain penalties if the defendant “during and in relation to 166 Scalia (1997), 24. See also Manning (2005), 434. 167 Here, I will not be concerned with the question whether pragmatic presupossitions or implicatures can form part of the content of the law. For a discussion, see Marmor (2011a). 168 507 U.S. 197 (1993). 181 . . . [a] drug trafficking crime[,] uses . . . a firearm.” 169 In the case of a firearm of the sort used for bartering by Mr. Smith (a MAC-10), the mandatory sentence is 30 years. The majority held that Smith had indeed violated the statute, arguing that its content did not specify any particular way in which the firearm in question had to be used. All that was required was that Smith used it in some way or other. And if the legal content of a statute is confined to its literal content – perhaps including semantic presuppositions – then the majority’s position appears to be correct. Scalia, on the other hand, argued that the communicative content of the statute was a specific pragmatic enrichment of its literal content, and that in order to violate the Code, Smith would have had to use the firearm as a weapon. “When someone asks “Do you use a cane?” he is not inquiring whether you have your grandfather’s silver handled walking stick on display in the hall; he wants to know whether you walk with a cane. Similarly, to speak of “using a firearm” is to speak of using it for its distinctive purpose, i.e., as a weapon.” 170 What textualists are after, then, is not literal content but the content that a rational member of the relevant audience would, knowing the relevant context and conversational background, associate with the legislature’s utterance. In pragmatic terms, Smith was a case of so-called required enrichment – a case in which the literal content of the sentence uttered has to be pragmatically enriched in order 169 18 U.S.C. § 924(c)(1). 170 507 U.S. 197 (1993); Scalia dissenting. 182 for the speaker to count as having successfully conveyed a complete proposition. 171 A verb like ‘use’, for example, may occur in a sentence with a so-called purpose parameter, but may also occurs without such a parameter – compare for example ‘He used a hammer to pound in the nails’ and ‘He used a hammer’. In the latter case, the literal content of the sentence does not express a complete (i.e. truth-evaluable) proposition and pragmatic enrichment is thus required. The general explanation for this systematic necessity of pragmatic enrichment is that speakers routinely compress intended communicative content into linguistic structures, or forms, which subsequently require pragmatic enrichment in order to be expanded. 172 Now, with a term like ‘use’, there will of course be a range of possible enrichments, corresponding (at a minimum) to the range of possible parameters: He used a hammer {to pound in the nails}/He used a hammer {as a door prop}/He used a hammer {as a weapon to fight off the burglar}/etc. In all these cases, we look to context and the rules of conversation – i.e. to pragmatic factors – in order to determine the speaker’s illocutionary intention. Some enrichments, of course, may be more “natural” than others, due perhaps to the way in which an object is typically used. For example, if someone says to me that a person needs to use a hammer, I will probably infer that she needs to use it in the way that people normally use hammers – to pound in nails. Similarly, if someone tells 171 Bach (1994) calls this completion; Recanati (2004) calls it saturation. For a helpful discussion of pragmatic enrichment in general, and of the difference between required and optional enrichment in particular, see Recanati (in press). 172 For similar discussions of Smith v. United States, see e.g. Neale (2007), 1–10, and Geis (1995), 1136. 183 me that a person used a firearm during a crime, I will take the speaker to be saying that the person used the firearm as a weapon. And although we cannot really say that ‘use’ – unaccompanied by any noun phrase – has a “default” enrichment of this sort, it is still sensible to say that in cases in which a speaker uses a verb phrase of the form ‘[use] an F’, the hearer is justified in taking the speaker to have in mind the normal use of an F (if there is one), unless of course the context provides significant evidence to the contrary. 173 I should note that it may seem reasonable to ask whether we aren’t talking about ambiguity, or even polysemy, rather than pragmatic enrichment. The significance of settling this issue is that if what is going on in Smith has to do with ambiguity or polysemy, then Scalia’s reasoning in the case is simply a matter of disambiguation, in which case Smith does not count as evidence that contemporary textualists are willing to count anything beyond literal meaning as contributing to legal content. That is, if Smith is really a case of ambiguity or polysemy, then the function of contextual inference is – in this case – to give us the relevant literal meaning of the phrase ‘use a firearm’, rather than to provide us with a pragmatic completion of such content. There are two good reasons, however, to believe that the relevant issue in Smith does in fact concern pragmatic enrichment rather than ambiguity or polysemy. 174 The first rests on the fact that semantics is in the business of describing those features of 173 This notion of default enrichment is perhaps sufficiently captured by Grice’s notion of Generalized Conversational Implicature, if we allow the relevant considerations to affect the primary content of the utterance. The idea is that a default enrichment is an enrichment e of the literal content q of a sentence ‘s’ uttered such that e would normally – i.e. in the absence of special circumstances – be associated with utterances of ‘s’. 174 These considerations e.g. motivate Kratzer (1981) to provide a monosemic account of modals. They arguably also motivate Grice (1989) to postulate a unified semantics for ‘or’. 184 expressions that are invariant between contexts of use. And considering the various uses of the verb ‘use’, there really is a feeling of invariance present. This invariance can be exemplified for example by paraphrasing simple sentences containing the verb so that the purpose parameter is made explicit: ‘He used a hammer’ might e.g. be paraphrased as ‘He used a hammer to pound in the nails’. But if the (here, stereotypical) “sense” of the verb has been given by the infinitive clause ‘to pound in the nails’, then – since the sentence still contains the verb ‘use’ – the meaning of ‘use’ must be neutral between various types of purposes. Or else what is said by the infinitive clause would be redundant – and it clearly is not. This strongly suggests that the verb ‘use’ is not ambiguous; rather, it is “skeletal” and requires a piece of information to be provided by the context of use, information comparable to the infinitive clause. The second reason for thinking that we are dealing with pragmatic enrichment rather than polysemy comes from the fact that we don’t want to postulate indefinite polysemy. Given the uncountable variety of uses to which we can put a verb like ‘use’, a proper linguistic account of the expression would require an indefinite number of lexical entries, plus an explanation of the relation between all of them. Further, since we have already established that ‘use’ has a neutral sense, a lexical entry would also be required for that. But then the other (indefinitely many) entries become redundant. A single lexical entry for the neutral ‘use’ provides a much more plausible semantics. 175 175 The general principle underlying the second part of my argument is known as the Modified Occam’s Razor: Senses are not to be multiplied beyond necessity. See Grice (1989), 47. 185 Getting back to textualism, it seems that Scalia’s reasoning in Smith allows us to conclude that contemporary textualists think that required pragmatic enrichment can form part the content of the law. But there also seems to be clear evidence that they want to allow for (at least) the possibility that legal content includes so-called optional enrichment, i.e. contextual enrichment of literal content that does constitute a complete proposition. 176 Implicit quantifier domain restriction is a good example of such enrichment: a speaker may e.g. utter the sentence ‘Everyone is coming to the party’ – which expresses a complete proposition – intending to communicate that everyone in the department is coming to the party. In chapters three and four, I discussed several legal cases involving optional enrichment, including Ali v. Federal Bureau of Prisons, Small v. United States, and United States v. Gonzales (all of which concerned quantifier restriction). 177 For our purposes here, what is important is that in those cases Scalia did not argue that it was irrelevant or inappropriate to ask whether or not particular restrictions were intended on certain quantifiers contained in the relevant statutes. He just happened to think that there was insufficient evidence of such restrictions (and also that precedent weighed against such readings). There are also cases, however, in which Scalia seems to think that there is sufficient evidence that the legislature intended to enact something other than the literal content of the statutory text, even if that content is not necessarily incomplete. In his 176 Bach (1994) calls this expansion; Recanati (2004) calls it strengthening; others have called it free enrichment. 177 552 U.S. 214 (2008), 544 U.S. 385 (2005), and 520 U.S. 1 (1997), respectively. 186 concurring opinion in Green v. Bock Laundry Machine Co., 178 for example, Scalia argued that Rule 609(a)(1) of the Federal Rules of Evidence just could not – on pain of irrationality – be construed literally and therefore had to receive a non-literal interpretation. Bock Laundry was a civil product liability case in which petitioner Paul Green sued Bock Laundry Machine Company (hereafter ‘Bock’) after having been injured by one of its machines while he was on work-release from a county prison. He testified that he had received inadequate information about the machine’s operation and risks of use. Bock, however, impeached Green’s testimony on the basis that he was a convicted felon. Green had filed a pretrial motion for the exclusion of impeaching evidence, which was denied, and, on appeal, Green argued that the trial court had erred in denying his pretrial motion, citing Rule 609(a) of the Federal Rules of Evidence, which – at the time – stated the following: For the purpose of attacking the credibility of a witness, evidence that the witness has been convicted of a crime shall be admitted if elicited from the witness or established by public record during cross-examination but only if the crime (1) was punishable by death or imprisonment in excess of one year under the law under which the witness was convicted, and the court determines that the probative value of admitting this evidence outweighs its prejudicial effect to the defendant, or (2) involved dishonesty or false statement, regardless of the punishment. 179 178 490 U.S. 504 (1989). 179 Federal Rule of Evidence 609(a); my emphasis. The rule has since been amended to address the issue that arose in Bock Laundry. 187 Green claimed that the probative value of admitting the testimony of his prior convictions did not outweigh its prejudicial effect. The problem with Green’s argument was that – read literally – Rule 609 does not apply to him, since he was a plaintiff and not a defendant. So the question was whether or not it was reasonable to read the rule non- literally, so as to refer not just to civil and criminal defendants but also to civil plaintiffs. If so, then the trial court did in fact err in allowing testimony regarding Green’s prior convictions. The Supreme Court majority – led by Justice Stevens – held that Rule 609 could indeed not be interpreted literally. The reason was that on a literal interpretation, the rule would establish an odd asymmetry between the rights of the disputing parties in civil cases: the defendant, but not the plaintiff, would have special protection from impeaching evidence. But neither party in a civil case, the Court said, enjoys any protection over the other party and so the rule “can’t mean what it says.” Given what can be taken for granted in the relevant legislative context regarding “law’s preferences” about the relative status of parties to a civil case, the attitude associated with the literal reading was – to use Scalia’s phrase – an “unthinkable disposition.” 180 We can say that, given the common ground between lawmakers and the courts (concerning the proper way in which to treat parties to a civil case), it would be unreasonable for the court to attribute to the legislature the intention to establish special protection against impeaching evidence for civil defendants but not for plaintiffs. Assuming that the legislature did not just make a 180 490 U.S. 504, 527B (1989). 188 mistake, it must therefore have intended to communicate something other than what it literally said. Or so the pragmatic version of Scalia’s reasoning goes. Now, if this is all correct, what has been shown is just that it is very unlikely that the legislature intended to enact the literal content of the legislative text. It remains to be seen what inference – if any – can be legitimately drawn about what content the legislature did intend to enact. In order to determine the proper non-literal content of Rule 609, the majority turned to a close examination of legislative history, concluding that it did not provide sufficient evidence that the lawmakers intended the protection provided by rule to extend to parties other than criminal defendants. The Court therefore held that Rule 609 did not apply to Green, thereby affirming the Court of Appeal’s decision that the trial court had not erred in admitting evidence regarding Green’s prior convictions. Scalia, in his concurring opinion, agreed with the majority’s conclusion and with certain aspects of its reasoning. He disagreed, however, with the majority’s use of legislative history in determining the non-literal content of the rule, arguing that there was only one relevant non-literal interpretation that the language of the rule could bear; the term ‘defendant’ as it occurs in Rule 609, he said, was obviously being used to mean criminal defendant. As Scalia rightly notes, it makes sense to say that speakers sometimes use ‘defendant’ to mean criminal defendant, but makes little sense to say that speakers sometimes use ‘defendant’ to mean civil plaintiff, civil defendant, prosecutor, and criminal defendant or civil plaintiff and defendant and criminal defendant. To quote his dissent in Johnson v. United States, a word or phrase will bear a suggested meaning 189 only if “you could use the word in that sense at a cocktail party without having people look at you funny.” 181 Scalia’s reasoning in Bock Laundry, then, is easy enough to model on the pragmatic framework underlying my argument for the textualist thesis. And note that it does not really make a difference whether the word ‘defendant’ was being used intentionally to mean criminal defendant or whether it was a “scrivener’s error” (an inadvertent omission of the word ‘criminal’). As a general matter, there is often no way to tell when it comes to the suspected omission of words and phrases – due simply to the way in which we use language. But the pragmatic reasoning is the same – in both cases, the primary question is whether or not there is sufficient evidence that the speaker intended to communicate something other than what she literally said. Based on Scalia’s opinions, then, it is reasonable to think that, according to the contemporary textualist, both types of pragmatic enrichment can in principle form part of the content of the law, although actual cases of optional enrichment will perhaps be rare. This is in line with my own conclusions in the previous two chapters, although I am inclined to think that the content of the statutes in Ali, Small, and Gonzales is indeterminate between their literal content and some relevant enrichment thereof. Scalia, on the other hand, thinks that in each of these cases, the legal content of the statute just is its literal content. 181 529 U.S. 694, 718 (2000). 190 4. Contemporary textualism and the problem of legislative context From what I have said so far, it seems that contemporary textualists have a pretty plausible view regarding the legal content of statutes. As many have pointed out, 182 however, the textualist’s insistence on the legal relevance of pragmatic inference about utterance content has potential to get her into trouble. One reason is that the context and conversational background presumably matter to utterance interpretation only insofar as these provide evidence of the speaker’s communicative intention. But, as I explained above, textualists think that very few intentions can in good faith be attributed to the legislature as a collective. So although textualists don’t hesitate to say that the legal content of a statute is determined by its meaning and that “meaning depends on the way speakers use words in context,” 183 the fact that the speaker’s illocutionary intentions determine how she is using the relevant words seems to present an obvious problem: it follows that the speaker’s illocutionary intentions determine legal content – a very bad result for the textualist. The standard textualist response to this problem, embraced in one form or another by the main champions of contemporary textualism (Scalia, Easterbrook, and Manning), is to say that the meaning of an utterance is determined not by how the words are actually being used (by the speaker) but rather – as we have seen – by the inferences that a rational hearer, knowing the context and conversational background, would be warranted 182 See e.g. Soames (2011) and Marmor (forthcoming). 183 Manning (2003), 2461; Scalia (1997), 23–25, 37, 135, 144. 191 in making regarding the way in which the relevant words are being used. If the objective theory of communicative content is correct, this makes good sense, philosophically. It seems, then, that – much of their rhetoric notwithstanding – textualists are perhaps best construed to be claiming not that intentions are irrelevant tout court but that the only intentions that judges should be concerned with are intentions of the illocutionary kind; and – further – that they should be concerned not with the actual illocutionary intentions of the lawmakers but rather with those illocutionary intentions that a rational hearer would, knowing the context and conversational background, be warranted in ascribing to the legislature, vis-à-vis the relevant enactment. This position, however, seems to get the textualist into at least one of two problems, depending on what – and how much – a person knows when she knows the context and conversational background of an enactment. According to Easterbrook (1990), for example, “[t]extualists … want to know [the] context [of statutory language], including assumptions shared by the speakers and the intended audience.” 184 And, he continues, “it may be essential to mine the context of the utterance out of the debates [recorded in legislative history],” which “may be useful in showing the scope of an enactment.” 185 On Easterbrook’s view, then, the common ground between the legislature and its intended audience may include information found in legislative history. In particular, information about bargains made by the lawmakers is 184 Easterbrook (1990), 443. 185 Easterbrook (1990), 443. 192 supposed to be able to shed light on the communicative content of the resulting enactment. 186 I think, however, that it would be a significant stretch to say that the information contained in legislative history generally forms part of the actual conversational background of enactments, since such information is – as a matter of fact – not generally “shared by the speakers and the intended audience,” especially when it comes to law addressed to ordinary citizens. But more importantly, perhaps, if information about actual bargains made in the legislative process is supposed to be able to shed light on the communicative content of legislative utterances, then Easterbrook must after all care about the actual illocutionary intentions of lawmakers. But this seems inconsistent with the critical aspect of textualism outlined in the first section of this chapter. If, on the other hand, legislative history does not form part of the conversational background – as Scalia e.g. appears to think – then the common ground between the legislature and its intended audiences seems to become quite thin, although its richness may of course vary significantly depending on the audience. As a result, the content that a rational hearer, knowing the context and conversational background, would associate with a particular enactment will presumably rarely go beyond its literal content. This, as we saw, was indeed my own conclusion from chapters three and four. But this seems to significantly undermine the contemporary textualist’s proud rejection of literalism. In addition, as Marmor (2008, 2011a) notes, other aspects of the conversational context are often fairly indeterminate. There seems to be significant indeterminacy, for example, 186 Easterbrook (1990), 443–5. 193 about what members of Congress are actually parties to the relevant conversation (bill initiators/sponsors, indifferent supporters, opponents, everyone?). 187 In general, Marmor thinks, “[t]he conversational situation of an act of legislation is often much more open- ended [than in ordinary conversation], without clear boundaries.” 188 Manning (2001, 2003), however, believes that worries like these should not be exaggerated; due attention to the contemporary textualist’s notion of conversational context, he thinks, shows that even if legislative history does not form part of the common ground between the legislature and its intended audience, the context is often rich – and determinate – enough to facilitate pragmatic enrichment. Two aspects of the common ground make the difference, he says. First, a nuanced competence with the English language is assumed (or whatever language is relevant). Such competence e.g. includes knowledge of how words are normally used; as Manning notes, the term ‘vehicle’, for example, is normally used to refer to (something like) “[t]hat in or on which a person or thing is or may be carried,” although it can also be used to refer to “[a]n art form or device used to convey an effect.” 189 Similarly, people normally use phrases of the form ‘[use] an F’ to communicate that the F is/was/will be used for its intended purpose (if it has one); for example, people normally use the phrase ‘[use] a gun’ to communicate that a gun is/was/will be used as a weapon. In addition, Manning says, if the intended audience is a specialized subcommunity – such as lawyers, scientists, or people working 187 See Marmor (2008), 434–435 , and (2011a), 98. 188 Marmor (2011a), 435. 189 Manning (2003), 2458, citing Webster’s New Collegiate Dictionary 953 (1953). 194 in the financial or commercial industry – competence with the technical terms particular to those communities is assumed. Second, the common ground includes firmly established common law doctrines – that is, the audience is supposed to be “conversant with [substantive] legal conventions.” 190 Criminal statutes, for example, that, per their language, do not contain any mens rea requirements are nevertheless taken to contain implicit requirements (as part of their communicative content). And according to Easterbrook (1999), “the legal system’s accepted procedures, evidentiary rules, burdens of persuasion – and defenses” also form part of the common ground. 191 And given the “abundance of such conventions in a legal system as old as [the US system],” Manning says, “textualists have at their disposal a collection of well-settled background norms.” 192 This rich collection of norms, then, forms part of the “background … against which Congress presumptively legislates.” 193 What this means for the textualist is that established legal conventions do not kick in after the communicative content of the law has been determined, but can actually affect the communicative content of relevant legislative utterances, for example by adding the equivalent of an implicit unless-clause to the literal content of a statute. 194 190 Manning (2003), 2467. 191 Easterbrook (1999), 1913. 192 Manning (2003), 2468; my emphasis. 193 Manning (2003), 2468. 194 For a view similar to this, but expressed in more nuanced linguistic terms, see Goldsworthy (2010), 238–243. 195 Now, I have no problem with the first element of the supposed common ground. Presumably, for example, any competent speaker will have good intuitions about the normal uses of words (assuming they have one). And if such uses are default enrichments rather than just relatively frequently occurring “senses” of certain ambiguous or polysemic terms, then the context may indeed sometimes be rich enough to facilitate pragmatic enrichment of the required kind. I should also note that although Manning does not discuss the so-called linguistic canons of construction, 195 it is not implausible to assume that these principles attempt to make explicit certain general pragmatic rules of thumb, competence with which can be ascribed to any sufficiently skilled language user. I think, however, that the second element is more problematic. A great deal of law – criminal law in particular – is arguably addressed to ordinary citizens, who are by no means “conversant with legal conventions.” As a result, knowledge of substantive legal doctrines on behalf of the hearer cannot always be assumed. This variation between contexts regarding what can reasonably be taken for granted has significant consequences for the contemporary textualist’s position. Consider for example the majority’s reasoning in Staples v. United States (Scalia joining). 196 In Staples, the question was whether or not Mr. Staples had – by possessing an unregistered machinegun – violated The National Firearms Act, 26 U.S.C. § 5861(d), which states that “It shall be unlawful for any person … to receive or possess a firearm which is not registered to him in the National Firearms Registration and Transfer Record.” It was an undisputed fact that Mr. Staples did not 195 See e.g. Scalia (1997), 25–29; see also Sunstein (1989), 454–460. 196 522 U.S. 398 (1998). 196 know that the firearm he possessed required registration, due to the fact that the weapon was a semiautomatic model inconspicuously modified to operate as an automatic one. Now, the literal content of the Act does not stipulate any requirements regarding the epistemic state of the “possessor”, and the district court therefore concluded that the Act did cover Staples’s situation and he was sentenced to probation and a fine; the court of appeals affirmed. The Supreme Court majority, however, argued that the relevant part of the Act had to be construed “in light of the background rules of the common law … in which the requirement of some mens rea for a crime is firmly embedded.” 197 For the textualists on the court (Justices Thomas and Scalia), this meant that the communicative content of the legislative utterance was something roughly equivalent to the following: It shall be unlawful for any person … to receive or possess a firearm which is not registered to him in the National Firearms Registration and Transfer Record, unless the person does not know that the object in question belongs to a type of firearm that requires such registration. In other words, due crucially to the background assumptions shared by the legislature and its intended audience, the actual communicative content of § 5861(d) is a pragmatic enrichment of its literal content. Or so contemporary textualists maintain. The problem with this line of reasoning, however, is that if ordinary citizens are generally the intended audience of criminal law, then the doctrine of mens rea cannot reasonably be taken to form part of the actual common ground and, consequently, there is no linguistic basis for claiming that the communicative content of § 5861(d) is a corresponding pragmatic enrichment of its literal content. This means that if the textualist 197 511 U.S. 600, 605 (1994). 197 wants to avoid a conviction for Mr. Staples without stepping outside the boundaries of the law, then she will have to concede that the mens rea doctrine is a source of law in virtue of its continued and widespread application by the courts – something that textualists do not particularly like to do, evidenced by their reference to common law doctrines as “judicial power-grabs.” 198 What Staples and similar cases indicate, I think, is that textualists are – despite their rhetoric – not simply concerned with the actual communicative content of legislative utterances. If they were, they would be more sensitive to the significant variation in the common ground between the legislature and its various audiences. To be sure, the textualist framework is heavily constrained by highly plausible ideas about language and communication, but to some extent their notion of context and conversational background appears to be designed or constructed in order to obtain certain legal results and block others. I suspect that one reason for this stems from the textualist’s desire to ultimately ground the authority of established legal doctrine in the authority of Congress. If established common law doctrines form part of the background of legislative utterances, and thereby affect their communicative content, then such utterances arguably count as implicit recognition of the relevant doctrines; as a result, judges still act as faithful agents of Congress when they recognize established doctrine, despite the fact that doctrine originates with the courts. As we have seen, this may well be the case when the relevant audience is in fact “conversant with legal conventions” – such as in the case of law addressed to courts or 198 Scalia (1997), 29. 198 administrative agencies. But there will also be a significant set of cases in which the textualist – on pain of having to permit a severely problematic result – will have to recognize that the authority of established legal doctrine is grounded in its having being established by judicial practice. In those cases, doctrine does not affect the communicative content of the law, but rather kicks in after that content has been established. Staples, I think, was an example of such a case. In the next section, I want to put these worries aside and examine a little further just how far the pragmatic framework underlying the current version of the textualist thesis can take us, vis-à-vis legal content. It is of course interesting in its own right to figure out some of the limits of the framework, but I think the following discussion sheds at least a little bit of light on two important things. First, I think that the framework allows us to explain how it is possible to make a principled distinction between the different textualist reasonings in Bock Laundry and US v. Marshall, both of which can be said to be cases in which the problem of unintended consequences arises. This would allow textualists to respond to the charge that they are inconsistent in their willingness to go beyond literal content in adjudication. Second, I think that it can also help to shed some light on the debate about the plausibility of what John Perry has recently dubbed conception- textualism. If what I say is correct, then this view is – contra Perry – not necessarily “confused, implausible, and unworkable.” 199 199 Perry (2011), 106. 199 5. Pragmatic inference and unforeseen consequences: US v. Marshall Cases like Bock Laundry – whether they are cases of pragmatic enrichment or misexpression (i.e. intention or error) – are often referred to as cases of unintended consequences. This seems appropriate enough to me. They differ significantly, however, from another class of cases that are also commonly categorized as cases of unintended consequences – cases in which the legislature simply did not foresee a certain problematic result. In such cases, I will argue, the pragmatic framework cannot be used to ground a non-literal interpretation of the relevant statute. Consider the case of United States v. Marshall. 200 In Marshall, the question was whether or not Mr. Marshall, who had sold around 12,000 doses of LSD, should be taken to fall under a statute triggering a mandatory minimum punishment if the sold items contain “a mixture or substance containing a detectable amount” of the relevant drug. 201 The issue was that LSD requires a particularly heavy carrier, so the ratio of drug to carrier is extremely low. On the basis of a literal reading of the statute, Marshall was sentenced to a long time in prison for selling what constituted relatively few doses. In order, for example, to receive a similar sentence in the case of heroin, Marshall would have had to sell around 10 kilograms of it. Thus, it makes sense to say that, on a literal reading, the statute has the unintended consequence that it obligates judges to impose disproportionate sentences on people who sell LSD, due to the fact that the lawmakers did not realize how the drug is sold. 200 908 F.2d 1312 (7 th Cir. 1990). 201 21 U.S.C. §960. 200 For our purposes, the relevant question is whether or not it would make sense to apply the same sort of pragmatic reasoning in Marshall as in Bock Laundry. The upshot of being able to do so would be that pragmatic textualism could mitigate a pretty wide range of problematic results. It would be a significant cost of such success, however, that the actual application of textualism would appear to be somewhat inconsistent or opportunistic, since the majority in Marshall, headed by textualist champion Judge Easterbrook, in fact held that despite the problematic result, the content of the relevant statute was in fact its literal content, and that it was not the Court’s job to rectify legislative mistakes. The framework underlying the textualist thesis, however, does not predict that the statutory content relevant in Marshall was something other than the literal content of the statute. Moreover, the framework allows us to explain in a principled way the main difference between the two cases. As it turns out, pragmatic reasoning can be applied only to a subset of cases in which the problem of unintended consequences arises, which undercuts the common objection that Bock Laundry and Marshall show that – as applied – textualism is inconsistent, opportunistic, and unprincipled. The relevant difference between Bock Laundry and Marshall is that the former is about warranted inference regarding communicative content, while the latter is not. In Bock Laundry, the unintended consequence appropriately prompted Scalia to infer that the communicative content of the legislative utterance was something other than its literal content, due in large part to the fact that it was unreasonable to think that the lawmakers intended to enact the literal content and simply did not foresee the relevant consequence. 201 In Marshall, however, the unintended consequence has no such role to play in reasoning about utterance content – and hence legal content – because the most likely explanation of the problematic result is that it was simply not foreseen. Thus, the main question in Marshall is about how to appropriately react to a problematic result that follows from the content of the legislative utterance. To fix these ideas better, let us take another example. Say that the legislature enacts a statute that reads “No dogs on public buses”. The literal content of the statute, then, applies to guide dogs. But if it is – given the context and conversational background – sufficiently likely that (i) the legislature foresaw this result, (ii) the legislature considers this a bad result, and (iii) the legislature does not have a competing preference the satisfaction of which outweighs the disutility of this result, then it is sufficiently likely that the legislature intended the law not to cover that subset of dogs. If so, then it is correct to say that the communicative content of the statute does not say anything about guide dogs, and, hence, that the content of the law does not entail a prohibition to bring guide dogs on public buses. If, however, any of the three relevant conditions is sufficiently lacking in probability, then there will be insufficient reason to believe that the legislature intended not to cover guide dogs with its enactment. Consequently, guide dogs would be prohibited from public buses. It should be fairly clear, then, how textualists can use the pragmatic framework to explain in a principled way the difference between Scalia’s reasoning in Bock Laundry and Easterbrook’s reasoning in Marshall. In Bock Laundry, the set of conditions mentioned above is satisfied, whereas in Marshall it is not. And although I do not 202 endorse the majority’s decision to refrain from changing the content of the law in Marshall, it is quite coherent for the textualist to argue that despite the fact that the result is both unintended and bad, the legal content of the relevant statute does not exceed the limits of Congress’s constitutional authority (and so should be left unchanged). 6. On the plausibility of “conception-textualism” Perry (2011) distinguishes between two types of textualism that can be extracted from Scalia (1997): meaning-textualism and conception-textualism. Meaning-textualism, says Perry, is the view that “the content of a statute is determined by the words in the text of the statute, given the meaning that those words had at the time of enactment or ratification.” 202 Conception-textualism, on the other hand, is the view that “the conceptions that the enactors had of the states, conditions, phenomena, and the like referred to by their words, used in the operative senses, are determinative [of statutory content].” 203 According to Perry, what makes it the case that meaning-textualism is a sensible philosophy of legal interpretation while conception-textualism is “totally implausible” is the fact that the former – but not the latter – applies to statutes “the same apparatus we use to determine what some individual says when they are talking to us.” 204 My aim in this section is to show that it is far from obvious that this is true. In at least some cases of linguistic communication, it is not implausible to argue that a speaker’s 202 Perry (2011), 106. 203 Perry (2011), 106. 204 Perry (2011), 106. 203 (mutually known) beliefs about that to which she wishes to refer, or quantify over, manage to affect utterance content. 205 Recall the two types of cases I have just discussed – exemplified by Bock Laundry and Marshall. It is my contention that somewhere in between those types of cases, there exists an interesting type of intermediate cases – one that puts significant pressure on Perry’s assessment of conception-textualism. These are cases in which (i) the lawmakers enact a statute the literal content of which covers – among other things – a certain class of “objects” (broadly construed), (ii) the lawmakers don’t have a positive intention that the statute not cover the relevant class, 206 but (iii) the lawmakers clearly believe that the relevant predicate of the statutory text does not apply to the class. The question is whether or not – in such cases – the resulting law covers the relevant class of objects? Unlike Perry, I think it is far from obvious what the answer to this question is. Say, for example, that Congress enacts a statute according to which “All Fs ought to φ” and that Congress also falsely believes that ‘F’ does not apply to a certain subset X of objects. Accordingly, in uttering the sentence, Congress neither intends ‘F’ to cover X nor intends ‘F’ to not cover X. Given its false belief, it simply has no intention regarding the relationship between ‘F’ and X, vis-à-vis this particular utterance. But if Congress’s belief that the members of X are not within the extension of the predicate ‘F’ is part of the conversational background, it becomes difficult to say whether or not this can affect the communicative content of the relevant speech act. 205 For a similar argument, see Ostertag (unpublished). 206 That is, the set of conditions mentioned in the previous section is not satisfied. 204 Consider, for example, the case of FDA v. Brown & Williamson Tobacco Corp., 207 in which the central question was whether or not the Food, Drug, and Cosmetic Act (FDCA) gave the FDA authority to regulate tobacco products. 208 In the first instance, the FDCA gives the FDA authority to regulate drugs; and ‘drugs’, according to the Act, means – among other things – “articles (other than food) intended to affect the structure or any function of the body of man.” 209 So the literal meaning of the Act does cover nicotine. But it is not implausible to claim that the legislature lacked a positive intention to not cover nicotine with its particular use of the term ‘drugs’, 210 because the lawmakers believed that nicotine was not in the extension of that term, as defined in the Act. The question, then, is whether or not this belief affects the content of the legislature’s utterance in such a way that the FDCA does not count as giving the FDA the authority to regulate tobacco. As we will see, I don’t have a firm opinion either way, but I think that we should at least not be too hasty to rule out the possibility that it does affect utterance content. Generally speaking, my intuition is that while it may be implausible to think that a speaker’s false belief about the extension of a predicate affects the content of speech acts with a mind-to-world fit (such as assertions), it is not so implausible to think that they 207 529 U.S. 120 (2000). 208 21 U.S.C. §301-399(a). 209 21 U.S.C. §321. 210 That is, it is not the case that the lawmakers thought that the term ‘drug’ actually covered nicotine and that they just intended to implicitly exclude nicotine from the relevant domain of quantification. 205 sometimes play a part in determining the content of speech acts with a world-to-mind fit (such as directives). What seems to make the difference is that a speaker’s beliefs may well have special relevance when it comes to interpreting utterances in which the speaker represents the world not as she thinks it is but rather as she would like it to be. As I mentioned above, one reason I’m interested in these kinds of cases is that Perry thinks that it is a “rather bizarre and hopeless” idea that the conception that a speaker has about that to which he wishes to refer, or quantify over, can affect the primary content of the speech act. Contra Perry, I think it is possible to make a decent case for this position, at least in the case of speech acts with a world-to-mind fit – including enactments. First, consider the difference between assertions and directives, with respect to intuitions about how information about speaker beliefs may affect what gets communicated. Say that S utters the following: ‘All Fs are G.’ Say also that it is part of the common ground that S believes that a class of objects, C, is not within the extension of the predicate ‘F’. Finally, say that C is in fact within the extension of ‘F’. The question, then, is this: In uttering ‘All Fs are G.’, did S predicate G-hood of the objects in C? The answer is not entirely clear, but it seems more reasonable to say that the erroneous belief does not affect the communicative content in such cases. Say for example that someone utters the following: ‘No mammal weighs more than 15,000 pounds.’ And that it is part of the common ground that the speaker believes that whales are not mammals. Does it follow from the communicative content of the speaker’s utterance that she attributed the property of not-weighing-more-than-15,000- pounds to whales (among other things)? My intuition is that it does, since I am inclined to 206 judge the speaker to have asserted something false. And if that is correct, then the speaker’s false belief did not affect what was communicated. Assuming that the scenario is unexceptional, it is reasonable to generalize from the case and say that in the case of assertion, it does not make a difference to the communicative content of a speaker’s utterance of ‘…F…’ that the speaker falsely believes of a class of objects that they are, or are not, within the extension of ‘F’. And I think it is not unreasonable to explain the intuition in part by reference to the fact that the speaker misrepresented the world. If that is correct, then I would like to generalize further and claim that in the case of speech acts with mind-to-world fit, false beliefs that a speaker has about that to which she wishes to refer, or quantify over, do not affect utterance content. It is important to emphasize that this example is different from examples in which there is a clear intention to restrict the domain of the quantifier ‘All Fs’. In such cases, it is part of the communicative intention to not say anything about the relevant set of objects. In the cases under discussion here, there is no positive intention to restrict the quantifier, but there is a clear (false) belief that the predicate doesn’t cover the relevant class of objects. Next, consider a case in which a speaker utters these words: ‘Get me all the Fs!’ Say also that it is part of the common ground that S believes that a class of objects, C, is not within the extension of the ‘F’. Finally, say that C is in fact within the extension of ‘F’. The question, as before, is this: In uttering ‘Get me all the Fs!’, did S direct the addressee to get her (among other things) the objects in C? Again, the answer is not 207 entirely clear, but it doesn’t seem totally unreasonable to say that the erroneous belief manages to affect the communicative content. Say for example that a boss utters the following to her assistant: ‘Get me all the ashtrays you can find in the building!’ Now, one issue concerns what further restrictions – if any – the speaker intended on the quantifier ‘all the ashtrays in the building’. 211 This is a relevant issue, but it is not the one I am interested in here. The issue I’m concerned with is whether it would affect the communicative content of the speaker’s utterance if it were part of the common ground between speaker and hearer, for example, that the speaker believes that the ashtrays that were ordered only last week are not in the building, when in fact they have just arrived. Here, I am not so inclined to say that it follows from the communicative content of the boss’s utterance that the new ashtrays belong to the set of things that the assistant was told to collect. However, I do not have a robust intuition that they don’t belong to that set either. Nevertheless, what this indicates is that – contra Perry – it is not entirely implausible to think that the conception that a speaker has about that to which she wishes to refer, or quantify over, can affect the primary content of the speech act – at least not in the case of directives. And since it is not unreasonable to think that this judgment is explained in part by reference to the fact that the speaker’s utterance represented the world as she would (in some sense) like it to be, I think it is sensible to suggest that in the case of speech acts with world-to-mind fit, false beliefs that a speaker has about that to which she wishes to refer, or quantify over, can affect utterance content. 211 For a discussion, see e.g. MacCallum (1966) and Manning (2003). 208 Coming back to FDA v. Brown & Williamson, it seems, then, that it is not out of the question to argue that the communicative content of the FDCA does not entail that the FDA has authority to regulate nicotine, despite the fact that nicotine is in the extension of the term ‘drug’ as it is defined in the Act itself. This cautious conclusion also has some significant counterparts in the domain of constitutional interpretation. On one reading of Scalia’s argument about the legal content of the Eighth Amendment, for example, he can be taken to claim that – indeed – the conception that the framers had about the extension of the term ‘cruel and unusual punishment’ somehow affects the content of the Amendment. 212 In particular, Scalia thinks that the death penalty – which may in fact be within the extension of the term ‘cruel and unusual punishment’ – is not among the punishments that are prohibited by the Amendment, because it is clear that the framers believed that it was not in the extension of that term. As I mentioned earlier, Perry thinks that conception-textualism has no linguistic basis at all. But if what I have said above is correct, then we should not be so hasty to rule it out. I should note that in his discussion, Perry does not really distinguish between statutory interpretation and constitutional interpretation. It is important to note, however, that Scalia actually endorses meaning-textualism for statutes but conception-textualism for constitutions. And although it may be possible to show that conception-textualism is defensible from a linguistic perspective, Scalia (1997) provides – in his reply to Dworkin – the ingredients to a non-linguistic argument for the claim that the framers’ beliefs about those things to which they intended to refer, or quantify over, affect the content of the 212 Scalia (1997), 144–149. 209 Constitution. According to Scalia, it is among the central functions of constitutions that they “freeze” contemporary conceptions of morality; this is one of the things that make a constitution what it is. If Scalia is right, then he may have a non-linguistic explanation of why the content of, say, the Eighth Amendment is affected by the beliefs that the framers had about cruel and unusual punishment. It’s their beliefs about right and wrong that the Constitution – in virtue of being a constitution – “freezes” (insofar as these beliefs can be taken to have been representative of popular morality). One of the distinct benefits that this functional argument for the legal relevance of the framers’ mental states has over the linguistic argument for conception-textualism is that – if it works out – its conclusion appears to be decisive. Another benefit is that Perry’s linguistic worries about conception-textualism would have little or no bite. In contrast, my above linguistic defense of conception-textualism guarantees only that the view should not be immediately ruled out as implausible. One of the major advantages of the linguistic argument, however, is that it does not rely on any controversial claims about the constitutive functions of constitutions. As a final note, it is worth mentioning that although conception-textualism may ultimately be defensible from a linguistic point of view, it may still be morally problematic to defer to the framers’ beliefs when decisions are made in cases concerning constitutional matters. 213 That is, even if conception-textualism were to provide us with the content of the Constitution, it is far from clear that constitutional courts ought not to change that content under certain circumstances. Here, however, is not the place to 213 For a related discussion, see e.g. Marmor (2005), 157, and Goldsworthy (2003), 177. 210 discuss the authority of constitutional courts to change the content of constitutions. But I do take what I have said about the linguistic basis of conception-textualism to suggest that if there is in fact a problem with this view, then it is primarily a moral one, rather than a linguistic one. This concludes my discussion of textualism and legal content. Before I go on to the final chapter of this dissertation and discuss an intentionalist model for resolving borderline cases, I want to briefly explain my reasons for rejecting the textualist approach to such cases. 7. Textualism and legal interpretation At the level of interpretation, textualism tells us that judges should play a very limited role. In case the communicative content of a statute yields a result that is undesirable without being unconstitutional, the court is obligated to follow the law in its decision. Marshall – as we saw – illustrated this line of thought well. And in case the content of the law is indeterminate with respect to the case before the court, the court is in the first instance bound by precedent. Have similar enough cases been settled in the past? If so, then the court should treat the current case in the same way. If not, the court should decide in favor of the defendant. So, basically, a judicial decision against a (criminal or civil) defendant is legitimate only if the case falls determinately under the relevant statute 211 or if there is precedent for applying the statute to sufficiently similar cases. 214 In all other cases, the court’s decision should be in favor of the defendant. It is beyond the scope of this dissertation to discern under what conditions courts should or should not follow the law, and so I will not venture to critically analyze the textualist position on the authority of judges in cases in which the content of the relevant statute yields a result that is undesirable but not unconstitutional, or in which clear precedent does. Perhaps the standard textualist considerations – democratic legitimacy, legislative supremacy, separation of powers, and the Rule of law – support their position, and perhaps not. Instead, I will focus on the question whether or not these considerations support the textualist’s view that the role of the judiciary should also be so limited when it comes to cases of vagueness. The answer to this question, I think, is no. Due to the nature of borderline cases, none of the traditional textualist considerations suffice to support a universal rule to favor the defendant in borderline cases. The short answer is that these considerations count against judicial lawmaking only insofar as the courts have a choice between following the law and changing the law (as the court did in Marshall, for example). Borderline cases, however, necessitate a change in the relevant law – the courts usually have no choice but to precisify the law, and in doing so they are engaged in a limited form of lawmaking. But if there is no choice but to make law, then the considerations that otherwise count against judicial lawmaking have no bite. Even the Rule of Law virtue of predictability is 214 See e.g. Easterbrook (1983), 549–50. 212 promoted equally well by making it a rule to decide against defendants in borderline cases. Appealing to considerations of fairness (such as fair notice) will not do the trick either, since such appeal presupposes that somehow the defendant’s legitimate expectations would be frustrated by a decision against her. But – as many have pointed out 215 – it does not make much sense to claim that the resolution of a borderline case frustrates legitimate expectations, since it is partly constitutive of such cases that it is impossible to justifiably form the beliefs that are necessary for having such expectations (although I don’t want to exclude the possibility of excusably forming such beliefs). Thus, since it is impossible to have legitimate expectations regarding borderline cases, it is also impossible that legitimate expectations are frustrated in the resolution of such cases. If this is correct, then fairness is not a concern in borderline cases and cannot ground a rule in favor of the defendant (or the plaintiff, for that matter). What seems left for the textualist, then, is to argue that limiting government is independently desirable. In fact, Marmor (2007) argues that this is the real motivation behind contemporary textualism. 216 Of course, if what matters – with respect to borderline cases – is primarily to “move” these cases determinately outside law’s reach, and hence outside the reach of the government, then the intentions of lawmakers don’t matter for resolving such cases. It might therefore seem that the neo-conservative agenda of contemporary textualism can rationalize their view about how to settle borderline cases. 215 See e.g. Hart (1994), 276, and Marmor (2007), 23. 216 See Marmor (2007), chapter 8. 213 Now, although I am not sympathetic to using the judiciary to reign in the government, I will not argue against the claim that limiting government is desirable. Instead, I will suggest that even if it were desirable, respect for legislative bargaining/compromise is more desirable. And if that is correct, then the intentions of lawmakers do in fact matter, vis-à-vis legal interpretation. Sometimes, the intentions of lawmakers ought to dictate which way borderline cases are decided. The reason that respect for legislative bargaining matters more than the judicial limitation of government is that such bargaining is fundamental to our legislative system. If it is not respected, incentive to bargain is lost and legislation is severely impeded. So when it comes to borderline cases, respect for legislative bargaining directs the courts to find out if a relevant bargain was made and if so, what it was. Now, the way in which the courts find this out would presumably be by consulting legislative history, which of course raises the worry that legislative history is an unreliable source for determining the intentions of lawmakers, due to the fact that many recorded statements have been strategically placed in order to sway judges in favor of a desired result. However, while this is indeed a worry worth taking seriously, legislative history is not entirely unreliable, as we will see in the next chapter. In fact, it contains a fair amount of reliable evidence of the intentions of lawmakers relevant to legislative bargaining. And if it can be shown that there exists some efficient method of discerning when statements indicating legislative intentions can be taken to be credible, then there is – at least in borderline cases – no reason for the courts not to consult 214 legislative history in the aim of finding evidence of the lawmakers’ intentions. As I will argue in the next chapter, such a method is indeed available. 215 CHAPTER SIX: RESOLVING CASES OF VAGUENESS In chapter two, I introduced Roy Sorensen’s epistemic account of vagueness and his distinction between absolute and relative borderline cases, and explained his argument that vagueness, understood as the possession of absolute borderline cases, does not have a constructive power-delegating function in law. 217 I summarized his argument in the following way: P1. Delegation of decision-making authority, vis-à-vis borderline cases, is valuable only if the relevant delegates are in a better position to answer the question whether x is F than those delegating the power. P2. When x is an absolute borderline case of ‘F’, it is not the case that the relevant delegates are in a better position to answer the question whether x is F than those delegating the power. C. Therefore, when x is an absolute borderline case of ‘F’, it is not the case that delegation of decision making authority, vis-à-vis x, is valuable. My reply was to argue that P1 is false. Delegation of decision-making authority can be valuable even if the relevant delegates are not in a better position to answer the question whether x is F than those delegating the power. The key to seeing why is to acknowledge that when faced with absolute borderline cases, the courts must engage with a related normative question – whether x ought, relative to the purposes of the law, to count as an F. 217 See Sorensen (2001a). 216 In my discussion of Sorensen’s argument, however, I was not discriminating regarding what sort of legislative purposes can settle the normative question. I assumed, for the sake of simplicity, that it is settled if any background justification of the law dictates a decision either way. I assumed this because there is no need to discriminate between types of legislative rationale in order to reply to Sorensen’s argument. In this current – and final – chapter, however, I want to consider what restrictions, if any, there may be on appealing to background justifications in order to appropriately settle the normative question. To be more specific, I want – in the first instance – to examine some of the conditions under which such deference can reasonably be said to maximize fidelity to law. I think a good case can be made that, insofar as the aim is to maximize fidelity to law, the resources for answering the normative question are – at least normally – restricted to rationale expressed by lawmakers during the legislative process under conditions that deter so-called cheap talk. It is one thing, however, to say that borderline cases can be settled by reference to legislative purposes and quite another to say that they should. I want, therefore, also to consider what reasons judges might have to resolve cases of vagueness on the basis of the relevant statute’s background justification. As we will see, I think that there are at least two types of reasons for judges to defer to the intentions of lawmakers – one general and one dependent on the legal system in place. As Marmor (2005) argues, judges can – irrespective of the legal system in place – have authority-based reasons to defer to legislative intentions, if the legislature can be expected to possess expertise regarding the relevant domain of conduct. In fact, Marmor thinks that expertise-based reasons exhaust 217 the kinds of reasons for which judges ought to defer to legislative intentions. But, as I will argue, judges in legal systems in which the legislative process depends crucially on legislative bargaining – like the US – may also have non-expertise-based reasons for such deference. As I indicated at the end of the previous chapter, I think that when it comes to legal systems that rely on legislative bargaining, one of the strongest reasons in favor of settling the normative question by reference to legislative rationale is that doing so respects such bargaining and thereby helps preserve a fundamental feature of the legislative institution. To sum up, then, there are two primary questions with which I will be concerned in this chapter. First, appeal to what sort of legislative rationale can reasonably be said to maximize fidelity to law, vis-à-vis judicial decision-making in borderline cases? And, second, under what conditions should judges aim to maximize fidelity to law in deciding such cases? I should emphasize that the aim is not to propose a complete decision procedure for resolving vagueness in the law. Far from it. The goal is simply to show that, under a modest set of circumstances, judges have strong reasons to appeal to legislative intentions in settling borderline cases. If that is correct, then a constrained form of intentionalism will sometimes be the best theory of legal interpretation (on the understanding that legal interpretation is the act of developing the law in the face of indeterminate/inconsistent legal content or a particularly problematic result). I should also note that what I say in this chapter applies not only to the resolution of borderlines cases associated with vagueness of content (discussed in chapters one and 218 two), but also to the resolution of borderline cases associated with vagueness about content (discussed primarily in chapter four). 1.1 Expressly offered rationale and the notion of commitment It seems clear enough that legislative purposes can sometimes help settle a verdict in borderline cases. It is less clear, however, what restrictions there are on appeal to legislative purposes if such appeal is to reasonably count as appeal to law’s purposes. I hope to show that insofar as appeal to legislative purposes in borderline cases is to count as maximizing fidelity to law, such appeal is – at least within the US legal system – normally restricted to rationale that certain lawmakers have in their official capacity expressly offered, “on record”. At a very general level, the argument is a relatively simple two-premise argument. The first premise is that what it is for judges to decide a borderline case in accordance with fidelity to law is for them to decide the case in accordance with (the proper balance of) those purposes to which the “legislatively relevant” actors have – in their official capacity – incurred a commitment. The second premise is that the primary way in which such actors incur the relevant kind of commitment is through the express offering of legislative rationale. The process by which many UK laws get enacted naturally suggests this simple picture, for example. In most cases, the government will outline proposals for new laws in so-called White Papers, official documents that – among other things – contain statements regarding the ends that the government hopes to achieve by the relevant piece 219 of legislation. The government itself – rather than members of Parliament – will normally introduce the bills outlined and justified in the White Papers, and on many occasions the bills pass more or less unamended; that is, the legislative process in Parliament is – more or less – a matter of voting ‘yes’ or ‘no’ on a bill introduced by the government. In these cases, it seems to make particularly good sense both to talk about expressly offered rationale (the rationale offered in the government White Papers) and to say that this rationale counts as the law’s purposes, since (in these cases) the government reasonably counts as a legislatively relevant actor. A lot depends, however, on the contingent structure of the lawmaking process in the relevant system. In the US, for example, bills are introduced by members of the legislature, rather than the government, and the legislative process is very rarely just a matter of voting ‘yes’ or ‘no’; conflicts of interest are prominent and legislative bargaining is a fundamental feature of the system. These considerations complicate the picture outlined above a great deal. One reason is that there are usually no official documents that can count as clear equivalents of White Papers, and it will be fairly contingent whether – during the legislative process – lawmakers offer any statements of what they take to be the purpose of the relevant piece of legislation. Consider for example the companion cases of Curran v. Mount Diablo Boy Scouts and Randall v. Orange County Council, 218 in which the question was whether or not the Boy Scouts should count as a business establishment under California’s Unruh Civil 218 17 Cal. 4th 670 (1998) and 17 Cal. 4th 736 (1998), respectively. 220 Rights Act 219 On the reasonable assumption that the Boy Scouts share enough characteristics with paradigm business establishments to count as a borderline case of the term ‘business establishment’, the question is whether the courts could have settled these cases by appealing to the rationale for the Act offered by lawmakers in the legislative process. As it happens, however, there simply is no legislative history for the relevant piece of legislation, except for different versions of the bill that ultimately passed as the Unruh Civil Rights Act. As a result, there were no official statements of legislative purpose to which the courts could have appealed in settling these cases. On the assumption that the remedial purpose of the Act wasn’t sufficiently obvious so as to not need explicit expression (more on such cases below), the court’s decision in Curran and Randall therefore arguably had to be justified by considerations beyond fidelity to law (in my sense of the term). The main complication for the simple picture, however, comes from the prominence of strategic factors in systems of the relevant sort. The reason is that that in order for speakers – in this case lawmakers – to genuinely incur commitments via utterances, a certain amount of cooperation is normally needed. In a game of Bullshit, for example, speakers are not taken to be committed to the truth of the statements that they make, because it is mutually known that the speech context is not a cooperative one; in particular, speakers cannot be relied on to assert a proposition only if they believe it and 219 Cal. Civ. Code §51. 221 have adequate evidence for their belief. Or, for those familiar with Gricean jargon, the strategic nature of the game makes it the case that the Maxim of Quality is suspended. 220 It is interesting – and relevant – to note that the US legal system appears to recognize that strategic factors have this undercutting effect on linguistic commitment. Under federal perjury law, for example, witnesses are not taken to be committed to the implied content of their testimony, due to the adversarial nature of cross-examination. Consider the case of Bronston v. United States, 221 in which the question was whether Mr. Bronston had correctly been found guilty of perjury, in virtue of having provided an “unresponsive” reply to a question that in ordinary conversation would be taken to trigger a so-called relevance implicature (the content of which would be false). Mr. Bronston, who for a five year period had a personal bank account in Switzerland, was asked the following question: ‘Have you ever [had a bank account in Switzerland]?’ His reply was this: ‘The company had an account there for about six months.’ Now, if the purpose of the relevant conversation had been the cooperative exchange of information, Mr. Bronston would have counted as having – via implication –incurred commitment to the false proposition that he himself had not had an account in Switzerland. The Supreme Court, however, unanimously argued that – for the purposes of the federal perjury statute – Mr. Bronston was not committed to this proposition, and that he should therefore not have been convicted of perjury. And although the Court’s decision was not framed in linguistic terms, the decision nevertheless has a good linguistic justification; in the 220 On the suspension of conversational maxims, see e.g. Martinich (1984), 33. 221 409 U.S. 352 (1973). 222 context of cross-examination, the conflict of interest between speaker and hearer makes it the case that speakers cannot be relied on to assert a proposition only if it is relevant. Or, again for those familiar with Gricean jargon, in the context of cross-examination, the Maxim of Relevance is suspended. Assuming otherwise would be a mistake on behalf of the examining lawyer. The relevance of these examples to the discussion here is that when it comes to legal systems in which strategic factors are as prominent as they are in the US and similar systems, it is a legitimate concern whether lawmakers ever incur the right kind of commitment via their statements regarding legislative rationale (assuming they sometimes offer such statements). The concern is that we get information regarding the relevant statements primarily from legislative history, i.e. from bills, committee reports, floor debates etc. As I mentioned earlier, some of the problems concern the availability and/or relevance of legislative documents. Do we have any statements regarding legislative purpose “on record”? Do all official legislative documents count? And do those that count all count equally? However, the most critical issue is due to the strategic nature of lawmaking and concerns the frequent manipulation of official documents by lawmakers in order to increase the likelihood of certain outcomes in high-level court cases. As Rodriguez & Weingast (2003) note, the incentive to influence or manipulate legislative history can be great; ardent supporters of a bill, for example, might be moved to influence legislative history in such a way that future interpretation fails to acknowledge significant compromise made in the legislative process (more on this 223 below). This means that even if we have clear statements regarding legislative purpose, there is a good chance that these statements are not sincere. In addition to all this, it is usually also far from clear – unlike in the standard UK lawmaking case – whose statements are most relevant. That is, in the US and similar systems, it is very often not clear which members of Congress count as the most legislatively relevant actors (in my sense of the term). Taking all these issues into account, there seems, then, to be a significant tension between my claim that borderline cases can be settled by deference to law’s purposes, understood as the rationale to which lawmakers have expressly incurred commitment, and the strategic nature of the legislative environment. It is important to point out, however, that this tension is not a theoretical one; that is, the tension is not evidence of a problem internal to the model that I am arguing for. Rather, the tension concerns the application of the model – it is an indication of a limitation on the number of cases in which it might be possible to resolve vagueness in such a way as to maximize fidelity to law. This sort of limitation, however, should be expected. In fact, as I will argue below, there is actually a further limitation on the applicability of the model, due to the fact that it is not always the case that judges ought to maximize fidelity to law (even if they can). But, as I said at the beginning of the chapter, the aim here is not to propose a complete decision procedure for resolving vagueness in the law. It is simply to show that, under a modest set of circumstances, judges have strong reasons to appeal to legislative intentions in settling borderline cases. 224 Before I go on to talk about this further limitation on (all things considered) appeal to legislative rationale, however, I want to consider a number of ways in which the limitations discussed so far can be mitigated. Interestingly, the most important way in which these limitations can be mitigated turns out to be grounded in the legislative system itself, in that certain parts of the institutional design manage to remedy the incentives that lawmakers have to engage in cheap talk. 1.2 Institutional remedies to non-cooperation One of the most significant limitations of the model I have been discussing is that even if we have statements about legislative rationale on record, legal systems with an adversarial legislative process and widespread conflict of interest among lawmakers typically offer more opportunity and incentive for cheap talk than for sincere utterances. Since insincerity undercuts the relevant kind of commitment, the question is whether or not we can identify any conditions in the legislative process that remedy these incentives. Taking a lead from Boudreau et al. (2007), I think we can, which – if correct – means that lawmakers sometimes do in fact manage incur the right kind of commitment in offering legislative rationale. In addition, the resulting framework – at least in some cases – offers a plausible answer to the question whose utterances count, as far as fidelity to law is concerned. The primary issue here is that since legislators may have significant incentive e.g. to exaggerate or misdescribe the content and/or purposes of a bill, interpreters must be wary of indiscriminate appeal to legislative history. Given the strategic environment in 225 which bills (if lucky) become laws, sincerity is presumably only guaranteed when these incentives are somehow remedied. And they are remedied, say Boudreau et al., when it is sufficiently costly to make a statement or when the product of the penalty of insincerity and the subjective probability of verification is sufficiently great. In general, legislative history may be taken to be reliable, or credible, to the extent that the likelihood of cheap talk is minimized. Based on the empirical work of Lupia & McCubbins (1998), the idea is that, under certain conditions, the non-cooperative nature of the legislative conversation can be remedied by “institutional intervention.” That is, the design of the legislative process (in the US and similar systems) is such that, sometimes, the institutional structure substitutes for common interests, which are assumed in cooperative conversational settings. As I said above, the relevant institutional structures are those that generally make it costly to make statements or, as I will focus on here, make the product of the probability of verification and the penalty of lying sufficiently weighty. 222 Boudreau et al. point out that the legislative process is structured in such a way that it yields certain benefits to the majority party (especially since the late 19 th century) and that it rewards and punishes certain behavior on behalf of officials. 223 These rules, the authors argue, affect the communicative incentives that lawmakers have. In particular, these rules give members of the minority party an incentive, for example, to grandstand and to describe proposed statutes and their rationale in ways that do not reflect their 222 See Lupia & McCubbins (1998), 17–96. 223 Boudreau et al. (2007), 959–60. 226 genuine understanding of it (typically, exaggeration and misdescription are common). But these rules also give some actors, especially members of the majority party, incentive to “speak their mind” or describe things how they genuinely understand them. If a committee’s proposal, for example, is not representative of the majority party’s collective interest, then it is highly probable that the proposal will be killed by either the Speaker or by the Rules Committee (i.e. it is highly probable that it will not advance to the floor). 224 Given these different incentives, we ought generally to take communication among majority members as significantly more credible than those of the minority. This does not mean that we should simply ignore the minority and take everything produced by the majority as credible sources of justification. We have to discriminate between records of majority communications as well. In examining majority communication – such as committee reports – we must, then, look for two things: penalties for lying and probability of verification. Boudreau et al. only roughly point out what kinds of situations in the legislative process count as providing penalties for lying; a more detailed identification is presumably further work for positive political theory. But they do identify some, such as the possibility of loss of leadership. 225 In the case of legislative rationale, I take it that e.g. exaggeration or significant misdescription of the justification behind a proposed statute might lead to losing a position as committee chair or as party whip etc. But only under certain circumstances is the probability of such a loss sufficient. If, for example, remarks about justificatory aims are made during open floor 224 Boudreau et al. (2007), 979. 225 Boudreau et al. (2007), 978. 227 time while the House is in recess, there are no penalties to speak of. 226 Hence, such remarks should be ignored – as they frequently are, if Boudreau et al. are correct. Other obvious penalties for misdescription might be loss of necessary support for passing the relevant bill. Such penalties are prominent, for example, in cases in which significant legislative bargaining is required in order to pass a bill. These sorts of penalties, as we will see, have a special sort of relevance to the discussion in this chapter. Matters of verification are a little trickier here, since I am concerned with justifications rather than factual statements (as I take to be the main focus of Boudreau et al.). Still, we can make relevant sense of the concept of verification by focusing on situations in which a possible statement grossly misrepresents what is commonly taken to be the justifying goals of a proposed law, i.e. the ends for which the legislation is intended to be a means. If the audience of the relevant remark does not include anyone who is able to verify whether the justification is misrepresented or not, or if there is little chance that a third party later checks the records, then conditions for sincerity are not met and the statement should be ignored. In general, whether or not the conditions for sincerity hold will be determined by whether the penalties for lying are sufficiently high given certain likelihood of verification and whether the likelihood of verification is sufficiently high given certain penalties for lying – i.e. by whether or not the product of penalties for lying and probability of verification is sufficiently weighty. In addition, it is important to privilege only statements that are made relatively late in the legislative process, or else the 226 Boudreau et al. (2007), 978. 228 statements – although sincere – may not reflect the ultimate bargains that actually facilitated the legislation. Thus, there are three main aspects to look for, vis-à-vis statements made by lawmakers in the legislative process: (i) the type of speaker (majority member or pivotal minority member), (ii) the expected utility of cheap talk, and (iii) when the statement is made. Now, as I mentioned already, it complicates matters even further that not only do we need to know under what conditions lawmakers’ utterances in general count as sincere, we also need to know whose utterances – if sincere – count, vis-à-vis fidelity to law. Boudreau et al. appear to assume that – as a general matter – it is the majority party that counts and that certain utterances by certain majority members can be taken as representative of the party’s view of the purposes of a particular piece of legislation. Perhaps this is true, and perhaps not. However, at least in cases in which significant legislative bargaining between the majority and moderate members of the minority is required in order for a bill to pass, the context of bargaining provides incentive for sincerity and the need for bargaining determines whose utterances count. To fix this last idea, let us consider the following scenario, borrowed and adapted from Marmor (2008). 227 Say that there are 100 members in a parliament. And that a bill has been initiated with the ardent support of 20 legislators. They need 31 additional votes, then, in order for the bill to pass. Assume that these 31 legislators are not ardent supporters of the bill, but that they have negotiated certain amendments to the initial bill in return for their support. Further, assume that the 20 ardent supporters have expressed 227 See Marmor (2008), 436–7. 229 an intention that would count in favor of extending the law to cover some borderline case x and that the 31 non-ardent supporters have expressed an intention that would count in favor of not extending the law to cover x. Given the fact that the commitments that the lawmakers have incurred recommend different decisions, vis-à-vis x, whose commitments should be taken to represent the commitments of the law? That is, insofar as the aim is to maximize fidelity to law, whose expressed rationale should guide judicial decisions in possible legal cases concerning x? This is not an easy question. As Marmor points out, giving effect to the expressed intentions of the 20 ardent supporters would be giving effect to a clear minority, which seems problematic. Not only are they a minority, they are the smallest minority with respect to this particular bill, with 31 moderate supporters and 49 that do not support it. Giving effect to the expressed intentions of the 31 moderate supporters also seems problematic, since they presumably know and care less about the relevant issues and also form a minority. Lastly, giving effect to some aggregate intention won’t help at all since what created the problem in the first place were conflicting intentions between members of the majority. The lesson that Marmor draws from cases of this sort is that in many circumstances involving borderline cases, the legislature should not be seen to have any particular intention that could settle whether or not x ought to be included in the relevant rule’s domain of application. If that is correct, then that points to another significant limitation of the model that I am proposing. In cases that involve significant legislative bargaining, however, the model is able avoid Marmor’s problem and may lead to slightly different conclusions (although this 230 will vary from case to case). The reason, explained above, is that when such bargaining is involved, the need for bargaining determines whose utterances count and the context of bargaining guarantees sincerity in those utterances, which in turn guarantees commitment to their content. The utterances that count are those made by the bargaining parties (in the context of bargaining) and these utterances can be taken to be sincere due to the presence of severe penalties for insincerity (here, loss of necessary support for passing the relevant bill). In the hypothetical case above, then, it seems clear – from the perspective of the present model – that, insofar as the aim is to maximize fidelity to law, judges ought to consult records of conversations between the 20 and the 31, since the hypothetical law is taken to be the result of significant legislative compromise. And it seems equally clear that both statements of the 49 (the minority) and the ardent supporters – when not made under conditions of sincerity – should be ignored. The model thus gives certain priority to the moderate 31 and, in particular, to statements by ardent supporters addressed to the moderates (in the context of bargaining). To fix these ideas even further with an actual example, let me describe very briefly an analysis, by Rodriguez and Weingast, of the use of legislative history in interpreting the Civil Rights Act of 1964. 228 Very long story short, significant compromise was needed in order to have the bill pass, both in the House and Senate, mainly because of a Southern Democrats’ filibuster. In particular, this required gaining the support of moderate Republicans, which generated a number of accepted amendments, mainly 228 42 U.S.C. §1981–2000(h). 231 fashioned to protect the interests of American businesses (especially in the North). 229 This resulted in a somewhat weakened bill, although the basic framework was unaltered, Rodriguez and Weingast argue. Litigations following the final bill’s enactment were numerous, and the courts often looked to legislative history as an aid in their decision-making process. History shows that, in the 60s, 70s, and 80s, judges frequently disregarded the compromises made at the later stages of the legislative process, relying primarily on statements made by ardent supporters, like the bill’s manager (and Democratic whip) Senator Humphrey and Senator Clark. 230 The result was a string of expansive interpretations of certain portions of the Act. In Griggs v. Duke Power Co., 231 for example, the court disregarded a crucial non- cheap statement made by ardent supporters aimed at Republicans in order to secure their support. The so-called Clark-Case memorandum includes a statement of one of the aims of Title VII of the Act: it is meant to “[protect] the employer's right to insist that any prospective applicant, Negro or white, must meet the applicable job qualifications. Indeed, the very purpose of Title VII is to promote hiring on the basis of job qualifications, rather than on the basis of race or color.” 232 In Griggs, the issue was whether or not Duke Power Co. had violated the Civil Rights Act by requiring a high school diploma and an IQ test for the higher paying jobs in the company. Since the 229 Rodriguez & Weingast (2003), 1472–73. 230 See also Rodriguez & Weingast (2007), 1210. 231 401 U.S. 424 (1971). 232 110 Cong. Rec. 7247 (1964). 232 statement quoted from the Clark-Case memorandum presumably qualifies as rationale offered publicly under sincerity-inducing conditions, it is arguable that in Griggs, the court should – insofar as its aim was to maximize fidelity to law – have decided for the defendants, although it is an open question whether or not fidelity to law should have been maximized in this case. If what I have said so far is correct, then judges should – insofar as the aim is to maximize fidelity to law – sometimes appeal to rationale expressed by lawmakers under sincerity-inducing conditions in order to settle borderline cases. One of the obvious limitations of this model is that it is limited to rationale that has actually been expressed, and so it makes sense to ask if there is any way in which to reasonably expand the model. In the next two sections, I will therefore consider two ways in which the model might be expanded to include rationale that is unexpressed but to which lawmakers are nevertheless committed (in the right way). 1.3 Is expression required? In this section, I want to address a worry regarding the requirement that the relevant rationale be explicitly offered by lawmakers. The worry is that in some cases, the purpose of an enactment may be quite obvious, and in that sense publicly available. If that is correct, then there seem to be cases in which appeal could reasonably be made to rationale that has not been explicitly offered, which may seem to indicate a problem for my view. If what I say below is correct, however, then one of the upshots of addressing 233 this worry is that although expression is still the normal means by which lawmakers incur commitment to legislative rationale, the applicability of the model is significantly expanded by allowing for the possibility that lawmakers are committed to such rationale in virtue of its being obvious. On the assumption that there really are cases in which there is an obvious and non-vacuous remedial purpose (that is, a sufficiently specific description of the mischief sought to be remedied), I think such cases are, on my account, best dealt with by saying that if the relevant rationale can in fact be legitimately appealed to, then that is because the legislature’s commitment to it has been incurred in a way that is sufficiently similar to the way in which commitment is incurred in cases of expressly offered justification. Or, to put it more concisely, I think that the way in which commitment to a rationale is incurred in virtue of its being obvious is simply a special case of the way in which commitment is incurrent by expression. On the account of assertion assumed throughout this dissertation, assertions are made (in part) against a set of background propositions that may be called the common ground between the speaker and hearer. Assertions are seen as attempts to add propositions to the common ground, which accurately predicts that in ordinary situations speakers will not assert propositions that are already part of it. The participants in the relevant conversation, however, still count as being committed to the propositions that form the common ground. This can be fairly well explained by looking at cases of presupposition accommodation – cases in which the speaker will utter a sentence the 234 assertive content of which presupposes something that is not yet part of the common ground. Imagine, for example, that you don’t know much about your co-worker and that she tells you that she has to pick up her sister at the airport. Her utterance presupposes that she has a sister, yet this information is not part of the common ground between the two of you. In order for the conversation to continue smoothly, you must therefore “accommodate” the presupposition, by adding the relevant information to the conversational background without objection. If you do not object, you (and, of course, the speaker) will now count as being committed to the truth of the proposition that she has a sister – and you will do so in virtue of the fact that this information has become part of the common ground. Presuppositions, then, are propositions taken for granted in a conversation. Moreover, it is mutually recognized that every participant in the conversation takes those propositions for granted. The relevance of this to my discussion here is that if a proposition is in fact genuinely obvious (and sufficiently salient), then it will be part of the common ground and thus the object of mutual commitment. Thus, although lawmakers can indeed be committed to genuinely obvious – but unexpressed – legislative rationale, such commitment will have been incurred in virtue of the rules of conversation, the same set of rules that explain expressly incurred commitment. It might be objected that the commitment I am talking about here is the wrong kind of commitment – that it is a special sort of linguistic commitment that falls short of the kind of the commitment required for my reply to the worry to work. I don’t think that 235 this is the case. The reason is that the obviousness of the relevant propositions gives participants to the conversation a warrant for treating each other in a certain way, namely as being committed to the truth of these propositions, unless otherwise indicated. Linguistic commitment is no less normative than any other form of commitment, or so I submit. Moreover, to see that there is nothing uniquely linguistic about this phenomenon, imagine that Bobbi and Clive form a joint plan to rob a bank. Among the things that form part of the common ground between Bobbi and Clive are the following propositions: Bobbi intends to rob a bank with Clive, Clive intends to rob a bank with Bobbi, Bobbi recognizes that Clive intends to rob a bank with her, and so on. Usually, several further propositions go into the common ground – some general and some specific. In Bobbi and Clive’s case, the following propositions might for example be part of it: Banks have security measures, This particular bank stands on a narrow one-way street, etc. Bobbi and Clive are each committed to these propositions. Further, this commitment is mutually recognized and they are therefore warranted in treating each other accordingly. If so, then each is, for example, entitled to expect the other not to spend part of the preparation budget on a wide and unwieldy getaway vehicle – exactly because their commitment to the proposition that the bank stands on a narrow one-way street is mutually recognized. There is nothing linguistic about this explanation, yet the phenomenon seems to be essentially the same, which suggests that the linguistic case is simply a special case of a 236 more general practical phenomenon. If that is correct, then the relevant commitment is no less normative than the practical commitment described in the bank robbery case. 233 Although perhaps not decisive, this reply offers a plausible explanation for why genuinely obvious, but unexpressed, rationale can be appealed to in settling borderline cases (again, assuming the aim is to maximize fidelity to law). It may be appealed to in virtue of being presupposed, i.e. of being mutually accepted as part of the common ground in the “legislative conversation”. However, it is still the case that, normally, commitment to the relevant legislative rationale is incurred via explicit statements made by lawmakers. 1.4 Commitment and counterfactuals There is another way in which we might expand the model. The idea is that in cases in which commitments expressly incurred by (the appropriate set of) lawmakers do not suffice to settle the normative question (‘Ought x count as an F?’), the notion of commitment may be flexible enough to include counterfactual expression of intention, which may dictate a verdict. If that is correct, then perhaps it makes sense to expand the model in such a way that a court also counts as deciding in accordance with fidelity to law if the decision is in accordance with the rationale that (the appropriate set of) lawmakers would be willing to offer under sincerity-inducing conditions. 233 This example is inspired by Lewis (1979), who also suggests a close connection between linguistic presupposition and joint plans, although his concern is with the analogy between the accommodation of presuppositions and of (sub)plans. 237 As far as the notion of commitment is concerned, I think that this is a reasonable expansion of the model. There are some significant limitations, however, on the scope of the resulting expansion. First, it is often very difficult to answer questions about counterfactuals, not least when they concern human action. To be sure, there will probably be certain highly general counterfactuals that can be considered plausible descriptions of the relevant lawmakers’ commitments, but the more abstract such descriptions are, the less likely they are to bring anything to bear on verdicts in particular borderline cases. In other words, in the case of counterfactuals relevant to settling borderline cases, the likelihoods of truth and practical application seem to be inversely related. A second limitation on this possible expansion of the model is that counterfactual propositions are themselves significantly vague. As Stoljar (2001) points out, standard theories of counterfactuals, such as those proposed e.g. by Lewis (1973) and Stalnaker (1968), analyze such propositions partly in terms of the comparative overall similarity of possible worlds. 234 The relation of comparative overall similarity, however, is vague, as Lewis himself points out. 235 This means that many counterfactuals that might be relevant with respect to settling borderline cases in courts are simply indeterminate (or false, depending on the theory). 236 Thus, although I think it does make sense to expand the model by including commitment analyzable in terms of counterfactual claims about the 234 See Stoljar (2001), 457–58. 235 Lewis (1973), 94. 236 For a good discussion of these issues as they relate to intentionalism in general, see Stoljar (2001), 457–63. 238 willingness of lawmakers to provide express rationale under sincerity-inducing conditions, we have good reason not to overestimate the extent of the resulting expansion. Before I continue on to the next section, I think it is in order to give an example of how legislative rationale of the sort I have discussed might help settle a borderline case, in particular given the time that I have spent explaining the limitations of the model. To take a rather mundane – and thus unexceptional – example, consider a standard pet ordinance that requires pets to be spayed and neutered. Imagine also that a homeowner’s backyard is visited by a male ferret and that the ferret eventually more or less comes to live there and is in many ways treated as a pet. Under these circumstances, it is easy enough to see how the ferret might be a borderline case of the term ‘pet’. The question, then, is whether or not the rationale of the pet ordinance can help settle whether the ferret ought to count as a pet (and thus ought to be neutered). I think it can. It seems safe enough to say that one of the rationales behind any spaying and neutering ordinance is to lower the risk of aggressive behavior in male animals. In case this rationale is not expressly offered by lawmakers, they therefore still have a counterfactual commitment to it, in the sense that it is rationale that they would be willing to offer expressly (even if they do not actually do so). Since a semi-domesticated unneutered ferret is likely to be dangerous, particularly in urban family areas, it seems reasonable to say – assuming the case reaches the courts – that, insofar as the aim is to maximize fidelity to law, the judges ought to count the ferret as a pet. There are cases, 239 therefore, in which the model does dictate a decision, despite all the limitations mentioned above. And since the example is a rather mundane one, we have reason to believe that the model can dictate a result in a significant – albeit modest – number of cases. In the next section, I want to address an important issue that is left open by the present model. The issue is that legislative rationale can be more or less general or specific, depending on how lawmakers choose to describe it, which invites the following question. If a specific rationale recommends a different decision than a general one, do we have reason to believe that lawmakers are – as a rule – more committed to one than to the other? This answer, I think, is ‘no’; it will vary hugely with circumstances whether lawmakers are more committed to one than to the other, or equally committed to both. 1.5 Legislative rationale and levels of abstraction It is clear that background rationales come in different levels of generality. Using the well-worn “No vehicles in the park” example, we can e.g. imagine the following plausible hierarchy of rationale: Statute: No vehicles in the park. Specific rationale: Reduction of noise, pollution, danger, etc. in the park. Intermediate rationale: Making the park a pleasant place to spend time. General rationale: Promote the well-being of the public. 240 For the sake of argument, let us assume that lawmakers have – in one way or other – incurred commitment to each of these rationales. Say that a case turns on whether or not an electric wheelchair ought to count as a vehicle for the purposes of the law. Each of the three rationales above arguably recommend the same decision, since counting electric wheelchairs as vehicles seems to help satisfy none of them. This invites at least two important questions. The first question is whether a court would – insofar as the aim is to maximize fidelity to law – be warranted in appealing to any one of these rationales in order to justify its decision that wheelchairs should not count as vehicles. The second question is whether the answer to the first question generalizes to all borderline cases. In the fictional example above, I think the answer to the first question is probably ‘yes’, on the assumption that the lawmakers are genuinely committed to each aim and regard the relatively specific aims as means to achieve the relatively general aims. But not all cases are so straightforward, which means that the answer to the second question is ‘no’. In some cases, the different rationales may recommend different decisions, in which case it makes sense to ask whether the lawmakers can be taken to be more committed to the specific rationale than to the general one, or vice versa. One way to go about answering this question is to ask which aims the lawmakers would be willing to give up if it turned out that the specific aims did not really promote the general ones. As Marmor (2005) notes, it would – as a matter of pure instrumental rationality – be incoherent to stick to the means if those means are inappropriate for achieving the ends, and so he holds that lawmakers’ specific aims should be taken into 241 account “only if, and to the extent that, their realization is likely to enhance [the legislature’s general aims].” 237 If that is correct, then considerations of instrumental rationality seem to indicate that lawmakers are more committed to their general aims than to their specific aims. However, not all questions regarding lawmakers’ commitment to general vs. specific legislative rationale can be answered by appeal to considerations of instrumental rationality. In some cases, the relationship between the general aims and the specific aims is not fundamentally a means-ends relationship at all. Rather, the general aim is a more abstract redescription of the specific aim. Since the commitment on behalf of the lawmakers is – in these cases – arguably just commitment to a single aim described in more or less abstract terms, it would make little sense to say that they are more committed to one aim than to the other. In other cases, the relationship between the general aims and the specific aims is indeed a means-ends relationship, but a complex one. At the very general level, lawmakers often have several aims that need to be balanced. These general aims will typically move them to draft bills, but as the proposed legislation is worked on, more specific aims usually emerge. Subaims are adopted, which are taken to be the best available means by which to attain the original aim while taking into account other relevant aims. Thus, the more specific aims work somewhat like complex subplans, intended to promote certain general aims while minimizing the danger of frustrating other general aims. In these cases, it becomes very difficult to determine whether the 237 Marmor (2005), 131. 242 lawmakers can be regarded as more committed to the general aims (taken individually) than to the specific aims, or vice versa. In yet other cases, the relationship between the general aims and the specific aims is also a means-ends relationship, but the lawmakers’ commitment to the specific aims is affected by considerations beyond instrumental rationality. In many cases, the specific aims will be the result of legislative bargaining, in which case the moderate members of the minority may have “signed off” on the general aim of the relevant bill (proposed by the majority) but not entirely on the means by which to achieve this end (also proposed by the majority). Thus, the moderates may negotiate a qualification on the means, for example by securing exceptions for certain things etc. In these cases, it also becomes difficult to determine whether it makes sense to say that the relevant lawmakers are more committed to the general aims than to the specific ones (or vice versa). The above considerations are admittedly very abstract, but their role is just to illustrate a limited but important point: It will vary hugely with the circumstances whether considerations of instrumental rationality warrant that we take lawmakers to be more committed to their general aims than to their specific aims. We should be careful, therefore, in claiming that – insofar as fidelity to law is concerned – general aims trump specific ones (or vice versa for that matter). As a result, fidelity to law alone will often not suffice to settle a borderline case if two or more of the relevant aims (at different levels of generality) recommend different decisions. To illustrate this point further with a concrete – albeit fictional – example, let us borrow one from Soames (2011). Imagine a community – Plainsboro – whose legislature 243 enacts the following statute: “It shall be a misdemeanor in the Township of Plainsboro for children on their way to or from school to accept rides in automobiles from strangers.” 238 The background rationale of the statute is to “reduce the danger of sexual assaults against the town’s children.” 239 But, as we have seen, justifications come in different levels of abstraction, one of which can appropriately be characterized as the aim of “reducing the danger of harm to [the town’s] residents.” 240 Let us, for the sake of argument, say that the lawmakers are committed to both rationales. To illustrate how the two rationales may recommend different decisions in borderline cases, Soames invites us to imagine that Susan – to whom the statute applies – works after school in a dangerous part of Plainsboro and that working there makes her a likely target of robbery. In case Susan accepts a ride to work from a borderline stranger who is clearly not dangerous (and so the more specific background rationale would not warrant a conviction), appealing to the more general aim could be used to justify convicting Susan. The reason being that doing so would discourage people from getting rides to dangerous parts of town, thereby reducing the danger of harm to the town’s residents. Soames argues that this example shows that whereas appeal to the more specific aim can be appropriately used in determining borderline cases, the more general one cannot. If what I have said so far is correct, however, this is not guaranteed by fidelity to 238 Soames (2011), 36. 239 Soames (2011), 55. 240 Soames (2011), 55. 244 law considerations alone. In particular, Soames’s case seems to be one in which the relationship between the general and the specific aim is not fundamentally an instrumental one. Rather, the general aim is a redescription of the specific aim, in which case I have said that it does not really make sense to say that the lawmakers are more committed to one than to the other. As a result, the judgment that judges ought to decide the case by appeal to the more specific aim has to be grounded in substantive considerations beyond fidelity to law. To sum up the discussion so far, I have proposed that insofar as the aim is to maximize fidelity to law in settling a borderline case, judges ought to decide in accordance with those legislative purposes to which lawmakers have incurred a commitment, and that the primary way in which lawmakers incur the relevant kind of commitment is via the express offering of legislative rationale under conditions that deter cheap talk. I also argued, however, that lawmakers may be committed to certain legislative rationale in virtue either of the fact that it is just obvious what that rationale is, and so there is no need to express it, or in virtue of the fact that they would be willing to offer the relevant rationale under sincerity-inducing conditions. And although such rationale may not always be available or, if it is, it may not always suffice to settle the relevant case, we should nevertheless expect that in a significant number of cases, borderline cases can be settled in accordance with fidelity to law. Now, even if what I have said so far is correct, all that this establishes is that in order to maximize fidelity to law, judges sometimes ought to defer to a law’s operative 245 rationale, i.e. the rationale to which the (relevant set of) lawmakers have incurred commitment. Since this is an instrumental claim, it doesn’t tell us anything about the value of the relevant end, i.e. of maximizing fidelity to law. Next, therefore, I want to turn to the second main question of the chapter, the question whether judges have reason to maximize fidelity to law, and if so, then under what conditions? 2.1 Authority, legislative bargaining, and maximizing fidelity to law As I indicated in the introduction, I think that there are primarily two types of reasons for judges to defer to the operative rationale of a statute. First, as Marmor (2005) has argued, judges can have authority-based reasons for such deference, if in doing so, they are more likely to comply with right reason than if they try to figure out the best course of action on their own. Second, judges can have function-preserving reasons to do so. In legal systems in which there is an adversarial legislative process and widespread conflict of interest – such as the US legal system – legislative bargaining is a fundamental feature of the system; without compromise among lawmakers, the system is not able to do what it is supposed to do, which is to produce legislation. On the safe assumption that the system is a valuable thing to have, there will therefore in many cases be a need to respect compromises reached in the legislative process, which in turn gives rise to reasons on behalf of judges to decide in accordance with the operative rationale of the relevant statute. Marmor (2005) argues that “generally, the primary way of justifying reasons for complying with the intentions of the legislator involves the very same considerations 246 which are taken to vindicate compliance with an authority’s directives in the first place.” 241 On the assumption that Joseph Raz’s “service conception” of authority captures the nature of legitimate authority – in particular, what he calls the normal justification thesis – it follows that deference to legislative intentions must be justified in one of the two following ways: 242 such deference is justified if the lawmakers can be expected either to have expertise with respect to the relevant domain of conduct or, in case the matter at hand is a collective action problem, if they can be expected to be in a superior position to solve the relevant problem. Following Marmor, call these the expertise justification thesis and collective action justification thesis, respectively. Marmor further argues that there is rarely reason to believe that lawmakers are in a superior position to judges with respect to solving collective action problems. “When the legitimacy of the legislator’s authority … derives from the collective action justification thesis,” he says, “judges have no particular reason to defer to the legislators.” 243 Thus, if deference to operative rationale is ever justified, according to Marmor, this justification must be grounded in the expertise justification thesis. If what I say below is correct, then this conclusion is too strong – respect for legislative bargaining can also give judges a reason to defer to a statute’s operative rationale. Moreover, if such reasons conflict with authority-based reasons, then – as I will argue – the former often outweigh the latter. 241 Marmor (2005), 134. 242 See Raz (1986), 53, for the basic idea. 243 Marmor (2005), 136. 247 As Marmor acknowledges, it is presumably often sensible – on the basis of the expertise justification thesis – to defer to a law’s operative rationale (consider, for example, FDA and EPA regulations). But if expertise (with respect to a certain domain of conduct) is lacking, then judges do not have an authority-based reason to defer to the expressed intentions of the lawmakers. Further, if a particular judge would do better to try to figure out the best course of action on her own, then she would actually have a positive reason to refrain from such deference. It is important, however, to note that there may be factors beyond considerations of instrumental rationality that are relevant to determining reasonable deference to a law’s operative rationale. Although instrumental considerations play a major role in determining the extent to which the aims of lawmakers ought to be taken into account, there are certain institutional factors that may be crucially relevant as well. In fact, Marmor (2011b) argues that we cannot fully understand the concept of legitimate authority unless we give due consideration to the institutional setting in question. Most importantly, the reasons for complying with the directives of an authority are intimately connected to the reasons for having the institution in virtue of which the relevant authority possesses its normative power. That is, whether or not an authority is legitimate cannot, à la Raz’s service conception, be determined solely by whether or not the subject, on average, does better – vis-à-vis conformity with right reason – by following the authority’s directives (i.e. better than she would do on her own). Legitimacy, says 248 Marmor, must also involve good reasons to have the kind of institution in virtue of which the relevant normative power exists. 244 Something similar, I think, can be said with respect to a significant subset of cases in which judges lack authority-based reasons to defer to operative rationale, including cases in which they actually have authority-based reasons to refrain from doing so. On the assumption that there is good reason to have the relevant legislative institution, in particular some particular aspect of it, there may often be good reason to respect the aims expressed by lawmakers, even when the recommendations of those aims conflict with the recommendations of the expertise justification thesis. As I will illustrate with an example in the next paragraph, a very strong case can be made that operative rationale should often prevail even if adopting the relevant aims is – strictly speaking – not the best course of action (from the perspective of the expertise justification thesis). The primary reason is that if courts do not, as a rule, respect the essential compromises made in the legislative process, then similar agreements will be harder to reach in the future, and the valuable institution of legislative bargaining is thereby undermined. Since such compromise is at the foundation of the institutional setting of many contemporary legal systems, it seems plausible to say that there are often weighty institutional reasons to defer to operative rationale, weighty enough to sometimes outweigh competing instrumental reasons. Recall, for example, the case of Griggs v. Duke Power Co., in which the question was whether or not Duke Power Co. had violated the Civil Rights Act by requiring a high school diploma and an IQ test for the higher paying jobs in the company. One strong 244 This line of though is further developed in Marmor (2011c). 249 reason for thinking that the court should – all things considered – have decided for the defendants is, as Rodriguez and Weingast (2003) point out, that ignoring essential compromise and thus important parts of the operative rationale of a law, will result in subsequent difficulties in reaching important agreements. Why would legislators have any incentive to offer support on moderate premises if the courts will interpret the law expansively according to the cheap talk of ardent supporters? Broad and expansive interpretations are sure to make subsequent legislative compromise very difficult. Since legislative compromise is an essential part of an institution that is good to have, courts therefore have a weighty reason to respect legislative bargaining in their decisions. And they respect such bargaining by deferring to the operative rationale of the law (if it is available). Granted, Griggs was not a case of vagueness, or was in any case not treated as one. It seems safe enough, however, to say that if the judges in this case had a strong bargaining-based reason to defer to the operative rationale of Title VII of the Civil Rights Act, then surely judges can have such reasons in borderline cases too. If that is correct, then the normative argument offered in the preceding couple of paragraphs suffices for the philosophical purposes of this chapter. There is of course plenty work to do in identifying the exact conditions under which the need to respect legislative bargaining provides the most weighty reasons. This, however, is presumably work for positive political theory and related fields, most notably behavioral law and economics. As a last remark, I should note that bargaining-based considerations provide judges with a strong reason to defer to specific rationale rather than general rationale (in 250 case they conflict). To see this, consider again cases in which the legislature’s specific aims fail to promote its general aims. As I explained above, it would – as a matter of pure instrumental rationality – be incoherent to stick to the means if those means are inappropriate for achieving the ends. Respect for legislative bargaining, however, sometimes requires that deference be made to the legislature’s specific aims even if they do not sufficiently promote its general aims. The reason is that in some cases the primary concern is not rational coherence on behalf of lawmakers, but rather whether or not ignoring their subplans – i.e. their own views about how to achieve their general legislative goals – runs the risk of impeding future legislative compromise. Insofar, then, as respect for legislative bargaining gives judges a reason to defer to the operative rationale of a law, it gives them a reason to privilege rationale that is relatively specific. Before I conclude this chapter by addressing the worry that what I have been calling respect for legislative bargaining actually undermines, rather than promotes, the institution of legislative compromise, I would like to briefly sum up the most significant similarities and differences between authority-based considerations and considerations based on respect for legislative bargaining, vis-à-vis deference to such rationale. The chief difference, of course, is that authority-based reasons are – unlike bargaining-based reasons – not dependent on a particular legal system being in place. A second important difference is that whereas – as we saw – respect for legislative bargaining privileges specific rationale, authority-based considerations, as Marmor notes, privilege general 251 rationale. 245 As for the main similarities, both types of considerations provide non- absolute reasons to defer to legislative intent and in both cases the weight of the relevant reasons is significantly affected by the dimension of time. As Marmor also notes, expertise changes over time and thus the weight of an expertise-based reason tends to decrease as time goes by. 246 Similarly, the importance of respect for legislative bargaining lessens as the bargains get older, since the negative “feedback effect” on the legislature arguably decreases over time. 2.2 Expressly acknowledged compromise vs. tacitly acknowledged compromise I have claimed – and spent a significant portion of this chapter arguing – that one of the main reasons that judicial appeal to background justification is (all things considered) appropriately limited to operative rationale is that it respects, and thus helps to preserve, a fundamental feature of our legislative system – the valuable institution of legislative compromise. I want, therefore, to conclude the chapter by addressing the worry that the prescriptive account that I have been arguing for might actually undermine the ability to reach such compromise. The worry is worthwhile addressing because it illustrates something significant about the different kinds of strategic compromise at play in the legislative environment. 245 See Marmor (2005), 139. 246 Marmor (2005), 138. 252 In order to see why this worry is worth taking seriously, consider the following argument about the strategic mechanics of legislative compromise, presented by Marmor (2011a): Consider … two legislators striving to achieve a compromise on a particular legislative act. … Part of what enables opposing parties to reach a compromise consists in their ability to conceal, or at least not make very explicit, the overall implications of their collective speech. … Had the conversational maxims that govern a strategic conversation been as clear and determinate as they are in the case of an ordinary conversation, it would have been very difficult to make strategic moves in the conversation. Precisely because there is some degree of uncertainty about the norms that govern the conversation, parties can exploit implications of what they say in ways which they would not be willing to make very explicit ex ante. 247 The thought behind this passage is, I take it, something like the following. Lawmakers are often able to reach compromises in virtue of the fact that it is uncertain how the courts will interpret certain bits of the legal text. Each party to the compromise hopes – and usually does what it can, in and out of Congress, to ensure – that its favored interpretation will prevail. That is, a compromise may seem like a good bargain if each party has a reasonable chance to fight for their interests on another front. The worry that I have in mind, then, is this: if the courts were to follow my recommendation that appeal to legislative purpose be limited to operative rationale, legislators might lose a great deal of the incentive that they have to compromise, since it will in many (reasonably foreseeable) cases be evident which way the operative rationale leans. In other words, it would seem 247 Marmor (2011a), 94. 253 that certainty at once facilitates and impedes compromise! Something appears to be amiss. I think that the context of the Civil Rights Act can help to dispel this worry, because it highlights a general feature of legislative compromise, a feature that may have been abstracted away from in Marmor’s example. The feature is this: the bargaining positions of the two legislators in Marmor’s example cannot – if the example is to reflect the legislative situations that I have been discussing – be seen to be equal. That is, we must recognize that the moderate legislator has the upper hand. Thus, although it is, I believe, correct that legislators in an equal bargaining position would lose their incentive to compromise in case the interpretive principles employed by courts were clear and determinate, what matters in normal cases of the sort of legislative compromise I have been discussing is that the moderates, whose support is required in order to pass the relevant bill, believe that it is sufficiently likely that any compromises will be respected by the courts. Now, this is not to say that it cannot happen, with respect to some parts of a bill, that ardent supporters and moderates are in roughly equal positions. In some cases, the parties will both be sufficiently eager to not let some issue, irresolvable by explicit agreement, sink the legislative effort. If so, uncertainty regarding judicial interpretation would offer a way to resolve the matter “tacitly”, as Marmor calls it, by leaving “some of the implications of their collective expression undetermined or deliberately vague or 254 ambiguous.” 248 And here it really is a pressing question whether the account I have been espousing undermines the possibility of such a resolution. While I do recognize that there is potential trouble here for my view, I think there is a fairly natural strategy that lawmakers can, and probably would, employ in order to retain mutual incentives to compromise in the way suggested by Marmor, even if the courts were to consider operative rationale the only authoritative rationale. The strategic norm would be something like this: In order to ensure sufficient uncertainty regarding outcomes in the relevant types of cases, take care not to reveal any compromissory legislative purpose that might settle the matter. In fact, it is unlikely that lawmakers will even need any strategic instruction of this sort, since it is highly unlikely that any such explicit rationale will emerge, exactly because of the strategic considerations that Marmor identifies. Lawmakers will indeed have an incentive to not make explicit anything that might dictate a decision either way. It seems, then, that the tension between the certainty required for the sort of compromise I discussed and the uncertainty required for the sort of compromise discussed by Marmor is only superficial, or in any case can be – and typically is – avoided by giving due attention to what the operative rationale does and does not contain. Thus, the notion of operative rationale does after all allow both for Marmor’s “tacitly acknowledged” type of compromise and for the sort of compromise highlighted by laws like the Civil Rights Act, which we can call expressly acknowledged compromise, both of which seem centrally important to the strategic practice of democratic lawmaking. 248 Marmor (2011a), 94. 255 2.3 Conclusion In this chapter, I have discussed some plausible restrictions on what can count as an answer to the question whether – from the perspective of the law – x ought to count as an F. If x is an absolute borderline case of ‘F’, I argued, x ought to count as an F only if this is dictated by the operative rationale of the relevant law, at least insofar as the aim is to maximize fidelity to law. The operative rationale of a law, I said, is normally the legislative rationale to which lawmakers have incurred commitment by expressly offering it during the legislative process under sincerity-inducing conditions, although they may also count as being committed to such rationale either in virtue of the fact that they would be willing to offer it under such conditions or in virtue of the fact that the relevant rationale is just obvious. I also acknowledged, however, that there are significant limitations to this model. In many borderline cases, appeal to legislative rationale will not dictate a decision, either because the relevant kind of rationale isn’t available or because, if available, it doesn’t suffice to pull the decision one way rather than the other. These sorts of limitations on decision procedures, however, are to be expected, especially if such procedures depend heavily on contingent matters, as they do in this case. I also argued that although appeal to operative rationale can help settle a verdict in a borderline case, it is not always the case that it should, which suggests a further significant limitation on the model. That is, it is one thing to determine what fidelity to law requires and quite another to determine whether or not fidelity to law should prevail over other competing considerations relevant to legal interpretation. I argued that judges 256 can have at least two kinds of reasons to maximize fidelity to law in borderline cases, i.e. to defer to operative rationale. They may have expertise-based reasons to do so, as discussed by Marmor (2005), or they may have reasons grounded in the need to respect legislative bargaining. What we get from all this is the following, rather complex, generalization about legal interpretation in borderline cases (for the sake of relative simplicity, I’m ignoring obvious commitment and counterfactual commitment in the following statement): In cases of vagueness, judges (pro tanto) ought – if the relevant legislation pertains to a domain of conduct with respect to which lawmakers possess relative expertise or if the legislation was reached through significant legislative bargaining – to settle the question whether x ought to count as an F by reference to the rationale (general or specific, depending on the type of legislation involved) offered expressly under sincerity-inducing conditions late in the legislative process by majority members or by members of the minority pivotal to the legislation. As I have stressed, this does not describe a complete decision procedure for resolving vagueness in the law. Nor should it. What it does describe is a modest but significant the set of circumstances under which (a constrained form of) intentionalism is the most appropriate approach to legal interpretation in cases of vagueness. 257 BIBLIOGRAPHY Atlas, J. D. & Levinson, S. (1981). It-clefts, Informativeness and Logical Form. In P. Cole (Ed.), Radical Pragmatics (pp. 1–62). New York: Academic Press. Audi, R. (1973). Intending. Journal of Philosophy, 70, 387–403. Bach, K. (1994). Conversational Impliciture. Mind and Language, 9, 124–162. Bach, K. & Harnish, R. M. (1979). Linguistic Commuication and Speech Acts. Cambridge, Mass.: MIT Press. Bar-Hillel, M. & Wagenaar, W. A. (1991). The Perception of Randomness. Advances in Applied Mathematics, 12, 428–54. Boudreau, C., Lupia, A., McCubbins, M., & Rodriguez, D. (2007). What Statutes Mean: Interpretive Lessons from Positive Theories of Communication and Legislation. San Diego Law Review, 44, 957–992. Broome, J. (1997). Is Incommensurability Vagueness? In R. Chang (Ed.), Incommensurability, Incomparability, and Practical Reason (pp. 67–89). Cambridge, Mass.: Harvard University Press. Calfee, J. & Craswell, R. (1984). Some Effects of Uncertainty on Compliance with legal Standards. Virginia Law Review, 70, 965–1003. Calfee, J. & Craswell, R. (1986). Deterrence and Uncertain Legal Standards. J.L. Econ. & Organization, 2, 279. Chang, R. (1997). Introduction. In Chang (Ed.), Incommensurability, Incomparability, and Practical Reasoning (pp. 1–34). Cambridge, Mass: Harvard University Press. Craswell, R. (1993). Property Rules and Liability Rules in Unconscionability and Related Doctrines. The University of Chicago Law Review, 60(1), 1–65. Dan-Cohen, M. (1984). Decision Rules and Conduct rules: On Acoustic Separation in Criminal Law. Harvard Law Review, 97, 625–677. Davis, W. (1984). A Causal Theory of Intending. American Philosophical Quarterly, 21(1), 43–54. Duxbury, N. (2008). The Nature and Authority of Precedent. Cambridge: Cambridge University Press. Easterbrook, F. H. (1983). Statutes’ Domain. University of Chicago Law Review, 50, 533. 258 Easterbrook, F. H. (1988). The Role of Original Intent in Statutory Construction. Harv. JL & Pub. Pol'y, 11, 59. Easterbrook, F. H. (1990). What Does Legislative History Tell Us. Chi.-Kent L. Rev., 66, 441. Easterbrook, F. H. (1999). The Case of the Speluncean Explorers: Revisited. Harvard Law Review, 112, 1913. Endicott, T. (2000). Vagueness in Law. Oxford: Oxford University Press. Endicott, T. (2011). The Value of Vagueness. In A. Marmor & S. Soames (eds.), Philosophical Foundations of Language in the Law (pp. 14–30). Oxford: Oxford University Press. Feldman, Y. & Teichman, D. (2009). Are All Legal Probabilities Created Equal? NYU Law Review, 84, 980–1022. Fodor, J. (1991). You Can Fool Some of The People All of The Time, Everything Else Being Equal; Hedged Laws and Psychological Explanations. Mind, 100, 19–34. Franke, M., de Jager, T. & van Rooij, R. (2009). Relevance in Cooperation and Conflict. Journal of Logic and Computation. Gallie, W. B. (1956). Essentially Contested Concepts. Proceedings of the Aristotelian Society, New Series, Vol. 56, 167–198. Geis, M. L. 1995. The Meaning of Meaning in the Law. Washington University Law Quarterly, 73, 1125. Gilovich, T., Vallone, R., & Tversky, A. (1985). The Hot Hand in Basketball: On the Misperception of Random Sequences. Cognitive Psychology, 17, 295–314. Goldsworthy, J. (1994). Implications in Language, Law, and the Constitution. In G. Lindell (Ed.), Future Directions in Australian Constitutional Law (pp. 150–184). Sydney: Federation Press. Goldsworthy, J. (2003). Raz on Constitutional Interpretation. Law and Philosophy, 22, 167–193. Goldsworthy, J. (2010). Parliamentary Sovereignity. Cambridge: Cambridge University Press. Greenberg, M. (2011). Legislation as Communication? Legal Interpretation and the Study of Linguistic Communication. In A. Marmor & S. Soames (Eds.), The Philosophical 259 Foundations of Language in the Law. (pp. 217–256). New York: Oxford University Press. Grice, H. P. (1971). Intention and Uncertainty. Proceedings of the British Academy, 5, 263–279. Grice, H.P. (1989). Studies in the Way of Words. Cambridge, MA: Harvard University Press. Griffin, J. (1986). Well-Being: Its Meaning, Measurement and Importance, Oxford: Clarendon Press. Guttel, E. & Harel, A. (2008). Uncertainty Revisited: Legal Prediction and Legal Postdiction. Michigan Law Review, 107, 466–500. Hadfield, G. (1994a). Weighing the Value of Vagueness. California Law Review, 82, 541. Hadfield, G. (1994b). Judicial Competence and the Interpretation of Incomplete Contracts. The Journal of Legal Studies, 23(1), 159–184. Hahn, U. & Warren, P. A. (2009). Perceptions of Randomness: Why Three Heads Are Better Than Four. Psychological Review, 116, 454–61. Harman, G. (1986). Willing and Intending. In R. E. Grandy & R. Warner (Eds.), Philosophical Grounds of Rationality (pp. 363–380). Oxford: Oxford University Press. Hart, H. L. A. (1994). The Concept of Law (2nd Ed.). Oxford: Clarendon Press. Hart, H. M. & Sacks, A. M. (1994). The Legal Process: Basic Problems in the Making and Application of Law. W. N. Eskridge, Jr. & P. P. Frickey (Eds.) New York: Foundation Press. Hawthorne, J. (2004). Knowledge and Lotteries. Oxford: Oxford University Press. Hurd, H. M. (1990). Sovereignty in Silence. The Yale Law Journal, 99, 945-1028. Husak, D. (2008). Overcriminalization: The Limits of the Criminal Law. Oxford: Oxford University Press. Kahneman, D. & Tversky, A. (1972). Subjective Probability: A Judgment of Representativeness. Cognitive Psychology, 3, 430–54. Kim, J. (1974). Noncausal Connections. Noûs, 8(1), 41–52. Kolodny, N. (unpublished). Toward a Principle of Instrumental Transmission. 260 Kratzer, A. (1981). The Notional Category of Modality. In H. J. Eikmeyer & H. Rieser (Eds.), Words, Worlds, and Contexts (pp. 38–74). Berlin: Walter de Gruyter. Lewis, D. (1973). Counterfactuals. Oxford: Blackwell. Lewis, D. (1979). Scorekeeping in a Language Game. Journal of Philosophical Logic, 8, 339–359. Lupia, A. & McCubbins, M. (1998). The Democratic Dilemma: Can Citizens Learn What They Need to Know? Cambridge: Cambridge University Press. MacCallum. G. (1966). Legislative Intent. The Yale Law Journal, 75(5), 754–787. Manning, J. F. (2001). Textualism and the Equity of the Statute. Columbia Law Review, 101, 1–127. Manning, J. F. (2003). The Absurdity Doctrine. Harvard Law Review, 116(8), 2387– 2486. Manning, J. F. (2005). Textualism and Legislative Intent. Virginia Law Review, 91, 419– 450. Manning, J. F. (2006). What Divides Textualists from Purposivists? Columbia Law Review, 106, 70–111. Marmor, A. (2001). Positive Law and Objective Values. Oxford: Oxford University Press. Marmor, A. (2005). Interpretation and legal theory (2nd Ed.). Oxford: Hart Publishing. Marmor, A. (2007). Law in the Age of Pluralism. Oxford: Oxford University Press. Marmor, A. (2008). The Pragmatics of Legal Language. Ratio Juris, 21, 423–452. Marmor, A. (2011a). Can the Law Imply More Than it Says? In Marmor, A. & Soames, S. (Eds.), Philosophical Foundations of Language in the Law (pp. 83–104). Oxford: Oxford University Press. Marmor, A. (2011b). The Dilemma of Authority. Jurisprudence, 2(1), 121–141. Marmor, A. (2011c). An Institutional Conception of Authority. Philosophy & Public Affairs, 39(3), 238–261. Marmor, A. (forthcoming). Truth in Law. In Freeman, M. & Smith, F. (Eds), Current Legal Issues: Law and Language. Oxford: Oxford University Press. 261 Martinich, A. P. (1984). Communication and Reference. Berlin: de Gruyter. Neale, S. (2005). Pragmatism and Binding. In Z. Szabó (Ed.), Semantics versus Pragmatics (pp. 165-285). Oxford: Oxford University Press. Neale, S. (2007). On Location. In M. O'Rourke & C. Washington (Eds.), Situating Semantics: Essays on the Philosophy of John Perry (pp. 251–393). Cambridge, MA: MIT Press. Ostertag, G. (unpublished). Cruelty and Kinds: Scalia, Dworkin, and Constitutional Pragmatics. Perry, J. (2011). Textualism and the Discovery of Rights. In A. Marmor & S. Soames (Eds.), Philosophical Foundations of Language in the Law (pp. 105–129). Oxford: Oxford University Press. Pietroski, P. & Rey, G. (1995). When Other Things Aren’t Equal: Saving Ceteris Paribus Laws from Vacuity. British Journal for the Philosophy of Science, 46, 81–110. Raz, J. (1972). Legal Principles and The Limits of Law. Yale Law Journal, 81, 823–854. Raz, J. (1986). The Morality of Freedom. Oxford: Oxford University Press. Raz, J. (2001). Sorensen: Vagueness has no function in law. Legal Theory, 7, 417–419. Recanati, F. (2002). Does Linguistic Communication Rest on Inference? Mind and Language, 17, 105–126. Recanati, F. (2004). Literal Meaning. Cambridge: Cambridge University Press. Recanati, F. (in press). Pragmatic Enrichment and Conversational Implicature. In D. Graff Fara & G. Russell (Eds.), The Routledge Companion to Philosophy of Language. London: Routledge. Redmayne, M. (1999). Standards of Proof in Civil Litigation. Modern L. Rev., 62, 167– 195. Rodriguez, D. B. & Weingast, B. R. (2003). The Positive Political Theory of Legislative History: New Perspectives on the 1964 Civil Rights Act and Its Interpretation. University Of Pennsylvania Law Review, 151, 1417–1542. Rodriguez, D. B. & Weingast, B. R. (2007). The Paradox of Expansionist Statutory Interpretations. Nortwestern University Law Review, 101, 1207. Ross, J. (2009). How to be a Cognitivist about Practical Reason. Oxford Studies in Metaethics, 4, 243–281. 262 Ross, J. & Schroeder, M. (forthcoming). Belief, Credence, and Pragmatic Encroachment. Philosophy and Phenomenological Research. Ruben, D. (2010). W.B. Gallie and Essentially Contested Concepts. Philosophical Papers, 39, 257–70. Ruger, T. W., Kim, P. T., Martin, A. D., & Quinn, K. M. (2004). The Supreme Court Forecasting Project: Legal and Political Science Approaches to Predicting Supreme Court Decisionmaking. Colum. L. Rev., 104, 1150. Scalia, A. (1997). A matter of interpretation: Federal courts and the law. Princeton N.J.: Princeton University Press. Schauer, F. (1987). Precedent. Stanford Law Review, 39, 571. Schauer, F. (1991). Playing by the Rules. Oxford: Oxford University Press. Schauer, F. (2010). Is There a Psychology of Judging? In D. E. Klein & G. Mitchell (Eds.), The Psychology of Judicial Decision Making (pp. 103–20). Oxford: Oxford University Press. Shaer, Ben. (unpublished). Toward a Cognitive Science of Legal Interpretation. Soames, S. (1999). Understanding Truth. Oxford: Oxford University Press. Soames, S. (2008). Drawing the Line Between Meaning and Implicature — and Relating both to Assertion. Nous, 42(3), 529-554 Soames, S. (2009). The Possibility of Partial Definition. In S. Soames, Philosophical Essays, Vol. II (pp. 362–381). Princeton: Princeton University Press. Soames, S. (2010). Philosophy of Language. Princeton, NJ: Princeton University Press. Soames, S. (2011). What Vagueness and Inconsistency Tell Us About Interpretation. In A. Marmor & S. Soames (Eds.), Philosophical Foundations of Language in the Law (pp. 31–57). Oxford: Oxford University Press. Sorensen, R. (2001a). Vagueness Has No Function in Law. Legal Theory, 7, 385–415. Sorensen, R. (2001b). Vagueness and Contradiction. Oxford: Oxford University Press. Sperber, D. & Wilson, D. (1995). Relevance (2nd Ed.). Oxford: Blackwell Publishers. Stalnaker, R. (1968). A Theory of Conditionals. Studies in Logical Theory, American Philosophical Quarterly Monograph Series, 2, 98. 263 Stanley, J. (2005). Knowledge and Practical Interests. Oxford: Oxford University Press. Stoljar, N. (2001). Vagueness, Counterfactual Intentions, and Legal Interpretation. Legal Theory, 7, 447–465. Väyrynen, P. (2009). A Theory of Hedged Moral Principles. Oxford Studies in Metaethics, Volume 4, 91–132. Vermeule, A. (1998). Legislative History and the Limits of Judicial Competence: The Untold Story of Holy Trinity Church. Stanford Law Review, 50, 1833–1896. Waldron, J. (2011). Vagueness and the Guidance of Action. In A. Marmor & S. Soames (Eds.), Philosophical Foundations of Language in the Law (pp. 58–82). Oxford: Oxford University Press. Williamson, T. (2002). Knowledge and its Limits. Oxford: Oxford University Press. Yaffe, G. (2009). A Procedural Rationale for the Necessity Defense. Journal of Value Inquiry, 43(3), 369–390. 264 APPENDIX: LIST OF LEGAL CASES Ali v. Federal Bureau of Prisons, 552 U.S. 214 (2008) Bradshaw v. Richey, 546 U.S. 74 (2005) Bronston v. United States, 409 U.S. 352 (1973) Brown v. Board of Education, 347 U.S. 483 (1954) Church of the Holy Trinity v. United States, 143 U.S. 457 (1892) Curran v. Mount Diablo Boy Scouts, 17 Cal. 4th 670 (1998) FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120 (2000) Green v. Bock Laundry Machine Co., 490 U.S. 504 (1989) Griggs v. Duke Power Co., 401 U.S. 424 (1971) Johnson v. United States, 529 U.S. 694, 718 (2000) Kosak v. United States, 465 U.S. 848, 854 (1984) Randall v. Orange County Council, 17 Cal. 4th 736 (1998) Small v. United States, 544 U.S. 385 (2005) Smith v. United States, 508 U.S. 223 (1993) Staples v. United States, 522 U.S. 398 (1998). Texas v. Johnson, 491 U.S. 397 (1989) United States v. Bacto-Unidisc, 394 U.S. 784 (1969) United States v. Eichman, 496 U.S. 310 (1990) United States v. Gonzales, 520 U.S. 1 (1997) United States v. Jewell, 532 F.2d 697 (1976) United States v. Laws, 163 U.S. 258 (1896) United States v. Marshall, 908 F.2d 1312 (7 th Cir. 1990)

Abstract (if available)

Abstract Lawmaking is - paradigmatically - a type of speech act: people make law by saying things. It is natural to think, therefore, that the content of the law is determined by what lawmakers communicate. However, it is sometimes vague what content they communicate, and even when it is clear, the content itself is sometimes vague. In my dissertation, I examine the nature and consequences of these two linguistic sources of indeterminacy in the law. I argue that vagueness in the law is sometimes a good thing, although its value should not be overestimated. I also propose a strategy for resolving borderline cases, which - if correct - sheds significant light on the debate about legal interpretation.

Linked assets