* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download pdf version - Universität Leipzig
Modern Greek grammar wikipedia , lookup
Udmurt grammar wikipedia , lookup
Construction grammar wikipedia , lookup
Japanese grammar wikipedia , lookup
Chinese grammar wikipedia , lookup
Swedish grammar wikipedia , lookup
Old Irish grammar wikipedia , lookup
Portuguese grammar wikipedia , lookup
Preposition and postposition wikipedia , lookup
Distributed morphology wikipedia , lookup
French grammar wikipedia , lookup
Arabic grammar wikipedia , lookup
Georgian grammar wikipedia , lookup
Spanish grammar wikipedia , lookup
Ancient Greek grammar wikipedia , lookup
English clause syntax wikipedia , lookup
Latin syntax wikipedia , lookup
Cognitive semantics wikipedia , lookup
Old English grammar wikipedia , lookup
Determiner phrase wikipedia , lookup
Esperanto grammar wikipedia , lookup
Polish grammar wikipedia , lookup
Scottish Gaelic grammar wikipedia , lookup
Serbo-Croatian grammar wikipedia , lookup
Transformational grammar wikipedia , lookup
Yiddish grammar wikipedia , lookup
Junction Grammar wikipedia , lookup
Antisymmetry wikipedia , lookup
Welcome to the 25th Conference of the Student Organization of Linguistics in Europe It is our pleasure to welcome you to ConSOLE XXV! For twenty-five years now, the Conference of the Student Organization of Linguistics in Europe has provided a platform for young researchers to present their research in formal linguistics to an international audience. We are very proud that the 25th anniversary of ConSOLE takes place at the University of Leipzig. As is tradition, there are four invited speakers to cover the four main disciplines in theoretical linguistics. The first one is Željko Bošković, a syntactician from the University of Connecticut. The field of phonology is represented by Ricardo Bermúdez-Otero from the University of Manchester. Seth Cable from the University of Massachusetts Amherst will stand up for semantics. Greville Corbett, University of Surrey, represents the field of morphology. In addition, for the 25th anniversary, we have invited two more speakers: Martin Haspelmath from the Max-Planck-Institute in Jena and the University of Leipzig will give us a typological perspective and Doreen Georgi, a young professor from the University of Potsdam, working in the field of morpho-syntax. It is a pleasure for us that they all are with us and we hope to learn a lot from their suggestions and ideas. In addition to the scientific part of the conference, we also will have some social events where we can get to know each other better. There will be a wine reception on the first evening. It will take place at the Institut für Linguistik in the Geisteswissenschaftliches Zentrum (GWZ) which was built on the spot where the Gewandhaus, the famous concert hall, was situated from 1884 until 1944. The conference dinner is on Thursday: We are going to dine at the Barfusz, a popular café – bar – restaurant right in the heart of the bustling Barfußgäßchen. We would like to take the opportunity to thank our sponsors for their financial support that enabled us to organize the conference. Those are the Walter de Gruyter Stiftung, which generously covered our printing costs, the Research Academy Leipzig, which sponsored three of our invited speakers, the Vereinigung von Förderern und Freunden der Universität Leipzig e.V., which contributed to travel expenses, the Deutsche Forschungsgemeinschaft (DFG), which granted 2 us financial means to cover miscellaneous parts of the conference, and the J. B. Metzler Verlag, Brill Verlag, Buske Verlag and Linguistik Portal für Sprachwissenschaft, which all added their share to make ConSOLE XXV happen. In addition, we thank the Institut für Linguistik and the graduate school Interaktion grammatischer Bausteine (IGRA) for their support and help with various issues concerning the practical details of organization. We would like to give special thanks to Sabine Tatzelt and Jochen Trommer for their assistance with the bureaucratic, legal and financial hurdles. Thanks also go out to Ana Mitrović for her help with the graphic design of the posters and name tags. Last but not least, we are grateful to the many researchers in Leipzig and all over the linguistic world who invested their time in reviewing abstracts for ConSOLE and provided us and the speakers with helpful comments and criticisms. We hope you enjoy ConSOLE’s anniversary and your stay in Leipzig! Sincerely, Katja Barnickel Laura Becker Siri Gjersøe Matías Guzmán Naranjo Johannes Hein Sampson Korsah Local organizers Yuriy Kushnir Andrew Murphy Jude Nformi Ludger Paschen Zorica Puškar Joanna Zaleska Local organizers Kate Bellamy (Leiden) George Saad (Leiden) Anastasiia Ionova (Leiden) SOLE board Sponsors Deutsche Forschungsgemeinschaft (DFG) Research Academy Leipzig (RAL) Vereinigung von Förderern und Freunden der Universität Leipzig e.V. Walter de Gruyter Stiftung J. B. Metzler Verlag Brill Verlag Buske Verlag Linguistik Portal für Sprachwissenschaft Universität Leipzig Contents Welcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Program Finding the Venue Wednesday . . . . Thursday . . . . . . Friday . . . . . . . 1 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 . 10 . 12 . 14 Ricardo Bermúdez-Otero Željko Bošković . . . . . . Seth Cable . . . . . . . . . Greville C. Corbett . . . . Doreen Georgi . . . . . . Martin Haspelmath . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Daria Bikina . . . . . . . . . . . . . . . . . . Imke Driemel . . . . . . . . . . . . . . . . . Ragnhild Eik . . . . . . . . . . . . . . . . . Anastasia Gareyshina . . . . . . . . . . . . Ryoichiro Kobayashi & Ryan Walter Smith Sabine Laszakovits . . . . . . . . . . . . . . Mora Maldonado . . . . . . . . . . . . . . . Sofia Nikiforova . . . . . . . . . . . . . . . . Roberto Petrosino . . . . . . . . . . . . . . Cora Pots . . . . . . . . . . . . . . . . . . . Rong Yin . . . . . . . . . . . . . . . . . . . . Javier Sanz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Keynote Abstracts 17 Speaker Abstracts 19 21 22 24 26 27 29 31 35 39 44 49 54 58 62 65 69 73 77 4 Contents Ollie Sayeed . . . . . . . . . . Ui-Jong Shin & Sunjoo Choi Philip Shushurin . . . . . . . Jolijn Sonnaert . . . . . . . . Jelena Stojković . . . . . . . . Michael Wilson . . . . . . . . Anqi Zhang . . . . . . . . . . Ema Živković . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Beate Bergmann . . . . . . . . . . . Ana Bosnić . . . . . . . . . . . . . . Anya Chalupová . . . . . . . . . . . Patrick D. Elliot . . . . . . . . . . . . Henry Zamchang Fominyam . . . . Anastasia Gerasimova . . . . . . . . Ryosuke Hattori . . . . . . . . . . . . Hyunjung Lee . . . . . . . . . . . . . Samuel Alhassan Issah . . . . . . . . Katsumasa Ito . . . . . . . . . . . . . Júlia Keresztes . . . . . . . . . . . . . Ryoichiro Kobayashi . . . . . . . . . Elizaveta Kuzmenko . . . . . . . . . A. Marlijn Meijer . . . . . . . . . . . Daria Mordashova . . . . . . . . . . Takanobu Nakamura . . . . . . . . . Louise Raynaud . . . . . . . . . . . . Laura Vela-Plo . . . . . . . . . . . . Maria Zarifyan & Anastasia Melnik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Poster Abstracts 115 Practical Information Internet Access . . Copy Shops . . . . Lunch . . . . . . . Coffee Shops . . . . Restaurants & Bars 81 85 89 93 98 102 106 110 . . . . . 117 120 122 126 130 131 135 139 143 146 150 154 160 164 168 172 177 181 185 189 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 191 192 193 193 Program 7 Finding the Venue Address: University of Leipzig, Hörsaalgebäude/Seminargebäude Rooms: HS 8/S203/S205 Universitätsstraße 1, 04109 Leipzig Hörsaalgebäude Seminargebäude ConSOLE XXV ConSOLE XXV Entrance ConSOLE XXV Figure 1: Plan of the second floor of the Hörsaalgebäude. The conference will take place in the “Hörsaalgebäude” [hø:5za:lg@bOId@] and the adjacent “Seminargebäude” [zemina5g@bOId@]. Both buildings are part of the main campus of Leipzig University. They are situated in the city centre and are most easily reached from the Augustusplatz [PaUgUstUsplaţ] tram station. It is also only a 10-minute walk from the main station (Hauptbahnhof), where most trams also stop. To enter the Hörsaalgebäude, you will need to enter the courtyard. One way to do it is to use the passage between the “Neues Augusteum” (a building that looks like a church) and the VaPiano restaurant, and then turn left. The talks will take place on the second floor of the Hörsaalgebäude, in room HS8 and the poster session will take place in the area in front of room HS8. The registration will take place in room S205 in the Seminargebäude and coffee and snacks for the coffee breaks will be provided in room S203. Both rooms can be accessed from the second floor of Hörsaalgebäude (see plan of the building). 8 Program Wednesday, 4 January 2017, Hörsaal 8 08:00 – 09:00 09:00 – 09:10 09:10 – 10:10 10:15 – 10:45 10:45 – 11:05 11:05 – 11:35 11:40 – 12:10 12:15 – 12:45 12:45 – 14:00 14:00 – 14:30 14:35 – 15:05 15:10 – 16:25 16:30 – 17:00 17:05 – 18:05 19:30 – Registration (Seminarraum 205, continues after 9:00) Welcome The phonological lexicon, usage factors, and rates of change: Evidence from Manchester English HRicardo Bermúdez-Otero (University of Manchester)H A stratal analysis of truncation in Spanish: Morphological and phonological evidence Javier Sanz (Universität Trier) Coffee break (Seminarraum 203) Epenthesize a mora, but pronounce a vowel – The Serbo-Croatian language game of šatrovački Jelena Stojković (University of Leipzig) Some mathematical phonology – An extension of ‘delete-and-unify’ Ollie Sayeed (Christ’s College, Cambridge) The atoms of person: Limitations on concept formation Jolijn Sonnaert (KU Leuven) Lunch break The resultative passive is agentive: A response to Embick (2004) Michael Wilson (University of Massachusetts Amherst) Agreement with person mismatches in coordination Imke Driemel (University of Leipzig) Poster session with coffee Norwegian compounds: The category of non-heads Ragnhild Eik (NTNU) Nominal classification: New perspectives from Canonical Typology HGreville Corbett (University of Surrey)H Wine reception GWZ, Beethovenstraße 15 (Room H1 5.16 [fifth floor]) Program 9 Poster session, Wednesday, 4 January 2017 ¬ Feature inheritance and the syntax of lexical VV compounds Ryoichiro Kobayashi (Sophia University/JSPS) Parametrizing intralinguistic variation: Case assignment strategies in Russian event nominalizations Anastasia Gerasimova (Lomonosov Moscow State University, MSPU) ® Clausal pied-piping in Basque wh-questions and syntactic optionality Louise Raynaud (Georg-August-Universität Göttingen) ¯ Superiority, economy, and information structure Patrick Elliott (University College London) ° Pied-piping of adjuncts in Hungarian Júlia Keresztes (Pázmány Péter Catholic Univesity) ± Similarities and differences in lexical synonymy: Far and its synonyms in European languages Maria Zarifyan and Anastasia Melnik (National Research University Higher School of Economics) 10 Program Thursday, 5 January 2017, Hörsaal 8 08:00 – 9:00 09:00 – 10:00 10:05 – 10:35 10:35 – 10:55 10:55 – 11:25 11:30 – 12:00 12:05 – 12:35 12:35 – 13:50 13:50 – 14:20 14:25 – 14:55 15:00 – 16:15 16:20 – 17:20 19:30 – Registration (Seminarraum 205, continues after 9:00) Harmonic Agreement HDoreen Georgi (University of Potsdam)H The Left Periphery fragmented: Evidence from Italian Roberto Petrosino (University of Connecticut) Coffee break (Seminarraum 203) The Syntax of U/AX Right Dislocaction Ui-Jong Shin and Sunjoo Choi (Dongguk University, Seoul, Korea) Syntax of Finnish numerical constructions Philipp Shushurin (New York University) The de re reading and the universal quantificational phrase in Mandarin Rong Yin (University of Massachusetts Amherst) Lunch break Conjunctive or disjunctive? On the syntax/semantics of -toka and -tari in Japanese Ryoichiro Kobayashi (Sophia University/JSPS) and Ryan Walter Smith (The University of Arizona) Plurality and specificity in Spanish interrogatives Mora Maldonado (LSCP – Institut Jean Nicod/DEC – École Normale Supérieure) Poster session with coffee Negation and antonymy in Tlingit HSeth Cable (University of Massachusetts, Amherst)H Conference dinner Barfusz, Markt 9 (city centre) Program 11 Poster session, Thursday, 5 January 2017 ¬ Verbal aspect in Russian: Scale or continuum? Elizaveta Kuzmenko (National Research University Higher School of Economics) Complementizer agreement in Busan Korean Hyunjung Lee (Sogang University) ® Believing it vs. believing so A. Marlijn Meijer (Humboldt-Universität zu Berlin) ¯ The multifunctional morpheme /LE/ in Awing: A wh-focus conspiracy Henry Zamchang Fominyam (University of Potsdam) ° Mechanics of variation – A case study: Genitive and derived possessive in prenominal position in West Bohemian dialect of Czech Anya Chalupová (Palacký University in Olomouc) ± How causal modal particles influence the mental representation of discourses – experimental evidence Beate Bergmann (Humboldt-Universität zu Berlin) ² Phase, plurality and floating Takanobu Nakamura (Sophia University) 12 Program Friday, 6 January 2017, Hörsaal 8 09:00 – 10:00 10:05 – 10:35 10:35 – 11:55 10:55 – 11:25 11:30 – 12:00 12:05 – 12:35 12:35 – 13:50 13:50 – 14:20 14:25 – 14:55 15:00 – 16:15 16:20 – 16:50 16:55 – 17:55 18:00 – 18:30 Do aliens have UG? On efficiency of grammatical coding as an explanation for grammatical universals HMartin Haspelmath(MPI Jena/University of Leipzig) H Cognate adverb construction in Moksha Mordvin Sofia Nikiforova (National Research University Higher School of Economics) Coffee break (Seminarraum 203) Discourse-linked indefinite pronouns in Moksha Mordvin Daria Bikina (National Research University Higher School of Economics) The at-issue status of appositive relative clauses: Evidence for a discourse-based approach Ema Živković (University of Niš) Western Mari correlatives: Between two types of A′ -phenomena Anastasia Gareyshina (Independent researcher) Lunch break A configurational account of Turkish DSM Sabine Laszakovits (University of Connecticut) Displaced morphology in Dutch: Variation in non-finite verb clusters Cora Pots (KU Leuven) Poster session with coffee The syntax and semantics of event measurements in Mandarin Anqi Zhang (University of Chicago) Spelling-out phases HŽeljko Bošković (University of Connecticut)H SOLE business meeting Program 13 Poster session, Friday, 6 January 2017 ¬ The presupposition of exclamatives at the syntax-semantics interface: Evidence from German and Japanese Katsumasa Ito (University of Tokyo) Standards of comparison and the case of Spanish “que – de alternation” Laura Vela-Plo (University of the Basque Country) ® Labeling and two types of null operators in English Ryosuke Hattori (University of Connecticut) ¯ The question system of Dagbani Samuel Alhassan Issah (Goethe-Universität Frankfurt am Main) ° Hill Mari verbal constructions on ˆl’ˆ / ˆlˆn: Discontinuous past and beyond Daria Mordashova (Lomonosov Moscow State University) ee e e ± What is PO doing? Spatial distribution and group forming in Serbian Ana Bosnić (University of Groningen and University of Nantes) Keynote Abstracts 17 The phonological lexicon, usage factors, and rates of change: Evidence from Manchester English Ricardo Bermúdez-Otero (University of Manchester) [email protected] In classical modular feedforward architectures of grammar, the phonetic realization of lexical items depends solely on discrete phonological information encoded in surface representations. This hypothesis accounts for fundamental facts of human language such as double articulation and the existence of neogrammarian change, but it fails to make sense of the observation that fine phonetic detail is also affected by gradient usage-related properties of lexical items such as token frequency and neighbourhood density. Exemplar Theory seeks to explain the phonetic effects of usage factors by abandoning the classical hypothesis that lexical phonological representations consist solely of categorical information. Less radical approaches, however, continue to uphold this assumption: some, such as Baese-Berk & Goldrick’s (2009) account of neighbourhood density effects, rely on the notion of gradient symbolic computation, according to which lexical phonological representations are made up of symbols that are discrete but exhibit continuously varying degrees of activation (Smolensky & Goldrick 2016). These two approaches to the phonetic effects of usage factors differ in their diachronic predictions. In the case of lexical token frequency, in particular, it has been repeatedly observed that, synchronically, high-frequency words exhibit more lenition than low-frequency words. According to Exemplar Theory, this is because during historical language change high-frequency words undergo reduction at a relatively faster rate due to greater exposure to reductive phonetic biases, whose effects are registered in phonetically-detailed lexical representations. Pace Hay & Foulkes (2016), however, this diachronic pattern has never been reliably observed, and these accounts fail to consider another logical possibility: that high-frequency words are ahead synchronically but actually change at the same rate as low-frequency words. In this talk I report the findings of an investigation into the effect of lexical token frequency on the glottal replacement of word-medial /t/ in Manchester English, using apparent-time data from 62 speakers born between 1926 and 1985 (2131 tokens). Two stringent tests (mixed effects logistic regression and comparison between curve-fitting models) show that lexical frequency gives 18 Keynote Abstracts rise to a ‘constant rate effect’ in the sense of Kroch (1989). This is consistent with modified versions of classical modular architectures in which lexical phonological representations encode purely categorical information and in which the impact of token frequency is produced by time-invariant orthogonal mechanisms. References Baese-Berk, Melissa & Matthew Goldrick. 2009. Mechanisms of interaction in speech production. Language and Cognitive Processes 24 (4), 527–554. Hay, Jennifer & Paul Foulkes. 2016. The evolution of medial /t/ over real and remembered time. Language 92 (2), 298–330. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1 (3), 199–244. Smolensky, Paul & Matthew Goldrick. 2016. Gradient symbolic representations in grammar: the case of French liaison. Ms, Johns Hopkins University and Northwestern University. Available as ROA 1286 at the Rutgers Optimality Archive, http://roa.rutgers.edu. Keynote Abstracts 19 Spelling-out phases Željko Bošković (University of Connecticut) [email protected] An appealing property of the phase theory is that it is relevant to many phenomena, i.e. many domain-based mechanisms are stated in terms of phases. However, although phasal complements have no theoretical status in the phase theory (only phases do), they are taken to define spell-out units. The talk argues for an approach where phases define spell-out domains, which means that what is sent to spell-out is the phase itself. Several arguments to this effect will be presented regarding syntax-phonology interaction (concerning cliticization, raddoppiamento fonosintattico, tone sandhi, and stress assignment) as well as more theoretical issues such as labeling. The assumption, however, has significant consequences for successive-cyclic movement. If phases are sent to spell-out and what is sent to spell-out is inaccessible to the syntax, successivecyclic movement cannot target phases. Under the account explored in the talk, successive-cyclic movement therefore does not proceed via phases (i.e. phasal edges). As a result, the account also eliminates the Phase-Impenetrability Condition. 20 Keynote Abstracts Negation and antonymy in Tlingit Seth Cable (University of Massachusetts Amherst) [email protected] The central focus of this talk is the curious morpho-syntactic structure of certain negative predicates in the Tlingit language (Na-Dene; Alaska, British Columbia, Yukon). In Tlingit, there is a small but highly frequent set of stative, gradable antonym pairs, where the negative antonym is formed from: (i) the root of the positive antonym, (ii) the negation marker tlél (or hél), (iii) an additional (unproductive) morphological operation. To illustrate, the following are the expressions in Tlingit meaning ‘it is good’, ‘it is bad’, and ‘it is not good’. (1) a. yak’éi 0cl.good ‘It is good.’ b. tlél ushk’é neg irr.shcl.good ‘It is bad.’ c. tlél uk’é neg irr.0cl.good ‘It is not good.’ Note that while (1b) and (1c) both contain the negation marker, (1b) differs in that the so-called ‘verbal classifier’ prefix has shifted from ‘0’ to sh-. The primary goal of this talk is to develop and defend a formal syntactic and semantic analysis of negative predicates like (1b), one that both elucidates their morpho-syntactic structure and explains how that structure is mapped onto their observed meaning. In particular, I will show that: (i) the negation appearing in (1b) is VP-external, clausal negation, and is not an incorporated negation (unlike English un- or non-); (ii) the meaning of (1b) is indeed that of a gradable negative predicate, and is not simply the propositional negation of (1a) (unlike the meaning of (1c)). The case for these two claims will be based upon a variety of facts and phenomena surrounding these structures, particularly their interactions with degree modifiers. Under my proposed analysis, the morphological operation observed in (1b) – i.e., the change in the verbal classifier – is the effect of a special Degree-Operator, one that can only be licensed by (clausal) negation, and must undergo movement to SpecNegP (in effect, a negative-concord item). We will see that this analysis predicts a variety of facts concerning (1b), especially its syntactic/semantic contrasts with (1c). Furthermore, I show that the proposed analysis of (1b) has consequences for our understanding of negative gradable adjectives in English. In brief, so-called ‘Cross-Polar Nomalies’ (CPNs) have been argued to show that all negative adjectives in English contain an underlying negation (Büring 2007, Keynote Abstracts 21 Heim 2008). I show that similar ‘CPNs’ can be found in Tlingit. However, due to idiosyncrasies of Tlingit morpho-syntax, Büring’s (2007) analysis of CPNs has an advantage over Heim’s (2008) with respect to the Tlingit facts. 22 Keynote Abstracts Nominal classification: New perspectives from Canonical Typology Greville C. Corbett (University of Surrey) (joint work with Sebastian Fedden) [email protected] There are types of data which are handled by some as representing a single (complex) system and by others as two systems operating side by side. Nominal classification, including gender and classifiers, is one such domain. However, particular analyses are often assumed rather than argued for, which can leave important questions unanswered. Instances of explicit argumentation for one system versus two include Goddard (1982) on case and Round & Corbett (2017) on tense-aspect-mood. Our aim is a general typology of nominal classification, and a crucial component will be the application of explicit arguments for determining the number of systems involved in a given language. Hence a focus of the talk will be interesting languages which – arguably – have more than one system of nominal classification. The idea of an opposition between gender and classifiers was articulated clearly by Dixon (1982, 1986). He used a set of criteria to oppose gender systems and classifier systems, (his terms were ‘noun class’ and ‘noun classification’, respectively), and this approach was adopted in, for instance, Corbett (1991). While some of his criteria have stood the test of time, others have to be jettisoned or at least revised. Seifart’s (2005) account of Miraña presented a system with clear characteristics of gender and of classifiers, making it harder to maintain a divide between the two. And Reid (1997) on Ngan’gityemerri provided another reason against maintaining a clear gender-classifier divide, since classifiers can grammaticalize into gender systems, giving rise to a range of intermediate types. And recent research has uncovered more and more languages that combine gender and classifiers. These languages can be found mainly in South America, for example Tariana (Arawakan & Aikhenvald 1994, 2000), and Ayoreo and Chamacoco (Zamucoan & Bertinetto 2009, Ciucci 2013). A key language for us will be the Papuan language Mian, which is analyzed as having four genders as well as six classifiers that appear as prefixes on a subset of verbs (Fedden 2011). All this suggests that the sharp divide between gender and classifiers that seemed reasonable and attractive cannot be maintained. Once we see nominal classification like this, we can get a clearer picture of the range of possible systems. If we pull apart the characteristics we traditionally Keynote Abstracts 23 associate with gender systems, and those of classifier systems, we see that they combine in many ways. This is a cue to adopt a canonical perspective, in which we define the notion of canonical gender, and use this as an idealization to calibrate from. This allows us to situate the interesting combinations of properties present in some of the challenging systems we present, for instance in Mawng (Singer 2016). It also leads to a typology of concurrent systems, like that of Mian. 24 Keynote Abstracts Harmonic agreement: On prominence scale interactions in agreement Doreen Georgi (Universität Potsdam) [email protected] I provide an argument for the use of weights à la Harmonic Grammar (HG) in morphosyntax. It is based on multidimensional scale effects in agreement where several prominence scales (e.g. the person and the number scale) interact in determining (i) the agreement controller or (ii) the order of agreement affixes. I present examples for (i) and (ii) in which the individual scales are ranked, i.e. scale S1 outranks scale(s) S2 (S1 ≻ S2 ) in case of conflicting preferences. Such interactions are unexpected under the widely held assumption that person and number probe separately. Furthermore, the data seem to require the reverse scale ranking (S2 ≻ S1 ) in certain contexts. An implementation of (reverse) scale rankings in Optimality Theory (OT) leads to a ranking paradox; possible solutions require (a) context-sensitive constraints for particular scenarios or local conjunction. Neither of them derives the exceptions; local conjunction undermines the strict dominance property of OT and introduces a complex type of constraints. I show that the apparent exceptions fall out as cumulative effects, known from phonology, once the individual scales (and their members) are weighted according to their prominence, and scale interactions are modeled by adding the weights of individual scales. The agreement / ordering rules can refer to the harmony score resulting from the added weights. As a result, the rules are simple and without exception. There is no need for reversing scale rankings, context-sensitive rules or concepts like local conjunction to capture the facts. Keynote Abstracts 25 Do aliens have UG? On efficiency of grammatical coding as an explanation for grammatical universals Martin Haspelmath (Max-Planck-Institut Jena/Universität Leipzig) [email protected] Many linguists seem to assume that species-specific properties of human cognition constrain the possible grammatical systems of human languages, so that non-human languages would not be subject to these constraints, and might be subject to other kinds of constraints. In this talk, I make the claim that many of the most interesting universal findings are not necessarily human-specific, and would thus be expected to be found in alien languages as well. This is because they are due to highly general efficiency principles that constrain any system which spends energy but has limited resources. I exemplify the efficiency basis of grammatical universals by familiar and less familiar patterns of argument coding, and I partially contrast the efficiency explanation with Baker’s recent ideas concerning dependent case assignment. Speaker Abstracts 29 Discourse-linked indefinite pronouns in Moksha Mordvin Daria Bikina (NRU Higher School of Economics, Moscow) [email protected] The most widespread approach to describe the semantics of indefinite pronouns is the semantic map proposed by Haspelmath (1997). This map is based on five binary features; these binary distinctions divide the conceptual space into nine functions. However, it has been mentioned many times that there are some finer syntactic and semantic distinctions within Haspelmath’s functions, e.g. type of comparative consruction, counterfactual vs. hypothetic condition, subordinate vs. implicit negation and so on (Tatevosov 2002; Aguilar-Guevara et al. 2010; Kozhanov 2015). In Moksha Mordvin (Finno-Ugric < Uralic), there is a series of indefinite pronouns which is derived by reduplication of the corresponding interrogative pronoun: (1) m z’ard son sa-j? when he come-npst.3[sg] ‘When will he come?’ (2) son m z’ard m z’ard sa-j he when when come-npst.3[sg] ‘He will come someday.’ e e e e e e This series can occur in any irrealis context, comparative constructions, clauses with subordinate negation, but there is an additional restriction which is related to the semantics of partitivity or discourse-linking (D-linking) (Pesetsky 1987, 2000; Enç 1991). In such a context the referent either is supposed to be drawn from a set of individuals previously introduced into the discourse, or it belongs to the ‘common ground’ shared by speaker and hearer. Discourse linking of indefinite pronouns affects the acceptability of sentences in Moksha Mordvin – non-discourse-linked reduplicated pronouns are not allowed in imperative contexts (3, 4) and comparative sentences (5, 6): (3) t’εrt’-k kin’ kin’ lijε-n’ call-imp.sg.3[3.o] who.obl who.obl another-gen ‘Call somebody else.’ 30 Speaker Abstracts e dumanda-k mez’ -v k / *mez’ mez’ think-imp.sg what-add what what ‘Think of something.’ (5) mon kel’k-sa-jn’ š kaladnaj kanfeta-t’n’ -n’ čem kodam I love-npst-1sg.s[3.o] chocolate sweet-def.pl-gen than which kodam lijε-t’n’ -n’ which another-def.pl-gen ‘I love chocolates more than any other sweets.’ (6) a. e (4) e e e e e e e e e mon kel’k-sa mora-ma-z’ -n’ čem I love-npst[1sg.s.3sg.o] sing-nzr-1sg.poss.sg-gen than mez’ -v k what-add b. *mon kel’k-sa mora-ma-z’ -n’ čem I love-npst[1sg.s.3sg.o] sing-nzr-1sg.poss.sg-gen than mez’ mez’ what what ‘I love singing more than anything.’ e e e e e It is also significant that reduplicated pronouns with partitive meaning can be used even in specific contexts (7), while an equivalent indefinite pronoun which is not discourse-linked is impossible in any specific context (8): e e (7) mon oš-st rama-n’ vet’ panar-t, kona-nc ˚ which-3sg.poss.sg.gen I city-el buy-pst.1sg five shirt-pl kona-nc t’ej -t rama-jn’ which-3sg.poss.sg.gen pron.dat-2sg.poss buy-1sg.s[pst.3.o] ‘I bought five shirts in the city, one of them I bought for you’. (8) učit’el’-s’ mez’ b d’ / *mez’ mez’ s’ormac’ teacher-def what indef what what write.pst.3[sg] doska-t’i, no mon ičk z’-d’ iz’-in’ n’εj board-def.dat but I far-abl neg.pst-1sg.s[3.o] see.cn ‘The teacher wrote something on the board, but I could not see it from afar.’ e e e e e e e e e e e The reduplication of indefinite pronouns is not the only case in Moksha Mordvin when discourse linking of an indefinite pronoun series expands its distribution 31 Daria Bikina on the semantic map. The discourse linking of pronouns with indefiniteness marker b d’ evokes similar semantic effects: only discourse-linked b d’ pronouns are possible in imperative contexts (9, 10): e e e e e e e e e sa-k m z’ard -ng / *m z’ard b d’ when indef come-imp.sg when-add ‘Come someday.’ maks -t’ kodam -v k / OK kodam b d’ šava-n’ε bufet -st give-imp.sg which-add which indef plate-dim cupboard-el ‘Give me a plate from the cupboard.’ e e e e e e e e (10) e e (9) The data from Moksha Mordvin shows that the semantics of indefinite pronouns depends on its discourse-linking. I suppose that the requirement for discourselinking of indefinite pronouns in some contexts is associated with semantic weakening or desemanticization as it was described in (Haspelmath 1997: 146–154). Indefinite pronouns extend their original free-choice function to the left of the map and lose non-specificity. Whereas specificity can be explained in the terms of partitivity (Enç 1991; von Heusinger, Kornfilt 2005), the data from Moksha Mordvin allow us to fixate an intermediate stage of this process. In the talk, we will take a closer look at the cases when discourse-linking affects the acceptability of an indefinite pronoun series. Whereas the direction of the extension is not obvious in the case of Moksha Mordvin, I will also discuss what the problems are with dealing this language data with Haspelmath’s theory of desemanticization. Abbreviations 1, 3 – 1st, 3rd person, add – additive particle, dat – dative, def – definite, dim – diminutive, el – elative, gen – genitive, imp – imperative, indef – indefiniteness marker, npst – nonpast, nzr – nominalizer, o – object, obl – oblique, poss – possessive, pron – personal pronominal stem, pst – past, s – subject, sg – singular References Aguilar-Guevara, A., M. Aloni, A. Port, R. Šimík, M. de Vos and H. Zeijlstra. 2010. Indefinites as fossils: a synchronic and diachronic corpus study. Ms, University of Amsterdam. Enç, M. 1991. The semantics of specificity. Linguistic Inquiry 22, 1–25. Haspelmath, M. 1997. Indefinite pronouns. Oxford: Clarendon Press. 32 Speaker Abstracts von Heusinger, K. and J. Kornfilt. 2005. The case of the direct object in Turkish: Semantics, syntax and morphology. Turkic languages 9, 3–44. Kozhanov, K. 2015. Lithuanian indefinite pronouns in contact. Contemporary Approaches to Baltic Linguistics, eds. Arkadiev P., A. Holvoet and Bj. Wiemer. Berlin: Walter de Gruyter, 465–490. Pesetsky, D. 1987. Wh-in-situ: Movement and inselective binding. The Representation of (In)definiteness, eds. Reuland E. and A. ter Meulen. Cambridge, MA: MIT Press, 98–129. Pesetsky, D. 2000. Phrasal Movement and Its Kin. Cambridge, Mass.: MIT Press. Tatevosov, S. 2002. Semantika sostavjajushchih imennoj gruppy: kvantornye slova (The semantics of the constituents of the NP: quantifier words). Moscow: IMLI RAN. 33 Agreement with person mismatches in coordination Imke Driemel (Universität Leipzig) [email protected] Introduction Coordinations of nouns create problems for agreement if conjuncts differ in φ-features in that they call for additional resolution rules to determine the values the agreement target has to copy. Person resolution typically follows a hierarchy of the form 1 ≻ 2 ≻ 3 (Zwicky 1977, Corbett 1983), with one well-known exception, namely German verbal agreement with coordinated subjects (Findreng 1976, Timmermanns et al. 2004). The pattern in (1) shows consistent resolution agreement in number. Person resolution, however, cannot account for 3pl in (2). (1) a. Ich und mein Freund trag-en zu viel Verantwortung. I and my friend carry-1pl/3pl too much responsibility ‘I and my friend carry too much responsibilities.’ b. Du und dein Freund trag-t/trag-en zu viel Verantwortung. you and your friend carry-2pl/carry-3pl too much responsibility ‘You and your friend carry too much responsibilities.’ c. Ich und du *trag-t/trag-en zu viel Verantwortung. I and you carry-2pl/carry-1pl too much responsibility ‘I and you carry too much responsibilities.’ None of the judgements change if the order of the conjuncts is switched. Hence, an alternative strategy along the lines of closest conjunct agreement (Bhatt and Walkow 2013, Marušič et al. 2015) cannot provide a solution. I propose an analysis, couched in the framework of distributed morphology (Halle and Marantz 1993, 1994), that derives resolution agreement without stipulating a hierarchy but with the help of impoverishment (Bonet 1991, Frampton 2002). Morphological Decomposition In order to derive the paradigm, shown in table 1, I follow Müller (2006), considering the vocabulary items given in (2), with one important exception: following recent assumptions in Albright and Fuß (2012), I assume that the 2pl/3sg syncretism is the result of accidental homophony (see also Sauerland and Bobalijk 2013). This assumption is due to a consistently observed intervention effect of verbs which prohibit 2pl agreement in data of type (2) if the full form of 2pl is 34 Speaker Abstracts Table 1: Verbal inflection markers in German. 1st 2nd 3rd strong pres past sg pl sg pl pres sg pl -@ -st -t -@ -st -t -n -t -n -∅ -st -∅ -n -t -n -n -t -n weak past sg pl -te-∅ -te-st -te-∅ -te-n -te-t -te-n syncretic with 3sg form. Compare (2): 2pl trag-t, 3sg träg-t to (3): 2pl geh-t, 3sg geh-t. If there is no stem allomorphy, the affix is ambiguous between (2a) and (2e), and since pl agreement is obligatory in contexts like (1), an intervention effect is predicted. The difference between (2) and (3) cannot be explained with -t maximally underspecified as [−speaker]. (2) a. /-t/ ↔ [−speaker,−hearer,−pl,+pres] b. /-te/ ↔ [−pres,−strong] c. /-st/ ↔ [+hearer,−pl] d. /-n/ ↔ [−hearer,+pl] e. /-t/ ↔ [+hearer,+pl] f. /-(e)/ ↔ [ ] (3) Du und dein Freund *geh-t/geh-en viel ins Kino. you and your friend go-2pl/go-3pl often in.the cinema ‘You and your friend often go to the cinema.’ Resolution Agreement I assume that the coordinator und, being the head of its own functional projection and taking the conjuncts as its arguments (Munn 1993, Zhang 2009), bears an already valued number feature for plural and separate unvalued person probes that gather the person features of its arguments via cyclic agree (Béjar 2003, Řezáč 2003, Béjar and Řezáč 2009). The valued φ-features project to the root node &P which acts as the closest goal for agreement with T, see (6). (4) [&P [∗num:pl∗,∗pers:1,2∗] [DP ich[∗num:sg∗,∗pers:1∗] ] ... ... [&′ [& und[∗num:pl∗,∗pers:∗,∗pers:∗] ] [DP du[∗num:sg∗,∗pers:2∗] ]]] This mechanism opens the doors for resolution agreement. Mimicking the denotation of the sum operator ⊕, proposed for non-clausal coordination (Link 1983, Hoeksema 1983, Krifka 1990), I will assume that person resolution is performed by the set union of the person features of the conjuncts (see also Imke Driemel 35 Darymple and Kaplan 2010). With the use of the decomposed person features [±speaker] and [±hearer] (Noyer 1992, Wiese 1994), the following functional morphemes for T are derived (ignoring for simplicity the weak/strong and present/past distinction): (5) a. 1sg ∪ 3sg = [−speaker,+speaker,−hearer,+pl] b. 2sg ∪ 3sg = [−hearer,+hearer,−speaker,+pl] c. 1sg ∪ 2sg = [−speaker,+speaker,−hearer,+hearer,+pl] Fission and Impoverishment The marker (e) is abstract in that its phonological realization depends on whether there is stem alternation from present to past: (e) → /∅ if there is no stem alternation (weak forms) and (e) → ∅ if there is stem alternation (strong forms). Moreover, fission (Noyer 1992) ensures that weak past tense inflectional markers can contain the additional vocabulary item: -te (see Müller 2006). In order to derive the agreement pattern for person mismatch coordinations we need one impoverishment rule, shown in (6). e (6) [+hearer] → ∅ /[±speaker] Notice that (6) will not change anything for non-coordinated controllers, as [±speaker] is only present if person mismatch coordination is involved. The vocabulary items in (2), together with (6), will now derive the pattern observed in the data above. For (5a) only (2d) -n is compatible, impoverishment applies vacuously. Contexts (5b) and (5c) are compatible with both (2d) -n and (2e) -t, respectively, but only in the latter does impoverishment apply and leave (2d) -n as the only exponent compatible. Outlook A major part of the analysis is contingent on the type of the coordinator. That this rationale is on the right track is supported by diverging judgments on person mismatch coordination with another type of coordinator such as either or, see (7). (7) a. Entweder ich oder mein Freund *lauf-e/läuf-t/lauf-en either I or my friend walk-1sg/walk-3sg/walk-1pl/3pl nach Hause. to home ‘Either I or my friend walk home.’ 36 Speaker Abstracts b. Entweder mein Freund oder ich lauf-e/*läuf-t/lauf-en either my friend or I walk-1sg/walk-3sg/walk-1pl/3pl nach Hause. to home ‘Either my friend or I walk home.’ An exclusive disjunctive operator allows on the one hand for closest conjunct agreement and on the other hand for resolution and/or a default value. The pattern in (7) carries over to other languages like English (see Zwicky and Pullum 1986) and requires further investigation. Selected References Albright and Fuß (2012) Syncretism, in: The Morphology and Phonology of Exponence, OUP: Oxford, 236–288. Corbett (1983) Resolution rules: agreement in person, number, and gender, in: Order, Concord, and Constituency, Foris: Dordrecht, 175–206. Marušič et al. (2015) The Grammars of Conjunction Agreement in Slovenian, Syntax 18:39–77. Müller (2006) Pro-Drop and Impoverishment, in: Form, Structure, and Grammar, Berlin: Akademie Verlag, 93–115. Sauerland and Bobalijk (2013) Syncretism Distribution Modeling: Accidental Homophony as a Random Event, Proceedings of GLOW in Asia IX., 31–53. 37 Norwegian compounds: The category of non-heads Ragnhild Eik (Norwegian University of Science and Technology, NTNU) [email protected] Introduction In this paper, I investigate the linguistic nature of non-heads in Norwegian compounds. More specifically, I ask whether non-heads are formally categorized as belonging to a word-class. Josefsson (1997) argues for Swedish that non-heads of compounds do not have formal word-class. The paper will show that for Norwegian, a language very closely related to Swedish, non-heads do need a category. However, this finding proves hard to analyze within existing models of syntactic categorization (i.e. Distributed Morphology, exoskeletal syntax as in Borer 2013). Background A compound can be defined as the concatenation of two lexical elements, a head and a non-head, to form a single word. In the Germanic languages, the head of the compound is the right hand element. The issue addressed in this paper concerns the grammatical properties of the non-head of the compound, the left-hand element. According to Harley (2009), non-heads in English compounds are categorized for word-class. For Dutch, De Belder (2013) argues that only some non-heads are categorized. Others are simply bare roots without any formal category. For Swedish, Josefsson (1997) argues that non-heads are not categorized at all. Dutch and Swedish It is difficult to determine the nature of non-heads using traditional criteria, seeing as the compound as a whole usually only displays properties of the head. However, the category of non-heads seems to be closely related to linkers. A linker is a semantically empty element appearing between the head and non-head of a compound (Lieber & Stekauer 2009:13). In the Germanic languages, linkers are mainly derived from genitive and plural markers. In Modern Dutch and Swedish, linkers appear after certain nominal non-heads. (1) Dutch (De Belder 2013:3) varken.s.hok pig.link.pen ‘pig’s pen’ (2) Swedish (Josefsson 1997:61) stol.s.rygg chair.link.back ‘back of a chair’ 38 Speaker Abstracts The observation that linkers only appear after nominal non-heads suggests that linkers are sensitive to the category of the non-head. This is in part what leads De Belder to claim that Dutch compounds with linkers have categorized non-heads, whereas compounds without linkers have bare roots as non-heads. Josefsson, although noting that linkers are sensitive to noun-hood in Swedish, does not really account for the observation in her analysis of Swedish compounds. Norwegian Based on data from Dutch and Swedish and on previous analyses, we would expect to find the same pattern in Norwegian. But the expectation is only partly borne out. We do find linkers on nominal non-heads, as in dag.s.lys ‘day light’. However, in Norwegian, unlike Dutch and Swedish, a linker also appears after verbal non-heads: (3) Dutch slap.pil sleep.pill ‘sleeping pill’ (De Belder 2013:4) (4) Swedish skriv.maskin write.machine ‘type writer’ (Josefsson 1997:57) (5) Norwegian skriv.e.bok write.link.book ‘note book’ (Vinje 1973:63) (6) N. Leksvik dialect skriv.ar.bok write.link.book ‘write book’ (Johannessen 2001:678) The same is true for Icelandic where verbal non-heads get -i or -u (Harðarson 2016). To see that the ending on verbal non-heads is really a linker and not the infinitive, which is also realized as -e in some varieties of Norwegian, consider (6). In this dialect the linker is -ar and the infinitival ending is Ø. Following the reasoning whereby linkers are sensitive to word-class, the data above suggest that in Norwegian both verbal and nominal non-heads are formally categorized as verbs and nouns respectively. There are alternative ways of interpreting the Norwegian data. For example, one could argue that what seems like verbal non-heads in Norwegian are in fact nominalized forms. I will discuss this and other options in my talk, arguing that they are inadequate. Analysis If the generalization above is correct, namely that both nominal and verbal non-heads are categorized as such, how should this be analyzed? 39 Ragnhild Eik For reasons of space, I will focus the discussion on verbal non-heads, as these are the ones that pose problems for previous accounts. A first analytic option, taking a Distributed Morphology approach, is that a root is merged with a categorizing v-head, and with the linker merged on top of this, as in (7). The v-head is phonologically empty. A second option is that the linker is in fact the categorizer, a phonological realization of the v-head, as in (8). A third option, based on Borer (2013), is that categorization is contextual. According to this view, something about the structural context makes the non-head verbal. The structure in (9) illustrates this idea, without make explicit exactly what that context might be. A fourth option is that the non-head is inherently categorized as a verb, as in (10). (7) (8) v √ link -e v √ skriv v -e v (9) √ skriv (10) skrivv link -e skriv v Ø De Belder considers the structure in (7) for Dutch verb-like non-heads, only without the linker. She argues that the structure is ruled out based on the observation that verbal (and adjectival) non-heads, unlike nominal non-heads, are disallowed with phonologically realized categorizers. The argument goes as follows: if there is indeed a categorizing v-head, this head should be able to host phonologically realized categorizers as well, seeing as vocabulary insertion is post-syntactic. This is not possible. De Belder’s observation seems to be valid for Norwegian as well: (11) Dutch (De Belder 2013:12) *menstru.eer.pijn menstru.atev.pain (12) Norwegian (my example) *konstru.ere.arbeid construct.v.work For De Belder, this shows that verbal non-heads in Dutch are not categorized by a zero v-head, but rather that these non-heads are bare roots. Applying the same argument to Norwegian leads to the conclusion that verbal non-heads are not verbal by virtue of their merging with a zero v-head. Thus, the structure in (7) is ruled out. 40 Speaker Abstracts The second option, (8), also runs into problems. If linkers were categorizers, that would predict that more than one linker could appear at a time. Consider the compound in (13). (13) [ N [ N/LINK [ N [ V forsk ] ing ] s ] opphald ] ‘research stay’ This compound has both a nominalizer, -ing, and a linker, -s. Now, if the linker were a realization of n, then we would be left with no explanation to the fact that linkers are in complementary distribution. Once an n is allowed to merge with an n, we predict that multiple n linkers can be merged. The distribution of linkers can however be explained if we assume that the linker has its own projection. Moving to the option in (9), this structure states that some other part of the context determines the category of the non-head. The problem for this account is that if we consider the compounds in (14)–(16), the non-heads all appear to have identical contexts. (14) skriv.eV .bok write.link-bok (15) papirN .bok paper-book (16) rosaA .bok pink-book That would lead to the conclusion that they are all formally the same – either they have the same category or they have no category. Yet, as has already been established, verbal and nominal non-heads have different properties, presumably because they are of different categories. We are left with the option that non-heads in Norwegian compounds have their category inherently, as illustrated in (10). An inherent category is also what Borer (2013) suggests for adjectives, but not for verbs and nouns in English. Conclusion Non-heads in Norwegian compounds must be categorized as belonging to a word-class. The data favor an approach where elements appearing as non-heads are inherently categorized. Selected References Borer, H. (2013). Taking Form. Structuring sense, b. 3. Oxford: Oxford University Press. De Belder, M. (2013). The root and nothing but the root: primary compounds in Dutch. Ms. Ragnhild Eik 41 Harðarson, G. (2016). Peeling away the layers of the onion: on layers, inflection and domains in Icelandic compounds. The Journal of Comparative Germanic Linguistics, 19 (1): 1–47. Harley, H. (2009). Compounding in distributed morphology. I. Johannessen, J. B. (2001). Sammensatte ord. Norsk lingvistisk tidsskrift, 1: 59–92. Josefsson, G. (1997). On the principles of word formation in Swedish, 51. Lund: Lund University Press. Lieber, R. & Stekauer, P. (2009). The Oxford handbook of compounding. Oxford University Press. 42 Western Mari correlatives: Between two types of A′ -phenomena Anastasia Gareyshina (Independent Researcher) [email protected] Overview In this presentation, I aim at illustrating explicitly by means of my own fieldwork data that Western Mari correlatives (< Mari < Uralic) in (1) mirror conditionals in (2) and not relative clauses exemplified in (3), despite their traditional label as a non-local relativization strategy (cf. Srivastav 1991, Bhatt 2003, Nikolaeva 2005). I propose an analysis where Western Mari correlatives are structurally alike to hanging topics, with demonstrative phrases serving as resumptive pronouns, effectively explaining their similarity. Data A correlative is a construction that consists of a matrix clause containing one or more demonstrative constituents (Dem-XPs) that are ‘associated’ with one or more relative constituents (Rel-XPs) contained in a subordinate clause, which precedes the matrix clause. Devices of ‘association’ are co-indexation and matching requirement (further MR), whereby categorial features and case features match. Typically, Dem-XPs and Rel-XPs are involved in bijection: the number of Dem-XPs is equal to that of Rel-XPs and their denotations are (pairwise) the same. I concentrate on simple correlatives only due to multiple analyses theoretically possible for them, unlike for multiple correlatives (cf. Bhatt 2003). In (4), compared with (1), we observe that when embedded under a propositional attitude verb like “want”, the verb form in the former matrix clause shifts from indicative to optative with respect to mood, the verb in the former subordinate clause left unchanged. This suggests that correlatives are subordinate rather than coordinate structures. In (5) and (6), respectively, we provide a group of correlatives, showing that a Dem-XP and a Rel-XP must occupy left peripheral positions within their respective clauses and that a wh-word is inserted immediately to the right of the Dem-XP, the other options being strongly dispreferred; we note that correlatives in Western Mari are parallel to conditionals in these respects. Analysis There are two camps of analyses regarding the structure of correlatives, labeled below as Option 1 and Option 2. Option 1 suggests low (DP-)adjunction and subsequent movement to an IP adjunct position, as in (Bhatt 2003). A CorCP is intrinsically an adjunct within a Dem-XP, and thus intuition of similarity to restricted relative clauses is Anastasia Gareyshina 43 captured. But for multiple correlatives, a different structure with high (IP-level leastwise) base-generation has to be assumed, thus counterintuitively different structures for simple/multiple correlatives have to be posited, even though semantics is conceivably identical. Option 2 entails that correlatives are formed via high (IP-)adjunction, as in (Izvorski 1996). A CorCP is intrinsically adjoined to a clause at the TP/CP-level, therefore we retain the uniform structure for simple and multiple correlatives. The uniform semantics is another advantage of this analysis. Cross-linguistically, it has been shown that correlatives tend to fall into the two aforementioned types. Thus, the first question is whether a subordinate clause in Western Mari correlatives exhibits locality properties with respect to the main clause or not, i.e. do we choose Option 1 or Option 2? Or, alternatively: are Western Mari correlatives more like relative clauses (Option 1) or topics/conditionals/anything that is not a relative clause (Option 2)? The (non-)locality issue can be resolved considering the data from three syntactic tests: (1) the question and answer test; (2) (im)possibility of coordination of [CorCP + Dem-XP] phrases; (3) presence/absence of reconstruction effects. Argument 1: Question-Answer Test. According to the question-answer test, a CorCP and a Dem-XP do not form a constituent, since not only a Dem-XP can be freely omitted in fragment answers, but answers with a correlative followed by a Dem-XP are unanimously regarded by native speakers as ill-formed (examples in (7)). Argument 2: Coordination Test. The second argument against the [CorCP Dem-XP] phrase is evidence from coordination. In Western Mari, coordination of [CorCP Dem-XP] is ungrammatical, but two CorCPs alone can coordinate (examples in (8)). Argument 3: Presence/absence of reconstruction effects. Another evidence against a CorCP moving out of a Dem-XP is absence of reconstruction effects. We can test variable binding to prove this. Consider three following examples, the first modeled after ex. (43a) in (Bhatt 2003), all of them construed so as to have an underlying representation as in ex. (43a′ ) (ibid.) (our examples in (9) have been slightly modified along the lines of basic Western Mari syntax). Should we suppose a CorCP being base-generated as a Dem-XP adjunct (per Option 1), we would expect (9ib) to be severely ungrammatical. Yet this is only slightly worse than (9iib). In sum, both examples are acceptable, and our tentative explanation derives them, assuming two different structures for LFs of (9ib) and (9iib) respectively: 44 Speaker Abstracts A. [ ForceP [ QP QPi [ TopP [ CorCP Rel-XPj . . . Proni . . . tj . . . ] j [ MatrixCP [ TopP Dem-XPj [ QP QPi [ TP . . . ti . . . tj . . . ]]]]]]] B. [ ForceP [ TopP [ CorCP Rel-XPj . . . Proni . . . tj . . . ] j [ MatrixCP [ TopP Dem-XPj [ QP QPk [ TP . . . tk . . . tj . . . ]]]]]] In A, a QP first raises to obtain scope over matrix TP and then raises to a position higher than the ‘overall’ TopP to obtain scope over the whole construction and to bind a pronoun contained within a CorCP. In B, there is a single QR instance, which clearly entails the impossibility of co-reference of a QP and a Pronoun. The fact that variable binding in Western Mari correlatives is to some extent controversial for speakers themselves follows from the observation that the (covert) configuration derived in A is the one with Weak Crossover. In sum, our findings favored Option 2. Comparing simple correlatives to left dislocation and conditionals, we propose an even more fine-grained structure for them: [ForceP [TopP [CorCP [TopP Rel-XPi [TP . . . t′i . . . ]]]i [MatrixCP [TopP Dem-XPi [FocP (Wh-XPj ) [QP (QPk ) [TP . . . (tk ) . . .¨ (tj ) . . . ti . . . ]]]]]]] e Examples e e e (1) t¨d¨ kü k¨čäl-eš, so mo-eš. who.nom search-prs.3sg that/she/he.nom always find-prs.3sg ‘He who looks for something will always find it.’ (2) k¨čäl-eš t¨d¨ g¨n’, t¨näm so mo-eš. that/she/he.nom search-prs.3sg if then always find-prs.3sg ‘If he looks for something, he will always find it.’ (3) t¨d¨, kˆdˆ k¨čäl-eš, so mo-eš. that/she/he.nom rel.nom search-prs.3sg always find-prs.3sg ‘That person [that one] who looks for something will always find it.’ (4) m¨.län.¨m kel-eš, kü k¨čäl-eš, I.dat.poss.1sg want-prs.3sg who.nom search-prs.3sg mo-žˆ. t¨d¨ that/she/he.nom find-opt.3sg ‘I want that whoever looks for something always find it.’ e e e e e e e e e e e e e e e e 45 Anastasia Gareyshina e e e (5) kü p¨täri (?kü) tol-eš jažo (*kü), t¨d¨ who.nom first that/she/he.nom good come-prs.3sg vär-¨m (?t¨d¨) näl-eš (*t¨d¨). place-acc take-prs.3sg ‘Whoever comes first will get good places.’ (6) (*mam) kü p¨täri tol-eš, (?mam) t¨d¨ who.nom first come-prs.3sg that/she/he.nom ma-m näl-eš (*mam)? what-acc take-prs.3sg ‘Whoever comes first, what will he get?’ (7) a. e e e e e e e e e e e b. Q: kü zvon’-en? who.NOM call-pst2.3sg A1: t¨d¨, [RelCP kü gišän ten’gec¨ that/she/he.nom who.nom about yesterday jad-ˆn-at]. ask-pst2-2sg A2: [CorCP kü gišän ten’gec¨ jad-ˆn-at]. A3: *[ [CorCP kü gišän ten’gec¨ jad-ˆn-at], [Dem-XP t¨d¨] ]. ‘Who called? – The person whom you asked yesterday about.’ e e e e e e a. *[ [CorCP kü lekci-m lˆd-eš], [Dem-XP who.nom lecture-acc read-prs.3sg dä [ [CorCP kü čaj-ˆm jämd¨l-ä], t¨d¨-m]], that/she/he-acc and who.nom tea-acc prepare-prs.3sg [Dem-XP t¨d¨-m] ] väslimäš-¨šk¨ ik c¨š anžˆc that/she/he-acc meeting-ill one hour before pˆrt-at. let-prs.3pl b. [CorCP kü lekci-m lˆd-eš] dä [CorCP kü čaj-ˆm jämd¨l-ä], väslim䚨šk¨ ik c¨š anžˆc pˆrt-at. ‘Whoever delivers the lectures and prepares tea will be let in an hour before the meeting.’ e e e (8) e e c. d. e e e e e e e e e e e e e e e (?)OK [CorCP . . . Proni . . . ]j [MatrixCP QPi [ tj Dem-XPj ] . . . ] ma-m [ t¨d¨ ]i už-eš, t¨d¨-m [ what-acc that/she/he.nom see-prs.3sg that/she/he-acc e e e e a. b. e (9)i. 46 Speaker Abstracts e e e každˆj orod-ˆš ke-š¨ ]?i lačok-eš every stupid-ill go.away-prtcp.act.nom truth-el šotl-a. consider-prs.3sg OK [CorCP . . . Proni . . . ]j [MatrixCP QPk [ tj Dem-XPj ] . . . ] ma-m [ t¨d¨ ]i už-eš, t¨d¨-m [ každˆj orod-ˆš ke-š¨ ]k lačok-eš šotl-a. ‘[Every madman]?i /OK k takes for truth what hei sees.’ e e e e e e a. b. e ii. References Bhatt R. 2003. Locality in Correlatives. Natural Language and Linguistic Theory, 21(3), 485–541. Izvorski R. 1996. The Syntax and Semantics of Correlative Proforms. In: Kusumoto K. (ed.) Proceedings of NELS 26, GLSA Amherst, Massachusetts, 133–147. Nikolaeva I. 2005. Relative Clauses. In: Brown K. (ed.) Encyclopedia of Language and Linguistics (2nd ed). Amsterdam, Philadelphia: John Benjamins, 501– 508. Srivastav V. 1991. The syntax and semantics of correlatives. Natural Language and Linguistic Theory, 9(4), 637–686. 47 Conjunctive or disjunctive? On the syntax/semantics of -toka and -tari in Japanese Ryoichiro Kobayashi & Ryan Walter Smith (Sophia University/JSPS & University of Arizona) [email protected] & [email protected] Synopsis In this paper, we investigate the nature of the Japanese nonexhaustive particles -toka and -tari. At first glance, the distribution of these particles is very similar to that of the focus particles -mo ‘also’ and -sae ‘even’: they are both used as coordinators and stand-alone particles, are incompatible with topic, and induce intervention effects. However, while -sae and -mo carry presuppositions that project out of non-veridical contexts, -toka and -tari do not, and instead receive disjunctive interpretations in this context. We analyze -toka and -tari as items that introduce alternatives, which, once they expand into propositions, are either universally or existentially quantified depending on the veridicality or non-veridicality of their environment, respectively, and derive their coordinative use from their basic use as single particles. Data In Japanese, the particles -toka and -tari are used in veridical contexts to provide non-exhaustive conjunctions of nominal and verbal structures, respectively, as shown in (1). (1) a. b. Taro-toka Hanako-toka-ga kita T-toka H-toka-nom came ‘Taro, Hanako, and others came.’ Taro-wa heya-o soojisi-tari eigo-o benkyoosi-tari sita T-top room-acc clean-tari English-acc study-tari did ‘T. cleaned his room, studied English and did other things.’ At first glance, -toka and -tari seem to pattern very closely with the focus particles -mo and -sae. First, -toka and -tari can stand on their own, acting much like focus particles themselves. (2) a. John-wa Nihongo-toka-o benkyoosita J-top Japanese-toka-acc studied ‘John studied Japanese among other things.’ 48 Speaker Abstracts b. Taro-wa Eigo-o benkyoosi-tari suru T-top English-acc study-tari do ‘Taro studies English among other things.’ Additionally, -mo and -sae can themselves act as coordinators, as demonstrated in (3). (3) a. b. Taro-mo Hanako-mo paatii-ni kita T-mo H-mo party-to came ‘Taro and Hanako also came to the party.’ Kare-wa nusumi-mo/sae, korosi-mo/sae suru he-top rob-mo/sae murder-mo/sae do ‘He also/even robs and murders.’ Moreover, all of these items are unacceptable with topical -wa in (4). Likewise, the nominal particles induce focus(/LF) intervention effects (Hoji 1986) in (5). Note that they become grammatical when the wh is overtly scrambled over the intervener. (4) a. *Taro-mo/sae-wa kita T-mo/sae-top came ‘As for also/even Taro, they came.’ b. *Taro-toka-wa kita (ok contrastive/*topic) ‘As for also Taro, came.’ c. *Soojisi-tari-wa Taro-ga sita (ok contrastive/*topic) clean-tari-top T-nom did ‘As for also cleaning, Taro did.’ (5) a. *?Hanako-mo/sae dare-o hometa no? H-mo/sae who-acc praised Q ‘Who did also/even Hanako praise?’ b. Darei -o Hanako-mo/sae t i home-ta no? c. *?Taro-toka-ga nani-o tabeta no? T-toka-nom who-acc ate Q ‘What did also Taro eat?’ d. Nanii -o Taro-toka-ga t i tabe-ta no? However, -toka and -tari differ from -mo and -sae in one crucial respect: although -mo and -sae are conjunctive regardless of their environment and Ryoichiro Kobayashi & Ryan Walter Smith 49 carry additive presuppositions that project out of non-veridical contexts, such as the conditionals in (6a–b), -toka and -tari lack such presuppositions, and instead gain a disjunctive-like interpretation in such contexts (6c–d). For instance, -toka and -tari (6c–d) do not entail that Taro himself comes to the party along with someone else, or that Taro actually eats broccoli, unlike -mo ‘also’ and /-sae ‘even’ (6a–b), which do. (6) a. Taro-mo kita-ra, Ryo-wa ocha-o dasu. T-mo come-if R-TOP tea-ACC serve ‘If Taro also comes to the party, Ryo serves tea.’ b. Taro-ga burokkori-o tabe-sae sur-eba, mama-wa yorokobu. T-NOM broccoli-ACC eat-sae do-if mom-TOP be.happy ‘If Taro even eats broccoli then his mom will become happy.’ c. Taro-toka (Hanako-toka)-ga kita-ra, Ryo-wa ocha-o dasu. ‘If Taro (or Hanako or someone else) comes to the party, Ryo serves tea.’ d. Taro-ga burokkori-o tabe-tari gyuunyuu-o non-dari su-reba, mama-wa yorokobu. ‘If Taro eats broccoli (or drinks milk or does something else) his mom becomes happy.’ To summarize, -toka and -tari exhibit many parallels with focus particles syntactically, but they differ from them in their semantic properties. Analysis Syntactically, we follow the spirit of previous analyses of -mo, such as Mitrović & Sauerland (2014), and claim that -toka and -tari are actually not the coordinator head, but are focus particles that attach to each conjunct coordinated by a silent coordinator J, as in (7a–b). Since the appearance of the second -toka is optional and does not affect semantics, we assume that it is syntactically always there, but optionally has phonetic content in (7a). (7) a. b. [ JP [ tokaP NP-toka ] [ J′ J [ tokaP NP-(toka) ]]] [ JP [ tariP VP-tari ] [ J′ J [ tariP VP-tari ]]] On its own, -toka selects for an NP complement. As for -tari, given parallels between it and the -mo/-sae, we propose that it selects a projection below TP. Semantically, we propose that sentences with -toka and -tari simply denote a set of individual and predicate alternatives, respectively, as in (8a–b), with no additive presuppositions like those that come with -mo and -sae. 50 (8) Speaker Abstracts a. b. J Taro-toka K = {Taro, Ryoichiro, Ziro, . . . } J heya-o soojisi-tari K = {λx.λw.x clean the room, λx.λw.x study English, λx.λw.x eat dinner, . . . } In the case of coordination with -toka/tari, we depart from Mitrović & Sauerland’s treatment of J by analyzing it as simply collecting alternatives introduced by each conjunct in exactly the same way that or does in the analysis of AlonsoOvalle (2006, 2008). This allows the alternatives to be composed with other elements of the sentence in the same way regardless of whether coordination is present or not. (9) Where JXPK and JYPK ⊆ Dτ , J [[ XP ] [ J [ YP ]]] K ⊆ Dτ = JXPK ∪ JYPK The alternatives are composed with other elements of the sentence via Pointwise Functional Application (Hamblin 1973), ultimately yielding a set of propositional alternatives. (10) J Taro-toka-ga kita K = {λw.Taro came, λw.Ryoichiro came, λw.Ziro came, . . . } (11) J Taro wa heya-o soojisi-tari sita K = {λw.Taro cleaned the room, λw.Taro studied English, λw.Taro ate dinner, . . . } Once the alternatives become propositional, they can be manipulated by one of two propositional quantifiers, defined below (Kratzer & Shimoyama 2002; Alonso-Ovalle 2006, 2008). (12) a. b. J∃Kw (A) = {λw′ .∃p ∈ A & p(w′ )} J∀Kw (A) = {λw′ .∀p ∈ A → p(w′ )} In non-veridical contexts, the set of alternatives is existentially quantified as in (12a), which gives rise to the interpretation that at least one of the propositions in the alternative set is true, but not necessarily the one overtly mentioned, the interpretation required for (6c–d). In veridical contexts, the alternatives are instead universally quantified as in (12b), which makes all of the propositions in the alternative set true, and thus gives rise to the conjunctive interpretation observed in (1a–b) and (2a–b). Moreover, because the J head merely denotes the union of the alternatives generated by each conjunct, the analysis requires no extensions to derive the correct interpretation of the cases involving coordinate structures. Ryoichiro Kobayashi & Ryan Walter Smith 51 Conclusion In this paper, we have shown that -toka and -tari pattern much like the focus particles -mo and -sae in terms of their syntactic distribution: they can be used as stand-alone particles and as polysyndetic coordinators, are incompatible with topical -wa, and induce intervention effects. However, they differ from other focus particles in lacking additive presuppositions and having interpretations sensitive to the (non-)veridicality of their environment. This paper proposes an analysis of these particles as introducing alternatives, which are then manipulated by propositional quantifiers higher in the structure, and unifies their use as single particles and as coordinators. References Alonso-Ovalle, L. (2006). Disjunction in alternative semantics. Doctoral dissertation, University of Massachusetts Amherst. Alonso-Ovalle, L. (2008). Innocent exclusion in an alternative semantics. Natural Language Semantics, 16(2), 115–128. Hamblin, C. L. (1973). Questions in montague English. Foundations of language, 10(1), 41–53. Hoji, H. (1986). Scope interpretation in Japanese and its theoretical implications. In Proceedings of the West Coast Conference on Formal Linguistics, Vol. 5, pp. 87–101. Kratzer, A. & Shimoyama, J. (2002, March). Indeterminate pronouns: The view from Japanese. In 3rd Tokyo conference on psycholinguistics. Mitrović, M. & U. Sauerland. (2014). Decomposing coordination. NELS 44, 39–52. 52 A configurational account of Turkish DSM Sabine Laszakovits (University of Connecticut) [email protected] Synopsis This paper presents a purely configurational approach to differential subject marking (DSM) in Turkish and potentially other Turkic languages. Introduction DSM is a well-studied phenomenon in the literature on Turkic whereby some subjects of nominalized embedded clauses are marked with nominative and others with genitive. There are three factors that seem to be relevant to this case-marking: 1. specificity of the subject NP, 2. location of the subject NP, 3. type of embedded clause. Following Diesing (1992) a.o., 1. and 2. can easily be connected: specific NPs have the tendency to appear in higher structural positions than non-specific ones, and it is high subjects that are marked as genitive. Regarding 3., it seems to be a robust cross-Turkic generalization that argument and relative clauses allow this specificity-driven DSM, but adjunct clauses are restricted to nominative-marking, at least with ‘factive’ nominalizers. This has been used to argue for hybrid case-assignment approaches (Baker & Vinokurova 2010 for Sakha, Gribanova 2016 for Uzbek); i.e. case assignment varies within the same language between licensing by a functional head (Chomsky 2000, 2001) for the DSM-facts and a configurational system (Marantz 1991) for all other case assignments. Claim This paper aims to show that a configurational case theory can also account for the DSM patterns (cp. Levin & Preminger 2015 for Sakha), thereby making hybrid accounts unnecessary. Following Aygen (2007), I assume that all argument clauses are headed by a noun that usually remains covert. It creates a case-assignment domain that will trigger genitive instead of nominative as unmarked case on the topicalized subject in its complement clause. All argument clauses are therefore complex NPs. In contrast, adjunct clauses are not headed by a covert noun and their case-assignment domains are not nominal. I propose that the availability of DSM is not due to features on some functional head, but to the presence or absence of an external noun. Data Regular argument clauses in Turkish contain a nominalized verb that retains all its matrix clause properties except for finiteness and that the subject may be differentially marked as genitive, (1). Sabine Laszakovits (1) 53 Ben [ Hasan-ın bu kitab-ı oku-duğ-un ]-u biliyorum. 1sg [ Hasan-gen this book-acc read-nmlz-3sg ]-acc I.know ‘I know that Hasan read this book.’ As has been noted by Aygen (2007), this construction can be transformed into a complex NP by inserting a head noun such as ‘fact’, (2). This head noun is marked by 3SG agreement similar to the head noun in compounds. (2) Ben [ [ Hasan-ın bu kitab-ı oku-duğ-u ] gerçeğ-in ]-i I [ [ Hasan-gen this book-acc read-nmlz-3sg ] fact-3sg ]-acc biliyorum. I.know ‘I know the fact that Hasan read this book.’ An adjunct clause is given in (3). The embedded subject must bear nominative (morphologically unmarked) and cannot bear genitive. (3) [ Hasan-(*ın) söylenti-yi duy-duğ-un-a göre ] herkes [ Hasan-(*gen) rumor-acc hear-ptpl-3sg-dat because ] everybody duyacak. she.will.hear ‘Since Hasan heard the rumor, everybody will hear it.’ The complementizer heading the adjunct clause in (3) is göre ‘because’, which appears also as a postposition in the meaning ‘according to’. As such, it may take a headless relative clause as in (4) (Aygen 2007). We can independently test this construction for its gap, and we observe that the head noun does not contain a compound-marking, making it different from the complex NPs that we postulated for argument clauses. (4) [ [ Hasan-ın duy-duğ-u ] (söylenti) ]-ye/na göre [ [ Hasan-gen hear-ptpl-3sg ] (rumor) ]-dat according herkes sevinecek. everybody she.will.be.happy ‘According to {the rumor/what} Hasan heard, everybody will be happy.’ When we insert an overt head noun into the adjunct clause in (3), we can observe that it stops patterning with adjunct clauses and starts patterning with 54 Speaker Abstracts argument clauses: 1. the subject must receive genitive case, and 2. göre in (5) has the postposition meaning, not the complementizer meaning. (5) [ [ Hasan-ın söylenti-yi duy-duğ-u ] gerçeğ-in ]-e göre [ [ Hasan-gen rumor-acc hear-ptpl-3sg ] fact-3sg ]-dat according ‘according to the fact that Hasan heard the rumor’ I suggest to deduce from these data that the adjunct clause in (3) does not contain a covert head noun. Analysis I propose that the head noun that can be made visible in argument clauses, is always present even though it usually cannot be seen. The structure for an argument clause is given in (6a) and for an adjunct clause in (6b). Both are parallel to Aygen (2007: ex. 26). (6) a. b. . . . [ [ subject-gen . . . verb-nmlz-agr ] {noun-agr / Ø} ]-kase . . . . . . [ subject-nom . . . verb-nmlz-agr-kase comp ] . . . Case assignment is configurational (Marantz 1991, a.o.), and genitive case is the unmarked case in the nominal domain. The outer NP creates the nominal domain for genitive to be licensed in the left periphery of the embedded clause on whatever NP does not yet have a case assigned. Non-topic positions are inside the complement of C and therefore not in a nominal domain and will receive nominative as unmarked case. (7a) shows the full structure for an argument clause and (7b) for an adjunct clause. (7) a. b. . . . [ NP [ CP topic-gen [ TP [ [ vP subject-nom [ VP V ] v ] Asp ] T-agr ] C ] N-agr ] . . . . . . [ CP topic-nom [ TP [ [ vP subject-nom [ VP V ] v ] Asp ] T-agr ] C] Conclusion This paper argues for a purely configurational approach to DSM in Turkish and similar Turkic languages, such as Uzbek (against Gribanova 2016). It proposes to account for the availability of genitive case on the subject of embedded clauses by adding a nominal case-assignment domain above the CP that is projected by a usually covert head noun. Whenever genitive case is not available, we can show that there is no covert head noun present. The structure of an adjunct clause is therefore a proper subset of the structure of an argument clause, contra Colley & Davis (2016). Sabine Laszakovits 55 Selected References Aygen, Gulsat (2007): Syntax and Semantics of Genitive Subject-Case in Turkic. California Linguistic Notes 32 (Spring 2007): 1–39. Baker, Mark & Nadya Vinokurova (2010): Two Modalities of Case Assignment: Case in Sakha. NLLT 28: 593–642 Colley, Justin & Colin Davis (2016): A new approach to Turkish nominalized clauses, WAFL 12. Gribanova, Vera (2016): Case, agreement, and differential subject marking in Uzbek nominalized clauses. Ms., Stanford. Levin, Ted & Omer Preminger (2015): Case in Sakha: are two modalities really necessary? NLLT 33(1): 231–250. 56 Plurality and specificity in Spanish interrogatives Mora Maldonado (Institut Jean Nicod, École Normale Supérieure, Paris) [email protected] 1 Background The meaning of interrogative sentences is contained in their answerhood conditions (Kartunnen, 1977). Singular wh-questions in English differ from plural and neutral interrogatives in that they trigger an uniqueness effect: they require a single true answer which names a singularity, and lead to a presupposition failure otherwise. Most accounts explain these differences by assuming that (a) only singular marking has semantic import (i.e. weak account of plurality, Link 1983, Sauerland 2003, 2005, Spector 2007); and (b) interrogatives admit a maximally informative answer (Dayal 1996, Heim 2004, Fox 2013, a.o.). By (a), a question such as ‘Which students called?’ will denote a set of singular and plural propositions, while ‘Which student called?’ will only contain singularities in its denotation. By (b), a complete answer to each of these questions should specify all the students who called. This maximality requirement has been captured by proposing an answerhood operator, which presupposes the existence of a maximal true proposition in the question denotation, and it returns this proposition as an answer in w (e.g. Dayal’s ANS in (1)). (1) JANSK(Q)(w)= ιp.p∈Q ∧ p(w) ∧ ∀p′ ∈Q: p’(w)→ p’⊂p Since a matrix question Q can be uttered in w iff ANS(Q) is defined in w, the uniqueness effect is explained. The fact that plural questions typically get plural answers is accounted for as the result of a pragmatic competition between singular and plural alternatives (Dayal 1996). 2 Puzzle Spanish distinguishes morphologically between singular and plurals forms of interrogative pronouns or quantifiers (e.g. (2)). While quiénesinterrogatives typically require a plurality named in the answer, singular quién-questions (e.g. (2a)) can give rise to both singular and plural answers (i.e. no uniqueness effect), suggesting that the wh-element ‘quién’ is not semantically singular and it denotes both atomic and plural individuals (cf. English ‘who’). Once we assume an underspecified meaning for ‘quién’, keeping a weak account of plurality would make the system redundant: ‘quién’ and ‘quiénes’ would have exactly the same denotation and they should be interchangeable. Mora Maldonado 57 (2) w1 : Only John and Mary went to the party; w2 : Only John went to the party a. Quién fue a la fiesta? ANS(Q)(w1 )=John and Mary; ANS(Q)(w2 )=John. Whosg went to the party? b. Quiénes fue a la fiesta? ANS(Q)(w1 )=John and Mary; ANS(Q)(w2 )=# Whopl went to the party? In this paper, I will describe the distribution of quién and quiénes interrogatives in Spanish, and claim that ‘quién’ and ‘quiénes’ can be only understood under a strong account of plurality, such that ‘quiénes’ is semantically plural. 3 Distribution of quién and quiénes interrogatives The availability of quién and quiénes interrogatives is restricted by the utterance context: the possibility of using each of these questions depends on the information that speaker and hearer share about the expected answer (i.e. common ground). Quiénes interrogatives can be felicitously uttered iff it is common knowledge that a plurality will be named in the exhaustive answer (e.g. (3b)). Conversely, whenever (i) the speaker is ignorant about the exact cardinality of the answer (e.g. ‘at least one’ scenarios in (4)) or (ii) the speaker targets a singular maximal answer (e.g. ‘exactly one’ scenarios in (3a)), quiénes-interrogatives will lead to a presupposition failure, and only the alternative with ‘quién’ would be available. (3) a. b. (4) a. b. Una de mis amigas fue a la fiesta pero no me acuerdo quién (# quiénes). ‘One of my friends went to the party but I don’t remember whoSG (# whoPL )’. Varias amigas fueron a la fiesta pero no me acuerdo quiénes (?? quién). ‘Several friends went to the party but I don’t remember whoPL (?? whoSG )’. Quién fue a la fiesta? Alguien debe haber ido. WhoSG went to the party? Someone must have been there. # Quiénes fueron a la fiesta? Alguien debe haber ido. WhoPL went to the party? Someone must have been there. 3.1 Cardinality requirement: ‘cuáles NP’ vs ‘quiénes’ Quiénes-interrogatives only have plural propositions in their denotation and, therefore, they give rise to a plurality effect (mirror image of the uniqueness effect in English questions). This contrasts with the denotation of plural cuáles-interrogatives (cf. which- 58 Speaker Abstracts interrogatives). For instance, in ((5)), the interrogative with ‘quiénes’, but not the one with ‘cuáles’, lacks a cumulative reading where for each day there is a single friend that John invited. (5) Each day Juan invited at least one friend. a. Juan sabe a cuáles/qué amigos invitó cada día de la semana. ‘Juan knows which friends did he invite each day of the week.’ b. #Juan sabe a quiénes invitó cada día de la semana. ‘Juan knows whopl did he invite each day of the week.’ 3.2 Specificity or D-linking constraint In certain cases, speaker’s knowledge about answer cardinality is not enough to license quiénes-interrogatives (e.g. (6)). On top of the plurality requirement, these questions seem to require specific contexts, where the speaker can identify the elements in the domain, and ask for a choice among them. This notion corresponds to the D-linked property (Pesetsky 1987, 2000) attributed to which-phrases in English. (6) Mary and John arrive at their apartment, where there is supposed to be no one. They hear two people whispering inside. Mary says to John: Quién está ahí? / # Quiénes están ahí? ‘Who is in there?’ 4 Account The use conditions of quién and quiénes interrogatives can be summarized as follows: A quiénes-interrogative Q can be uttered in w iff (i) ANS(Q) is defined in w; and (ii) for all x ∈ D e , JxKw is known. Otherwise, the alternative with ‘quién’ should be used. This distribution can be accounted by assuming that ‘quiénes’ contains in its lexical entry both a plurality and a D-linked requirement, whereas ‘quién’ is underspecified. Let us assume a known domain D e , such that D e = {m, b, j, m ⊕ b, m ⊕ j, b ⊕ j, m ⊕ j ⊕ b}. The denotations for ((2a)) and ((2b)) are given in ((7)). (7) a. J(2a)Kw = λp.∃x ∈ D e . x is human & p = λw *went-to-the-party(w)(x) = {m went, j went, b went, m ⊕ j went, m ⊕ b went, j ⊕ b went, m ⊕ j ⊕ b went} w b. J(2b)K = λp.∃x.∣x∣ > 1& x are human & p = λw went-to-theparty(w)(x) = {m ⊕ j went, m ⊕ b went, j ⊕ b went, m ⊕ j ⊕ b went} Mora Maldonado 59 Employing the ANS operator will yield to two LFs with different presuppositional strength: the worlds where ANS(J(2b)K) is defined are included in the worlds ANS(J(2a)K) is defined. In every world where the maximal true answer is predicated from a singularity, ANS(J(2a)K) will correctly yield to a presupposition failure. Conversely, in any scenario where more than one person went to the party and the speaker knows it, ANS(J(2a)K) and ANS(J(2b)K) will be strawson equivalent. A pragmatic principle such as Maximize Presupposition! (Heim 1991, Schlenker 2012) would then select the latter, since this is the LF carrying a stronger presupposition. When a quién-interrogative is uttered, one should infer that it’s not presupposed that the question has a plural answer, or the alternative with ‘quiénes’ should have been used. Questions such as ((2a)) are therefore preferred in ‘at least one’ type of situations, triggering an ignorance inference regarding the cardinality of the answer. However, if the speaker is assumed to be opinionated, the alternative with ‘quiénes’ can be safely negated, deriving an exhaustive, ‘exactly-one’, implicature. The concrete implementation of this derivation depends, crucially, on the particular account of scalar implicatures that we choose. Selected References Dayal, V. 1996. Locality in wh quantification: questions and relative clauses in Hindi. Kluwer Academic Publishers. Fox, D. 2013. Mention-some readings of questions. Class notes, MIT Seminars. Heim, I. 1991. Articles and definiteness, In: Semantics. An international handbook of contemporary research. Berlin: De Gruyter. Karttunen, L. 1977. Syntax and semantics of questions. Linguistics and philosophy, 1, 3–44. Pesetsky, D. 1987. Wh-in-situ: Movement and unselective binding. The representation of (in) definiteness, 98, 98–129. Pesetsky, D. 2000. Phrasal movement and its kin. MIT press. Sauerland, U. 2003. A new semantics for number. Semantics and Linguistic Theory, 13, 258–275. Sauerland, U., Anderssen, J. and Yatsushiro, K. 2005. The plural is semantically unmarked. Linguistic Evidence, 4, 409–430. Schlenker, P. 2012. Maximize presupposition and Gricean reasoning. Natural Language Semantics, 20, 391–429. Spector, B. 2007. Aspects of the pragmatics of plural morphology: On higherorder implicatures, In: Presupposition and implicature in compositional semantics. Springer, 243–281. 60 Cognate adverb construction in Moksha Mordvin Sofia Nikiforova (NRU Higher School of Economics, Moscow) [email protected] This study focuses on semantics of the cognate adverb construction in Moksha, a Finno-Ugric language (Uralic language family) spoken in Mordovia, Russia. This phenomenon has never been looked at closely before, nor has it been compared to similar constructions in other languages. In this study the Moksha data is viewed in the typological perspective; we provide a thorough description of the construction’s semantics and propose an analysis which can be applied to other languages as well. The cognate adverb construction (CAC) is formed by a conjugated form of a verb and a deverbal adverb derived from the same stem. Such constructions, unlike usual ‘verb + adverb/gerund’ combinations that have more or less compositional semantics, possess their own meaning, see (1). c’ora-s’ ud- z’ ud-i boy-def.sg sleep-conv.atd sleep-npst.3sg (lit.: ‘The boy sleeps (how?) sleeping.’) 1. ‘The boy sleeps soundly.’ 2. ‘The boy is sleeping (and not just lying with his eyes shut).’ (1) e Meanings of the Moksha cognate adverb construction fall into two categories: intensification and ‘accuracy of description’. Intensification Different groups of verbs get intensified in different ways. Many verbs get intensified with respect to the most prominent parameter in their semantic class. For example, for most of the motion verbs it is the speed parameter: las’k z’ las’k ms ‘to run very quickly; lit.: to run running’; šačt z’ šačt ms ‘to crawl very slowly; lit.: to crawl crawling’. For most of the sound verbs it is the loudness parameter: ivad’ z’ ivad’ ms ‘to shout very loudly; lit.: to shout shouting’; toškaz’ toškams ‘to whisper very softly; lit.: to whisper whispering’ In the case of verbs with an incremental theme, the CAC indicates that the whole theme is affected by the event. e e e e e e Sofia Nikiforova kuc pal- z’ pal-i house.def.sg burn-conv.atd burn-prs.3sg ‘The whole house is burning.’ e (2) 61 Not all verbs get intensified when they appear in a CAC; however, given a context of implicit or explicit contrast, all Moksha verbs in this construction can receive the next meaning we are going to discuss– the ‘accuracy of description’. Accuracy of description The use of the construction can also indicate the speaker’s certainty that the chosen verb is the most suitable one for this particular situation. s’orma-t’ son s’ormad- z’ s’ormad- z’ he write-conv.atd write-pst.3sg.s.3sg.o letter-def.gen ‘He wrote the letter (he didn’t type it).’ e e e (3) This meaning of the CAC is often found in metaphors and exaggerations. e e l’ij- z’ l’ij-s’ od ping -z’ fly-conv.atd fly-pst.3sg young time-1sg.poss.sg ‘My young years have really flown by (not just gone by).’ e (4) Constructions with similar form and meaning can be found in other languages as well (Uralic, Semitic, Slavic, Nilo-Saharan). The examples below provide a mere illustration: • Almost identical (both in form and meaning) CACs can be found in languages closely related to Moksha Mordvin: Erzya Mordvin, Shoksha Mordvin, Hill Mari, etc. • In Hungarian there is a syntactic reduplication construction formed by a deverbal adverb derived from the stem by the suffix -va/-ve, followed by the conjugated form of the same verb: (5) Hát kér-ve kér-ünk titek-et well ask-adv ask-1pl 2pl-acc ‘We beg you very much.’ (Brdar et al. 2014:5) It is used to express intensification. • So called ‘tautological infinitive’ constructions are found in many, if not all, Semitic languages (Biblical Hebrew, Old Aramaic, Syriac, Ugaritic, 62 Speaker Abstracts Akkadian, Classical Arabic, Maltese and others). These constructions are formed by a conjugated verb and a special verbal form usually called ‘the infinitive absolute’. (6) Biblical Hebrew sakol visakel stone:inf.abs stone:npst.3m.sg ‘He will be surely stoned.’ It is also used to show the completeness of an occurrence (as in horeš lo horišo ‘did not utterly drive them out’) or to intensify the idea of the verb in some other way (as in šim’u šamoa ‘listen attentively’). The hypothesis Our hypothesis is that the two meanings of the Moksha cognate adverb construction are in fact two manifestations of a single meaning. This meaning is defined as follows: the construction narrows down the verb denotation leaving only its nuclear part, its most prototypical ‘representatives’. In the semantics of some verbs there is a parameter that is crucial for distinguishing between prototypical and non-prototypical situations denoted by the verb. It is, for example, the speed parameter for the verb ‘to run’: fast running is prototypical and slow running is not. When a speaker uses a verb like ‘to run’ in a CAC, only prototypical (i.e. fast) running is meant. And so it brings about the intensification effect. If there is no such parameter in the verb’s semantics, there is no intensification effect. But the use of a CAC still means the same thing – that the situation the speaker is describing is prototypical for the verb in the construction. This creates an implicature: ‘if the situation is prototypical for the verb V, there can be no mistake in choosing V for its description’ and thus leads to the ‘accuracy of description’ meaning. Data from different Uralic, Semitic and other languages mentioned above can be analyzed in a similar way. The proposed analysis puts Moksha CACs next to other cases of lexical and syntactic reduplication that are sometimes described within the prototype theory. 63 The left periphery fragmented: Evidence from Italian Roberto Petrosino (University of Connecticut) [email protected] Introduction The purpose of this paper is to provide new evidence shedding light on the status of cartographic projections. The cartographic approach assumes that the underlying syntactic structure of sentences is more complex than the usual functional projections – i.e., CP, IP/TP, vP, VP, DP/NP. Thus, regarding traditional CP, Rizzi (1997) argues that the traditional CP is actually made up of several projections, where scope-discourse features (such as focus and topic) are licensed. One of the issues that is currently debated regarding Rizzi’s (1997) is whether the full structure in (1) is always projected (i.e. whether CP is always fully split). (1) [[ ForceP Fore [ TopP Top [ FocP Foc [ TopP Top [ FinP Fin ] IP ]]]]] By analyzing data regarding anaphor binding across clauses in Italian, the present contribution bears on two issues. First, I show that an anaphor in an embedded clause can be bound from a higher clause only if it is at the outmost edge of the embedded clause, in compliance with the phasal approach to Condition A. Second, based on such anaphor-binding data, I show that the full CP cartography from (1) is not always projected, thus providing a new perspective on the issue. Discussion In Italian, anaphor binding across clauses is generally disallowed, as illustrated by (2) (all the unacceptable examples in the abstract are fine if they are modified so that they do not contain an anaphor). (2) *Giannii si chiede [ se Maria ha comprato [ il Gianni refl ask.3sg whether Maria aux.3sg buy.ppt.msg [ the ritratto di [ se stesso ]i ]]. picture of [ refl same.msg ] ‘Johni wonders whether Mary has bought the picture of himselfi .’ However, such examples improve if the anaphor-containing DP is fronted. Crucially, the anaphor must be fronted to the embedded clause initial position, as the examples in (3–4) show, where the anaphor-containing DP is fronted to the embedded clause initial position in (3), but not in (4). 64 Speaker Abstracts (3) ?Giannii si chiede [ [ quale ritratto di [ se stesso ]i ] Gianni refl ask.3sg which picture of refl same.msg Maria Maria ha comprato ]. aux.3sg buy.ppt.msg ‘Gianni wonders which picture of himself Mary bought.’ (4) *Giannii si chiede, [ Maria, [ quale ritratto di [ se stesso ]i ], Gianni refl ask.3sg Maria which picture of refl same.msg ha comprato ]. aux.3sg buy.ppt.msg ‘Gianni wonders which picture of himself Mary bought.’ The effect is confirmed by anaphor-containing topics, as in the examples in (5–6), both of which contain a topic and a wh-phrase in the embedded clause. Although both orders of these two elements are in principle possible, when the topic contains an anaphor, the topic must precede the wh-phrase. (5) Giannii si chiede, [ [ [ il ritratto di [ se stesso ]i ]j , chi loj Gianni refl ask.3sg the picture of refl same.msg who cl ha comprato ]. aux.3sg buy.ppt.msg ‘John wonders, the picture of himself, who bought.’ (6) *Giannii si chiede, [ chi, [ [ il ritratto di [ se stesso ]i ]j , loj Gianni refl ask.3sg who the picture of refl same.msg cl ha comprato ]. aux.3sg buy.ppt.msg ‘John wonders, the picture of himself, who bought.’ In the talk I will discuss a number of other constructions that involve interaction between different types of fronting and anaphor binding which all confirm the pattern that was exhibited by the above constructions: cross-clausal binding of an anaphor is possible only if the DP that contains the anaphor is clauseinitial. I argue that this pattern, i.e. the data given in examples (3–6), provides support for the phase-based conception of Condition A (see Bošković 2016b, Canac-Marquis 2005, Despić 2013, Hicks 2009, Lee-Schoenfeld 2008, Safir 2014, among others), where an anaphor may be bound outside of its clausal phase only if it is located at the edge of that phase, under the assumption that the highest clausal projection is a phase (see Bošković 2014, 2015; Wurmbrand Roberto Petrosino 65 2014, for a number of arguments to this effect). The anaphor-containing DP is at the edge of the embedded clause phase in (3), but not in (4). It is also at the embedded clause phase edge in (5), but not in (6). Crucially, this is the case only if the full CP cartographic structure is not always projected: if it were, all the cases where the anaphor se stesso is not in [Spec, ForceP] should be unacceptable. Being the highest clausal projection in (1), ForceP would always be a phase if present: the anaphor-containing DP, which is located in [Spec, FocP] in (3) and in [Spec, TopP] in (5), then would not be located at the phasal edge, hence these examples should also be unacceptable. The data presented above then indicate that the finely-articulated sequence of functional projections in the traditional CP field is not always projected – in fact, only the projections with overt morphological manifestation are projected in the examples discussed above (for recent arguments to this effect from very different considerations, see Bošković 2016a, Erlewine 2016). Conclusion This talk provides data regarding anaphor binding across clauses in Italian which show that an anaphor may be bound cross-clausally only when it is located at the phasal edge of the clause. It is shown that the data can be captured under the phase-based approach to Condition A. The data also provide evidence that the full left periphery may not be always projected. The talk will also discuss more complicated constructions where the relevant phase has more than one edge, which will be used to test Bošković’s (2016b) claim that in the case of a phase which has multiple edges, only the outmost edge counts as the phasal edge for the purpose of the Phase-Impenetrability Condition. References Bošković, Željko (2014). Now I’m a phase, now I’m not a phase: On the variability of phases with extraction and ellipsis. Linguistic Inquiry 45, 27–89. Bošković, Željko (2015). ‘From the Complex NP Constraint to everything: On deep extractions across categories’, The Linguistic Review 32, 603–669. Bošković, Željko (2016a). On the timing of labeling: Deducing Comp-trace effects, the Subject Condition, the Adjunct Condition, and tucking in from labeling. The Linguistic Review 33, 17–66. Bošković, Željko (2016b). Getting really edgy: On the edge of the edge. Linguistic Inquiry 47(1), 1–33. Canac-Marquis, Réjean (2005). Phases and binding of reflexives and pronouns 66 Speaker Abstracts in English. In Proceedings of the 12th international conference on head-driven phrase structure grammar, 482–502. CLSI Publications Stanford, CA. Despic, Miloje (2013). Binding and the structure of NP in Serbo-Croatian. Linguistic Inquiry 44, 239–270. Erlewine, Michael Yoshitaka (2016). ‘Anti-locality and optimality in Kaqchikel Agent Focus.’ Natural Language & Linguistic Theory 34(2), 429–479. Hicks, Glyn (2009). The derivation of anaphoric relations, vol. 139. John Benjamins Publishing. Lee-Schoenfeld, Vera (2008). Binding, phases, and locality. Syntax 11, 281–298. Rizzi, Luigi (1997). The fine structure of the left periphery. In: Elements of grammar, pp. 281–337. Safir, Ken (2014). One true anaphor. Linguistic Inquiry 45, 91–124. Wurmbrand, Susi (2014). Tense and aspect in English infinitives. Linguistic Inquiry 45, 403–447. 67 Displaced morphology in Dutch: Variation in non-finite verb clusters Cora Pots (KU Leuven) [email protected] Variation in te-placement in Dutch The morphosyntactic variation in Dutch finite verb clusters has been studied extensively (Barbiers et al. 2005, 2008; Wurmbrand 2015), but their non-finite counterparts have received much less attention (though see Vanacker (1969) for an early descriptive account). This paper presents new data on the placement of the infinitival marker te ‘to’ in three-verb clusters in Dutch (1). Furthermore, it shows that te placement happens post-syntactically and that regional differences can be accounted for by different PF mechanisms. (1) a. . . . om dat te hebben1 kunnen2 kopen3 in.order that to have.inf can.inf buy.inf b. . . . om dat hebben1 te kunnen2 kopen3 in.order that have.inf to can.inf buy.inf c. . . . om dat hebben1 kunnen2 te kopen3 in.order that have.inf can.inf to buy.inf ‘. . . in order to have been able to buy that.’ (ok northern Dutch, ok southern Dutch) (*northern Dutch, ok southern Dutch) (*northern Dutch, ok southern Dutch) In (1), the complementizer om selects a te-infinitive: VP1 . In (1a) te shows up on V1 as expected. In (1b) and (1c), however, te appears on V2 and V3 respectively instead of on V1 . Northern varieties of Dutch only allow the structure in (1a), whereas southern varieties accept (1a)–(1c). Displcaed zu in German The pattern in (1) is very similar to displaced zu in German (Salzmann 2013, 2016). German shows displaced morphology in clusters that are not strictly descending (2): (2) a. . . . ohne das Buch lesen3 gekonnt2 zu haben1 321 order without the book read.inf can.ptcp to have.inf b. . . . ohne das Buch haben1 lesen3 zu können2 132 order without the book have.inf read.inf to can.inf ‘. . . without having been able to read the book.’ (Salzmann 2016: 406) 68 Speaker Abstracts In (2), ohne selects a zu-infintitive. In a strictly descending order, zu appears on the hierarchically highest verb haben (2a). In (2b) however, zu does not occur on V1 , but on V2 . These data show that zu placement depends on linear adjacency rather than on hierarchical structure. Salzmann (2013, 2016) therefore argues that zu is attached to the verb post-syntactically, by Local Dislocation (Embick & Noyer 2001). Local Dislocation takes place when the hierarchical structure is linearized: the infinitival marker is attached to the linearly most adjacent verb and is inverted with it. Salzmann (2016) assumes a head-final structure for verb clusters in German, meaning that zu takes its verbal complement to the left. Linearization is bottom-up (Embick & Noyer 2001): first the verb cluster is linearized and then zu is placed. Zu placement in (2a) and (2b) is given in (3a) and (3b): (3) a. b. 3[21] zu → 3[2 zu 1] ‘lesen gekonnt zu haben’ 1[32] zu → 1[3 zu 2] ‘haben lesen zu können’ Prerequisites for the analysis Salzmann (2016) argues that in Dutch, the infintival marker te is attached to the verb by Lowering. Lowering of a functional head applies directly after the syntactic structure is sent to PF: the infinitival marker thus attaches to the hierarchically highest verb. In (1a), te is lowered and attached to V1 hebben. Then, the structure is linearized into a 123 order, resulting in the structure ‘te hebben kunnen kopen’. Salzmann’s (2016) analysis cannot explain the position of te in (1b) and (1c) however. I therefore propose a different analysis: in northern varieties of Dutch, te is indeed attached to the verb by Lowering, but in southern varieties, te is attached by Local Dislocation. Furthermore, I do not take linearization in Dutch to be head-final but head-initial, meaning that te is linearized to the left of the verb cluster. The analysis The new analysis thus consists of two claims: i) linearization in Dutch is head-initial, and ii) in northern varieties of Dutch te is attached to the verb by Lowering, but in southern varieties by Local Dislocation. In northern varieties, only the structure in (1a) can be derived. Te is lowered onto V1 hebben before linearization takes place, and therefore cannot occur in any other position than the one in (1a). In southern varieties however, te is attached to the verb after linearization of the verb cluster. The structure in (1a) is thus derived in a different way in southern Dutch: the cluster is linearized and then te is attached to the linearly adjacent verb hebben. Te placement in (1a) in southern varieties is given in (4a). The structure in (1b) is derived by Local 69 Cora Pots Dislocation: te attaches to the complex head [12] of the cluster [12]3. Then Local Dislocation takes place, and te is attached to and inverted with hebben. Te placement in structure (1b) is given in (4b). (4) a. b. te 123 → te123 ‘te hebben kunnen kopen’ te [12]3 → [1 te 2]3 ‘hebben te kunnen kopen’ The two claims of this analysis lead to two extra structures that should occur: i) a structure in which V1 and V2 are inverted by Local Dislocation leading to a te[21]3 order, and ii) a structure in which V2 and V3 are inverted by Local Dislocation, leading to a te1[32] order. These structures are indeed grammatical in southern varieties of Dutch, given in (5a) and (5b) respectively. (5) a. b. . . . om dat te kunnen2 hebben1 kopen3 1 in.order that to can.inf have.inf buy.inf Dutch) (southern . . . om dat te hebben1 kunnen2 kopen3 (southern Dutch) in.order that to have.inf can.inf buy.inf ‘. . . in order to have been able to buy that.’ The structure in (1c) ‘hebben kunnen te kopen’ seems to be a problem for the present analysis, because Local Dislocation only attaches and inverts the infinitival marker with the linearly most adjacent verbal element, meaning that te should not be able to occur on V3 . However, southern varieties of Dutch also allow te-doubling in these types of clusters, as shown in (6). (6) . . . om dat te hebben1 te kunnen2 kopen3 in.order that to have.inf to can.inf buy.inf ‘. . . in order to have to be able to buy that.’ (southern Dutch) I therefore propose, in line with Zwart (1993) for similar clusters, that this structure is derived from one in which V1 and V2 are both te-infinitives: teV1 -te-V2 -V3 . The structure in (1c) is then derived by Local Dislocation of the second te and V3 , followed by deletion of the highest te. Te placement in the structure in (6) is given in (7): 1 Note that it is controlled for a reverse scope reading of V1 hebben and V2 kunnen by the context sentence given before the test sentence: Heb je het huis gezien dat hij gekocht heeft? ‘Have you seen the house he has bought?’, forcing a epistemic reading. 70 (7) Speaker Abstracts te 1 te [23] → te 1[2 te 3] ‘hebben kunnen te kopen’ The claim that the sentence in (1c) is derived from a te-V1 -te-V2 -V3 structure makes another prediction. That is, the structure te1te[32], in which both te’s occur and in which V2 and V3 are inverted by Local Dislocation, should also be grammatical. This structure is indeed allowed: (8) . . . om dat te hebben1 te kopen3 kunnen2 in.order that to have.inf to buy.inf can.inf ‘. . . in order to have been able to buy that.’ (southern Dutch) Conclusion and outlook This study reveals new data concerning te placement in Dutch non-finite clusters. I propose a new analysis in which te placement happens post-syntactically: by Lowering in northern varieties and by Local Dislocation in southern varieties. Time permitting, I will show how the present account can also be applied to other displacement phenomena in Dutch, such as displacement in present participle constructions in Dutch, in which the present participle morphology of V1 appears on V2 (Hoeksema 1993; Den Dikken 2004). 71 The de re reading and the universal quantificational phrase in Mandarin Rong Yin (University of Massachussetts Amherst) [email protected] Synopsis Examining the De Re reading with respect to the universal quantificational phrase (i.e., mei ge xuesheng ‘every student’) in Mandarin, I argue that data from Mandarin do not support the QR analysis or Keshet (2011)’s Split Intensionality Theory, but can be explained by a system incorporating World Pronouns. Data In Mandarin, when a universal quantificational phrase originates in object position in a simple sentence, it can scramble to pre-verbal position. This is shown in example (1). The syntactic and semantic properties of dou ‘all/each’ do not affect the topic at all, so I leave aside the issues of dou throughout the discussion. (1) wo mei-bu dianyingi dou hen xihuan ti 1.sg every-cl movie dou very like ‘I like every movie very much.’ In Mandarin, the verb renwei ‘to believe’ and juede ‘to think’ can both take a clausal complement, which is shown in examples (2)–(3). In (2), the universal quantificational phrase scrambles to the pre-verbal position in the embedded clause. In (3), the universal quantificational phrase is the subject of the embedded clause. A De Re reading is available in both (2) and (3). (2) Bucky renwei [β Steven mei-bu zhengzai shangying de Bucky believe Steven every-cl prog show de jilupiani dou kan le ti ] documentary.film dou watch asp ‘Bucky believes that Steven has watched every documentary film that is on.’ (3) Steven juede [γ mei-bu zhengzai shangying de jilupian Steven think every-cl prog show de documentary.film dou hen xiaren ] dou very frightening 72 Speaker Abstracts ‘Steven thinks that every documentary film that is on is very frightening.’ However, only the universal quantificational phrase in (3) but not (2) can scramble to the pre-verbal position in the matrix clause, as shown in (4) and (5). (4) Steven mei-bu zhengzai shangying de jilupiani dou juede Steven every-cl prog show de documentary.film dou think [γ ti hen xiaren ] very frightening ‘Steven thinks that every documentary film that is on is very frightening.’ (5) Bucky mei-bu zhengzai shangying de jilupiani dou Bucky every-cl prog show de documentary.film dou renwei [β Steven t′i kan le ti ] believe Steven watch asp ‘Bucky believes that Steven has watched every documentary film that is on.’ QR Analysis It is beyond the scope of the discussion here what exactly the syntactic properties of the clause β in example (5) and γ in example (4) are (e.g., whether β, γ are a CP or a TP, whether β, γ are finite or infinitival, etc.). However, regardless of β and γ’s syntactic properties, I propose that it is highly possible that the clause β in (5) is an island preventing the universal quantificational phrase from scrambling out of the embedded clause; while the clause γ in (4) is not an island and the universal quantificational phrase can scramble out of it to the matrix clause. Assuming that at LF, β is still an island that quantificational phrases cannot escape, a QR analysis cannot explain why a De Re reading is available in both (2) and (3). Under a QR analysis, in order to get the De Re reading in (2), the universal quantificational phrase has to move to the pre-verbal position in the matrix clause (i.e., a position that precedes the main verb renwei ‘to believe’). However, as shown in (5), QR cannot escape the clause β, and thus it is impossible for the universal quantificational phrase to QR to a position that precedes/scopes over the verb renwei ‘to believe’ in the matrix clause that possesses the intensional operator. In this sense, the QR analysis falsely predicts that (2) cannot have the De Re reading. Rong Yin 73 More Data When the intensional verb is embedded in an if-clause, as shown in (6), the universal quantificational phrase can still be evaluated in the real world (i.e., De Re reading). (6) Context: Bill is a professor. ruguo [TP Mary renwei [β mei-bu jiaoshou dou shi xuesheng ]]], if Mary believe every-cl professor dou be student ta hui renwei Bill shi xuesheng 3.sg would believe Bill be student ‘If Mary thought that every professor was a student, she would think that Bill is a student.’ Keshet (2011)’s Split Intensionality Theory Under Keshet (2011)’s theory, at LF, the universal quantificational phrase must move out of clause β and take scope over the intensional operator introduced by if. However, since β is an island for QR at LF, the universal quantificational phrase cannot move out of β. In this sense, Keshet (2011)’s theory cannot explain why the universal quantificational phrase in (6) can be evaluated in the real world without violating the syntactic rules. A system incorporating World Pronouns Under the system incorporating World Pronouns, sentence (6) can be interpreted as having the De Re reading without violating any syntactic rules. In the left structure, the NP jiaoshou ‘professor’ takes as an argument the world pronoun variable that is bound by the 74 Speaker Abstracts operator w1. In this sense, the property of being a professor can be evaluated in the real world and enables the De Re reading in (6). The De Re reading in sentence (2) can also be explained using World Pronouns: the NP zhengzai shangying de jilupian ‘documentary film that is on’ takes an argument the world pronoun variable that is bound by the highest operator, which makes it possible to interpret the property of “documentary film that is on” in the real world. Potential Problems In Mandarin, the universal quantificational phrase can also occur in pre-subject position, which is shown in (7). If this means that the universal quantificational phrase can scramble to a pre-subject position, both QR analysis and Keshet (2011) can explain the De Re reading in (2) and (6). (7) mei-bu zhengzai shangying de jilupiani Bucky dou every-cl prog show de documentary.film Bucky dou renwei [β Steven t′i kan le ti ] believe Steven watch asp ‘Bucky thinks that Steven has watched every documentary film that is on.’ However, I propose that it is very unlikely that the universal quantificational phrase can scramble to the pre-subject position from the post-verbal object position: In both (2) and (3), the universal quantificational phrases can occur in pre-subject position in the matrix clause and do not seem to be sensitive to the clausal boundaries at all. In other words, it is possible that universal quantificational phrase in pre-subject position is base-generated and probably binds a variable in the object position in the embedded clause, instead of being derived by movement. 75 A stratal analysis of truncation in Spanish: Morphological and phonological evidence Javier Sanz (University of Trier) [email protected] The most common truncation process in Spanish consists in the shortening of a prosodic word into a disyllabic trochee. This happens irrespectively of both the length and the stress configuration of the base word; e.g., amplificadór → ámpli ‘amplifier’, compañéro → cómpa ‘classmate’, Concepción → Cónce ‘FEM name’, Gertrúdis → Gértru ‘FEM name’ (Casado 1984, Prieto 1992). Hypocoristic truncation can be further subdivided into two different processes, one of which preserves the two leftmost syllables of the base name while the other preserves both the stressed and final syllables (Prieto 1992, Piñeros 2000): (1) Name Gender Ignácio Jeús Dolóres masc masc fem Left-anchored truncate Igna Jésus Dólo Stress-anchored truncate Nácho Chus~Chúso Lóles~Lóla The two patterns differ in (a) their productivity, (b) their degree of segmental faithfulness to the base name, and (c) the possibility to modify the ending of the truncate. Left-anchored forms are by far the most productive of the two types (Prieto 1992). They are also predominantly faithful to the segmental makeup of their respective base forms (Prieto 1992, Piñeros 2000), as well less likely to change their endings (Roca and Felíu 2003). Stress-anchored variants generally only occur as proper name truncates and, even then, they are less productive than left-anchored hypocoristics (Prieto 1992). As shown in (1), these type of hypocoristics tend to undergo a series of phonological processes including palatalization, plosivization, coda deletion and reduplication. Because of these processes, Stress-anchored forms have long being associated to child speech (Boyd-Bowman 1955). Previous analyses couched within Output-to-Output Correspondence approaches to Optimality Theory (McCarthy and Prince 1995, Benua 1997) have accounted for both types of truncation as the result of different Emergence of the Unmarked effects (Prince and Smolensky 1993/2004, McCarthy and Prince 1994). E.g., Piñeros (2000) assumes that, whereas Stress-anchored truncation 76 Speaker Abstracts would be caused by the domination of certain prosodic as well as certain melodic constraints over Faithfulness, in the case of Left-anchored truncates only prosodic constraints would be higher ranked. Indeed, the unmarked pattern of Stress-anchored hypocoristics is consistent with the cross-linguistic tendency observed in younger children to preserve the phonetically prominent syllables of adult words (Kehoe 2000, Demuth 2001). However, Left-anchored forms such as Jesús → Jésus, Rubén → Rúben or Miguél → Míguel show that the fully productive process that takes place in the adult grammar need not result neither in a less marked prosodic structure (the previous hypocoristics are more marked than their corresponding base names regarding the Weight-to-Stress principle) nor need they be truncates in the proper sense. A more fitting account of productive truncation must hence posit the existence of a fixed template consisting in a disyllabic trochee that maximally copies the leftmost segments of an already existing word while complying with the phonotactics of the language. Therefore, and contrary to the view of non-lexicalist frameworks such as Distributed Morphology (Marantz 2001), fully productive truncation cannot occur at the root level. Otherwise, we could not account for the different morphological behaviour observed in the two types of hypocoristics. On the one hand, Stress-anchored hypocoristics whose endings do not agree with the biological gender of their base names usually exhibit variant forms with the default gender marker (-o for masculine names; -a for feminine ones). They can obtain this marker either by transforming the ending of the truncate into the default vowel or by simply adding the marker (see e.g. Chus~Chúso, Lóles~Lóla in (1) above). This contrasts with Left-anchored truncates, which are much less likely to undergo such additions/changes (ØÍgno, ØJéso and ØDóla are not documented in the literature). This is indicative of the fact that, when truncation takes place on line, gender marker affixation has already taken place. On the other hand, since Stress-anchored forms are presumably stored in our lexicon as adult representations of child speech, they have the possibility to undergo the phonological changes required to comply to the default pattern. A further piece of evidence often ignored is the existence of templatic truncates made out of expressions consisting in one or more clitics attached to either a noun or a verb. Obviously, this type of forms cannot result from a process taking place neither at the root nor at the stem level. 77 Javier Sanz (2) Expression por favor for favour ‘please’ Truncate pórfa Expression sí le tengo yes it.dat (I) have ‘I’ve got it’ Truncate síle fin de semana end of week ‘weekend’ fínde no le tengo no it.dat (I) have ‘I haven’t got it’ nóle Nevertheless, the previous data can be explained if we adopt the architecture proposed by Stratal Optimality Theory (Bermúdez-Otero 2012, Kiparsky 2015). According to most versions of this framework, there are three hierarchically ordered strata: stems, words, and phrases. It is generally agreed that the Spanish lexicon stores stems with gender markers and other types of thematic vowels (Bermúdez-Otero 2013) and that fully productive morphological processes take place at the higher levels, which is precisely where cliticization is thought to happen (Bermúdez-Otero 2006). The particularities of Spanish truncates, and perhaps of truncation in general, can thus be better understood in light of some of the proposals made by Stratal Optimality Theory. References Benua, Laura. 1997. Transderivational identity: phonological relations between words. Doctoral Diss, University of Massachusetts, Amherst. Bermúdez-Otero, Ricardo. 2006. The phonology of cliticization in Stratal Optimality Theory. Handout of paper presented at Annual Meeting of the Linguistics Association of Great Britain, Newcastle upon Tyne, 31 August 2006. Available at www.bermudez-otero.com/clitics.pdf Bermúdez-Otero, Ricardo. 2012. The architecture of grammar and the division of labour in exponence. In The morphology and phonology of exponence, ed. by Jochen Trommer, 8–83. Oxford: Oxford University Press. Bermúdez-Otero, Ricardo. 2013. The Spanish lexicon stores stems with theme vowels, not roots with inflectional class features. Probus 25: 3–103. Boyd-Bowman, Peter. 1955. Cómo obra la fonética infantil en la formación de los hipocorísticos. Nueva Revista de Filología Hispánica 9: 337–366. Casado Velarde, Manuel. 1984. Acortamientos léxicos en el español actual. Iberoromania 20: 1–8. Demuth, Katherine. 2001. Prosodic constraints and morphological development. In Phonological, syntactic and neurophysiological aspects of early 78 Speaker Abstracts language acquisition, ed. by Jürgen Weissenborn and Barbara Höhle, 3–21. Amsterdam: John Benjamins. Kehoe, Margaret M. 2000. Truncation without shape constraints: the latter stages of prosodic acquisition. Language Acquisition 8: 23–67. Kiparsky, Paul. 2015. Stratal OT: a synopsis and FAQs. In Capturing phonological shades within and across languages, ed. by Yuchau E. Hsiao and Lian-Hee Wee, 2–44. Newcastle upon Tyne: Cambridge Scholars. Marantz, A. 2001. Words. WCCFL XX Handout, USC, February 2001. McCarthy, John and Alan Prince. 1994. The emergence of the unmarked: optimality in prosodic morphology. In Proceedings of the North East Linguistic Society 24, ed. by Mercè González, 333–379. Amherst, MA: GLSA. McCarthy, John and Alan Prince. 1995. Faithfulness and reduplicative identity. In University of Massachusetts Occasional Papers 18, ed. by Jill N. Beckman, Laura Walsh Dickey and Suzanne Urbanczyk, 249–384. Amherst, MA: GLSA. Piñeros, Carlos-Eduardo. 2000. Prosodic and segmental unmarkedness in Spanish truncation. Linguistics 38: 63–98. Prieto, Pilar. 1992. Truncation processes in Spanish. Studies in the Linguistic Sciences 22: 143–158. Prince, Alan and Paul Smolensky. 1993/2004. Optimality Theory: constraint interaction in Generative Grammar. Oxford: Blackwell. Roca, Iggy and Elena Felíu. 2003. Morphology in truncation: the role of the Spanish desinence. In Yearbook of Morphology 2002, ed. by Geert Booij and Jaap van Marle, 187–243. Dordrecht: Kluwer. 79 Some mathematical phonology – an extension of ‘delete-and-unify’ Ollie Sayeed (Christ’s College, Cambridge) [email protected] This is an attempt to formalize one implementation of (linear) Rule-Based Phonology based on set theory – the theory being formalized is an extension of Bale, Papillon and Reiss (2014; ‘BPR’) within Substance-Free Phonology to rules mapping whole strings to strings, needed to model insertion, deletion, metathesis, and other rules that don’t just involve mapping a single segment to a single other segment. To have a formal theory of phonology, we want to be able to define as much as possible from some collection of primitive objects using well-defined (in this case set-theoretic) operations. For phonology, our primitives are a set of feature values V – these are the usual [+voice], [−voice], [+high], [−high], etc. – and a set of timing slots I. Everything else can be built out of V and I by the normal operations of subsethood, intersection, union, complementation, power set. A segment is a subset (or ‘bundle’) of the set of feature values V ; a feature is an equivalence class of feature values from V ; a string is a function from I to the power set P(V ), assigning a segment to each timing slot; a rule is a set of ordered pairs of strings, mapping the first string to the second string; a feature geometry is a set of subsets of features, defining which sets of features can and can’t play a role in the rule component. The goal is that everything in the world of phonology should be expressible as a set of some kind, as has been successful in model-theoretic semantics (so this is ‘set-theoretic’ phonology). If we achieve this, we have a completely explicit picture of a phonological system as a formal object. I define a rule R in intension as an ordered triple (T, E, C): a structural description T (a string), an environment E (an ordered pair of strings), and a structural change C (another ordered pair of strings). The interpretation is that any string that ‘matches’ the string in the target T and is flanked by the pair of strings E changes in a way given by the strings in C. Like the BPR model, the idea of a string ‘matching’ a target is done by defining a relation of subsumption; but the limitation of the BPR model is that subsumption is only defined between pairs of segments, so rules can only map one segment to another. This can’t express insertion (where the number of segments needs to decrease), deletion (where it needs to decrease), or metathesis (where multiple 80 Speaker Abstracts segments are involved in the target). To fix this, I define an idea of subsumption between strings, rather than just segments: we say a string W1 subsumes a string W2 iff W1 can be mapped onto a subpart of W2 in an order-preserving way such that each segment in W1 is compatible with its corresponding segment of W2 . For a rule with structural description T to apply to a string W, T has to subsume W. The environment E is an ordered pair (E L , E R ): the other condition we require for a rule with environment (E L , E R ) to apply to a string is that E L (the ‘left environment’) subsumes the section of W2 immediately to the left of L(W1 ), and that E L (the ‘right environment’) subsumes the section of W2 immediately to the right of L(W1 ). In total, for a string W to be input to a rule R whose left environment, structural description, and right environment concatenate to a string Q, a necessary and sufficient condition is that Q subsumes W. Again following BPR, I take it that the application of a rule is a two-step ‘delete-and-unify’ process; first a subtraction of some features of segments of the target string, then a unification of those segments with some new features defined by the rule. So a structural change C = (C1 , C2 ) is defined by two strings: the material C1 to be deleted from the target string, and then the material C2 to be unified with it. We require that C1 can be mapped onto W in a way satisfying the second and third points under the definition of subsumption above. If a rule applies to a target string of length 3 and deletes [−voice] from the first segment, deletes [+high] and [−back] from the second one, and deletes the entire third segment, its C1 would look like this: C1 = {(i1 , {[−voice]}), (i2 , {[+high], [−back]}), (i3 , ∅)} The deletion part of the rule will map i1 , i2 , and i3 onto the target string in a way that preserves the ordering under < on I, and then deletes [−voice] from the segment at the image of i1 , [+high] and [−back] from the segment at the image of i2 , and then the whole timing slot at the image of i3 . In sum, given a target string {( j1 ,S1 ),( j2 ,S2 ),( j3 ,S3 )}, deleting {(i1 ,{[−voice]}),(i2 ,{[+high],[−back]}),(i3 ,∅)} gives a new string: {( j1 , S1 − {[−voice]}), ( j2 , S2 − {[+high], [−back]}), (∅, S3 )} The second part of the rule involves ‘unifying’ the string C2 with the target string W, once the material from C1 has all been deleted. As with the mapping from C1 to W, we map the timing slots of C2 by some function U onto the Ollie Sayeed 81 timing slots of the new W - and at each timing slot, we take the union of each set of feature values at each i in C2 with its corresponding set at each U(i) in W. If a rule unifies takes a string of length three and unifies {[−ATR]} with the first slot, nothing with the second, and {[−voice],[+ATR]} with the third, its C2 looks like this: C2 = {(i1 , {[−ATR]}), (i2 , ∅}), (i3 , {[−voice], [+ATR]})} If we unified this with a string {( j1 ,S1 ),( j2 ,S2 ),( j3 ,S3 )}, the output of the rule would be the string: {( j1 , S1 ∪ {[−ATR]}), ( j2 , S2 ), ( j3 , S3 ∪ {[−voice], [+ATR]})} I believe this is the simplest and most mathematically natural extension to the BPR ‘delete-and-unify’ model of rule semantics that also accounts for the rules the BPR model currently can’t handle. After presenting this model, I’ll discuss the theoretical motivation behind inferring a particular model of phonology from a set of attested languages. It might seem that the model isn’t restrictive enough, in that it generates languages that aren’t attested; I’ll argue that concepts like ‘restrictiveness’ and ‘generative power’ shouldn’t play a role in reasoning about phonology. Unattested rule patterns predicted to be computable by a particular grammar are more sensibly treated as unattested for diachronic reasons, rather than facts about the synchronic grammar (adopting the perspective argued by Blevins, 2004). On the other hand, it would also be wrong for our null hypothesis to be that phonology can generate any computable pattern; this would require proposing extra machinery not evidence by attested languages, like the tape of a Turing machine. I argue that a model like this one hits a middle level, by being general enough to avoid being stipulatively restrictive while also not proposing more complex machinery than we know phonology has access to. References Bale, Alan, Maxime Papillon, and Charles Reiss. 2014. Targeting underspecified segments: A formal analysis of feature changing and feature filling rules. Lingua 148: 240–253. Blevins, Juliette. 2004. Evolutionary Phonology: The emergence of sound patterns. Cambridge University Press. Hale, Mark, and Charles Reiss. 2008. The phonological enterprise. Oxford University Press, USA. 82 Speaker Abstracts Johnson, C. Douglas. 1972. Formal Aspects of Phonological Description. Mouton. Kaplan, Ronald M., and Martin Kay. 1994. Regular models of phonological rule systems. Computational Linguistics 20(3): 331–378. 83 The syntax of U/AX right dislocation Ui-Jong Shin & Sunjoo Choi (Dongguk University) [email protected] Merchant (2004), Griffiths and Lipták (2014), and Ko (2014), among many others, unanimously assume that in the question-answer pairs, the ‘short’ answer sentences below in English, Hungarian, and Korean are an instance of (contrastive) fragment. (1) Q: Does Abby speak GREEK fluently? A: No, ALBANIAN. – English Merchant (2004: 688) (2) Q: Mari BÉLÁT hívta meg mágahoz enni? – Hungarian Mari Béla.A invited PV herself.TO eat.INF ‘Did Mari invite BÉLA to eat?’ A: Nem, vacsorára mindig PÉTERT. no dinner.FOR always Péter.A ‘No, for dinner she always invited PETER.’ Griffiths and Lipták (2014) (3) Q: Mary-ka motwu ta an manna-ss-ni? Mary-Nom all not meet-Past-Q (lit.) ‘Didn’t Mary meet all/any of them?’ A: Ung, motwu ta. Yes, all (of them). (lit.) ‘Yes, Mary did not meet all of them.’ – Korean (all≫Neg, Neg≫all) (all≫Neg, *Neg≫all) Ko (2014: 293) In this paper we argue that these sentences are not such an instance, but that they are instead derived from right dislocation (RD). We begin to establish that the sentences like (3A) in Korean composed of the polarity answer particle (PAP) such as ung ‘yes’ or ani ‘no’, followed by one XP remnant (alias U/AX) involve RD, by showing that not just the PAP but also the immediately following covert structure ‘reconstructed’ from the preceding question clause is indispensible to licensing the following XP remnant, as in (4) and (5): (4) Q: chelswu-ka mek-ess-ni? Chelswu-nom eat-pst-q 84 Speaker Abstracts ‘Did Chelswu eat?’ A: Ani, amwu-kes-to. no, anything ′ A : *Ung, amwu-kes-to. yes, anything (5) Q: chelswu-ka mek-ci anh-ass-ni? Chelswu-nom eat-nm not-pst-q ‘Did Chelswu eat?’ A: Ung, amwu-kes-to. yes, anything ′ A : *Ani, amwu-kes-to. no, anything In (4A) and (4A′ ), the negative polarity item (NPI) remnant occurs after the PAP. The negative PAP is a licensor of the following NPI remnant. However, the situation is opposite in (5A) and (5A′ ) as answers to the preceding negative question, where not the negative but the positive PAP is an NPI licensor. This points to the fact that not only the overt PAP but also the postulated covert TP immediately following it is essential in licensing the NPI remnant. We suggest à la Kramer and Rawlins (2009) and Holmberg (2015) that the PAP is also a remnant derived by elision of the TP. For example, (4A) has the following underlying structures prior to such an elision: (6) A: Ani [TP chelswu-ka [e] mek-ess-e], amwu-kes-to. No Chelswu-nom eat-pst-decl anything In this structure, the right-edge remnant (including the RDed NPI) is linked to and may or may not be licensed in the preceding host clause, exactly in the same way as the RDed elements in the canonical RD construction. Another point can be made supporting the RD analysis of the XP remnant at issue. Ross (1986: 260) notes that the RDed element is associated with the possessive pronoun as in (7). This is also the case with the English counterpart of UAX as in (8). However, the run-of-the-mill fragment cannot be realized as a bare DP as in (9): 85 Ui-Jong Shin & Sunjoo Choi (7) I noticed his car in the driveway last night, your friend from Keokuk. (Ross 1986: 260) (8) Q: Did you notice his car in the driveway? A: Yeah, your friend from Keokuk (9) Q: Whose car did you notice in the driveway? A: Your friend*(’s) This conception of UAX as involving RD is adopted to resolve a controversy concerning the asymmetry between Sluicing and UAX-RD in light of the island sensitivity of contrastively focused remnants as in (10) and (11): (10) *Abby wants to hire someone who speaks GREEK, but I don’t remember what OTHER languages she wants to hire someone who speaks. Merchant (2008: 148) (11) Q: Did you hear they hired someone who speaks BULGARIAN fluently? A: (*)No, SERBO-CROATIAN. * in Griffiths and Lipták (2014); OK in Barros et al. (2014) Since RD is subject to the Right Roof Constraint (RRC) (Ross 1967), UAX-RD in (11) applies successfully in juxtaposition not with the whole sentence, but with the relative clause. The contrastive focused wh-element in (10), however, undergoes leftward movement, inducing island effects. However, if such a juxtaposition is not available because an intervening clause, UAX-RD fails to meet the RRC, ruling out examples like (12): (12) Q: Is the book that RINGO wrote on sale? A: *No, Lennon. In addition to UAX-RD with a single remnant, there is UAX-RD with multiple remnants, as in (13). One restriction on multiple UAX-RD is that their RDed elements need to be derived from the same clause, regardless of whether they receive informational or contrastive focus. (13) Q: JOHN talked to MARY. A: No, BILL to SUSAN. (14) A: Did someone hear that Mary talked about some issue in syntax at the conference last week? B: *Yeah, JOHN, about LABELING. (15) A: Did the MOTHER say that Mary was on a date with JOHN at the 86 Speaker Abstracts mall? B: *No, the FATHER, with BILL. This clause-mate requirement on RDed elements in multiple UAX-RD naturally follows from the RRC that regulates RD. Since two RDed elements in (14) and (15) are derived from the two different clauses, one of them is bound to violate the RRC. Selected References Barros et al. 2014. There is no island repairs. Ms. Griffiths and Lipták (2014) Contrast and island sensitivity in clausal ellipsis, Syntax 17(3). Holmberg. 2015. The Syntax of Yes and No. Ko. 2014. Right dislocation as specificational focus. Ms. Kramer and Rawlins. 2009. Polarity particles: An ellipsis account, NELS 39. Merchant 2004. Fragments and ellipsis; L&P 27(6). Ross. 1986. Infinite Syntax! 87 Syntax of Finnish numerical constructions Philip Shushurin (New York University) [email protected] Data This paper proposes an account of case-marking and number-marking in Finnish constructions with numerals. The problems that I address in this paper is why agreement pattern in numerical constructions where the numeral is marked nominative (henceforth direct constructions) is different from those where the head is marked with any other case (henceforth indirect constructions) and why numerals and all nominals following them (henceforth postnumerical nominals) are marked singular despite semantic plurality. In direct constructions all nominals following the numeral are marked singular partitive (1), while in indirect constructions all postnumerical nominals are also marked singular but agree with the numeral in case: (1) kaksi iso-a talo-a two.nom.sg big-part.sg house-part.sg ‘two big houses’ (2) kahde-ssa iso-ssa talo-ssa two-sg.iness big-sg.iness house-sg.iness ‘in two big houses’ Account In this paper I develop the idea proposed in Danon (2012) that numerals are merged as heads in some cases and as modifiers in others. More specifically, I assume that all Finnish numerals (except for yksi, ‘one’; this numeral never appears with partitive nominals) can occupy a specifier position or the position of the head of QP. In direct constructions, such as (1), the numeral occupies the position of the QP head, in which case the QP assigns its complement partitive case. If the noun in the complement is further modified, for instance by adjectives, the modifier exhibits concord with it in case and number (Fig. 1). In indirect constructions, such as (2), the Q head is empty and doesn’t assign partitive case. (By assumption, only overt Q heads can assign partitive case). In this case the Q head is empty and it does not assign the partitive case to its complement. Numerals in this case occupy the specifier of QP and demonstrate the modifier behavior: they agree with the noun both in case and in number (Fig. 2). Since the modifier option and the head option are 88 Speaker Abstracts QP Q kaksi QP NP [PART] Adj NP [PART] iso-a talo-a Figure 1 kahde-ssa Q′ NP [INESS] Adj NP [INESS] isso-ssa talo-ssa Figure 2 in strict complementary distribution one may wonder what conditions which option is selected in each particular case. I assume that (a) the head option is the default option (i.e. if the numeral can be both a head and a modifier the head option is chosen), the modifier option is chosen when the head option is blocked. The case where the head option is blocked is indirect constructions. I want to propose that (a) in all indirect constructions there is a local assigner of case: a PP, in the case of locative cases, or a DP, in the case of genitive DPs, and (b) that the numeral cannot occupy the head position in such cases because its complement, to which the Q assigns the partitive case, would be unable to agree in its case feature because it is already case-marked (by QP). In other words, I propose that no nominal in Finnish can get be case-marked more than once. (In direct constructions, nouns are marked nominative; since the morphological realization of nominative case is always null, no such violation may occur). I avoid look-ahead problem by suggesting that the modifier and head options are independent and the impossibility of one option results only in that the other becoming the only option. Thus, in indirect constructions the only remaining option is the modifier option. In this case the head of Q is empty, the noun remains unmarked in case until it is case-marked by a functional projection: A PP in the case of locative cases and a DP in the case of genitive. Number marking Another important aspect is the number marking. Postnumerical nominals in both head position and modifier position are always Philip Shushurin 89 marked singular. In contrast, in prenumerical position all agreeing nominals are obligatorily plural-marked: (3) ne mukava-*(t) kaksi pien-tä talo-a these nice-nom.pl two.nom.sg little-part.sg house-part.sg ‘these two nice little houses’ (When no numeral is present semantically plural nominals always get plural morphology). I assume that plurality in Finnish is encoded by a privative [+pl] feature which is usually (see the discussion of pluralia tantum nouns below for exceptions) contained on a NumP, a functional projection above QP. While all DP-internal elements usually must agree in this feature, this agreement is blocked when a QP is present. I hypothesize that the blocking effect is due to phasehood. I propose that the Q projection is a phase and thus the agreement between the Num’ bearing the [+pl] feature and its potential goal inside an NP is failed (Preminger 2011). I further hypothesize that although Num probe fails to probe down, it establishes an Agree relation with the phase head (or its Spec, when it is a modifier). This operation results in ‘unlocking’ the phase, now other probes can probe down into the NP. The case probe, once the PP or DP is merged, is able to establish the agree relation with the nominal in the NP. When there is no PP or DP case-assigner, but there is an overt Q head, its complement gets partitive case as usual (Fig. 3). This idea, with minor further developments, can also account for another complicating type of numerical constructions, namely, numerical constructions involving pluralia tantum nouns. In such constructions the noun, all its modifiers and the numeral are marked plural. (4) ne kahde-t/*kaksi piene-t/*pieni these two-nom.pl/two.sg.nom small-nom.pl/small.nom.sg hää-t wedding-nom.pl ‘two weddings’ Since the noun is marked nominative and not partitive I assume that the numeral here occupies the modifier position. The head option must be blocked in this case because a Q head must select for a singular complement. (This option is unavailable since pluralia tantum nouns are always plural). Adopting an idea proposed in Kramer (2016) for Amharic data, I assume that pluralia 90 Speaker Abstracts tantum nouns contain a [+pl] feature on n. This feature triggers QP-internal plural agreement according to general concord rules. QP-external plural agreement is due to a different [+pl] feature contained on a NumP. Strictly speaking, blocking occurs in this case as well, however, a QP has an internal source of plurality. Unlike accounts proposed earlier, such as (Brattico 2011) this account does not stipulate two kinds of case features, viz. strong and weak. DP [+PL] NumP [+PL] D ne Adj Num′ [+PL] QP mukava-t Q kaksi NP [PART] Adj NP [PART] pien-tä talo-a Figure 3 References Brattico, P. 2011. Case assignment, case concord, and the quantificational case construction. Lingua 121, 1042–1066. Danon, G. (2012). Two structures for numeral-noun constructions. Lingua 122(12), 1282–1307. Preminger, O. (2011). Agreement as a fallible operation. Doctoral dissertation, Massachusetts Institute of Technology. Kramer, R. A split analysis of plurality: Number in Amharic. Linguistic Inquiry 47, 527–559. 91 The atoms of person: Limitations on concept formation Jolijn Sonnaert (KU Leuven Brussels) [email protected] Claim I argue that inclusive person is the sum of the person atoms {sp} and {hr}, and show that the unlexicalisability of other sums of atoms is predicted by the concept formation constraint (cfc) in the kite framework (Seuren & Jaspers 2014). The atoms of Person Semantically, inclusive refers to a group including speaker and hearer. Morphologically, inclusive is most often independent from first (speaker) and second person (hearer) (1). Otherwise, it is not infrequently related to 1st person (2) and sometimes also to 2nd (3) (Daniel 2005). (1) Tümpisa Shoshone SG PL incl ta-mmü 1 nü nü-mmü 2 ü mü-mmü 3 (dem) (demonstr) (3) Tok Pisin SG incl 1 mi 2 yu 3 em (2) Quechua SG incl 1 nuxa 2 xam 3 pay PL nuxa-ñči(k) nuxa:-guna xam-guna pay-guna PL yu-mi(-pela) mi-pela yu-pela ol Based on these considerations, I consider inclusive to be the sum of the atoms for first and second person. The Hasse diagram below shows how atoms (represented by bitstrings) can combine into different sums (Smessaert 2009): every 1-bit represents exactly one atom (Fig. 1). Level 1 in Fig. 2 shows the atoms for 1st {sp}, 2nd {hr} and 3rd person {non-part}. Level 2 on the left in Fig. 2 shows the sum of speaker and hearer: inclusive {sp, hr}. I argue that the other two sets ({sp, non-part} and {hr, non-part}) are never lexicalised 92 Speaker Abstracts as a person morpheme. Fig. 3 shows how the person paradigm of Tümpisa Shoshone in (1) is mapped onto the Hasse-diagram for person. Figure 1: 3-atom Hasse diagram Figure 2: Person Hasse diagram Figure 3: Tümpisa Shoshone The Concept Formation Constraint The kite framework studies (mereo)logical relations between concepts and lexical items in closed lexical fields (e.g. the quantifiers) using geometrical figures to represent these relations (Jaspers 2012, Seuren & Jaspers 2014). The logical hexagon by Jacoby, Sesmat and Blanché (a.o. Blanche 1952) in Fig. 4 shows the relations between the lexical items with arrows for entailment, full lines for contradiction, and dotted and dashed lines for (sub)contrariety. Fig. 5 exemplifies this for the quantifiers. Figure 4: The hexagon Figure 5: The hexagon: Quantifiers 93 Jolijn Sonnaert The hexagon shows the following restriction for lexicalisation in closed lexical fields: (4) concept formation constraint (Seuren & Jaspers 2014, p. 621–626): Both the O- and U-corner never receive a simplex lexicalisation. This turns the hexagon into a kite of lexicalised concepts, shown in Figs. 4 and 5 with the bold lines. Mereologies The hexagon and cfc can be applied to logical (e.g. quantifiers above) and mereological lexical fields (e.g. colour, Jaspers 2012). Person is a case of the latter. Mereologies deal with parthood relations rather than the logical entailments. Note that these relations show a clear isomorphism (Smessaert 2009, Jaspers 2012). Both entailment (ϕ → ψ) and proper parthood (ϕ ⊂ ψ) between the bitstrings representing the corners can be calculated with the same formula (see (7) and (8) for an illustration): (5) a. ϕ∧ψ = ϕ b. ϕ∨ψ =ψ An important difference between the logical and the mereological hexagon is the nature of the I-O-U corners. Rather than a disjunction of their adjacent corners as for the quantifiers (6a), in mereologies they designate a new element based on the mereological sum of these (6b) (Jaspers 2012). (6) a. b. Logical kite: I = A ∨ Y (e.g. ‘some’ = ‘all or some’) Mereological kite: I = A ⊕ Y (e.g. incl {sp, hr} = speaker and hearer) The Person Kite The hexagon and kite below portray exactly the relations we expect to find between the attested person distinctions. Both {sp} and {hr} are proper parts of inclusive: {sp, hr}: (7) a. b. 100 {sp} ∧ 110 {sp, hr} = 100 {sp} 100 {sp} ∨ 110 {sp, hr} = 110 {sp, hr} (8) a. b. 010 {hr} ∧ 110 {sp, hr} = 010 {hr} 010 {hr} ∨ 110 {sp, hr} = 110 {sp, hr} Also, the other mereological sums of person atoms as seen in the Hasse diagram in Fig. 2, reside in the unlexicalised corners U and O. No language has a person 94 Speaker Abstracts morpheme to express a non-speaker, i.e. {hr, non-part}, or a non-hearer, {sp, non-part} (as exemplified in Fig. 8). This is confirmed by the 39 languages of my sample and the typological literature studied (a.o. Daniel 2005, Cysouw 2009, Forchheimer 1953, Harbour 2016, Ackema & Neeleman 2016). Figure 6: Person hexagon Figure 7: Person kite Figure 8: Tümpisa Shoshone References Ackema, Peter & Ad Neeleman. 2016. Features of person. To Appear. Blanché, Robert. 1952. Quantity, modality and other kindred systems of categories. Mind 61(243), 369–375. Cysouw, Michael. 2009. The paradigmatic structure of person marking. New York: Oxford University Press. Daniel, Michael. 2005. Understanding inclusives, 3–48. John Benjamins Publishing Co. Forchheimer, Paul. 1953. The category of person in language. Berlin: Walter de Gruyter & Co 2014th edn. Harbour, Daniel. 2016. Impossible persons. To Appear. Jaspers, Dany. 2012. Logic and colour. Logica Universalis 6, 227–248. Jolijn Sonnaert 95 Seuren, Pieter A. M. & Dany Jaspers. 2014. Logico-cognitive structure in the lexicon. Language 90(3), 607–643. Smessaert, Hans. 2009. On the 3D visualisation of logical relations. Logica Universalis 3, 303–332. 96 Epenthesize a mora, but pronounce a vowel – The Serbo-Croatian language game of Šatrovački Jelena Stojković (University of Leipzig) [email protected] Introduction I discuss novel data from a Serbo-Croatian (SC, henceforth) syllable reversal ludling, Šatrovački. Previous analysis (Rizzolo 2004) has failed to explain the existence of a schwa vowel in the ludling, since it is not present in the host language. Following Itô et al. (1996 on zuuja-go) that ludlings in principle do not violate OO-correspondence in terms of segments, I presuppose that Šatrovački exhibits epenthesis of a maximally underspecified element, specified only with a mora and feature [−cons] (McCarthy 1988, Zimmermann 2016). The schwa vowel on the surface is observed as a phonetic implementation of such nucleus (Polgárdi 1996). Background Šatrovački is a SC syllable reversal ludling, comparable to verlan in French or zuuja-go in Japanese. The basic principle of these ludlings can be captured by a single constraint CrossAnchor, which observes a PWd as two sections whose order must be reversed (Itô et al. 1996: 37). I focus here on outputs based on CVC monosyllabic source forms (SF), which are created by moving the consonantal material to the front of the PWd. The output forms (OF) are adapted to conform to the phonology of the host language, either via sonorant syllabification or epenthesis. Some analyses of syllable reversal games (Friesner 2005 for verlan, Rizzolo 2004 for Šatrovački) assume vowel epenthesis and thus fail to account for the vowel quality, different than the host language’s default vowel (/a/ in SC, according to Simonović 2015). Furthermore, schwa is not a licit vowel in SC. Data based on recorded conversations with 20 speakers of Šatrovački and supplemented with examples from Rizzolo (2004) reveal that there are two ‘dialects’ of Šatrovački. In the variety illustrated in (1) a schwa is added even when it is not phonotactically necessary (1-a) (Rizzolo 2004). Additionally, all consonantal material is moved (1-e), and the syllable headed by schwa is always accented. (1) Outputs from one variety of Šatrovački a. ‘muž’ [muZ] ↦ ["Z@mu] ‘husband’ b. ‘dop’ [dop] ↦ ["p@do] ‘dope’ c. ‘smor’ [smor] ↦ ["rsmo] ‘boredom’ " 97 Jelena Stojković d. e. ‘dlan’ [dlan] ↦ ["n@dla] ‘palm’ ‘bend’ [bend] ↦ [n."[email protected]] ‘band’ " Proposal Following Itô et al. (1996) and extending the empirical scope of their analysis, I offer an OT account of Šatrovački, showing that epenthesis is the result of minimal violation of prosodic faithfulness, but not of segmental correspondence. Therefore, I argue that this epenthesis involves an interplay of constraints regulating foot structure and accent assignment, resulting in two kinds of implementation: syllabification (when the segment is syllabifiable) and schwa formation (phonetic implementation of a segmentless mora and the feature [−cons]), as a result of the undominated Empty Category Principle (Polgárdi 1996). /muZ/ a. ["Zmu] *! + b. ["Zµ .mu] c. ["Za.mu] *! * * * (3) FtBin Dep-IO SylMarg Licemse-µ Dep-µ (2) FtBin Dep-IO SylMarg License-µ Dep-µ Analysis The reversal of phonological material is driven by the CrossAnchor constraint (not included here). The repair strategy for cross-anchoring monosyllabic forms is epenthesis of a mora, which is a result of ranking FtBin ≫ Dep-IO ≫ Dep-µ. Insertion of a mora is used in two ways when otherwise Dep-IO and SylMarg (onset clusters may not be of descending sonority, Zec 2002a: 251) would be violated: (i) by assigning a syllabic role to /r/ as the only syllabic non-vowel in SC, and (ii) by leaving a vowelless mora behind, which surfaces as a schwa. This is ensured by the constraint Government Licensing / GL and the Empty Category Principle (Polgárdi 1996) that phonetically realizes segmentless nuclei. The mora is assumed to carry a [−cons] feature that is normally assigned to sonorants when they become syllabic. /dop/ a. ["pdo] *! * + b. ["pµ .do] c. ["pa.do] *! * * The FtBin constraint makes sure the mora is always inserted, even though the rank of SylMarg would predict epenthesis in (3), but not in (2), because /Zm/ is an allowed onset, but /pd/ is not. Since the accent can only be assigned to the leftmost element (NonFinality; not included here) of the binary foot (FtBin), the vowelless mora is accented and pronounced as a schwa. In attested 98 Speaker Abstracts *! * * * *! * *! Son-ϕ[−cons] (5) Dep-IO SylMarg License-µ Dep-µ /smor/ a. ["rsmo] + b. ["r.smo] " c. ["ra.smo] d. ["rµ .smo] Dep-IO SylMarg License-µ *P-/r/ Dep-µ (4) Son-ϕ[−cons] cases SylMarg dominates *P-/r/, which prohibits syllabic /r/, and Dep-µ, but it is dominated by other faithfulness constraints, namely Dep-IO, w.r.t. to Son-ϕ[−cons]. Tableau (4) shows the evaluation of word-initial /r/ in a Šatrovački form. Since /r/ can be syllabified and accented, the epenthesized mora is adjoined with the sonorant (4). Candidate d. (with an empty nucleus) is penalized by License-µ. /dlan/ a. ["ndla] *! b. ["n.dla] *! " + c. ["nµ .dla] * * d. ["na.dla] *! Max-IO SylMarg License-µ Dep-µ Son-ϕ[−cons] FtBin (6) *Coda-WdFin The difference between (4) and (5) lies in the quality of the initial segment. As argued by Zec (2002b: 127–128), the syllable and the foot can be associated with different sonority thresholds – Son-ϕ[−cons] requires that the head of the foot be vocalic, which, together with the assumption that /r/ is a [−cons] segment (Zec 2002b: 127) means that only vowels and /r/ can be accented. Since a syllabic /n/ as a head of a foot violates Son-ϕ[−cons], the mora in ["nµ .dla] (5) is not adjoined with a segment, but only with [−cons] feature. /bend/ a. ["dµ .ben] *! * * b. ["ndbe] *! * c. ["dµ .be] *! * * µ d. ["n.d .be] *! * * " + e. [n."dµ .be] * ** " f. [nµ ."dµ .be] **! ** Tableau (6) shows the effects of Son-ϕ[−cons] in accent assignment: /n/ or /d/ Jelena Stojković 99 are not acceptable as accent-bearing segments, so two moras are inserted and the second one is accented. Compared to (5), the tableau (6) assumes that a low ranked License-µ that chooses candidate e. as optimal by not allowing two empty nuclei. If we follow Friesner’s (2005) claim that the epenthetic segment must be present in the grammar of the host language (and, by extension, must follow from the same ranking), we have to conclude that schwa is a licit (though least marked) vowel of Serbo-Croatian. Its absence in the lexicon of Serbo-Croatian would have to be an accidental gap. Since this is not the case, observed as a phonetic realization of an accented segmentless mora, the schwa in Šatrovački follows from the Empty Category Principle. References Friesner, M. 2005. A unified account of the French language game Verlan. Ito, J. et al. 1995. Prosodic faithfulness and correspondence: Evidence from a Japanese argot. McCarthy, J. & A. Prince. 1995. Faithfulness and identity in prosodic morphology. Polgárdi, K. 1996. Constraint ranking, Government Licensing and the fate of final empty nuclei. Rizzolo, O. 2004. Šatrovački: la construction et l’exploitation d’un corpus de verlan serbo-croate. Simonović, M. 2015. Lexicon immigration service: Prolegomena to a theory of loanwords . Zec, D. 2002a. The role of prosody in morphologically governed phonotactic regulatities . Zec, D. 2002b. Prosodic Weight. 100 The resultative passive is agentive: A response to Embick (2004) Michael Wilson (University of Massachusetts Amherst) [email protected] Introduction It is generally accepted that there are at least two kinds of passive structures in English, a verbal passive and an adjectival passive (see Wasow 1977). Embick (2004) argues that the adjectival passive should be bifurcated into a resultative passive and a stative passive, making three total passives (including the verbal passive, which he refers to as the “eventive” passive). An example of the distinction is in (1): (1) The door is closed. a. ‘Someone closes the door (habitually)’ b. ‘The door is in a state of having become closed’ c. ‘The door is in a state of closed-ness’ (= eventive) (= resultative) (= stative) Embick (2004) argues each reading corresponds to a distinct structure: the eventive passive is encoded by an aspectual head taking a vP complement headed by vAGENT , the resultative passive is encoded by an aspectual head taking a vP complement headed by vFIENT (see below for details), and the stative occurs with an aspectual head, but no v 0 . I argue that the structure Embick (2004) proposes for the resultative passive is incorrect. In particular, Embick (2004) claims that the resultative passive is non-agentive: it occurs with vFIENT , but not vAGENT . I argue that the resultative passive is agentive, and must contain a vAGENT in its structure. I tentatively propose a structure and semantics that can properly capture the resultative passive reading, while noting issues for my analysis. Detils of Embick (2004)’s Account Embick motivates a distinction between two types of adjectival passives based on the fact that the resultative passive can occur with adverbial modifiers whereas the stative cannot (2), and the fact that the resultative passive cannot occur following a verb of creation whereas the stative can (3). (2) The carefully closed package. . . a. ‘The package that is in a state of having been carefully closed. . . ’ (= resultative) Michael Wilson 101 b. *‘The package that is in a state of closed-ness carefully. . . ’ (= stative) (3) This door was built closed. a. *‘This door was built in a state of having become closed.’ (= resultative) b. ‘This door was built in a state of closed-ness’ (= stative) The difference in (2) is explained by the resultative passive encoding an event, which is amenable to adverbial modifiers; the stative passive does not encode an event, and so cannot be modified by adverbials. The difference in (3) is explained by the fact that the verb build requires that the door did not undergo a prior closing event, whereas the resultative reading would require that it did undergo a prior closing event, leading to a contradiction. Embick (2004) cashes out this difference in terms of the following structure for the resultative passive (I omit his structures for the eventive and stative passives for reasons of space): AspP Asp vP DP v′ vFIENT Root Embick arrives at a structure that eschews vAGENT because of the fact that the resultative passive does not license a by-phrase denoting an agent, as the clearly agentive eventive passive does, as in (4). (4) The metal is hammered by John. a. *‘The metal is in a state of having been hammered by John.’ (= resultative) b. ‘John hammers the metal (habitually).’ (= stative) Embick defines vFIENT as taking a root as an argument, and describing an event of becoming the state described by the root. The aspectual head predicates the result state of this event of the DP. 102 Speaker Abstracts The Resultative is Agentive I argue against Embick: the resultative passive obligatorily encodes agentivity. For one thing, the sole specific diagnostic he gives for agentivity is not a solid diagnostic for agentivity. Consider (5) (from den Dikken 1995:119): (5) To John was given a car. There is no place in this sentence that a by-phrase can be inserted without resulting in ungrammaticality. But the unavailability of a by-phrase does not mean the sentence is non-agentive. For another thing, resultative passive readings are not available for not-agentive predicates, even in cases where such an event should be salient, as in (6): (6) (I work in a cookie factory, and have to verify that all the cookies coming down the line have the proper fresh-baked smell. Before I clock out for the day, I cannot say:) *The cookies are smelled. Yet a minimally different sentence containing an agentive predicate is licit in this context: (7) (Following the context in (6)): The cookies are sniffed. Examples involving similar “perceptual” predicates show the same contrast. Furthermore, Embick’s proposal seemingly predicts that unaccusative verbs should be able to license resultative passives, since vFIENT requires an internal argument, but does not require vAGENT . But unaccusatives do not license resultative passives: (8) *Eliza is appeared. (= ‘Eliza is in a state of having appeared.’) (Embick also proposes a structure for resultative secondary predicates, which creates an incorrect prediction that resultative secondary predicates should entail resultative passives. I omit the details here for reasons of space.) Toward the Structure and Interpretation of the Resultative Passive I propose the following tentative structure for the resultative passive, assuming it is agentive: Michael Wilson 103 AspP Asp vP (resultative) vAGENT VP V DP I reduce the unavailability of a by-phrase to the semantics of the resultative head, which existentially closes the agent argument introduced by vAGENT (P is the proposition designated by vP): J (resultative) K = λP<e,st> .λes .∃xe :∃ss :P(x)(e) ∧ result-state(e, s) Putting this together with vP, we have the following for the resultative of “hammer the metal”: λes .∃xe :∃ss :agent(e, x) ∧ hammer(e) ∧ patient(e, the metal) ∧ result-state(e, s) This seems to predict that the end-state may hold of any participant in an event, which is prima facie incorrect. However, I explore some tentative steps showing that this is possibly a correct prediction due to a related construction (the (all) V-ed out construction). I also discuss other possible solutions. Work remains, but evidence shows the resultative passive is agentive—we must model it as such. 104 The syntax and semantics of event measurement in Mandarin Anqi Zhang (University of Chicago) [email protected] Introduction The telicity of an incremental change verb depends on the referential properties of its internal arguments (Krifka 1989, 1992; Tenny 1994; Ramchand 1997; Piñon 2005 among others). However, the compositional semantics of various subtypes, such as consumption verbs, and degree achievements, have been mostly analyzed differently. For a unified account, Kennedy (2012) proposes to extend the scalar analysis of degree achievements to incremental consumption verbs, such that the direct object denotes a measure of change function, which is unexpected for the semantics of noun phrases. To validate Kennedy’s analysis, it is crucial to show that the incremental theme NP is associated with an event variable. In this paper, I argue that the incremental theme direct object in Mandarin can indeed have an event type of meaning, because it can be directly modified by a duration phrase or a frequency phrase syntactically. Semantics-Syntax Mismatch In Mandarin, a duration phrase (DrP) or a frequency phrase (FP) can occur after the verb and before the direct object in the form of ‘V + DrP/FP + DP’. As (1) and (2) show, both the duration phrase san xiaoshi ‘three hour’ and the frequency phrase san ci ‘three times’ can precede the direct object book ‘shu’, occupying a position that other adverbs cannot. Additionally, unlike the frequency phrase, the duration phrase can be optionally marked with the modifier head ‘head’, which indicates that the duration phrase is modifying the following noun (Huang 2009). (1) Ta du le san xiaoshi (de) shu S/he read prf three hour (mod) book ‘S/he read books for three hours’ (2) Ta du le san ci (de*) shu S/he read prf three times (mod) book ‘S/he read books three times’ Both the DrP and FP have been argued to form a single constituent with the direct object for the following reasons (Sybesma 1999, Liao 2014): (1) a possessive/modifier marker de can intervene between the DrP and the direct Anqi Zhang 105 object, (2) the DrP or the FP with the following direct object pattern like massifiers and classifiers respectively (Sybesma 1999), and (3) evidence from tonal sandhi also suggests the DrP or the FP forms a constituent with the following direct object. However, this single-constituent analysis raises a problem of semanticssyntax mismatch: a DrP or an FP semantically should modify a verb, and yet syntactically modifies the direct object (Huang et al. 2009). To resolve this mismatch, Huang et al. (2009) proposes that in the deep structure the DrP or the FP still modifies the verb and through movements they end up appearing to form a constituent with the following direct object. In this paper, I offer new evidence from constituency tests and parallelism with different measure phrases, and further propose an explicit syntactic structure for the single-constituent analysis Constituency Tests ‘DrP DO’ pass tests such as topicalization in (3) and cleft in (4) that a double-object construction, a constituent dervied from verb movemen(cf. Huang et al. 2009), does not. This provides ample evidence that the duration phrase is base-generated in front of the direct object. (3) a. san-xiaoshi shu wo kan-le. three-hour book I watch-prf. ‘I read books for three hours.’ b. *Yuehan hua wo gei-le. John flower I give-prf. Intended: ‘I gave flowers to John.’ (4) a. b. wo kan le de shi san-xiaoshi shu. I watch prf rel cop three-hour book. ‘What I did was reading books for three hours.’ wo gei-le de shi Yuehan hua. I give-prf rel cop John flower. Intended: ‘What I did was giving John flowers.’ Parallel to Different Measure Phrases Jiang(2009) shows that the modifier head de is optional for a monotonic measure phrase (cf. Schwarchild 2007) and infelicitous for a classifier construction in Mandarin. The distribution of de between the DrP/FP and the direct object is exactly parallel.These parallels are easily explainable if the DO is interpreted with an event type meaning. The 106 Speaker Abstracts DrP measures the size of an event and the FP counts the number of occurences of an event. (5) a. san bang (de) yingtao three pound (mod) cherry ‘three pounds of cherries’ b. san ge (de*) yingtao three CL (mod) cherry ‘three herries’ Analysis I propose that the direct object can be type-shifted to an event type of meaning in this construction. In the syntax, when assigned a [+Theme] feature, the direct object is type-shifted to an event type meaning. This explains the syntactic and semantics mismatch without incurring any movement. A noun with a [+Theme] feature can select for a DrP as a monotonic measure phrase or an FP as a numeral classifier at its specifier as in (6) and (7). The [+Theme] feature also restricts the occurrence of DrP/FP as a nominal specifier to the direct object. (6) (7) NP[+Theme] MP[+Mon] Num M[+Mon] three N[+Theme] NP[+Theme] N[+Theme] CLP book Num hour book CL three times The ‘DrP + de’ is simply a ModP, like other adjectives, such as in (8), but because de semantically requires a concrete measure phrase as its complement, ‘Num+Classifier’ is infelicitous with de. (8) (9) * N[+Theme] N[+Theme] ModP MP[+Mon] Num M[+Mon] three hour N[+Theme] Mod de book N[+Theme] ModP Mod CLP Num CL three times de book Anqi Zhang 107 With an event type of meaning, the duration phrase and frequency phrase can directly modifies the direct object. My proposal not only explains the syntactic-semantic mismatch straightforwardly, but also directly supports Kennedy(2012)’s analysis by showing that the direct object can be associated with an event variable. Different from Kennedy’s analysis for English, however, the event measurement is encoded in the DrP or the FP instead in Mandarin, because if the noun phrase encodes a specific measure function, the DrP or the FP would not be able to modify the noun phrase. (10) Jbook[+theme] K = λe∃x[Theme′ (e)(x) ∧ book(x)] Jthree-hourK = λe[Duration′ (e) = 3-hour] Jthree-timesK = λe[Count′ (e) = 3] Conclusion My paper argues that in Mandarin the incremental theme argument can have an event type of semantics, directly supporting Kennedy(2012)’s proposal by providing syntactic evidence that the direct object can be syntactically modified by event measurement phrases, such as a durational phrase or a frequency phrase. Specifically, I offer new evidence to prove that the duration phrase and the frequency phrase can form a single constituent with the following direct object and develop a concrete syntax-semantic interface for this construction. This paper has further crosslinguistic implications that the event measurement function may be located in slightly different places. 108 The at-issue status of appositive relative clauses: Evidence for a discourse-based approach Ema Živković (University of Niš) [email protected] Background Utterances in discourse convey contents which may differ in whether they express the speaker’s main point or not. At-issue content expresses the speaker’s central message, whereas content which is secondary with respect to the main point of the utterance is standardly labeled as not-at-issue. The focus of this paper is on a particular class of expressions typically considered to be not-at-issue – appositive relative clauses (henceforth ARCs). The at-issue status of ARCs has been a subject of recent debate mainly due to an observation that sentence-final ARCs, unlike sentence-medial ARCs, can be directly rejected (AnderBois et al., 2010; Syrett & Koev, 2015). This means that final ARCs can sometimes be at-issue, if it is assumed that only at-issue content is susceptible to direct rejections. A discourse-based approach developed by Jasinskaja (2016) accounts for this observation by making use of general discourse mechanisms such as the Right Frontier Constraint (Polanyi, 1988) and Question Under Discussion (QUD) memory stack model (Grosz & Sidner, 1986). According to this approach, the interpretation of the at-issue status of final ARCs depends on rhetorical relations by which they connect with their main clauses. Rhetorical relations can be coordinating and subordinating. Discourse units connected by coordinating relations, such as Narration and Contrast, are considered to be equal, which allows the discourse to progress in a left-to-right manner. On the other hand, subordinating relations, such as Explanation and Elaboration, lead to hierarchical structures and they do not “push the discourse forward”. This kind of analysis is applied to final ARCs to account for their at-issue status. When a final ARC is attached to the main clause by a subordinating relation, as in (1), its QUD will be on top of the memory stack, which means that the ARC content can be at-issue. However, due to the subordinating relation, there is also an option of popping this QUD off the stack, which would leave the main clause QUD on top of the stack, making the main clause content at-issue. This means that either the main clause or the final subordinate ARC can be at-issue by the end of processing the sentence. Ema Živković (1) 109 The managers assigned the project to Mark, who was the only experienced engineer at the meeting. On the other hand, when a final ARC is attached to the main clause by a coordinating relation, as in (2), the main clause QUD needs to be resolved and popped of the stack before introducing the QUD corresponding to the ARC. The main clause QUD cannot be pushed on top of the stack any more, which means that only the ARC content can be at-issue by the end of processing the sentence. (2) The managers assigned the project to Mark, who finalized it three months later. An important prediction that follows from the analysis above is that in an experimental setting final discourse-structurally coordinate ARCs should express more at-issue behavior than final discourse-structurally subordinate ARCs. The main purpose of the experiment conducted in the present paper was to test this prediction. Experiment The experimental stimuli involved sets of sentences which consisted of a main clause and a sentence-final ARC. A direct rejection test (Tonhauser, 2012) was used to measure at-issueness in isolated sentences. A forced choice task was designed where the participants had to choose between a rejection of the main clause and a rejection of the ARC, as in (3) below. (3) Officer James Wilson arrested Lisa, who broke out of jail two days later. a. No, he didn’t. (target: main clause) b. No, she didn’t. (target: ARC) The test sentences were generated by manipulating the rhetorical relations between main clauses and ARCs. Each sentence underwent four modifications for the purpose of manipulating four types of rhetorical relations: two coordinating ones (Narration and Contrast) and two subordinating ones (Explanation and Elaboration). In each of these modifications the main clause remained the same, while the relative clause changed depending on the type of rhetorical relation by which it was connected with the main clause. An example set is given in table 1. 64 test items in total were distributed among four lists following a 4 × 4 Latin square design. The test items were presented to the participants together with 18 fillers in pseudorandomized order, which means there were 34 110 Speaker Abstracts items per participant. The experiment was administered in the form of an online survey. 59 native speakers of English took part in the experiment. Table 1: Example of a set of test items Type of relation Sentence with the relative clause underlined 1. Narration The managers assigned the project to Mark, who finalized it three months later. 2. Contrast The managers assigned the project to Mark, who, however, failed to finish it before the deadline. 3. Explanation The managers assigned the project to Mark, who was the only experienced engineer at the meeting. 4. Elaboration The managers assigned the project to Mark, who joined the team in March. Results The dependent measure in the statistical analysis was the percentage of ‘No’ responses targeting the main clause or the ARC. The results indicated that when the relative clause was connected to the main clause via a coordinating relation, the participants chose to reject the ARC in most cases (73.1% for the ARC v. 26.9% for the main clause). On the other hand, when the relation was subordinating, the participants opted for the ARC rejection approximately half of the time (47.5% for the ARC v. 52.5% for the main clause). A chi-square test showed that there was a significant association between the participants’ responses and the type of relation between the main clause and the ARC. This means that whether the participants chose to reject the main clause or the ARC depended on whether the relation between the two clauses was coordinating or subordinating. Furthermore, no significant difference was found in the distribution of the participants’ responses between the two groups of coordinate ARCs, as well as between the two groups of subordinate ARCs Discussion The results of the conducted experiment confirmed the prediction that final coordinate ARCs should express more at-issue behavior than final subordinate ARCs, provided that being a target of a direct rejection is a diagnostic for being at-issue. It can further be noticed that coordinate ARCs were clearly the preferred targets of direct rejections when compared to main clauses. However, given that final coordinate ARCs are always predicted to be at-issue by the end of the sentence, the number of rejections corresponding to Ema Živković 111 them should have been even higher. The paper addresses this issue by looking at two particular test items which had a low percentage of ARC rejections. Conclusion The paper examines whether sentence-final ARCs follow a uniform pattern with respect to their at-issue status. The results of the conducted experiment provide empirical support for the discourse-based approach to the at-issue status of ARCs and at the same time pose a problem for other approaches to ARCs, such as the one developed by AnderBois et al. (2010), as well as the syntactic approach assumed by Syrett and Koev (2015). References AnderBois, S., Brasoveanu, A., & Henderson, R. (2010). Crossing the appositiveat-issue meaning boundary. Semantics and Linguistic Theory, 20, 328–346. Grosz, B. J., & Sidner, C. L. (1986). Attention, intentions and the structure of discourse. Computational Linguistics, 12(3), 175–204. Jasinskaja, K. (2016). Not at issue any more. Manuscript submitted for publication. Polanyi, L. (1988). A formal model of the structure of discourse. Journal of Pragmatics, 12, 601–638. Tonhauser, J. (2012). Diagnosing (not-)at-issue content. Semantics of UnderRepresented Languages of the Americas, 6, 239–254. Syrett, K., & Koev, T. (2015). Experimental evidence for the truth conditional contribution and shifting information status of appositives. Journal of Semantics, 32(3), 525–577. Poster Abstracts 115 How causal modal particles influence the mental representation of discourses – experimental evidence Beate Bergmann (Humboldt-Universität zu Berlin) [email protected] Causality is crucial to establish coherence in discourse. (Fletcher 1986; Trabasso & van den Broek 1985). Previous studies have shown that propositions connected via a causal relation, e.g., by the means of the causal discourse marker ‘because’, are better integrated into the mental representation (Caron et al. 1984, Trabasso & Van den Boreak 1985) as well as easier to process (Millis & Just 1994, Sanders & Noordman 2000) than propositions in a non-causal relation (e.g., additive relation ‘and’) or propositions without overt marking (Sanders & Noordman 2000). Myers et al. (1987) found evidence that best recall performances can be achieved for sentences with a moderate degree of causality, compared to high and low degrees of causality. German modal particles (MPs) are also described as linguistic devices that establish coherence relations between two propositions p and q (e.g., Döring 2016). The present study investigates the meaning and function of two particular German modal particles, which have been argued to mark causal relations, namely eben (‘obviously’) and auch (lit. ‘too’). Eben denotes obviousness of the state-of-affairs characterized by the proposition eben scopes over (= peben ) (e.g., Weydt 1969), and is in a causal relation with a preceding proposition q (Thurmair 1989, Karagjosova 2004). Auch marks q as expected/known, with pauch giving a reason for the expectedness of q (Thurmair 1989), whereas q and pauch being causally related (Karagjosova 2003). Experiment In a delayed recall experiment (N = 28), 32 discourses consisting of two sentences q and p standing in a weak causal relation (norming study, N = 25) were manipulated by different discourse markers such that p contained the causal conjunction denn (‘because’), or the MP eben, or auch, or it was presented with no overt discourse marker (control) (1). Items were presented auditorily due to the high frequency of auch as focus particle (homophone). 16 fillers were added. Discourses were presented in 4 blocks of 12 items followed by a recall phase after each block: participants read q and recalled p as literally as possible in spoken form. Data were coded for correct ‘literal’ (lit) and ‘semantically equivalent’ (sem) recall of each of the 5 lexical constituents in p (subj/verb/adv/obj1/obj2; score 116 Poster Abstracts 0–5). Statistical analysis (LMM) revealed that both lit and sem recall were better if p contained eben compared to control: participants recalled more constituents (lit: t = 2.04, p < .05; sem: t = 2.36, p < .05). Recall was marginally better with denn than without (lit: t = 1.89, p = .06; sem: t = 1.76, p = 0.07). Auch did not differ from control (lit: t = 0.07, p = 0.95; sem t = 0.06, p = 0.95). The results suggest that eben and denn as clear markers of causal relations impact the mental discourse representation such that they facilitate cued recall, whereas auch does not. The findings are in line with the proposed semantics of eben and auch such that eben is the clearer marker of causality. (e.g., Karagjosova 2004) Further support comes from a post-hoc analysis, in which discourse markers in responses were annotated. Dennresponses contained significantly more often the target marker denn than the other responses contained their target discourse marker. More interestingly, in eben-responses, a non-target, but causal discourse marker (e.g., denn, weil) occurred significantly more often than in other responses. The current experimental findings combine the descriptive work of previous research on a semantic and/or pragmatic level with experimental methods and results from a psycholinguistic point of view, which is to my knowledge a novel approach to the research field of modal particles. Thus, the findings shed new insights into the function and meaning of (causal) MPs. (1) Stefan hat sich schon wieder eine Erkältung eingefangen. ‘Stefan has caught a cold again.’ control Er ernährt sich täglich durch Fertiggerichte vom Supermarkt. denn Denn er ernährt sich täglich durch Fertiggerichte vom Supermarkt. eben Er ernährt sich eben täglich durch Fertiggerichte vom Supermarkt. auch Er ernährt sich auch täglich durch Fertiggerichte vom Supermarkt. ‘He eats convenience food from the supermarket every day (, that’s why!).’ Beate Bergmann 117 Figure 1: Mean number of recalled constituents with 95% CI (lit = literally correct, sem = semantically correct). Figure 2: Total number of responses with (non-)target discourse marker. Selected References Caron, J., Micko, H. C. & Thüring, M. (1988). Conjunctions and the recall of composite sentences. Journal of Memory and Language 27, 309–323. Karagjosova, E. (2004). The Meaning and Function of German Modal Particles. Doctoral dissertation. Universität des Saarlandes, Saarbrücken. Myers, J. L., Shinjo M. & Duffy, S. A. (1987). Degree of causal relatedness and memory. Journal of Memory and Language 26. Thurmair, M. (1989). Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Trabasso, T. & van den Broek, P. (1985). Causal thinking and the representation of narrative events. Journal of Memory and Language 24, 612–630. 118 What is PO doing? Spatial distribution and group forming in Serbian Ana Bosnić (University of Groningen and University of Nantes) [email protected] A series of truth-value judgment experiments were conducted to investigate two competing characterizations of the Serbian distributive marker PO: Does PO distribute over spatial/temporal units via being associated with a distributive operator (Balusu 2006), or are distributive readings obtained because of event plurality (sum of events)? (Knežević 2015). The latter view then claims PO is just a plurality marker and not a universal quantifier. Exhaustivity is a requirement for universal quantification (Zimmermann 2002). This means that the distributive Key (in Choe’s 1987 terminology), be it an overt NP or an event argument, has to be exhaustively distributed over by the distributive Share (an argument that is being distributed). We thus tested the competing claims of the nature of PO by checking whether event arguments must be exhausted. If PO is a universal quantifier, then it will require exhaustivity. If, however, it simply marks event pluralization, exhausting event arguments (spatial units) is not required. We used variations of sentences like (1) with four different scenarios (pictures). We contextually and visually made cages/caves as relevant spatial temporal units (the restrictor) over which the jumping-monkeys had to be distributed: (1) Skače PO jedan majmun. jump-3sg dist one-nom.m monkey-nom.m ‘One monkey jumps at different locations/times.’ The results suggest arguments are required to be exhausted on two crucial conditions (which had about 85% rejection from the speakers) (see Fig. 1 below). What is more, Korean results revealed the same pattern. We believe the results are generalizable to other cases of distributive markers (e.g., Japanese, Korean, Peruvian Quechua) and we expect the analyses of those markers will be relevant for our own. Ana Bosnić 119 Figure 1: Conditions in which two groups of monkeys are not jumping. The rejection of the given scenarios indicates the exhaustivity is required over the relevant groups of participants (monkeys) and not over the spatial units (cages). The generalization of PO is then as follows: “For every relevant group of participants (monkeys), one participant (monkey) has to jump. This holds when the contextual method of division is not salient enough or it is not explicitly given”. We therefore arrived at a group reading of PO, which is something that has been discussed for Korean (McKercher and Kim 1999). However, the claim that PO is simply a group-making device would not hold for PO because PO forces distributive readings and unmarked sentences force collective (Knežević 2015). Thus, group readings are just a special variant of readings that a distributive PO has. This was observed for the distributive marker -nka in Quechua and we at this point adopt the suggested analysis (Faller 2001). In our talk we will discuss further theoretical implications and possible analyses, as well as experimental methods for other instances with PO. References Balusu, R. (2006). Distributive reduplication in Telugu. Proceedings of NELS 36, 39–53. Choe, J.-W. (1987). Anti-Quantifiers and a Theory of Distributivity. Ph.D. dissertation, University of Massachusetts, Amherst. Faller, M. (2001). The problem of Quechua -nka-distributivity vs. group forming. Proceedings of SULA, 38–46. Knežević, N. (2015). Numerals and Distributivity in Serbian: at the syntaxsemantics-acquisition interface. Ph.D. Thesis, University of Nantes. McKercher, D. & Kim, Y. (1999). What does ssik in Korean really mean. Japanese/Korean Linguistics 9, 239–252. Zimmermann, M. (2002). Boys Buying Two Sausages Each: On the Syntax and Semantics of Distance-Distributivity. Ph.D. Thesis, University of Amsterdam. 120 Mechanics of variation – A case study: Prenominal genitives and derived possessives in the West Bohemian dialect of Czech Anya Chalupová (Palacký University Olomouc) [email protected] This paper explores Parrott and Nevins’s (2010) approach to intra-individual variation in the case of West Bohemian possessive nouns, as italicized in 4. In the authors’ formalized theory of intra-individual variation, the variant forms are phonological exponents of underspecified Vocabulary items which are inserted into syntactic nodes with impoverished feature specifications. In Czech, there are three constructions expressing possession within the boundaries of a noun phrase, according to prescriptive grammar: (1) Derived possessives a. Jan-in dům Jana-poss.fem house.nom ‘Jana’s house’ b. Petr-ův dům Petr-poss.masc house.nom ‘Peter’s house’ (2) Genitive case (postposed) a. dům Jany house.nom Jana.gen ‘the house of Jana’ (3) Possessive pronouns můj dům my.nom house.nom ‘my house’ The derived possessive endings -ův, -in are consided to be a special sort of adjectival ending in traditional treatements, even though the endings are morphologically distinct, and not found elsewhere in the language’s grammar. The West Bohemian dialect (WB) offers yet another construction, the prenominal genitive, which is not as widespread as the previous ones. It is considered ungrammatical by Czech native speakers outside of this dialect. (4) Genitive case (preposed) b. Jardovo babičky kůň a. sestry dům Jarda.poss grandma.gen horse.nom sister.gen house.nom ‘the grandma of Jarda’s horse’ ‘my sister’s house’ c. mojí sestry manžel d. táty kamarád my.gen sister.gen husband.nom dad.gen friend.nom ‘my sister’s husband’ ‘my dad’s friend’ Anya Chalupová 121 The WB prenominal genitive shows mixed characteristics of all the other possessive elements. In the paper, I argue that the prenominal possessive noun and prenominal genitive are in the state of intra-individual variation in terms of Nevins and Parrott (2010), i.e. within an I-language grammar of a WB speaker who uses both these constructions. The appearance of underspecified forms is problematic since the phi-feature deletion occurs variably instead of categorically through the application of Impoverishment rules. The impoverished object is the WB Prenominal genitive. Intra-individual variation arises from the properties of objects and operations in the postsyntactic morphological component. (5) Characteristics of the base nominal for possessive elements Construction Prenominal Prenominal Postnominal Prenominal Characteris- possessive genitive genitive possessive tics pronouns (WB only) pronoun Gender/ +FEM/+MASC; +ANIM +/−FEM/+/−MASC; Animacy +/−ANIM Number +SG +/−SG Word/phrase WORD WORD/ PHRASE PHRASE Supporting arguments for intra-individual variation of prenominal genitives and derived possessives in West Bohemian are summarized in the following points: (i) With the speakers who have both constructions, these constructions can be coordinated. (ii) Both constructions exhibit the same type of restrictions with the exception of the possibility “light (non-lexical) premodifier” in the case of derived possessive. (iii) They are linearized in the same way, i.e. they are in surface prenominal position. The Impoverishment of the syntactic terminal allows insertion of a less specific vocabulary item into it. This means that the items which are in competition must share some features. In case of the derived possessive, it has been argued 122 Poster Abstracts that it moves from postnominal genitive (Veselovská 1998; Kozánková 2015). The featural representation based on this hypothesis is the following: (6) a. Derived possessive: [+SG, +ANIM, +FEM/+MASC, +GEN, +POSS] b. Prenominal genitive: [+SG, +ANIM, +FEM/+MASC, +GEN] The +POSS feature is the reason of idiosyncratic behaviour of the element in Czech, i.e. the impossibility of any modification. The presence of the feature +POSS blocks +GEN agreement. There is a salient cross-linguistic argument supporting the hypothesis that there is an underlying genitive in the derived possessive, or in other words that the derived possessive bears +GEN. In Upper Sorbian, the derived possessive can be premodified. Otherwise it exhibits the same restrictions as the derived possessive in Czech. Any element premodifying the derived possessive always comes as genitive no matter what case is on the head noun of the whole phrase. The mechanism of intra-variation of prenominal genitives and derived possessives in West Bohemian is then the following: (i) The syntactic terminal gets Variably Impoverish for the feature [+POSS]. (ii) In case the Impoverishment takes place, the inserted element is the prenominal genitive, in other cases it is the derived possessive. (iii) This variation is not dependent on extra-linguistic context, e. g. situation. References Embick, David. 2007. Variation and morphosyntactic theory: Competition fractionated. Language and Linguistics Compass 2(1): 59–78. Embick, David & Noyer, Rolf. 2007. Distributed Morphology and the Syntax/Morphology Interface. In: The Oxford Handbook of Linguistic Interfaces, edited by G. Ramchand and C. Reiss, Oxford: Oxford University Press. Honeybone, Patrick. 2011. Variation and linguistic theory. In: Analysing variation in English, edited by W. Maguire and A. McMahon. Cambridge: Cambridge University Press, 151-177. Kozánková, Anna. 2015. Slavic prenominal possessive structure in a crosslinguistic perspective. In: Proceedings of the 5th Central European Conference in Linguistics for Postgraduate Students, edited by M. Janebová and L. Veselovská. Anya Chalupová 123 Křen, M., Cvrček, V., Čapka, T., Čermáková, A., Hnátková, M., Chlumská, L., Jelínek, T., Kováříková, D., Petkevič, V., Procházka, P., Skoumalová, H., Škrabal, M., Truneček, P., Vondřička, P., & Zasina, A.: SYN2015: reprezentativní korpus psané češtiny. Ústav Českého národního korpusu FF UK, Praha 2015. Dostupný z WWW: http://www.korpus.cz Nevins, Andrew & Parrott, Jeffrey K. 2010. Variable rules meet Impoverishment theory: Patterns of agreement leveling in English varieties, Lingua 120(5): 1135–1159. Veselovská, Ludmila. 1998. Possessive Movement in the Czech Nominal Phrase. Journal of Slavic Linguistics 6(2): 255–300. 124 Superiority, economy, and information structure Patrick D. Elliot (University College London) [email protected] We assess accounts of superiority effects in terms of economy, originally proposed by Golan (1993) and Reinhart (1998), and recently revived by Fox (2012). We find that economy-based accounts are faced with insurmountable obstacles, and instead argue for an account of superiority effects in terms of Information Structure, building on ideas developed in Kitagawa, Roehrs & Tomioka (2003). Superiority Multiple questions give rise to superiority effects (Chomsky 1973) ((1a) vs. (1b)). The standard intuition is best captured in Kuno & Robinson’s (1972: 474) generalization (2). (1) a. I know whoi ti bought whatj . b. *I know whatj whoi bought tj . (2) Wh crossing constraint: A wh word cannot be preposed over another wh. Economy Reinhart (1998: 46) (citing Lasnik & Saito 1992) notes that the contrast between (3) and (4) is problematic for any narrowly syntactic account of superiority; in both examples, movement of whj violates the generalization in (2). (3) *I know whatj whoi bought tj . (4) Whoh th knows whatj whoi bought tj ? Building on Golan (1993), Reinhart (1998) argues that, in order to understand this contrast, one should compare the superiority-obeying question (4) (repeated in (6)) to the minimally-distinct superiority-obeying question in (5). (5) Whoh th knows whoi ti bought whatj ? – Mary knows who bought caviar (6) Whoh th knows whatj whoi bought tj ? – Mary knows what John bought. As indicated by the focusation of the answers (Krifka 2004, a.o.), (5) and (6) ask different questions. Reinhart couches her analysis in terms of global Patrick D. Elliot 125 interface economy – the derivations underlying (5) and (6) only compete if they express the same global meaning. (1b) is unacceptable, because its minimallydiffering superiority-obeying competitor (1a) expresses the same meaning. (6) is acceptable because its minimally-differing superiority-obeying competitor (5) expresses a different global meaning. Pair-list vs. single-pair It has been know since at least Kuno (1982) that the presuppositions of multiple questions differ, depending on which wh overtly moves. Relevantly, multiple questions can have two readings: Pair-List (PL) and Single-Pair (PL) (Dayal 1996, 2002; Nicolae 2013; Kotek 2014). (5) and (6) are examples of the SP reading. (7) and (8) exemplify the PL reading. (7) Which waiterh th served which tablei ? – Jeff served table A, Troy served table B, and Abed served table C. (8) Which tablei did which waiterh serve ti ? – table A was served by Jeff, table B was served by Troy, and table C was served by Abed. The answers in (7) are sorted according to the waiters, and the answers in (8) are sorted according to the tables. This intuition is formalized by Dayal (1996, 2002) as two distinct presuppositions carried by multiple questions under the PL reading. (9) Domain exhaustivity: every member of the set quantified over by the overtly moved wh is paired with a member of the set quantified over by the insitu wh. (10) Pointwise uniqueness (functionhood): every member of the set quantified over by the overtly moved wh is paired with a unique of the set quantified over by the in-situ wh. Semantically sensitive shortest move Fox (2012) explores the predictions of Reinhart’s global interface economy account in light of this more fine-grained understanding of the semantics of questions: Prediction: Superiority-violating multiple questions should be acceptable, but only under the PL reading. This is because Dayal’s presuppositions only affect the PL reading. Fox argues that 126 Poster Abstracts this is correct on the basis of examples like the following (adapted from Kotek 2014): (11) Context: scientists have discovered a new planetary system consisting of just two stars. a. Whichh th revolves around whichi ? superiority obeying b. *Whichi does whichh revolve around ti ? superiority violating A PL reading of (11) is ruled out, since both whs have the same domain of quantification: {star1 , star2 }, and the revolve around relation is non-reflexive and antisymmetric. It follows that the presuppositions of the PL reading can never be satisfied. Global interface economy makes the right prediction here – superiority may not be violated, since the SP reading is the only one available. Response: we show that global interface economy cannot be what is responsible here. (12) Context: there are three linguists at the party – Ad, Hans, and Klaus. a. Whichh th admires whichi ? superiority obeying b. *Whichi does whichh admire ti ? superiority violating Global Interface Economy predicts (12b) to be acceptable under the PL reading. This is because (12b) should presuppose that for each linguist, there is a unique linguist who admires them, whereas (12a) presupposes that for each linguist, there is a unique linguist they admire. In both examples, the domains of quantification of the two whs were the same. Global interface economy predicts that if the domains of quantification are distinct, but the superiority-obeying and -violating interrogatives express equivalent question meanings, superiority-violations should lead to unacceptability. (13) Context: you attend a social event for married couples only, and encounter a group consisting of three men and three women. a. Which of these three menh th is married to which of these three womeni ? sup. obeying b. Which of these three womeni is which of these three menh married to ti ? sup. violating Patrick D. Elliot 127 Since the married to relation is symmetric and nonreflexive, and ∣the three men∣ = ∣the three women∣, there is no situation in which the presuppositions of one question could be satisfied, where the other isn’t – the interrogatives express equivalent questions. It is still possible to violate superiority in (13b). Analysis We reject global interface economy, and propose that constraints on superiority-violations should be accounted for in terms of an informationstructural asymmetry between the whs. Our generalization is as follows: (14) In a superiority-violating multiple question, whh and whi must have distinct domains of quantification. Building on Kitagawa, Roehrs & Tamioka (2003), we argue that when superiority is violated, the overtly moved wh is interpreted in a dedicated Contrastive Topic position (Constant 2014), and this imposes an information-structural asymmetry between the two whs, not present when superiority is obeyed. References Dayal, Veneeta. 1996. Locality in WH Quantification. Studies in Linguistics and Philosophy. Fox, Danny. 2012. The semantics of questions. Class notes, MIT seminar. Kuno, Susumo. 1982. The focus of the question and the focus of the answer. Papers from the parasession on nondeclarative sentences, Chicago Linguistics Society. Kuno, Susumo and Robinson, Jane J. 1972. Multiple wh questions. Linguistic Inquiry. Reinhart, Tanya. Wh-in-situ in the framework of the minimalist program. 1998. Natural Language Semantics. 128 The multifunctional morpheme /l´/ in Awing: A wh-focus consipracy e Henry Zamchang Fominyam (Universität Potsdam) [email protected] An overwhelming assumption in the generative framework is that wh-words and focalized elements share a common feature – [+focus]. Awing, like most Grassfields Bantu languages spoken in the North and Western regions of Cameroon have a morpheme that often shows up with wh-words and focalized elements. In these languages, such a morpheme is commonly labelled F(ocus) M(arker) FM (e.g., Biloa 2014, Fominyam 2012, among others). However, very little is known about the types of foci (i.e., new, contrastive or exhaustive). Moreover, the exact role of this morpheme that optionally occurs with wh-words is still unclear. The /l´/ morpheme (glossed here as LE) in Awing, in particular, functions as: a copula, an exhaustive focus marker, a conjunction (opposing the predication of two clauses) and a wh-‘operator’. The aim of this paper is to use the Awing language, in particular, and set the pace for a rethinking of focus marking in Grassfields Bantu, in general. First, it will be shown that the LE morpheme is used with wh-words only if there is a presupposition of other explicit elements. Although such contrast militates for a focus oriented analysis, the syntax of embedded questions in Awing, will reveal, however, that the LE morpheme can best be equated to a morphological wh-operator, in parallel to a wh- EPP/edge feature responsible for wh-movement in languages like English. The paper conclude with an analysis of preposed materials in Awing by showing that such materials can be best accounted for in a biclausal (cleft) structure. Using the different negation morphemes in Awing, among other diagnoses, I will show that the LE morpheme in sentence-initial position is a copula and as such, it bears no focus feature. e 129 Parametrizing intralinguistic variation: Case assignment strategies in Russian event nominalizations Anastasia Gerasimova (Lomonosov MSU, MSPU) [email protected] The idea of parameter is foundational in linguistics. Since the Theory of Principles and Parameters it has been common to see the differences between languages as the result of parameter realization (Chomsky & Lasnik 1993; Pesetsky 2003). This approach was preserved through the Minimalist framework, where formal features of lexical and functional heads serve as parameters predetermining the outcome of the derivation. Consequently, cross-linguistic morphosyntactic variety results from language-specific values of general parameters. However, variation commonly exists inside a single language and falls into two options of variance: either grammatical variants of one language are distributed among the speakers, or they coexist within the grammatical scope of one individual. The first type of variation prevails within speech communities and is dictated by a number of factors such as area, social class, gender, genre etc., e.g. was/were variation across English dialects and sociolects (Anderwald 2001). Another type occurs in the speech of one individual and is caused by the presence of several variants for one linguistic configuration. This talk addresses the issue of intralingual variation in the context of Russian event nominalizations (Alexiadou 2001; Grimshaw 1990). Russian event nominalizations belong to ergative-possessive type (Koptjevskaja-Tamm 2002). This means that the arguments of intransitives and internal arguments of transitive stems are marked with possessive case GEN, while external arguments of transitives are assigned INSTR. (1) a. b. c. padenie kursa rublja fall course.gen rouble.gen ‘fall of rouble course (weakening of rouble)’ vorchanie sosedei grumbling neighbors.gen ‘neighbors’ grumbling’ ispolnenie arii Shaljapinyim performance aria.gen Chaliapin.instr ‘performance of aria by Chaliapin’ 130 Poster Abstracts Nevertheless, the corpus data shows that instrumental case marking is not limited to prototypical transitive stems (contra Engelhardt & Trugman 1998; Rappaport 2002). The external argument of transitive nominalizations with lexically governed internal argument can be marked both GEN and INSTR. Two alternatives are possible, which means that external argument demonstrates differential case marking. (2) torgovlja evreev skotom trading Jews.gen cattle.instr ‘trading in cattle by Jews’ (Russian National Corpus) (3) torgovlja tserkov’yu nebesnoi blagodat’yu trading church.instr grace of God.instr ‘trading in grace of God by church’ (Russian National Corpus) As previously was shown in (Gerasimova, Lyutikova & Pereltsvaig 2016), the two existing case theories make distinct predictions about case marking for nominalizations whose internal argument is lexically governed. Inherent case theory (Woolford 2006) expects the external argument to be marked with INSTR, an inherent case, which is assigned independently from internal argument and which is associated with Agent θ-role. On the contrary, Dependent Case Theory (Marantz 2000) predicts that external argument is assigned GEN, an unmarked adnominal case, which appears under the lack of another caseless DP within a case competition domain. Hence we see that each theory can justify only one alternative. Consequently, it is essential to find out, which model of intralingual variation is represented in the case of Russian event nominalizations. If the ability to mark the external argument with INSTR is individual, then the two modalities of case assignment are distributed among speakers. Alternatively, if both strategies are equally available to any speaker, we should account for how the two case configurations can coexist in one individual grammar. In other words, the case assignment mechanisms have to allow the choice between GEN and INSTR. The problem is in particular interesting in the context of recent discussion of whether the coexistence of two modalities of case assignment is possible within one language (Baker & Vinokurova 2010; Levin & Preminger 2015). In this talk I will present new data that expands the knowledge about case assignment strategies in Russian event nominalizations. I conducted two linguistic experiments focusing on what cases speakers choose in the process Anastasia Gerasimova 131 of speech production and how they estimate grammaticality of cases when reading sentences. In survey A data was collected from 120 participants, who were asked to generate arguments of nominalizations assigning cases that sounded most natural to them. Then 78 respondents from the first experiment took part in survey B: they were asked to evaluate the grammaticality of event nominal constructions with external argument marked GEN or INSTR using the Likert scale (Likert 1932). Relying on the experimental data we can conclude that within the experiment A Russian native speakers are inconsistent in using INSTR. In accordance with the usage speakers can be grouped into two clusters depending on how frequently they used INSTR during the survey A. However, speakers’ evaluation of the INSTR acceptability is not consistent with the grouping based on actual usage in speech. Thus, the testee who constantly assign INSTR to external arguments of action nominal constructions may estimate sentences with INSTR much less preferable than with GEN and vice versa. The findings might prove that two strategies of case assignment to arguments of nominalization coexist within case grammar of one speaker. So the hypothesis that predicts in Russian two modalities of case assignment distributed among the speakers must be rejected. Therefore, the experimental study shows that there is a need for a new model that would take into account an eventual coexistence of two alternatives for case marking and explain speakers’ preferences when choosing the case of external argument. References Alexiadou, Artemis. 2001. Functional Structure in Nominals: Nominalization and Ergativity. Amsterdam: J. Benjamins Publishing. Anderwald, Lieselotte. 2001. Was/were-variation in non-standard British English today. English World-Wide 22.1: 1–21. Baker, Mark C. and Nadya Vinokurova. 2010. Two modalities of case assignment: Case in Sakha. Natural Language & Linguistic Theory 28.3: 593–642. Chomsky, Noam and Howard Lasnik. 1993. The theory of principles and parameters. Syntax: An international handbook of contemporary research 1: 506–569. Engelhardt, Miriam and Helen Trugman. 1998. D as a source of adnominal genitive in Russian. Proceedings of the Sixth Annual Workshop on formal approaches to Slavic linguistics. Ann Arbor: Michigan Slavic publications. 132 Poster Abstracts Gerasimova, Anastasia, Ekaterina Lyutikova and Asya Pereltsvaig. 2016. Case Marking in Russian Eventive Nominalizations: Inherent vs. Dependent Case Theory. Formal Approaches to Slavic Linguistics 25, 14 May, Cornell University, Ithaca, NY, USA. Grimshaw, Jane. 1990. Argument Structure. Cambridge: MIT Press. Koptjevskaja-Tamm, Maria. 2002. Nominalizations. London: Routledge. Levin, Theodore and Omer Preminger. 2015. Case in Sakha: are two modalities really necessary? Natural Language & Linguistic Theory 33.1: 231–250. Likert, Rensis. 1932. A technique for the measurement of attitudes. Archives of Psychology 140: 1–55. Marantz, Alec. 2000. Case and licensing. Arguments and case: Explaining Burzio’s generalization: 11–30. Pesetsky, David. 2003. Principles & Parameters Theory. Oxford International Encyclopedia of Linguistics. Rappaport, Gilbert C. 2002. Numeral phrases in Russian: A minimalist approach. Journal of Slavic linguistics 10: 327–340. Woolford, Ellen. 2006. Lexical case, inherent case, and argument structure. Linguistic Inquiry 37.1:2 111–130. 133 Labeling and two types of null operators in English Ryosuke Hattori (University of Connecticut) [email protected] Introduction Stowell (1986) shows (also noted by Chomsky 1973, Stowell 1987, Browning 1987, Cinque 1990; among others) that the null operator (Op) movement in tough constructions is clause-bounded (movement from subject/object position of an embedded finite clause is degraded) as in (1). (1) a. *Betsy is easy [ Opi [ PRO to expect [ ti fixed the car ]]] b. *John is easy [ Opi [ PRO to believe [ ti kissed Mary ]]] c. ??This car is hard [ Opi [ PRO to claim [ Betsy fixed ti ]]] d. ??That language is impossible [ Opi [ PRO to say [ Greg will learn ti ]]] (Stowell 1986:477) However, in other structures where Op movement involves, e.g. comparative/temporal clauses, the extraction of Op from the embedded finite clause is possible (Bresnan 1975, Larson 1990): (2) a. Mary read more books [ PP than [ CP Opi everyone thinks [ CP ti Tom believes [ CP ti that it is said [ ti that John read ti ]]]]] b. I saw Mary in New York [ PP before [ CP Opi she claimed [ CP ti that she would arrive ti ]]] (Interpretation: “prior to the time t that she alleged would be the time of her arrival”) The clause-bounded effect in (1) is explained in the previous works based on the Empty Category Principle under the proper binding relation (e.g. Chomsky 1981); however, this effect has not been discussed in the recent years. This paper, under the labeling theory (Chomsky 2013), explains the above difference between (1) and (2) by assuming that a Complex Null Operator is involved in English tough sentences, following Hicks (2009). Labeling and Freezing Chomsky (2013) claims that the algorithm for Labeling is basically stated in the following way: when a head and a phrase merge, the head projects as in (3a); when two phrases are merged, a shared feature of the two phrases is projected as in (3b) or if one of the phrases is a trace it gets ignored and the other phrase is projected as in (3c). 134 Poster Abstracts (3) a. X X YP b. f XP[f] YP[f] c. YP t YP Now, based on the labeling algorithm (Chomsky 2013), successive cyclic movement of a wh-phrase is forced since in an intermediate position where a whP and a CP are merged, creating {α whP, CP}, the syntactic object α cannot be labeled unless whP raises up further, its trace being ignored so that α is labeled as CP based on (3c). In some cases, a whP shares a feature with a CP and thus stops raising. For instance, this happens in an indirect question where a wh-phrase moves to merge with a CP in the complement position of a verb like wonder. (4) they wondered [α in which Texas city [β C [ JFK was assassinated ]]] (Chomsky 2013:45) Here, the α is of the form {α whP, CP}, but whP does not raise. Based on the algorithm in (3b), the most prominent feature of whP and of CP, namely the interrogative feature Q (Cable 2007, 2010; Narita, 2011), being shared, it projects as the label of α. Thus, the Freezing effect (Rizzi 2006, 2007) is now interpreted based on the labeling theory, i.e. an element undergoing A′ -movement gets frozen for further movement when it participates in feature-sharing for labeling. Explanation : I claim that Op-movement involved in tough constructions is an XP in terms of labeling operation as in (3b). Hicks (2009) claims that a null operator in tough constructions is a wh-phrase with more complex internal structure than is typically assumed, i.e. a complex DP with the internal DP as the tough subject as shown below. DP1 [iφ, uCase, iQ, uWH] (5) D NP N DP2 [iφ, uCase] Op John Based on this complex null operator (henceforth, CNO) analysis, when the CNO merges with the V as an object, the patient θ-role is assigned to the whole Ryosuke Hattori 135 complex DP1, and after the CNO merges with a CP, the inner DP2 is smuggled (Collins 2005a, b) into the matrix subject position without being assigned an accusative Case. (6) TP DP2j John Q DP1[Q] i . . . tj . . . CP[Q] . . . ti . . . What is important here, which Hicks does not discuss, is that assuming this CNO analysis, the CNO (=DP1) shares the Q feature with a CP when it is internally merged with it in (6). Thus, A′ -movement of the CNO gets frozen when it is merged with a finite clause and shared feature projects as the label, just as the case of indirect question. The ungrammaticality of the sentences in (1), therefore, can be explained based on the labeling theory in Chomsky (2013). Consequences Since there is no freezing effect in Op-movement in comparatives or temporal clauses as in (2), the above analysis of Op involved in tough constructions has the consequence that there is no feature-sharing in Op movement in comparative or temporal clauses. This paper pursues the possibility of the Op to be a head (=X) and thus projects as a label after movement as in (3a). Cecchetto and Donati (2015) argue that there are cases in which overt operators can project as a label after they move as a head to internally merge with a CP, e.g. in the free relative construction as shown below. (7) I read [ DP [ D what ] [ CP you read twhat . ]] (Cecchetto & Donati 2015:1) Here, the wh-word moves as a D head and provides a label when it merges with a CP, so that resulting structure what you read becomes a DP. Looking at null operators in this regard, the Op in comparative/temporal clause seems to project as a label in the same way, since crucially we get the same DP interpretation of a clause in such cases (e.g. “the amount in which John read books.” or “time of 136 Poster Abstracts Mary’s arrival”). In fact, a number of authors have claimed that comparative clauses should be analyzed as a kind of free relatives (e.g. Donati 1997). Selected References Bresnan, J. 1975. Comparative Deletion and Constraints on Transformations. Linguistic Analysis 1, 25–74. Cecchetto, C. & C. Donati. 2015. On (re)labelling. Linguistic Inquiry Monographs. Cambridge, Mass : The MIT Press. Chomsky, N. 2013. Problems of projection. Lingua 130, 33–49. Collins, C. 2005a. A smuggling approach to raising in English. Linguistic Inquiry 36, 289–298. Donati, C. 1997. Comparative Clauses as Free Relatives: a Raising Analisis. Probus 9, 145–166. Hicks, G. 2009. Tough-Constructions and their Derivation. Linguistic Inquiry 40, 535–566. Larson, R. 1990. Extraction and Multiple Selection in PP. The Linguistic Review 7, 169–182. Rizzi, L. 2006. On the Form of Chains: Criterial Positions and ECP Effects. in L. Cheng, N. Corver, eds, Wh Movement: Moving on. MIT Press, Cambridge, Mass., 97–134. Stowell, T. 1986. Null antecedents and proper government. Proceedings of NELS 16, 476-493. 137 Complementizer agreement in Busan Korean Hyunjung Lee (Sogang University) [email protected] Nutshell I investigate the phenomenon of complementizer agreement in Busan Korean (BK) interrogatives under the feature-checking system. Unlike Seoul Korean, various interrogative complementizers such as -ka, -ko, -na, and -no are observed. The four distinctive forms depends on (i) the categorial feature of the predicate and (ii) the type of question (polar vs. content). Overall, it is shown that the complementizer agreement of BK interrogatives is analyzed by general computational properties such as probe-goal Agree. Background BK, a dialect of Korean spoken in the southern tip of the Korean peninsula, has an unusual case of allomorphy on its interrogative complementizers. The complementizer encodes, among other things, whether the predicate is nominal or verbal. I have found only two other cases of allomorphy conditioned by lexical category in the literature before (Rezac, 2004, Wilson, 2014), thus the current study adds a vital typological component to agreement possibilities in the world’s languages. Data The basic paradigm for complementizer allomorphy in questions is described in (1) (So, 1984). (The complementizers are shown in boldface.) (1) a. b. c. d. Ni-ka chayk-ul ilk-n-a you-nom book-acc read-Kv -Vq ‘Are you reading a book?’ Ni-ka mwe-lul ilk-n-o you-nom what-acc read-Kv -Vwh ‘What are you reading?’ Ce salam-i Swumin-i-k-a that man-nom Swumin-cop-Kn -Vq ‘Is that woman Swumin?’ Ce salam-i nwu-Ø-k-o that man-nom who-cop-Kn -Vwh ‘Who is that woman?’ Observe that the consonant (hereafter K) varies with respect to the categorial feature of the predicate (Kv versus Kn ), and that the vowel (V) differ with respect 138 Poster Abstracts to the kind of question (polarity, Vq vs. content, Vwh). I thus propose that the complementizer is actually a sequence of two morphemes, K-V. Consider the following data, however, where unexpected forms are observed. In the first example Kv is found on a copular construction (rather than the expected Kn ). In the second example Kn is found on a verbal predicate (rather than the expected Kv ). (2) a. b. Ce salam-i Swumin-i-yess-n-a (*k-a) that man-nom Swumin-pst-cop-Kv -Vq (Kn -Vq) ‘Was the woman Swumin?’ Ni-ka chayk-ul ill-ul-ke-k-a (*n-a) you-nom book-acc read-irr-nmz-cop-Kn -Vq (Kv -Vq) ‘Will you read a book?’ Note that the future forms are built with the copula plus a nominalized form of the verbal predicate. Consider the following future form of the verbal predicate in (2b). (1) eat-irr-nmz-cop-Kn -Vq irr – irrealis; nmz – nominalizer Discussion The observation above is that the BK interrogative complementizer is a bimorphemic complex. The consonant (hereafter K) encodes interrogative Force and co-varies with the lexical category of the predicate (with the exceptions noted above) and the vowel (hereafter V) co-varies with the type of question: polarity versus content. The puzzling case of allomorphy is that found on the consonant. Given the cyclic nature of vocabulary insertion, higher morphemes should not be morphologically conditioned by morphemes lower on the tree. Thus, when lexical insertion takes place, K should not be able to see the categorial features on the predicate underneath it. This forces that conclusion that an Agree relation holds between K and the categorial feature of the predicate (See Rezac, 2004, for a similar phenomenon in Breton). K and V are present only in questions, and V indicates the kind of question (I capture this with the following lexical entries and the following Vocabulary Items (δ = categorial feature). Recall that K appears in questions only, so must be specified as [iQ]. V also appears in questions only, so is also specified [iQ]. V preferentially agrees with a wh-feature. However, if no wh-feature is found, Agree fails (in the sense of Preminger, 2014), and [uwh] is deleted. 139 Hyunjung Lee (2) K [iQ, uδ:] /k/ ↔ [Q, δ:n] /n/ ↔ [Q, δ:v] V [iQ, uwh] /a/ ↔ [Q, wh] /o/ ↔ [Q] I assume that K and V are distinct probes in the C layer on Int and Force, respectively (Rizzi, 1997, 2001), although the precise location is not vital. In the core cases K probes for the closest lexical category and finds either n or v, giving rise to the forms above. Here are the derivations for (1a, c), respectively. I assume the DP inside the copular construction contains the full range of DP-internal functional material, including nP. The [uwh] Probe on Force, failing to enter into an Agree relation, is simply deleted. Crucially, [uwh] is deleted before Spell-Out, while [uδ:n/v] survives at PF for Vocabulary Insertion. (3) a. [ForceP [IntP [TP [DP you ]i [vP ti [VP rice [V eat ]] v ] T ] Int[iQ,uδ:v] ] Force[iQ,uwh] ] b. [ForceP [IntP [TP [DP that man ]i [RP ti [DP Mincwu ] [R COP ]] T ] Int[iQ,uδ:n] ] Force[iQ,uwh] ] Unexpected Cases Recall that predicate noun constructions in the past tense give rise to verbal agreement on the complementizer. I argue that overt tense marking requires a (phonologically null), active v for T to be licensed. It well known in the traditional literature on Korean grammar that putting past tense on a non-active verb gives rise to active properties (Yeon & Brown, 2011). Thus, I assume an active (but non-agentive) v appears in past tense copular constructions. (See Harley, 2013 on the separation of the external argument introducing property from v.) Now, when K probes for a categorial feature, it finds v and selects the /n/ allomorph. Here is the derivation for (2a). (4) [ForceP [IntP [TP [DP that woman ]i [vP [RP ti [DP Swumin ] [R COP ] ] v ] TPST ] Int[iQ,uδ:v] ] Force[iQ,uwh] ] When K probes for a categorial feature it find the nominalizer ke (a reduced form of the noun kes ‘thing’). Thus, both unexpected cases fall out from general properties of Korean grammar—namely, that past tense requires an active v and that the future is constructed from a nominalized form of the verb. 140 Poster Abstracts Conclusion I have investigated BK interrogative complementizers which displays a typologically interesting and theoretically challenging form of agreement where allomorphy is based on categorial feature. I have proposed that it is derived by an Agree relation with the closest categorial feature. Selected References Harley, H. 2013. External arguments and the Mirror Principle: On the distinctness of Voice and v. Lingua 125:34–57. Preminger, O. 2014. Agreement and Its Failures. Cambridge, MA: MIT Press. Rezac, M. 2004. The EPP in Breton: An Unvalued Categorial Feature. In Triggers, eds. A. Breitbarth and H. Van Riemsdijk, 451–492. Berlin, New York: Mouton de Gruyter. Rizzi, L. 1997. The Fine Structure of the Left Periphery. In Elements of Grammar: A handbook in generative syntax, ed. L. Haegeman, 281–337. Dordrecht: Kluwer Academic Publishers. 141 The question system of Dagbani Samuel Alhassan Issah (Goethe-Universität Frankfurt am Main) [email protected] Introduction This paper examines the grammar of question phrases in Dagbani, a relatively little researched Gur language spoken in Northern Ghana. I focus on the inventory of the question phrase system, their internal structure, and grammatical features such as the semantic distinction between human/nonhuman question phrases, specification for number value and lexical ambiguity. Description Though Olawsky (1999) discusses Dagbani question phrases, there are some shortcomings such as inexhaustive list of the semantic domains of question phrases, failure to distinguish between simplex and complex question phrases and the lack of discussion on their grammatical properties. Contrary to Olawsky, I propose that the inventory of the question phrases of Dagbani is as given in table 1. I further argue that question phrases are independent elements in that they do not need to be bounded to any other syntactic category within the sentence. The proposal is made that the question phrases are independent syntactic elements chosen within the numeration. Table 1: Dagbani question phrases question phrases Semantic category Gloss bò díní yà wúlà ŋúní álá ŋún+NP NP+álá sáhá díní/bóǹdálí bòzùyú/wúlàzùyú nonhuman thing location manner/instrument human cost possession quantity time reason what which where how who how much whose how many when why I pursue an analysis according to which there is a distinction in the question 142 Poster Abstracts phrases of Dagbani based on the semantic features of the referents. Whereas ŋúní ‘who/whom’ substitutes for human referents, bò ‘what’ substitutes for non-human referent. This semantic distinction based on human/nonhuman referents is exemplified in (1). (1) a. b. ŋúní/*bò ń dá-Ø yílí máá? who/what foc buy.perf house def ‘WHO has bought the house?’ bò/*ŋúní ká bí-hí máá sá dá-rá? what/who foc child.pl def trm buy.imperf ‘WHAT were the children buying yesterday?’ Though ambiguities in question phrases have been argued to be typologically rare as articulated in Cysouw (2005), I argue that it is a property of the question phrase system Dagbani in the sense that wúlà ‘how’ has multiple interpretation as either a manner adverb or an instrument as explicated in (2). (2) a. b. c. Q: Wúlài ká á dí-Ø bìndírígù máá ti ? how foc 2sg eat.perf food def ‘HOW did you eat the food?’ A: Dírígù ká ń záŋ dí-Ø bìndírígù máá. [Instrument] spoon foc 1sg use eat.perf food def ‘A SPOON, I have used to eat the food.’ A: Bíεlábíεlá ká ń dí-Ø bìndírígù máá. [Manner] slowly foc 1sg eat.perf food def ‘SLOWLy, I ate the food.’ The question phrase wúlà ‘how’ in (3a) is ambiguous in the sense that it could be substituting for the instrument of the action as in dírígù ‘spoon’ (3b) or for a manner non-argument element, which is an adverb-like expression, bíεlab´εla ‘slowly’ as in (3c). Number marking is also identified as a grammatical property of question phrases in Dagbani. I postulate that number marking is a morphological phenomenon considering the fact that it is coded using the inflectional suffix -nímá, a plural marker in Dagbani. However, not all the question phrases inflect for number. Analysis In addition to the typological description of the Dagbani question system, I provide a theoretical analysis couched within Minimalism Samuel Alhassan Issah 143 (Chomsky 1995 et seq.). I therefore show that the question phrases that are specified for [+human], [−human] and [+thing] are sensitive to number, given that they alternate for plurality (ŋúní∼bánímá (who.sg∼who-pl), bò∼bònímá (what.sg∼what-pl), díní∼dínnímá (what.sg.thing∼what.thing-pl). I propose that these question phrases carry an uninterpretable number feature that undergoes feature checking with an interpretable number head also part of the question phrase. Of particular interest is the alternation of suppletion observable in ŋúní∼bánímá, which seems to represent a case of genuine root suppletion. Conclusion This paper gives a systematic analysis of an aspect of Dagbani grammar has not received any systematic linguistic attention. It therefore, adds to our knowledge of grammar of questions phrases in African languages, especially on an otherwise under described one. Selected References Chomsky, Noam (1995). The Minimalist Program. Cambridge. MIT Press. Cysouw, Michael. (2005). The typology of content interrogatives. Presentation at 6th meeting of the Association for Linguistic Typology 24 July 2005, Padang, Indonesia. Olawsky, Knut J. (1999). Aspects of Dagbani grammar, with special emphasis on phonology and morphology. Unpublished PhD dissertation. Munich: Lincom. 144 The presupposition of exclamatives at the syntax-semantics interface: Evidence from German and Japanese Katsumasa Ito (University of Tokyo, JSPS Research Fellow) [email protected] This paper proposes a mechanism of how presupposition occurs in German exclamatives with a complementizer (German dass-exclamatives) like (1a), which can also explain the nature of Japanese exclamatives with a nominalizer (Japanese koto-exclamatives) like (1b). German dass-exclamatives show interesting behavior regarding presupposition. Von Fintel’s (2004) wait-a-minute-test suggests that so schön ‘so beautiful’ in (1a) is not a part of presupposition (2). Following Zanuttini & Portner (2003), I assume that the presupposition trigger of exclamatives is a FACT-Op (factivity-operator). This operator also exists in clauses embedded by factive predicates. Haegeman (2014) suggests that the FACT-Op is base-generated at IP-field and moves to CP-field. Reinterpreting her idea under the framework of Pesetsky & Torrego (2007), I assume that the FACT-Op at IP-field has uFCT[val] (uninterpretable valued fact-feature), which agrees with iFCT[ ] (interpretable unvalued fact-feature) at C. After the Agree-operation, the uF[val] of FACT-Op is deleted and iF[ ] at C is valued. Note that iF[ ] at C is a probe and uF[val] of FACT-Op is the goal. Following Corver (1990) and Rett (2008), I further assume that the degree predicate so schön ‘so beautiful’ in (1a) constitutes DegP (degree phrase). The wh-phrase and adjective of exclamatives like (3) also constitute DegP, which agrees with C and moves to Spec-CP. In analogy with (3), we can assume that the DegP so schön also agrees with C in (1a), though there is no movement. The DegP has iDEG[val] (interpretable valued degree-feature) and the exclamative C has uDEG[ ] (uninterpretable unvalued degree-feature). In (1a), the DegP so schön is in-situ since the uDEG[ ] at C does not have EPP-property. In order to implement compositional semantics under the syntactic framework of Pesetsky & Torrego (2007), I propose following rules: (i) After Agree-operation between iF[ ] at X and uF[val] at Y, the meaning of Y is interpreted at X. (ii) After Agree-operation between uF[ ] at Head-XP and iF[val] at Y, the Katsumasa Ito 145 meaning of Y is interpreted at Spec-XP and Y leaves a trace in its original position. The rule (ii) is applied regardless of EPP-property. This rule enables to hold the LF-movement analysis by Heim & Kratzer (1998), which allows traces to be various types. In order to simplify the discussion, however, I ignore traces in this abstract. The denotation of FACT-Op is (4). (5) illustrates a simplified version of the denotation of the DegP so schön in (1a). The degree argument is type d and introduced by the context (cf. Rett 2008). Under these assumptions, the derivation of (1a) proceeds like in (6) and its semantic composition is (7). The result correctly predicts that so schön ‘so beautiful’ in (1a) is not a part of presupposition. Our theory also provides an account for intriguing nature of Japanese kotoexclamatives. Japanese koto-exclamatives (1b) become ungrammatical when the predicate is not an adjective (8), while German dass-exclamatives have no such restriction. When koto-exclamatives are embedded, the restriction on predicates disappears and the subject is marked (not as genitive but) as nominative (9). According to Miyagawa (2011), the genitive subjects in Japanese are due to the absence of C and its feature inheritance. The obligatory genitive subject in (1b) suggests that there is no C in Japanese koto-exclamatives. Without C, the FACT-Op cannot agree with iFCT[ ] and its uFCT[val] remains until the end of the derivation. This is the reason why (8) is ungrammatical. In (1b), on the other hand, there is no FACT-Op and the derivation converges. The wait-a-minute-test (10) indicates the absence of the FACT-Op in kotoexclamatives. In (10), only the existence of the subject referent is presupposed. This presupposition is triggered by the definite description sono ‘the’ and the effect of FACT-Op is not observed, which suggests that there is no FACT-Op. But why is the numeration for (8) without Fact-Op impossible? The answer may be because the pragmatics requires non-gradable predicates to be presupposed in order to interpret a clause as exclamative. It will be shown in this talk that this generalization seems to be valid cross-linguistically. (1) a. b. Dass die so schön getanzt hat! that she so beautifully danced has Sono keshiki-no/*-ga kireidatta koto! the scenery-gen/-nom beautiful.past nmlz 146 Poster Abstracts (2) Dass die so schön getanzt hat! a. —Hey, wait a minute. She didn’t dance. b. —#Hey, wait a minute. It was not so beautiful. (3) [ DegP How tall ] i he is ti ! (4) J FACT-Op K = λf<s,t> : f ∈ CG.f (5) (6) (7) J DegP K = J so schön K = λf<s,t> .λdd .λws .beautifully(f)(d) in w J FACT-Op K (J IP K) = λws :she danced ∈ CG.she danced in w J CP K = J DegP K {J FACT-Op K (J IP K)} λdd .λws :she danced ∈ CG.beautifully(she danced)(d) in w (8) *Lisa-no (kireini) odotta koto! Lisa-gen beautifully danced nmlz (9) Ken-wa [ Lisa-ga/*-no kireini odotta ] koto-ni Ken-top Lisa-nom/-gen beautifully danced nmlz-dat odoroita. was.surprised ‘Ken was surprised that Lisa danced beautifully.’ (10) Sono Furansuoohi-no kireidatta koto! the queen.of.France-gen beautiful.past nmlz a. —Hey, wait a minute. The queen of France doesn’t exist. b. —#Hey, wait a minute. She was not beautiful. Selected References Corver, N. (1990). The Syntax of Left Branch Extractions. Doctoral dissertation, Katholieke Universiteit Brabant. von Fintel, K. (2004). Would you believe it? The king of France is back! Presuppositions and truth-value intuitions. In: Reimer, M. & Bezuidenhout, A. (eds.), Descriptions and Beyond, 269–296. Oxford University Press. Haegeman, L. (2014). Locality and the distribution of main clause phenomena. In: Aboh, E. O., Guasti, M. T. & Roberts, I. (eds.), Locality, 186–222. Oxford University Press. Katsumasa Ito 147 Heim, I. & Kratzer, A. (1998). Semantics in generative grammar. Blackwell. Miyagawa, S. (2011). Genitive subjects in Altaic and specification of phase. Lingua 121(7), 1265–1282. Pesetsky, D. & Torrego, E. (2007). The syntax of valuation and the interpretability of features. In: Karimi, S., Samiian, V. & Wilkins, W. K. (eds.), Phrasal and clausal architecture: Syntactic derivation and interpretation, 262–294. Benjamins. Rett, J. (2008). A degree account of exclamatives. Proceedings of SALT, 18, 601–618. Zanuttini, R. & Portner, P. (2003). Exclamative clauses: At the syntax-semantics interface. Language 79(1), 39–81. 148 Pied-piping of adjuncts in Hungarian Júlia Keresztes (Pázmány Péter Catholic University) [email protected] In this paper I present the findings of a series of experiments studying piedpiping in Hungarian prenominal adjuncts. According to Horváth (2000) pied-piping is not acceptable in wh-movement and relativization in prenominal adjuncts while it is unrestricted in focus-movement. I suggest contra Horváth, that Hungarian pied-piping can violate the constraint on the position of the pied-piper inside the phrase in both wh- and focus-movement. Theoretical background on pied-piping It has been observed crosslinguistically that certain movement operations may move a large constituent. Ross (1967) coined the term pied-piping to refer to constructions in which a constituent properly contained in a bigger phrase is moved (as in (1b,c)) – the element targeted by the transformation rule is embedded inside a bigger phrase. (1) a. b. c. reports [which] the government prescribes the height of the lettering on. . . reports [the covers of which] the government prescribes the height of the lettering on. . . reports [the lettering on the covers of which] the government prescribes the height of. . . Pied-piping can be observed in questions (in (2)), relative clauses (in (3)) in English. (2) [Whose picture] did you buy yesterday? (3) This is the actress [whose picture] I bought yesterday. There are two main constraints on pied-piping that apply in most languages. These restrictions concern the position of the pied-piper (that is the element that triggers the movement) inside the phrase. Webelhuth (1992) claims that an element that is situated in the specifier or the head position can trigger the movement of the phrase, while complements and adjuncts cannot. Heck (2008) observes that the element that moves the phrase must be on the (left) edge of the phrase. Júlia Keresztes 149 Hungarian pied-piping and its problems For Hungarian, Horváth (2000, 2010) obverves that in prenominal adjuncts pied-piping is unrestricted in focus-movement, but it is ruled out in prenominal adjuncts in relativisation and wh-movement. This contrast is illustrated in (4), (5), and (6). First, in (4), the whole adjunct [BARACKPÁLINKÁT követelő vendégek] is moved. The pied-piper in this case is the focused accusative noun barackpálinkát ‘apricot brandy’. As (5) and (6) show pied-piping is not available in relativization and whmovement. (4) [BARACKPÁLINKÁT követelő vendégektől] fél a apricot.brandy.acc demanding guests fear.3sg the pincér waiter.nom ‘It is [ customers demanding APRICOT BRANDY ] that the waiter is afraid of.’ (5) *az ital, [amit követelő vendégektől] fél a the drink which.acc demanding guests fear.3sg the pincér waiter.nom intended: ‘the drink [ customers demanding which ] the waiter is afraid of. . . ’ (6) *[mit követelő vendégektől] fél a pincér? what.acc demanding guests fear.3sg the waiter.nom ‘[ Customers demanding what ] is the waiter afraid of? ’ (Horváth 2000) Horváth (2000 et seq.) accounts for this difference by proposing that an operator attaches to the focused phrase and that it is responsible for the acceptability of pied-piping. Horváth claims that there is no syntactic feature involved in focusmovement. She suggest that the exhaustive identification operator adjoined to the focused phrase triggers movement. She claims that wh-movement and relativization both involve strong syntactic features and that is the reason why wh-elements and relative pronouns cannot pied-pipe the phrase they are contained in. 150 Poster Abstracts New data and proposal New data, however, shows that Horváth’s generalizations and analysis do not capture the Hungarian data correctly. The data come from close to 100 native speakers of Hungarian who were asked to rate sentences involving focus-movement in (7), relativization in (8), and wh-movement in (9). The results show that pied-piping is as acceptable in embedded questions as it is in focus-movement contrary to the predictions of Horváth’s approach. (7) Péter furcsállta, hogy [pont a BORSODBÓL származó Peter surprised that precisely the Borsod.from originating gyerekeket] fogadják örökbe leggyakrabban. children adopt VM most.often ‘Peter was surprised that it is [precisely the children coming from Borsod] that people adopt most frequently.’ (8) Ede elmondta, hogy melyik az a megye, [ahonnan származó Ed said that which the the county where.from originating gyerekeket] szívesen örökbefogadnak. children gladly adopt ‘Ed told me which is the county [children coming from where] people like to adopt .’ (9) Laci megkérdezte, hogy [melyik országból származó gyerekeket] Leslie asked that which country originating children fogadják örökbe leggyakrabban. adopt VM most.often ‘Leslie asked [children coming from which country] people adopt most frequently.’ To account for these data, I propose an analysis bulding on Cable’s (2010). He suggests that a Q-operator attaches to the phrase containing the wh-element, which triggers the movement of the phrase. Cable’s analysis cannot derive the Hungarian pattern completely either. My proposal relies on the insight that pied-pipers in Hungarian do not need to Agree with the operator adjoined to the larger phrase containing the pied-piper. This means that the wh-feature bearing element does not need to Agree with the Q operator in Hungarian. Such an analysis conforms to the judgments collected in the surveys I conducted. The constraint on the position of the pied-piper inside the phrase can be violated in Hungarian in wh-movement and focus-movement. Júlia Keresztes 151 References Cable, Seth. 2010. The grammar of Q: Q-particles, wh-movement, and pied-piping. Oxford: Oxford University Press. Heck, Fabian 2008. On pied-piping – Wh-movement and beyond. Berlin, Boston: Mouton de Gruyter. Horvath, Júlia 2000. Interfaces vs. the Computational System in the Syntax of Focus. In: Hans Bennis, Martin Everaert, and Eric Reuland (eds.) Interface Strategies, 183–206. Amsterdam: Holland Academic Graphics. Ross, John Robert. 1967. Constraints on Variables in Syntax. Doctoral dissertation MIT. Webelhuth, Gert 1992. Principles and Parameters of Syntactic Saturation. Oxford: Oxford University Press. 152 Feature inheritance and the syntax of lexical VV compounds Ryoichiro Kobayashi (Sophia University/JSPS) [email protected] The main aim of this paper is to point out the correlation between productive lexical VV compounds and the lack of object φ-agreement in comparative perspectives. Japanese abounds in lexical VV compounds, in which neither heads are functional or auxiliarized, unlike syntactic compounds (Kageyama 1993). Lexical VV compounding is highly productive (Fukushima 2005) in (1). (1) a. tobi-orir jump-drop ‘jump off ’ b. d. tataki-tubus e. strike-smash ‘knock to pieces’ nomi-aruk drink-walk ‘go bar-hopping’ c. naguri-koros hit-kill ‘beat to death’ tabe-kuraber eat-compare ‘compare the taste’ I focus on such lexical VV, as defined in (2), assuming that they are formed in Syntax (Nishiyama 2008). (2) Lexical VV: Endocentric VV that satisfy the non-interruptibility principle of lexical integrity, which behave as indivisible X0 units in the phrasal syntax. (Kageyama 2016:278) Although they have sparked numerous studies in Japanese syntax/semantics, it has not been an-swered why some languages allow productive VV while others do not. The aim here is twofold: Observing 10 languages, I first demonstrate that a language can form productive VV if it lacks ob-ject-verb φ-agreement. Next, I propose a formal account of the syntax of lexical VV. Data Let us first observe languages with productive VV: Korean like Japanese has VV (3) with a phono-logical linker. In the same vein, Mongolian (4) (Khurelbat 1992), and Malayalam (5) (Krishnamurti 2003) have a wide variety of lexical VV compounds. What these languages have in common is that they lack overt subject/object φ-feature agreement altogether (Siewierska & Bakker 1996 a.o.). 153 Ryoichiro Kobayashi (3) a. ttwi-e-nem jump-go.over ‘jump over’ b. kwulm-e-cwuk starve-die ‘starve to death’ (4) a. dza:j-ögöx teach-give ‘show’ b. avc-irex take-bring ‘bring’ (5) pookuwaan-anuwadicc go-permit ‘permit to leave’ Next, languages with both overt subject/object φ-agreement: In Welsh (5), Estonian (6) and Swahili (7), the overt φ-morphemes appear on the verbs that undergo agreement with the arguments. As expected, these languages lack endocentric productive VV compounds (cf. Veldi 2010), as in (9–11). (6) Mae Steffan yn-dy-garu di. be-pres.3sg Steffan prog-2sg-love.inf 2sg.you ‘Steffan loves you.’ (Borsley et al. 2007) (7) Arvasin mehed vanemad olevat. thought.1sg man.3pl older.3pl to.be ‘I thought that the men were older.’ (8) Ume vi-ona vi-tabu? 2sg.subj 3pl-see 3pl-book ‘Have you seen the books?’ (9) Welsh *diod-gyrru drink-drive (10) Estonian (11) *hüppama-sõitma jump-ride (Ertlt 1999) Swahili *pig-ua hit-kill In order to further refine the observation, some languages with only subject φ-agreement are in order: Turkish (12), Bangla (13) and Igbo. If the lack of object φ-agreement correlates with productive VV compounds, then these languages should allow such VV. Indeed, the prediction is borne out that they are rich in endocentric lexical VV, as in (14) (Turkish: Kuribayashi 2006), (15) (Bangla: Soma 2003), and (16) (Igbo: Ihionu 1992). From these observations, we can draw a descriptive generalization (17). 154 Poster Abstracts (12) Ben bu makale-yi yavaş-yavaş oku-yacağ-ım 1sg this article-acc slowly-slowly read-fut-1sg ‘I will read the article slowly.’ (Şener & Takahashi 2010) (13) Ami æk-ţa boi-ke por-l-am ˙ 1sg one-cl book-acc read.past.1sg ‘I have seen two books.’ a. gelince-şaşır b. ağlayarak-gel come-surprise cry-come ‘come to surprise’ ‘walk crying’ (14) (15) (16) (17) a. kheTe-mor work.hard-die ‘work to death’ kú-wá ˙ hit-break ‘break by hitting’ b. (David 2015) ghumiye-poR sleep-fall ‘fall asleep’ Generalization: If the object and verb undergo φ-agreement, productive lexical VV is blocked. The crucial observation for my argument comes from Chicheŵa (Bresnan & Mchombo 1987), in which object φ-agreement is completely optional. The sentences (18) are both grammatical, but those in (19) with VV become ungrammatical only when the object and verb agree (Sam Mchombo p.c.). This naturally derives if object φ-agreement somehow blocks the endocentric productive VV, as in (17). The next question is, how we can analyze this correlation most naturally on the syntactic basis. (18) a. b. (19) a. b. Njûchi zi-ná-lúm-a alenje bees 3pl.su-bit hunters Njûchi zi-na-wá-lúm-a alenje (wá ‘3.pl’ obj-agree marker) ‘The bees bit the hunters.’ Ndi(*-wa)-ka-pemp-a pamanga 1sg(-3pl)-go-beg-asp maize.3pl ‘I go beg maize.’ Kati madzi banu (*ndi-)dza-man-e-ni ine if water your (1sg-)come-refuse-asp-imp 1sg ‘If it is your water, come to refuse me.’ 155 Ryoichiro Kobayashi Analysis In order to capture the descriptive generalization in (17), I assume Chomsky’s (2008) v-to-V Feature Inheritance (FI): φ-features on non-phasal V are inherited from the phasal v in (20), which occurs obligatory, not optionally unlike C-to-T (Chomsky 2008:149). I follow Nishiyama and Ogawa (2014) for the structure of lexical VV compounds, in which√they are √ base-generated via the set-merge. VV compounds are formed when V1 and V2 merge, which is immediately dominated by v in (21). (20) Obligatory v-V Inheritance: v[uφ] (21) a. √ V[uφ] IA[vφ] √ √ Merge( √ √V1 , V2 ) ={ V1 , V2 } b. Introduce v (order irrelevant) v √ V1 √ √ √ V2 V1 V2 Why is productive VV compounding blocked in languages with object agreement? I argue that FI is a one-to-one relation since it’s a prerequisite √ for the Agree, which is generally the one-to-one relation. After merging V1 and √ √ V2 , FI of [uφ] from v becomes ambiguous since there√are two Vs. It makes Minimal Search unable to unambiguously relate v to V, failing in FI; hence the derivation cannot proceed and the structure is deemed ungrammatical. Even if FI occurs, [uφ] becomes unable to probe into on the IA in (22), √ [vφ]√ since IA is no longer in the search domain (Note, V1 or V2 is the probe, not v). Such problem does not arise in languages without object φ-agreement, since there is no FI (23). (22) *NG: φ on v (23) OK: NO φ on v v[uφ] v IA[vφ] IA No φ/No Inheritance *Inheritance √ √ V2 V1 √ V1 √ V2 156 Poster Abstracts As for v without [uφ] in object-agreeing languages, I argue that v still has φwith default values; hence FI occurs. Such FI also fails to be one-to-one, since there √ are two Vs for only one v. To sum up, the FI and the mechanism in (22) correctly capture the descriptive generalization in (17). Abstract Case checking as a reflex of φ-agreement One may wonder why English does not allow productive VV, though it does not show any overt evidence on v for object agreement. Lieber (1992:80) has shown that NN, NA, AA, AN are productive in English, whereas root compounds containing V as one or both members are frozen expressions (e.g. stir-fry, sleep-walk) and barely productive (Lieber 2005:378). I argue that in English, objects undergo covert φ-agreement, along the lines of Chomsky (2000, 2008), who argues that v licenses accusative Case, entering into Agree with the IA. An object’s Case is checked only when it values a φ-complete set of checking features on v. Although there is no overt evidence on the verbs, DPs in English have φ-features; hence any Agree relation with it has to involve φ-features. Thus, I conclude that the accusative Case is checked as a reflex of object φ-feature agreement in English; hence the blocking mechanism in (22) correctly precludes productive lexical VV compounds. So far, I have assumed implicitly that Turkish, Bangla, and Igbo lack object agreement. If English has covert φ-agreement, what are the arguments that Japanese, Korean, Turkish, Bangla, etc. DO NOT have covert φ-agreement? My answer is that they employ a different mechanism for Case licensing from English. Note that these languages have a rich system of overt Case particles. Kuroda (1988) and Fukui & Takano (1998) argue that the existence of such a rich system correlates with the absence of Case-feature checking via φ-agreement. Specifically, I assume Fukui & Takano’s (1998:58) KP-analysis, which follows Kuroda’s (1988) that nominals can be licensed either by abstract Case (via φ-agreement) or by morphological case (case particles) as in (24). In this line of argument, I propose that languages like Turkish, Bangla, Japanese, Korean (case particles), and Igbo (tones) utilize the morpho(phono)logical case system for accusative licensing. Therefore, they make no recourse to object φ-agreement unlike English. (24) Arguments must be licensed by Case (abstract) or case (morphological). (Kuroda 1988:40) Ryoichiro Kobayashi 157 Consequence The proposal is compatible with the fact that Germanic languages like English, German and Dutch (Booji 1992) allow productive root compounds of other categories such as NN (Roeper et al. 2002). Verbs induce φ-agreement and the blocking mechanism in (22) precludes productive VV compounds. On the other hand, NN, NA, AA, or AN are not restricted in such a way; hence productive root compounding of these categories correctly receives a natural analysis as the syntactic merger of heads. In languages like German, expressions such as kennen lernen ‘get to know’, spazieren gehen ‘take a walk’ exist (Neef 2009), which seem to be VV compounds. They are separable verbs and indeed separated due to the verb-second phenomenon in (25), which violates the non-interrptibility principle of lexical integrity in the definition of the endocentric lexical VV compounds in (2). (25) a. b. Ich lerne keinen Mann kennen I learn no man know Ich (*kennen-)lerne keinen Mann ‘I get to know nobody.’ Conclusion The overall discussions pointed out the correlation between productive lexical VV compounds and the lack of object-verb agreement: VV can be formed in syntax if a language lacks objectverb agreement. The current study supports the Chomsky-Borer conjecture: Cross-linguistic variation is limited to differences in the properties of certain functional elements in the Lexicon (Fukui 2006). Selected References Fukui, N. & Y Takano. 1998. Symmetry in syntax: Merge and demerge. JEAL 7, 27–86. Lieber, R. 2005. English word-formation processes. In Handbook of Wordformation. 375–427. Nishiyama, K. & Y Ogawa. 2014. Auxiliation, atransitivity, and transitivity harmony in Japanese VV compounds. IIS 20, 71–101. 158 Verbal aspect: Scale or continuum? Elizaveta Kuzmenko (NRU Higher School of Economics, Mosco) [email protected] Introduction The thesis explores the grammatical category of aspect in the Russian language. Aspect can be described as “different ways of viewing the internal temporal constituency of a situation” (Comrie 1976:3). Historically, there are two aspectual values: imperfective and perfective. Imperfective aspect typically denotes states or activities that cannot, or have not yet been, completed, whereas perfective aspect implies that the state or action has been completed. This is a very complicated category and it is not very comprehensible for the speakers of languages that do not have aspect. There have been no attempts to describe the category of aspect as a continuum with two poles and a variety of cases between them. Additionally, there are few works that describe morphological properties from a statistical point of view (Kuznetsova 2013, Janda & Lyashevskaya 2011). In my research I focus on the integrity and continuity of this category. In other words, my research objective is to establish whether verbal aspect is a scalar or continuous value. There have been many works dedicated to irregularities in the Russian aspectual system, and comparisons with other Slavic languages such as Czech, where aspect is expressed optionally, show that verbal aspect is not a strict or discrete system; nevertheless, the distinction of two aspectual values is still generally recognized. Methods In my research I want to apply statistical and computational methods to the description of aspect in Russian. I work with a disambiguated part of the Russian National Corpus. This subcorpus consists of 5.4 million words. The analysis is performed on the basis of 7952 verb pairs. 1981 of them are prefixal pairs: in these pairs the base verb is imperfective, and perfectives are formed with the help of prefixes. These verb pairs are taken from the database of the “Exploring emptiness project”1 . An example of a verb from this list is the pair fotografirovat’ <i> – sfotografirovat’ <p> ‘take a photo’. Another 5971 pairs are extracted from the Zaliznyak’s Grammatical Dictionary of Russian, and in these pairs the simplex verb is perfective, and imperfectives are formed via suffixation. An example of such pair is ocharovat’ <p> – ocharovyvat’ <i> ‘to 1 http://emptyprefixes.uit.no 159 Elizaveta Kuzmenko Table 1: Sample grammatical profiles. lemma aspect mode nonpast past inf imper gerund bormotat’ ipf ‘mutter’ abs rel 39 0.203 108 11 0 0.563 0.057 0 32 0.167 vypolnit’ pf ‘complete’ abs rel 46 0.148 73 91 4 4 0.235 0.293 0.013 0.013 be charming’. Also this verb list contains pairs that are formed through stem alternation, like javlyat’ <i> – javit’ <p> ‘present, show’. From the corpus data there were extracted 549 513 occurrences of verbs from the two lists with their grammatical characteristics: tense (past, nonpast), form (finite verb, infinitive, gerund, participle), mood (imperative, treated as a separate from), voice (active, passive, only for participles). The further analysis is performed using this matrix. Analysis I set out to further explore the interdependencies that can be found in the interaction of tense and aspect. My goal is to find out which grammatical forms are most characteristic not for perfectives and imperfectives in general, but for smaller groups of verbs as well as for individual verbs. I extracted from the corpus 549 513 verb occurrences with all their grammatical characteristics. After that, I calculated how many occurrences of each grammatical form were found for each verb – these figures constitute grammatical profiles of these verbs. Both relative and absolute figures were estimated. Absolute figures in grammatical profiles are just raw numbers reflecting how many times each form of a verb was found in the corpus. Relative figures show what percentage each grammatical form constitutes with regard to all usage cases of a verb. Examples of grammatical profiles for two verbs with absolute and relative figures are presented in Table 1. As I have a matrix with all grammatical profiles, which can also serve as a 9-dimensional space characterizing the distributions of verbs, it would be appropriate to sort the verbs out according to their aspect, and analyze the position of each verb in the constructed space. I project the matrix with relative grammatical profiles into the two-dimensional space using the Principal Component Analysis (PCA). The resulting plot can be seen on Figure 1. 160 Poster Abstracts Figure 1: PCA on grammatical profiles. Conclusion There is indeed a notable distinction between perfective and imperfective verbs that was widely discussed in the literature. It can be seen visually and proven by the means of statistics. However, not all the verbs with the same aspectual value are alike. They can differ more or less from the verbs with the other aspect. This is reflected on the plot by their position: some verbs are put close to the verbs with another aspect, or even overlap with that aspect. Some other verbs can be found far away not only from the verbs from the opposite aspectual class, but also from other verbs with the same aspect. This gives us insights into the nature of the category of aspect: it is not a binary value with strict division, but a continuous category, and verbs manifest their aspectual affiliation in different ways. Also there is an overlap between perfective and imperfective verbs with regard to their grammatical profiles: some verbs have different aspectual values but similar distributions. Almost all biaspectual verbs fall in this category for obvious reasons – as they are used in both aspects, their grammatical profiles are not prototypical neither for imperfective verbs nor for perfective verbs. Verbs that are not biaspectual can also be found in the overlap area. These verbs are obsolete or they have some discourse peculiarities that motivate their appearance in the overlap area. Elizaveta Kuzmenko 161 Also there are some outliers whose distributions are very different from other verbs within that aspect. Mostly such verbs have some predominant grammatical form that is used in the majority of verb occurrences. For example, there are verbs that are used mostly as imperatives. There are also some verbs that are used as participles only. References Comrie, Bernard. 1976. Aspect: An introduction to the study of verbal aspect and related problems, vol. 2. Cambridge University Press. Janda, Laura A. & Olga Lyashevskaya. 2011. Aspectual pairs in the Russian national Corpus. Scando-Slavica 57, 201–215. Kuznetsova, Julia. 2013. Linguistic profiles: Correlations between form and meaning. Tromsø: Universitetet i Tromsø, PhD thesis. 162 Embedded polar responses: the case of English so A. Marlijn Meijer (Humboldt-Universität zu Berlin) [email protected] This study investigates the use of so in English embedded constructions like (1B). In this use, so competes with that-clauses and propositional anaphors it or that. Native speakers report that in embedded responses to questions such as (1A) the use of so or a that-clause is fine. In response to assertions such as (2A), the use of so is degraded. In line with these intuitions, Needham’s (2012) corpus study shows that antecedents of so mostly are questions. (1) A: Is John coming tonight? B: I believe {so ∣ he is ∣ ? it ∣ ? that }. (2) A: John is coming tonight. B: I believe {? so ∣ he is}. The distribution of so seems to be restricted in other ways too. It cannot combine with certain clause taking predicates (e.g. regret or resent), whereas other anaphors, such as it, do not have that restriction; see (3). (3) We should listen to him sometime. We wouldn’t regret {it ∣ * so}. In other respects, the distribution of so seems to be wider than that of it: so can occur in conditional clauses (see (4)) or together with sentential adverbs (see (5)), whereas it cannot. (4) A: Is John coming to the party? B: If {* it ∣ so}, he would have a great time. (5) Would we have felt the same if this had been our first stay? Likely {so ∣ *it}. This paper only focuses on so and provides an explanation for its preferred use in response to questions and its distribution. Kiparsky & Kiparsky (1971) have suggested that so can only occur with non-factive predicates, due to their syntax. Cushing (1972) argued that so can only combine with predicates that indicate that the speaker does not take a definite stance (e.g. think, but not claim). Cornish (1992) proposed that so is ‘intensional’ and does not presuppose that the referent is true. More recently, Needham (2012) suggested that so denotes the proposition corresponding to A. Marlijn Meijer 163 the current polar question under discussion (QUD). She furthermore argues that the subject of the so-utterance is not committed to this QUD. However, Bhatt’s (2010) finding, that so can occur with know in certain contexts, e.g. in (6)–(7), is problematic for these theories. Furthermore, Moulton has shown that so can also occur with predicates like claim (as shown in (8)) or admit. Examples (6)–(8) show that the subject associated with the so-utterance can in fact use so to refer to a proposition that s/he takes to be true and that s/he is committed to. Note that so can also combine with tell in I told you so responses, in which the speaker clearly commits to the referent of so. (6) It will rain tomorrow. I know so, because I checked the weather report. (7) Rooney knew he was special from a young age. And those who nurtured a talent that comes along rarely in any sport knew so, too.1 (8) A: Is he in the market for a bride? B: He claims so. I argue that so presupposes that its referent is still under discussion, i.e. ‘on the Table’ in Farkas and Bruce (2009), and thus is not part of the common ground (CG), at the time of the occurrence of the eventuality of the predicate that so combines with. For it, I follow Moulton (2015) in assuming that it refers to salient propositional content, crucially, without speculating on whether the associated proposition is part of the CG or not. Now, with Maximize Presupposition, it follows that so is preferred when its antecedent is not CG. This explains two properties of so mentioned above: (i) its context-sensitive distribution and, (ii) its preferred used in response to questions. Let’s look at (ii) first. Following Farkas & Bruce (2009:24), I assume that propositions denoting polar questions are not part of the CG until the ‘asker’ (implicitly) signals agreement with the answer (whereas assertions can be accepted ‘unsignalled’). Therefore, at the moment of answering a question, the answer nor the question is CG. (Note that we predict an I know so-response to be out here, since an affirmative response with high certainty can also be expressed by using yes.) Now, let’s look at (i): the context-sensitive distribution of so. On the present account, so can combine with verbs that are contextually fitting in the so-utterance and that, in the specific context and discourse, do not presuppose the acceptance of the referent into the CG. In (6), on this account, know can combine with so, because the previous assertion was not accepted into the 1 http://www.dailymail.co.uk/sport/article-389647/Walking-miracle.html, 2006, from Bhatt 2010 164 Poster Abstracts CG yet. If another speaker would affirm the previous assertion before the continuation, the know so-response becomes degraded, as the account predicts. This account makes one further major prediction: a so-utterance and its antecedent discourse move never both entail the referent of so, such that the referent is never CG at the utterance time of the so-utterance. A small corpus study, focusing on think so, was conducted to investigate this prediction. The results are shown in Table 1. Table 1: Corpus study (search: [think] so . ) – Corpus for Contemporary American English Antecedents questions (n = 70) assertions (n = 27) embedded prop. prop. below neg/modal 7/70 8 (+1?)/70 11/27 5 (+1?)/27 so-utterances involving negation involving modality 39/70 3/70 18/27 5/27 The corpus study shows that so is mostly used to respond to questions. When it is used to respond to assertions, most of its antecedents (16/27) were introduced below a propositional attitude verb (e.g. believe or think) or a modal item indicating possibility (e.g. might or maybe/perhaps) or negation. Furthermore, the responses to assertions, mostly (222 /27) contained such a modal item or a negation itself. The four remaining so-responses to assertions, without a hedge or negation, all involved a quantified subject, such as at least some students or even some of the most vocal and long time critics. This shows that – as predicted – in this small set of data the so-utterance and the antecedent assertion or question never both lead to the acceptance of the proposition referred to into the CG. Note that the definition above explicitly states that the CG-status of the proposition that so refers to, should be evaluated at the time corresponding to the thinking or believing event (depending on the predicate in question). Therefore, this account can explain I told you so-responses as well as occurrences with past tense like in (9). 2 One so-response involved negation and modality. Thus, a total of 22 so-responses involved negation and/or a modal. 165 A. Marlijn Meijer (9) A: Do you want coffee? B: Yes. A: I thought so. At the time of A’s response in (9), the referent of so is CG. However, at time of A’s thinking about B wanting coffee it was not. At that time, B was not committed to the proposition expressing that s/he wants coffee. This study thus suggests that past tense can take us back to the past CG of a discourse, giving us insight in commitments of the interlocutors of the discourse in question at that past time. In order to make this work, this account must assume that interlocutors keep track of the CG and the commitments speakers made or, crucially, did not make (yet) (cf. Krifka 2012, Stalnaker 2014). This study proposed a novel analysis for the use of English so in embedded responses and presented a corpus study that confirms the prediction made by the analysis. Therefore, it makes an important contribution to the literature on propositional anaphora and propositional attitudes. References Cornish, F. (1992). So Be It: The Discourse-Semantic Roles of So and It. Journal of Semantics 9(2), 163–178. Cushing, S. (1972). The semantics of sentence pronominalization. Foundations of Language, 186–208. Farkas, D. F. & Bruce, K. B. (2009). On reacting to assertions and polar questions. Journal of semantics, ffp010. Kiparsky, P. & Kiparsky, C. (1971). Fact. Linguistics Club, Indiana University. Moulton, K. (2015). CPs: Copies and compositionality. Linguistic Inquiry. Needham, S. M. (2012). Propositional Anaphora in English: The relationship between so and discourse (Doctoral dissertation, Carleton University Ottawa) 166 Hill Mari verbal constructions on ˆl’ˆ / ˆlˆn: Discontinuous past and beyond ee e e Daria Mordashova (Moscow State University) [email protected] This research deals with a “frozen” form ˆl’ˆ / ˆlˆn (3sg.pst / 3sg.pf forms of the verb ‘to be’) in combination with conjugated verbal forms in Hill Mari (Uralic). The data was collected mainly by elicitation during a field trip to the Gornomari district of Mari El, Russia in 2016. Hill Mari tense system includes no formal distinction between present and future (further on marked as .npst), but at the same time contains two synthetic past tense forms. One of them is marked as simple past (.pst), and the other one gets the label of perfect (.pf), although its semantics is much broader. The difference between the two past forms lies in the temporal distance: both of them equally function as (non-)hodiernal past markers, but only the perfect form can indicate remote past. The marker ˆl’ˆ / ˆlˆn combines with verbs in both present and past tense (only in perfect), which follow the rules of agreement in person and number with the subject. The grammar of Mari (Alhoniemi 1993) considers these forms as periphrastic Imperfect and periphrastic Perfect respectively, but does not pay much attention to their grammatical semantics. In typological papers (Plungian & van der Auwera 2006, Sitchinava 2013) devoted to similar verbal constructions (and including some Mari data) combinations ‘verb-pf + to be-3sg.pst’ are described as Pluperfect forms which can encode “discontinuous past” and ‘verb-npst + to be-3sg.pst’ constructions are considered analogous to Past Progressive. My research presents new data on what meanings can be expressed by constructions on ˆl’ˆ / ˆlˆn, contributing to the typology of Pluperfect and its polysemy, as well as to the sphere of “discontinuous past” (see Plungian & van der Auwera 2006) and the markers of this grammatical category. Firstly, verbal constructions on ˆl’ˆ / ˆlˆn cannot be described as prototypical Pluperfect, as they unwillingly encode taxis relations – perfect forms and non-finite clauses are the most widespread strategies. In (1) the action ‘to come’ (Event, according to Reichenbach 1947) happens before another action ‘to sit down’ in the past (Point of Reference), but the Event here cannot be expressed by the pluperfect. ee e e ee e e ee e e ee e e 167 Daria Mordashova (2) provides a possible context for the form on ˆl’ˆ / ˆlˆn, although perfect is more preferable here: ee e e (1) Pet’a tol-ˆn (*ˆl’ˆ/*ˆlˆn) – cilän š¨nz-¨n-nä Pet’a come-pf.3sg be-3sg.pst/be-3sg.pf all sit.down-pf-1pl čaj-ˆm jü-äš tea-acc drink-inf ‘After Pet’a had come, all the people sat down to drink tea.’ (2) Kˆnam mä tol-ˆn-na, keč¨ uže š¨nz-¨n when we come-pf-1pl sun already sit.down-pf.3sg (OKˆl’ˆ/OKˆlˆn) be-3sg.pst/be-3sg.pf ‘When we came, the sun had already gone down.’ e e e ee e e e e e e e e ee e e Still ˆl’ˆ / ˆlˆn can be viewed as “discontinuous past” markers, denoting the non-existence of consequent state or result of an action at the moment of speech (see also the notions of “cancelled”/“reversed” result (Squartini 1999) or “anti-resultative” (Plungian 2001)): ee e e ˆl’ˆ/ˆlˆn Van’a š¨rg¨-št¨ jam-ˆn vara t¨d¨m Van’a forest-in get.lost-pf.3sg be-3sg.pst/be-3sg.pf then he-acc mo-n-ˆt find-pf-3pl ‘Van’a got lost in the forest, but was found then.’ e e ee e e e e e e (3) e Secondly, ˆl’ˆ / ˆlˆn forms Past Habitual in combination with a verb-PF. This construction encodes an action frequently repeating in the past but no longer occurring: ee e e Mä š¨nz-en-nä ˆl’ˆ/ˆlˆn okn’a anzˆlnˆ dä we sit-pf-1pl be-3sg.pst/be-3sg.pf window in.front.of and anž-en-nä lˆm-ˆm look-pf-1pl snow-acc ‘(It often happened that) we sat in front of the window and looked at the snow. (But we don’t do this anymore.)’ e e ee e e e e (4) e Interestingly, it is possible to form a kind of Past Habitual with the help of 168 Poster Abstracts ‘verb-NPST + ˆl’ˆ / ˆlˆn’ construction, but in this case an action or a state may still take place: ee e e Pet’a izi-ž¨ godˆm každˆj keč¨-n lem-¨m kačk-eš Pet’a small-poss.3sg during every day-gen soup-acc eat-npst.3sg ˆl’ˆ/ˆlˆn be-3sg.pst/be-3sg.pf ‘Pet’a ate soup every day in his childhood (and maybe he still eats soup).’ e e e e e (5) ee e e However, probably the most challenging is the mirative (see DeLancey 1997) semantics of constructions on ˆl’ˆ / ˆlˆn. It is an expansion of semantics of a “discontinuous past” marker, which has not yet got enough attention in typology. It can appear in both present and past tense: ee e e (6) ˆl’ˆ/ˆlˆn Pet’a k¨z¨t zabor-ˆm čiält-ä Pet’a now fence-acc paint-npst.3sg be-3sg.pst/be-3sg.pf ‘Pet’a is painting the fence. (But I thought he should have been in some other place.)’ (7) M¨n tengeč¨ omˆn-ˆstˆ kašt-ˆn-am ˆl’ˆ/ˆlˆn I yesterday sleep-in walk-pf-1sg be-3sg.pst/be-3sg.pf ‘It turned out that I walked while sleeping yesterday. (But I don’t remember anything.)’ e ee e e ee e e e e e e e e e e Thus, my research provides new typological data on the polysemy of “discontinuous past” markers, which develop meanings (habitual and mirative) in the sphere of present as well. More detailed data from Hill Mari with supporting typological and theoretical background will be adduced in the talk. Abbreviations 1, 3 – 1, 3 person, acc – accusative, gen – genitive, in – inessive, inf – infinitive, npst – non-past tense, pf – perfect, pl – plural, poss – possessive, pst – simple past, sg – singular References Alhoniemi, A. 1993. Grammatik des Tscheremissischen (Mari): mit Texten und Glossar. Hamburg: Buske. DeLancey, S. 1997. Mirativity: The grammatical marking of unexpected information. Linguistic Typology 1.1, 33–52. Plungian, Vladimir A. 2001. Antirezul’tativ: do i posle rezul’tata [Antiresultative: before and after result]. Plungian, Vladimir A. (ed.), Issle- Daria Mordashova 169 dovanija po teorii grammatiki. I: Glagol’nye kategorii. Moscow: Russkie slovari, 50–88. Plungian, Vladimir A. & van der Auwera, J. 2006. Towards a typology of discontinuous past marking. Sprachtypologie und Universalienforschung – Language typology and universals 59,4: 317–349. Reichenbach, H. 1947. Elements of Symbolic Logic. N. Y.: McMillan. Sitchinava, Dmitriy V. 2013. Tipologija plyuskvamperfekta. Slavyanskij pluskvamperfekt. [Typology of Pluperfect. Slavic Pluperfect]. Moscow: AST-PRESS KNIGA. Squartini, Mario. 1999. On the semantics of the Pluperfect: Evidence from Germanic and Romance. Linguistic Typology 3.1, 51–89. 170 Phase, plurality and floating Takanobu Nakamura (Sophia University, Japan) [email protected] Overview This paper tries to unify event-based theories of plurality and derivation by phase; distributivity is obtained when [plural] feature is at the edge of D phase, cumulativity is obtained otherwise. Essentially, I follow Kratzer’s (2008) two pluralization mechanisms: lexical cumulativity and [plural] feature that is interpreted as *-operator. Her account of plurality directly leads to a system where element at the phase edge is interpreted at the next phase-level. This phase-based composition is independently motivated by the fact that event partitioning quantifiers (EPQs) in Japanese (Nakamura 2016) requires the semantic composition of argument nominals and verbal predicates to be done within different domains. An EPQ is exemplified below, being marked by the wavy line. I refer to a unit of a numeral e.g., san (3), and a classifier e.g., nin (humanCL ), as a numeral quantifier (NQ) and CL in gloss stands for a classifier. (1) San-nin-no gakusei-ga go-satsu-no hon-o 3-humanCL -gen student-nom 5-bookCL -gen book-acc hito-ri-ni-satsu ka-ta. 1-humanCL -2-bookCL buy-past ‘Each of the three students read two of the five books.’ Consequently, the proposed account shed light on how distributivity and floating quantifiers (FQs) is related: only QPs that can host [plural] feature can be a FQ. Analysis Kratzer (2008) argues that lexical predicates are plural in default (see also Krifka 1998, Landman 1996, 2000 among others), refuting the free insertion of *-operator, which is a generalized pluralizing operator. Instead of the free insertion of generalized *-operator, she claims that there are two [plural] features in a DP (see also Sauerland 2005) and phrasal cumulativity is obtained by high [plural] feature in plural DPs, pointing out that phrasal cumulativity arises only when the phrase is a sister of a plural DP (see also Schwarzschild 1993). Her DP structure is cited below. The low [plural] is for nouns and the high [plural] is for D. She argues that the higher [plural] appear even when a DP does not contain any plural NP. The fact that DPs like Spencer Takanobu Nakamura 171 and Webster can trigger verbal agreement and have distributivity shows that this is the case. Then, she argues that D’s [plural] feature is, unlike that of noun, uninterpretable within DPs and thus moves out of its DP. As a result, it functions as *-operator on the verbal phrase right below its DP. Its coverage is superior to the prediction of lexical cumulativity alone and the free insertion of generalized *-operator. [plural] D [plural] classifier N However, the reason why DPs allow apparent redundancy: two [plural] feature, is obscure. Moreover, why one of them is uninterpretable in DP but interpretable out of DP is in nature stipulative. Lastly, the detail of how [plural] feature of D moves out of its DP and related with verbal phrase is not explicit. To solve these shortcomings without spoiling its coverage, I suggest a phase-based modification: D is a phase head (see Chomsky 2000, 2001 among others) and the edge of D is interpreted at v phase, not at D phase. In syntactic literature, it is argued that syntactic structures are cyclically interpreted at each phase, which is defined by some functional heads (e.g., v, C). When a phase is activated, the complement of phase head is sent to semantic component and phonological component. Thus, in this conception, semantic composition is not done all at once, but cyclically phase-by-phase. In this explanation, we have no need to stipulate that [plural] feature of D is uninterpretable only within DPs. As [plural] feature of D is always higher than D, it is always at the edge of D phase. Thus, it is always interpreted out of the DP. If D’s [plural] feature is not uninterpretable, we have no need to postulate two distinct [plural] features. So, [plural] feature is unique and its interpretive distinction is attributed to the different structural position: above classifier or above D. Thus, this proposal overcomes the shortcomings of Kratzer’s (2008) account without spoiling its value. [plural] feature is unique and when it is above D, it is interpreted within verbal phrase because D is phase head. Three shortcomings are solved in this way. 172 Poster Abstracts Now, we can resort to Krazter’s (2008) pluralization mechanism without positing two distinct [plural] features. This modification allows the proposed account to explain Gil’s (1995) generalization that distributive universal quantification requires singularities. In the proposed account, distributivity is obtained when [plural] feature is at the edge of D phase. Since there is one [plural] feature within a DP, if [plural] feature is at the edge of D phase, no [plural] feature is interpreted within the DP. On this point, Krazter (2008) denies the existence of [singular] feature and argues that singularity is semantically specified by covert classifier in English. Thus, absence of [plural] feature directly leads to singularity of the DP. Hence, Gil’s (1995) generalization is straightforwardly predicted from the proposed account. This architecture is applicable for numeral classifier languages such as Japanese because she assumes that English has covert classifier (see also Krifka 1995 and Borer 2005). I assume that when there is an NQ, it hosts [plural] feature. Actually, it is independently claimed that there are multiple positions for NQ in Japanese DPs, e.g., Nomura (2013) argues that NQs can be either at the spec of #P or QP and distributivity is obtained by Q head, which is higher than # head. Here, I assume Japanese has a K head instead of D head at the top of nominal projection, though I leave it open whether Japanese lacks D (see Kuroda 1992 and Fukui 1995). K is realized as a case particle and can host [plural] feature. In this sense, K head corresponds to Q head in Nomura (2013). Then I assume that K in Japanese functions as a phase head (see Narita 2010 and Bošcović 2014 among others, for discussion that the complement of K is transfer domain). The notion of phase-based composition is independently motivated by the observation on EPQs in Japanese, which is illustrated in Nakamura (2016). (2) San-nin-no gakusei-ga go-satsu-no hon-o 3-humanCL -gen student-nom 5-bookCL -gen book-acc hito-ri-ni-satsu yon-da. 1-humanCL -2-bookCL read-past ‘Each of the three students read two of the seven books.’ As Nakamura (2016) discuss, plurality of agent and theme that are associated with an event is specified by the EPQ; every event has an agent comprising one human and a theme comprising two books. Importantly, interpretation of argument nominals is independent of its verb and its thematic predicates. Espe- Takanobu Nakamura 173 cially, the plurality of the internal argument must be specified independently of that of theme. I assume that Agent (e, x) is introduced by v head. (4) is a derivation before v is introduced. (3) VP: λe<st> .λy<e> .[*read(e, y) & 2(y)] / IA: λz<e> .[*book(z) & 5(z)] Here, plural entity z and y must be pluralized in different number specification. This seemingly strange treatment is, however, necessary to account for the scopeless reading of IA. Then following Oh (2001), Nakamura (2016) argues that EPQs are “distributive polarity item”, which must be within the scope of *-operator in whatever syntactic reason, just like NPIs which must be within the scope of a negator. (4) vP: *λe<st> .λx<e> .[*read(e, y) & 2(y) & Agent(e, x)] / EA: λx<e> .[*student(x) & 3(x)] IA is also distant from the scope domain of *-operator on vP. In this way, the interpretation of EPQs are properly described. What is odd so far is that (i) plural entity z and y comprise different number of atoms, (ii) IO must not be within the scope of any *-operator other than that of nominal predicate, which is pluralized along with lexical cumulativity. However, in the proposed account, these are not strange at all. (i) is natural because every KP has two distinct positions for [plural] feature, i.e., NQs, and these positions are independent of each other. (ii) is also natural because [plural] feature with go-satsu (5 books) is at the lower [plural] feature position and interpreted within KP, not at vP. For the detail of this composition, I propose that phase heads in nominal domain, e.g., D, K, denote a choice function of type <e, e> and every complement of nominal phase denotes type <e> entity, by the application of existential closure at the phase level. So, EPQ’s oddity is explained under the proposed theory, without serious complication. Consequence It has been observed that floated quantifiers (FQs) allow only a distributive reading (e.g., Junker 1995, Nakanishi 2007). Nakanishi (2007) argues that FQs are adverb and quantify over events. However, this account obscures the reason why distributive reading is also available with non-floated QPs. My proposal can provide unified account for it: distributivity is obtained only when [plural] feature is interpreted out of D(or K)P. Thus, we can maintain the insight that distributivity is attributed to plurality of thematic role and event and to do this, it is enough that [plural] feature is at the edge of nominal 174 Poster Abstracts phase. Note that this proposal is compatible both the notion of FQs as adverbs and FQs as remnants of DP movement. Moreover, this account has potential for answering why only a specific class of quantifiers, in particular, universal quantifiers allow floating in English, French and so on. If FQs allow only a distributive reading, we can obtain a generalization that only distributive universal quantifiers allow floating. Coupled with an account for Gil’s (1995) generalization, I suggest that only QPs that can host [plural] feature can be a FQ. This can support for the claim that FQs are pronounced *-operator. Selected References Kratzer, Angelika. 2008. On the plurality of verbs. In Event structures in linguistic form and interpretation. Berlin: Mouton de Gruyter. Nakamura, Takanobu. 2016. Events, Distributivity and Numerals: Evidence from Event Partitioning Quantifiers in Japanese. In Proceedings of the SICOGG 18th. Nomura, Junya. 2013. DP-internal Movements and the Semantics of the Floating Quantifiers in Japanese. In Proceedings of the WAFL 8th. Bošković, Željko. 2014. Now I’m a phase, now I’m not a phase: On the variability of phases with extraction and ellipsis. LI 45. 175 Clausal pied-piping in Basque wh-questions and syntactic optionality Louise Raynaud (Georg-August-Universität Göttingen) [email protected] Introduction Basque embedded wh-questions exhibit apparent optionality between long-distance extraction of the wh-word and clausal pied-piping. This paper attempts to account for this pattern of optionality found in Basque in a way that addresses the issue of syntactic optionality. The central theoretical claim here is that Cable’s QP-based analysis of pied-piping is compatible with a Minimalist approach to optionality of the sort pursued in Biberauer & Richards (2006). Data There is in Basque an uncommon alternative to long-distance extraction (1) in wh-questions, namely clausal pied-piping (henceforth CPP) (2). (1) Nork esan du Jonek [ edan duela ura ]? who.erg say aux Jon.erg drink aux.comp water.abs (2) [ Nork edan duela ura ] esan du Jonek? who.erg drink aux.comp water.abs say aux Jon.erg ‘Who did John say drank water?’ (Duguine & Irurtzun 2014:2) These constructions appear to be optional with respect to one another. They can freely alternate in a number of syntactic contexts and display no difference in meaning or interpretation (e.g. in scope or presupposition, cf Arregi 2003). There are some instances where one option is not available and the other prevails, like with adjuncts or relative islands that constitute independent syntactic constraints to movement. But it is not the case that CPP is a default option that occurs if and only if the pied-piped element cannot be extracted from the clause. CPP constructions in Basque are thus able to optionally alternate with their wh-movement counterparts, a phenomenon that has also been observed of Quechua (Cable 2010) and Bangla (Simpson & Bhattacharya 2006). The Puzzle Although true optionality could be defined as ‘semantically vacuous alternations in surface order’ (Biberauer & Richards 2006:35), its syntactic and conceptual implications remain under discussion as it poses a theoretical problem in the context of the Minimalist Program (MP). True 176 Poster Abstracts optionality, of the sort that is observed in the above examples, appears to be contrary to economy principles of the MP, namely Last Resort (LR) and Full Interpretation (FI), which regulate operations. This is the so-called Fox/Reinhart intuition on optionality, captured by Chomsky (2001:34) as a ‘general economy principle’: “An optional rule can apply only when necessary to yield a new outcome”. Furthermore, pied-piping poses a problem for the received theory of wh-movement, since it does not match the generalization that only elements bearing a wh-feature can move (Heck 2009). How can the pattern of optionality found in Basque be accounted for in terms that are compatible with economy principles of the MP? A Q-based Analysis (Cable 2010) Our data can be successfully accounted for in Cable’s (2010) framework. Cable proposes a novel account of wh-movement in which the target of wh-movement is a Q-particle that takes as its complement the wh-phrase and must agree with CQ . Pied-piping constructions are cases where the targeted Q-particle is not directly structurally adjacent to the whword. In the case of Basque CPP, the QP takes the CP as its complement (3). This account has two advantages. Firstly, it accounts for restrictions on CPP in embedded wh-questions that we find in Basque. Basque classifies as what Cable calls a Q/wh-agreement language, where the wh-word and the Q-particle must agree. Thus in virtue of the Phase Impenetrability Condition, in order to Agree the wh-word must be in the same phase as the Q-particle and to do so must move to the left edge of the clause. This accounts for the impossibility in Basque for an embedded CP to be pied-piped by wh-words internal to it (5) – the wh-word has to move to the left edge of the clause (4). CP1 (3) C′ QP CP2 wh-word . . . (4) Q CQ TP tQP [ CP Nor1 [ IP joango dela t1 ] ] 2 esan du Jonek t2 ? who go aux said aux John 177 Louise Raynaud (5) *[ CP [ IP Joango dela nor ] ] 2 esan du Jonek t2 ? go aux who said aux John ‘Who did John say will go?’ (Cable 2010:154) Furthermore, the distribution of CPP is not totally unrestricted, as mentioned earlier. Cable’s framework allows us to account for these cases in which longdistance extraction of the wh-word is ungrammatical because of an independent syntactic constraint, e.g. an adjunct island (6) and CPP is the constrained alternative (7). (6) *Zer izutu zen erregea entzun zenuenean? what frightened aux king hear aux.comp.when (7) %Zer entzun zenuenean izutu zen erregea? what hear aux.comp.when frightened aux king ‘What did the king become frightened when he heard?’ Last but not least, theoretically nothing in Cable’s framework requires Q to be as close as possible to the wh-word it is associated with. So in the same conditions where there are no barriers and no intervention effects, CPP and wh-extraction can theoretically occur in free variation. Cable’s account, unlike others like Heck’s (2009) that could potentially account for the other properties above, has the advantage of predicting cases in which both wh-extraction and CPP are viable options and can potentially alternate freely. Moreover, we find that this analysis is compatible with a global approach to syntactic optionality such as Biberauer & Richards’ (2006). Syntactic Optionality Biberauer and Richards (2006) legitimate optional variation on the basis that it is actually admitted on system-internal grounds in Minimalism. Whereas optionality is excluded from the functional motivation of movement, it is not with respect to how a given feature is formally satisfied. In other words, sometimes “the grammar doesn’t mind” as long as formal requirements are fulfilled, and then true optionality arises. In absence of grounds to privilege one option over the other, it can be more efficient to leave the choice open. Efficiency in the syntax is determined in terms of cost. Keeping in mind the analysis of Basque data outlined above, the CPP option is no more costly than wh-extraction: it involves the same number of steps (one movement of the QP and the elements it contains to Spec,CP), the Agree operation is just as local and the obligatory feature +Q triggering movement is equivalently 178 Poster Abstracts satisfied. As in both constructions Move is motivated by an obligatory feature that is already present in the derivation no matter what, they need not yield for a new interpretation, and can result in a semantically vacuous alternation. This approach to optionality that relies on the equal satisfaction of formal and featural requirements rather than on the identity of the derivations/lexical array allows us to account for the Basque data in a satisfying way. The Interpretation Problem While the above accounts for the possibility of a semantically vacuous alternation, one might expect that there might be semantic consequences associated with which element the Q-particle merges with. The absence of difference in interpretation thus needs to be accounted for. One solution that is compatible with our account of CPP in Cable’s terms is proposed by Arregi (2003). He claims that the pied-piped complement reconstructs at LF in its base position, resulting in an LF structure identical to the long-distance extraction construction such as [CP1 Wh1 . . . [CP2 . . . t1 . . . ]]. This analysis in terms of covert movement is moreover supported by the fact that whereas long-distance movement is allowed across negation (9), CPP is not (8), the reason for that being that negation blocks reconstruction of pied-piped material (Beck 1996). (8) *Nork edan duela ardoa ez du esan Mirenek? who.erg drink aux.comp wine.abs neg aux say Miren.erg (9) Nork ez du esan Mirenek edan duela ardoa? who.erg neg aux say Miren.erg drink aux.comp wine.abs ‘Who didn’t Miren say drinks wine?’ 179 Standards of comparison and the case of Spanish “que – de alternation” Laura Vela-Plo (University of the Basque Country (UPV/EHU)) [email protected] The syntactic and semantic properties of standards of comparison in inequality comparative sentences are still under discussion. In this paper, we analyse the syntactic complexity and semantic nature of standards of comparison in Spanish and other Romance languages, in order to offer a more extensive analysis of the cross-linguistic properties and variation of these components of comparative constructions. With this in mind, we first examine Romance standards (e.g. Spanish que – de or Italian che – di alternation) and then reconsider the syntactic and semantic formalisation of standards of comparison cross-linguistically to accommodate the observed data. Finally, the analysis of comparatives in several typologically different languages will support a parametric classification that is grounded on the idea that a clausal (CP) complement with wh-movement is not the unique strategy employed to express degree comparison in natural languages. Previous Accounts There are cross-linguistic differences regarding the syntactic structure of standards of comparison. Languages like Japanese have been argued to allow only phrasal structures in the position of standard of comparison (a DP, which can be modified by a relative clause (RC), cf. Shimoyama 2012); whereas both phrasal and clausal structures are available in languages like English (cf. Kennedy 2007). (1) a. John is smarter [than [DP Bill]]/[than [DP him]] phrasal standard b. John is smarter [than [CP Bill is]] clausal standard (2) This mountain is higher [than [DegP 1200 meters]] phrasal standard With respect to the semantic type of standards of comparison, Kennedy (2007) proposes that the comparative marker (-er in English) can select a standard that expresses either individual or degree comparison. On the one hand, individual denoting standards express orderings between arbitrary individuals, and they are therefore of type e in the semantics. On the other hand, degree denoting standards express orderings between individuals and arbitrary degrees and are already of type d in the semantics. There is cross-linguistic variation regarding 180 Poster Abstracts the semantics of standards of comparison: some languages allowing only individual standards, e.g. Turkish (Hofstetter 2009); whereas both individual and degree standards are available in Japanese (Shimoyama 2012) or English (Heim 1985). Implicit Assumption in the Literature Phrasal standards (e.g. “Bill” or “him”) have been related to individual comparison, while it is generally assumed that clausal standards (“Bill is”) involve operator movement, similar to a whmovement, in which a trace or variable is left that corresponds to a degree of a gradable predicate (Chomsky 1977, Kennedy 2007, a.o.). Hence, the following correlation is frequently implicitly assumed: (3) A phrasal complement implies an individual denoting standard, and a clausal complement implies a degree denoting standard. Proposal We argue that this correlation does not hold and that a new reformulation of the parameters is necessary. Japanese only allows for phrasal complements (DPs, which may include a RC) in standards of comparison. Crucially, however, this language can have both individual comparison as well as degree comparison (as in “Kono matidewa [gozyuufifty-meetorumetersyorithan] takai biruwa kinsisareteiru” ‘In this town, buildings higher [than 50 meters] are prohibited’, Shimoyama 2012:85). After examining the que – de alternation in Spanish comparatives, also found in other Romance languages, we also conclude that the generalization in (3) is not correct. (4) a. Mide más {*que/de} dos metros. Degree DE ‘It measures more than 2 metres.’ b. Tu hija es más alta más {*que/de} lo que esperaba. Degree DE [Lit.] ‘Your daughter is taller than that which I expected.’ c. María mide más {que/*de} Ana. Individual QUE ‘María measures more than Ana.’ (5) a. Amelia envía más flores de{[DP las [CP que tú recibes]/*[CP tú recibes]} 7DE[CP] /3DE[DP+RC] b. Amelia envía más flores en una semana que [CP tú recibes en un mes]. 3QUE clausal[CP] ‘Amelia sends more flowers in a week than you receive in a month.’ 181 Laura Vela-Plo The data in (4) and (5) suggest that standards with ‘que’ may denote individual comparison ((4c), with the denotation proposed in (6a)), but, interestingly, do not allow the expression of degree comparison (see (4a,b)). Curiously, both phrasal and clausal structures are found in the position of complement of ‘que’ (see (4c) and (5b)). In contrast, standards with ‘de’ allow the expression of degree comparison (see (4a,b)) and, crucially, ‘de’ standards can only have a phrasal complement, either a DP or a RC whose head refers to the degree or quantity employed as reference of the comparison (see (4b) and (5a)). In terms of the properties that we are checking, what follows from our analysis is that Spanish ‘que’ standards are non degree denoting and allow clausal structures; whereas ‘de’ standards can denote degrees and disallow clausal structures. (6) a. J másque K = λG<e,d> λyλx[G(x) > G(y)] INDIVIDUAL denoting standard b. J másde K = λG<e,d> λdλx[G(x) > d] DEGREE denoting standard In sum, building on data from typologically different languages such as Spanish, Italian (Donati 1997), Japanese (Shimoyama 2012) or Turkish (Hofstetter 2009), a.o, we show that the correlation outlined in (3) assumed so far does not hold. If this is correct, the conclusion that follows from here is that degree abstraction in standards of comparison does not depend on a clausal structure with wh-movement, and that some languages resort to phrasal structures with degree or quantity relativization for that purpose. In this respect, we propose that the nature and complexity of standards of comparison is cross-linguistically determined by syntactic and semantic properties that correspond to two independent parameters, not to a single criterion as presumed. Extending Kennedy’s 2007 proposal, we argue that there is a syntactic parameter [+/− clausal] by which either phrasal, or both phrasal and clausal structures are allowed in the complement position of the standard marker; and a semantic parameter [+/− degree denoting] that restricts the semantic type of the standard, which can be either non-degree (6a) or degree denoting (6b). 182 Poster Abstracts Table 1: Proposed parametric classification and some sample languages. SYNTACTIC PARAMETER Are clausal structures allowed? −clausal standard SEMANTIC PARAMETER Is degree comparison allowed? +clausal standard −degree Turkish (abl), Japanese (yori), Italian (che), denoting English (than), Italian (che), Spanish (que) Spanish (que) +degree Japanese (yori), English (than), English (than) denoting Italian (di), Spanish (de) Key Contributions of this Paper • A formal syntactic and semantic analysis of the que – de alternation in Spanish comparatives that can be applied to other Romance languages. • The idea that a clausal (CP) complement with wh-movement is not the unique strategy employed to express degree comparison in natural languages (Chomsky 1977): the use of either DegPs directly or DPs which may be modified by RCs is also a widespread option, especially (but not limited to) among Romance languages (also Donati 1997). • Two autonomous parameters, one syntactic and the other semantic, are able to account for the variation observed on the expression of standards of comparison cross-linguistically. References Chomsky, N. 1977. On Wh-Movement. In Culicover, P. et al. eds. Formal Syntax, 71–132. NY: Academic Press. Donati, C. 1997. Comparative clauses as free relatives. Probus 9, 145–166. Heim, I. 1985. Notes on comparatives and related matters. Ms., U. Texas, Austin. Hofstetter, S. 2009. Comparison in Turkish. PSuB 13, 187–202. Kennedy, C. 2007. Modes of comparison. In PCLS 43, 1(1), 141–165. Shimoyama, J. 2012. Reassessing cross-linguistic variation in clausal comparatives. NatLangSem 20(1), 83–113. 183 Similarities and differences in lexical synonymy: far and its synonyms in European languages1 Maria Zarifyan & Anastasia Melnik (RNU Higher School of Economics, Moscow) [email protected] & [email protected] The current study is conducted within the framework of the research group “Multilingual database of synonyms”, that has several purposes: research, descriptive and practical. The research part is based on lexicographic and typological approaches: we highlight similarities and differences between the words. The results of the research part formed the basis of a database with the descriptions of the synonyms of Russian, English, German, Italian, French and Polish (the technical developer of the database is Nikolay Mikulin). The practical part of our study consists of developing online drills that will help a language learner understand the differences between synonyms of both native and target language. This paper demonstrates the analysis of a semantic field far as an example of our study. The study was conducted in three stages. First of all, we create descriptions for each item in a set of synonyms in order to juxtapose their semantic and syntactic features. Secondly, we highlight idiomatic meanings of each word, because acquisition of idiomatic meanings is usually the most difficult part of language learning. The aim of our work is a development of online drills that are based on the descriptions. This paper focuses on the concept ‘far’ and its synonyms of Russian and English languages. The polysemy of spatial words is a well-studied area. First of all, some spatial adjectives can be used both in deictic (reference to a speaker position) and non-deictic modes (see (1) and (2), respectively): (1) My school is far away and I get very tired walking there. (2) My school is very far from the city centre. 1 The article was prepared within the framework of the Academic Fund Program at the National Research University Higher School of Economics (HSE) in 2016 (grant №16-05-0054 Multilingual database of synonyms: theoretical and computer models) and supported within the framework of a subsidy granted to the HSE by the Government of the Russian Federation for the implementation of the Global Competitiveness Program 184 Poster Abstracts Secondly, lexemes with ‘far’ semantics are characterized by having two kinds of usage due to the type of landmark valence: on the one hand they can be used in deictic contexts, in which the observer acts as a reference point (see Paducheva 1993), on the other hand they can form constructions with narrative meaning, in which the reference point is an object outside of the speaker’s scope. Moreover, it is well-known, that spatial lexemes can develop temporal meanings. “Space-time” metaphor is considered to be one of the language universals of semantics, mostly manifested in spatial prepositions and adverbs (Fillmore 1971, Lakoff & Johnson 1980, Clark 1973, Donati 2010), yet also found in Russian adjectives (Bulygina & Shmelev 1997): the authors observe Russian antonymous adjectives blizkij ‘close, nearby’ and dalekij ‘far’ and point out that blizkij can denote only future events, but not the past ones: blizkoje buduščeje – ‘near future’ (*blizkoje prošloje – ‘near past’). Dalekij most likely denotes distant events in the past (dalekoje prošloje – ‘remote past’), however it can be also used when one is talking about the future (dalekoje buduščeje – ‘distant future’). While ‘space-time’ metaphorical mapping is universal, there are also metaphors that are less widespread across the languages and thus they are quite complicated for language learners. In order to facilitate language learning, we have highlighted the some parametres of ‘far’ that can influence lexicalization patterns across languages: for example, the type of an oriented object can be relevant. For instance, all languages are sensitive to animacy or inanimacy: when one denotes a kinship of two or more people, English distant and remote can be used (distant <remote> relative, but not ? far relative) as well as Italian parente lontano (distant relative), German entfernter Verwandter (distant relative), Russian dalnij rodstvennik (distant relative), French cousin éloigné (distant cousin), Polish daleki krewny (distant relative). Such differences between languages are the most difficult part for those who study a foreign language, thus, our practical aim is to develop online exercises, based on the descriptions. While exercises and tests from language learning textbooks give the information about a foreign language without considering the structure of learner’s native language, our multilingual exercises allow a user to establish connections between grammatical and semantic structure of both target and native language. The initial step of a drill is to provide a learner with a set of rules for using the words in different contexts. Such rules must be short and clear, but at the same time they must encompass each aspect of. Here is the example of English rules: Maria Zarifyan & Anastasia Melnik (3) a. b. c. d. e. f. g. h. i. 185 An object or place, located at considerable distance (far, distant, remote): a far country. Located at considerable distance from the present on the time axis (far, distant): far future, distant (remote) past. Coming from a distance; perceptible from afar (distant): a distant sound; a distant telephone call. Located at more considerable distance than another (about two things) (far). He lives on the far side of the lake. the far corner. Extensive or lengthy (far): a far journey A person, distantly related by blood or marriage (distant): distant cousin. Minimally similar (distant): a distant likeness. Operating or controlled from a distance (remote): remote sensors, remote job, remote control Hidden away; secluded (remote): remote village When a learner looks through these rules and sees the differences between each lexeme he/she can start with filling the gaps with the suitable form. Here is the example of the first task: (4) This ____ town is so out of the way that mail comes only once a week. Thus, a learner must read a sentence and fill the gap with one or several suitable adjectives. The context helps to choose the correct lexeme: for instance, from the sentence above, a learner can infer that the town is secluded and hard reach, therefore the correct answer is remote. References Apresjan Yu. D. 1980. Anglo-russkij sinonimicheskij slovar’ [English-Russian Dictionary of Synonyms]. Moscow: Izdatel’stvo Russkij Jazyk Apresjan Yu. D. (ed.) Novyj ob"yasnitel’nyj slovar’ sinonimov russkogo jazyka [The New Explanatory Dictionary of Synonyms of the Russian Language]. Clark, H.H. 1973. Space, time, semantics, and the child. In T.E. Moore (ed.), Cognitive development and the acquisition of language (pp. 27–63). New York: Academic Press. Donati M. 2010. The space of address between deixis and metaphor. In G. Marotta, A. Lenci, L. Meini, and F. Rovai (eds.), Space and language 2009. 186 Poster Abstracts Proceedings of the Pisa International Conference, 299–315. Pisa: Edizioni ETS. Fillmore, C. J. 1971. The Santa Cruz lectures on deixis. Bloomington, In Indiana University Linguistic Club. Lakoff, G. & Johnson, M., 1980. Metaphors we live by. Chicago: University of Chicago Press. Practical Information 189 Internet Access You will be able to access the internet via Wi-Fi at the conference venue and in most buildings of the university. If your university login allows you to access eduroam, this is also available at University of Leipzig. If not, please connect to eduinfo using the following credentials. Network name Event name (Veranstaltungsname) Password IP-Address DNS server eduinfo ConSole XXV 22c45-2016 obtained automatically (DHCP) obtained automatically (DHCP) Copy Shops Unfortunately, there is no copying facility right in the Seminargebäude, but some copy shops are quite close: Printy Ritterstraße 5 Monday to Friday, 8:00 – 18:00 http://www.printy-leipzig.de/ Universitätsdruckzentrum Ritterstraße 10 Monday to Friday, 8:00 – 18:00 Copy Café – Copyhouse Universitätsstraße 18 Monday to Friday, 8:30 – 19:30 http://www.copyhouse-leipzig.de Campus Copy next to Copy Cafe 190 Practical Information Lunch Here are some suggestions for the lunch break. All locations are close to the conference venue. Mensa am Park Universitätsstraße 5 (ground floor of the conference venue) The University’s central cafeteria on the campus provides a pasta and salad buffet and it offers dishes with meat or fish, and vegetarian dishes. Diner No. 1 Universitätsstraße 16 A small American-style diner serving burgers, sandwiches, and pancakes. Just the right choice if you’re yearning for a savoury meal. Maza Pita Neumarkt 9 Excellent Syrian food in a small shop with a copy shop in the basement. Subway Neumarkt 9–19 The well-known sandwich chain. Regular sandwiches start at around 3 €. Handbrotzeit Nikolaistraße 12–16 Hand-sized freshly baked bread filled with melted cheese and other ingredients. They also offer a salad of the season. Soup&Nem Nikolaistraße 18 A Thai bistro. You can get various soups and other europeanized Asian food like fried noodles or roasted duck. Bagel Brothers Nikolaistraße 42 Bagels with lots of different toppings; coffee, tea, pastries. Vapiano Augustusplatz 11 A well-known German casual restaurant chain offering Italian food. You order your food directly from the cooks. The pizza and pasta start at 7 €. Coffee Shops 191 In addition, you find a lot of snack bars and bakeries all over the city centre, especially in the Nikolaistraße and the main station (German, Italian and Asian diners). Coffee Shops Starbucks Grimmaische Sraße at the east end of the campus. Coffee culture Opposite the Seminargebäude at the corner of Universitätsstraße and Gewandgäßchen. Restaurants & Bars If you are looking for a restaurant or bar to go to in the evening, here are some suggestions. Acapulco Karl-Liebknecht-Straße 42 A nice bar/restaurant on the ‘Karli’, a street in the south of the city centre, which features a lot of different bars, pubs, clubs and restaurants. The Acapulco offers a lot of Mexican specialties for moderate prices and may serve as the perfect starting point for a tour through the bars in the south of Leipzig. Piccola Italia Rosentalgasse 12 A small, cosy Italian restaurant right at the north end of the city centre and a ten-minute walk from the main station. 100Wasser Barfußgässchen 15 A nice café/bar in the colourful style of the Austrian architect/designer/artist Friedensreich Hundertwasser. Especially the rich breakfast buffets, coffee specialties and large selection of cocktails 192 Practical Information are worth a try. The 100Wasser is located in the city centre in the ‘Barfussgässchen’. Auerbachs Keller Mädlerpassage, Grimmaische Straße 2-4 The famous Auerbachs Keller is one of the oldest bars and restaurants in Leipzig. It is, of course, a magnet for tourists from all over the world but nevertheless it has maintained the special flair which inspired Goethe to devote a scene from ‘Faust’ to his favourite wine tavern. Not a cheap dinner place but definitely worth a look. Bayerischer Bahnhof Bayrischer Platz 1 The Bayrischer Bahnhof is a famous bar/beer garden a few minutes southeast of the city centre. It features some typical German or rather Bavarian specialties and the ambience of a historic German brewery. Thüringer Hof Burgstraße 19 The Thüringer Hof is located in the city centre, right next to the famous Thomaskirche. A variety of traditional Thuringian and Saxonian dishes are served. India Gate Nikolaistraße 10 A nice restaurant in the middle of the city centre, right next to the Nikolaikirche and only a two-minute-walk from the university. It offers a lot of interesting food in a cozy Indian ambience. Shady Körnerstraße 2 This small restaurant serves Middle Eastern cuisine in a very nice atmosphere. The food is not that cheap, but absolutely delicious. Fra Diavolo Burgplatz 2 A stylish Italian restaurant at the north end of the city centre that serves delicious pizza and pasta. Restaurants & Bars 193 Barfussgässchen, Karli and Gottschedstraße There are three locations where a lot of bars and restaurants are situated. One is the Barfussgässchen, a small alley, right in the city centre near the market place, where you find pubs, bars, restaurants and a lot of people every night. This is also where the conference dinner on Thursday will take place. The second is the Karli (Karl-Liebknecht-Straße) a bit south of the city centre. It is quite a long street with many different bars and is especially popular among students. The third place is the Gottschedstraße at the west side of the city center where you can find a lot of bars, pubs, restaurants and coffee bars. Pilot Bosestraße 1 A very nice bar/restaurant next to the ‘Centraltheater’ where you might meet a bizarre stage actor. The food is not that cheap, but excellent (e.g. rather creative salads or sandwiches). Sol y Mar Gottschedstraße 4 A bar/restaurant with relaxing atmosphere where you can sit on couches and even get a massage. An Nam Gottschedstraße 13 An excellent Vietnamese restaurant serving fresh dishes, both vegetarian and meaty, at reasonable prices.