University of Southern California Dissertations and Theses

The grammar of correction

(USC Thesis Other)

The grammar of correction

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content The Grammar of Correction by Ulrike Steindl Dissertation Presented to the Faculty of the Graduate School of the University of Southern California In Partial Fulfillment of the Requirements for the Degree of DOCTOR OF PHILOSOPHY (LINGUISTICS) December 2017 Contents 1 Introduction 1 Introduction 7 1.1 Formalizing the difference between sondern and aber 1.1 Formalizing the difference between sondern and aber . . . . . . . . . . . . . . . . 12 1.1.1 Different scopal configurations 1.1.1 Different scopal configurations . . . . . . . . . . . . . . . . . . . . . . . . 12 1.1.2 The role of exhaustivity 1.1.2 The role of exhaustivity . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.2 Two discourse relations shared by sondern and aber 1.2 Two discourse relations shared by sondern and aber . . . . . . . . . . . . . . . . . 13 1.3 Guide 1.3 Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2 The syntax of German correction 2 The syntax of German correction 16 2.1 Important generalizations, literature about correction 2.1 Important generalizations, literature about correction . . . . . . . . . . . . . . . . 20 2.2 German specific considerations 2.2 German specific considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.2.1 Verb second word order 2.2.1 Verb second word order . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.2.2 Scrambling and the syntax of topic and focus 2.2.2 Scrambling and the syntax of topic and focus . . . . . . . . . . . . . . . . 23 2.2.3 Different types of negation and their syntax 2.2.3 Different types of negation and their syntax . . . . . . . . . . . . . . . . . 24 2.2.4 Phonological omissions 2.2.4 Phonological omissions . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.3 Sondern/aber in single corrections 2.3 Sondern/aber in single corrections . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.3.1 Subcategorization evidence 2.3.1 Subcategorization evidence . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.3.2 Deletion evidence 2.3.2 Deletion evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 2.3.3 Islands 2.3.3 Islands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 2.3.4 Agreement 2.3.4 Agreement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 2.3.5 Scope 2.3.5 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 2.3.6 Adverbial scope 2.3.6 Adverbial scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 2.3.7 Locating negation 2.3.7 Locating negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 2.3.8 Interim conclusion 2.3.8 Interim conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 2.3.9 Consequences of this proposal for negation 2.3.9 Consequences of this proposal for negation . . . . . . . . . . . . . . . . . 68 2.4 Conclusion 2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 2.4.1 What is the specific syntactic structure for simple correction? 2.4.1 What is the specific syntactic structure for simple correction? . . . . . . . . 70 2.4.2 Subcategorization tests: simple correction, local negation 2.4.2 Subcategorization tests: simple correction, local negation . . . . . . . . . . 70 2 Contents 3 2.4.3 Subcategorization tests: simple correction, distal negation 2.4.3 Subcategorization tests: simple correction, distal negation . . . . . . . . . 73 3 Semantics and pragmatics ofsondern 3 Semantics and pragmatics ofsondern 77 3.1 Introduction 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 3.1.1 Observation 1: C1 contradicts C2 3.1.1 Observation 1: C1 contradicts C2 . . . . . . . . . . . . . . . . . . . . . . 78 3.1.2 Observation 2: Conjunct one shows polarity discourse contrast 3.1.2 Observation 2: Conjunct one shows polarity discourse contrast . . . . . . . 79 3.1.3 Observation 3: Corrections with sondern do not trigger scalar implicatures 3.1.3 Observation 3: Corrections with sondern do not trigger scalar implicatures 82 3.2 Rooth (1992) essentials 3.2 Rooth (1992) essentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 3.2.1 Regular semantic values and focus semantic values 3.2.1 Regular semantic values and focus semantic values . . . . . . . . . . . . . 84 3.2.2 The operator and the context pronoun 3.2.2 The operator and the context pronoun . . . . . . . . . . . . . . . . . . . 85 3.2.3 Discussion of discourse coherence in Rooth (1992) 3.2.3 Discussion of discourse coherence in Rooth (1992) . . . . . . . . . . . . . 87 3.3 Contrast 3.3 Contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 3.4 Polarity focus in discourse 3.4 Polarity focus in discourse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 3.5 Interpretive components formalized 3.5 Interpretive components formalized . . . . . . . . . . . . . . . . . . . . . . . . . 93 3.5.1 Exhaustivity in the scalar interpretation of corrections 3.5.1 Exhaustivity in the scalar interpretation of corrections . . . . . . . . . . . 94 3.5.2 Hurford’s constraint and exhaustivity in coordinations 3.5.2 Hurford’s constraint and exhaustivity in coordinations . . . . . . . . . . . 95 3.5.3 Conceivable Hypotheses about Exhaustivity 3.5.3 Conceivable Hypotheses about Exhaustivity . . . . . . . . . . . . . . . . . 98 3.5.4 Interim Conclusion 3.5.4 Interim Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 3.6 Discourse relation 3.6 Discourse relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 3.6.1 Background: verum focus 3.6.1 Background: verum focus . . . . . . . . . . . . . . . . . . . . . . . . . . 103 3.6.2 Possible analyses for corrections 3.6.2 Possible analyses for corrections . . . . . . . . . . . . . . . . . . . . . . . 105 3.6.3 Focused negation 3.6.3 Focused negation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 3.6.4 Verum Focus operator 3.6.4 Verum Focus operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 3.7 The contrast condition between two conjuncts 3.7 The contrast condition between two conjuncts . . . . . . . . . . . . . . . . . . . . 112 3.8 Conclusion 3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 4 Semantics and pragmatics ofaber 4 Semantics and pragmatics ofaber 119 4.1 The puzzle 4.1 The puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 4.2 Data 4.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 4.2.1 Aber can occur in two constructions, sondern cannot 4.2.1 Aber can occur in two constructions, sondern cannot . . . . . . . . . . . . 121 4.2.2 Observation 1: contradiction between contrasts prohibited 4.2.2 Observation 1: contradiction between contrasts prohibited . . . . . . . . . 122 4.2.3 Observation 2: simple scalar contrast requires negation 4.2.3 Observation 2: simple scalar contrast requires negation . . . . . . . . . . . 123 4.2.4 Comparison with sondern 4.2.4 Comparison with sondern . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4.3 Proposal 4.3 Proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 4.3.1 Aber encodes likelihood 4.3.1 Aber encodes likelihood . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 4.3.2 The scalar examples 4.3.2 The scalar examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 4.3.3 Proposal: aber requires exhaustification 4.3.3 Proposal: aber requires exhaustification . . . . . . . . . . . . . . . . . . . 138 3 4 Contents 4.3.4 [even [negation [ scalar exhaustivity ] ] ] as an NPI configuration 4.3.4 [even [negation [ scalar exhaustivity ] ] ] as an NPI configuration . . . . . 139 4.3.5 Negation is obligatory 4.3.5 Negation is obligatory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 4.4 Summarizing 4.4 Summarizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 5 Conclusion 5 Conclusion 148 5.1 Open questions 5.1 Open questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 Bibliography Bibliography 151 4 Acknowledgements One of the hardest things about writing a dissertation is that you start this task before knowing where the journey will take you, but at the same time you must trust you arrive where you were meant to go. That means that you have to be comfortable with traveling. There were many people in my life who played a part in making it possible for me to find this way, by giving me directions, by helping me find strength for the way both physically or mentally. My advisor Roumyana Pancheva knew the way long before I did. I am indefinitely grateful to her for tirelessly giving me direction, but also giving me space when I needed it, and encouraging me not to give up. My co–advisor Khalil Iskarous, while maybe not knowing the exact path I would take, was determined that each of the paths I could possibly take would lead me to a good place. I am grateful to him for being excited about this project, even when I sometimes couldn’t be, and giving me space to explore the beginning of many different paths. I am grateful to Andrew Simpson for always having an open ear, giving me advice when I was asking for it, and quite literally preparing me for walking long ways. Thanks to Kevin Knight for serving as my external committee member, and letting me audit his Computational Linguistics class, which in retrospect proved very important. I am also grateful to all other members of the USC faculty, especially Maria Luisa Zubizaretta for being interested and always having an open ear, and providing me new directions, Hajime Hoji for his enthusiasm in philosophy of science, and forcing me to take a wider perspective, and Louis Goldstein for the perspective on life and food he provided. Thanks also to the countless people who were willing to discuss the current stage of my work with me, and share their ideas, intuitions and opinions. They include but are not limited to David Beaver, Daniel B¨ uring, Jesse Harris, Katharina Hartmann, Kyle Johnson, Friedrich Neubarth, Jason Merchant, Alan Munn, Winfried Lechner, Michael Wagner, David Pesetsky, Benjamin Spector and Anna Szabolsci. I want to thank my cohort, Huilin Fang, Alfredo Garcia–Pardo, Peter Guekguezian, Dasha Henderer and Catlin Smith, for walking alongside me, and showing me ways to manage this project, and always being willing to help and support me. I hope I could do the same for you. 5 6 Contents Then I want to mention the many other great friends I made at the linguistics department. Many of them helped me in completing this dissertation with academic guidance. All of them helped me to make my life a bit better, and helping me develop mental and often physical strength (in the form of dinners mostly), and diversions and support when needed. , Priyanka Biswas, Mary Byram-Washburn, Monica Do, Bet¨ ul Erbas ¸ı, Ana Besserman, Bhamati Dash, Emily Fedele, Samantha Gordon–Danner, Jessica Harmon, Brian Hsu, Canan Ipek, Cynthia Lee, Mairym Llorens, Mythili Menon, Ngo Nha Binh, Sarah Ouwayda, Iris Ouyang, Daniel Plesniak, Joshua Segui, Alif Silpachai, Saurov Syed, Barbara Tomaszewicz, Yoona Yee, Hailey Wei, Chia–fen Wu, and Xiao He. I want to thank everybody at the Normandie House for making sure that I didn’t just have a house, but a home from the first day I moved to Los Angeles. Mehmet Aykol,, Matilde Cordeiro, Brad Foley, Tugba Koker, Eliseo Melero, Wendy Vu, Amanda Wagenbach and also Chad Geib, Barret Philips, Sarah Signor, I could have never found the determination to complete this project without sharing meals with you, your support, the food, (sometimes pointless) discussions, and perspective. Other sources of energy and stability in my life were all the people and workouts at the Los Angeles Rowing Club, specifically the members of the boats captained by Susan Casey, Greg Holwitt, Diana Allen, Lily Lee, Shelley Pearce, and Marygrace Coquia. I’d like to thank my parents Hilda and Franz, and my siblings Petra, Christoph and Hannes, and all other members of my family for supporting me though this task, and raising me in a way that allowed me to face this challenge and come out on the other side. The person replaced with in almost all of the lists above is my partner Thomas Borer. The ways he contributed to me being able to compete this work are so numerous and diverse I have no way of naming them all. They range from just being there and enlightening me with his positivity to being the first point of discussion for linguistic ideas. I love you so very much and I am grateful that you are willing to share your life with me. This thesis is dedicated to the memory of my friend Wolfgang ‘Bruno’ Rieder. I want to thank him for being by my side in my adolescent life, and being part of so many good memories. Thank you, Bruno, for your perspective and unwavering support, for your opinions and sharing your teenage music choices with me. 6 Chapter 1 Introduction When linguists think of utterances, often they think of them as being prototypically assertions; utterances used to state a speaker’s view about a situation. How assertions are related to each other often draws less attention. Assertions are only one aspect of speech acts: they represent a single speaker’s opinion. Merely uttering an assertion does not mean the speaker’s view is accurately received and agreed upon by the listener. An assertion is only one step in construing a consensus in a conversation. This dissertation focuses on one of the other steps, a speech act called correction. Correction is a type of coordination found in a discourse situation in which a previous assertion is drawn into question. An example of correction is given with the sentence (1B 1B) in in the sample discourse (1 1) below. It consists of three major parts, the antecedent, the target and the consequent. The correction refers back to a previous utterance or an assumption implicit in a discourse, the antecedent. In the case at hand, this is the assertion She went to the same school as I. The corrective utterance in (1B 1B) then repeats and negates the antecedent explicitly with the phrase she didn’t go to USC, the target of correction. Finally, a new assertion, the consequent is made. It is the communicative goal of correction that the speaker asks for the consequent to be accepted as an agreed upon proposition. Both the target and the antecedent entail or imply a contradiction with the antecedent utterance. These three parts of correction express some overlapping information (for example, all three are about Sandra, the speaker A, and the school they went to), and they identify which parts of the information should be considered non–overlapping. (1) A. i. The other day I ran into Sandra. ii. I hadn’t seen her in two years. iii. You know she went to the same school as I. B. She didn’t go to USC, but to UCLA. In this discourse, a number of assertions are made. Most of these assertions are false upon close 7 8 reflection. For example, two years is likely not the duration of the time since the speaker has last seen Sandra. This demonstrates that being the antecedent of a correction is not related to objective truth or falsity of said sentence. In this discourse, it is the assertion that Sandra went to the same school as the speaker prompts the other discourse participant to utter the correction in (1B 1B). The speaker of the correction may have chosen this utterance since this is where they want to lead the conversation, for example. While a discourse like (1 1) may be a very prototypical way of expressing a correction, it is by no means the only one. Comparing multiple ways of expressing corrections with one another, it is found that an objection to a previous assertion or entailment thereof can be made with as little phonologically realized information as the consequent. The mini–discourses in (2 2) show that objections to assertions or propositions which are discour- sively established may be expressed without phonological realization of the antecedent or the target. In the sentence (2a 2a), several reasoning steps are necessary to deduce from the statement I know Sandra from school which school she went to. In fact, the same is true for the original example (1B 1B), there is no requirement for any type of phonological identity between the antecedent and the target, in fact, very often the antecedent is not phonologically realized. Nonetheless, (2a-B 2a-B) is an appropriate response. Likewise in (2b 2b), it is not necessary that the target is spelled out and negated: this discourse shows that the relation between the antecedent, the target and the consequent can be established in different ways, for example via pitch accent and focus in the example. (2) a. A. I know Sandra from school. B. Sandra didn’t go to USC, but to UCLA. b. A. Sandra went to USC. B. She went to UCLA. 1 1 In this work, correction is presented as a specific relation between the antecedent, the target and the consequent. The target and the antecedent are in a specific relation. This relation is argued to be a relation of presupposition, which represents the fact that the antecedent may be present only implicitly. Another relation is formed between the target and the antecedent. This relation is argued to be a result of the semantics of the operator but. These two parts of correction must be in a contrastive relation. The relation between the antecedent and the consequent follows from the definition of the former two relations. Taking a wider perspective, this means that correction is part of a class of utterances used to mark when an utterance is no longer agreed upon as an optimal description of a given situation. It is also a type of discourse reference. This work is specifically interest in answering whether there are restrictions on the linguistic means which the construction of consensus and discourse reference 1 Small caps mark a pitch accent. 8 Chapter 1. Introduction 9 uses, specifically given the large degree of variation in expressing correction. Is there a uniformity to the way in which utterances which are no longer the optimal descriptions of a situation must be linguistically marked? Must the be linguistic marking when two utterances are related and condition each other in discourse and if so, what is it? And most importantly, for each of these questions, are there common traits across languages, which have the capacity to inform us about the human language faculty in general. There are four reoccurring linguistic means instrumental in corrections. They are repetition on both a semantic and a phonological level, negation or an expression downward entailingness (since the antecedent was repeated and negated), prosodic focus, and a coordinator like but. These means can be combined in various ways to express the correction of an antecedent statement. For example, the discourse in (3a 3a) uses all means available: repetition, prosodic marking, negation and the coordinator but. The example (3b 3b) uses only the means of repetition and prosody, whereas the example (3c 3c) employs repetition, prosody and negation. (3) a. A: Erica ate a pear. B: Erica didn’t eat a PEAR, but a FIG. b. A: Erica ate a pear. B: No. Erica ate a FIG. c. A: Erica are a pear. B: Erica ate a FIG, not a PEAR. Each of these grammatical means can be used independently in sentences which do not express correction. This is demonstrated in the examples in (4 4). In these examples it is shown that negation can be used in a way unrelated to a discourse antecedent, repetition and the resulting parallelism are also used in questions, pitch accents may also be used in other types of coordinations, and but may also be used in sentences where no antecedent statement is drawn into question. (4) a. It didn’t rain today. b. What did Erica eat? — Erica ate a fig. c. Erica ate FIGS, PEARS, and APPLES. d. Erica ate FIGS, but she also ate APPLES. The current work investigates the interplay between all of these functions, but specifically focuses on the syntax and semantics of coordinators like but. It focuses on corrections which have the complete set of linguistic means, that is, repetition, negation or other expressions of downward entailingness, prosodic stress or focus, and finally, use a corrective coordinator. It also focuses on corrections where all parts of correction (antecedent, target, consequent) are phonologically realized. In cases where parts of corrections are phonologically omitted, this work strives to give an explanation about the mechanism in which the omission happens. It is the objective of this work to create insight into structures which have all these semantic and phonological parts, but also, to distinguish it from its paradigmatic variants. 9 10 A specific interest in corrective coordinators is due to the fact that across languages, but is used in remarkably parallel ways. In many languages, English and French among them, a single lexical item, but or mais ‘but’ can take over the same two functions. Examples for these functions are given in the following sentences, they will be called corrective and counterexpectational. Example (5 5) shows corrective sentences, whereas example (6 6) shows counterexpectational sentences. (These observations were originally made in Anscombre and Ducrot Anscombre and Ducrot, 1977 1977, the terminology due to Vicente Vicente, 2010 2010) (5) a. Sandra didn’t eat FIGS, but PEARS correction b. Sandra Sandra n’a not-has pas not mang´ e eaten des pl.indef figues, figs, mais but des pl.indef poires. pears. (6) a. It rained but Sandra went to the park counterexpectational b. Il It a has plu, rained, mais but Sandra Sandra est is all´ ee gone au to-the parc. park. The fact that two unrelated languages have the same distinct functions for the same grammatical items suggests a grammatical base for these parallels. However, it is not a general trait of but to allow for these two functions. Other languages, like Spanish, use different coordinators in these two discourse types. In the case of corrections, they use sino ‘but’, whereas with counterexpectationals, they use pero ‘but’ (Vicente Vicente, 2010 2010). This suggests that the two readings are grammatically distinct. German also uses two different coordinators, sondern and aber. However, in German the distinction between these readings is less clear. In the counterexpectational example (7 7), only the coordinator aber is available, whereas in the corrective (8 8), both sondern and aber are possible. (7) Es It hat has geregnet, rained, aber/*sondern but/but Sandra Sandra war was im in-the Park park. (8) Sandra Sandra hat has nicht not Feigen, figs, sondern/aber but/but Birnen pears gegessen. eaten. These facts are briefly summarized with the following table. (9) correction counterexpectational sondern 3 7 aber 3 3 At first glance, these facts alone may not seem like a large problem for distinguishing between corrections and counterexpectationals and defining their relation. One may be tempted to analyze them as aber being the equivalent of but or mais, and sondern fulfilling a subset of the functions of aber. A closer look at the data shows that this is not an option. The two coordinators differ crucially in their scalar functions (these facts are initially discovered by Anscombre and Ducrot Anscombre and Ducrot, 10 Chapter 1. Introduction 11 1977 1977). Moreover, there are situations where only sondern is grammatical, and others where only aber may occur. The two coordinators are not interchangeable. These facts are presented in examples (10 10) through (12 12) below. The examples in (10 10) show that aber may not occur in situations where the second conjunct entails the target of correction without negation. Sondern shows no such restrictions. These facts alone show that sondern cannot be a subcase of aber, since sondern can occur in a wider range of situations. (10) a. Sandra ate not two sondern/*aber three figs. 2 3 b. Sandra ate not three sondern/aber two figs. 3! 2 Additionally, independent of the coordinator, for both sondern and aber, an expression of downward entailment is obligatory in correction. This is shown in example in (11 11). Independent of the coordinator, a downward entailing expression must be present in correction. (11) Sandra Sandra hat has *(nicht) not Feigen, figs, sondern/aber but/but Birnen pears gegessen. eaten. On the other hand, aber may also occur in other non–corrective situations where a downward entailing expression is not required. Under the assumption that sondern fulfills a subset of the functions of aber, this fact remains mysterious. It raises the question what it is in correction specifically that makes an expression of downward entailingness obligatory. Furthermore, there are certain subtle semantic–pragmatic differences between the two coordi- nators. The example in (12 12) demonstrates that only sondern allows for the two conjuncts to be contradictions with respect to each other. (12) Diese This Couch couch ist is nicht not bequem, comfortable, sondern/*aber but/but unbequem. uncomfortable. Given these subtle differences, a simple explanation, that is, labeling sondern corrective, and aber corrective and counterexpectational, does not account for the differences that these two coordinators show. The functions of correction and counterexpectational are not exclusively anchored to a lexical item, but rather a lexical item in a context licensing the respective readings. Defining these contexts is a main goal of the current work. In light of these data and considerations, this dissertation sets out to answer these question com- plexes: 1. How to derive a formal account of the difference between aber and sondern? a) How do the linguistic domains involved, that is syntax, semantics and pragmatics specifi- cally interact? 11 12 1.1. Formalizing the difference between sondern and aber b) What are the discourse conditions of correction and how do the the components in syntax, semantics and pragmatics interact to express the discourse semantics of correction? c) Which elements are obligatory? Of the linguistic means mentioned so far, how can they be expressed and how can variability be allowed for? 2. How to formalize the discourse requirements of correction? How can the relation between the target and the antecedent be defined? 3. Explain the need for downward entailingness in correction: Since an expression of downward entailingness is not obligatory for all types of denial, how do the syntactic operators interact to force an expression of downward entailingness in correction? In the remainder of the introduction, I will give a short overview of the answers to these questions given in this work. 1.1 Formalizing the difference betweensondern andaber 1.1.1 Different scopal configurations The characteristic semantics of sondern and aber are due to a complex interaction of different focus operators. Sondern and aber are each a focus sensitive operator, and they interact with negation or an expression of downward entailingness and different types of exhaustivity operators. As a consequence, the two types of correction, (correction with aber and sondern) allow for derivations with distinct scalar presuppositions: in corrections with sondern, the scopal configurations are less rigid, and more orderings of the relevant operators are available. The possible scopal configurations are given in (2.3.5 2.3.5). (13) sondern a. sondern (exhaustivity (negation)) b. sondern (negation (exhaustivity)) aber aber (negation (exhaustivity)) This is due to the fact that aber is has a presupposition about likelihood, and therefore it resembles the operator even. An example to demonstrate the intuitive parallelism between but and even is given in (14 14). These two operators share the property that among their contextual alternatives, there are other alternatives which are more likely than the alternative phonologically realized. 12 Chapter 1. Introduction 13 (14) a. Sandra even went to the park. ! Other alternatives were more likely than Sandra going to the park. b. It rained, but Sandra went to the park. ! Other alternatives where more likely than it raining and Sandra going to the park. The presuppositions for this operator in a scalar configuration can only be fulfilled when it occurs in a certain scopal configuration with respect to two of the other operators that can be involved, that is, a downward entailing operator and an operator of exhaustivity. 1.1.2 The role of exhaustivity Exhaustivity is a crucial part in the derivation of both the lexical entry of sondern and aber. Both of them require exhaustification as part of their conjuncts. In the case of non–scalar corrections, this operator is indistinguishable from the interpretation of prosodic focus. This dissertation ascribes to the view of exhaustification as a covert syntactic operator, which is optional in many contexts, but obligatory in order to derive corrective readings. It is argued that exhaustification is necessary for both corrections with sondern and aber. It is proposed that different exhaustivity operators are proposed in corrections with sondern and aber. In corrections with sondern, the process of optional embedded exhaustification proposed in Chierchia Chierchia (2004 2004) and related literature is assumed. In the case of aber, it is assumed that an exhaustification operator which introduces a necessary ordering in the set of alternatives is at work. This is the operator SOLO defined in Crniˇ c Crniˇ c (2011 2011). This operator also plays a role in licensing Negative Polarity Items, and a similar interaction is found in the case of aber. The fact that aber can occur in counterexpectational and correction is not an effect of lexical polysemy, but rather it is explained as a focus effect. Specifically, when aber occurs in corrections, on account of aber’s similarity with even, its presuppositions are incompatible with the presup- positions of a proposed exhaustivity operator, which in turn requires an expression of downward entailingness to occur in such sentences. A similar analysis was proposed for even in Crniˇ c Crniˇ c (2011 2011). This means, when there is particularly narrow scope in each of the conjuncts of aber, negation will be required. 1.2 Two discourse relations shared bysondern andaber The core semantics of correction are encoded in presuppositions of focus operators that are common to both sondern and aber when used in correction. On the one hand, the expressions below the scope of negation in one of the conjuncts must be mutually exclusive in the given situation. That 13 14 1.2. Two discourse relations shared by sondern and aber means in a correction Sandra didn’t eat an apple but a pear, it must be excluded that Sandra ate both of these fruits. This restriction is modeled with the notion of contrast proposed in Rooth Rooth (1992 1992), for both coordinators it is necessary that the focus alternatives of C1 are a subset of C2 and vice versa. In addition, both ways of expressing correction share a covert polarity focus operator, which is called CORR in the discussion here. This operator resembles a verum focus operator, in that it presupposes a certain relation of the correction target with the discourse. By virtue of this operator, the relation between the target and the antecedent is defined. Consider the example in (15 15) to give an intuition about the communalities between correction and verum focus. They share that they both refer to the common ground and correct the polarity of an assertion contained in it. They differ with respect to the polarity of the assertion: while verum focus refers back to an utterance with negative polarity, CORR refers to an assertion which necessarily has positive polarity. (15) a. i. Sandra isn’t writing a novel. The speaker believes that Sandra isn’t writing a novel ii. Sandra IS writing a novel. Counter to the speaker’s belief, Sandra is writing a novel. b. i. Sandra is writing a novel. The speaker believes that Sandra is writing a novel. ii. Sandra isn’t writing a NOVEL, but a collection of poems. Counter to the speaker’s belief, Sandra is not writing a novel. She is writing a collection of poems. This proposal is sketched in the tree in (16 16). The two coordinators differ in the fact that sondern requires a relation between negation and CORR at all times, whereas aber only has polar focus in corrections. The relationship between the target sentence and the antecedent sentence is described as a presupposition, where the target sentence presupposes that the nonnegated version of the target sentence, C1 in the tree in (16 16), is part of the common ground. (16) sondern/aber C2 CORR C1 neg a. CORR presupposes that C1 is part of the common ground. b. JC1K F JC2K F , JC2K F JC1K F 14 Chapter 1. Introduction 15 Another observation is that in corrections, there two conjuncts stand in a complex relation. Considering a sentence like (17 17) below, it is manifest that there is a notion of contrast in this sentence, since it is not possible to utter such a sentence in a situation where Sandra could have eaten both an apple and a pear. (17) Sandra didn’t eat an apple, but she ate a pear. To account for these facts, the present work proposes that there is a necessary evaluation of focus lower than negation in the first conjunct, and at the highest syntactic level in the second conjunct. In addition, the focus semantic alternatives at this evaluation must be subsets of each other. In the previous discussion, the question about the relation between discourse semantics and grammar is brought up. In the view of the current work, there are certain minimal discourse operators which position an assertion against the utterance background. In corrections, these operators are arranged in a way that their semantics relies on each other’s output. 1.3 Guide Chapter 2 2 discusses the syntax of corrective coordinations. Being coordinations, there is large syntactic variety in the types of phrases that can occur in correction. Furthermore, due to the flexible word order in German, a number of transformations are possible, rendering the underlying syntactic structure rather opaque. It is argued that sentences with sondern and aber have different syntactic restrictions, with the negation scope in sondern being necessarily lower than the scope of aber. It is argued that polar negation is derived by a syntactic process of syntactic agreement within the negated conjunct of correction. Chapter 3 3 focuses on the semantics and pragmatics of sondern. It discusses its interaction with focus, and derives the mutual exclusivity restriction in a formal way. It also introduces two ways of deriving scalar corrections, allowing for both of the readings found in scalar corrections with sondern. Furthermore, it introduces the polarity operator corr used in defining the discourse relation corrections must occur in. Chapter 4 4 proposes that aber is a focus associative coordinator which involves a likelihood presuppositon, thereby resembling the coordinator even. This resemblance makes a number of additional predictions regarding the relationship of aber with negation. It is shown that the parallelism between aber and even also extends into this domain, thereby explaining the optional presence of a downward entailing operator in sentences with aber. This suggests a view where counterexpectational and corrective coordinations are not part of a lexical ambiguity, but rather a result of different scopal configurations. Chapter 5 5 concludes the dissertation, and proposes future avenues of research in the form of open questions. 15 Chapter 2 The syntax of German correction This chapter investigates the syntactic representation of correction structures in German. There is considerable variability in the targets of correction: correction can apply different syntactic levels like DPs, VPs, TPs, NPs, but at the same time it the semantic target or correction need not be identical to these syntactic levels. In terms of their interpretation, they can also correct pragmatic inferences or pronunciation (Geurts Geurts, 1998 1998), that is, language components the syntactic level of which is not obvious. Furthermore, they can be realized with two coordinators that have similar but not identical semantics, sondern and aber. Given this variability, a crucial question is whether this variability in other domains than syntax is also mirrored in their syntactic representations. Correction is a type of coordination that has a number of grammatical features distinguishing it from more widely studied types of coordinations like those with and and or. German is a particularly interesting and challenging language to consider, since it has a more complex paradigm than other languages, which is due to the different lexical expressions of corrective coordinators. It will be shown that some of these variabilities are caused by the particular features of German syntax, like Verb Second word order, others are due to more general syntactic restrictions. Yet another set arises as an interaction with the lexical and discourse semantics of correction. It will be proposed in this chapter that there is a direct mapping between the discourse function of a coordinator like aber, which can serve two distinct discourse functions, and its syntactic representation. It has been pointed out in Anscombre and Ducrot Anscombre and Ducrot (1977 1977); Vicente Vicente (2010 2010) and much subsequent literature that coordinators like but, which convey a notion of adversativity, have two essential functions. These functions have been called correction and counterexpectational in the literature. Consider the examples in below. 16 Chapter 2. The syntax of German correction 17 (18) a. Anna didn’t eat an apple, but a fig. correction b. There was a sound from the telephone, but Anna slept well. counterexpectational There are two essential grammatical differences between these two constructions. On the one hand, negation is obligatorily present in a correction sentence like (18a 18a), but not in the counterexpectational sentence, (18b 18b). On the other hand, the conjuncts in (18a 18a) share much more syntactic material than the conjuncts in (18b 18b). This is reflected not only in the fact that the second conjunct contains little phonological material in (18a 18a), since there might be an ellipsis process, but also in the fact that (18b 18b) conjoins two essentially different sentence types, an expletive construction and an intransitive sentence, which ascertains that it is possible that the two conjuncts conjuncts share no material at all. Now consider how these two types of adversative sentences are realized in German in example (19 19). In some ways, the German sentences are subject to similar restrictions as their English counterparts: (19a 19a) requires negation to be present, whereas (19b 19b) does not have the same restriction. While in (19a 19a), much of the sentential material is shared between the conjuncts, the same is not true in (19b 19b). A crucial feature of the grammar of German, that will constitute a main question in the discussion of the syntax of these constructions is that a sentence synonymous to English (18a 18a) can be expressed with two different coordinators, sondern, and aber. In a traditional view of the separation between semantics and pragmatics, these sentences have the same truth conditions, since they differ in their pragmatic inferences. In chapters 3 3 and 4 4, I will propose that there are crucial differences in the discourse semantics of these two coordinators which are due to a lexical specification of their pragmatic restrictions. Only aber ‘but’ can occur in a counterexpectational sentence. Both coordinators, sondern ‘but’ and aber ‘but’ require negation in the correction case. This means that in certain environments, aber requires negation, while sondern requires negation in all contexts. It can be seen that a core property of correction itself is that it is dependent on negation. In this dissertation it is argued that this dependence is not part of a lexical subcategorization of the coordinator aber, but rather follows from the pragmatics of the construction. This question is discussed in depth in chapter 4 4. (19) a. Anna Anna hat has *(nicht) not einen an Apfel, apple, sondern/aber but SONDERN /but ABER eine eaten. Feige gegessen. Anna didn’t eat an apple, but a fig. b. Es It kam came ein a Ger¨ ausch noise vom from-the Telefon, telephone, *sondern/aber but SONDERN /but ABER Anna Anna hat has gut well geschlafen. slept. There was a noise from the telephone, but Anna slept well. 17 18 In this chapter, the focus will be on investigating whether there are syntactic differences between corrections with sondern and corrections with aber. The questions investigated will be whether sentences like (19a 19a) and (19b 19b) are indeed as syntactically related as they look, and whether there are any syntactic differences between the two. Specifically, it will be proposed that sondern and aber have different restrictions on coordination size and the location of negation within their conjunct. While sondern requires its second conjunct to match the semantic scope of negation, aber does not have such a requirement. As a consequence, sondern is more restrictive with respect to how much other material can occur in corrections. I will also argue that both sondern and aber have few categorial restrictions, they can coordinate conjuncts of many different sizes. It will be suggested, although somewhat briefly, that there is a relation between the location of negation in a sentence and its discoursive interpretation. The default location is the locus of sentential negation, a position adjacent to the verb phrase which is marked by the projection NegP in many analyses. When negation is realized in a location away from its default position in NegP, it is discoursively marked. This captures the observation that negation with rather narrow scope have more perceivable discourse presuppositions, while sentential negation does not have such an obligatory presupposition. To illustrate the hypotheses considered, a number of tree schematics are given in table 2.1 2.1 on page 18 18. In this chapter, it will be evaluated whether or not there are any syntactic restrictions which Table 2.1: Hypotheses about the syntactic representation of corrections Low Coordination Hypothesis sondern aber XP Y F sondern X F not shared material XP Y F aber X F not shared material High Coordination Hypothesis XP shared material, Y F sondern shared material X F not XP shared material, Y F aber shared material X F not confine either sondern or aber to be only a high or a low coordination. It will be found that there are no such restrictions. There is, however, a coordination level that sondern cannot exceed. In corrective sentences like (18a 18a) and (19a 19a), the information that is being corrected will be called the target. In (18a 18a), this is the DP an apple. I will call the information by which it is replaced in 18 Chapter 2. The syntax of German correction 19 the correction the consequent. In (18b 18b), this is the DP a fig. The notion of target and consequent are terms to describe the type of information which is minimally different. This is distinguished from notions like ‘the left conjunct’ and ‘the right conjunct’, which refer to the syntactic phrases coordinated, and may involve larger constituents than just the target. A property of German that is relevant for other languages is which disambiguation mecha- nism is employed to distinguish between correction and counterexpectational readings of aber. Since aber can be used in two discoursive functions, the distinction between corrective aber and counterexpectational aber is expected be due to either polysemy, or a type of structural ambiguity (syntactic, discoursive). This makes a disambiguation process necessary. In the current disserta- tion, evidence supporting a structural ambiguity view of the difference between correction and counterexpectational reading is presented. Such an structural ambiguity might also be found in languages which do not have a lexical distinction between the two discourse functions of correction and counterexpectationals, like English. Investigating the grammar of correction in German can give insight into the discourse features of other languages, like English and many others. Therefore, developing tests determining how the relation between correction and counterexpectationals presents itself will be an important contribution of this analysis. This will provide insight into questions about why corrections have different restrictions across languages. In particular, since the difference between correction and counterexpectationals in other languages is not necessarily lexically marked, as it is in German, it is important to determine how to distinguish between correction/counterexpectationals in the first place. This chapter addresses the question what the syntactic representation of correction in German is, with particular focus on the differences between the two correctors sondern and aber. Clearly this will be give deeper insight into the grammar of German, but also beyond that, it will be also for the grammars of other languages. Two major ways in which corrections and counterexpectationals differed were identified in (18 18). On the one hand, corrections appear to have much more shared ma- terial compared to counterexpectationals. It will be argued that for German, this means specifically that sondern, which only occurs in corrections, and aber when it occurs with obligatory negation, have grammatical restrictions that will cause them to coordinate on different syntactic levels. The second observation is that negation is obligatory in corrections, but not in counterexpec- tationals. It will be proposed in chapter 4 4 that the underlying reason for this restriction is that the negation must license a particular scalar configuration. This is expected to have syntactic correlates. Although this question will not be discussed in a focused manner in this chapter, there is a number of polarity and negation restrictions which manifest themselves in syntax, and a detailed syntactic description will be invaluable in exploring the underlying nature of these restrictions. I will argue that the reason for these observations are explained by making explicit the syntactic 19 20 2.1. Important generalizations, literature about correction differences between the two coordinators. It will be argued that sondern can only coordinate conjuncts that are the same size or smaller than the scope of its negation. Since sentential negation in German can be realized in a syntactic location lower than the subject, single corrections with sondern can have smaller conjuncts. Aber on the other hand can coordinate conjuncts which are larger than the scope of negation. Due to this property, there is no “pressure” to coordinate only small conjuncts, conjuncts of all sizes can be successfully conjoined. In the considerations of these constructions, I will first consider what I call simple corrections, that are corrections where there is a single pair of target and consequent. The evidence considered comes mainly from subcategorization (section 2.3.1 2.3.1), deletion (section 2.3.2 2.3.2), island tests (section 2.3.3 2.3.3), and scope (section 2.3.5 2.3.5). 2.1 Important generalizations, literature about correction Two important generalizations about corrections were introduced in the previous section. On the one hand, negation is obligatory in these constructions. On the other hand, we saw that corrections have a tendency to be small, that is to share much material with the respective other conjunct. An additional feature of the corrective construction is that it is tightly intertwined with focus. (cf. Toosarvandani Toosarvandani, 2013 2013) Consider the paradigm in (20 20) below. For ease of exposition, this example is given in English, but it is replicable in German. (20) a. Sandra didn’t give Susi an APPLE, but A BOOK. b. ?? Sandra didn’t give SUSI an apple, but a BOOK. These data demonstrate that both the correction target and the consequent need to be prosodically focused, and furthermore, these foci need to be matched in syntactic position. Prosodically focusing any constituent other than the target and the consequent results in an utterance that is not well formed. A complete description of correction needs to account for how this restrictions is represented in grammar. This restriction will also be part of the semantic discussion, but in this chapter the focus will be the impact of such a restriction on the syntactic representation. The goal of this chapter to find out about the syntactic properties of and differences between the coordinators sondern and aber, and their lexical specifications. The two differ fundamentally in “selecting” for negation, which sondern does, but aber does not. This chapter is also striving to distinguish the properties specific to sondern, aber and correction in general, from other unrelated properties of German syntax. German is a scrambling V2 language, so interference between these factors and the syntactic representation of correction are expected. The main complexes of syntactic facts that are the objective of this chapter are summarized in the following paragraphs. 20 Chapter 2. The syntax of German correction 21 What is the syntactic size of correction Corrections can target all types of different syntactic components, and even multiple syntactic components. Given this versatility, should the coordination size of correction be considered uniform, or different in corrections that have different phonological form? In multiple corrections, under the assumption that coordination is only possible for constituents, a mechanism of phonological suppression (be it VP ellipsis or deletions with other licensing conditions like Right Node Raising, PF deletions) is obviously active. In analogy, the question whether phonological suppression also applies in simple corrections suggests itself. We can also not assume that corrections with sondern and aber necessarily behave the same, since the reason for different coordination sizes might lie in different syntactic subcategorizations for each of these coordinators, so it is necessary to accomodate the fact that the two coordinators might have different syntactic specifications. Different sentence types after scrambling? Are all types of deletion the same? German has large syntactic variability due to scrambling and V2. For example, a German embedded clause has very different syntactic structure from a main clause, where the verb and another constituent move into the CP. Even in a main clause, it may make an important difference whether a constituent is realized (fronted) in Spec of CP, or in its in situ position. These movements are often associated with reorderings due to the discourse requirements and information structure. In addition, German has frequent reorderings syntactic levels below TP. Interactions between correction and these other syntactic processes are expected. Therefore, it needs to be made explicit whether certain syntactic features are due to independent other phenomena, or corrections specifically. Multiple Focus A particular trait of German is that correction constructions with multiple targets are freely available. This is in contrast to languages like English. It is a goal of this chapter to find out about the underlying reasons for this difference, and explore potential loci of crosslinguistic variation which might explain this difference. In section 2.2 2.2, I will introduce the syntactic representation of German sentences. It is investigated whether there are any immediate selectional restrictions in terms of which phrases can occur in these coordinations, then I investigate whether or not deletions and movement are active in the syntactic derivation of these corrections. I will also consider adverbs as a test of coordination height. These considerations will help account for the observations made above, that the target and consequent of correction need to correspond to each other, and that corrections tend to have small conjunct size compared to counterexpectational sentences. 21 22 2.2. German specific considerations 2.2 German specific considerations Before delving into the investigation of the syntax of the constructions in question, some background regarding the syntax of German is necessary. In particular, I will focus on common assumptions generally made about verb second position, movement processes that are assumed to happen in a German sentence, and assumptions commonly made about the realization and representation of negation, and known findings about the types of deletion. 2.2.1 Verb second word order A particular feature of the syntax of German is that a finite verb always has to occur in the second position of a main clause. In general, this is explained in syntax by assuming that an inflected full verb head moves through T to C. In case of a periphrastic construction, like in perfect tense, it is assumed that the auxiliary moves from its in situ position, T, to C. This leaves us with some degree of expected opacity regarding any sentences where correction occurs in a position after the moved verb, i.e. following the second position in the sentence. Consider a sentence like (21 21) below. In a sentence like this, the syntax is coherent with the correction being a coordination of DPs, VPs, or vPs, even without assuming any types of deletion processes, since either the verb or subject has moved through these phrases. The latter two possibilites require across-the-board movement, but do not rely on ellipsis. (21) Sandra Sandra aß ate nicht not ¨ Apfel, apples, sondern but Birnen. pears. a. Sandra ate i [ VP [ DP not Apples but pears] t i ] b. Sandra ate i [ VP [ VP not Apples t i ] [ VP but pears t i ] ] c. Sandra j ate i [ vP [ vP t j DP not Apples t i ] [ vP but t j pears t i ] ] A periphrastic structure, like the one in (22 22) below, is less opaque in this sense. Without making any additional assumptions, this sentence can only be a coordination of DPs. This is illustrated below. (22) Sandra Sandra hat has nicht not ¨ Apfel, apples, sondern but Birnen pears gegessen. eaten. a. Sandra has i [ TP [ VP [ DP not Apples but pears] eaten ] t i ] b. Sandra has i [ TP [ VP [ VP not Apples eaten ] [ VP but pears eaten ] ] t i ] c. Sandra has i [ TP [ vP [ vP Sandra not Apples eaten ] [ vP but Sandra pears eaten ] ] ] A similar argumentation applies to embedded clauses. For German, it is assumed that in embedded clauses, the auxiliary remains in its position in T, which is also where an inflected verb will move 22 Chapter 2. The syntax of German correction 23 in a clause where tense inflection is expressed synthetically. Due to this, embedded clauses would require ellipsis in all types of coordinations other than DP coordinations. Due to this difference between sentence types, I take special care to consider sentences with a periphrastic verb form, and also embedded clauses. In these sentences, there are limited possibilities for processes without phonological realization, and only ellipsis needs to be ruled out, but across the board movement is not considered an issue in these sentences. 2.2.2 Scrambling and the syntax of topic and focus Another prominent type of movement, and therefore source of opacity, in German is scrambling. Scrambling is often considered a movement that takes place due to reordering to correspond with information structure, which is why scrambling is expected to be very intertwined with corrections, which have specific information structure requirements. There are two types of movement need to be distinguished. On the one hand, there is movement from in situ positions to Spec of CP, so–called “long” scrambling. This movement is obligatory and generally regarded as A’ movement (Grewendorf and Sabel Grewendorf and Sabel, 1999 1999). Some authors believe that the reason for this reordering is to mark the moved constituent as topic (e.g. Frey Frey, 2006 2006). However, there are also reorderings that take place below the C–domain. The following sentences are all well–formed permutations of a sentence. We can see that there are considerable movement possibilities in the part of the sentence dominated by the CP. In general, I will treat scrambling as adjunction to the minimal phrase possible. (23) a. Sandra Sandra hat has Erich Erich gestern yesterday Birnen pears gegeben. given. b. Sandra Sandra hat has gestern yesterday Erich Erich Birnen pears gegeben. given. c. Sandra Sandra hat has Erich Erich Birnen pears gestern yesterday gegeben. given. d. Es It hat has gestern yesterday Sandra Sandra Erich Erich Birnen pears gegeben. given However, there is one notable exception that was discussed in the literature. In several works, Frey and Pittner Frey and Pittner (1998 1998) proposes that there is a specific topic position in the middle field, that is, lower than TP and higher than vP. An important consequence of this assumption is that only certain high adverbs can scope over this topic position (gl¨ ucklicherweise ‘fortunately’, anscheinend ‘apparently’, sicherlich ‘certainly’). This is important since it will be shown that in corrections with sondern, negation cannot scope over these adverbs, indicating a possible interference between the topic movement to the lower position and licensing of correction. 23 24 2.2. German specific considerations These movement possibilities have consequences for the creation of data in diagnosing correction. Possibly interfering topic movement to a position below TP is a factor that needs to be controlled for. As a result, it is important to apply separate tests to detect this movement, and also control for the discourse structure of the data, to control for topic–comment structure in the data. Frey and Pittner Frey and Pittner (1998 1998) argue this type of topic movement to a lower position also occurs in embedded clauses. While I will consider embedded clauses separately in any case, it is important to know that we do not have reason to believe that embedded clauses behave differently from matrix clauses as regards their information structure. 2.2.3 Different types of negation and their syntax German has many expressions of negation, which differ syntactically as well as morphologically. For the syntactic analysis, it needs to be established that the negative adverb nicht, as well as the negative determiner kein(e/r) ‘no’, literally ‘not one’ can both occur in correction, and so can other expressions of negation and downward entailingness like kaum ‘barely’. Since there is no immediately discernible difference in the interpretation of these negations in correction, any hypothesis made about the syntactic structure of correction must also have a way of unifying negations which are clearly realized in different syntactic positions. Syntactically, the location of the negation with kein/e/r ‘no’ is unambiguous. As the morphology (not–one) of the expression makes clear, this is a determiner. Despite its syntactic location, it typically takes sentential scope, and also interacts scopally with modals and other scope taking expressions that are higher than CP. Consequently, kein/e/r is considered a Negative Concord Item by a number of authors (Zeijlstra Zeijlstra, 2004 2004; Penka and Zeijlstra Penka and Zeijlstra, 2005 2005), which has to be licensed by a higher Negation Phrase. These authors suggest that the reason for this behavior is that, just like morphology suggests, kein/e/r ‘no’ actually consists of two parts, a negation operator and an existential, which makes the expression more prone to scopal interaction. The negative adverb nicht appears to occur in various positions in a sentence. Generally, in a discourse neutral sentence, this adverb has to follow the auxiliary in a periphrastic construction, and is typically taken to be dominated by TP (Jacobs Jacobs, 1991 1991) and adjoined to vP. In sentences which are not discourse neutral however, negation can also occur in positions typically referred to as constituent negation, that is, adjoined to other constituents, since nicht ‘not’ is not limited to sentential level scope, and can occur in other positions, namely adjoined to DPs, and other maximal phrases. For example, nicht+XP can occur in Spec of CP, preceding the main verb, which is a common constituency test in German, since it implies constituenthood, since this phrase was moved together. It can also occur as the answer to a question, which also implies that this is a constituent that occurs in Spec of CP, accepting an analysis of fragment answers like Merchant Merchant 24 Chapter 2. The syntax of German correction 25 (2005 2005), who proposes that fragment answers are moved to Spec of CP with subsequent deletion of the complement of CP. (24) a. What happened at the exam yesterday? b. Drei Three Studenten student sind are nicht not zur to-the Pr¨ ufung exam gekommen. come. Three students did not come to the exam. (25) a. How many students came to the exam? b. Nicht Not alle all Studenten students sind are zur to-the Pr¨ ufung exam gekommen. come. Not all students came to the exam. c. Nicht Not alle. all. To illustrate why this creates difficulty in the analysis of correction, consider the sentence in (26 26) below. In this sentence, we know negation has to be lower in scope than sondern because of the interpretation (negation cannot take scope over the second conjunct), but in principle the negation could adjoin to any of the constituents in its scope, that is the DP, but also the vP, with subsequent movement of the subject. Since here, the correction target is the object, there is no independent reason to believe that that negation scopes over the Verb phrase, or adjoins to the verb phrase here. Due to this variability of the location of nicht, I will assume that just like with kein* ‘not–one’, negation is contributed by a negation phrase which takes scope within the first conjunct of corrections, although it need not be realized in this position. (26) Sandra Sandra hat has nicht not Birnen pears gegessen, eaten, sondern but ¨ Apfel. apples. As a consequence, there are corrections which are expressed with nicht in a sentential negation location, but also, there are corrections where nicht is adjacent to the correction target. In the creation of a testing paradigm, both cases will be tested, and furthermore, some space will be dedicated to sorting out the ambiguous cases, for example with the help of adverbials. There is also evidence that nicht, when it does not occur in its sentence level position, has a particular discourse function. In chapter 3 3, I will discuss the nature of this function in a more detailed fashion. For now, hopefully it suffices to say that there is interaction with the phonological locus of negation and certain discourse phenomena. Han and Romero Han and Romero (2004 2004) observe that across many languages, realizing negation high in the sentence introduces a bias for a negative answer in questions. (27) a. Didn’t Chris go to class? biased towards a positive answer b. Did Chris not go to class? not biased 25 26 2.2. German specific considerations (28) Han and Romero Han and Romero (2004 2004, 7) a. Hat has (nicht) Neg Hans Hans (nicht) Neg Maria Maria gesehen? seen Didn’t Hans see Maria? biased towards a positive answer b. Hat Has Hans Hans Maria Maria nicht not gesehen? seen? Did Hans not see Maria? not biased To account for these phenomena, they propose that there is a higher projection in these sentences. When negation is moved into this projection, it allows for certain discourse implicatures. As I will argue in more detail in chapter 3 3, a similar process is the case for corrections, therefore the Negation Phrase that will be assumed in the highest level of the negated conjunct will be argued to have certain discourse semantic presuppositions. 2.2.4 Phonological omissions An issue that I will dedicate more attention to in the course of this chapter is where phonological omissions can occur. The term “omissions” here is used at first as a theory–neutral term for all type of phonology–syntax “mismatches”, be they due to deletion, movement, ellipsis, right node raising, fragmentation, sluicing, and whatever other types of processes there are that suppress the pronunciation of a component realized at LF. In this subsection, tests are executed which allow me to narrow down the deletion processes found in correction. Part of the reason why there is large scale potential for ambiguity in correction is that phonological deletion and their provenience must be made explicit. Considering the examples in (29 29), it can be seen that in the first sentence, an omission from the right conjunct must have been made, whereas in the second sentence, material was omitted from the left conjunct. The omission is obvious since the remaining material are non–constituents, which can otherwise not occur in coordinations. (29) a. Sandra Sandra hat has den the Apfel apple gegessen, eaten, und and Erich Erich die the Birne. pear. b. Sandra Sandra kann, can, und and Erich Erich wird will die the Birne pear essen. eat. It is argued in the literature that rightward deletion, where the antecedent for the deleted constituent is realized to the left, and leftward deletion, where the inverse is the case, is not due to the same process, and that these processes have different licensing conditions and occur in different parts in the syntactic derivations (Hartmann Hartmann, 2000 2000). This alone requires us to conduct all tests separately for omissions in the left or the right conjunct. Furthermore, many times the question whether omissions are due to deletions or are only artifacts of the syntactic structure or movement is 26 Chapter 2. The syntax of German correction 27 not as clear, especially if one considers lower coordinations as a syntactic possibility. Determining the status of correction given all these possibilities will be a major component of the following discussion. As a consequence of these considerations, all tests will be conducted separately for sondern and aber, in a periphrastic verbal tense in matrix and embedded clauses, with negation in sentential as well as non–sentential locations, and with deletions in the left as well as the right conjunct. 2.3 Sondern/aber in single corrections In this section, I will consider various types of evidence in order to establish the coordination height. I will consider subcategorization evidence in subsection 2.3.1 2.3.1, deletion tests such as island violations in subsection (2.3.2 2.3.2), agreement and semantic evidence from quantifier raising and reconstruction. 2.3.1 Subcategorization evidence In this subsection, I am investigating whether there are any categorial restrictions in the subcatego- rization of corrections. Although in the literature, generally correction is treated as coordination, which depending on one’s view do not subcategorize, there are other views of coordination (Munn Munn, 1993 1993) according to which coordination only selects for one of the conjuncts with the other one adjoined, and another view where corrective coordinations are argued to be a two–place quantifier (Ed Keenan, pc.). Another possibility is that corrections are not coordinations but rather subordi- nations (for example, B¨ uring and Hartmann B¨ uring and Hartmann (1998 1998) argue that there are certain cases of apparent coordination which are more justly analyzed as subordination). As a first test, I am investigating which phrases and heads can be superficially coordinated with sondern and aber in corrections, and whether any conclusions can be drawn from these facts. As we will see, the construction is very versatile, and almost all types of phrases can occur in correction, pointing towards a coordination structure without subcategorization restrictions. A factor that has been pointed out in the literature of English correction is that constituent and sentential negation have different properties (cf. Horn Horn, 1985 1985; McCawley McCawley, 1991 1991; Toosarvandani Toosarvandani, 2013 2013). Comparing corrections with constituent negations, like not+DP, with sentential negation like n’t cliticized to an auxiliary, it was found that in the former case, corrections were not subject to island violations, whereas in the latter case, island violations were found. (These findings of course have implications for the discussion of ellipsis, which will be discussed in detail in subsection 2.3.2 2.3.2.) This was taken as evidence that coordinations with the sentential negation n’t involved ellipsis, whereas coordinations with the constituent negation not+DP did not. Consequently, also in German, the location of negation could either be an indicator about or 27 28 2.3. Sondern/aber in single corrections could affect the construction of a sentence, with sentential negation being indicative of a higher coordination height. Given that we are testing a hypothesis that the coordination height for sondern has a more direct relation with negation scope compared to aber, it will be important to have detailed data with both constituent negation and sentential negation. Unlike in English, the lexical choice of the expression of negation in German does not tell us unambiguously whether or not a negation has sentential or constituent scope. Consider the examples of negation in (30 30) below. These sentences show that a negation which is realized as a negative determiner can occur in both sentential and constituent negation. I consider constituent negations these cases in which further explication is necessary, that cannot stand as a statement of their own without making the context explicit, and some inferences of the negated sentence continue to remain valid, for example that Sandra eats something in the given examples. We can see that the negation expressed as a determiner, kein(er) ‘not-a’ is interpreted as a wide scope negation in general, and mostly the position of focus is indicative of whether negation is interpreted as constituent negation or not. In general, it can be observe that nicht is ambiguous between a constituent negation reading and a sentential negation reading. Keine, however, it only sentential, which is why (30a 30a) is acceptable. (30) a. Sandra Sandra isst eats keine no Feigen. figs. b. Sandra Sandra isst eats keine no FEIGEN. figs. Sie She isst eats BIRNEN. pears. c. Sandra Sandra isst eats nicht not FEIGEN. figs. Sie She isst eats BIRNEN. pears. d. Sandra Sandra isst eats Feigen figs nicht. not. The predicting factor with respect to how the negation is interpreted is focus on the one hand, but also where in the sentence the adverbial nicht ‘not’ is realized. Considering the data below for negation expressed in a German main clause, the sentence in (31b 31b) is the one which can most easily be said out of the blue. The other sentence are uttered in contexts where other constituents of the sentence are targets of correction, for example whether or not the money was given as opposed to other contextual alternatives, like for example lent. (31) a. Sandra Sandra hat has nicht nicht dem the M¨ adchen girl zwei two Euro euros gegeben given. (sondern (but dem the Buben)/ boy) (sondern (but drei) three) b. Sandra Sandra hat has dem the M¨ adchen girl nicht not zwei two Euro euros gegeben. given. Natural continuation: . . . but three. 28 Chapter 2. The syntax of German correction 29 c. Sandra Sandra hat has dem the M¨ adchen girl zwei two Euro euros nicht nicht gegeben given Natural continuation: . . . but lent. These data demonstrate that correction can be uttered with the negation realized close to the correction target, and farther removed, in a discoursively neutral position where no additional specification about the context is necessary. When negation is realized away from its neutral position, a discourse bias is registered, for example, if it is realized preceding a verb, the correction is expected to be verbal. Since it is unclear that these different usages of negation have identical properties, an understanding about what role the syntactic location of negation plays in correction. Due to this concern, corrections with different category targets will be tested, both with negation in an adjacent position, and a position that is reconcilable with the negation adjoined to vP, the hypothesized neutral position. In the following, I will call negation that is realized away from the default location of negation and local to the correction target “local” negation, and negation realized at its default location, distant from the correction target “distal” negation. With local negation In this section, I am investigating whether or not there are any subcategorization restrictions in corrections with local negation. Here, it is established how much material needs to be phonologically realized to express corrections, and whether there is a minimum and a maximum boundary. There is an implicative relation between the syntactic coordination level and the choices of correction target and consequent: the target of correction must be expressible within a conjunct of the coordination. As an example, imagine the tempus of a verb is the target of correction. We know that the tempus of the verb is expressed on the auxiliary, which is mostly realized in T or C. As a consequence, the syntactic coordination level must at least be T, and is typically TP. At the same time, I am investigating the hypothesis that the second conjunct in corrections with sondern may not contain any material which is realized higher than negation scope. This will make the prediction that in coordinations with sondern, especially in these low scope negation cases, there are increased restrictions about the nature of the second conjunct, whereas we expect less restrictions from corrections with aber. German also has additional complexity that needs to be taken into consideration when performing syntactic tests: in German verb second clauses, there is obligatory movement of the finite verb to C, and a constituent to Spec of CP. Due to this, low scope coordinations will only be possible when attempting to coordinate subparts of sentences that are lower than CP and where the target or consequent of correction is not moved to Spec of CP. Furthermore, since the verb as well as the auxiliary are commonly assumed to be base generated to the right of the direct and indirect object, generating its complements to their left, which might complicate the analysis or coordination in 29 30 2.3. Sondern/aber in single corrections correction. Due to these potential complications, I looked at three types of sentences. With few exceptions, the tempus was kept constant to be the perfect tense, so the auxiliary would mark C, and the main verb the right edge of the sentence. Type 1 are sentences where the left conjunct is fully realized and any possibly missing material is part of the right conjunct. In these sentences, the second conjunct appears following the main verb. B¨ uring and Hartmann B¨ uring and Hartmann (2001 2001) argue that causal and CP complements of a verb move from a position left of the main verb to a position to the right of the main verb. Consequently, in the example below, (32a 32a) is the base position, and (32b 32b) is the derived position. A similar possibility will be evaluated for correction. Consider these sentences for illustration. Here, it needs to be established whether (33a 33a) is the base order of (33b 33b). (32) a. Kris Kris ist is nicht, not weil because Erika Erika gekommen came ist, gone. gegangen. Kris didn’t leave because Erika came. b. Kris Kris ist is nicht not gegangen, gone weil because Erika Erika gekommen came ist. is. Kris didn’t leave because Erika came. (33) a. Kris Kris ist is nicht not mit with Erika, Erika sondern but Edith Edith gekommen. come. Kris didn’t come with Erika, but with Edith. b. Kris Kris ist is nicht not mit with Erika Erika gekommen, come, sondern but Edith. Edith. Kris didn’t come with Erika, but with Edith. In the second type of sentences, the right conjunct is fully realized and any missing material would be part of the left conjunct. Such sentences are consistent with the suspected base conditions in (32a 32a) and (33a 33a). If there are very small corrections they are expected to be found in these orders. Type three are sentences where the entire corrective coordination appears in the Vorfeld, or is superficially consistent with such an analysis. It will need to be discussed whether these sentences should form a category in their own right, or should be subsumed under the second type, namely sentences where the right conjunct is fully realized. In the sentences I am investigating in this section, the negation is “local” to the coordination, that means, it appears immediately before the target of correction. In the interest of brevity, I will only present sentences with ambiguous status here, the remainder of these sentences can be found in the appendix to this chapter in section 2.4.1 2.4.1 on page 70 70. Specifically, all data with local negation and simple correction are here. In example (34 34) below, examples of the three different sentence types used for the subcategoriza- tion tests are given. While superficially, the examples are consistent with a DP level coordination, this is of course not a necessity. All of these sentences have the same denotation, but might be used 30 Chapter 2. The syntax of German correction 31 in different discourse environments. I will call the coordination type 1 in (34a 34a) full left correction, since if one assumed any deletions, they would have to occur in the right conjunct, and the left conjunct appears fully realized. An example of type 2 is given in (34b 34b). It will be called full right correction: it appears that the right conjunct is fully realized, whereas the left conjunct contains deletions. In the sentence in (34c 34c), both the correction target and the correction consequent occur preceding the auxiliary in verb second position. This will be called Fronted corrections. This is consistent with either the entire correction Nicht den jungen Hund, sonder/aber den Wolf ‘not the young dog but the wolf’ being fronted as a constituent, which would make the examples in (34c 34c) essentially different from the other sentence types. Alternatively, the fronted cases might be a subcase of full right conjuncts, since potentially we are dealing with two coordinated sentences in which the object have been fronted, with a large part of the left conjunct phonologically surpressed. (34) I didn’t see the young dog, but the wolf. a. Ich I habe have nicht not den the jungen young Hund dog gesehen, seen, sondern/aber sondern/aber den the Wolf. wolf. Full left correction b. Ich I habe have nicht not den the jungen young Hund dog sondern/aber sondern/aber den the Wolf wolf gesehen. seen. Full right correction c. Nicht Not den the jungen young Hund, friend sondern/aber sondern/aber den the Wolf wolf habe have ich I gesehen. seen. Fronted correction I summarize the data in the table 2.2 2.2 below. The actual examples are given in (94 94)–(98 98) in the Appendix. In this summary, we can see that generally, all phrasal constituents can be levels of coordination in correction, whereas not all heads can occur in corrections. Lexical heads like N and A cannot be targets of corrections in which the left conjunct is fully realized, and they are only borderline acceptable as targets of corrections where the right conjunct is fully realized. They can also not be fronted. Other functional heads like P and T heads are also not or barely acceptable as targets of these corrections, but C and D heads can occur as targets of corrections with a full left conjunct. C can even be fronted. These facts provide an argument in favor of correction being a coordination, since it inherits the categorical information of its coordinates. Since in German only constituents can occur in sentence initial position, a coordination of two heads is not expected to occur in this position. The sharp contrast between heads and phrases can be taken as evidence that no deletion processes are involved in fronting of corrections, since in case these heads were embedded in a larger, phrasal structure, parts of which might be elided or otherwise suppressed, then a grammatical contrast between heads and phrases would be unexpected. It is arguable that the exceptions to this rule, C 31 32 2.3. Sondern/aber in single corrections and D, being functional categories, cannot occur without a lexical category, even if said category is not phonologically realized. With respect to the phrasal constituents, there are no restrictions as to which phrases can be coordinated. The only notable exception is TP. Consider the example of TP level correction provided in (35 35). The reason for the ungrammaticality of this sentence is simply an artifact of the test. The reason is that negation is in an unnaturally high location in the example given. In the example, a modal is contrasted with an auxiliary, so that the the main verbs would both be infinitive and hence phonologically identical. The reason for this ungrammaticality is nature of the tests: negation is positioned initial to the coordinated constituent, but in the case of TP, negation does not sit naturally at the left edge of TP. On the other hand, when put in its natural position, we cannot clearly say whether TPs or smaller constituents are coordinated. From this, the accurate conclusion to draw is not necessarily that TP level contrast has different grammaticality restrictions, but rather that negation does, since it cannot be adjoined to TP in these cases. These facts will be considered more deeply in subsequent discussion. (35) TP level correction: a. ?? Ich I habe have geh¨ ort, heard, dass that nicht not Erika Erika den the jungen young Hund dog sehen see wird, will, sondern/aber sondern/aber kann. can. b. Ich I habe have geh¨ ort, heard dass that nicht not Erika Erika den the jungen young Hund dog sehen see wird, will, sondern/aber sondern/aber Erika/sie Erika/she den the jungen young Hund dog sehen see kann. can. c. * Nicht Not Erika Erika den the jungen young Hund dog sehen see wird, will, sondern/aber sondern/aber Erika Erika den the jungen young Hund dog sehen see kann can habe have ich I geh¨ ort. heard. Overall, we find that when negation is expressed locally, that is, unambiguously with the target of correction, there are few syntactic restrictions on what can occur in correction. Mostly, we see that there are restrictions on heads, or potentially small phrases like APs or NPs, which these tests do not necessarily distinguish accurately. For the moment, I am disregarding processes like deletion or movement, which will be discussed later, and will make the relationship between the phonological form and the underlying syntactic structure a bit more opaque. Clearly, this test are not sufficient to investigate conclusively the syntactic structure of corrections, but it is a good test of what must at least be part of a correction. We found that the correction targets must at least be phrasal. In these data, no dramatic differences between the two coordinators sondern or aber were encountered. Although there were some cases where one of the coordinators seemed more natural or preferred, no cases were encountered in which sondern was grammatical, whereas aber was not or vice versa. 32 Chapter 2. The syntax of German correction 33 Table 2.2: Grammaticality of corrections with different coordination height and local negation Category Full left Full right Fronted Lexical Head N * ? * Lexical Head A * ? * Functional Head P * ? * Functional Head D X X * Functional Head T ? ? * Functional Head C X ?? X Phrase DP X X X Phrase PP X X X Phrase VP X X X Phrase vP X X X Phrase TP * X * Phrase CP X X X Corrections with distal negation In this section, it will be tested to what extent correction can be expressed when negation is realized away from the target of correction. In particular, any differences found between cases with local negation and distal negation will be interesting to investigate lateron and will help us understand the grammatical restrictions involved in the composition of focus structures and negation. We saw above that negation or at least a downward entailing marker is necessary for correction to be grammatical. In the local negation examples, the negation adverb nicht ‘not’ was realized immediately preceding the target of correction. Since also the hypothesis is entertained that aber exercises looser control over its second conjunct, since the second conjunct can contain material that is not part of the scope of negation of the first conjunct, it is predicted that since the scope of negation here is higher and encompasses the entire proposition, this construction should be less susceptible to differences between sondern and aber. This paradigm consists of constructions where the negation appears in a position in consistent with Spec of vP. In corrections, negation need not, and in some cases cannot be realized adjacent to the target of correction. It was briefly mentioned above that cases with distal negation were found to trigger island effects, which was taken to be evidence for these cases to have larger conjuncts. Parallel to these findings, also in German we suspect that these cases might be more susceptible to ellipsis, since the conjuncts might be larger than they phonologically appear; if the negation actually has a position in Spec of vP, that means we expect the second conjunct to be at least of this size. The crucial cases in this paradigm will be the cases where the targets of corrections are syntacti- cally larger than vP: in these examples, unlike in the local negation cases, the target of correction 33 34 2.3. Sondern/aber in single corrections is outside the overt c–command domain of negation. It will be instructive to see if the relative scope of negation and focus exert any grammatical restrictions onto each other. For the tested sentences, sentences identical to the ones with “proximal” negations sentences were used, with the only difference that negation was kept at a constant location following the auxiliary, that is generally adjoined to vP. All sentences were in a tense with obligatory auxiliaries. The actual examples can be found in the appendix. Like above, I tested full left conjuncts, full right conjuncts, and fronted coordinations. Some of the sentences used for the proximal negation cases did not have to be changed to construct the sentence for distal negation, which stresses the importance of applying independent tests for ellipsis, since these sentences, albeit looking identical, might still have different underlying structures. The findings are summarized in table 2.3 2.3 below. Table 2.3: Gramamticality of different size correction with distal negation Category Full left Full right Fronted Lexical Head N * X * Lexical Head A * ? * Functional Head P * X * Functional Head D X X * Functional Head T X – * Functional Head C X X * Phrase DP X X * Phrase PP X X * Phrase VP X ? * Phrase vP X X * Phrase TP X X * Phrase CP X X * A major finding here is that the coordination can absolutely not be fronted without fronting the negation along with it, as a consequence, all fronted examples (numbers (108c 108c) through (121c 121c) in the appendix) are ungrammatical. In Vicente Vicente’s (2010 2010) analysis of English and Spanish, this was treated as an argument in favor of a sentential coordination analysis for fronted sentences, and these cases were analyzed as involving deletion in the left conjunct. An alternative explanation might be that that there is a syntactic dependency between the negation and the coordinators (both sondern and aber), hence the coordinator need to stay overtly in the surface scope of negation. In chapter 4 4, it will be argued that this syntactic dependency is Negative Polarity Licensing, which, in addition to requiring a Downward Entailing environment, is known to be rather PF sensitive. Since it will be said that the negation licenses an operator within the conjunct, they cannot be raised outside the scope of negation. 34 Chapter 2. The syntax of German correction 35 It is important to observe that the prohibition against fronting occurs only in correction, and is not generally found in other types of sentences where focused constituents that are associated with a lower operator are fronted. Focused constituents by themselves (in a non–corrective environment) can be realized within and outside the overt scope of negation. For example, a sentence like (36 36), in which a focused constituent moves from within the surface scope of the particle it is associated with to a position where the particle does not c–command it, is entirely grammatical. The same is possible with negation, however, as soon as correction comes into play, the sentence becomes ungrammatical. (36) JULIANE Juliane hat has nur only die the Blumen flowers gegossen. watered. It was only Juliane that watered the flowers. (37) JULIANE Juliane hat has nicht not die the Blumen flowers gegossen. watered. It was not Juliane that watered the flowers (but somebody else). Another generalization to be drawn from this data set is that while “lexical” heads like N, P and A are still restricted to occur in in “full left” sentences, they can easily occur in “full right” correction structures. This shows that going forward in investigating correction, it is important to understand the restrictions of both processes of generating ellipses. Consequences The main contribution from the subcategorization test were that overall, there is no subcategorization evidence that speaks against a traditional coordination analysis for correction. We saw that there are certain heads and phrases that have restricted distribution. I will discuss these phenomena more indepth in the following paragraphs. 1. Heads cannot occur freely, and even less so with low negation. 2. Fronting is only possible when negation is fronted along. Heads In this section, underlying reasons for the restrictions on heads in corrections are considered. Here, it will be argued that these heads resist being in the immediate scope of the negative adverb nicht ‘not’, because negation is interpreted differently with phrases as opposed to heads. By virtue of the negation being placed in such a low scope environment, its interpretation is qualitatively different from when it occurs in a higher scope environment. For this purpose, Sentences with the adverb nicht were construed. This adverb is free to occur in various locations in a sentence, and used it as head negation. With these examples, it is considered whether its interpretation is the same in all these different environments. 35 36 2.3. Sondern/aber in single corrections In the examples in (38 38) below, I considered the interpretation of negation when it is applied to heads in a neutral context that is not adversative. Specifically, I construed examples of nicht+N in (38a 38a), nicht+DP in (38b 38b), nicht+A(P) in (38c 38c), and nicht+P(P) in examples (38d 38d) and (38e 38e). (38) a. ?? Renate Renate hat has einen a jungen young nicht not Hund dog gef¨ uttert. fed. b. Renate Renate hat has nicht not einen a jungen young Hund dog gef¨ uttert. fed. c. Renate Renate hat has einen a nicht not jungen young Hund dog gef¨ uttert. fed. d. ?? Renate Renate hat has den the Schl¨ ussel key nicht not UNTER under das the Bett bed fallen fall lassen. let. e. ?? Renate Renate hat has den the Schl¨ ussel key nicht not ZWISCHEN between die the St¨ uhle chairs fallen fallen lassen. let. There are some differences that become apparent. On the one hand, the interpretation of the sentence (38a 38a) is ‘Renate fed something, and that something was not a dog’. The interpretation is somewhat similar to correction, but the sentence is grammatically at least unusual, and not all speakers accept it. Likewise, the interpretation of the sentence (38c 38c) with the negation adjacent to the adjective is “Renate fed an old dog”. From the contrast between the sentence in (38a 38a) and the one in (38c 38c), it seems that in sentences where negations apply to heads, especially when these heads express predicates, like nouns and adjectives do, the entailments or implicatures from these sentences are more straightforward and often need not be specified. For example, a not–dog is a creature other than a dog, not young is old, etc. In these sentences, apparently a straight forward mechanism to construe alternatives is available to the speaker. Without providing an explicit semantic analysis for these cases, the interpretation that these sentences have can be easily modeled if it is assumed that this type of negation is focus sensitive. One might assume that this negation takes the context set of the constituents it dominates as an argument, asserts that the prejacent phrase is negated, but presupposes that another one of the focus alternatives are true. Since the focus alternatives in these cases are very restricted, these specific presupposition are the result. What we retain for the future discussion is that negation scoping lower than its default location in Spec of vP might need to be assigned a different interpretation as constituent negation. Other cases, like the sentences where nicht cooccurs with focused spatial prepositions like (38d 38d) and (38e 38e) are less acceptable, since there are a large, maybe indefinite number of spatial alternatives to choose from, and there is no entailment relations between different directions, which would help exclude any of the other alternatives. As such, without contextual restriction, this sentence lacks informativity. These sentences become acceptable when the alternatives contextually restricted, with the focused proposition being one of them. 36 Chapter 2. The syntax of German correction 37 (39) Possible contexts a. Hat Has die the Renate Renate den the Schl¨ ussel keys zwischen between die the Sessel chairs fallen fall lassen, let, oder or neben beside die the Sessel? chairs Did Renate let the keys fall between the chairs or next to the chairs. (40) Answer: Die Renate hat den Sch¨ ussel nicht ZWISCHEN die Sessel fallen lassen. Given the assumption that very low scope negation potentially has a different interpretation than sentential negation, the apparent restrictions on heads in correction might really be a pragmatic restriction. Negation is easily acceptable in cases where there are a salient, limited number of focus alternatives to the phrase negation applies to, and there are contextually rather few alternatives to choose from, so negating one will reduce this number by an informationally significant degree. Fronting Another syntactically very significant restriction that was encountered is that in (possibly only apparent) cases of fronting, negation needs to be fronted along, that is, it needs to be realized sentence initially. This is a restriction that we need to account for in the grammar of correction. Here I want to consider this restriction more closely. We found that sentences like the one below are ungrammatical. Corrections display a restriction against the entire coordination, that is both the target and the consequent of correction, appearing outside the overt scope of negation. In the example below, negation appears in a position overtly c–commanding the VP, but the object has moved out, which is prohibited. Two possible underlying structures for these sentences are entertained, which are made explicit by the bracketing in the example. (41) * Den The Hund dog sondern/aber sondern/aber den the Wolf wolf hat has Sonja Sonja nicht not gestreichelt. petted. a. [Den Hund sondern/aber den Wolf] i hat Sonja nicht t i gestreichelt. b. [Den Hund] i hat Sonja t i gestreichelt *sondern/aber [den Wolf] j hat Sonja nicht t j gestreichelt. Right away, we see some problem with the assumption that the underlying structure is (41b 41b) in these structures. The reason is that (41b 41b), when realized with aber, would be entirely grammatical if it were realized without any deletions in the first conjunct. It also raises problems with sondern. Sondern is expected ungrammatical here, since the negation would be in the second conjunct, which is excluded. This is shown in (42 42) below. (42) * Sonja Sonja hat has den the Hund dog gestreichelt, petted, sondern but Sonja Sonja hat has nicht not den the Wolf wolf gestreichelt petted. 37 38 2.3. Sondern/aber in single corrections However, the interpretation where negation scopes only in the second conjunct is not intended here. Furthermore, it is known (and will be discussed in more detail in the following section) that Right Node Raising, one of the process with which a structure like (41b 41b) might be derived, requires the material it reduces to be phonologically identical to its antecedent, which is not the case in the present constellation, since the second conjunct contains negation the negation nicht, but the first conjunct does not (cf. Hartmann Hartmann, 2000 2000). It is also interesting that this grammatical restriction applies to correction specifically, and not to coordination in general. This indicates that there is a syntactic dependency between the negation and the corrective coordination. This can be deduced because fronting sentences are grammatical with coordinators like und ‘and’, as the example (43 43) exemplifies. Furthermore, this sentence can have both a wide scope and a narrow scope reading, but assuming a derivation analogous to (41b 41b) would predict that only a reading where the coordination takes high scope would be possible. However, this sentence can never have a reading indicating Sonja petted the wolf, and Sonja didn’t pet the dog, that is, a coordination type reading where negation takes scope within the second conjunct, which again is indicative of a low coordination analysis here. (43) Den The Hund dog und and den the Wolf wolf hat has Sonja Sonja nicht not gestreichelt. petted. Sonja didn’t pet the dog and the wolf. Further probing an analysis like (41b 41b), one might consider whether such a structure could arise as a result of a movement restriction within the conjuncts. One might question whether there might be any restrictions which prohibit movement within each of the conjuncts. The structure in (41b 41b) assumes that the subject is individually fronted in each of the conjuncts, which, one might suspect, could violate a PF condition. When considering this idea further, it is found that the same restriction would also need to occur for deletions in the right conjunct, which is not the case. When entertaining a high coordination analysis like (41b 41b), generally it is unexpected that movement would interfere with the grammaticality, since examples (44 44) and (45 45) demonstrate that movement within the conjuncts is possible. In (44 44) we can see that fronting of the object of the first conjunct, the wolf, is possible. Similarly, in (45 45) aber has no restriction about where negation occurs, it can also occur in the second conjunct. When negation occurs there, it only takes scope over the second conjunct, but the object is free to be fronted within the second conjunct. Judging by the relative position of the second conjunct object and the negation, movement within the second conjunct is possible, since the object can occur preceding the negation. Movement from the second conjunct to a preverbal position is impossible. In example (46 46) below only fronting from the first conjunct is possible. The last example, (47 47), shows that this restriction also applies when the coordination is realized after the main verb, so for both full left and 38 Chapter 2. The syntax of German correction 39 full right coordinations. These last two examples are easily explained by assuming the coordinate structure constraint holds for corrections. (44) Den The Wolf i wolf hat has Sonja Sonja nicht not t i gestreichelt, petted, sondern/aber sondern/aber den the Hund. dog Sonja didn’t pet the wolf, but she petted the dog. (45) Den The Wolf wolf hat has Sonja Sonja gestreichelt, petted, *sondern/aber sondern/aber fnichtg the den not Hund dog fnichtg. not. Sonja petted the wolf, but she didn’t pet the dog. (46) * Den The Hund i dog hat has Sonja Sonja nicht not den the Wolf wolf sondern/aber sondern/aber t i gesteichelt petted. (47) * Den The Hund i dog hat has Sonja Sonja nicht not den the Wolf wolf gesteichelt petted sondern/aber sondern/aber t i From the asymmetries between full left and full right coordinations, we can draw the conclusion that deletion in the right and the left conjunct obey very different restrictions. The coordination level in the sentence in (44 44) below is in fact larger than meets the eye, since movement across negation which ends up fronting the object across the subject is possible. This is taken to be evidence in favor of a coordination of CPs in “full left” coordination types. It also shows that as long as negation is within the first conjunct, the grammaticality of the sentence is established. Also the origin of the fronted constituent makes a difference, which is consistent with other sentence types in corrections, like the “full right”–sentences being smaller coordinations. When the coordinator sondern occurs preceding the verb, extraction like we saw in (44 44) is no longer possible. Assuming that it is the coordinate structure constraint that prohibits this movement, it follows that the movement in (48 48) is movement originating from within a coordination, and targeting a location outside. The same is not true for example (44 44), where the movement is within a coordination. (48) * Den The Wolf wolf i hat has Sonja Sonja nicht not t i sondern but den the Hund dog gestreichelt. petted. These positional test surfaced some underlying restrictions in the grammar of correction. Ev- idence was presented which indicated that fronted corrections are likely CP level coordinations in “full left” structures, but DP coordinations in “full right” structures. On the one hand, we saw that the location of negation plays a crucial role in the grammaticality, and that negation has to surface c–command corrections. In later sections, we will see that this restriction exhibit differences between sondern and aber. However, we also saw that the positional data was not uniform for “full left” and “full right” structures, that is, structures with deletion assumed in the right and left conjunct, respectively. In order to understand the grammar of correction comprehensively, it is necessary to consider independent evidence about the process of deletion. This question will be tackled in the following section. 39 40 2.3. Sondern/aber in single corrections 2.3.2 Deletion evidence The preceding sentences in (44 44) through (48 48) showed that different syntactic configurations have very different restrictions when it comes to movement and interpretation. Based on the coordinate structure constraint, we also saw evidence of differing coordination sizes in different syntactic positions. In the present section I will investigate some means which have been proposed in the literature to differentiate between small coordinations, which need not rely on deletion processes and large coordination, which rely on deletion to derive the phonological form. I will look at differences between Right Node Raising and gapping, at some evidence from verb stranding, and the possibility of backward gapping, evidence from Islands, Agreement, scope, and further movement arguments. Right Node Raising vs. Gapping One of the questions that needs to be exhausted is what the underlying structure of “full right” sentences, as opposed to “full left” sentences is. In the preceding discussion about fronting, we saw evidence suggesting that although they allow some similar deletions, their underlying structure is quite different. I think there are two conceivable independent origins for these differences. On the one hand, it might be that “full right” and “full left” corrections have different coordination sizes. On the other hand, it has been brought up in the literature (Hartmann Hartmann, 2000 2000) that rightward deletion, in which the antecedent is preserved, and leftward deletion, in which the consequent is preserved, have different licensing restrictions. Both of these factors might work together in creating the restrictions on correction. It might not be immediately obvious why Right Node Raising is brought up, since the examples seen above are not canonical examples of Right Node Raising. In the example in (49 49), a typical example of Right Node Raising is shown and also how some correction examples could possibly be analyzed as Right Node Raising by analogy. For Right Node Raising cases in German, Hartmann Hartmann (2000 2000) has argued that they generally involve coordination of at least the level of C’. On the other hand, for the correction in (49b 49b), such an analysis seems much less obvious, since another simple alternative is that they simply involve coordinations of DPs in object position. (49) a. Sandra Sandra weiß, knows dass that Erich Erich schon already angekommen arrived ist is und and Markus Markus glaubt, believes dass that Erich Erich schon already angekommen arrives ist. is. b. Sandra Sandra hat has nicht not den the Hund dog gestreichelt petted sondern but den the Wolf wolf gestreichelt petted. 40 Chapter 2. The syntax of German correction 41 c. Sandra Sandra weiss knows nicht, not sondern/aber sondern/aber Markus Markus glaubt, believes dass that Erich Erich schon already angekommen arrived ist. ist. A sentence like (49b 49b) is generally not at all considered a typical instance of Right Node Raising. For example, Sabbagh Sabbagh ( 35 35, 24) gives two criteria for the definition of Right Node Raising: On the one hand, there needs to be an argument position that appears displaced from two sentences it is associated with (the complement clause in 49a 49a), and secondly, this argument is typically realized to the right of the right conjunct. In sentence (49b 49b), his criterion that Right Node Raising involves an argument position realized on the right is not given, since the constituent shared between the conjuncts is the main verb. There are, however, grammatical and unambiguous examples of Right Node Raising in correction, like in example (49c 49c). “Prototypical” Right Node Raising corrections like (49c 49c) underly the same restrictions as general RNR structures. One of these restrictions is that Right Node Raising constructions can only occur with a raised verb in a main clause, since otherwise the argument would occur in the Nachfeld, that means to the right of the main verb, which is a very restricted position for arguments in German. Evaluating the hypothesis that “full right” corrections involve Right Node Raising, it is predicted that their “pivots”, that is, material which needs to be interpreted in both conjuncts, will need to occur to the right of the coordination. This is the restriction that is violated in the ungrammatical sentences in (50 50) below. Since the object in the sentences in (50 50) are base generated to the left of the main verb, the objects are not available for RNR in these sentences. The sentence (50b 50b) shows that Right Node Raising in corrections is also not available when a possible structure as a coordination of a larger constituent (VP in the example) is conceivable, since the shared material (called pivot in the literature) does not occur to the right of the coordination. This condition is fulfilled in (50c 50c), where the verb has moved to C 0 , and the shared object occurs to the right of the conjunction. In all of these criteria, Right Node Raising in correction patterns along with Right Node Raising in coordinations like and. (50) a. * Sandra Sandra hat has nicht not gewaschen washed sondern/aber sondern/aber gestreichelt petted den the Hund. dog. b. * Sandra Sandra hat has nicht not gewaschen washed sondern/aber sondern/aber den the Hund dog gestreichelt. petted. c. Sandra Sandra wusch washed nicht, not, sondern/aber sondern/aber streichelte petted den the Hund. dog. Returning to the question whether a “full–right” sentence like (49b 49b) above involves Right Node Raising, it was seen that generally, such a sentence is not thought to involve Right Node Raising. Above, the hypotheses considered for the derivation of (49b 49b) are that the sentence was either derived 41 42 2.3. Sondern/aber in single corrections by Right Node Raising, or without ellipsis. For the case at hand, we can see that it is not derived by plain Right Node Raising, since the verb is not raised to C. Might there be a process of exceptional Right Node Raising, which can also apply with non–arguments as a pivot, in this case, with v 0 ? Hartmann Hartmann (2000 2000) analyzes Right Node Raising to be a coordination of at least C’, which is only licensed at PF, and not at LF. If this is true, one might consider V as a RNR pivot as an option, which would result in the prediction that these sentences are very high coordinations. However, there is also insufficient evidence in corrections for solely phonological licensing, which is argued characteristic of RNR. If it is in fact only the phonological form that counts, it is predicted that all phonologically identical pivots are acceptable. For example, items that are phonologically the same, but have different interpretation or syntax should be acceptable if the sentences in question are indeed instances of Right Node Raising. The sentence in (52 52) below uses this prediction to provide evidence in favor of a small conjunct approach of full right object coordinations like (49b 49b), and against the idea that the process involved in Right Node Raising might potentially be extended to non–arguments like V positions. In example (51 51), a coordination is attempted with arguments of kalt ‘cold’. This adjective is lexically ambiguous, and can occur with two different case assignments. On the one hand, it can be part of a quirky case assignment in a sentence like Mir ist kalt. ‘to-me.dat is cold’, but it can also be an adjectival predicate, assigning nominative to non–sentient arguments, like in Das Schlafzimmer.nom ist kalt ‘the bedroom is cold’. The predicate kalt ‘cold’ remains phonologically identical. However, the arguments of these two instances of kalt cannot be coordinated, presumably since they assign different case. If we analyzed this type of coordination as Right Node Raising with the analysis of Hartmann Hartmann (2000 2000), we expect this to be possible, since Right Node Raising is argued to be licensed phonologically, and the two instances are phonologically identical. (51) * Es It ist is nicht not mir, to-me.dat sondern but das the Schlafzimmer bedroom.nom kalt. cold. (52) * Nicht Not mir, to-me.dat sondern but das the Schlafzimmer bedroom.nom ist is kalt. cold. Intended: I am not cold, but the bedroom is cold. Furthermore, outside of correction, a sentence which is a coordination of two different inter- pretations of an ambiguous lexical item is entirely possible. For example, Right Node Raising can coordinate conjuncts with two interpretations of the well–known ambiguity Bank ‘financial institution’ and Bank ‘long chair’. This highlights the fact that a different leftward deletion process must apply in cases where V 0 is the remnant of a leftward deletion. (53) Ich I sitze sit nicht not auf on der the sondern but gehe walk zur to-the Bank. bank. 42 Chapter 2. The syntax of German correction 43 As a consequence, I consider cases of “full right” corrections with non–argument remnants not to be instances of Right Node Raising, unless they clearly involve a pivoted argument. Since this concerns only very few cases, it is not discussed any further. Another option that has been mentioned in the literature is verb stranding, which is discussed below. In the continuing attempt to determine the coordination size and type of deletion of sentences like (49b 49b), that is, full right object position corrections, I am further considering if this process could be derived by virtue of assuming across-the-board movement of the verb to a position to the right. Verb-stranding is a process assumed to apply in English VP ellipses. It is described for a number of languages in Goldberg Goldberg (2005 2005). Goldberg Goldberg (2005 2005) discusses sentences like the one in (54 54). Her work accounts for restrictions that certain languages like English, but also Hebrew and Swahili have with respect to realizing the verb in the second conjunct. In the case of English, the analysis is meant to explain the fact that the main verb cannot be realized in English, but the adverb too or similar adverbs are necessary. (54) Arthur [VP brought a present to Hall], a. and Julia did [bring a present to Hall] too b. * and Julia brought too; *and Julia will bring too. Goldberg Goldberg (2005 2005, 6) I do not think that the German data can be described by this type of verb stranding, since the data described cover the relation between the auxiliary and the adverb. This is not the type of case we are dealing with here, so I believe this body of literature is not relevant. Under the assumption that the negation needs to scope lower than the coordinators sondern/aber, the following three possible structures for these sentences are considered. (55) a. Sandra Sandra hat has [ DP nicht not den the Wolf wolf sondern but den the Hund dog ] gestreichelt. petted b. Sandra Sandra hat has [ VP nicht not den the Wolf wolf gestreichelt sondern but den the Hund dog gestreichelt]. petted c. Sandra Sandra hat has [ VP nicht not den the Wolf wolf t i sondern but den the Hund dog t i ] gestreichelt i . petted These three structures are different from each other since the structure (55a 55a) in does not involve ellipsis, the structure in (55b 55b) involves a deleted verb, whereas the sentence (55c 55c) involves Across The Board– movement to the right. In the last section it was argued that the sentence in (55b 55b) does not involve Right Node Raising, but at this point it was not discussed whether other deletion/ellipsis process could be involved. Due to the deletions taking place, this process might be argued to be a reverse pseudogapping structure, since the verb is deleted, whereas the auxiliary is realized in the conjunct where deletion takes place, that is the left conjunct in (55b 55b). It is reversed since in pseudogapping, the auxiliary is typically 43 44 2.3. Sondern/aber in single corrections retained in the right conjunct. Both verb stranding and “reverse pseudogapping” are processes of VP ellipsis (Johnson Johnson, 2009 2009 for English). However, independently, pseudogapping in its “forward” form is ungrammatical in German, so this process would be an exception within the German grammar, but also across languages. Backward Gapping While not very widely discussed for German, in SOV languages, due to their word order, a process of backward gapping is in principle possible. We are interested if this process is also applicable to the full right sentences discussed above. The possibility of backward gapping was already predicted in Ross Ross (1970 1970). Since gapping can occur in the second conjunct in an S V O & S O configuration, why should backwards gapping not occur in a S O & S O V configuration? Koster Koster (1975 1975), who argued that both German and Dutch have underlying SOV word order, discusses the option of backward gapping in these structures. However, as him and others (Maling Maling, 1972 1972) point out, backward gapping is only possible when the main verb is realizes at the end of the second conjunct. It is not possible in any other configuration. At this point, backwards gapping is mentioned as a possibility. One of the criteria used to distinguish ellipsis structures from non–ellipsis structure, that is, possible backwards gapping from DP coordinations, are island constraints, which will be evaluated in the next section. 2.3.3 Islands In the literature on correction, but also ellipsis in general, island effects have been used as tests. For English, it has been pointed out that correction targets which are negated using constituent negation do not trigger island effects, whereas correction targets which involve sentential negation do (Toosarvandani Toosarvandani, 2014 2014). As I will argue in this section, similar facts are reproducible in German (although island effects are not only conditioned by negation type in German, but also by sentence type, that is, the difference between full left and full right coordination). In this section, I will mostly consider evidence from islands as discussed in the literature of ellipsis and other deletions. We will see that sondern and aber differ in certain situations with respect to whether they allow island violations. In much of the data, sondern does not exhibit island effects. This indicates that the two coordinators have different underlying syntactic structure, and is a main reason for suggesting that the conjuncts of aber are larger than the ones of sondern. Whether or not island effects arise in corrections has been treated as a test for the presence of hidden material in some types of deletion. The underlying motivation for this test, given a theory of ellipsis derivation along the lines of Merchant Merchant (2005 2005), lies in the way that certain types of focus sensitive ellipses are assumed to be derived. Specifically, this has been argued for gapping and 44 Chapter 2. The syntax of German correction 45 stripping, and other types of ellipses, like fragment formation. The example below illustrates in broad strokes the mechanics in Merchant Merchant (2005 2005). In a fragment answer, Merchant Merchant (2005 2005) argues, movement licensed by an information structural feature (topic or focus) is triggered to a fronted position, with the remainder of the sentence being elided. (56) What does Sandra eat? — A pear i Sandra eats t i Merchant Merchant (2005 2005) puts forward this type of analysis for fragments, but also gapping and stripping. Since this derivation relies crucially on movement, which is subject to island condition violations, island violations will diagnose whether such a derivation of ellipsis is at work. This was a crucial argument that was brought forward to shed light on the question of the coordination size of correction in English. In the preceding sections, we saw that the position of negation is closely linked to coordination level, since sentences where negation are realized low are (at least possibly) compatible with low level coordinations, whereas the same is impossible for higher scoping negation. Along these lines, Toosarvandani Toosarvandani (2013 2013) argues that in English, coordination size covaries with the type of negation. Where the sentence–adverbial negation n’t occurs in a sentence, he finds evidence of ellipsis due to island violations, whereas in case of lower adverbials, no evidence of ellipsis is found. This argument is based on island effects in the spirit of analyses described above: the adverbial negation exhibits island effects whereas the negative determiner no does not exhibit such effects. In German, I will show that whether or not we encounter island effects in corrections is modulated by negation as well as sentence types. While “full right” correction does not show island effects and therefore does not involve ellipsis, sentence–initial/fronted and extraposed/full left coordination do. Here is Toosarvandani Toosarvandani’s (2013 2013) argument presented in more detail. Just like in German, corrective coordinations can occur with both constituent negation and sentential negation in English. Relevant examples are given in (57 57). In (57b 57b) and (57c 57c), the negative adverb not is used to modify a determiner and a Determiner Phrase, respectively. Also in (57a 57a), DPs under the scope of constituent negation are the target and consequent of correction. In English, this sentence is not perfectly acceptable with the constituent negation not. Many speakers accept this sentence, but report as sounding antiquated. The same sentence is perfectly acceptable when expressed with the sentential negation n’t in (57d 57d). (57) a. ?/% Miriam ate not an apple but a pear. b. Miriam ate not one but two pears. c. Not Miriam but Norbert ate a pear. d. Miriam didn’t eat a fig, but peel a fig. 45 46 2.3. Sondern/aber in single corrections Toosarvandani Toosarvandani (2013 2013), basing himself on work by Vicente Vicente (2010 2010), who in turn bases himself on Drubig Drubig (1994 1994), shows that only sentential negation exhibits island effects in English, but constituent negation does not. Below, some of the relevant examples are given. He also considers other types of island violations, but in this work, only coordination islands and relative clause islands are presented because they give the clearest intuitions in German. The nature of the argument is that since ellipsis triggers island effects, the unacceptable sentences (in the relevant readings) are thought to have ellipsis. The examples are cited literally, that is, including the strikeouts. However, it is doubtful that Toosarvandani Toosarvandani (2013 2013) means the strikeouts necessarily represent the deleted material, since he later goes on to argue that English corrections are low conjuncts, which appears at odds with the strikeouts. The strikeouts here are meant to highlight the relevant reading, but not to make a claim about the syntactic structure. The examples in (58a 58a) and (58c 58c) are ungrammatical, since it is assumed the constituent remaining in the second conjunct is assumed to raise from a position from which movement is generally prohibited, namely a coordination and an adjunct respectively. In the sentences with constituent negation in (58b 58b) and (58d 58d), such a movement is not required. (58) a. * Alfonse didn’t cook rice and beans, but potatoes. (=. . . but Alfonse cooked [rice and potatoes].) b. Alfonse cooked rice and not beans but potatoes. c. * Alfonse didn’t smash the vase that Sonya had brought from China, but from Japan. (=. . . but Alfonse smashed [the vase that Sonya had brought from Japan].) d. Alfonse smashed the vase that Sonya had brought not from China but from Japan. Now let’s consider the German examples. First of all, I want to show that German also has the same possibility of corrective negation being expressed on a sentential and a constituent level. The negations used to express this need not be morphologically different. They do, however, occur in a different syntactic position. Also, the cases where DPs are corrected, like English (57a 57a), are entirely okay in German. When the negative adverb nicht occurs immediately before the verb, it has the most neutral reading, but if stress is put on the main verb, it can also be interpreted as a correction of the verb. Notice also that there is some awkwardness in saying nicht ein “not a”, possibly since there exists a very salient negative quantifier kein(e/r) ‘no’. I circumvent this by using plural DPs in the example. (59) a. Miriam Miriam hat has nicht not Feigen figs gegessen, eaten, sondern but Birnen. pears. Miriam ate not figs, but pears. b. Nicht Not Miriam Miriam sondern but Norbert Norbert hat has eine a Birne pear gegessen. eaten. Not Miriam but Norbert ate a pear. 46 Chapter 2. The syntax of German correction 47 c. Miriam Miriam hat has eine a Feige fig nicht not gegessen, eaten, sondern but gesch¨ alt. peeled. Miriam didn’t eat a fig, but peel a fig. Already in previous sections evidence was presented arguing that different German sentence types (“full left”, “full right” and “fronted”) have different underlying syntactic structure. We expect that sentences with a full left conjunct will differ from the ones with a full right conjunct, with fronted corrections remaining an open question, particularly when considering the discussion of different deletion types in German presented in the preceding sections. Consequently, we potentially expect different island effects in the three model sentence structures. We can see that the full left conjuncts are not acceptable in their island reading, which indicates that the consequent of the correction, Kartoffeln ‘potatoes’ has been moved, which is indicative of ellipsis in these cases. In (60 60) below, you can see the examples of sentential negation in a “full left” correction. Sondern and aber do not differ in this case, island effects are triggered with both of the coordinators. (60) a. * Alfons Alfons hat has nicht not Reis rice und and Bohnen beans gekocht, cooked sondern/aber but Kartoffeln. potatoes Alfons didn’t cook rice and beans, but rice and potatoes b. * Alfons Alfons hat has nicht not die the Vase Vase die that ihm him Sonja Sonja aus from China China gebracht brought hat, has, zerst¨ ort, destroyed, sondern/aber but aus from Japan. Japan Alfons didn’t destroy the vase that Sonja brought him from China, but the one that she brought him from Japan This pattern is not surprising. It is often argued that main verbs are base generated at the end of a German sentence, and it select their objects to the left. (B¨ uring and Hartmann B¨ uring and Hartmann, 2001 2001) There are several restrictions at play here. On the one hand, we cannot coordinate constituents of different type, so the DP Kartoffeln ‘potatoes’ cannot be directly coordinated with gekocht ‘cooked’, or nicht Reis und Bohnen gekocht ‘not rice and beans cooked’. Already from this, it can be deduced that the second conjunct likely contains a suppressed verb, and its structure must underlyingly contain a verb, and predict the island effect. (61) Alfons Alfons hat has nicht not Reis rice und and Bohnen beans gekocht, cooked, sondern/aber sondern/aber Kartoffeln i potatoes t i gekocht. cooked. Another relevant question is whether the deleted phrases contain coordination. The answer is no, which triggers the island effect and makes the reading requiring a deleted coordination unavailable. As an illustration, please consult the schematic in (62 62). This possibility is excluded since it violates the coordinate structure constraint. 47 48 2.3. Sondern/aber in single corrections (62) * Alfons Alfons hat has nicht not Reis rice und and Bohnen beans gekocht, cooked, sondern/aber sondern/aber Kartoffeln i potatoes Reis rice und and t i gekocht. cooked. Island effects are also witnessed with sentential negation in full left sentence structures, with the same argumentation at play as in full left sentences with constituent negation above. The smallest constituent that can be right–extraposed without predicting a violation of the coordinate structure constraint is at least VP. In addition, we also know that the level of coordination in this sentence is higher than the scope of negation in the first conjunct. (63) a. * Alfons Alfons hat has Reis rice und and Bohnen beans nicht not gekocht, cooked, sondern/aber but Kartoffeln. potatoes. Alfons didn’t cook rice and beans, but rice and potatoes. b. * Alfons Alfons hat has die the Vase, Vase die that ihm him Sonja Sonja aus from China China gebracht brought hat, has, nicht not zerst¨ ort, destroyed, sondern/aber but aus from Japan. Japan. Alfons didn’t destroy the vase that Sonja brought him from China, but the one that she brought him from Japan It can be deduced that independent of negation scope, all “full left” coordinations involve ellipses. In the fully realized right conjuncts, we find no evidence for islands with either sentential or constituent level negation. In (64 64), “full right” examples with sentential negation are presented. Sondern does not trigger island effects. (64) a. Alfons Alfons hat has nicht not Reis rice und and Bohnen, beans sondern/*aber but Kartoffeln potatoes gekocht. cooked. Among available readings for sondern: Alfons didn’t cook rice and beans, but rice and potatoes. b. Alfons Alfons hat has nicht not die the Vase, Vase die that ihm him Sonja Sonja aus from China, China sondern/*aber but aus from Japan Japan gebracht brought hat, has, zerst¨ ort. destroyed Alfons didn’t destroy the vase that Sonja brought him from China, but the one that she brought him from Japan This is one of the constructions where sondern and aber diverge from one another, since aber triggers an island violation, that is, it can only denote the non–island reading. This might be consistent with two potential analyses: either a backwards deletion/ backwards gapping analysis, or a low coordination analysis of sentences with sondern. Under the former 48 Chapter 2. The syntax of German correction 49 assumption, this would mean that the underlying structure is best approximated by (65a 65a) below. In the case of low coordination, we simply assume that there are no deletions, and the structure is “what you see is what you get”. This option is represented in (65c 65c). Another variation of this sentence is presented in (65b 65b). (65) a. Alfons Alfons hat has nicht not Reis rice und and Bohnen beans gekocht, cooked sondern/*aber but Kartoffeln i potatoes t i und and Bohnen beans gekocht. cooked. Among available readings for sondern: Alfons didn’t cook rice and beans, but rice and potatoes. b. Alfons Alfons hat has [[ nicht not [ DP Reis rice und and Bohnen beans ]],[ sondern/*aber but [Kartoffeln i potatoes [ DP t i und and Bohnen beans ]]]] gekocht. cooked. Among available readings for sondern: Alfons didn’t cook rice and beans, but rice and potatoes. c. Alfons Alfons hat has nicht not [ DP Reis rice [ &P und and [ DP Bohnen, beans sondern/*aber but Kartoffeln]]] potatoes gekocht. cooked. Among available readings for sondern: Alfons didn’t cook rice and beans, but rice and potatoes. Contemplating the possibility of backwards deletion further, it can be seen that this assumption does not provide a satisfying explanation. If deletion applies to the head sketched out in the example (65a 65a), we would expect the structure to continue to be ungrammatical, since there is still movement (as a process of deriving ellipsis) violating the Coordinate Structure Constraint in the second conjunct. Hence, example (64 64) is predicted to be ungrammatical across the board, independent of coordinator, which is counter to fact. The same problem afflicts (65b 65b), since there is a movement violating the Coordinate Structure Constraint in the right conjunct. However, if one does not assume this movement, the underlying reason for the island violation is unexplained. Consequently, something entirely new is learned about the structure in (65c 65c): if indeed this is the underlying structure, negation can sometimes c–command sondern and continue to be a grammatical sentence, and need not be part of the left conjunct of the construction. The converse is true for aber, it must never be c–commanded by negation, or has an entirely different syntactic structure. Likewise, in the constituent negation cases, object corrections also do not produce island effects. However these examples are still crucially different from the ones in (64 64), since aber is acceptable in these cases, which indicates that there is no inherent problem for aber to occur in a structure without ellipsis, as long as the negation scope is not too high. 49 50 2.3. Sondern/aber in single corrections (66) a. Alfons Alfons hat has Reis rice und and nicht not Bohnen, beans sondern/aber but Kartoffeln potatoes gekocht. cooked. Among available readings: Alfons cooked not rice and beans, but rice and potatoes. b. Alfons Alfons hat has die not Vase, the die Vase ihm that Sonja him nicht Sonja aus from China, China sondern/aber but aus from Japan Japan gebracht brought hat, has, zerst¨ ort. destroyed Alfons didn’t destroy the vase that Sonja brought him from China, but the one that she brought him from Japan The contrast in the aber cases exhibits that the real difference between the constituent negation cases and the sentential negation cases lies in the locality restrictions that sondern and aber exert with respect to the placement of negation. The last syntactic configuration of interest is preverbal correction. A crucial question with respect to preverbal correction/ fronting is whether the negation preceding the sentence applies on a sentential or a constituent level. For the time being, the null hypothesis given the phonological form is that negation in the Vorfeld applies to DPs in the examples given. In case there are underlying ellipsis in these sentence, island effects are the test that would detect them, as we would expect an island violation if ellipsis applies (backwards) to the left conjunct. So far only ellipsis in the right conjunct was tested for, and deletion in the left conjunct was not excluded. 1 1 If we find island effects for the left conjunct, we will be able to take this as evidence of deletion, but we cannot take the absence of island effects as evidence for the absence of deletion, since some types of backwards deletion do not cause island violations. I argued that focus movement was the underlying reason for island effects, but we do not have independent evidence for such movement in the left conjunct. However, if such island effects are indeed triggered, this would provide independent evidence for the assumption of focus movement. (67 67) presents the relevant data. (67) Sentential negation with island in right conjunct Ich I habe have nicht not gesehen, seen, dass that Alfons Alfons und and Lisa Lisa sondern/*aber but Miriam Miriam Kartoffeln potatoes gekocht cooked haben. have Among available readings: I didn’t see that Alfons and Lisa cooked potatoes, but that Alfons and Miriam cooked potatoes. (68) Sentential negation with island in the left conjunct 1 Vicente Vicente (2010 2010) excludes backwards ellipsis altogether on grounds of processing grounds. This does not seem compelling to me, since Natural Language has a number of non–linear dependencies, cataphoric pronouns being one of them. 50 Chapter 2. The syntax of German correction 51 Ich I habe have nicht not gesehen, seen, dass that Alfons Alfons sondern/*aber and Lisa Lisa und but Miriam Miriam Kartoffeln potatoes gekocht cooked haben. have Among available readings: I didn’t see that Alfons and Lisa cooked potatoes, but that Alfons and Miriam cooked potatoes. (69) Constituent Negation with island in the right conjunct Nicht Not Alfons Alfons und and Lisa Lisa sondern/*aber but Miriam Miriam haben have Kartoffeln potatoes gekocht. cooked Among available readings: I didn’t see that Alfons and Lisa cooked potatoes, but that Alfons and Miriam cooked potatoes. (70) Constituent Negation with island in left conjunct Nicht Not Alfons Alfons sondern/aber but Lisa Lisa und and Miriam Miriam haben have Kartoffeln potatoes gekocht. cooked. Among available readings: Not Alfons and Miriam cooked potatoes, but Lisa and Miriam cooked potatoes. As we saw above, the surface data indicates it is possible to realize an entire correction in the Vorfeld, but only granted the negation is moved along with the correction. As a result, in (67 67), two configurations are tested, those with the negation moved along, which are possibly constituent negation, and others which are embedded under negation in a matrix clause, which is more likely to be sentential negation. I am not presenting the relative clause extraction islands since they are independently excluded, since extraction of relative clauses to the V orfeld is ungrammatical. Aber is ungrammatical in all sentences in (67 67), which is unexpected. The reason it is unexpected is that (66 66) indicated that aber can in fact occur in small coordinations. Consequently, it is unexpected that aber is ungrammatical due to being forced into a syntactic location where it cannot take a large enough syntactic size, for example. The sentences in (67 67) and (68 68) indicate that aber is sometimes not licensed to occur in correction sentences where negation is realized in a matrix clause or from locations higher up in the structure. The verb sehen ‘see’ is used as an embedding verb so as to exclude potential issues with neg raising, so there is no reason to assume the negation reconstructs into the embedded clause. Verb agreement is additional evidence that can be taken into consideration for the existence of the target reading in the cases in (67 67). As I will discuss later in more detail, with corrections, there is only ever agreement with a single conjunct. Since in the cases in question, the reading that violates the island would have a singular second conjunct, the fact that there is plural agreement in these sentences provides some more support for the existence of the reading in question. 51 52 2.3. Sondern/aber in single corrections Recap: Islands From the examples presented in this section, it can be concluded for German that whether or not an island violation is encountered is modulated by different sentence types. The data are summarized in table 2.4 2.4 below. We saw that in preverbal position, aber triggers island effects, whereas sondern does not, and also sometimes shows additional interaction with negation in the matrix clause. In “full right” structures, that is, potential Backwards Gapping examples, aber triggered an island effect with sentential negation, but not constituent negation. Independently, the explanation of Backwards Gapping was found to be unconvincing. In “full left” conjunct structures, constituent negation did not trigger island effects, and therefore ellipsis is not assumed to be present with sondern, but not with aber. Table 2.4: A summary of the island data sondern aber preverbal constituent No Island Effect No Island Effect sentential No Island Effect Island Effect embedded (full right) constituent No Island Effect No Island Effect sentential No Island Effect Island Effect right–extraposed (full left) constituent No Island Effect Island Effect sentential Island Effect Island Effect The island data allows important inferences about ellipsis, but it highlights interesting observa- tions about the difference between sondern and aber. In the discussion of the possible derivation of the perceived gaps in “full right” conjuncts, I discussed two hypotheses. One of them, a backward deletion mechanism, was discarded since it did not provide an explanation for the pattern of island effects observed. As a consequence, we know that both sondern and aber can occur in coordinations as small as DPs. Furthermore, sondern and aber differed with respect to whether or not they triggered island effects in certain configurations, namely in DP coordinations, and CP level coordinations where negation was assumed to be interpreted outside of the coordination. We found that aber was ungrammatical in these configurations, whereas sondern was not. Taking together these observations, two conclusions were drawn. On the one hand, there were exceptions to out assumptions that negation needs to be in the scope of the first conjunct of correction. Assuming conclusion of section 2.3.2 2.3.2 that “full right” correction with constituent level coordination is coordination of DPs that does not involve ellipsis, and the argumentation about fronting, we saw that a crucial difference between sondern and aber was that sondern happily appeared within the scope of negation which was higher up, whereas aber did not. This appears exceptional and is in need of additional explanation. The author takes these facts as evidence for 52 Chapter 2. The syntax of German correction 53 further LF movement involved in the derivation of sentences with sondern. Another consequence of these observed data is that there may differences between the negations involved in corrections with sondern and aber. The island tests provided some evidence regarding the coordination size of the two coordinators. Sondern can coordinate DP conjuncts independent of syntactic location and negation type (local or distal negation). For sondern, the only configuration where negation type played a central role are the full–left constructions. A reason for this was given: since there is good evidence that in these constructions, the sondern+XP component is moved to the right from a position preceding the verb, the smallest component that can be moved without any violation of the Coordinate Structure Constraint is at least the VP, which is followed by a deletion of the main verb. 2.3.4 Agreement For Spanish corrections, Vicente Vicente (2010 2010) uses agreement of the conjuncts as an argument in favor or against sentence level coordinations. I will consider the same line of argumentation for German, concluding that it is not a fully insightful reasoning for this language, since although although data consistent with low and high coordination is found, it is not possible to exclude either of the possibilities based on agreement. The rationale works as follows: often in coordinations of DPs with and and other coordinators, plural agreement is found although the individual coordinates appear in singular. For example, in the sentence Stef and Marygrace are working hard, even though Stef and Marygrace are each singular, the agreement is consistent with Stef and Marygrace forming a plural entity. Possibly we could expect to find the a similar pattern for corrections. However, a more suitable coordinator to compare is or, since it is more prone to agreement with the conjuncts. Also, whether the coordinations appear in Spec of CP or following the finite verb makes a crucial difference for agreement. For German specifically, agreement with the entire correction is generally impossible, unlike with or. (71) a. Es It hat/*haben has/have nicht not Erich Erich sondern/aber sondern/aber Magda Madga die the Torte cake gegessen. eaten. Erich didn’t eat the cake, but Magda did. b. i. Die The Torte cake hat/*haben has/have nicht not Erich Erich sondern/aber sondern/aber Magda Magda gegessen. eaten Erich didn’t eat the cake, but Magda did. ii. Die The Torte cake hat/haben has/have nicht not Erich Erich oder or Magda Magda gegessen. eaten Neither Erich nor Magda ate the cake. 53 54 2.3. Sondern/aber in single corrections c. i. Nicht Not Erich Erich sondern/aber sondern/aber Magda Madga hat/*haben has/have die the Torte cake gegessen. eaten ii. Nicht Not Erich Erich oder or Magda Magda hat/haben has/have die the Torte cake gegessen. eaten. Neither Erich nor Magda ate the cake. We can see that while oder ‘or’ is mostly free to agree in singular and plural, generally sondern and aber agree only in singular when each of the conjuncts are singular. Also, in general, in sentences with oder ‘or’, negation always takes high scope over the coordinator. Without further indepth analysis of the interaction of syntax and semantics involved in creating the scope relations between negation and coordination, the conclusion that can be drawn about sondern and aber is that their scope is not variable. As a consequence we have no way of knowing if agreement is a matter of syntactic structure or is constrained by the semantics of the construction. Vicente Vicente (2010 2010) also applies another, related test: he uses correction to coordinate SG+PL and PL+SG conjuncts, in hopes of answering the question which conjunct agrees with the verb by making the conjuncts a different number. As such, if plural agreement is possible, we would think that agreement is either with the entire conjunct, which is independently excluded, as the examples in (71 71) show, or with the plural conjunct. This test gives us further insight into the different syntactic structure of the three sentence types (full left, full right, fronted). For the sentences with a fully realized left conjunct in (72 72), we find that agreement occurs with the left conjunct, although the contrast is more stark in cases where a plural DP in the left conjunct precedes a singular DP in the right conjunct. This may be construed as an argument in favor of an adjunct structure of coordination, since in such a structure, the first conjunct is closer to the verb with which agreement takes place. However, in the sentences where a singular DP in the left conjunct precedes a plural DP in the right conjunct, really both singular and plural are possible. Since the intuitions are not very strong, I will use a number of different grammaticality judgments below, namely, in descending order of ungrammaticality *, ??, ?. (72) It is/are notfone/threeg trombinistfsg came, butfone/threeg pianistfsg. a. Es ist/?sind nicht ein Posaunist gekommen, sondern drei Pianisten. 1 sondern 3 b. Es *ist/sind nicht drei Pianisten gekommen, sondern ein Posaunist. 3 sondern 1 c. Es ist/??sind nicht ein Posaunist gekommen, aber drei Pianisten. 1 aber 3 d. Es *ist/sind nicht drei Pianisten gekommen, aber ein Posaunist. 3 aber 1 When eliding the second conjunct, as in (73 73), but at the same time fronting the negated subject, agreement occurs strictly with the first conjunct. The judgments in (72 72) are reiterated, but are much clearer when the negation is fronted to Spec of CP. (73) Notfone/threeg trombonistfsg is/are come, butfone/threeg pianistfsg. 54 Chapter 2. The syntax of German correction 55 a. Nicht ein Posaunist ist/*sind gekommen, sondern drei Pianisten. 1 sondern 3 b. Nicht drei Pianisten *ist/sind gekommen, sondern ein Posaunist. 3 sondern 1 c. Nicht ein Posaunist ist/*sind gekommen, aber drei Pianisten. 1 aber 3 d. Nicht drei Pianisten *ist/sind gekommen, aber ein Posaunist. 3 aber 1 In the cases with a full right conjunct, with the entire correction preceding the participle in (74 74), agreement happens preferentially with the plural conjunct, no matter whether plural is expressed in the first or the second conjunct. The contrasts are much starker with aber than with sondern, since with the latter, really both singular and plural are possible, but plural agreement is preferred. This is really the case that stands out from the paradigm. (74) It is/are notfone/threeg trombonistfsg, butfone/threeg pianistfsg come. a. Es ?ist/sind nicht ein Posaunist, sondern drei Pianisten gekommen. 1 sondern 3 b. Es *ist/sind nicht drei Pianisten, sondern ein Posaunist gekommen. 3 sondern 1 c. Es *ist/sind nicht ein Posaunist, aber drei Pianisten gekommen. 1 aber 3 d. Es *ist/sind nicht drei Pianisten, aber ein Posaunist gekommen. 3 aber 1 Agreement is with the right conjunct in the cases in (75 75), which are fronted coordinations with a full left conjunct. Independently, these cases speak against Right Node Raising on a level higher than the auxiliary, since this process was described in the literature to require phonological identity, which is not given in the cases below, since the auxiliaries are not identical. (75) Notfone/threeg trombinistfsg, butfone/threeg pianistfsg are/have come. a. Nicht ein Posaunist, sondern drei Pianisten *ist/sind gekommen. 1 sondern 3 b. Nicht drei Pianisten, sondern ein Posaunist ist/*sind gekommen. 3 sondern 1 c. Nicht ein Posaunist, aber drei Pianisten *ist/sind gekommen. 1 aber 3 d. Nicht drei Pianisten, aber ein Posaunist ist/*sind gekommen. 3 aber 1 The paradigm is summarized in Table 2.5 2.5 below Table 2.5: Summary of the agreement paradigm full left full right fronted full left (stripping) fronted full right sondern SG<PL SG/?PL ?SG/PL SG/*PL *SG/PL PL<SG *SG/PL *SG/PL *SG/PL SG/*PL aber SG<PL SG/??PL *SG/PL SG/*PL *SG/PL PL<SG *SG/PL *SG/PL *SG/PL SG/*PL It is a bit of an open question what these data tell us. In accordance with the literature on Closest Conjunct Agreement, we might expect that agreement can either happen with the closest conjunct or 55 56 2.3. Sondern/aber in single corrections with the entire coordination. However, the data in (71 71) speaks against the possibility of agreement with the entire coordination. Also the observation that both singular and plural agreement are possible in some cases gives us further support not to consider agreement as a test of coordination size. If number can never be created from both conjuncts in cases where singular and singular are coordinated, the same is to be expected in coordinations of plural and singular or singular and plural. Given that this is the case, there is still variations with respect to which conjunct is in agreement with the verb, which is not an issue in the Spanish data presented in Vicente Vicente (2010 2010). On the other hand, the data where a singular object is coordinated with a plural object is less clear, and the contrast in grammaticality is perceived strongest in the preverbal cases. In these cases, agreement is strictly with the second conjunct in “full right” cases, which indicates that the auxiliary that is phonologically realized stems from the second conjunct. In all cases a categorical difference between sondern and aber is not found, it is more relevant in which sentence configuration the coordination occurs. Overall, the contrasts are a bit more stark with aber. From this we might conclude that sondern generally has more ambiguity for the speaker when it comes to coordination level. We can also conclude (to my knowledge in accordance with the Closest Conjunct literature) that movement to Spec of CP is the process that imposes strict agreement. In the cases with assumed extraposition to the right (the “full left cases”) in (72 72), agreement tends to be with the first conjunct, which indicates that coordination occurs at a level higher than verbal agreement, since agreement can happen with either conjunct. In the comparable preverbal negation case, (73 73), the intuition of ungrammaticality is much more categorical, meaning that the auxiliary must necessarily agree with the subject of the first conjunct, whereas in (72 72), only a tendency is recognizable. In the embedded cases (the “full right” cases), agreement tends to be plural. The strong interpretation of agreement data in terms of coordination height suggested by Vicente Vicente (2010 2010) is not warranted. His interpretation of this type of data is that whenever we see singular agreement, we necessarily deal with high coordination, whereas whenever there is plural agreement, there is low coordination. The reason I do not think this argumentation applies to German is that on the one hand, correction never agrees plurally, that is, there is never evidence that correction will construe the number from both conjuncts like other coordinators do. However, we find significant syntactic differences between different types of correction in agreement and many other syntactic tests, which would remain unaccounted for of agreement is taken to necessarily refer to sentential vs. nonsentential correction, as Vicente Vicente (2010 2010) intends. This being said, we can recognize three different patterns in the agreement data. In cases where the left conjunct is fully realized, agreement is with the left conjunct. In cases where the verb is realized in the right conjunct an correction remains in situ, agreement can happen with both the singular and the plural conjunct. In those cases where correction appears fronted, agreement takes 56 Chapter 2. The syntax of German correction 57 place in the fully realized right conjunct. These data are taken as evidence that in both “full left” and “fronted” corrections the left conjunct and right conjunct is realized in full, hence agreement happens in these sentences. This is either evidence for a high coordination level of these sentences, or for agreement being restricted to the closer conjunct. It is more remarkable is that the agreement restrictions disappear with the embedded “full right” sentences. In these sentences, suddenly both singular and plural agreement is an option. On the one hand, this indicates that these sentences are structurally crucially different form the others, and also it likely means that both conjuncts are more accessible to T, meaning they are structurally lower than T. I take this to be evidence for these embedded structures to really be small coordinations without ellipses. In terms of the questions of coordination height, these data do not call in question the analysis of “full right” sentences as small coordination without ellipsis, and the “full left” conjuncts as higher coordination, involving ellipsis. These also have interferences with the semantics of the construction. As will be discussed in later chapters, sondern and aber differ with respect to whether they allow the second conjunct to entail the first. These judgments are also an influence in the grammaticality judgments here. 2.3.5 Scope The island and agreement data give a first indication about the differences between sondern and aber. On the one hand we saw some differences related to coordination size, on the other hand, we saw differences in how the coordinators related to the syntactic location of negation. In certain configurations, ellipsis was found in aber clauses and could not be established with sondern, indicating that the cases with ellipsis coordinate larger structures. We have seen that in some syntactic configurations, for example the preverbal fronting and embedded cases, sondern has the option of coordinating strictly DPs, whereas aber must coordinate larger structures that involve deletion. We find a difference between sondern and aber substantiated in other constructions as well. Another significant difference we encountered was related of the coordinators to negation. The discussion in section 2.3.3 2.3.3 showed that when we take considerations from research in deletions into account, we must conclude that sondern can sometimes occur within the scope of negation, whereas aber cannot occur in a such position. In this section I will introduce a related observation, namely that the two coordinators allow for different negation scope. With both types of coordinators, the lexical item nicht ‘not’ can be used, which, judging from the data in the last sections, can adjoin to verbal constituents as well as DPs and other phrases. Consider the sentences in (76 76) below. In these examples, sondern and aber are used in the same sentence as the universal quantifier, jeder ‘every’. Given the right intonation, namely a “rise–fall” intonation with rising prosody on jeder, and falling prosody on the negation 57 58 2.3. Sondern/aber in single corrections nicht, even in a scope–rigid language like German, inverted scope is predicted possible. However, the inverted scope reading is not available in the example with sondern in (76a 76a), but only in the example with aber in (76b 76b). These sentences intentionally do not have the structurally disambiguating auxiliary+participle structure, which leaves open the question of whether they should be “full left” or “full right”, that is elliptical or nonelliptical conjuncts. (76) a. Jeder Everyone isst eats nicht not ¨ Apfel, apples, sondern but Birnen. pears. Nobody eats apples, but everybody eats pears. 8x(:A(x)^ B(x)) 2 2 b. Jeder Everyone isst eats nicht not ¨ Apfel, apples, aber but Birnen. pears. Nobody eats apples, but everybody eats pears. 8x(:A(x)^ B(x)) Some people eat apples, but everybody eats pears. :(8x(A(x))^8y(B(y))) There are two conceivable hypotheses which could explain the differences between these sentences. On the one hand, it is possible that aber can coordinate larger conjuncts. So that this option can account for the differences in (76 76), the size of these conjuncts must be at least as high as the landing site of quantifier movement in German. Inverted scope in German has been argued to be a property of contrastive topics (B¨ uring B¨ uring, 1997 1997), which might be associated with a higher syntactic position independently. If sondern coordinates sub–sentential conjuncts, that would result in quantifier movement from only one conjunct (the negation quantifier raising to take scope over the sentence) to be ungrammatical since it might require movement from only one conjunct, which is violated by the Coordinate Structure Constraint. In addition, negation as an adverbial is independently not expected to undergo movement, supporting the assumption that the inverted scope is due to movement of the subject. Under the hypothesis of different coordination sizes, the sentence (76b 76b) is not a small coordina- tion of DPs, but rather a constituent containing the quantificational determiner jeder ‘every’. Due to this syntactic structure, inverted scope is possible. This assumption may have consequences in the available information structures of the sentence: it has been proposed that inverted scope is only possible with contrastive topics, and that the availability of the contrastive topic interpretation is due to a higher landing site available for focus movement in sentences like (76b 76b). Another hypothesis is that the pattern above is a reflection of the distinct relations to negation the two coordinators have. Sondern requires negation to be very local, it may not allow for covert movement of the negation, since sondern and the negation are required to be in a specific syntactic relation. This required close syntactic relation different or does not exist with aber, which is 2 In German, inverted scope is made accessible though a rise–fall intonation. (B¨ uring B¨ uring, 1997 1997) 58 Chapter 2. The syntax of German correction 59 evidenced by the fact that sentences with aber, albeit with a different interpretation, can occur without negation. Comparing sondern/aber with the scope relations on other coordinations like und ‘and’, it is found that in the latter case, only low negation scope is possible when it is ascertained that the negation is realized within a conjunct. (When realizing negation in the first conjunct, we cannot be sure whether negation is realized outside the coordination, therefore, it is realized in the second conjunct with und). This parallelism makes it likely that a movement restriction such as the coordinate structure constraint is the underlying reason for the lack of the wide negation reading in (76a 76a) above. (77) Jeder Everybody isst eats Birnen pears und and nicht not ¨ Apfel. apples. Everybody eats pears, and nobody eats apples. Unavailable: Not everybody eats pears and apples. Since this establishes the availability of inverted scope as a criterion, immediately the next question opens up: is the availability of inverted scope reading restricted by the different syntactic structures we considered? The paradigm in (78 78) through (80 80) investigates this question. At the time of writing, these judgments are not fully established. (78) “Full left” a. Jeder Everybody hat has nicht not ¨ Apfel apples gegessen, eaten, sondern but Birnen. pears. X inverted scope b. Jeder Everybody hat has nicht not ¨ Apfel apples gegessen, eaten, aber but Birnen. pears. X inverted scope (79) “Full right” a. Jeder Everybody hat has nicht not ¨ Apfel, apples sondern but Birnen pears gegessen. eaten. * inverted scope b. Jeder Everybody hat has nicht not ¨ Apfel, apples, aber but Birnen pears gegessen. eaten. X inverted scope (80) a. Sandra Sandra hat has nicht not gesehen, seen dass that jeder every Student, student sondern but jeder every Professor professor ¨ Apfel apples gegessen eaten hat. has. X inverted scope 59 60 2.3. Sondern/aber in single corrections b. Sandra Sandra hat has nicht not gesehen, seen dass that jeder every Student, student aber but jeder every Professor professor ¨ Apfel apples gegessen eaten hat. has. X inverted scope (81) a. Nicht Not jeder every Student, student sondern but jeder every Professor professor hat has ¨ Apfel apples gegessen. eaten. b. Nicht Not jeder every Student, student aber but jeder every Professor professor hat has ¨ Apfel apples gegessen. eaten. (82) a. Nicht Not jeder every Student student sondern but Professor professor hat has ¨ Apfel apples gegessen. eaten. b. Nicht Not jeder every Student student aber but *(jeder) every Professor professor hat has ¨ Apfel apples gegessen. eaten. Again, these data are a testimony for the distinction between the “full right” structures and the “full left structures”. In the Verb second sentence above, it is obscured whether or not the structure is derived from a “full left” or a “full right” structure, but we see that it patterns with the “full right” sentences. The full left and full right sentences support the observations made so far. Since the coordination in the full left sentences are derived from movement from a location preceding the verb to a location after the verb, the negation in the left conjunct can move without inhibition after the sondern and aber clauses have been right–extraposed. In case of the full right conjuncts, these data support the hypothesis that corrections with aber have a larger conjunct size and are derived with deletion, which is why inverted scope is an option, since either the coordinate structure constraint is not violated, or the conjunct size is large enough. The data regarding inverted scope support the conclusions drawn in the earlier discussion. There is good evidence that “full right” corrections behave differently from the other types of correction. There are several hypotheses that were entertained about the mechanism behind this restriction. On the one hand, the reason might be coordination size, whereas on the other hand, the movement restriction might be due to the coordinate structure constraint preventing an operator from moving from within the conjunct. 2.3.6 Adverbial scope So far most tests which were geared towards verifying which phrases can minimally occur in correction, and there was much evidence in favor of small conjuncts, that is coordination of DPs, occurring in corrections without the need for ellipsis in certain constructions. It is also important to find out whether or not there is a maximal boundary for phrases that can occur in corrections. 60 Chapter 2. The syntax of German correction 61 Generally, data was presented showing that Complementizer Phrases could occur in corrections, with the negation being realized in the matrix clause. However, this type of data is unable to address the question of how high the negation can scope within a single sentential domain. As a consequence, so far it is an open question to which extent high scoping operators like adverbs can be targets of corrections. Under a cartographic view of adverbials, this will indicate that correction can coordinate phrases at least as large as the projection in question. In the paradigm below, I present data with manner adverbs, epistemic adverbs, and temporal adverbs respectively. All examples under a) are examples where negation is realized immediately following the adverb, whereas in the examples in b), negation is realized immediately preceding the verb. This adverbial paradigm investigates whether these adverbs can be construed as correction or not, implying that the conjuncts are at least as large as the projections that host the respective adverbials. Since the negation in correction must scope higher than said adverb, it also indicates that the negation is or can be raised to this high position. It is hypothesized that this means negation is realized in a structurally higher position in the examples in a) as opposed to in b), since in the latter examples negation likely only takes scope over the verb (barring other causes of opacity like reconstructions). Consequently, negation can be relatively low in the a) examples of the manner adverbs, since it only needs to c–command the vP, medium height with the a) examples of the temporal adverb cases, since it needs to c–command TP, and very high with the epistemic adverb cases, that is c–commanding the CP or a higher projection. While not all the word orders in the upcoming examples are unmarked, they are grammatical, as this paradigm serves to illustrate the possibility of correcting certain phrases. Some of the resulting word orders are not possible out of the blue, but being restricted to a particular context is a trait of correction overall. First, manner adverbs are considered. Nothing out of the ordinary is found with manner adverbs, they can occur in all sentence types independent of negation scope. Like all other multiple foci, they do not allow fronting (or obvious gaps in the first conjunct). Under the assumption that negation is the sister of sondern, we will therefore have to assume that polarity or the relevant feature in question is expressed in a higher projection. (83) Manner Adverbs a. i. Sigrid Sigrid hat has nicht not komplett completely das the Buch book gelesen, read, sondern/aber sondern/aber teilweise. partly. ii. Sigrid Sigrid hat has nicht not komplett, completely, sondern/aber sondern/aber teilweise partly das the Buch book gelesen. read iii. Nicht Not komplett, completely, sondern/aber sondern/aber teilweise partly hat has Sigrid Sigrid das the Buch book gelesen. read. 61 62 2.3. Sondern/aber in single corrections b. i. ? Sigrid Sigrid hat has komplett completely das the Buch book nicht not gelesen, read, sondern/aber sondern/aber teilweise. partly. ii. ? Sigrid Sigrid hat has komplett completely nicht, not, sondern/aber sondern/aber teilweise partly das the Buch book gelesen. read. iii. Nicht Not komplett, completely sondern sondern/aber teilweise partly hat has Sigrid Sigrid das the Buch book gelesen. read Interestingly, we see there is a clear contrast between aber and sondern cases with epistemic adverbials, as the examples in (84 84) demonstrate. (84) Epistemic Adverbs a. i. Sigrid Sigrid hat has nicht not anscheinend seemingly das the Buch book gelesen, read, sondern/aber sondern/aber sicherlich. certainly ii. Sigrid Sigrid hat has nicht not anscheinend, seemingly, sondern/aber sondern/aber sicherlich certainly das the Buch book gelesen. read. iii. Nicht Not anscheinend, seemingly, sondern/aber sondern/aber sicherlich certainly hat has Sigrid Sigrid das the Buch book gelesen. read. b. i. Sigrid Sigrid hat has anscheinend seemingly das the Buch book nicht not gelesen, read, *sondern/aber sondern/aber sicherlich. certainly ii. * Sigrid Sigrid hat has anscheinend seemingly nicht, not, *sondern/aber sondern/aber sicherlich certainly das the Buch book gelesen. read. iii. ? Anscheinend Seemingly nicht, not, sondern/aber sondern/aber sicherlich certainly hat has Sigrid Sigrid das the Buch book gelesen. read Where negation is realized to the right of the epistemic adverbial anscheinend ‘seemingly’, sondern is barred from appearing as a coordinator. The sentences can be rendered grammatical by adding another negation to the first conjunct. We know that sondern has a specific semantic relation with negation, which will be discussed more explicitly in chapters 3 3 and 4 4. But from these data we can conclude that negation has to be realized in a specific syntactic configuration with sondern, otherwise the sentence is not well–formed. Other reminiscent data with aber were shown in the preceding discussion, in which aber could not be construed as a correction when negation was part of a matrix clause, for example. Here we see that negation cannot at the same time scope over epistemic adverbs and be in the left conjunct of sondern. What specifically causes the ungrammaticality in sentence (84b-ii 84b-ii) compared to its grammatical counterpart in (84a-ii 84a-ii)? In the version with sondern, in the ungrammatical cases, the auxiliary in C is followed by the auxiliary, from which we can conclude that the adverb is realized in a position lower than this. We can conclude from acceptability of the other adverbs (temporal adverbs, for example), that it is not problematic to establish the relation between sondern and the negation (whatever the specific nature of this relation), so it is either a problem of how high the negation can 62 Chapter 2. The syntax of German correction 63 covertly be raised, or that an epistemic adverb specifically intervenes in a necessary semantic or syntactic movement or agreement process. However, since aber has a different negation requirement, this requirement is fulfilled in the sentences in (84 84). Specifically, since the adverb seemingly ‘anscheinend’ is located in a syntactically very high position, and the negation would need to c–command this position to take scope over it, apparently this is not possible in the examples given. Another possibility of explaining these facts is that the corrective coordination occurs at a very low level, and the adverb cannot reconstruct into this coordination. Finally, here are the temporal adverbs. They do not show any unexpected pattern and are thus in line with the manner adverbs. (85) Temporal Adverbs a. i. Sigrid Sigrid hat has nicht not gestern yesterday das the Buch book gelesen, read, sondern/aber sondern/aber heute. today. ii. Sigrid Sigrid hat has nicht not gestern, yesterday sondern/aber sondern/aber heute today das the Buch book gelesen. read. iii. Nicht Not gestern, yesterday, sondern/aber sondern/aber heute today hat has Sigrid Sigrid das the Buch book gelesen. read. b. i. Sigrid Sigrid hat has gestern yesterday das the Buch book nicht not gelesen, read, sondern/aber sondern/aber heute. the today. ii. Sigrid Sigrid hat has gestern yesterday nicht, not, sondern/aber sondern/aber heute today das the Buch book gelesen. read. iii. Nicht Not gestern, yesterday, sondern/aber sondern/aber heute heute hat has Sigrid Sigrid das the Buch book gelesen. read. These data show clearly that there is a structural restriction which concerns exclusively sondern, but some additional background on these structural positions must be considered. In particular, Frey Frey (2004 2004) argues that these types of adverbs, due to being adjoined to TP, prevent scrambling that is otherwise possible. 2.3.7 Locating negation Frey Frey (2004 2004) argues that there is an intermediate topic position located immediately higher than the sentence level adverbs which show interaction with sondern. Adverbs occurring higher than this position are gl¨ ucklicherweise ‘fortunately’, anscheinend ‘apparently’, sicherlich, ‘certainly’. Temporal and locative adverbs, which are also realized high in the clause, do not interact in this way with topicality. He assumes these adverbs are adjoined to TP, but at the same time he does not assume a vP projection. I will assume these adverbs are adjoined immediately higher than 63 64 2.3. Sondern/aber in single corrections default negation position. By adjoining in this location, they prevent the expression of negation immediately below it to take higher scope. According to him, these adverbs form a natural class. They adjoin to a sentence at a level without presuppositional content. Consider the example in (86 86) that he provides (Frey Frey, 2004 2004, ex.7). The conclusion that he draws from this example is that topics must not be in the surface scope of these types of adverbs. I follow him in the observation that there is a scopal restriction between these adverbs and negation, but assign it a different explanation. From his analysis it is concluded that the adverbs can only take scope over a particular component if said component is in focus. He ensures that these components are discourse new, which contributes one type of focus. (86) Ich I erz¨ ahle tell dir you etwas something ¨ uber about Maria. Maria. a. N¨ achstes Next Jahr year wird will Maria Mary wahrscheinlich probably nach to London London gehen. go Next year Mary will probably go to London b. # N¨ achstes Next Jahr year wird will wahrscheinlich probably Maria Maria nach to London London gehen. go. As such, I disagree with his judgment presented here. The example (86b 86b) can be acceptable if there is prosodic focus on Maria. According to my judgment, the prosodic focus in example (86a 86a) is on London. There is a parallelism between these adverbs and nicht: all of them can associate with focus, which Frey Frey (2004 2004) also observes and credits Jacobs Jacobs (1986 1986) with the observation. It is diagnosed by them being able to form a constituent and move to the Vorfeld. For a similar insight and analysis also see B¨ uring and Hartmann B¨ uring and Hartmann (2001 2001). Consider these examples to illustrate. The example in b) is provided by myself. (87) a. Dem theDAT OTto Otto anscheinend/anscheinend apparently dem the OTto Otto hat has sich REFL Maria Maria anvertraut. confided–(in) Frey Frey (2004 2004, 6) b. Nicht Not dem the.dat Otto Otto hat has sich self Maria Maria anvertraut, confided-in, sondern but dem the.dat Erich. Erich. It is proposed that the reason that negation cannot outscope these adverbs is because both of these components are necessarily focus–associative particles. This means they both necessarily take a context set as their argument, and introduce a presupposition which situates the prejacent, that is, the sentence realized without the focused particle, with respect to the focus alternatives of this sentence. The theoretical background of these assumptions is discussed extensively in chapter 3 3. For now, suffice it to sketch this analysis: 64 Chapter 2. The syntax of German correction 65 (88) a. JnichtK (C, p, w): i. asserts not p in w ii. presupposes an alternative in C is true in w b. JwahrscheinlichK (C, p, w): i. asserts likely p in w ii. presupposes that other alternatives in C are unlikely true in w Due to these presuppositions, using focus associative particles like wahrscheinlich and nicht to associate with the same instance of focus introduces two presuppositions which are at odds with each other: while nicht presupposes another alternative than the prejacent is true, wahrscheinlich ‘likely’ presupposes that other alternatives are unlikely true. Looking back at the examples in (84 84), here repeated as (89 89) (with additional marking of the pitch accents), there is now an explanation for the pattern that was encountered. Assuming both anscheinend ‘seemingly’ and not ‘nicht’ obligatorily associate with focus, in the examples in (89a 89a), nicht associates with seemingly, and seemingly itself associates with the VP, so they do not have conflicting presuppositions since they do not apply to the same context set. In the examples in (89b 89b), seemingly is attempting to associate with nicht, but the second conjunct does not provide a relevant consequent, and additionally nicht appears to be not in the relevant position for its semantics to be construed together with sondern. (89) Epistemic Adverbs a. i. Sigrid Sigrid hat has nicht not ANSCHEINEND seemingly das the Buch book gelesen, read, sondern/aber sondern/aber sicherlich. certainly ii. Sigrid Sigrid hat has nicht not ANSCHEINEND, seemingly, sondern/aber sondern/aber sicherlich certainly das the Buch book gelesen. read. iii. Nicht Not ANSCHEINEND, seemingly, sondern/aber sondern/aber sicherlich certainly hat has Sigrid Sigrid das the Buch book gelesen. read. b. i. Sigrid Sigrid hat has anscheinend seemingly das the Buch book NICHT not gelesen, read, *sondern/aber sondern/aber sicherlich. certainly ii. * Sigrid Sigrid hat has anscheinend seemingly NICHT, not, *sondern/aber sondern/aber sicherlich certainly das the Buch book gelesen. read. iii. ? Anscheinend Seemingly nicht, not, sondern/aber sondern/aber sicherlich certainly hat has Sigrid Sigrid das the Buch book gelesen. read A further question is why aber is not subject to these restrictions. The idea proposed here and more in more depth in chapter 4 4 is that when aber occurs as a correction, it also necessarily corrects a proposition in discourse. When it does not occur with correction, it does not refer back 65 66 2.3. Sondern/aber in single corrections to discourse. As such, negation in aber has two possible interpretation. Sondern does not have this optionality. With sondern, all negations must be interpreted as the correction of a discourse proposition. As will be explained in more detail, the construal of a discoursively active negation involves syntactic movement. This type of movement is blocked for the sentences with sondern due to the interaction with anscheinend ‘seemingly’, and due to a restricted coordination size. In the example in (89b-ii 89b-ii) above, since negation need not necessarily relate to discourse, no movement is necessary. 2.3.8 Interim conclusion Two major goals were set at the beginning of this chapter. On the one hand, it was a major goal to find out as precisely as possible about the syntactic structure, with particular focus on coordination height. On the other hand, finding the differences between sondern and aber was another major goal. In general, for both types of correction, it was necessary to take into account known facts about German grammar, since different syntactic structures had different restrictions and underlying structures for correction. I will discuss the syntactic findings by construction type. “full right” corrections sondern When the verb was in final position, and the negation was constituent negation, the data for this structure was consistent with low coordinations which did not involve ellipses. It appears this configuration is the underlying structure for many of the other sentence types. Only in this configuration were there certain exceptions, causing negation to occur exceptionally outside the scope of sondern. Furthermore, it was shown that this type of coordination needed to be at a syntactic level lower than CP level, since negation could not scope over CP–level adverbs. In the kind of “full right” coordination where the verb was not in final position, like the unambiguous Right Node Raising structures, there was no evidence discovered showing that a standard analysis of Right Node Raising as offered in Hartmann Hartmann (2000 2000) is not applicable. However, in corrections where non–arguments were the target of correction, the deletion process was shown to be different from Right Node Raising. aber In “full right” sentences with aber, the exceptions with sondern were not found, aber was never ambiguous about scoping lower than negation. For example, aber could not occur in this position when negation had sentential scope. full left corrections In general, in this configuration, no differences between sondern and aber were encountered. However, we found that this type of correction generally involved ellipsis, 66 Chapter 2. The syntax of German correction 67 suggesting a higher coordination level than “full right” corrections. In the literature, it was suggested than clauses following the main verb are moved from a position preceding the verb. In parallelism with the literature I assume that sondern/aber+XP is moved from a position in the left conjunct, hence derived from the “full right” correction type. fronted In the discussion so far, the underlying structure about the fronted data pointed in the direction of the fronted correction involving ellipsis. This was the result of the island tests, and also the agreement phenomena. However, assuming such a syntactic structure relies on the assumption of a backwards deletion process, which was only described for verb final structures. This is a reason for rejecting a high coordination analysis for these sentences. Overall, the differences between sondern and aber were mostly found in the “full right” types of correction. Setting aside that the two conjuncts have subtly different discourse requirements, two types of configuration were found in which a negation which was realized farther away could be construed into a correction with sondern, and not with aber. The relevant island example, repeated from (64 64), is repeated below. (90) Alfons Alfons hat has nicht not Reis rice und and Bohnen, beans sondern/*aber but Kartoffeln potatoes gekocht. cooked. Among available readings for sondern: Alfons didn’t cook rice and beans, but rice and potatoes. Under the assumption that the island effect is due to focus movement in the second conjunct, it was argued that in these cases, sondern appeared to be in the scope of negation. There were two possible explanations suggested for this fact. Either aber could not occur in these exceptional constructions, because it has grammatical restrictions against being in the scope of negation (and hence is a PPI), or it is due to a different coordination size of sondern and aber in this construction. It is possible that only sondern can scope low enough to occur in a coordination of DPs, whereas aber needs to take higher scope. Furthermore, the scope relations in sondern and aber corrections differed in that aber allowed for inverted scope in certain coordinations. Likewise, this fact can also be explained if we assume a lower coordination level from sondern than from aber. However, another explanation that may be just as illuminating is to assume that the negation is scope rigid with respect to sondern since there is a polarity licensing relation between the two. The same two solutions are open for the adverbial scope differences. On the one hand, one might argue that there is a coordination height difference between sondern and aber, but on the other hand, one could argue that there is polarity licensing that is not met in these cases. 67 68 2.3. Sondern/aber in single corrections 2.3.9 Consequences of this proposal for negation In the current work, it is proposed (although the evidence presented is by no means complete) that discourse–active negation, that is, negation which either needs to refer back to a nonnegated sentence, or must presuppose such a sentence (for more details please refer to chaper 3 3) has a different syntactic representation than sentential negation. It has been observed in the literature Han and Romero Han and Romero (2004 2004) that there is a relation between where the negation is syntactically realized in a sentence, and its discourse interpretation. This is considered a crucial observation in the analysis of discourse negation. The relevant examples are repeated below. The insight about alternative questions in English is originally due to Ladd Ladd (1980 1980). These examples show that for alternative questions in German, when negation is realized in a location different than a location adjoined to VP, a discourse bias is introduced. (91) a. Didn’t Chris go to class? biased towards a positive answer b. Did Chris not go to class? not biased (92) Han and Romero Han and Romero (2004 2004, 7) a. Hat has (nicht) Neg Hans Hans (nicht) Neg Maria Maria gesehen? seen Didn’t Hans see Maria? biased towards a positive answer b. Hat Has Hans Hans Maria Maria nicht not gesehen? seen? Did Hans not see Maria? not biased The same type of variability is found in assertive clauses. Where negation is realized away from its default location in a sentence, it is discoursively marked, that means, it conveys a presupposition about discourse. The nature of this presupposition is described in chapter 3 3. (93) Sandra Sandra hat has Erich Erich ¨ Apfel apples gegeben. given a. Sandra hat Erich keine ¨ Apfel gegeben. neutral b. Nicht Sandra hat Erich ¨ Apfel gegeben. Presupposition: It is assumed that (93 93) c. Sandra hat nicht Erich ¨ Apfel gegeben. Presupposition: It is assumed that (93 93) d. Sandra hat Erich nicht ¨ Apfel gegeben. Presupposition: It is assumed that (93 93) e. Sandra hat Erich ¨ Apfel nicht gegeben. Presupposition: It is assumed that (93 93) I assume that this presupposition has a syntactic and a semantic correlate. Syntactically, this presupposition arises when a negation which is not realized in its base location covertly moves to its base location, that is, adjoining Spec of vP. Whether or not negation occurs in 68 Chapter 2. The syntax of German correction 69 its base position is a predictor of whether the discourse interactive component is activated. When negation is realized adjoining to Spec of vP without movement, it is unmarked. When movement to Spec of vP happens, we can see discoursive markedness, that is, the introduction of a presupposition which is dependent on the base location of negation. When the conjunct encompasses Spec of vP, this movement can happen only within the negated conjunct, but in small coordinations like DP coordinations, the entire correction (e.g. not apples but pears) may move, since they presupposition required for the interpretation of correction is clearly composed at a higher level. For example, a sentence like Sandra hat keine ¨ Apfel, sondern Birnen gegessen ‘Sandra ate no apples, but pears’, has a presupposition Sandra ate apples, despite the fact that there is no component Sandra ate apples, but rather, due to the corrective coordination occurring at a higher location, the presupposition is construed at a higher level. Semantically, this presupposition is computed by association with focus, that means, the context set within the negated conjunct in a correction will be evaluated. Independently, association with focus has also been argued to rely on syntactic movement (Wagner Wagner, 2006b 2006b,a a). Since it was convincingly argued that not all corrections are propositional in the preceding sections, and there is good evidence that some DP level coordinations contain no ellipsis, corrections which are uninterpretable in their embedded object position move to a position higher in the sentence, at which point the presupposition will be calculated. (As example (93 93) indicates, the presupposition is a complete sentence. ) In the case of sondern, the negation movement is required, whereas in sentences with aber, it is only required when narrow scope is realized within the first conjunct. These assumptions give a specific syntactic interpretation to a classical distinction made in the theory of negation, namely the distinction between metalinguistic negation and descriptive negation proposed in Horn Horn (1985 1985). 2.4 Conclusion In this chapter, the question why corrections appear to involve smaller coordinations than counterex- pectationals was addressed, and how this type of intuition should be captured in the grammar of these two constructions. It was argued that coordinations with sondern were smaller than coordinations with aber. Related questions were whether differing grammatical properties of the two resembling coordi- nators sondern and aber could be reasons for this impression. In the course of this discussion, it was considered which deletion processes were involved (VP ellipsis, gapping, Right Node Raising), and how correction interfaces with known properties of German syntax. It was argued that there were some restrictions as regards the negation represented in correction. When negation was realized away from its default position, it needed to move to the default position 69 70 2.4. Conclusion to agree. 2.4.1 What is the specific syntactic structure for simple correction? In the following table 2.6 2.6 on page 76 76, I summarize the major arguments for the grammatic structure of Object DP–level correction with sondern and aber respectively, and propose some syntactic structures. Overall, it needs to be maintained that a crucial difference between coordinations with sondern and aber is that the former cannot coordinate phrases larger than vP, which was evidenced by the adverbial data. Corrections with aber can coordinate also larger constituents, which is not pictured here. In these examples, where a higher copy of negation is crossed out, this is meant to represent a movement process on LF, and therefore a discoursively marked negation. The trees in the table are a simplification, in fact it is likely that the entire corrective coordination moves to establish this syntactic relation. Appendix 2.4.2 Subcategorization tests: simple correction, local negation (94) N(P): I have seen the young not dog but wolf. a. * Ich I habe have den the jungen young nicht not Hund dog gesehen, seen, sondern/aber sondern/aber Wolf. wolf. b. ? Ich I habe have den the jungen young nicht not Hund dig sondern/aber sondern/aber Wolf wolf gesehen. seen c. */?? Nicht Not Hund dog sondern/aber sondern/aber Wolf wolf habe have ich I den the jungen young gesehen. seen. (95) D: I have seen not three but two young dogs. a. Ich I habe have nicht not drei three sondern/aber sondern/aber zwei two junge young Hunde dogs gesehen. seen. b. Ich I habe have nicht not drei three junge young Hunde dogs gesehen, seen, sondern/aber sondern/aber zwei. two. c. * Nicht Not drei, three sondern/aber sondern/aber zwei two habe have ich I junge young Hunde digs gesehen. seen. (96) P: I haven’t voted for but against the proposal. a. Ich I habe have nicht not f¨ ur, for, sondern/??aber sondern/aber gegen against den the V orschlag proposal gestimmt. voted. b. * Ich I habe have nicht not f¨ ur for den the V orschlag proposal gestimmt voted sondern/aber sondern/aber gegen. against 70 Chapter 2. The syntax of German correction 71 c. * Nicht Not fur for sondern/aber sondern/aber gegen against habe have ich I den the V orschlag proposal gestimmt. voted. (97) T: I believe that Sandra will not play but can play. a. Ich I glaube, believe dass that Sandra Sandra nicht not spielen play wird, will, sondern/??aber sondern/aber kann. can. b. ? Ich I glaube, believe dass that Sandra Sandra spielen play nicht not wird, will, sondern/aber sondern/aber kann can. okay if spielen is a contrastive topic. c. * Nicht Not spielen play wird, will, sondern/aber sondern/aber kann can glaube believe ich I dass that Sandra. Sandra d. * Nicht Not wird will sondern/aber sondern/aber kann can glaube believe ich I dass that Sandra Sandra spielen. play. (98) C a. Ich habe nicht ob sondern/aber warum Sandra den Hund gef¨ uttert hat gefragt. b. ?? Ich habe gefragt nicht ob, sondern/aber warum Sandra den Hund gef¨ uttert hat. c. Nicht ob sondern/aber warum Sandra den Hund gef¨ uttert hat, habe ich gefragt. d. Nicht ob Sandra den Hund ge¨ uttert hat, sondern/aber warum habe ich gefragt. (99) Object DP a. Ich habe nicht den jungen Hund gesehen, sondern/aber den Wolf. b. Ich habe nicht den jungen Hund sondern/aber den Wolf gesehen. c. Nicht den jungen Hund, sondern/aber den Wolf habe ich gesehen. (100) Indirect Object DP a. Ich habe nicht dem jungen Hund einen Knochen gef¨ uttert, sondern/aber dem Wolf. b. Ich habe nicht dem jungen Hund sondern/aber dem Wolf einen Knochen gef¨ uttert. c. Nicht dem jungen Hund, sondern/aber dem Wolf habe ich einen Knochen gef´ uttert. (101) Subject DP a. Nicht der junge Hund hat die Maus gejagt, sondern/aber der Wolf. b. Nicht der junge Hund sondern/aber der Wolf hat die Maus gejagt. c. Erika hat gesehen, dass nicht der junge Hund, sondern/aber der Wolf die Maus gejagt hat. d. Erika hat gesehen, dass nicht der junge Hund die Maus gejagt hat, sondern/aber der Wolf. (102) PP 71 72 2.4. Conclusion a. Ich habe nicht f¨ ur den V orschlag, sondern/?aber gegen den V orschlag gestimmt. b. Ich habe nicht f¨ ur den V orschlag, sondern/?aber gegen den V orschlag gestimmt. c. Nicht f¨ ur den V orschlag, sondern/?aber gegen den V orschlag habe ich gestimmt. (103) AP a. ? Ich habe den nicht jungen sondern/aber alten Hund gesehen. b. * Ich habe den nicht jungen Hund gesehen sondern/aber alten. c. * Nicht jungen sondern/aber alten habe ich den Hund gesehen. (104) VP a. Ich habe den jungen Hund nicht gesehen, sondern/aber gerochen. b. Ich habe geh¨ ort, dass Anna den jungen Hund nicht gesehen, sondern/aber gerochen hat. c. Ich habe geh¨ ort, dass Anna den jungen Hund nicht gesehen hat, sondern/aber gerochen. d. Nicht gesehen, sondern/aber gerochen habe ich den jungen Hund. (105) vP Here, the idea is that if the auxiliary is high and everything else is in situ, presumably the remainder is in vP a. Es hat nicht Erika den jungen Hund gesehen, sondern/aber Erich. // Ich habe geh¨ ort, dass nicht Erika den jungen Hund gesehen hat, sondern/aber Erich. b. Es hat nicht Erika, sondern/aber Erich den jungen Hund gesehen. // Ich habe geh¨ ort, dass nicht Erika, sondern/aber Erich den jungen Hund gesehen hat. c. * Nicht Erika den jungen Hund gesehen, sondern/aber Erich hat d. * Nicht Erika hat den jungen Hund gesehen, sondern Erich schon.//X aber Erich schon. e. Nicht Erika sondern/aber Erich hat den jungen Hund gesehen. (106) TP level correction: a. ?? Ich I habe have geh¨ ort, heard, dass that nicht not Erika Erika den the jungen young Hund dog sehen see wird, will, sondern/aber sondern/aber kann. can. b. Ich I habe have geh¨ ort, heard dass that nicht not Erika Erika den the jungen young Hund dog sehen see wird, will, sondern/aber sondern/aber Erika/sie Erika/she den the jungen young Hund dog sehen see kann. can. c. * Nicht Not Erika Erika den the jungen young Hund dog sehen see wird, will, sondern/aber sondern/aber Erika Erika den the jungen young Hund dog sehen see kann can habe have ich I geh¨ ort. heard. (107) CP 72 Chapter 2. The syntax of German correction 73 a. ? Ich habe geh¨ ort, nicht dass Erika nicht den jungen Hund gesehen hat, sondern/aber dass sie wird. / *sondern/aber dass Erich. *sondern/aber dass den Wolf. b. Ich habe nicht geh¨ ort, dass Erika den jungen Hund gesehen hat, sondern/aber (dass) sie es wird. / sondern/aber (dass) der junge Hund Erika gebissen hat./ sondern (*dass) Erich einen Wolf. I believe this is parallel to what is found in regular coordinations. c. nicht, dass Erika den jungen Hund gesehen hat, sondern/aber dass sie es wird, habe ich geh¨ ort. 2.4.3 Subcategorization tests: simple correction, distal negation (108) N(P) a. * Ich habe nicht den jungen Hund gesehen, sondern/aber Wolf. b. ? Ich habe nicht den jungen Hund sondern/aber Wolf gesehen. c. * Hund sondern/aber Wolf habe ich nicht den jungen gesehen. (109) D a. Ich habe nicht drei sondern/aber zwei junge Hunde gesehen. b. Ich habe nicht drei junge Hunde gesehen, sondern/aber zwei. c. * Drei, sondern aber zwei habe ich nicht junge Hunde gesehen. (110) P In the ditransitive sentences, it becomes clear that reordering could make the sentence better, but I stuck with this sentence since it replicated the word order of the out–of–the–blue sentence. a. Ich habe nicht f¨ ur, sondern/??aber gegen den V orschlag gestimmt. b. * Ich habe nicht die Schl¨ ussel unter, sonder/aber auf den Stoß Papier gelegt. c. * Ich habe nicht f¨ ur den V orschlag gestimmt sondern/aber gegen. d. Ich habe nicht die Schl¨ ussel under den Stoß Papier gelegt, sondern auf. e. * Fur sondern/aber gegen habe ich nicht den V orschlag gestimmt. f. * Unter sondern/aber auf den Stoß Papier habe ich nicht den Schl¨ ussel gelegt. (111) T a. Ich glaube, dass Sandra nicht spielen wird, sondern/??aber kann. b. * Spielen wird, sondern/aber kann glaube ich nicht dass Sandra. c. * Wird sondern/aber kann glaube ich nicht dass Sandra spielen. (112) C 73 74 2.4. Conclusion a. Ich habe nicht gefragt ob Sandra den Hund gef¨ uttert hat, sondern/aber warum . b. Ich habe nicht ob sondern/aber warum Sandra den Hund gef¨ uttert hat gefragt. c. */# Ob sondern/aber warum Sandra nicht den Hund gef¨ uttert hat, habe ich gefragt. d. */# Ob Sandra den Hund ge¨ uttert hat, sondern/aber warum habe ich nicht gefragt. (113) Object DP a. Ich habe nicht den jungen Hund gesehen, sondern/aber den Wolf. b. Ich habe nicht den jungen Hund sondern/aber den Wolf gesehen. c. */?? Den jungen Hund, sondern/aber den Wolf habe ich nicht gesehen. (114) Indirect Object DP a. Ich habe nicht dem jungen Hund einen Knochen gef¨ uttert, sondern/aber dem Wolf. b. Ich habe nicht dem jungen Hund sondern/aber dem Wolf einen Knochen gef¨ uttert. c. * Dem jungen Hund, sondern/aber dem Wolf habe nicht ich einen Knochen gef´ uttert. (115) Subject DP ** a. Der junge Hund hat nicht die Maus gejagt, sondern/aber der Wolf. b. * Der junge Hund sondern/aber der Wolf hat nicht die Maus gejagt. c. * Erika hat gesehen, dass der junge Hund, sondern/aber der Wolf nicht die Maus gejagt hat. d. Erika hat gesehen, dass der junge Hund nicht die Maus gejagt hat, sondern/aber der Wolf. (116) PP a. Ich habe nicht f¨ ur den V orschlag, sondern/?aber gegen den V orschlag gestimmt. b. Ich habe nicht f¨ ur den V orschlag, sondern/?aber gegen den V orschlag gestimmt. c. * F¨ ur den V orschlag, sondern/?aber gegen den V orschlag habe ich nicht gestimmt. (117) AP a. # Ich habe nicht den jungen sondern/aber alten Hund gesehen. b. * Ich habe den nicht jungen Hund gesehen sondern/aber alten. c. * Jungen sondern/aber alten habe ich nicht den Hund gesehen. (118) VP a. Ich habe nicht den jungen Hund gesehen, ??sondern/??aber gerochen. b. Ich habe geh¨ ort, dass Anna nicht den jungen Hund gesehen, sondern/aber gerochen hat. c. ? Ich habe geh¨ ort, dass Anna nicht den jungen Hund gesehen hat, sondern/aber ge- rochen. 74 Chapter 2. The syntax of German correction 75 d. * Gesehen, sondern/aber gerochen habe ich nicht den jungen Hund. (119) vP Here, the idea is that if the auxiliary is high and everything else is in situ, presumably the remainder is in vP a. Es hat nicht Erika den jungen Hund gesehen, sondern/aber Erich. // Ich habe geh¨ ort, dass nicht Erika den jungen Hund gesehen hat, sondern/aber Erich. b. Es hat nicht Erika, sondern/aber Erich den jungen Hund gesehen. // Ich habe geh¨ ort, dass nicht Erika, sondern/aber Erich den jungen Hund gesehen hat. c. * Erika den jungen Hund gesehen, sondern/aber Erich nicht hat d. * Nicht Erika hat den jungen Hund gesehen, sondern Erich schon.//X aber Erich schon. e. * Erika sondern/aber Erich hat nicht den jungen Hund gesehen. (120) TP: I put the contrast in this example in the tense. However, the more natural way of expressing this contrast is to contrast the VPs against each other, e.g. “Nicht gesehen hat, sondern sehen wird”. Also, this might be multiple focus, since gesehen hat contrasts with sehen wird. nicht gesehen hat, sondern (*gesehen) worden ist. a. Ich habe geh¨ ort, dass Erika nicht den jungen Hund gesehen hat, sondern/aber wird. b. Ich habe geh¨ ort, dass Erika nicht den jungen Hund gesehen hat, sondern/aber Erika/sie den jungen Hund sehen wird. Since the second clause is still verb end we see it is still embedded, in such sentences, a coordinator cannot be omitted. (* Ich habe geh¨ ort Erika den jungen Hund sehen wird.) c. * Erika den jungen Hund gesehen hat, sondern/aber Erika nicht den jungen Hund sehen wird habe ich geh¨ ort. (121) CP a. ? Ich habe geh¨ ort, dass Erika nicht den jungen Hund gesehen hat, sondern/aber dass sie wird. / *sondern/aber dass Erich. *sondern/aber dass den Wolf. b. Ich habe geh¨ ort, dass Erika nicht den jungen Hund gesehen hat, sondern/aber (dass) sie es wird. / sondern/aber (dass) der junge Hund Erika gebissen hat./ sondern (*dass) Erich einen Wolf. I believe this is parallel to what is found in regular coordinations. c. Dass Erika den jungen Hund gesehen hat, sondern/aber dass sie es wird, habe ich nicht geh¨ ort. 75 76 2.4. Conclusion Table 2.6: Syntactic structures SONDERN ABER Full Left Correction Sandra has not A eaten sondern B Inverted scope possible Island effects with Sentential Negation CP C hat vP vP t Sandra B eaten sondern vP VP nicht A eaten t Sandra nicht hat Sandra CP C hat vP vP t Sandra B eaten sondern vP VP A eaten t Sandra nicht hat Sandra Sandra has not A eaten aber B Island effects independent of Negation type Inverted scope possible Smallest coordination: vP, larger ones possible CP C hat vP vP t Sandra B eaten aber VP A eaten t Sandra nicht hat Sandra Full Right Correction Sandra has not A sondern B eaten No Island effects No inverted scope Exceptional position of Negation CP C hat vP eaten nicht A sondern B t Sandra nicht hat Sandra Sandra has not A aber B eaten Inverted Scope possible Island effects with sentential negation Here both DP coordination and vP+ coordination possi- ble Fronted Correction Not A sondern B has Sandra eaten No island effects No inverted scope Exceptional position of nega- tion Not A aber B has Sandra eaten Island effects with sentential negation May be object coordination or larger coordination. 76 Chapter 3 Semantics and pragmatics ofsondern 3.1 Introduction In this chapter, an in depth analysis of the discourse properties of correction with sondern is presented. It is shown that correction has three crucial minimal discourse properties. First, it must presuppose a proposition that has the inverse polarity of the negated conjunct, second, at first glance it appears to be lacking scalar implicatures, and lastly, the two conjuncts must contradict each other. Its near–synonym aber need not obligatorily presuppose a proposition of the inverse polarity, and displays scalar implicatures as expected. Aber also only requires that the conjuncts contradict each other in certain narrow scope corrective environments, but not in all environments. These phenomena are accounted for as an interaction between a polarity focus operator, an exhaustification operator, and an obligatory presupposition requiring that the second conjunct be minimally different from the first. The first conjunct contains a polarity focus operator, which is modeled after the semantics of verum focus. Furthermore, the pragmatic effects of embedded exhaustification, a process which is argued necessary to understand the inferences in coordination plays an important role. The second conjunct has a presupposition which ensures contrastivity. This presupposition is part of the lexical entry of sondern, which associates with two focused constituents. The current discussion mainly follows the framework of alternative semantics in the version of Rooth Rooth (1992 1992), with some additional adjustments to account for pragmatic effects. I also rely on discussions of polarity focus in Han and Romero Han and Romero (2001 2001, 2004 2004). 78 3.1. Introduction An example of a corrective discourse is presented in (122 122). There are two significant discourse requirements in corrections: the first requirement is that there is evidence that the stressed con- stituents have to be in contrast in a syntactic position within the scope of negation. The second requirement is that the first conjunct needs to have opposite polarity to an actually realized or accommodated discourse proposition. Another puzzling fact is that scalar implicature effects, which might be predicted a priori for scalar items, are not encountered. The reason the effects are expected is that they occur with related coordinators like and. In this chapter, the presence and scope of focus operators, as well as other semantic and pragmatic processes, is inferred from these observations. (122) A. Sandra ate apples for breakfast yesterday. B. i. Sandra didn’t eat APPLES, but PEARS. ii. Sandra Sandra hat has nicht not ¨ APFEL, apples, sondern but BIRNEN pears gegessen. eaten. (123) A. A(ntecedent) B. (: C1), but/sondern C2 In the following sections, I will show that these three observations are crucial components of corrections. A generalized form of correction is given in (123 123). A terminological clarification needs to be made at this point. It will be useful to be able to refer to the first conjunct including negation, but also the first conjunct excluding negation. When using the term “first conjunct”, I refer to the entire conjunct including negation. When using the abbreviation C1, I refer to the first conjunct without negation. Analogously, “second conjunct” refers to the coordinator in addition to the other content, whereas C2 refers only to the content without the coordinator. In this chapter, the author remains impartial as to what the syntactic category of the conjuncts and C1 and C2 is, that is, whether or not they are realized syntactically as phrases or sentences with subsequent deletions. In chapter 2 2 evidence was presented and it was shown this question needs to be discussed on a case by case basis. 3.1.1 Observation 1: C1 contradicts C2 One hallmark of correction is that C1, that is, the first conjunct without negation, necessarily contradicts C2. We also know that the first conjunct necessarily contains negation. As a consequence, the statement (:C1)&C2 will always be considered true and maximally informative by person uttering a correction if they adhere to the Gricean maxims. In addition, it will be shown that additional contrast conditions will need to be introduced to restrict corrections. The semantic description(:C1)&C2 without any additional assumptions restricting the inter- pretation of C1 and C2 has very weak truth conditions. Consider the following example in support of this point: 78 Chapter 3. Semantics and pragmatics of sondern 79 (124) ? Steve didn’t eat an apple, and Steve ate a pear. Without considering the context of utterance, the truth conditions of this sentence are very easy to fulfill, and the sentence barely means anything more than Steve ate a pear without a coordination. It might, for example, be uttered in a situation where Steve grew up in a region of the world where apples cannot be grown, and will thus be true in virtually all situations where Steve ate a pear. Naturally, the two conjuncts must be interpreted as context for the respective other conjunct: Steve didn’t eat an apple in a situation where Steve ate a pear. Without this assumption, the first conjunct is almost trivially true in almost all types of situations, as is a problem for negated sentences in general. As such, that the speaker needs to consider (:C1)&C2 true and maximally informative are very weak felicity conditions. Additional mechanisms need to be implemented ensuring that C1 and C2 are more relevant to each other. The reason a sentence like (124 124) is not a correction is that it does not have one of the hallmark properties of correction. In a discourse licensing correction, only one of C1 and C2 can be true in the context where a correction is uttered. What makes the sentence Steve didn’t eat an apple but a pear different from a mere coordination like in (124 124) is that at the utterance time, the speaker assumes that Steve could have only eaten either an apple or a pear, and not both. More formally, it is a characteristic trait of correction that C1&C2 is false, that is, that C1 (the proposition Steve ate an apple in 124 124) and C2, (Steve ate a pear) contradict each other. It is proposed that a suitable way to explain these facts is to introduce a locality restriction on the interpretation of focus, which will ensure that the contradiction condition is always met in corrections. The following paragraphs show that there is a second restriction that a successful analysis has to account for. 3.1.2 Observation 2: Conjunct one shows polarity discourse contrast In example (122 122), the corrective sentence is uttered in the context of an antecedent sentence, in the example this sentence is Sandra ate apples for breakfast yesterday. One of the characteristic features of correction is that there is a negation, necessarily expressed in the first conjunct in the case of sondern, and this negation is the opposite polarity as its antecedent sentence. Corrections with aber lack the restriction that negation must occur in the first conjunct, but the negated conjunct in these corrections appears to be subject to the same restriction nevertheless. 1 1 This characterization consists of two substatements: on the one hand, corrections necessarily refer back to a statement in discourse. On the other hand, corrections need to have opposing polarity with respect to this statement. 1 A question not discussed here is whether or not in a sentence with aber where the second conjunct carries the negation, like I ate an apple ABER not a pear, the discourse requirement of the second conjunct might be fulfilled by the first conjunct. 79 80 3.1. Introduction To substantiate the first claim, consider this example conversation. The first question is generally viewed as a prototypical question that starts a discourse, that is, a question that does not presuppose any common information between the speaker and the addressee. In the example, it is attempted to answer such a question with a correction. (125) A. What happened? B. # Sandra didn’t eat an apple, but a pear. Accepting the assumption that this question starts a discourse, the answer in B is infelicitous as an answer to the question What happened?. The only way this dialogue can become natural is if it is assumed that the speaker and the addressee have some previous common assumptions which entail the proposition Sandra ate an apple. If this condition is not given, the sentence is infelicitous. The second claim regarding opposing polarity is evaluated in the paradigm in (126 126) below. In this dialogue, a discourse antecedent including a negation is assumed. Three sentences are considered as possible continuations. A correction without negation, a correction with a single negation, and a correction with multiple negation. The examples below show that only one of the utterances is a felicitous follow up to (126 126)A, and it is the utterance with multiple negation that adds a single negation with respect to the already negated antecedent proposition. (126) A: Sandra didn’t eat an apple. a. * B: Sandra ate an apple, but a pear. b. * B: Sandra didn’t eat an apple, but a pear. c. B: Sandra didn’t not eat an apple, but a pear. I have discussed extensively in chapter 2 2 that a sentence without any expression of downward entailingness is ungrammatical as a correction, so the ungrammaticality of (126a 126a) is unsurprising. What is surprising is the grammatical distinction between (126b 126b) and (126c 126c). If (126b 126b) is to be construed as a correction in the context of (126 126)A, it is ill–formed when expressed with only a single negation. This sentence is acceptable in a non–corrective reading, where Sandra didn’t eat an apple is repeated by speaker B, and then additional detail is provided about this situation. In this non–corrective construal, but a pear is an infelicitous continuation in this context. It is more natural utter the uncoordinated clause She ate a pear. Crucially, when attempting to express a correction of a negated sentence, a double negation is required in the first conjunct. This shows very clearly that a discourse as schematized in example (127 127) below constitutes a well–formed correction. (127) A. P B. (: P), but Q A further question that suggests itself is whether it must be this way. In example (126 126), it was demonstrated that a polarity contrast is a sufficient condition for a well–formed correction. It also 80 Chapter 3. Semantics and pragmatics of sondern 81 needs to be shown that is a necessary condition that the only contrast between the first conjunct and the antecedent is in polarity. The example in (128 128) below shows just that. An example where it is attempted to construe a contrast where polarity in addition to the nature of the object is contrasted is judged as infelicitous. Example (129 129) shows the same for a discourse where it is attempted to contrast only the nature of the object, but not polarity. (128) A. Sandra ate an apple. B. # Sandra didn’t eat a pear but a pineapple. (129) A. Sandra didn’t eat a pear. B. # Sandra didn’t eat an apple, but a pineapple. From these examples it can be deduced that it is necessary that correction contrasts with an antecedent statement in discourse in only polarity, and no other properties. A simplification made in examples (126 126) through (129 129) is that such an antecedent statement is phonologically realized in these context, and is word–for–word identical to C1. In actual discourse, this is rare. Rather, the proposition in C1 is entailed by utterances accepted in discourse (that is, propositions in the common ground) but not spelled out explicitly. Specifically in the parliamentary discourse data that I have considered for this work, the speaker utters the correction as a way of suggesting that C1 should be accommodated into the common ground. Below, there are two examples from the corpus of parliamentary proceedings that will illustrate these configurations. (130) Minister Hundstorfer: You must take note of the fact that we aggressively take measures, despite rising unemployment. You must take note of the fact that, whether you like it or not, there has been a reduction in unemployment with the 15 through 19 year olds. Whether you like it or not, you must take note of that. (Interruption by deputy Kickl: I am really not quite sure if all that you are telling us is true!) Sie You m¨ ussen must zur to-the Kenntnis notice nehmen, take, dass that wir we 500 500 Millionen million investieren invest f¨ ur for die the 15- 15 bis through 24-J¨ ahrigen, 24-year-olds, dass that wir we nicht not wegschauen, look-away, sondern but hinschauen. look-at. You need to acknowledge that we invest 500 million for the 15 through 24 year-olds, that we don’t look away [from the problem of youth unemployment], but that we look right it [that we address the problem]. (131) Deputy Sabine Oberhauser, PhD, MAS: What was done in the past few years in the federal government? There were a number of things, for example the employment guarantee. The employment guarantee caused people between 15 and 18 years to be guaranteed to find or 81 82 3.1. Introduction receive an education or a training on a high level. In the year 2013 there were just short of 6700 — interrupted by deputy Belakowitsch–Jenewein: Training on a high level? — deputy Oberhauser: Training is better in any case! Dear deputy Belakowitsch, I prefer every single youth who attends a training which furthers their prospects. Sie You wissen know ganz completely genau, exactly, dass that wir we in in ¨ Osterreich Austria nicht not davon from-it leben, live dass that wir we Billig-Arbeitskr¨ afte cheap-laborforce produzieren, produce, sondern but dass that wir we gut well ausgebildete, educated, gut well geschulte schooled Jugendliche youths haben. have. You know precisely that in Austria we don’t live off producing cheap laborforce, but that we have well educated and well schooled youths. In example (130 130), a previous speaker made the accusation against the government representatives that they take insufficient measures regarding youth unemployment. This accusation was expressed through a number of speeches. Referring back to these speeches, the minister of labor and social affairs utters the sentence in (130 130). As the example shows, there is no literal mention of the Minister “looking away” from the problem, rather, the antecedent proposition is entailed by previous utterances. In example (131 131), there was never a literal utterance stating that Austria lives off producing cheap labor force. However, the (same) speaker uses correction as a rhetorical device to so that their opinion, namely that instead of producing cheap labor force Austria has well–educated youths, is stated against the background of other speakers claiming this fact. 3.1.3 Observation 3: Corrections with sondern do not trigger scalar impli- catures Compared to other types of coordinations, it appears correction with sondern is lacking the effect of scalar implicatures. We can see that corrections with sondern can increase or decrease in informativity. Informativity is clearly defined when propositions stand in an entailment relation with each other. If proposition a entails a proposition b, proposition a is considered more informative. Corrections with aber show sensitivity to the informativity of the coordination. In corrections with aber, C1 (sans negation) must be stronger than C2, that is, C1 without negation must entail C2, and hence be more informative. The same sensitivity is not given with sondern. The reason that it is unexpected is observation 1, the requirement that the conjuncts contradict each other. Observation 1 will be shown to be an effect based on an exhaustivity implicature. While the formal definition for this implicature will be given later on, intuitively, it means that the sentence seen in (124 124) is in fact interpreted with some additional interpretive effects, marked with italics in 82 Chapter 3. Semantics and pragmatics of sondern 83 (132 132) below. The operator responsible for this effect, the exhaustivity operator, is known to interact with scalar arguments. In the corrections with sondern, it appears this scalar interaction with the exhaustivity operator is missing. (132) Steve didn’t eat an apple and nothing else, and Steve ate a pear and nothing else. Consider the following paradigm in (133 133). First, it needs to be established that the two conjuncts have no entailment relationships with one another, that is, neither does eating not some strawberries entail eating all strawberries, not does eating not all strawberries entail eating some strawberries. As such, entailment relations can not be utilized state this generalization. However, we find that in corrections with sondern and aber, there are different restrictions between which entailment relations are allowed between C1, without negation, and C2. While sondern allows both that C2 entail C1 in (133a 133a), and that C1 entail C2, as in (133b 133b), aber requires that C1, without negation, entails C2. (133) a. Sandra Sandra aß ate nicht not einige, some, sondern/*aber but alle all Erdbeeren strawberries. b. Sandra Sandra aß ate nicht not alle, apple, sondern/Xaber but einige some Erbeeren. strawberries. Furthermore, sensitivity to scalar strength across disjunctions is well–described in the literature as Hurford’s constraint, examples are given in example (134 134). It was observed that generally, disjuncts may not entail each other. However, scalar items can be saved by exhaustification of one of the conjuncts. Exhaustification will also be at the root of the lack of scalar implicatures in corrections like (133 133). (134) a. * Sandra is French or Parisian. Sandra is Parisian or French. b. ?? Sandra ate some or all strawberries. Sandra ate all or some strawberries. In the analysis proposed, a process like exhaustification will play a crucial role. This process modulates the interpretation of statements which are entailed by other statements. Consider a sentence like Sandra ate four strawberries. This sentence is entailed by infinitely many alternative statements like Sandra ate five/six/seven . . . strawberries. Consequently, there is an inference made in the contextualized interpretation of this sentence: Sandra ate four and not more than four strawberries. A process like this is also proposed to be instrumental in the interpretation of scalar corrections. These are the three characteristic properties of corrections that need to be accounted for. Firstly, C1 and C2 are contradictory. Secondly, the negated first conjunct is in polarity contrast with a 83 84 3.2. Rooth (1992) essentials discourse statement. Thirdly, corrections with sondern appear to lack scalar implicatures. These cru- cial properties as an interaction of focus semantics and the concrete semantics and presuppositions of correction. In section 3.2 3.2 background on the alternative semantics formalism proposed by Rooth Rooth (1992 1992) is given, and some assumptions about discourse coherence are made explicit. In section 3.3 3.3 and 3.6.4 3.6.4, additional data about the contrastivity requirement and the polarity requirement are given. Finally in 3.5 3.5, a formalization is proposed. For this discussion I will largely set aside the question whether or not corrections are large coordinations or small coordinations, and assign more importance on the relative scope and discourse relations of the foci involved. The reason is that, as was presented in chapter 2 2, there is not a uniform coordination height for all types of corrections. 3.2 Rooth (1992) essentials Before analyzing correction specifically, the system of focus semantics that this analysis will largely be based off, proposed in Rooth Rooth (1992 1992), will be introduced. His analysis is appealing since it proposes a unified, formalized focus operator and it makes explicit assumptions about how this operator interacts with contrast and certain discourse types, which will help account for the observations made in section 3.1 3.1. His account of focus is an alternative semantics account, since it derives the meaning of focus from the interaction of what and how an utterance is phonetically realized with salient contextual alternatives. The three essential components to achieve this are a mechanism to incrementally calculate a possible set of alternatives, a semantic operator which marks the scope of focus, and a pronoun which is anaphoric to propositions which were salient in the preceding discourse, which is required to be in a specific relation to the utterance in question. I will provide more detailed introduction into the basic motivations and mechanisms of these three entities in the following sections 3.2.1 3.2.1 through 3.2.3 3.2.3 3.2.1 Regular semantic values and focus semantic values Rooth Rooth (1992 1992) argues that each syntactic node has two layers of meaning. One type is the semantic value, which is abbreviated by J.K 0 in Rooth Rooth (1992 1992). This value is the regular denotation of any node, with the general semantic composition rules like functional application and predicate abstraction applying to it. The second layer of meaning is what is called the focus semantic value, abbreviated by J.K f . Intuitively, the focus semantic value gives insight about alternatives to the focused expression that 84 Chapter 3. Semantics and pragmatics of sondern 85 could have been uttered. For any given node, it is derived by locating a constituent that bears a focus feature, which is the syntactic correlate for bearing pitch accent, replacing it with a variable of the same semantic type, accessing the relevant domain of quantification, and introducing all members in the domain of quantification in place of the variable. For example, to form the focus semantic value of Sandra F ate an apple, Sandra F is replaced by a variable x e , yielding x e ate an apple. In a situation where D e containsfSandra, Erich, Sigrid, Annag the set of alternatives isfSandra ate an apple, Erich ate an apple, Sigrid ate an apple, Anna ate an appleg. For nodes which do not contain an F–feature, the focus semantic value is singleton set containing the denotation of the F–marked constituent. Focus alternative sets are combined by the mathematical operator followed by the application of the usual rules of composition like functional application or predicate abstraction, as demonstrated in the tree below. (135) 8 > < > : eat a pear eat a peach eat a banana 9 > = > ; a pear F 8 > < > : a pear a peach a banana 9 > = > ; eat n eat o 3.2.2 The operator and the context pronoun One of the major insights in Rooth Rooth (1992 1992) is to provide a unified analytical base for different uses of prosodic accent and their semantic effect. He proposes a single mechanism for association with focus as involved in sentences with only, question–answer coherence, the dependency of scalar implicatures on contextual information, as well as contrast in parallel syntactic structures. This is challenging, since some of these phenomena are due to syntactic context, whereas others are implicit in the discourse context. It is implemented by assuming that in addition to the focus semantic value, each of the conjuncts have context pronouns which are related to other sentences by an anaphoric dependency, and the interpretation of other syntactic nodes within a given sentence can be taken into account when evaluating focus in a comparable way. To make explicit the syntactic location of focus evaluation, but also to mediate between these two restriction on focus alternatives, Rooth Rooth (1992 1992, ex.40) proposes the operator. It is defined with the following two conditions. In Rooth Rooth (1992 1992), G is used as a context pronoun which is in 85 86 3.2. Rooth (1992) essentials an anaphoric relation to previous contexts, whereasg is intended as another syntactic node in the sentence. He assumes that the set case and the individual case occur interchangeably depending on the syntactic context (Rooth Rooth, 1992 1992, 86). This definition is applied to a number of different cases, that is the interpretation of only, the pragmatic restriction of focus interpretation, and the notion of contrasting individuals. Specifically, the former two use the set case, and the latter uses the individual case. (136) Definition as given in Rooth Rooth (1992 1992, 93) (his example 40) a. set case: jG presupposes thatG is a subset of the focus semantic value forj and contains both the ordinary semantic value of j and an element distinct from the ordinary semantic value ofj b. individual case: jg presupposes thatg is an element of the focus semantic value forj distinct from the ordinary semantic value ofj The examples given by Rooth Rooth (1992 1992) are discussed in more detail below. The example given in Rooth Rooth (1992 1992, ex.7) for the former phenomenon is Mary only read F The Recognitions. If one were to take into account only the alternatives provided in the sentence, one might expect the set of alternatives to be the unrestricted set of predicates of the form x–ing The Recognitions, like dusting The Recognitions, printing The Recognitions, i.e. , the focus variable. In order to predict the intended interpretation for this sentence, namely the one where only reading The Recognitions is opposed to understanding The Recognitions, both reading and understanding need to be provided in the context set, which is a set of propositions that is part of the denotation of the pronoun anaphorically related to the previous context. This is ensured by positing a pronoun of sets of alternatives containing these values, the context pronoun C, which is required to be subset of the focus semantic value per the set clause (136a 136a) in Rooth Rooth’s (1992 1992) definition. Rooth Rooth (1992 1992) gives another example for how focus sets allow for pragmatic inferences. His examples are given in (137 137) below. These examples are unrelated to the semantics of only. (137) a. Well, I [passed] F . Suggests I did not ace the exam. b. Well, [I] F passed. Suggests other people did not pass the exam. For these examples, Rooth Rooth (1992 1992) describes how the context pronoun denotes a set that contains members that are ordered by virtue of entailment relations (for example: Steve and I passed entails I passed, and acing an exam entails passing an exam). Since the context pronoun must be a subset 86 Chapter 3. Semantics and pragmatics of sondern 87 of the focus semantic value, the relevant interpretation of focus is ensured. Rooth Rooth (1992 1992) argues that by only committing to a lower item on the scale, like passing the exam, a cooperative hearer will infer that acing the exam is not true. Crucially, this scale has to be negotiated or be evident in the context of an utterance. It is a result of the anaphoric relation with a context set in discourse. Another crucial case is the case of contrast in examples like the hypothesized beginning of a joke. If focus were evaluated at sentence level, in this example, it would remain unclear why both adjectives, that is, American and Canadian, bear a pitch accent. In this example, the individual clause of (136 136) comes into play: to license this type of contrast, focus needs to be evaluated in the subject and object, no higher than the DP level; the ordinary semantic value of an American farmer needs to be a member of the focus semantic set of a Canadian farmer, and vice versa. Rooth Rooth (1992 1992) argues that since Ja Canadian farmerK 0 2Jan American farmerK f , these phrases are interpreted as in contrast, since it fulfills condition (136b 136b). This is the formal definition that Rooth Rooth (1992 1992) provides for contrasting individuals. In the analysis of the even stronger contradiction condition in corrections, this condition is at work, although it does not suffice to formulate the restrictions in correction. Rooth Rooth’s (1992 1992) condition ensures that the conjuncts are construed in parallel in a relevant way, which will be a subpart of the formalism of the contradiction condition that will be proposed as part of the analysis of the discourse conditions of correction. (138) An American F farmer was talking to a Canadian F farmer. . . (Rooth Rooth, 1992 1992, ex.11) 3.2.3 Discussion of discourse coherence in Rooth (1992) Rooth Rooth (1992 1992) only covers a particular case of discourse coherence, the one between question and answer pairs. He argues that the set case of (136 136) is applied in question and answer pairs. The denotation of a question in alternative semantics was argued to be the set of its answer choices (Hamblin Hamblin, 1973 1973). Rooth Rooth (1992 1992) proposes that for a question and a (focused) answer to be cohesive, the denotation of the question, that is, the set of answer choices, must be subset of the focus alternatives in the answer. Schematically, this is exemplified in (139 139) below. In this example, since the the denotation of the question is a subset of set of focus alternatives, the question answer pair is considered cohesive. (139) A. What did Sandra eat? Hamblin denotation: H=fSandra ate an apple, Sandra ate a pear, Sandra ate a fig . . .g B. Sandra ate a fig F . Rooth’s set of focus alternatives: R=fSandra ate an apple, Sandra ate a pear, Sandra ate a fig . . .g C. Since HR, this question–answer pair is discoursively cohesive 87 88 3.2. Rooth (1992) essentials Here, I want show that although the discourse focus condition as formulated in Rooth Rooth (1992 1992) potentially covers question answer coherence quite well, in a narrative discourse additional config- urations must be allowed for. Consider the examples below as an illustration. It will be assumed that focus is interpreted at sentence level here, since there are no other focus interactive operators in these sentences. The shape of the alternatives is given in (140b 140b). Crucially, although this is a simple narrative discourse relation, in no way can the proposition (140a 140a) be construed to be a subset of the focus alternatives of (140b 140b), since all of them have the shape Saskia is eating x. Note that although counterintuitive, it is the preceding phrase in discourse that is required to be a subset of the following phrase, as is well motivated in Rooth Rooth (1992 1992). Using variable names in analogy to the definition in (136 136), this would mean, (140a 140a) is f, which has to be a subset of the focus semantic value of (140b 140b),G. Similarly with example (141 141): assuming the alternatives of this sentence are described as given in the example, the proposition Loitering is prohibited (presumably with the addition of some kind of generic operator) cannot be a subset of the focus semantic value of this sentence. This means that additional discourse requirements and relations have to be formalized to account for cases like (140 140) and (141 141). Other accounts, like the Question Under Discussion account proposed by Roberts Roberts (2012 2012), argue that whenever a focused constituent is uttered, it presupposes an implicit question that has an appropriate form, for example for (140b 140b) this would be What is Saskia eating?, or it might be argued that in sentence (140a 140a), there is an implicit question What is everybody eating?. This account defines a notion of complete answerhood. In the case at hand, since “Saskia” does not make a complete answer, it is interpreted as a partial answer to this question. The take away at this point is that while all in all this analysis need not have a much more complex apparatus than the one proposed in Rooth Rooth (1992 1992) to account for the data, in this particular respect additional formalisms for antecedent–target relations will need to be introduced. (140) a. Saskia, Erik and Arno are at dinner. b. Saskia is eating an apple F . fx2 D e : Saskia is eating xg (141) a. Loitering is prohibited. b. The homeless man was asked to leave F fx2 D he;ti : The homeless man was asked to x.g These examples demonstrate that there are many forms of discourse coherence, and although there are without a doubt many rules that govern them, these rules do not make use of the Roothian focus conditions as defined in (136 136) above. While the Roothian notion of discourse coherence is a start, some further refinement is needed. This does not mean that the Roothian rule is never applicable. Other than in question–answer pairs, there are examples of types of discourses where the ordinary semantic value can be an element in the set of focus alternatives. These examples need to be pairs of sentences which differ minimally, 88 Chapter 3. Semantics and pragmatics of sondern 89 with only the focused constituent changing. They need to be composed from the presupposition of focus, combined with one of the other alternatives. The sentences in the versions in the a examples of (142 142) and (143a 143a) below fulfill the Roothian subset relation, since they share a comparable set of focus alternatives. A further interpretive component of these sentences is that just like corrections, these objects are conceptualized as mutually exclusive. The fact that these sentences are perceived as a contradiction will be discussed in more detail later, for the time being, suffice it to say that in the literature (Fox and Katzir Fox and Katzir, 2011 2011; Chierchia et al. Chierchia et al., 2009 2009) this meaning component is analyzed to be due to an exhaustivity implicature. Due to this implicature, such sentences will typically be understood as mutually exclusive. This is the configuration is also encountered in corrective sentences. (142) a. Erik ate apples. b. Erik ate pears F . (143) a. Erik washed the dishes. b. Erik [ peeled apples ] F . 3.3 Contrast Based on Rooth Rooth (1992 1992), we can determine the location of focus interpretation in correction sentences. As described in section 3.1.1 3.1.1, the constituents pronounced with pitch accents are clearly contrastive in the sense of Rooth Rooth (1992 1992) with respect to each other. In accordance with Rooth Rooth (1992 1992), occurring in comparable syntactic environments is one impor- tant factor in being contrastive. Lack of syntactic parallelism precludes a contrastive interpretation, and furthermore also precludes correction. This can be seen in example (144 144) below. This example has two potential reasons for ungrammaticality. On the one hand, the semantic type in the second focused phrase does not match the type in the first phrase, which would preclude contrast in the sense of Rooth Rooth (1992 1992) from being established, since the semantic value cannot be a subset of the focus semantic value of the respective other conjunct. On the other hand, correction also introduces an understanding that the two conjuncts need to be in a salient relation with each other, which apples and sleep are not. (144) * Sandra didn’t eat APPLES, but SLEEP. As was made explicit in section 3.1.1 3.1.1, there is an additional condition in corrections, which is that the alternative provided in the first conjunct and the alternative in the second conjunct cannot both be true in the communication situation where correction is uttered. The question I want to answer in this section is whether based on this information we can determine where focus must be interpreted to determine this type of contrast. 89 90 3.3. Contrast Specifically, there are two salient locations that I consider. Either this type of contrast is construed at a location outscoping negation, or in a location outscoped by negation. The two possibilities are given in (145 145) below. These trees disregard questions of coordination height and assume proposition level coordination for all sentences. (145) a. Sandra ate an APPLE. C but Sandra didn’t eat a PEAR C b. Sandra ate an APPLE. C but C Sandra ate a PEAR not In option (145a 145a), focus alternatives are formed at a level lower than the coordinator but, since also in examples of contrastive focus in coordinations, like a PRETTY and LARGE apartment, coordinators are nonactive in the formation of alternatives. Granted this assumption, assuming a domain of individuals D e fa pear, an apple, a sandwichg, the set of alternatives for the left conjunct are J:C1K f =fSandra didn’t eat a pear, Sandra didn’t eat an apple, Sandra didn’t eat a sandwichg. For the right conjunct, it is JC2K f =fSandra ate an apple, Sandra ate a pear, Sandra ate a sandwichg. Like this, the phrases cannot be in contrast with each other, since Rooth’s contrast requirement states that the semantic value of at least one of the conjuncts must be an element in the set of alternatives of the respective other conjunct. In the present configuration, this is impossible, since all alternatives in the right conjunct do not contain negation, whereas the ones in the left conjunct do. Clearly the semantic value of the left conjunct must contain negation, and the one of the right conjunct must not, so they cannot be a membership relation with the focus alternative value respective other side. Consider the second candidate, (145b 145b). Here, assuming the same domain of individuals, the alternatives to the left conjunct and the right conjunct are identical, that is JC1K f =JC2K f =fSandra ate an apple, Sandra ate a pear, Sandra ate a sandwichg. As such, each of the semantic values can easily be an element in the respective other focus semantic value. From these consideration I deduce that focus must be evaluated below negation to capture the contrast observation about correction. 90 Chapter 3. Semantics and pragmatics of sondern 91 A further question regarding the locations of these foci is how it is ensured that they be interpreted within the scope of negation, that is, which mechanisms prevent any other configurations. There are two ways this might come about: on the one hand, it is possible that all other scope configurations of focus evaluation simply fail, since they cannot meet the contrastivity requirement. On the other hand, it may be that the lexical entry in each of the conjuncts makes explicit reference to a context set, as is the case in association with focus. In the discussion about the contrastivity requirement in section 3.7 3.7, I present evidence showing that focus in the second conjunct must be evaluated at the highest possible syntactic level in conjunct two, which would remain unexplained if one were to assume that they fail on account of not fulfilling the contrastivity requirement. As such, these considerations constitute an argument in favor of an association with focus analysis of correction. 3.4 Polarity focus in discourse Another crucial characteristic of correction is that is stands in a particular relation to a discourse antecedent, and this proposition is the first conjunct without the contribution of negation. It was shown in section 3.1 3.1 that the way in which correction refers back to this antecedent is rather restricted: the only way in which variation between the antecedent and the first conjunct is permitted is by adding negation to it, no more (not negation and an additional contrast), and no less (negation must be added, and cannot be present in the sentence accidentally). It was also mentioned that the antecedent is typically not expressed phonologically, and may be an entailment of a preceding sentence. Since this characteristic affects the negated conjunct only, I want to consider only the properties of the first conjunct. Due to this, I will analyze a situation where speaker A and speaker B disagree about a proposition p, irrespective of this specific realization of correction. This section will discuss the nature of pitch accent on a negated auxiliary. It will be discussed whether or not this type of focus could possibly be composed compositionally from a rule set proposed in Rooth Rooth (1992 1992). As was mentioned, his framework does not discuss discourse coherences at great length, so it is discussed whether it might be easily extended to the data at hand. It will be argued that a strictly compositional account is not possible, hence, an operator is proposed to introduce the relevant component. Consider the sentences in (146 146), which are meant to be read as two sequential discourse statements expressing disagreement about a proposition. (146) A. Sandra ate an apple. B. Sandra DIDN’T eat an apple. In this situation, the Rooth Rooth’s (1992 1992) set condition in (136 136) applies, as was was motivated in (142 142) above. The most natural prosodic realization of sentence (146B 146B) has a pitch accent on didn’t. 91 92 3.4. Polarity focus in discourse Under this reading, the focus semantic alternatives of sentence (146B 146B) might be expected to be F1=fSandra ate an apple, Sandra didn’t eat an appleg, since this intuitively represents well the type of contradiction found in examples like (146 146), and it fulfills Rooth’s subset condition for discourse coherence. Since negation is a functional component, only observing the locus of the pitch accent is insufficient evidence, since on the one hand, it is unclear what the focus alternatives of negation might be, and also since it seems more plausible that pitch accent is on the auxiliary rather than the negation, since the negation is contracted. However, it may be argued that the negation in (146B 146B) is an instance of broad focus, not requiring a specific prosodic realization. In addition, negation need not be explicit to express disagreement, which the sentences presented in (142 142) demonstrated. A crucial question here, just like with contrast, is whether in discourse (146 146), it is necessary for focus to be evaluated higher or lower in the structure than negation. Initially, both appear to be possible. When the auxiliary with the cliticized negation receives a pitch accent, one might assume the set of alternatives in negation is given by F1 above, in this case the membership condition is fulfilled. At the same time, by virtue of the crucial properties of correction identified above, we must require that other types of contrast, and therefore other types of focus alternatives are also present in correction. Evidence was presented in the previous section showing that a focus evaluation must also be present at a level lower than negation to account for the contrast between the conjuncts. Composing this set of alternatives with negation without any additional assumptions predicts there be focus semantic alternatives like F1’=fSandra didn’t eat an apple, Sandra didn’t eat a pear, Sandra didn’t eat a sandwichg. This focus alternative set does not yield the desired contrast relation, since J146A 146AK 0 62 F1’, that is,fSandra ate an appleg62fSandra didn’t eat an apple, Sandra didn’t eat a pear, Sandra didn’t eat a sandwichg. Another possibility to consider is that maybe the validity of the discourse relation is also checked within the scope of negation, as is the discourse relation for contrast across the conjuncts. Consider a case where an additional pitch accent is placed on apple in the consequent sentence in (146B 146B), so it becomes Sandra DIDN’T eat an APPLE. In such a sentence, the set of alternatives of this sentence might be formed at a level lower than syntactic negation, in accordance with the discussion in the previous section. In this case, the set of alternatives might be F1”=fSandra ate an apple, Sandra ate a pear, Sandra ate a sandwichg, assuming this set, the discourse membership condition is met, and J146A 146AK 0 62 F1”. We would encounter the same configuration of focus values also if the the consequent sentence in (146B 146B) were something different, like in sentence (147B 147B) below, which might well have the same set of contextual alternatives. Just like in the previous sentences, the semantic value of Sandra ate an apple is a member of 92 Chapter 3. Semantics and pragmatics of sondern 93 the set of alternatives of the second sentence if we are continuing that the discourse coherence requirement is evaluated lower than negation. However, this is contrary to fact, since when realized with a pitch accent on the object, this discourse is infelicitous as a correction. (147) # A. Sandra ate an apple. B. Sandra didn’t eat a sandwich F . This argumentation shows that there appear to be conflicting requirements on corrective sen- tences. For contrast between the conjuncts, it must be assumed that there is an evaluation of focus lower than negation. For polarity contrast, which is contrast across discourse, assuming no additional mechanisms or operators, a focus evaluation higher than negation is required, but at the same time it is necessary that negation not interact with the realization of narrow focus which is due to contrast between the conjuncts. In the following sections I will show that even if additional assumptions are made about this type of focus interpretation, it runs into severe problems, which provides evidence in favor of a focus sensitive operator rather than a strict consequence of the interpretation of focus. 3.5 Interpretive components formalized Resulting from the previous discussion, we can see that we need to account for three crucial interpretive components, namely the discourse level contrastivity, the contrastivity across conjuncts, and the fact that corrections with sondern appear to lack scalar implicatures. I propose to account for contrastivity across conjuncts by assuming on the one hand that the conjuncts must be contrastive in the sense of Rooth Rooth (1992 1992), and also that the first conjunct necessarily involves pragmatic exhaustification. Since exhaustification is assumed to be able to occur freely at different levels of embedding in the a sentence, and thus might apply higher than and lower than negation, it can produce felicitous readings with both pragmatically stronger as well as pragmatically weaker second conjuncts. Contrastivity across conjuncts is proposed to be ensured as a presupposition in the second conjunct. The second conjunct has an implicit reference to the exhaustive statement in the first conjunct. This is made sure by sondern being focus sensitive, and imposing a presupposition about the second conjunct. I will first discuss how the exhaustivity requirement in the first conjunct arises in subsection 3.5.1 3.5.1, and then I will discuss how the contrast requirement arises in the second conjunct based on the interpretation of the first conjunct in 3.7 3.7. 93 94 3.5. Interpretive components formalized 3.5.1 Exhaustivity in the scalar interpretation of corrections In this subsection, I propose that pragmatic exhaustification is a relevant process to consider in the interpretation of corrections. 2 2 Based on scalar implicatures, I will argue that this is a better representation of the data than assuming that multiple evaluations of focus. Pragmatic exhaustification has been argued to be a grammaticalized process, which applies at a particular syntactic level, and not necessarily only to utterances. (Chierchia et al. Chierchia et al., 2009 2009) When there are no scalar implicatures involved, this process is not distinguishable from focus. The process of exhaustification is crucial to deriving the readings that we see speakers assuming for many quantifiers in discourse, which are often not fully predicted by their lexical entry, but rather occur as an interaction from a sentence with its pragmatic alternatives. As we saw in the discussion of Rooth Rooth (1992 1992) above, also focus alternatives can be conceptualized to have pragmatic ordering, so the discussion about exhaustification equally applies to the interpretation of focus. Consider the following sentence to illustrate the concept of exhaustification. (148) Sandra ate three pears. Pragmatic alternatives:fSandra ate one pear, Sandra ate two pears, Sandra ate three pears, Sandra ate four pears. . . .g Interpreting this sentence without taking into account its alternatives in discourse, and considering purely its semantics, there is an infinity of situations in which this sentence is true. The reason is that this sentence is true in situations where Sandra ate any natural number of pears greater or equal to three. Since each of these numbers entails three, without making any further assumptions about the appropriateness of different true sentences, a speaker might be expected to truthfully utter the sentence above, although the situation might be more informatively described as Sandra ate x pears, where x represents the exact number of pears. This problem has of course long been addressed in linguistic pragmatics. Ever since Grice Grice (1975 1975), it is taken for granted that cooperative participants in a conversation not only generally utter true statements, but among the statements that do not contradict their beliefs, they choose the most informative statement that they could utter. Being informative here is defined as being entailed by as few other alternatives as possible. Expecting this behavior, a hearer will typically not interpret the sentence above according to its strictly semantic reading, and think it is conceivable that Sandra ate pears in a quantity that is larger or equal to three upon hearing (148 148), but they will assume that the speaker uttered the most informative sentence that they could, that means, the sentence that entails the most other sentences. 2 The term pragmatic as used in this discussion is merely intended to mean not necessarily affecting truth conditions. In the present work, it is assumed that one and the same syntactic operator may have syntactic and pragmatic effects, since it can affect truth conditions as well as the pragmatics of a sentence. 94 Chapter 3. Semantics and pragmatics of sondern 95 That is, a hearer will typically interpret the sentence as Sandra ate three pears and the speaker does not believe Sandra ate more than three pears. The second conjunct of this sentence, and the speaker does not believe she ate more than three pears is called exhaustivity implicature. 3.5.2 Hurford’s constraint and exhaustivity in coordinations Exhaustivity has been a crucial topic in pragmatics. An important area of discussion has been how the process of introducing exhaustivity is best characterized. Traditionally, exhaustivity has been thought of as a consequence of the Grice Grice’s (1975 1975) maxim of quantity, which states that a cooperative speaker has to be as informative as they can. This type of maxim is thought to apply to speech acts. However, Chierchia et al. Chierchia et al. (2009 2009) make a compelling argument that the introduction of exhaus- tivity must be a grammaticalized process, which can have distinct scopal positions. They propose that exhaustivity does not apply at the level of speech acts, but that it can occur in distinct syntactic positions which might be outscoped by a quantifier. As a consequence, they propose an operator which can occur embedded under the scope of certain operators. The lexical entry given for this operator in Chierchia et al. Chierchia et al. (2009 2009, 51) is presented in (149 149). (149) Jexh ALT (S)K W = 1 iff JSK W =1 and8f2 ALT(f(w)=1!(JSKf)) It maintains that S, the argument of the exhaustivity operator, is true, and that all other alternatives which are true are those entailed by S. They make this claim based on arguments involving Hurford’s constraint, which, independent of exhaustivity, is also relevant for the current discussion, since it helps us formulate predictions about inferences in coordinations. Hurford’s constraint is relevant to the present discussion in two ways: on the one hand, embedded exhaustification will be an important tool in deriving corrections, with Hurford’s constraint being an important argument in favor of such an operator. On the other hand, it is important to recognize that there are independent expectations about the strength of the conjuncts in coordination, which will be a base line for corrections. Therefore, the argumentation in Chierchia et al. Chierchia et al. (2009 2009) using Hurford’s constraint is briefly introduced here. Hurford’s constraint states that in disjunctions, neither disjunct is allowed to entail the respective other, which is the reason for the ungrammaticality of (150a 150a). Exceptions to Hurford’s constraint are found in sentences with scalar implicatures, where the scalar implicature of the first conjunct contradicts the second conjunct. Consider the example from Chierchia et al. Chierchia et al. (2009 2009, 5) in (150b 150b). This example is expected to violate Hurford’s constraint, but Chierchia et al. Chierchia et al. (2009 2009) make the argument that within the first conjunct, exhaustivity has to be introduced. Based on this sentence and sentences with embedding, they argue that exhaustification must be due to a localized grammatical operators, and not merely a process applying to speech acts. (150) a. * Peter is either French or Parisian. 95 96 3.5. Interpretive components formalized b. Peter either solved both the first and the second problem or all of the problems. Chierchia et al. Chierchia et al. (2009 2009, 5) c. Peter either solved both the first and the second problem and no other problems or all of the problems. own annotation On the one hand the discussion about Hurford’s constraint is important since it sheds light on the nature and locality restrictions of the exhaustivity implicature. On the other hand, it is also important to create predictions about how we expect implicatures and mutually entailing constituents to behave in coordinations. This is why some consideration to how entailment in coordination behaves as a base line is needed. Consider first these cases of coordination. I am considering two cases of entailment across conjuncts which cause Hurford’s constraint, one where the entailment is due to a superset – subset relation, and another case of scalar implicatures. Also, I am establishing a base line about whether or not entailments in coordinations are possible. This case is important for the discussion of correction, since correction is a coordination, and it needs to be established whether or not there could be any general prohibitions against entailments across the conjuncts of coordination. These examples are crucially different since while the latter forms a scale which often yields an exhaustivity implicature (uttering two implies not three), the former does not trigger such an implicature, since uttering in Paris does not imply not in France. In coordination, all these sentences are acceptable to varying degrees. In case of the examples (151a 151a), the first conjunct entails the second conjunct, whereas the inverse is true for (151b 151b). The sentences in (151a 151a) are without a doubt a bit unnatural, and there is a strong sense that the second conjunct contributes redundant information, there are contexts which might license these usages. For example (151a-i 151a-i) might be licensed in a disagreement about whether or not Paris is a part of France. (151a-ii 151a-ii) might be uttered in a situation where it is maintained that everybody must eat at least a single apple. It is vastly improved by adding a causal adverb I ate two and hence one apple. In the scalar example (151a-ii 151a-ii) it appears there is exhaustification in the second conjunct, contradicting the first, and this contradiction needs to be justified. (151) a. i. ? Ich I fahre drive durch through Paris Paris und and durch through Frankreich. France. ii. ?? Ich I aß ate zwei two und and einen one Apfel. apple. The reading of interest is the reading where the speaker ate two apples total. b. i. Ich I fahre drive durch through Frankreich France und and durch through Paris. Paris. ii. Ich I aß ate einen one und and zwei two ¨ Apfel. apples. 96 Chapter 3. Semantics and pragmatics of sondern 97 The reading of interest is the reading where the speaker ate two apples total. In these examples, we see that in coordinations, cases where the second conjunct entails the first are more readily acceptable, but also cases where the first conjunct entails the second can be licensed in specific pragmatic contexts. There is an overall expectation that the second conjunct will be the stronger on an informational level. Consider the analogous examples with disjunction. These are examples all predicted to be potentially unacceptable due to Hurford’s constraint. As was shown, Hurford’s constraint can be circumvened by use of an exhaustivity operator. Generally, Hurford’s constraint is expected to be independent of linear order in the sense that whether or not the first or second disjunct entails the respective other one is expected to be insignificant. In the actual sentences however, there is a significant difference found between the two realizations. Examples where the first conjunct entails the second, namely (152a 152a), are much less acceptable. We can observe that exhaustification is much more readily available in the first conjunct, and like above there is an expectation that the stronger conjunct is realized in the second conjunct. (152) a. i. ?? Ich I fahre drive durch through Paris Paris oder or durch durch Frankreich. France. ii. ??? Ich I aß ate zwei two oder or einen one Apfel. apple. b. i. ? Ich I fahre drive durch through Frankreich France oder or durch through Paris. Paris. ii. Ich I aß ate einen one oder or zwei two ¨ Apfel. apples. A constant factor across coordination and disjunction is that the stronger utterance is realized in the second conjunct. This is a generalization about how scalar material is structured in coordinated utterances. It will be important to keep in mind that generally sentences with stronger statements in the first conjunct are less acceptable. The key point about exhaustification is that there is good evidence that it may apply in an embedded position, that is, lower than a logical operator. Another important generalization from the argumentation is that implicatures generated from one conjunct can influence the interpretation and acceptability of the second conjunct. In the following section, I will discuss in detail the influence of exhaustification onto correction. It is also important to note that without additional assumptions, there are no entailment relationships between the conjuncts in correction. This is of course not to say that the propositions are unrelated, but if the conjuncts influence each other’s likelihood, it must be modeled with a more complex semantics or pragmatics. 97 98 3.5. Interpretive components formalized 3.5.3 Conceivable Hypotheses about Exhaustivity Since the exhaustivity operator is a silent operator which can occur in an embedded position, there are a number of positions where it could potentially occur. Consider the following tree structuring the hypotheses. In the following discussion, a number of hypothesis are considered: on the first, each of the conjuncts may be interpreted exhaustively or without exhaustivity (that means there are four combinations, since in principle the conjuncts could be interpreted independently). Furthermore, in the first conjunct exhaustification might apply higher than or lower than negation. Figure 3.1: Hypothesis Space First Conjunct Exhaustivity outscoped by Negation outscoping Negation No Exhaustivity Second Conjunct Exhaustivity No Exhaustivity It also needs to be considered that the second conjunct may represent an item higher on the scale or lower on the scale than the non-negated first conjunct. In the case of sondern, this does not result in an acceptability difference, but it will be an important point in the discussion of aber. Consider first the interpretive effects in the interpretation of a scalar item. In the following discussion, I will use a type of enriched truth table for clarity of exposition. Note that these truth tables assume cross-categorial entailment. Consider the case of exhaustification on a scale presented in table 3.1 3.1 on page 99 99. A set of alternatives is given by the leftmost column, so in the case at hand it isf1, 2, 3, 4, 5g. Each item in the set of alternatives is annotated with a truth value or implicature value, which is given in the second column for “3”, and in the third column for “exh(3)”. The antecedent is marked with gray background, whereas the the entailments of the antecedent are marked with pink background. Implicatures are marked with lower case letters. As can be seen, while the sentence non–exhaustively does not have any consequences for the interpretations of the alternatives 3 and 4, the same is not true in the exhaustive case. Since exhaustive interpretation implies that all alternative propositions entailing the prejacent are not true, the alternatives 3 and 4 are implied to be false in the exhaustive case. (153) Sandra ate three apples. There are two major cases to be considered. There is a case where C1, the first conjunct without negation, is entailed by the second conjunct, and another case where C1 entails the second conjunct. 98 Chapter 3. Semantics and pragmatics of sondern 99 Table 3.1: Exhaustification on a scale ”three” 3 EXH(3) 1 T T 2 T T 3 T T 4 f 5 f It is proposed that these different continuations correspond to different scope configurations of the negation with respect to Exhaustivity. These different readings do not occur right away from the assumptions made so far about the truth conditions and implicatures of the sentence. This is demonstrated using the enriched truth tables in Table 3.2 3.2. Table 3.2 3.2 presents this table for a negated scalar sentence like (154 154). (154) Sandra ate not three apples. As a side remark it should be added there is reason to believe that the interpretation of negation is dependent on its realization as sentence or constituent negation. I will set these differences aside for the moment, since I am only considering the logical possibilities here. Consider the table in 3.2 3.2 for this discussion. A clarification on the table: the T/F designations are given based on the non–negated alternatives. So, in this table, if NEG(3) is T, 3 is listed as F. Right away, we can see that of course monotonicity is reversed. The column NEG(3) is straight forward, since NEG(3) entails NEG(4) and NEG(5). The set of alternatives considered for this case isfNEG(1), . . . , NEG(5), . . .g There are no further implicatures about the cases that are lower on the scale. Note that in this interpretation, uttering a negation does not affect the likelihood of non–negated alternatives on the same scale. Said differently, negation without exhaustification does not construe any expectations based on the entailments of negation which affect the non–negated items on the scale, NEG(1) and NEG(2) in the given example. Now I want to consider the two cases NEG(EXH(3)) and EXH(NEG(3)). Of these cases, the straightforward case is EXH(NEG(3)). Since the definition of exhaustification indicates that all alternatives which entail the prejacent (not 3 in the example) are implied false, in the case at hand, the two propositions which entail the prejacent, namely not 1 and not 2, are implied to be false. As a consequence of exhaustification, these two alternatives are implied to be false. Without any more assumptions about negation, if the alternatives not 1 and not 2 are implied to be false, consequently 1 and 2 are implied to be true. It is noteworthy that by virtue of exhaustification, implicatures are generated about alternatives which are not negated. A case where there is no evident solution along the lines sketched so far is the middle column, NEG(EXH(3)). Here, it is hypothesized that negation applies to a set of alternatives like the column 99 100 3.5. Interpretive components formalized Table 3.2: Exhaustification on a scale with negation “not three” NEG(3) NEG(EXH(3)) EXH(NEG(3)) 1 t 2 t 3 F F F 4 F F F 5 F F F Table 3.3: Not only three not only three 1 T 2 T 3 T 4 t 5 t EXH(3) in table 3.1 3.1. However, the effect of implicature is necessarily overridden by the semantics of negation and the reversal of monotonicity. As a consequence, without any other assumptions, we expect the semantics to be not different form negation without exhaustification, as in the NEG(3) column of table 3.2 3.2. A different proposal is made here. It is proposed that negation associates with the interpretation of the exhaustivity operator itself. The exhaustivity operator has similar semantics to scalar only. An example of a sentence with scalar only is given in (155 155) below. (155) Sandra only ate three pears. Both operators quantify over a set of alternatives, and imply in the case of the exhaustivity operator, or assert in the case of only that no alternative which entails the prejacent is true. In the case in (155 155), this means the alternatives four, five . . . are presupposed wrong. The current proposal is that sentences with a relative operator scope of NEG(EXH(3)) are interpreted like the negation of sentence (155 155), which is given in (156 156). This sentence implies that Sandra ate more than three pears. (156) Sandra didn’t only ate three pears. Given in table form, according to intuition the resulting state after uttering (156 156) is given in table 3.3 3.3. It is proposed that this is a reading that the first conjunct can have in cases of correction. I will not provide a compositional way in which this reading was derived. I believe in cases like (156 156) what is likely happening is negation associating with only, with the alternatives for only being other quantificational adverbs, presumably also and others. In the case of (156 156), this generates a 100 Chapter 3. Semantics and pragmatics of sondern 101 Table 3.4: “Not three but two” 1 t 2 t 3 F 4 F 5 F 1 T 2 T 3 f 4 f 5 f presupposition that one of the alternatives is true. Since this means an alternative sentence with also can be true, it is presupposed that the alternatives with 3 and 4 can be true. This means that there is focus on both the scalar element three in example (156 156), as well as the quantificational adverb only. Without committing to the proposal that the derivation of this reading with the exhaustivity operator works in a similar fashion, I want to propose the exhaustivity operator has an alternative like also. When negation is combined with a set of alternatives which has both an exhaustivity operator and no exhaustivity operator (or another operator), an implicature is generated which indicates that some cases that the prejacent entails can be true. To sum up, it is proposed that the exhaustivity operator may take scope lower than or higher than negation. When it takes scope higher than negation, it generates the implicature that alternatives entailing the prejacent may be true. When it takes scope lower than negation, it generates the implicature that alternatives entailed by the prejacent are true. This allows us to refine our view on the function of sondern. This work proposes that the function of sondern is to deny implicatures. Also, it licenses the exhaustivity operator to take scope lower than negation, a scopal configuration that is not univerally available, and crucially unavailable in minimal pairs of the left conjucts with aber. In the following paragraphs, I will introduce this proposal in more detail. As was mentioned above, the interpretation of the left conjunct may be represented with two truth tables, that is, one where negation scoped over a scalar set of alternatives, and another one where it scoped only over the exhaustivity operator. The two candidates available for the second conjunct are ones where the second conjunct entails C1, the first conjunct without negation, or ones where it does not. Note that exhaustification in the way proposed saves us from putting things in this roundabout manner, since now inferences are made from negated to nonnegated alternatives. It is proposed that the reading (EXH(NEG(3))) combines with the continuation where the continuation is entailed by C1, and (NEG(EXH(3))) combines with one where C1 entails C2. Looking at the tables, this means that in both readings, the contradiction is only at the level of implicatures, and there is no truth–conditional correction. The generalization that sondern is a coordination in contradictions at the level of implicatures 101 102 3.5. Interpretive components formalized Table 3.5: “Not three but four” 1 T 2 T 3 T 4 t 5 t 1 T 2 T 3 T 4 T 5 f only fits in well with many of the nonscalar cases of corrections with this coordinator. A typical example of a correction with sondern may be the following example (157 157). In this sentence, the contradiction arises from an exhaustive implicature: C1 (the left conjunct without negation) does not entail that she did not eat anything else, and neither does C2. However, the use of focus creates such an implicature, the very implicature that is being denied by virtue of using sondern. (157) Sandra Sandra aß ate nicht not einen an Apfel, apple, sondern but eine a Birne. pear. At this point I am not fully committed to claiming that this is a general restriction for sondern, since I cannot provide sufficient argumentation on the basis of a wide data study here. However, it may be an explanation for some of the restrictions that was shown for sentences with sondern, namely that it was not allowed in direct contradictions. Be reminded that sentences like the following sentence were not acceptable. (158) * Sandra Sandra aß ate keine no ¨ Apfel, apples, sondern but alle. all. The generalization that sondern is a contradiction only at the level of implicatures may give an explanation for these data: since here, the second conjunct contradicts the truth values of the first conjunct, it is not felicitous to use only in such a correction. 3.5.4 Interim Conclusion In this section, it was argued that it is a principle function of sondern to introduce a correction at the level or implicatures. Since implicatures are introduced with the covert exhaustivity operator (EXH), this operator taking different scope resulted in different types of corrections. It was proposed that the exhaustivity operator could take scope higher than and lower than negation, with different expected continuations. 102 Chapter 3. Semantics and pragmatics of sondern 103 3.6 Discourse relation In this section, an analysis of the observation that the first conjunct necessarily occurs in a specific discourse relation with a discourse proposition is offered. It was observed that the first conjunct needs to be the addition of a downward entailing operator to this discourse proposition. This discourse proposition is accepted as being part of the common ground. I will relate this discussion to the literature on verum focus, and the semantics of questions with polarity bias. I will discuss whether or not these factors could arise from a regular interpretation of focus with minimal additional assumptions. The discussion will show that a more appropriate approach considers polarity contrast as found in corrections the result of an operator which introduces a discourse presupposition. 3.6.1 Background: verum focus This analysis will account for the discourse requirements by proposing that correction with sondern necessarily relies on a type of discourse relation comparable to the one in verum focus. At the same time, it will agree with some of the relevant literature in that verum focus must be defined as an operator and does not come for free in focus semantics. The concept of verum focus is first introduced in H¨ ohle H¨ ohle (1992 1992). It describes that stress on certain verbal positions and also complementizers is interpreted as stressing that a proposition is true. Consider the example in (159 159) below. In this example, focus is placed on the main verb, however, this does not mean that this verb is contrasted against other verbs like reads, buys etc., which would be expected in a regular focus interpretation. Rather, the object of contrast is whether or not the proposition is indeed true. There is variability as regards to where this type of focus is phonologically realized, it can be realized on main verbs, auxiliary verbs, and complementizers. Simplifying slightly, H¨ ohle H¨ ohle (1992 1992) analyzes this phenomenon by defining a proposition level operator VERUM which takes a proposition p as its argument and asserts it is true that p. I will use this simplified version as a descriptive approximation for now. (159) A: Erich Erich bezweifelt, claims dass that Sigrid Sigrid ein a Buch book schreibt. is-writing. B: Sigrid Sigrid SCHREIBT is-writing ein a Buch. book. It is true that Sigrid is writing a book. There are essential similarities between examples of verum focus above and corrections. Nega- tions in corrections have a very similar interpretive effect: whereas in example (159 159), verum focus follows a downward entailing statement embedded in the verb bezweifelt, in the example (160 160), a focused negation asserts the falsity of a previously asserted statement. In both of these cases, a 103 104 3.6. Discourse relation discourse statement is contrasted against a following statement with inverse polarity. A focused negated statement like Sigrid writes NO book is related to a discursive statement without negation, the embedded sentence of Erich claims Sigrid is writing a book. Due to these similar restrictions and processes, it is desirable to deduce an analysis for corrections from the existing analyses of verum focus, since they share crucial properties, and also interact with polarity and discourse in a similar way. (160) a. Erich Erich behauptet, claims Sigrid Sigrid schreibt is-writing ein a Buch. book. b. Sigrid Sigrid schreibt writes KEIN no Buch. book. It is not true that Sigrid writes a book. The expression of verum focus and its interactions with polarity has been discussed in greater depth with respect to questions. Han and Romero Han and Romero (2001 2001) analyze the interpretive variations in questions with different types of negation, originally pointed out by Ladd Ladd (1980 1980). For the examples in (161 161) below, it is found that in English and many other languages, there is a relation between expected answer and the place and type of negation that is realized in a question. For English, they remark that negation realized in C and cliticized to an auxiliary corresponds with the speaker having the previous belief that the proposition about which the question is asked is true. In the example (161a 161a) below, this corresponds to the speaker having a belief that John drinks. However, when negation is realized in a more embedded position lower than the subject like in (161b 161b), there are no such epistemic restrictions regarding the speaker. (161) a. Doesn’t John drink? speaker believes John drinks b. Does John not drink? speaker has a neutral expectation Han and Romero Han and Romero (2001 2001) propose that this should be modeled as the relation between negation scope and scope of a verum focus operator. This operator is roughly synonymous to the adverb really, and defined in (162 162) below. Applied to a proposition p, this operator asserts that the speaker is certain p should be part of the common ground, which they call a conversational epistemic operator. (162) lp hs;ti lw:8w’2Epi x (w)[8w”2 Conv x (w’)[p2 CG w” ]]= FOR-SURE-CG x Presence or absence of negative bias in questions in their system depends on whether negation outscopes this operator. If so, a negative question bias is encountered, otherwise, the question is interpreted neutrally. Gutzmann and Mir´ o Gutzmann and Mir´ o (2011 2011) argue that a major issue with Han and Romero Han and Romero’s (2001 2001) analysis of verum focus concerns the semantics that they give to verum focus, since Han and Romero Han and Romero (2001 2001) assume the conversational epistemic conclusions are part of the assertion. 104 Chapter 3. Semantics and pragmatics of sondern 105 These data are relevant to correction since on the one hand, the discourse relation is closely related. In a discourse characteristic to polar questions of the kind Han and Romero Han and Romero (2001 2001) describe, these questions follow a discourse state where a nonnegated assertion is part of the common ground of conversation. This set of alternatives is followed by a polar question, in alternative semantics this is a set of alternatives containing a negated proposition and a non–negated proposition as a popular view in alternative semantics (Hamblin Hamblin, 1973 1973) views the denotation of a question to be the set of possible answers. Correction is very similar. In a correction, a discourse state which is necessarily a nonnegated proposition (cf. observation 2 in the introduction of this chapter) is followed by a proposition which on the surface contains a negated proposition. This requirement must somehow be encoded in the semantics of this sentence. In this way, a correction is reminiscent of a polar question. There is evidence that although on the surface the first conjunct of a correction looks like a negated proposition, also a nonnegated proposition is encoded. Furthermore, we have also seen extensive evidence that there are positional restrictions for negation. For example, negation cannot outscope epistemic adverbials in corrections. The relevant examples are in chapter 2 2 on page 62 62 3.6.2 Possible analyses for corrections The other interpretive processes described in the previous subsections, namely the apparent lack of scalarity effects, and the contrast between the two conjuncts were argued to be an interaction between where (local) exhaustivity applies, and a presupposition of the second conjunct stating that its focus values must be a subset of those of the first conjunct. In order to account for the observed discourse requirements, there are some things that need to be maintained. As was described above, in the simplest case, correction refers back to a discourse proposition, and can be characterized by this form, repeated from (127 127). (163) A. P B. (: P), but Q This clearly not be explicitly the case in all corrections. In the example from the parliamentary data shown in (131 131), it was made clear that an antecedent sentence need not be explicitly uttered, but it can apparently be accommodated in a corrective sentence. Similar to the verum focus cases, the interpretive effect is that the speaker utters their wish to add this proposition, P in the schematic above, to the common ground, but while the speaker actually believes:P. I want to argue that this is a reason to treat this interpretive component as a presupposition, since accommodation is a typical feature of presupposition. When presuppositions are not explicitly satisfied in context, there are interpretive mechanisms available which ensure that the presupposition will still be interpreted. I will argue that the contextually accommodated proposition is presupposed. 105 106 3.6. Discourse relation One of the tests for whether or not an interpretive component is a presupposition is whether it can project past negation. Consider the example in (164 164). In this sentence, a correction is placed under a negated embedding verb. We can see that despite the negation, the sentence continues to convey that somebody claimed that Sandra ate apples, which is indicative of this component being a presupposition. (164) Erich Erich glaubt believes nicht, not dass that Sandra Sandra nicht not ¨ Apfel, apples sondern but Zwetschgen prunes isst. eats. Erich doesn’t believe that Sandra doesn’t eat apples but prunes. ! Somebody claimed Sandra ate apples. ! In the current discourse, it is believed that Sandra ate apples. Another requirement for the proposal of the discourse component here is that it should be able to interact sufficiently well with the proposed semantics so far. There are a number of proposals for the interpretation of verum focus already. Many ac- counts, including H¨ ohle H¨ ohle (1992 1992), Han and Romero Han and Romero (2001 2001) and Gutzmann and Mir´ o Gutzmann and Mir´ o (2011 2011), opt for an operator–based solution for verum focus. In addition, Gutzmann et al. Gutzmann et al. (2017 2017) present arguments against treating verum focus as a focus phenomenon. They argue that while focus has a relatively uniform appearance across languages, in the case of verum focus, the close relation between PF and LF/pragmatics is not uniformely given. While it is true that English and German are relatively uniform in marking verum focus with a pitch accent on the auxiliary, there exist a variety of other expressions of verum focus across languages. For example, Italian uses a stressed adverb bien, Spanish uses the polarity particle s´ ı. Due to these and more language comparisons, Gutzmann et al. Gutzmann et al. (2017 2017) argue that the interpretation of verum focus is not due to a specific case of focus, but rather must be a lexical operator. This is the view that will be adopted for the semantics of correction. Also in the corrective sentences that are the subject of this discussion, there is no uniform location of focus. Typically, there are pitch accents realized on the items that are in contrast. Also, there is no obligatory prosodic focus on negation. However, the contextual presupposition stays constant. This analysis will argue that there is verum focus, but that verum focus is due to a lexical operator with certain presuppositions. One of the goals for this analysis is to be explicit about how the discourse dependency effect arises from the given focus alternatives, so even if there is no direct correspondence between the phonological form and discourse interpretation, the effect and interaction of a verum focus–type operator with the focus structure of the sentence will be made explicit. In doing so, I will to evaluate two proposals: on the one hand, I will evaluate whether or not it might be the case that simply negation is focused in subsection 3.6.3 3.6.3, where it will be concluded 106 Chapter 3. Semantics and pragmatics of sondern 107 that this analysis does not cover the interpretive range of the construction. I will argue in favor of an operator–based solution for verum focus in subsection 3.6.4 3.6.4. 3.6.3 Focused negation In alternative semantics theories of focus, the placement of a syntactic feature on a certain phrase covaries with prosodic accent placements and alternative formation, which based on the semantic type of this phrase, forms a set replacing the focus marked constituent with all items of this type, which are later contextually restricted. In this subsection, it is considered whether or not the discourse effects on correction are the direct application of this view of focus semantics, with negation being focus marked. It is concluded that a number of additional assumptions would have to be made in order to justify this assumption, and that it is more desirable to formulate an operator. An alternative set formed may be formed with other functions of the type of negation, the alternative set may then be further contextually restricted. An open question in this context is what the semantic type of negation is, and whether there are different types of negation, or only a single kind. In a more logics–oriented semantics, often it is assumed that there is only one negation operator, which is a truth–value operator, with all other expressions of negation being derived from it, or their lexical entry making reference to the basic negation operator. In these semantics, negation is considered to be of typeht;ti, orhst;sti in intensional semantics. Its alternatives might thus be sentence–level operators, which might be adverbs like maybe, and there is one adverbial that has been considered quintessential to verum focus, namely something like it is true that. The specific nature of the quality of the modal operator is not discussed in the current work, it is simply maintained that whichever modal predicate is selected anchors the proposition in question in the common ground. If the negation in question is accepted as being a quantifier, one may consider the set of alternatives being other quantifiers. For the current discussion, the alternative of negation is considered to be the identity func- tion, l p hs;ti :p. In this exposition, I will assume that with focused sentential negation, the set of alternatives is minimallyf:;l p hs;ti :pg. An immediate problem for this account is that sentential negation is by far not the only expression of negation which can occur in correction, many other downward entailing expressions can also occur in the first conjunct of correction. They may be quantifiers like kein*+DP ‘not-a DP’, as well as adverbials like kaum ‘barely’. Considering these negations, without any modifications of the way alternatives are created, it is expected that polarity focus might be interpreted lower than these DP level negations, which is contrary to fact. Continuing to give this idea the benefit of the doubt, one might consider (and in fact there is 107 108 3.6. Discourse relation good evidence for this independently) that negation, especially with negative indefinites like kein* ‘not a’ is not interpreted locally, but rather is a case of negative concord. This was argued for kein* in Penka and Zeijlstra Penka and Zeijlstra (2005 2005); Zeijlstra Zeijlstra (2004 2004), and we might extend this assumption to other types of downward entailing expressions that are not realized in a typical sentential negation position. Negative concord is how the negation system in languages like French and Spanish are described (cf. example 165 165): in these languages, sentential negation is expressed in two places, but is not actually interpreted as multiple negation (::p), rather, negation is only interpreted once in a single location. This is very apparent with negative quantifiers in these languages, where negation is interpreted in a different location than the quantificational part. (165) a. Le The filme film ne not me to-me plaˆ ıt appeal pas. not. I don’t like the film. b. Je I ne not regrette regret rien. nothing. I don’t regret anything. In trying to apply this type of insight to correction, it needs to be assumed that for all quantifiers and other downward entailing expressions except the negative adverbial nicht ‘not’ in its position following the subject, there is an underlying Polarity Phrase which is the largest phrase in the left conjunct. This phrase is interpreted as a sentence level negation, with the other lexical contributions interpreted in situ. I will focus on sentences where the negation is nicht. How does this proposal account for a non–scalar case like Sandra didn’t eat a pear, but an apple, or its translation in German? The assumed LF of this sentence with sentence level negation is given in (166 166). (166) (not F (Sandra ate a pear F )) but Sandra ate an apple F . In accordance with previous proposals, we know that the contrastive component is contributed by a combination of localized exhaustification and presuppositions, and for the present purpose I will assume that all these processes are already complete. The focus structure of the left conjunct is then assumed as in Figure 3.2 3.2 on page 109 109. This is not an appropriate representation of the intuitions: we want to represent that Sandra didn’t eat a pear, but an apple discoursively follows a previous statement Sandra ate a pear. If this is meant to be a result of the focus on sentential negation, then we would like:(Sandra ate a pear) to be only contrasted with Sandra ate a pear, but in this type of semantics, it is contrasted with this and all types of other statements. This means that we conclude, backing up Gutzmann et al. Gutzmann et al. (2017 2017), that also in correction, the suspected polarity focus should not be taken literally. It is desired that the composition of polarity 108 Chapter 3. Semantics and pragmatics of sondern 109 Figure 3.2: Illustration of the focus composition 8 > > > > > > < > > > > > > : :Sandra ate a pear :Sandra ate a peach :Sandra ate a banana Sandra ate a pear Sandra ate a peach Sandra ate a banana 9 > > > > > > = > > > > > > ; Sandra ate a pear F 8 < : Sandra ate a pear Sandra ate a peach Sandra ate a banana 9 = ; not F : l p hs;ti :p focus does not take alternatives into account, and is therefore not a focus sensitive operator in the general sense, that is, an operator the semantics of which necessarily relies on the focus alternatives of its sister. (Suspected) polarity focus in correction does not fall into this category. This speaks in favor of giving the discourse reference in correction an analysis as a semantic operator with presuppositional content. 3.6.4 Verum Focus operator In accordance with many previous treatments of verum focus, the more promising way of accounting for the discourse referential component of correction is analyzing it as a lexicalized presupposition. As we saw, similar types of phenomena also occur in other sentence types, specifically, in yes/no questions of the type Doesn’t John drink? It is an important insight Guerzoni Guerzoni (2004 2004); Han and Romero Han and Romero (2004 2004) that there is certain vari- ability in the realization of negation. When negation is phonetically realized in a syntactically high position, it is interpreted as a discourse active component. In the current work, it is proposed that the negation in correction is this type of negation. At the same time, it was shown in the previous section 3.5 3.5 that negation also interferes with functions lower in the first conjunct. On the one hand, we saw that negation has scopal amboguity with the exhaustivity operator, and on the other hand, we saw that there was not only one realization of negation which could occur in correction, but all downward entailing expressions, including for example the adverb less. To account for these conflicting requirements, it is proposed that a type of negative concord is at play in the first conjunct of correction. In this type of negative concord, different projections contribute different facets of meaning. As such, the negation as realized in the first conjunct 109 110 3.6. Discourse relation marks it’s scope, but it is in concord with a higher polarity phrase, which contributes an epistemic component. A challenge about the epistemic contribution in correction that it is unclear where in the structure it shall be introduced. It was shown that negation in correction was not restricted syntactically, and could be DP negation as well as adverbial negation. On the other hand, the epistemic presupposition must be introduced lower than negation, since negation can never be part of it. For example, in a hypothesized correction Sandra ate an apple. – Sandra didn’t eat an apple but a pear, the epistemic presupposition might be it is believed that Sandra ate an apple, that is, an embedded clause without a negation. But this is also evidence to say that negation is not interpreted locally, since the presupposition of correction is the entire first conjunct without negation, and not only the part that is corrected in the first conjunct. As was shown in the syntax chapter 2 2, there is good evidence that preverbal corrections are in fact small coordinations of DPs. We also saw that kein* is considered to be a Negative Concord item in the literature. (167) Erika Erika hat has [keine no Erdbeeren, strawberries, sondern but ¨ Apfel] DP apples gegessen. eaten. Presupposition: Either Erika ate strawberries or It is accepted that Erika ate strawber- ries The interpretation of Negative Concord Items is proposed to proceed as follows in Zeijlstra Zeijlstra (2004 2004): many Negative Concord Items are negative indefinites. As such, their interpretation is divided into an indefinite and a negation operator, which each may be interpreted separately. This is argued based on the ambiguities which arise between negative indefinites and modals. (168) Du You musst must keine not-a Krawatte tie anziehen. wear. It is not obligatory that you wear a tie. :> must>9 There is no tie that you must wear. :>9> must It is obligatory that you do not wear a tie. must>:>9 Indeed we find that embedded in a correction, the latter readings which required that the modal move past negation is not available, which is evidence for the fact that negation must be interpreted as highest in its conjunct. This may be an effect of ellipsis, since the modal appears to be elided in the second conjunct, and the two conjuncts are required to share the same scopal makeup (cf. Fox Fox, 2000 2000). (169) Du You musst must keine not-a Krawatte tie sondern but eine a Fliege bowtie tragen. wear. It is not obligatory that you wear a tie. It is obligatory you wear a bow tie. 110 Chapter 3. Semantics and pragmatics of sondern 111 There is not a tie that you must wear. There is a bow tie you must wear. # You must not wear a tie. You must wear a bow tie. It is argued that by a process similar to Negative Concord Interpretation, the negation in corrections are interpreted as the highest element within their conjunct. The highest projection in the first conjunct is a polarity phrase, and as is proposed with Negative Polarity Items, there is polarity agreement with the PF–realization of negation and this highest phrase in the first conjunct. This also means the PF–realized negation, even if they are realized as sentential adverbs, are actually quantifiers. In addition, the Polarity Phrase which occurs as the highest phrase in the second conjunct bears an epistemic interpretation. We may assume a similar interpretation as the one proposed in Han and Romero Han and Romero (2001 2001). As a consequence, the LF based on which the above sentence is interpreted is like in the representation below. (170) : [ Erika ate some strawberries ] but [ Erika ate apples.] After this transformation, the component that is presupposed by the verum focus operator is the sister of negation, Erika ate some strawberries, C1 in the discussion below. It is proposed that the interpretation which anchors the sister of negation in the common ground is contributed by a factive epistemic operator(C1), as proposed in Han and Romero Han and Romero (2001 2001). The semantics for the operator in Han and Romero Han and Romero (2001 2001) is repeated again in (171 171) below. Some modification to their semantics are proposed. (171) JVERUMK gx/i =lp hs;ti lw:8w’2Epi x (w)[8w”2 Conv x (w’)[p2 CG w” ]] I am not providing a detailed analysis about whether this operator has the precise semantics required for corrections. Also, Han and Romero Han and Romero (2001 2001) discussion makes a number of particular assumptions which may not apply to this discussion. Han and Romero Han and Romero (2001 2001, 18) define Epi x (w) as the set of worlds that conforms to x’s knowledge in w, Conv x (w’) is the set of worlds where all the conversational goals of x in w’ are fulfilled, and where CG(w”) is the common ground. Adapted to the current situation, definitely negation must be taken into account. As a conse- quence, the operator is adapted to the following. This operator, while presupposing that its sister presupposition is in the common ground according to the speaker’s epistemic state, asserts that said proposition is negated. Admittedly this definition glosses over the Negative Concord processes sketched above, but it is a sufficient approximation of the sate of affairs. It is assumed that in fact, a different syntactic node contributes the negation. (172) JCORRK gx/i =lp hs;ti lw:8w’2Epi x (w)[8w”2 Conv x (w’)[p2 CG w” ]].: p As a consequence, the fact that the first conjunct is presupposed to be part of common ground is due to the operator CORR, which takes highest scope in the first conjunct. A syntactic structure is 111 112 3.7. The contrast condition between two conjuncts given below. (173) but a pear Sandra didn’t eat an apple CORR In this section, the discourse dependency by assuming a polarity focus operator. The function of this operator is to presuppose the content of the first conjunct minus negation, that is, C1, is part of the common ground. It was argued that Negative Concord in involved in the interpretation of correction, and the negation is actually contributed from within the conjunct. 3.7 The contrast condition between two conjuncts In section 3.1.1 3.1.1, it was identified that the second conjunct necessarily must be in a particular contrast relation with the first conjunct. In this relation, the nonnegated C1 and C2 must be mutually exclusive. I will argue that this is ensured since the second conjunct also has a lexicalized meaning component introducing this presupposition. It is fairly easy to produce “non–sequitur” examples where pitch accents are not realized in matching positions. This is shown again in (174a 174a). This demonstrates that there is a grammatical restriction which necessitates that the foci in the two conjuncts must match in their syntactic location. (174) a. # Sandra Sandra isst eats keine no BIRNE, pear, sondern but W ¨ ASCHT washes einen an Apfel. apple Rooth Rooth’s (1992 1992) mechanism to model contrastive focus as in the an AMERICAN farmer talks to a CANADIAN farmer–type of example, repeated from (138 138) above, is to ensure that the focus set of the latter is a subset of the former. For the case of correction, Rooth Rooth’s (1992 1992) definition of the contrast relation is not restrictive enough, since the subset relation can be established “accidentally”: in the example above, the focus operators are hypothesized to apply at the levels as marked here [An AMERICAN]C farmer talks to [a CANADIAN farmer]D. However, there are no restrictions mentioned about which syntactic relation the two foci must be in with respect to each other. As such, it is not clear whether Rooth Rooth (1992 1992) means for this process to apply across sentences or utterances. There are a number of cases where this type of contrast is wanted to occur across utterances. Example (175 175) is one of them. In this example, the two utterances refer to the same state of affairs, and the speakers intend for these sentences to be about the same set of alternatives. 112 Chapter 3. Semantics and pragmatics of sondern 113 (175) a. Edith wears a blue dress. b. Edith wears a GOLDEN dress. One of the open questions about Rooth Rooth’s (1992 1992) subset condition is that it is not specified specifically at which syntactic level the condition has to be met. In example (175 175), this subset condition could be met at the AP level, DP level, or sentence level. If the subset relation can be established across sentences, as is wanted, it can also be established across DPs in different syntactic positions across sentences. Consider the examples in (176 176) below. In these examples, the adjective in the object DP is read as contrastive to a subject of a preceding sentence. While these readings exist, a lot of arbitrariness is introduced since focus evaluation is possible on many different levels. In the third example, although the pitch accent may be well–placed, for the phrases to be interpreted contrastively and hence to fulfill the subset condition, a type of “reconstruction” of a pitch accent in the first sentence is necessary, that means, American in the first sentence must bear at least a secondary pitch accent. Although there is no pitch accent realized on American, it is still possible that Canadian is interpreted as contrastive to it. (176) a. An AMERICAN farmer milked a cow. The cow is owned by a CANADIAN farmer’s father. b. An AMERICAN farmer milked a cow. Its name is Resi. It is owned by a CANADIAN farmer’s father. c. An American farmer milked a COW. Its name is Resi. It is owned by a CANADIAN farmer’s father. This mechanism is taken to be at play in the case of corrections. It is not restrictive enough to match the constaints found in correction. The data show that the second conjunct of correction must be interpreted contrastively: neither is it acceptable that the second conjunct is realized without prosodic accent, as example (177 177) shows, nor that the second conjunct has a prosodic accent in a non–contrastive place, as is shown in (178 178). This restriction must be specified in the lexical semantics of sondern. (177) * Sandra Sandra hat has keine no ¨ APFEL apples sondern but Birnen pears gegessen. eaten. (178) * Sandra Sandra hat has keine no ¨ APFEL apples sondern but UNREIFE unripe Birnen pears gegessen. eaten. To ensure only well–formed focus configurations are created, it is not enough to specify the coordination height of the coordinator and that a focus semantic value it c–commands fulfills the subset relation; the entire C2 must be interpreted to be in contrast with C1. Rooth Rooth’s (1992 1992) system is 113 114 3.7. The contrast condition between two conjuncts quite permissive, and does not specify any further restrictions as to where in the syntax the contrast requirement must be fulfilled. Consider the sentences in (179 179), which are meant to be interpreted contrastively. This is a well–formed discourse, and there is a single focus evaluation in the second sentence which fulfills the contrast requirement, so its set of alternatives is a subset of the set of alternatives of the preceding sentences. Hypothetically, focus can be evaluated at multiple levels while still maintaining the contrastive requirement imposed by Rooth Rooth (1992 1992), as is meant to be represented by the different focus operators in this example. They represent the possibilities of syntactic heights for focus evaluation that are available in this example. In the context of the first sentence, the second sentence could be uttered and be interpreted contrastively with all of these different locations at which focus is evaluated. Therefore we can see that only to require that the second focus set is a subset of the first is not necessarily a restriction if we can freely choose at which level focus is evaluated in each of the conjuncts. This is a larger issue in focus evaluation, and is otherwise likely a desired consequence of Rooth Rooth’s (1992 1992) system of focus semantics, since contrast can indeed be established at different syntactic levels. (179) a. Erica bought these unripe apples. b. h Gi [ Sandrah Fi [ ateh Ei [ theseh Di [ freshh Ci [ PEARS ] ] ] ] ] In correction however, the coordination height fully determines the level at which focus/contrast is interpreted. In the following discussion, examples with different coordination heights are considered. For example, in example (180 180) below, CP level constituents are coordinated, at the same time placing prosodic focus on the subject in the second sentence. If focus in this sentence is interpreted at CP level, the two sentences are contrastive with respect to each other. However, the realization in (180 180) is not a good way of achieving this, and the better way of expressing this is to also put prosodic focus on the main verb gekauft ‘bought’, indicating that a larger constituent is preferred to be focus–marked. (180) ?? Ich I glaube believe nicht, not, dass that Erika Erika ¨ Apfel apples gegessen eaten hat, has, sondern but dass that ¨ APFEL apples gekauft bought worden have sind. been. Also, it is possible for the first conjunct to be reinterpreted as being contrastive, without the necessity for a pitch accent realized in this conjunct. This is demonstrated in the example in (181 181) below. In this example, a pitch accent is placed on the adjective sch¨ onen ‘beautiful’ in the second conjunct. Out of the blue, this sentence is not fully felicitous, demonstrating again that the entire phrase contained in the second conjunct must be contrastive. It is much improved by placing the pitch accent on Birnen ‘pears’. This acceptability contrast is not expected in Rooth Rooth’s (1992 1992) contrast condition since ugly pears may easily contrast with apples since they are of the same semantic type. 114 Chapter 3. Semantics and pragmatics of sondern 115 But if additionally it is assumed in the utterance context of (181 181) that apples are ugly, then the sentence becomes more acceptable, since if this assumption is made, ¨ Apfel ‘apples’ in the first conjunct can be accommodated as having the same syntactic structure as Birnen ‘pears’ in the second conjunct. Although I don’t want to provide an explicit account for this particular data point, it is evidence in favor of the entire second conjunct being perceived as in contrast with the first conjunct, and moreover, that contrast may still have a syntactic component, i.e. that syntactically similar constituents are construed in contrast more readily. (181) ?? Erika Erika hat has nicht not ¨ Apfel, apples, sondern but diese these SCH ¨ ONEN beautiful Birnen pears gegessen. eaten. A further restriction that needs to be introduced is that the second conjunct is necessarily in contrast with the first conjunct, not with any other discourse proposition. This is shown in example (182 182). The only way in which contrast with another discourse utterance is possible is when second occurrence focus is realized in the second conjunct. In example (182 182), two types of contrast are realized within the second conjunct. One type of contrast is with the first conjunct, ¨ Apfel. Another contrast is with a previous proposition in discourse, Sandra ate ugly pears. The grammaticality contrast between (182c 182c) and (182d 182d) shows that contrast between the conjuncts is required, and additional contrasts can only be introduced if contrast is realized between the conjuncts first. This is explained if the focus on sch¨ one ‘pretty’ in (182c 182c) is analyzed as second occurrence focus, which is realized on top of an existing focus structure. To satisfy the requirements of correction, this underlying focus structure must be in contrast with the first conjunct. Since this requirement is not fulfilled in (182d 182d), the sentence is ungrammatical. (182) a. Sandra Sandra hat has h¨ assliche ugly Birnen pears gegessen. eaten. b. Erika Erika hat has ¨ Apfel apples gegessen. eaten. c. Erika Erika hat has nicht not ¨ Apfel, apples, sondern but SCH ¨ ONE beautiful Birnen pears gegessen. eaten. d. * Erika Erika hat has nicht not ¨ Apfel, apples, sondern but SCH ¨ ONE pretty–ones gegessen. gegessen. From this discussion, it can be deduced that for the two conjuncts to only accidentally be in a relation where the context set of C1 is a subset of the context set of the focus in C2 is not a sufficient description for examples of correction. There are two possible implementations for this finding. Implementation 1 is to assume that the semantics of sondern include a presupposition that the context set of the sister of sondern be a subset of the sister of a previous constituent. Implementation 2 is that the context pronoun of the second conjunct is necessarily bound by the context pronoun of the first conjunct. 115 116 3.7. The contrast condition between two conjuncts An illustration of the two implementations is given in (183 183) below. (183) a. JsondernK=lC2:lC1: JC1K F JC2K F . C1^ C2 3 3 b. B i C2 sondern C i : C1 Both of these implementations ensure that the focus set of the second conjunct is a subset of the focus set of the first conjunct. However, it appears that the second formulation is on the one hand too restrictive, and otherwise also unprecedented in the literature, at least to my knowledge. It was shown that the second conjunct continues to be context dependent, and does not strictly share the context dependencies of the first conjunct, since it can receive additional second occurrence focus independently of the first conjunct. If its context set is necessarily bound, we would not expect this to be possible, since this implementation would predict for the two context sets to be identical. In addition, to my knowledge this type of approach has not been witnessed in any other type of construction. One might consider to explore this option a bit further, since the proposal has been made that presuppositions can be analyzed as pronouns (Beaver and Geurts Beaver and Geurts, 2014 2014), rendering the consequences of the first implementation and the second implementation very similar to each other. Therefore, a promising implementation is a lexicalized presupposition which requires the focus set of C1 to be a subset of C2. A subsequent question is whether the contrast requirement fulfills the usual criteria for presuppositionhood, to wit projection across presupposition holes like negation and certain predicates like m¨ oglich ‘possible’. A challenge in conducting this test is that it is a bit unclear how to verbally express the presupposition that is being tested. I propose Apples and pears are mutually exclusive, although this impression possibly arises from an interaction of the semantics of the first conjunct combined with the second conjunct. Under these assumptions, the projective behavior of the presuppositions is supported. (184) Es It ist is m¨ oglich, possible dass that Erika Erika nicht not ¨ Apfel apples gegessen eaten hat, has, sondern but Birnen. pears. ! Apples and pears are mutually exclusive. (185) Ich I habe have nicht not gesagt, said dass that Erika Erika nicht not ¨ Apfel apples gegessen eaten hat, has, sondern but Birnen. pears. ! Apples and pears are mutually exclusive. 3 In the previous discussion of the syntax, it was shown that corrections of bare Ns is less acceptable, all other constituents are acceptable. Hence the domains of C1 and C2 should be restricted to all types other than e 116 Chapter 3. Semantics and pragmatics of sondern 117 This presents evidence of a lexical entry for sondern that is minimally (183a 183a), introducing a presupposition about the relation of the focus set of C1 and C2. Other focus sensitive items also operate with presuppositions based on focus sets, for example, this is an important contribution of the semantics of even. 3.8 Conclusion In this chapter it was argued that the semantics of corrections with sondern must account for at least three properties. I) correction stands in a very particular discourse relation with an antecedent or accomodated sentence, and forms the first conjunct, which adds a single negation to said sentence. II) the content of the conjuncts of correction must be interpreted as mutually exclusive, and III) at first glance, correction appears to lack scalar implicatures. These properties were explained due to the following proposals. Correction is typically the denial of implicatures. For example, the fact that the two conjuncts may not be direct contradictions of each other was traced back to the fact that corrections are denials of implicatures. The scalar implicatures are contributed by an exhaustification operator which may scope higher than or lower than negation. As a consequence, scalar implicatures in the conventional sense are not detected right away, since there are ways of deriving a second conjunct which entails or is entailed by the first conjunct. Upon close observation, there are still effects of scalar implicatures detected. The relation of the first conjunct to a discourse proposition is an effect similar to verum focus. In accordance with the literature on verum focus, also here this discourse dependency is analyzed to be an operator introducing a presupposition that a given proposition is part of the common ground, which is most straightforwardly fulfilled by a preceding utterance of this proposition. The relation between the verum focus–type correction operator and the realization of negation is modulated by Negative Concord. The second conjunct is focus sensitive in that a presupposition is introduces, requiring the focus semantic value of the non–negated first conjunct to be a subset of the maximal focus semantic value of the second conjunct, and for the two proposition to be presupposed to be mutually exclusive. 117 118 3.8. Conclusion Briefly, this summarized in the following schematized tree which presents the relative scopal configuration of the operators. The operator$ in this tree represents that negation and exhaustivity may occur in either order. (186) B C2 sondern A C1 :$exh Corr-OP 118 Chapter 4 Semantics and pragmatics ofaber In this chapter, a discourse oriented semantics for the lexical item aber will be proposed. The major challenge in this undertaking is that aber is a coordinator which can occur both in counterexpecta- tional and corrective environments. It will be proposed that aber is a focus–sensitive coordination with a scalar presupposition, but otherwise has the semantics of the coordinator and. I will propose that the difference between the two readings arises due to different scopal configurations, and not due to polysemy, which differs from the analysis for but in English in Toosarvandani Toosarvandani (2014 2014). In addition to a scalar semantics for aber, it is also argued that a covert scalar focus associative particle SOLO is part of the semantics of correction with aber. Since this operator has contradictory presuppositions with the scalar component of aber, it relies on negation to reverse the entailment in the scalar presuppositions in one of the components. When negation outscopes this operator, it has an effect of reversing the scales, comparable to the process involved in licensing Negative Polarity Items, which provides an explanation for why negation is obligatory in corrections. 4.1 The puzzle In the literature for but, minimally two elementary readings are usually proposed, the corrective reading, and the counterexpectational reading. Examples for these readings are given in (187 187). (187) a. Erika didn’t eat an apple, but a pear. correction b. Erika went to the park, but it was raining. counterexpectational In many languages, these two readings are expressed with the same coordinator. At the same time, they are grammatically quite different. The most tangible grammatical difference is that corrections require negation or a downward entailing operator within the first conjunct, whereas counterexpectational sentences can be grammatical with and without negation in either conjunct. 119 120 4.1. The puzzle Based on evidence from languages like Spanish and German, which use two different lexical items in these two functions, it has been argued that these two communicative functions should be analyzed as distinct (Anscombre and Ducrot Anscombre and Ducrot, 1977 1977; Vicente Vicente, 2010 2010). Toosarvandani Toosarvandani (2014 2014) char- acterizes the relation between the two readings as the type of polysemy assumed in modals. He proposes that whereas corrections must have an empty ordering source, counterexpectationals need not. In languages like German, he proposes that there are different lexical items depending on the restrictions placed on ordering sources. He is the only one among the list of authors with an explicit proposal about the relation between the corrective and counterexpectational meaning. However, for German, this is not a complete picture. While German has a coordinator, sondern, which can occur only in correction, it has another coordinator, aber, which occurs in both func- tions, correction and counterexpectational sentences. That means there is an apparent element of optionality in the expression of correction, since correction can be expressed with both sondern and aber. From these observations, two questions arise: on the one hand, are corrections with sondern and aber different, and if they are, how so? On the other hand, can a uniform semantics for aber be given that will account for its use in corrections as well and counterexpectationals? In this work, I will present data that will show that the optionality that is sometimes witnessed between sondern and aber is only apparent, in fact, there are subtle contrasts between corrections with sondern and aber. I will propose a novel analysis of the semantics of aber which will help account for these differences. Aber will be analyzed as a focus sensitive coordinator with a scalar presupposition. As such, incorporates an obligatory silent component which resembles the focus sensitive particle even. A consequence of this analysis is that the polysemy between counterexpectational and corrective readings will be explained as an effect of scalar licensing or NPI licensing. This analysis fits in with an particular trait of German, namely that it has a comparatively rich inventory of scalar particles. The crucial data to support this analysis are cases where sondern and aber are not interchange- able. One such case is presented in (188 188) below. These cases involve scalar items and hence trigger scalar implicatures. It was shown in chapter 3 3 that sondern is apparently insensitive to scalar implicatures, and it appears the conjuncts are pragmatically independent of each other. Interestingly, the same is untrue for aber. In the first sentence, (188a 188a), the first conjunct has the implicature that Sandra also did not eat more than two apples. The second conjunct of sondern typically presents a proposition which violates the expectation created by the implicature of the first sentence. The same is not true for aber; aber ostensibly may not be in a position which contradicts the implicatures of the first conjunct. (188) a. Sandra Sandra hat has nicht not zwei, two, *aber/sondern aber/sondern drei three ¨ Apfel apples gegessen. eaten. 120 Chapter 4. Semantics and pragmatics of aber 121 b. Sandra Sandra hat has nicht not drei, three, Xaber/sondern aber/sondern zwei two ¨ Apfel apples gegessen. eaten. Although at first glance very comparable, there are subtle discoursive differences between the semantics of aber and the semantics of sondern. (188 188) demonstrates they differ in the relationship they allow for the scalar reasoning between their conjuncts, it will be shown that there are also a number of other significant differences between the two. While both of the particles will finally be analyzed as focus sensitive coordinations, it will be shown that their semantics and hence their discourse requirements are quite distinct. In the following section 4.2 4.2, I will present some relevant data about aber, and I will also compare if aber also shares the observations made about sondern in chapter 3 3. Then in section 4.3 4.3, I will present some previous findings made in the relevant literature, as well as a specific scalar semantics for aber will be proposed. 4.2 Data 4.2.1 Aber can occur in two constructions,sondern cannot One of the most significant properties about aber that this analysis needs to address is that this coordinator appears combines such distinct discourse functions. As was shown in (187 187), English but can occur in two readings that are quite different. An analysis that provides little insight into the regularities found in the distribution of the readings of aber might maintain that aber is ambiguous and has two distinct lexical entries. To the author, this is not a satisfying explanation since there continues to be a significant amount of grammatical and interpretational regularities distinguishing between these two readings. For example, aber–type of coordinations unify corrective and counterexpectational interpretations across many languages, which is a fact that needs additional explanation and is not expected if the two readings merely come about as a lexical ambiguity. In fact, this means that a number of languages share the same task of disambiguating between these two readings, and use remarkably similar tools in it. These types of large parallelisms are typically thought to originate from a grammatical process. The most crucial regularity across languages as well as the two German coordinators is that correction crucially relies on negation or an expression of downward entailment; the same is not true for counterexpectational sentences. This distinction is so definitive that sondern is always preceded by a downward entailing expression, whereas aber has no such restrictions. The examples in (189 189) show this difference clearly. Note that this is a feature of the speech act of correction rather than a lexical item: in certain circumstances, a downward entailing expression is also obligatory with aber, drawing doubt upon a simple ambiguity hypothesis. 121 122 4.2. Data When a downward entailing operator occurs in either of the conjuncts of aber, there is optionality as to whether such a sentence is interpreted as a correction or a counterexpectational sentence. The deciding factor here is not negation narrow scope within each of the conjuncts of aber. A sentence with aber must be interpreted as a correction only if negation or a downward entailing operator is present and narrow focus is realized in the conjuncts. (189) a. Erika Erika isst is-eating heute today (keine) no ¨ Apfel, apples, aber but sie she mag likes schon indeed ¨ Apfel. apples. b. # Erika Erika isst is-eating heute today keine no ¨ APFEL, apples, aber but sie she mag likes schon indeed ¨ Apfel. apples c. Erika Erika isst is-eating heute today *(keine) no ¨ APFEL, apples aber but BIRNEN. pears. In the following subsections, I will present some crucial features of aber which this analysis accounts for. 4.2.2 Observation 1: contradiction between contrasts prohibited It is a crucial property of corrections with sondern that C1 (the first conjunct without negation) and C2 (the second conjunct without sondern) are in contradiction with one another. (190) a. Sandra Sandra aß ate nicht not ¨ Apfel, apples, sondern but Birnen. pears. Inference: Sandra didn’t eat both apples and pears. In the previous chapter it was also found out that this also means that corrections with sondern can express contradictions that are not context dependent, for example, coordinations where C1 is the negation of C2. (191) Diese This Couch couch ist is nicht not bequem, comfortable, sondern but unbequem. uncomfortable. It not obvious to determine whether aber allows for the same inferences as sondern in (190 190). Since aber need not necessarily occur with negation, we can say that aber in its non–corrective use frequently coordinates items which are not contradictory. (192) Dieser This Raum room ist is sch¨ on, beautiful aber but dunkel. dark. These types of sentences can easily occur in corrections. Although sondern would be grammati- cal in this sentence, it conveys a very different meaning: this sentence with an added negation, with sondern instead of aber, can only be uttered if in the context of this utterance the room is expected to be either beautiful or light. The same sentence with aber does not have this connotation. 122 Chapter 4. Semantics and pragmatics of aber 123 (193) Dieser This Raum room ist is nicht not sch¨ on, beautiful, aber/sondern but hell. light. Sondern: the second conjunct implies that there is no valuation of the room in terms of being beautiful or not. Furthermore, aber is not acceptable in contexts with a contradiction that is not context dependent, such as the coordination of p and:p. Consider the example in (194 194) below. (194) # Diese This Couch couch ist is nicht not bequem, comfortable, aber but unbequem. uncomfortable. These examples show that aber, in contrast to sondern, requires that its conjuncts do not contradict each other. The example of differences in scalar contrasts with both sondern and aber was already given in example (188 188). This example fits in with the generalization presented in this section: the second conjunct of the infelicitous examples of (188 188) are in direct contradiction with the first conjuncts. In sentences with aber, the second conjunct may not contradict the entailments of the first conjunct, independent of whether this conjunct is negated or not, a requirement which sondern does not share. 4.2.3 Observation 2: simple scalar contrast requires negation In example (188 188), it was shown that within certain parameters, aber was felicitous as a coordinator in scalar correction. Given this observation, I want to step back and observe a factor that so far was taken for granted in the expression of corrections: in all cases of corrections, it is necessary that a downward entailing expression occur with aber. Given that this analysis’ goal is to provide a single lexical entry for aber to explain both the corrective and the counterexpectational reading, there is no explanation for the obligatoriness of negation in these contexts, since clearly, aber can occur without negation. Without assuming two lexical items or other mechanisms of polysemy for corrective as opposed to counterexpectational aber, the grammaticality contrasts in (195 195) remains a mystery. It is not clear why contrasting quantities is different from contrasting qualities, why the one would require negation, where the other one does not. (195) a. * Sandra ate three but two pears. b. * Sandra ate two but three pears. c. Sandra choose a light but cool room. Example (195 195) demonstrates that this is not an effect of the prohibition on contrast described as observation 1, since both orders of numerals are prohibited, and a contrast like (188 188) is not 123 124 4.2. Data encountered. The difference between the examples in (195a 195a) and (195b 195b) as opposed to (195c 195c) is also not obviously due to a syntactic restriction. What really sets the examples in (195a 195a) and (195b 195b) apart is that aber is used in a scalar comparison environment. On the other hand, to generalize that all examples with obligatory negation are necessarily scalar is too strong, since we find that some low coordinations of DPs, which do not have a scalar contrast, exhibit the same pattern. Consider example (196 196) below. In this example, the correction targets a DP, with no scalar inferences apparent. (196) Sandra Sandra hat has *(nicht) not einen an Apfel, apple aber but eine a Birne pear gegessen. eaten. Another crucial factor in corrections is narrow focus, that is, pitch accent placement in matching syntactic locations in both of the conjuncts. We can see that the vast majority of sentences where negation is obligatory also has narrow focus realized within the conjuncts, and contrast in a single location. The sentences in (197 197) show that in general, the correction consequent has to match the syntactic focus location of the correction target. In the examples in (197 197), the second conjunct is elided, other than a single constituent, providing evidence that the remnants of the ellipsis are in focus. However, the necessity for negation vanishes as the conjunct becomes larger, as is shown in (197e 197e), an example that Toosarvandani Toosarvandani (2014 2014) calls semantic opposition, to him a third discourse function of aber. (197) a. Sandra Sandra hat has *(nicht) not ERICH Erich einen an Apfel apple gegeben, given, aber but Erika. Erika. b. ?? Sandra Sandra hat has nicht not ERICH Erich einen an Apfel apple gegeben, given, aber but eine a Birne. pear. c. Sandra Sandra hat has Erich Erich *(nicht) not eine a BIRNE pear gegeben, given, aber but einen an Apfel. apple. d. ?? Sandra Sandra hat has Erich Erich nicht not eine a BIRNE pear gegeben, given, aber but Erika. Erika. e. Sandra Sandra hat has (nicht) (nicht) ERICH Erich eine a BIRNE pear gegeben, given, aber but Erika Erika einen an Apfel. apple. Toosarvandani Toosarvandani (2014 2014) characterizes the example in (197e 197e) as a separate category of the use of aber. In this analysis, it will be argued that these are special cases of counterexpectational aber. While these examples clearly have narrow focus, negation is not required in these sentences, and, in fact, is prohibited in many languages. While Toosarvandani Toosarvandani (2014 2014) treats these cases as distinct, I think their semantics will be sufficiently described by assuming multiple foci in these constructions. A fitting generalization is that scalar contrast in correction with aber must have negation, and so does a class of syntactically low coordinations. Narrow focus by itself is not a sufficient factor, the 124 Chapter 4. Semantics and pragmatics of aber 125 sentences that must have negation are characterized by narrow focus in addition to low coordination size. 4.2.4 Comparison withsondern In the previous chapter, three major properties of corrections with sondern were identified. It will be useful to know whether aber shares any of these properties, since such knowledge will make it possible to find a general description of the properties of correction across the influence of individual lexical items. One of these factors, the extent to which aber interacts with scalar quantification, was presented in previous sections. Contrast between conjuncts Since the goal is to propose a lexical entry for both the coordinator in corrective and counterex- pectational readings, one of the questions is whether it is possible to find a type of contrast that is well–defined for both types of readings. It was shown already that the two conjuncts of aber may not be contradictions. Although in scalar cases whether or not a sentence is a contradiction is sometimes hard to determine since scalar implicatures must be taken into account, such restrictions clearly exist in non–scalar cases. On the one hand, C1 and C2, that is, the conjuncts lower than negation and lower than aber, may not be contradictions. For aber to be used felicitously, it must be in principle possible that C1 and C2 are true. At the same time, when they are used in a correction, they are used in a particular situation where it is unlikely for them to be true at the same time. In the case of sondern, C1 and C2 needs to be false. In coordinations with aber where no negation occurs, C1 and C2 needs to be true. (198) a. Diese This Couch couch ist is nicht not ungem¨ utlich, uncomfortable, sondern but gem¨ utlich. comfortable. b. * Diese This Couch couch ist is nicht not ungem¨ utlich, uncomfortable, aber but gem¨ utlich. comfortable. (199) a. Es Is regnet, is-raining, aber but Sandra Sandra geht is-going in to den the Park. park. b. Die The Sonne sun scheint is-shining nicht, not, aber but Sandra Sandra geht is-going in to den the Park. park. Also in counterexpectational sentences that do not have any obvious narrow scope, (199 199) being one example, clearly it is the case that both propositions, C1 and C2, have to be true at the same time. However, negation is not a distinctive factor, since in counterexpectational sentences with negation like (199b 199b), there is no sense in which the C1 without negation, the sun is shining and C2, Sandra is going to the park are mutually exclusive. In fact, in a sentence with aber, often C1 and 125 126 4.2. Data C2 appears a more natural statement than not C1 aber C2. Given this paradigm, it can be deduced that there are crucial differences in aber and sondern regarding the availability of contradictory conjuncts. While felicitous with sondern, independent of negation, contradictory statements with aber are infelicitous. Polarity contrast In sentences with counterexpectational aber, there need not be an obligatory polarity contrast with a discourse proposition. In fact, both conjuncts can contain new information to the listener, without the sense that there are previous expectations how these pieces of information should be evaluated. This manifests in the fact that these sentences can be the answer to the question What happened?, which can start a discourse. That these propositions are new to the discourse is also communicated naturally with the addition of adverbs like gerade ‘just now’ and neuerdings lit. ‘newly’ in the example discourses. It can be seen in these discourses that even though there appears to be an expectation that the two conjuncts are in a salient, here causal or prohibitive relation with respect to each other, there is no expectation about the truth or falsity of each conjunct of these non–scalar/counterexpectational uses of aber. (200) A. What happened? B. fEs It hat has gerade just-now zu to regnen rain begonnen/ started Die the Sonne sun hat has gerade just-now zu to scheinen shine begonneng, started, aber but die the Sandra Sandra geht goes neuerdings newly trotzdem despite-it in to den the Park. park. 6! There was an expectation that it wasn’t going to start raining. 6! There was an expectation that the sun wasn’t going to start shining ! There was an expectation that given that it rained/that the sun shined, Sandra wouldn’t go to the park. When aber is used as coordinator in a scalar correction in the discourses in (201 201), the examples show that the discourses exhibit previous expectations about the truth and falsity of the conjunct containing negation. In addition, it is unnatural that the content of both of the conjuncts be discourse new information at the same time. (201) A. What happened? B. Die The Sandra Sandra hat has gerade just-now nicht not einige some ¨ Apfel apples gegessen, eaten, aber but alle. all. ! There was an expectation Sandra would eat (only) some apples. C. Die The Sandra Sandra will wants nicht not einige some ¨ Apfel apples essen, eat, aber but neuerdings newly alle. all. 126 Chapter 4. Semantics and pragmatics of aber 127 ! There was an expectation that Sandra would eat (only) some apples. Similarly, also with narrow scope nonscalar corrections, the discourse expectation is manifest. (202) A. What happened? B. Die The Sandra Sandra hat has nicht not ¨ Apfel, apples, aber but Birnen pears gegessen. eaten. ! There was an expectation that Sandra would eat apples. This indicates that corrections with aber, realized with narrow scope and negation, exhibit a similar discourse relation requirement than corrections with sondern. In the following paragraphs it will be tested if they share a relevant further property observed for sondern, namely whether these correction differ from the discourse proposition by a single negation added, like corrections with sondern do. The examples in (203 203) show that in corrections with aber, the restrictions are less strong. While, just like with sondern, negation is obligatory (cf. 203a 203a), it need not be in the first conjunct (cf. 203b 203b). Contrary to the examples with sondern, sentences with aber can also be a repetition of an antecedent sentence uttered in previous context, as in (203c 203c), although the sentence might be more natural with a concessive adverb like zwar ‘although’. It can also occur in configurations where the target of correction is a negated sentence, and double negation is needed to refer to this sentence in (203d 203d). (203) A: Sandra didn’t eat an apple. a. * B: Sandra Sandra aß ate einen an Apfel apple aber but eine a Birne. pear. b. B: Die The Sandra Sandra aß ate einen an Apfel, apple, aber but keine not-a Birne. pear. c. ? B: Sandra Sandra aß ate keinen not-an Apfel, apple, aber but eine a Birne. pear. d. B: Sandra Sandra hat has nicht not keinen not-an Apfel apple gegessen, eaten, aber but eine a Birne. pear. Although the reason for this difference is not immediately obvious, this indicates that aber has a more loosely grammaticalized relationship with negation. It also appears that the boundary between correction and counterexpectational is rather fluent. The sentence in (203c 203c) has the “feel” of a counterexpectational due to this lack of polarity contrast with the antecedent utterance. Comparable to sondern, when there is polarity contrast with a previous sentence, this contrast must be only in terms of polarity. Attempting to contrast a coordination with aber in more places than only polarity is likewise infelicitous, as shown in (204 204). Also, it is infelicitous to contrast in another property than polarity with a negated antecedent in (205 205). 127 128 4.3. Proposal (204) A. Sandra ate an apple. B. # Sandra Sandra hat has keine not-a Birne pear gegessen, eaten, aber but eine a Ananas. pineapple. (205) A. Sandra didn’t eat an apple. B. # Sandra Sandra hat has keine not-a Birne pear gegessen, eaten, aber but eine a Ananas. pineapple These data show that correction with aber mostly shares a crucial property with corrections with sondern, namely that they required exclusively polar contrast with a salient antecedent. However, since aber interacts more freely with negation, it was not necessary for negation to occur in the first conjunct. Furthermore, there is also the possibility of interpreting a sentence with a negation which follows an antecedent sentence with negation, and hence is not in this particular discourse relation with an antecedent, as a counterexpectational sentence. In cases where a downward entailing expression is obligatory, it is assumed that the same CORR operator that was argued for in the case of sondern is at work. 4.3 Proposal In the following section, I will present the proposal for the lexical semantics of aber that will account for these crucial properties. Building on some basic intuitions in Anscombre and Ducrot Anscombre and Ducrot (1977 1977), I propose that aber should be analyzed as a focus sensitive operator, taking two arguments. It will also be analyzed as having an obligatory scalar component. This will relate the semantics of aber with the literature about scalar particles and Negative Polarity Items. Crniˇ c Crniˇ c (2011 2011) argues that German is a language with a very detailed expression of scalarity. This work argues that aber should be added to this class. As is evident from the data shown so far, co-occurring with negation is a crucial feature of correction, whether it be expressed with aber, with sondern, or in another way. I argue that negation here is obligatory, since these corrective sentences share qualities of NPI licensing in a view like Chierchia Chierchia (2004 2004); Kadmon and Landman Kadmon and Landman (1993 1993); Krifka Krifka (1995 1995). These authors view the licensing of Negative Polarity Items as a rescue mechanism for scalar items which are are too low on the scale to make an informative contribution. I propose that the necessity of a downward entailing environment in correction should be explained in analogy to this view. Since in correction, the propositions which occur with a downward entailing operator otherwise are copies of an antecedent statement or presupposed statement, they contribute no new information to the discourse. Their informational contribution can be maximized by negating them. First, I will present the relevance of the scalar intuition with the example of the counterexpecta- tional sentences. Then, I will go into detail about the specificities of deriving the scalar/ narrow 128 Chapter 4. Semantics and pragmatics of aber 129 focus sentences. This will lead to the idea that there is not only one scalar sentence, but that scalar sentences have an exclusive scalar component. It will be shown that this makes the interpretation of the second conjunct of aber resemble the German scalar negative polarity item auch nur lit. ‘also only’, which is one of the translations of even in German, but demonstrably has different scalar presuppositions. As such, interaction with negation is expected, which will be presented in a following section. 4.3.1 Aber encodes likelihood In a well known analysis of the semantics of French mais ‘but’, Anscombre and Ducrot Anscombre and Ducrot (1977 1977) propose the following semantic representation for the reading of but that later came to be called counterexpectational. In a conjunction p but q, either p implies:q, or p implies a proposition which in turn implies the negation of q. They point out that the first conjunct necessarily has to be what they call “argumentatively stronger” than the second conjunct of pero and aber. 1 1 As shown in Anscombre and Ducrot Anscombre and Ducrot (1977 1977, 30, their ex.1-2), sentences where C2 entails C1 are ungrammatical in languages that have a specific counterexpectational coordinator like German and Spanish. (206) a. No es cierto, pero es probable. b. Das ist nicht sicher, aber das ist wahrscheinlich. It is not sure, but it is likely. (207) a. * No es probable, pero es cierto. 1 From Anscombre and Ducrot Anscombre and Ducrot (1977 1977, 29): Pour donner une description g´ en´ erale de PA (pero, aber), nous poserons les r` egles suivantes: Soient p et q deux phrases; ´ enoncer p PA q, c’est: 1. Pr´ esenter p comme un argument possible pour une ´ eventuelle conclusion r. 2. Pr´ esenter q comme un argument contre cette conclusion, i.e., dans le cadre de la th´ eorie argumen- tative pr´ esent´ ee dans Ducrot 1973, comme un argument pour:r. 3. Attribuer ` a q plus de force argumentative en faveur de:r que l’on n’en attribue ` a p en faveur de r. La suite p PA q, prise dans sa totalit´ e est donc argumentativement orient´ ee en faveur de:r. In oder to give a general description of PA (pero, aber) we establish the following rules: be p and q two phrases, realizing p PA q is to: 1. Present p as a possible argument for a conceivable conclusion r. 2. Present q as an argument against this conclusion, i.e. in the framework of the argumentative theory in Ducrot 1973, as an argument for:r. 3. Attribute more argumentative force to q in favor of neg r, than one attributes to p in favor of r. The conclusion of p PA q, in its entirety, is hence argumentatively oriented in favor of:r 129 130 4.3. Proposal b. * Das ist nicht wahrscheinlich, aber das ist sicher. It is not likely, but it is sure. In the present analysis, their insight is used to construct a semantic proposal using focus sensitive operators which also have scalar presuppositions. What is essential in Anscombre and Ducrot Anscombre and Ducrot’s (1977 1977) analysis is that the two conjuncts compete in argumentative strength, that is, they are ordered with respect to their pragmatic strength and contextual likelihood. Furthermore, although this ordering is most obvious in cases with scalar implicatures, or in cases where there is an entailment relations between the conjuncts, it contributes to the characteristic semantic and pragmatic inferences in all coordination uses of aber. Consider the following sentences with aber. This analysis shares the same intuitions as Anscombre and Ducrot Anscombre and Ducrot (1977 1977), but the means to achieve these facts are quite different. Toosarvandani Toosarvandani (2014 2014) presents an analysis, also in the vein of Anscombre and Ducrot Anscombre and Ducrot (1977 1977) in the question under discussion framework, which again is quite different from the current work. Both accounts do not account for the fact that German aber can occur in both corrections and counterexpectationals. (208) Es It regnete, rained, aber but Sandra Sandra ging went in to den the Park. park. In the present proposal, counterexpectational sentences with aber that occur without obligatory negation are thought to have propositional focus, like in (208 208), and on a constituent level in scalar cases and other examples where negation obligatory. In contrast to corrections with sondern, even in sentences with narrow scope, it cannot be said that the conjuncts in sentences with aber exclude each other. For example, in (208 208) both the predicate raining and not raining can easily be true in a situation where Sandra went to the park is true, in fact it is the conjunction with aber that appears to establish the relation between these two propositions. Focus alternatives are relevant As is implicit in Anscombre and Ducrot Anscombre and Ducrot (1977 1977), speakers also take other discourse propositions into account when interpreting sentences like (208 208). In the model of Anscombre and Ducrot Anscombre and Ducrot (1977 1977), these propositions are the negation of the second conjunct of counterexpectational coordinators, and further contextual entailments of the first conjuncts of a counterexpectational coordination. Furthermore, this is the case in general, be it in counterexpecta- tional or corrective sentences. I propose that the propositions taken into account for the evaluation of the second conjunct are created by a mechanism that is already in place, namely the creation focus alternatives of a context sentences and the first conjunct. In the case of wide focus, these alternatives might be propositional level or polar alternatives. If they are polar alternatives, a verum focus operator along the lines presented in chapter 3 3 must be assumed. In the case of narrow focus, 130 Chapter 4. Semantics and pragmatics of aber 131 they are minimally different propositions, derived via focus alternatives employed in Rooth Rooth (1992 1992) and other alternative semantics accounts. In counterexpectationals that have wide–scope focus, a discourse proposition that is taken into account is the negation of the second conjunct. In (208 208), “it rained” is contextually related with “Sandra went to the park”, and at the same time, “it rained” is related to “Sandra didn’t go to the park”. Clearly the conjuncts as phonetically realized in (208 208) are related to one another, but in addition, the first conjunct is also related with an alternative proposition to conjunct 2, namely its negation. In the given example, it is assumed that there is wide focus on each of the conjuncts, and their sets of alternatives are propositions without any requirement for semantic parallelism. An example scenario is given in (209 209) below. In this tree, the sentences enclosed byfg are chosen to serve as example for contextual wide scope alternatives. (209) Sandra didn’t go to the park ( Sandra stayed at home Edgar is cooking dinner ) It is raining ( Sandra is sleeping The sun is shining ) For the example (208 208), the context is such that only some of the alternatives are in a conditional relationship with one another. The proposition it is raining has a conditional relationship with Sandra didn’t go to the park and Sandra stayed at home, whereas it is not related to other propositions like Edgar is cooking dinner. In this situation, the prejacent of the second conjunct, Sandra didn’t go to the park, is in a contradictory relationship to some of the other alternatives, namely the alternative Sandra stayed at home in the example. It is the contribution of aber that this contradictory proposition is more likely in the given context. However, on the one hand it need not be that aber necessarily take wide scope, and on the other hand, it also need not be that the prejacent is in contradiction with the alternative proposition. The sentence in (210 210) illustrates both of these features. In this gapping sentence, the verb is elided, indicating that the content can be reconstructed from its syntactic context. In addition, the subjects and the objects in (210 210) bear pitch accents, indicating that at least the objects, if not also the subjects are in narrow focus (the subjects may also be contrastive topics). Furthermore, there is nothing inherent to the propositions Sandra ate strawberries and Miriam ate pears which makes them contradictory, they may well be true at the same time. However, it is the contribution of aber that in the context of evaluation of sentence (210 210), there was another alternative proposition in either the first or the second conjunct which is more likely. It may be Miriam never eats the same thing as 131 132 4.3. Proposal everybody else, for example. (210) SANDRA Sandra aß ate ERDBEEREN, Strawberries, aber but MIRIAM Miriam BIRNEN. pears. The current analysis proposes that aber imposes an order onto the set of alternatives of its conjuncts. It forms the coordination of all first conjunct alternatives with all second conjunct alternatives, and presupposes that one of the alternatives is more likely than the prejacent. Using the case in (208 208), the set of alternatives formed is given in (211 211) below. It is annotated with estimated likelihood, and the antecedent is printed in bold face. It can be seen that there are a number of alternatives which are more likely than the prejacent, that is, the propositions realized in the actual sentence. These are the conditions which license the use of counterexpectational, and, as we will see, also corrective aber. (211) 8 > > > > > > > > > > > > > > > > < > > > > > > > > > > > > > > > > : It is raining and Sandra didn’t go to the park (less likely) It is raining and Sandra stayed at home (very likely) Sandra is sleeping and Sandra didn’t go to the park (very likely) Sandra is sleeping and Sandra stayed at home (very likely) It is raining and Edgar is cooking dinner (may be) Sandra is sleeping and Edgar is cooking dinner (may be) The sun is shining and Sandra didn’t go to the park (may be) The sun is shining and Sandra stayed at home (may be) The sun is shining and Edgar is cooking dinner (may be) 9 > > > > > > > > > > > > > > > > = > > > > > > > > > > > > > > > > ; This type of argumentation makes counterexpectational sentences with aber resemble another focus sensitive operator, the adverb even. It is a common analysis of even that it introduces a presupposition with the content that the prejacent of a sentence with even is less likely than some or any (depending on the analysis) contextual alternatives. Consider the example in (212 212) for comparison: In this example, even associates with the focused VP, went to the park. It evokes the impression that it is unlikely for Sandra to go to the park. More formally, analyses of even (Karttunen and Peters Karttunen and Peters, 1979 1979; Crniˇ c Crniˇ c, 2011 2011) maintain that even introduces a presupposition which takes into account focus alternatives. The content of the presupposition is that the sentence without the contribution of even, the prejacent, Sandra went to the park in the example, is less likely than other focus alternatives, Sandra stayed at home in the example given. (212) Sandra even [ went to the park F ]. 8 > < > : Sandra went to the park (less likely) Sandra stayed at home (more likely) Sandra went to a caf´ e (may be) 9 > = > ; In this analysis, it is assumed that corrections have a likelihood presupposition similar to even. For (208 208), they maintain that the coordination C1 & C2, It rained and Sandra went to the park is not a 132 Chapter 4. Semantics and pragmatics of aber 133 likely proposition, given all the other things that Sandra could have potentially done, that is, given the focus alternatives of both conjuncts. The situation is less straightforward in scalar examples like (213 213) below. In analogy to the non–scalar cases, the proposition in the first conjunct, not eating three apples, is interpreted to be contextually related with whether the proposition in the second conjunct eating two apples is true, and just like in nonscalar cases, the proposition Sandra isn’t eating three apples and Sandra is eating two apples evaluated against the background of the alternatives of this proposition, characterized by the setfSandra isn’t eating three apples and Sandra is eating n applesjn2N + g, or potentially fSandra isn’t eating m apples and Sandra is eating n applesjm;n2N + g. In addition, contextual restriction must apply. (213) a. Sandra Sandra isst ate nicht not drei F three ¨ Apfel, apples, aber but zwei F . two. b. * Sandra Sandra isst ate nicht not drei F three ¨ Apfel, apples, aber but vier F . four. In the upcoming section 4.3.2 4.3.2, I will make explicit what the nature of these restrictions is. We can see that more refined semantic assumptions are necessary to characterize the implicatures and entailments between the first and the second conjunct. An important question is whether and to what extent the first conjunct counts as a contextual restriction to the first conjunct. 4.3.2 The scalar examples It the literature about but ever since Grice Grice (1975 1975), it is generally assumed that but should be analyzed as the coordinator and with an added interpretive component representing the expectation that these conjuncts are unlikely to co-occur. The present analysis also continues in this vein. A consequence of such an analysis is that the underlying semantic representation, barring other implicatures and presuppositions, is closely related to and. As such, it is expected that it is subject to the same constraints as and. In all examples so far, the scale considered is the scale of natural numbers. This scale is a clear example for argumentation since it has prevalent entailment relationships, which in turn makes predictions regarding likelihood. This is discussed in great detail in Crniˇ c Crniˇ c (2011 2011). In the sentence (213 213), semantic entailment imposes an order on the set of alternatives. An order likeh:::5;4;3;2;1i is manifest, with each item entailing all items that occur further towards the right on the list. This is different compared to the preceding examples, where we encountered rather contextualized orderings, like likelihood of coming to the party, etc, that is, ordering which can only be established with situational knowledge. A correction with aber must have a negation as part of one of its conjuncts. In general, in 133 134 4.3. Proposal negated sentences, the entailment relations are reversed. That means, for example, a sentence Robert doesn’t have two kids entails the propositions Robert doesn’t have three kids, Robert doesn’t have four kids, etc. As such, the scale appears reversedh:1;:2;:3;::::4i. It is an insight in Crniˇ c Crniˇ c (2011 2011) that entailments are related to likelihood, which he deduces from Kolmogorov’s third axiom, which states that the probability of mutually exclusive events in a given probability space sums to one. Specifically, he observes that If a proposition p entails a proposition q, q cannot be less likely than p. (Crniˇ c Crniˇ c, 2011 2011, 15). Taking this into account, we can deduce that 3 must not be less likely than 4, must not be less likely than 5 etc. As the same time, not 3 must not be less likely than not 2 must not be less likely than not 1 etc. With these insights in mind, I want to compare the grammatical example in (213a 213a) with the ungrammatical example in (213b 213b) in the same way I did before. The relevant set of alternatives is identical, with the contextual restrictions differing. They are represented as a tree for legibility. In each of the branches, entailment goes from top to bottom, which, following Crniˇ c Crniˇ c’s (2011 2011) account means that all items which are towards the top of the page are less likely, whereas the items towards the bottom of the page are equally likely or more likely than its alternatives which are higher up on the page. (214) a. not 3 but 2 not 2 &5 &4 & 3 & 2 & 1 not 3 &5 &4 & 3 & 2 & 1 not 4 &5 &4 & 3 & 2 & 1 b. * not 3 but 4 134 Chapter 4. Semantics and pragmatics of aber 135 not 2 & 5 & 4 & 3 & 2 & 1 not 3 & 5 & 4 & 3 & 2 & 1 not 4 & 5 & 4 & 3 & 2 & 1 What these trees show is that without further argumentation, the approach as sketched so far does predict the grammaticality difference between (213a 213a) and (213b 213b). The reason is that in both cases, alternatives at least as likely as the antecedent can be found. What is needed to account for this difference is a mechanism which will allow for conclusions across polarity contrasts. In chapter 3 3 we saw that since there are no entailment relations between an upwards monotonous scale and a downward monotonous scale (for example, not 2 does not have an entailment relation with 2 and vice versa), a further operator is needed to strengthen the sentence. The desired outcome for the set of alternatives of the second conjunct is that this conjunct be restricted to alternatives that are at an equal position on the scale or lower on the scale than the first conjunct. That is, in the example in (213 213), the alternatives in the second conjunct which are three or higher on the scale should be restricted. There are two available mechanisms are conceivable. On the one hand, one may use an operator encoding exhaustivity, or otherwise, since aber is sensitive to likelihood, the operator may also sway the likelihood of said alternatives in this way. Exhaustification of the second conjunct, in a position lower than the coordinator, has this effect. It is the effect of exhaustification to imply that all alternatives entailing the prejacent are entailed not to be true. In the data at hand, for example for the alternative two in the second conjunct, this means not three, four, five are implied. I will repeat the trees presenting the possible alternatives including the effect of exhaustification. The effects of exhaustification are given in parentheses (215) a. not 3 but 2 135 136 4.3. Proposal not 2 &5 (not 6+) &4 (not 5+) & 3 (not 4+) & 2 (not 3+) & 1 (not 2+) not 3 &5 (not 6+) &4 (not 5+) & 3 (not 4+) & 2 (not 3+) & 1 (not 2+) not 4 &5 (not 6+) &4 (not 5+) & 3 (not 4+) & 2 (not 3+) & 1 (not 2+) b. * not 3 but 4 not 2 &5 (not 6+) &4 (not 5+) & 3 (not 4+) & 2 (not 3+) & 1 (not 2+) not 3 &5 (not 6+) &4 (not 5+) & 3 (not 4+) & 2 (not 3+) & 1 (not 2+) not 4 &5 (not 6+) &4 (not 5+) & 3 (not 4+) & 2 (not 3+) & 1 (not 2+) Looking at this type of presentation, we can see that the implicature contributed by the exhaustifica- tion of the second conjunct in the grammatical case entails the information given in the first conjunct (not three or more entails not three), whereas in the ungrammatical case, it does not (not five or more does not entail not three). As such, the generalization may now be refined. Assuming that 136 Chapter 4. Semantics and pragmatics of aber 137 exhaustification is the underlying reason for the grammaticality difference, aber exhibits a restriction where the implicatures of the second conjunct may must entail the first conjunct. Furthermore, the tool for deriving these readings is once again embedded exhaustification. This also means that in the case of aber, this exhaustification is localized in the second conjunct. In accordance with the discussion of data in the chapter about sondern, it can be seen that these data can also be expressed with sondern. As was discussed there, the exhaustification of the first conjunct, not three, created the implicature not not one and not not two. With aber, it is desirable to have a restriction which does not rely on negation, since aber does not necessarily occur with negation. However, an alternative way of expressing this contrast may be that negation may not associate with exhaustification in the case of aber. Another relevant question is how this proposed exhaustification analysis interacts with the proposal that aber, since it incorporates a type of semantics of even, interacts with the likelihood component of even. This work proposes that the observation that aber only sometimes requires negation is due to an interaction of the likelihood component of aber with the exhaustification. Since it is one of the goals of this work to propose a unified semantics for aber, it is important to point out that the assumption of a scalar component like even as part of the second conjunct also runs into a problem here. The reason for it is the semantics of even that is generally assumed. In most analyses (Karttunen and Peters Karttunen and Peters, 1979 1979; Crniˇ c Crniˇ c, 2011 2011), it is maintained that even does not have a truth conditional contribution, but rather contributes a presupposition about the structure of the focus alternatives. The presupposition is that the prejacent of even, that is the sentence in which even is uttered without the contribution of even, is not the most likely among its alternatives. Felicity of even relies on an ordering in the set of alternatives, be it ordering via entailment, or via other functions like likelihood. Many orderings are inherent to a contexts, they may be likeliness of x to attend the party, scale of liked fruits, heaviness, or other kinds. When dealing with quantificational determiners like the natural numbers which were part of my examples so far, a salient ordering is entailment. Crniˇ c Crniˇ c (2011 2011) argues entailment is viewed to be an inverse relation to likelihood, since he maintains that if if a proposition p entails a proposition q, q cannot be less likely than p. The entailment relations among the natural numbers are described by:::5! 4! 3! 2! 1. This means there is one item on this scale that is entailed by all other items, namely 1. Since 1 is true whenever any of the alternatives of 1 is true, this means 1 is the either more or equally likely, but not less likely. Due to this intuition, likelihood is intuitively inverse to entailment, and a proposition that is entailed by many other propositions is therefore more likely. With this background, it can be seen that even is expected ungrammatical when it is entailed by all of its alternatives, that is, when it is the most likely item. Due to this restriction, the following sentence is ungrammatical. (216) * Sandra ate even one apple. The proposal for the semantics of aber has been to say that the coordination of C1 and C2 is 137 138 4.3. Proposal less likely that C1 and C2 alt , where C2 alt is any of the the focus alternatives of C2. A further option is that C1 and C2 may have to be less likely than C1 alt and C2 alt , or C1 alt and C2. Depending on where exhaustification occurs, the exhaustified component is, in parallel with sentences with even, not expected to the strongest item on the scale. Said restriction is also true for sentences with aber. There is a prohibition for both conjuncts to be the strongest item on the scale, which prohibits that two items which each are strongest on their scale are coordinated with aber. (217) * Sandra Sandra aß ate keine, none, aber but alle all ¨ Apfel. apples. However, Crniˇ c Crniˇ c (2011 2011) observes that there is a restriction preventing negation from occurring directly in the scope of even. When even and an exhaustifying component cooccur, negation necessarily must also occur, otherwise the scalar structure is infelicitous. The reason is that an even–like component is uninterpretable in the scope of exhaustification, since the latter maintains that the prejacent is the most likely, whereas the former maintains that the prejacent is the least likely. In this work, it is proposed that this process is at the root for explaining why negation is only sometimes obligatory with aber. This question will be reduced to a question of relative scope. Evaluating the focus for the scalar component of aber higher than at coordination level, and assuming a set of alternatives with two independent foci would solve this problem, since, for example, a proposition Sandra ate not two but one cucumber entails Sandra ate not three but one cucumber, and also, entailments for the second conjunct might be taken into account in other contexts for the evaluation of the first conjunct. For example, a sentence Sandra ate not four and three pears entails Sandra ate not four and two pears. Therefore, if the alternatives are assumed to be evaluated at a level higher than the coordination level, the scalar structures expected to be well–formed with scalar aber. In the following section, I will introduce some striking parallels between coordinations with aber and other scalar particles like even. 4.3.3 Proposal: aber requires exhaustification Other than negation, a crucial property that distinguishes the cases where negation is obligatory from the ones where it is not, is that narrow focus is obligatory in these sentences. (218) a. Sandra Sandra isst is-eating nicht not ¨ Apfel F , apples, aber but Birne F . pears. b. Sandra Sandra isst is-eating nicht not drei F , three, aber but zwei F two Birnen. pears I will propose that the fact that negation is obligatory in these sentences is due to an interaction between the semantics of aber, which have an even–type scalar component, and a further focus 138 Chapter 4. Semantics and pragmatics of aber 139 associating operator that occurs lower than the coordinator. I will show that when narrow focus is interpreted at a level lower than aber, the scalar component implicit in aber and the component representing focus find themselves in a configuration which resembles the licensing of the German NPI auch nur. I will follow Crniˇ c Crniˇ c’s (2011 2011) analysis, which models the contrast in NPI licensing between sondern, auch nur and einmal, all of which are NPIs, but they do not have identical licensing conditions. What is the difference between wide scope focus and narrow scope focus? One of the most perceivable differences between these two components is that narrow focus can have a pitch accent on a position that is different from sentence focus, whereas the other type of focus cannot. While both types of focus can be the answer to questions, wide focus is the answer to a question like What happened? whereas narrow focus is found when replacing wh pronouns like what and who in questions by their possible answers. This insight is capitalized on more explicitly in Roberts Roberts (2012 2012). (219) What happened? a. Sandra Sandra hat has ¨ Apfel apples gekauft. bought. b. * Sandra Sandra hat has ¨ Apfel F apples gekauft. bought. (220) What did Sandra buy? a. * Sandra Sandra hat has ¨ Apfel apples gekauft. bought. b. Sandra Sandra hat has ¨ Apfel F apples gekauft. bought. These sentences are also different in their interpretation: wide focus readings imply exhaustivity less explicitly. While in the wide focus reading, there is no connotation that the answer is exhaustive in the sense that all that happened is that Sandra bought apples, this reading is more clear for the narrow scope readings. In these readings, it is quite clear that the object is interpreted exhaustively. 4.3.4 [even [negation [ scalar exhaustivity ] ] ] as an NPI configuration The result from removing negation from corrections is ungrammaticality. This is unexplained if negation is simply a logical component that does not stand in any specific grammatical relation with correction. However, the situation that negation is grammatically required is unusual and requires explanation. An important fact that must be kept in mind for this is that there is no way in which disagreeing with a previous expression requires negation. Rather, this speech act, called denial (cf. Geurts Geurts, 1998 1998), is sometimes expressed with a downward entailing expression, but sometimes it is expressed 139 140 4.3. Proposal with other means. Consider the examples below for illustration. They show that there are a multitude of natural language means which can be used in denial, be they focus, or other adverbs. However, it is surprising that in the case of coordinative corrections expressed with sondern and aber, only one way is licensed. (221) A. Sandra wore a red F dress. B. She wore a green F dress. (222) A. Erica wants to go to the park. B. Erica rather wants to go to the zoo. Furthermore, using downward entailingness in correction is not just licensed, it is required. In the discussion of the relation of certain expressions with negation, there are two relations that are prominently discussed. On the one hand, there are Negative Polarity Items, that is, items which are either only grammatical within the scope of a downward entailing expression, or, which receive a different type of interpretation depending on whether they occur within the scope of negation or not, or Negative Concord, which is considered a subcase of Negative Polarity, which is described in the interpretation of negative indefinites. In this section, I want to propose that the operators at play in the semantics of aber form a Negative Polarity configuration in certain narrow scope environments. When such a configuration is encountered, a downward entailing expression is obligatory with aber. Negative Polarity Items are items which in many contexts require a downward entailing op- erator c–commanding them. They can also interact with other polarity operators, like operators licensing positive polarity items (Homer Homer, 2012 2012). One of the most discussed examples is English any. A sentence with any must contain a negation surface c–commanding any, otherwise it will be ungrammatical, as shown in (223 223). In this sense, one might say that in order to obtain certain interpretations for an NPI, negation is obligatory. I propose that an analogous process is at work in the case of the obligatoriness of negation in corrections with aber. (223) Ericfdidn’t buy/*boughtg any prunes. One of the reasons why there is a parallelism between NPI licensing and correction licensing is that indeed, a downward entailing operator is sufficient to license aber, and a literal expression of negation is not necessary. Consider following example: (224) Sandra Sandra aß ate kaum barely ¨ Apfel, apples, aber but Bananen. bananas. Sandra barely ate apples, but she ate bananas. Specifically, I propose that corrections with aber have a similar structure to the German Negative Polarity Item auch nur. Auch nur, lit. ‘also only’ is a German scalar NPI with one particular property, 140 Chapter 4. Semantics and pragmatics of aber 141 namely that it may not occur in the scope of clausemate negation. Consider these examples below. In these sentences, it can be seen that while auch nur ‘even’ is ungrammatical under clausemate negation, it can occur in the restrictor as well as the nucleus of a negative existential quantifier. (225) * Sandra Sandra hat has nicht not auch also nur only einen two Studenten students eingeladen. invited. Sandra didn’t even invite a single student. (226) Kein No Professor professor hat has auch also nur only einen one Studenten student eingeladen. invited. No professor invited even a single student. (227) Kein No auch also nur only eine one Minute Minute zu too spaet late gekommener coming Student student hat has die the Pr¨ ufung exam bestanden. passed. No student who came even a minute late has passed the exam. Another important interpretive component of auch nur are its scalar presuppositions. Auch nur can only associate with a component on the very bottom of a contextual scale. As a consequence, the following sentence is ungrammatical: (228) * Kein No Professor professor hat has auch also nur only f¨ unf five Studenten students eingeladen. invited. Crniˇ c Crniˇ c (2011 2011) analyzes auch nur to be a complex focus sensitive adverb, that is, a focus sensitive adverb that is composed of two adverbs which associate with the same constituent. In his analysis, auch nur is made up of two components, even on the one hand, and a component a bit similar to only, but with a scalar presupposition, which Crniˇ c Crniˇ c (2011 2011) calls SOLO. These two focus sensitive components, he argues, associate with the same item in a sentence. They introduce conflicting presuppositions: while even presupposes that its prejacent is less likely than its focus alternatives, SOLO presupposes that its prejacent is most likely among its focus alternatives. As such, they can never occur in their base position, since they have inherently infelicitous presuppositions. As a consequence for the licensing of auch nur, the component even has to move across negation, making the presuppositions combinable. This is necessary since the scalar contribution of one of the components must be reversed. This can happen through negation. Independently, it is maintained by Crniˇ c Crniˇ c (2011 2011) and many others (Karttunen and Peters Karttunen and Peters, 1979 1979) that EVEN must scope higher than negation, since it cannot be interpreted lower than negation. These authors argue that when it occurs on the surface as lower than negation, in fact it must move since otherwise the presuppositions are in contradiction. As a consequence, a sentence with auch nur has the schematized logical form as given in (229 229) below. (229) EVEN i not t i SOLO DP F 141 142 4.3. Proposal In this logical form, not reverses the direction of entailments provided by the component SOLO, making the presuppositions compatible with each other. The current analysis proposes that correction cases of aber end up with a similar configuration by accident. In previous sections, it was proposed that aber in counterexpectational is a scalar coordinator with the following lexical entry: (230) JaberK=ly 2 2 :lx : x^ y is contextually less likely thanJx^ yK f :x^ y This amounts to the effect of having a scalar component scoping higher than the level of coordination, and allows for the alternatives of both foci to be incorporated into the calculation of alternatives. When narrow scope is placed lower than the level of coordination, and especially in scalar components, I will assume that focus associates with an operator like the one proposed in Crniˇ c Crniˇ c (2011 2011, 134). Its definition is given in (231 231) below, slightly adjusted from Crniˇ c Crniˇ c’s (2011 2011) notation. (231) JSOLOK g,c (C, p, w) is defined only if8q2 C[q6= p! q is contextually less likely thanp] If defined, JSOLOK g,c (C, p, w) = 1 iff p(w) = 1 This creates a configuration of operators that resembles the unlicensed configuration of auch nur, that is, when it is not in the scope of negation. Different from the case of auch nur, even/aber already finds itself in the correct position without the necessity to move. According to Crniˇ c Crniˇ c’s (2011 2011) proposal, in this configuration the presuppositions of these two operators are irreconcileable, since EVEN/ABER requires the focus prejacent to be among the least likely, and SOLO requires its prejacent to be among the most likely among its alternatives. As also happens with the licensing of auch nur, a negation occurs, and this negation takes participates in reconciling the presuppositions of the alternatives. It is necessary here to be more explicit about the relative scope of the scalar operators assumed in corrections with aber. One scalar configuration to consider is whether or not there are focus operators with scalar SOLO in both of the conjuncts, that is, whether or not the syntactic structure might be approximated by the following tree structure. (232) ABER–EVEN SOLO YP F NEG SOLO XP F If aber is indeed parallel to auch nur, we would expect sentences with auch nur to be licensed with a realization of nicht that has polarity focus, like the realizations that we find in corrective 2 Both x and y have variable types 142 Chapter 4. Semantics and pragmatics of aber 143 sentences. As such, we would expect an example like (233 233) to be grammatical, but as the data show, this is contrary to fact. (233) A. Sandra Sandra hat has ¨ Apfel apples gegessen. eaten. B. * Sandra Sandra hat has nicht not auch also nur only einen one Apfel apple gegessen, eaten, aber but eine a Birne pear. This shows us that a component similar to auch nur cannot be construed across polarity negation. This might either be because the negation here is accepted as a sentence level negation, or because this type of negation provides interference for the focus associations, maybe because it associates with focus itself. Both of these hypotheses are plausible, the latter specifically since negation is known to be able to cause these types of interference for focus associations, as argued in Beck Beck (2006 2006). However, the cases in (233 233) are not a valid analogy to the case in (232 232) since in the former, no movement is necessary. The contribution of solo under the scope of negation in the second sentence of a dialogue like Sandra ate an apple — Sandra didn’t eat an apple? is described in (234 234). It represented the presupposition contributed by the operator solo. In a scalar reading, when realized within the scope of aber/even, this operator gives rise to contradictory presuppositions, illustrated by the presuppositions in (234c-i 234c-i), which have to be interpreted under negation. (234) a. Sandra Sandra hat has zwei two ¨ Apfel apples gegessen. eaten. b. Sandra Sandra hat has nicht not zwei F two ¨ Apfel apples gegessen. eaten. c. [ Sandra has not [ SOLO two F apples ] eaten ] i. JSOLO two F applesK is defined if for all alternative in the context other than “two apples”, they are less likely than p. ! “one apple” is more likely than “two apples”. The presupposition of SOLO is not fulfilled. ii. JSarah has not SOLO two F apples eatenK: it is not the case that Sarah ate two ap- ples. It is presupposed that Sarah hasn’t eaten one apple is more likely than Sarah hasn’t eaten two apples. When placing SOLO under the scope of negation, the presupposition is met. From this argumentation, we can deduce that if SOLO occurs anywhere in the first conjunct, it must occur higher than negation. However, this is independently excluded, since then the scalar component in aber-even and SOLO would have conflicting presuppositions. Furthermore, the only 143 144 4.3. Proposal way it can be saved is by being in the scope of even, as we can see that (234c-ii 234c-ii) is a suitable antecedent for even. Applied to (234c-ii 234c-ii), even will state that “It is not the case that Sarah ate two apples” is not the most likely among its alternatives. This presupposition is met since one of its alternatives “Sandra didn’t eat one apple” entails this sentence. As such, we expect that the following dialogue with focus on the numeral one should be ungrammatical: (235) Sandra Sandra hat has einen an Apfel apple gegessen. eaten. (236) # Sandra Sandra hat has nicht not einen F an Apfel apple gegessen. eaten. This is indeed the case. In order to save this sentence, it might be continued with sondern, or a similar continuation. Also a grammatical version of representing this interpretation is the sentence Sandra ate no apple. Reformulating it in this way has the consequence that there are no entailing alternatives created based off this sentence, since a potential alternative Sandra ate some/many/all apples does not entail this sentence. Another open question is whether or not the second conjunct is necessarily scalarly interpreted. Consider there may be a similar requirement as was posited in the interpretation of sondern is in place: there, a presupposition was assumed where the second conjunct was obligatorily in contrast with the first conjunct. However, evidence shows this is not the case with aber. It is not necessary that all instances of polarity focus have a corrective reading, and in fact, a corrective reading only arises when the second conjunct is interpreted as being in contrast (in the sense of Rooth Rooth, 1992 1992) with the first conjunct. Consider the paradigm in (237 237). In this example, it is shown that one of the characteristic requirements of correction, namely that there be different polarity across the conjuncts, only causes ungrammaticality in case the sentences have a comparable focus structure. This indicates that the example (237a 237a) is not a correction. (237) Antecedent: Sandra is eating two apples. a. Sandra Sandra isst is-eating nicht not zwei two ¨ Apfel, apples, aber but keinen nobody.acc k¨ ummert concerns das. it. Sandra is eating two apples, but nobody cares. b. Sandra Sandra isst is-eating nicht not zwei two ¨ Apfel, apples, aber but (*nicht) not einen. one. Sandra is-eating not two apples, but not one. Another configuration of scalar operators to consider is (238 238), which assumes that there is a scalar operator SOLO as part of the second conjunct. The clear problem with this type of representation is that the negation never c–commands solo, and as such it is expected ungrammatical. 144 Chapter 4. Semantics and pragmatics of aber 145 (238) ABER–EVEN SOLO YP F NEG XP F There are some exceptions for the generalization that narrow focus always needed negation, like in the example in (239 239) below. (239) A bright but cool room These examples are taken to be evidence that the relevant notion of alternatives must take into account version of exhaustification that takes into account the idea of “innocent exclusion”, which is discussed in the following paragraphs. Katzir Katzir (2013 2013) argues, following Fox and Katzir Fox and Katzir (2011 2011) and others, that the exhaustivity in focus sensitive particles like only only allows for conclusions about some, not all of the focus alternatives. He gives the example below (Katzir Katzir, 2013 2013, 338). From the sentence in (240 240), it can be concluded that John doesn’t own blue convertibles, but not that John doesn’t own expensive/cheap convertibles. That means, the presupposition of only, which excludes some of the alternatives does not apply to the entire set of alternatives. (240) John only owns RED convertibles. a. John owns red convertibles. b. John owns blue convertibles. c. John owns cheap convertibles. d. John owns expensive convertibles. This concept is formalized as the additional restriction on only that the propositions that are presupposed negated must be innocent exclusions. They are defined as follows. (Katzir Katzir, 2013 2013, 339) (241) Exhaustivity and innocent exclusion: a. JonlyK(p)(A)(w)() p(w)&8q2 IE(p;A)::q(w) b. IE(p, A) := T fB A : B is a maximal set in A s.t.:B[fpg is consistentg c. :B :=f:p : p2 Bg This gives a new terminology to describe the example in (239 239). Each of the conjuncts does not entail the negation of the respective other conjunct, that is, a bright room does not entail not a cool room, and a cool room does not entail not a bright room. As such, they can not be innocently excluded with respect to each other. We can see that in such cases, negation need not necessarily occur in sentences with aber. It is proposed that aber may be successfully defined 145 146 4.3. Proposal with a single lexical item with very restrictive presuppositions. Under certain circumstances, these presuppositions must cooccur with a downward entailing expression to be grammatical. These cases are interpreted as corrective. When the presuppositions of aber are met, a downward entailing expression is not necessary, yielding the counterexpectational cases. This means the notion of coherence in corrections must be characterized by innocent exclusions. Only items can occur in corrections when they are part of the set of innocent exclusions of the respective other conjunct. Furthermore, in addition to Crniˇ c Crniˇ c’s (2011 2011) proposal of the semantics of SOLO, a further presupposition must be added, namely that only innocent exclusions may be included in the focus alternatives. 4.3.5 Negation is obligatory The intended consequence of this analysis is that negation in aber now plays a role similar to the one negation plays in the licensing of a negative polarity item like auch nur. Is this analogy justified? In the sense that negation in both cases transforms an otherwise ungrammatical sentence into a grammatical sentence, it definitely is. I want to argued that the justification is deeper than that. NPI licensing in the view of Chierchia Chierchia (2004 2004); Krifka Krifka (1995 1995); Kadmon and Landman Kadmon and Landman (1993 1993); Crniˇ c Crniˇ c (2011 2011) and many others has the func- tion of maximizing the informativity of a given sentence. Minimizers are an intuitive example for this view. Intuitively, minimizers contribute a predication of the smallest possible quantity of help, in the case of (242 242). In the view of the above authors, on account of the small quantity, uttering this sentence would not be informative, since it is entailed by virtually all other sentences, and hence contributes the lowest end of a scale. However, adding negation makes this sentence the highest end of the negated scale, and makes it much more informative. (242) * Louise lifted a finger to help. How specifically can this analogy to aber be drawn? There are two possibilities to consider. The less insightful way is that we assumed for these cases an operator solo, which contributes this type of presupposition. More insightful will be to consider the necessary context for corrective aber. It must occur in a context or with a presupposition of the first conjunct (without negation), that means, in the case of (243 243), in the context of the first conjunct. Since these sentences are identical, it can be said that the left conjunct in (243 243) is a minimally informative contribution, which is made maximally informative by adding negation. (243) a. Antecedent: Sandra ate apples b. * Sandra Sandra aß ate ¨ Apfel, apples, aber but Birnen. pears. 146 Chapter 4. Semantics and pragmatics of aber 147 4.4 Summarizing In this section, I argued that the lexical entry of aber does not differ between corrections and coun- terexpectational sentences. In both cases, its contribution is a scalar focus associative coordination which asserts the coordination of the conjuncts is true, and presupposes that this proposition is not the most likely among the alternatives. It was argued that for narrow focus and scalar cases, another operator, SOLO, is introduced lower than the scope of negation in the correction. This operator, interacting with negation and solo, gives aber an interpretation similar to a Negative Polarity Item like German auch nur. It was argued that the view of this construction as “discoursive” negative polarity item is indeed wanted and informative. By contrast to sondern, aber need not necessarily have negation. Sondern does not necessarily have a scalar evaluation of the sort of even, and only an exhaustive component. As such, and because this exhaustive component cannot be interpreted by itself, also there, negation is necessary to license this component. 147 Chapter 5 Conclusion This dissertation proposes an analysis of the grammar of correction in German. German proved to be an interesting language to investigate in this respect, since it has a very fine–grained lexicon of corrective expressions, compared to other languages. In addition, it was shown that correction has a significant scalar component, and German also has a very explicit lexicon regarding scalarity. It is likely that the fine–grained corrective structure is a facet of this more general trait in German. The present work has presented a view in which the semantics of sondern and aber are a result of the combination of a number of operators, these operators are narrow focus, polarity negation, exhaustivity, certain scalar components, and coordination. Polarity Negation It is a crucial component of correction that there is either reference to a discourse antecedent or a presupposition. This dissertation provided a specific syntactic and semantic analysis for this interpretation. In syntax, the reading was triggered by negation agreeing with a syntactically higher polarity projection. In semantics, it was proposed that the lexical entry for this polarity projection is a polarity focus operator which brought about this interpretation. Since many types of negation have a sentential and a polarity focused reading, it is desirable for this process to be a derivational process, rather than an ambiguity. Lexical choices A detailed focus semantic analysis for the two corrective coordinators sondern and aber was proposed. Based in empirical differences regarding scalar implicatures, it was shown that aber was an inherently scalar coordination, with scalarity introduced at a level higher than the coordination. The interpretation of sondern and its apparent absence of scalar implicatures was due to a downward entailing operator and an exhaustivity operator. These operators could occur in a variable scope configuration. The shared components across these two lexical items are polarity negation, and the narrow focus 148 Chapter 5. Conclusion 149 scoping within the coordination. Furthermore, both types of correction share the presupposition that the focus alternatives have to be carried over between the two conjuncts. Another large component that these coordinators have in common is the mechanism in which negation was required to occur. It was argued that without negation, both corrections introduced that their negated conjuncts are the lowest item on a pragmatic scale, a configuration which is ruled out by the maxim of quantity. Due to this, in a process which was proposed for the licensing of Negative Polarity Items, negation had to be introduced to maximize informativity in corrections. 5.1 Open questions This type of analysis raises a number of avenues for future research. One of them is how these facts relate to correction in other languages. German, it was argued, has two different ways of expressing corrections, one of them with sondern, and another one with aber. There are many languages which only have one corrective coordinator, for example English. In these languages, can a similar process of disambiguating between corrective and counterexpectational readings be assumed? If these languages differ syntactically, does the syntactic representation interfere with the licensing of these readings? Or do languages which only have a single lexeme have ambiguity or polysemy to decide between corrections and counterexpectational readings, with corrections being like sondern and counterexpectational readings being like counterexpectational aber? An important criterion in addressing these questions is how scalar implicatures behave in such languages, and whether or not they behave uniformely. Another interesting extension of this work is the study of multiple corrections. A significant crosslinguistic difference in the realization of correction, which is not caused by any other cross– linguistic differences occurs in corrections with complex targets, that is, corrections where multiple items are corrected using just one sentence. In German, complex targets may involve the subject, whereas in English, such sentences are ruled out. (244) a. Sandra Sandra hat has keine not-a Birnen pears gegessen, eaten, sondern/aber but Erika Erika Feigen. figs. b. * Sandra didn’t eat pears, but Erika figs. Another group of languages, the Slavic languages, allow multiple corrections only if all targets are in the surface scope of negation. Here is an example from Bulgarian, provided by Roumi Pancheva (p.c.) (245) a. * Ivan Ivan jade ate ne not bob beans a but Maria Mary oriz. rice *’Ivan ate not beans but Mary rice’. 149 150 5.1. Open questions b. Ne Not Ivan Ivan jade ate bob beans a but Maria Mary oriz. rice ‘It’s not the case that Ivan ate beans but that Mary ate rice.’ These data are indicative of movement restrictions that only occur in multiple correction. It appears languages have different restrictions with respect to whether focused components can move past polar negation, or can be reconstructed under polar negation. These data also show some parallelism with the availability of pair–list reading in questions, which in turn points toward an explanation based on the grammatical restrictions of multiple focus construction, since it was argued that questions and focus use the same infrastructure (Beck Beck, 2006 2006). We also found multiple focus exceptional for sentences with aber. We saw that with single corrections, in general, when focus scoped lower than the coordination, correction was necessary. The same is not the case for multiple corrections. (246) Sandra Sandra isst is-eating ¨ Apfel, apples, aber but Erika Erika isst is-eating Birnen. pears. 150 Bibliography Anscombre, J.-C. and Ducrot, O. (1977). Deux mais en franc ¸ais? Lingua, 43(1):23–40. Beaver, D. I. and Geurts, B. (2014). Presupposition. In Zalta, E. N., editor, The Stanford Encyclope- dia of Philosophy. Metaphysics Research Lab, Stanford University, winter 2014 edition. Beck, S. (2006). Intervention Effects Follow From Focus Interpretation. Natural Language Semantics, 14(1):1–56. B¨ uring, D. (1997). The great scope inversion conspiracy. Linguistics and Philosophy, 20(2):175– 194. B¨ uring, D. and Hartmann, K. (1998). Asymmetrische Koordination. Linguistische Berichte, 174:172–201. B¨ uring, D. and Hartmann, K. (2001). The Syntax And Semantics Of Focus-Sensitive Particles In German. NLLT, 19(2):229–281. Chierchia, G. (2004). Scalar implicatures, polarity phenomena, and the syntax/pragmatics interface. Structures and beyond, 3:39–103. Chierchia, G., Fox, D., and Spector, B. (2009). Hurford’s constraint and the theory of scalar implicatures. In Presuppositions and implicatures. Proceedings of the MIT-Paris Workshop., volume 60, pages 47–62. MIT Press Cambridge, MA. Crniˇ c, L. (2011). Getting even. PhD thesis, Massachusetts Institute of Technology. Drubig, H. B. (1994). Island constraints and the nature of focus and association with focus. Arbeitspapiere des Sonderforschungsbereich, 340. Fox, D. (2000). Economy and semantic interpretation. Number 35. Cambridge, MA:MIT Press. Fox, D. and Katzir, R. (2011). On the characterization of alternatives. Natural Language Semantics, 19(1):87–107. 151 152 Bibliography Frey, W. (2004). A Medial Topic Position for German. Linguistische Berichte, pages 153–190. Frey, W. (2006). Contrast and movement to the German prefield. The architecture offocus [Studies in Generative Grammar 82], V . Molnar, 8:235–265. Frey, W. and Pittner, K. (1998). Zur Positionierung der Adverbiale im deutschen Mittelfeld. Linguistische Berichte, 176:489—–534. Geurts, B. (1998). The Mechanisms of Denial. Language, 74(2):pp. 274–307. Goldberg, L. M. (2005). Verb-stranding VP ellipsis: A cross-linguistic study. PhD thesis, McGill University. online at Link to part1 Link to part1 and Link to part 2 Link to part 2 last retrieved on May 1, 2017. Grewendorf, G. and Sabel, J. (1999). Scrambling in German and Japanese: Adjunction versus multiple specifiers. Natural Language & Linguistic Theory, 17(1):1–65. Grice, H. P. (1975). Logic and conversation. In Davidson, D. and Harman, G., editors, The logic of grammar, pages 64–75. Encino, CA: Dickenson. Guerzoni, E. (2004). Even-NPIs in yes/no questions. Natural Language Semantics, 12(4):319–343. Gutzmann, D., Hartmann, K., and Mathewson, L. (2017). Verum focus is verum, not focus. submitted for publication. Link to paper Link to paper, downloaded on 4/9/2017. Gutzmann, D. and Mir´ o, E. C. (2011). The dimensions of verum. Empirical issues in syntax and semantics, 8:143–165. Hamblin, C. L. (1973). Questions in Montague English. Foundations of language, 10(1):41–53. Han, C.-h. and Romero, M. (2001). Negation, focus and alternative questions. In WCCFL 20 Proceedings, pages 262–275. Citeseer. Han, C.-h. and Romero, M. (2004). Disjunction, focus, and scope. Linguistic Inquiry, 35(2):179– 217. Hartmann, K. (2000). Right node raising and gapping: Interface conditions on prosodic deletion. John Benjamins Publishing Company. H¨ ohle, T. N. (1992). ¨ Uber verum-fokus im Deutschen. In Informationsstruktur und Grammatik, pages 112–141. Springer. Homer, V . (2012). Domains of polarity items. Downloaded from Vincent Homer’s homepage Vincent Homer’s homepage on 4/1/2016. 152 Bibliography 153 Horn, L. R. (1985). Metalinguistic negation and pragmatic ambiguity. Language, pages 121–174. Jacobs, J. (1986). The syntax of focus and adverbials in German. In Abraham, W. and De Meij, S., editors, Topic, Focus and Configurationality, pages 103–127. Amsterdam: John Benjamins. Jacobs, J. (1991). Negation. In von Stechow, A. and Wunderlich, D., editors, Semantik: ein internationales Handbuch der zeitgen¨ ossischen Forschung, pages 560–596. Berlin/ New York: de Gruyter. Johnson, K. (2009). Gapping is not (VP-) ellipsis. Linguistic Inquiry, 40(2):289–328. Kadmon, N. and Landman, F. (1993). Any. Linguistics and Philosophy, 16(4):353–422. Karttunen, L. and Peters, S. (1979). Conventional Implicature. Syntax and Semantics: Presupposi- tion, 11:1–56. available at Link to paper Link to paper retrieved on 2017/3/12. Katzir, R. (2013). A Note on Contrast. Natural Language Semantics, 21(4):333–343. Koster, J. (1975). Dutch as an SOV Language. Linguistic Analysis, 1:111–136. Krifka, M. (1995). The semantics and pragmatics of polarity items. Linguistic analysis, 25(3- 4):209–257. Ladd, D. R. (1980). The structure of intonational meaning: Evidence from English. Indiana University Press. Maling, J. M. (1972). On ”Gapping and the Order of Constituents”. Linguistic Inquiry, 3(1):101– 108. McCawley, J. D. (1991). Contrastive negation and metalinguistic negation. In Chicago Linguistic Society, volume 27(2), pages 189–206. Merchant, J. (2005). Fragments and ellipsis. Linguistics and Philosophy, 27(6). Munn, A. B. (1993). Topics in the syntax and semantics of coordinate structures. PhD thesis, The University of Maryland. Penka, D. and Zeijlstra, H. (2005). Negative indefinites in Dutch and German. In 20th Comparative Germanic Syntax Workshop, Tilburg. Roberts, C. (2012). Information structure in discourse: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics, 5(6):1–69. This paper circulated as a manuscript since about 1996. 153 154 Bibliography Rooth, M. E. (1992). A theory of focus interpretation. Natural language semantics, 1(1):75–116. Ross, J. R. (1970). Gapping and the Order of Constituents. In Bierwisch, M. and Heidolph, K. E., editors, Progress in linguistics, pages 249–259. The Hague: Mouton. Sabbagh, J. (24–35). Right Node Raising . Language and Linguistics Compass, 8(1). Toosarvandani, M. (2013). Corrective but coordinates clauses not always but sometimes. NLLT, 31(3):827–863. Toosarvandani, M. (2014). Contrast and the structure of discourse. Semantics and Pragmatics, 7(4):1–57. Vicente, L. (2010). On the syntax of adversative coordination. Natural Language & Linguistic Theory, 28(2):381–415. Wagner, M. (2006a). Association by movement: evidence from NPI-licensing. Natural Language Semantics, 14(4):297–324. Wagner, M. (2006b). Givenness and locality. In Proceedings of SALT, volume 16, pages 295–312. Zeijlstra, H. (2004). Sentential negation and negative concord. Netherlands Graduate School of Linguistics. 154

Abstract (if available)

Abstract This dissertation proposes an analysis of the grammar of correction in German. ❧ The present work has presented a view in which the semantics of sondern `but' and aber `but' are a result of the combination of a number of operators, these operators are narrow focus, polarity negation, exhaustivity, certain scalar components, and coordination. ❧ Polarity Negation is a crucial component of correction there is reference to a discourse antecedent or a presupposition. This dissertation provides a specific syntactic and semantic analysis for this interpretation. In syntax, the reading is triggered by negation agreeing with a syntactically higher polarity projection. In semantics the lexical entry for this polarity projection is a polarity focus operator which brought about this interpretation. Since many types of negation have a sentential and a polarity focused reading, it is desirable for this process to be a derivational process, rather than an ambiguity. ❧ A detailed focus semantic analysis for the two corrective coordinators \emph{sondern} and \emph{aber} is proposed. Based in empirical differences regarding scalar implicatures, it was shown that aber was an inherently scalar coordination, with scalarity introduced at a level higher than the coordination. The interpretation of sondern and its apparent absence of scalar implicatures is due to the interactions between a downward entailing operator and an exhaustivity operator. These operators could occur in a variable scope configuration. ❧ The shared components across these two lexical items are polarity negation, and the narrow focus scoping within the coordination. Furthermore, both types of correction share the presupposition that the focus alternatives have to be carried over between the two conjuncts. ❧ Another large component that these coordinators have in common is the mechanism in which negation is required to occur. Without negation, both corrections introduced that their negated conjuncts are the lowest item on a pragmatic scale, a configuration which is ruled out by the maxim of quantity. Due to this, in a process which was proposed for the licensing of Negative Polarity Items, negation had to be introduced to maximize informativity in corrections.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Building phrase structure from items and contexts

PDF

Building adjectival meaning without adjectives

PDF

Narrowing the focus: experimental studies on exhaustivity and contrast

PDF

Where number lies: plural marking, numerals, and the collective-distributive distinction

PDF

Comparative iIlusions at the syntax-semantics interface

PDF

The case of a person: The person case constraint in German

PDF

Number marking and definiteness in Bangla

PDF

The grammar of individuation, number and measurement

PDF

Towards the unity of movement: implications from verb movement in Cantonese

PDF

Reference time in the dynamics of temporal dependency in Korean

PDF

Functional categories: the syntax of DP and DegP

PDF

Building a knowledgebase for deep lexical semantics

PDF

Decomposing Slavic aspect: the role of aspectual morphology in Polish and other Slavic languages

PDF

Soft biases in phonology: learnability meets grammar

PDF

The morphosyntax of states: deriving aspect and event roles from argument structure

PDF

Exploring the effects of Korean subject marking and action verbs’ repetition frequency: how they influence the discourse and the memory representations of entities and events

PDF

Syntax-prosody interactions in the clausal domain: head movement and coalescence

PDF

Constraining assertion: an account of context-sensitivity

PDF

A reduplicative analysis of sentence modal adverbs in Spanish

PDF

Prosody and informativity: a cross-linguistic investigation

Asset Metadata

Creator Steindl, Ulrike (author)

Core Title The grammar of correction

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Linguistics

Publication Date 10/03/2017

Defense Date 05/09/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag aber,but,correction,discourse,English,German,grammar,negation,OAI-PMH Harvest,pragmatics,semantics,sondern,syntax

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Iskarous, Khalil (committee chair), Pancheva, Roumyana (committee chair), Knight, Kevin (committee member), Simpson, Andrew (committee member)

Creator Email steindl.ulli@gmail.com,ulrike.steindl@gmx.net

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-441932

Unique identifier UC11265560

Identifier etd-SteindlUlr-5809.pdf (filename),usctheses-c40-441932 (legacy record id)

Legacy Identifier etd-SteindlUlr-5809.pdf

Dmrecord 441932

Document Type Dissertation

Rights Steindl, Ulrike

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA