Abstracts
Talks
Željko Bošković: On subject positions, the EPP, and contextuality of syntax - [handout]
This talk will examine a number of issues, starting with the derivation of subject questions like who left and a number of other constructions that involve local subject A’-movement. The over-arching concerns during the discussion will be what the constructions under consideration tell us about wh and subject positions, the EPP, and the syntax more generally. The answers to these concerns will lead to fine-tuning the position of various subject and wh-phrases, split IP, and a new conception of the EPP that situates it within a broader theoretical move toward contextuality in syntax in general, which also argues against regarding syntax as fully Markovian in nature.
Ewa Willim: A dedicated null human impersonal pronoun in Polish and the Null Subject Parameter - [handout]
In the (classical) Null Subject Parameter (Rizzi 1982), formulated in a framework in which a null subject pronoun (pro) is a minimally specified nominal whose φ-features are (or can be) supplied by INFL, a null subject pronoun can be referential/definite iff its licensing head [AGR] bears a grammatically specified feature of person. INFL lacking a specified feature of person can only license a non-referential null subject pronoun in a null subject language (NSL).
Addressing the question whether there is pro in natural language grammar, Holmberg (2005) observes that a null referential subject pronoun and a null speaker-inclusive/generic subject pronoun are in complementary distribution in rich agreement languages and that the (un)availability of a null referential subject pronoun and the (un)availability of a null generic subject pronoun correlates with the typological patterns of null subject languages. Consistent NSLs such as Italian, Spanish and European Portuguese allow a [3sg] referential null subject but disallow a null generic [3sg] subject pronoun and conversely, a null generic pronoun is licensed in partial NSLs like Finnish, Hebrew and Brazilian Portuguese, which do not license null antecedentless/discourse-anaphoric subject pronouns. In Holmberg (2010), this asymmetry is reduced to the presence or absence of a feature that can trigger a definite/referential interpretation of a null subject pronoun, viz. [uD], in T.
In this talk, I will revisit the syntactic derivation of two morphosyntactically distinct Polish impersonal constructions and drawing on the analysis of Turkish impersonal constructions offered in Legate et al. (2020), I will argue that they feature a minimal thematic null pronoun, which enters the syntactic derivation with an unvalued φ-feature [iφ: __]. The pronoun is licensed by a head with a valued uninterpretable feature [uφ: human], which confers upon the pronoun it selects for in its specifier a [+human] interpretation restriction via Agree. The heads are realized with distinct morphology at PF and thus, they also identify the pronoun they license.
If the analysis of the two Polish impersonal constructions that I will present is on the right track, Polish impersonals offer further evidence that impersonal pro reduces to a minimal pronoun entering the derivation with unvalued φ-features and that a φ-deficient null pronoun can be licensed and have its content identified in a consistent NSL in the absence of valued φ-features in T/AGR as long as the grammar makes available a functional head (or heads) dedicated to licensing and identifying a null human impersonal pronoun. To the extent that the difference between referential and non-referential thematic null pronouns crucially rests on the presence or absence of a specified feature of person in T, as originally suggested in Chomsky (1981) and Rizzi (1982), it is Person rather than D that plays a crucial role in licensing pronouns in rich agreement languages (cf. also Fassi Fehri 2009). The presence of the [uD] feature in T thus needs to be further investigated.
Robert Freidin: The cycle in syntax: rereading the past to understand the future - [handout]
From the fundamental perspective of principles and processes, the foundation of the generative enterprise for the past almost seven decades, this presentation will review the history of cyclicity in generative grammar to elucidate what conceptual shifts have occurred and what may be presently ongoing.
Mamoru Saito: In Defense of Covert Wh-movement (after 40 years) - [handout]
There are two main syntactic analyses for wh-questions in Japanese, exemplified in (1).
(1) [ CP [ IP Taroo-ga nani-o kat-ta] ka] osie-te kudasai.
Taroo-NOM what-ACC buy-Past ka teach please
‘Please tell me what Taroo bought.’
The first analysis, proposed by Lasnik and Saito 1984 and Richards 2001, among others, is in terms of covert movement of the wh-phrase to the specifier position of the question C, ka. The second was originally proposed by Kuroda 1965 and developed by Nishigauchi 1990 and Takahashi 2001, among others. According to this analysis, the wh-phrase is interpreted as a variable and is unselectively bound by ka, which serves as the question operator. I argue in this paper for the covert movement analysis. Tsai 1999 presents convincing evidence that argument wh-phrases in Chinese are interpreted as variables. I first compare data from Chinese and Japanese and show that wh-phrases in Japanese differ from their Chinese counterparts and are not interpreted as variables. Further, comparing data from the two languages again, I confirm the generalization of Nishigauchi 1990 and Watanabe 1992 that Japanese wh-questions are subject to the wh-island condition. I conclude then that wh-phrases in Japanese undergo operator movement.
In the second part of the paper, I examine the nature of this operator movement. It has been assumed that wh-phrases in Japanese must occur with a quantificational particle such as the question ka. However, I first point out that this assumption is not warranted, as (2) indicates.
(2) Daigaku-wa asu [ CP [ IP (sotugyoosiki-de) dare-ga kooensu-ru] to] happyoosu-ru.
university-TOP tomorrow commencement-at who-NOM lecture-Pres. to announce-Pres.
‘The university will announce tomorrow who the (commencement) speaker is.’
I show that the embedded clause in (2) is not a question and yet, the wh-island effect is observed in this kind of clauses. I propose then that the wh-phrase covertly moves to Spec, FocP of the clause. This fits well with Shimoyama 2006 and Kotek’s 2018 semantic analysis of Japanese wh-questions. Kotek proposes that a wh-phrase is interpreted as a focus with a set of alternatives but with no reference of its own. Then, a clause with a wh-phrase has an alternative semantic value (= the set of possible answers to the question) but no ordinary semantic value. Her analysis is that ka shifts the alternative semantic value to the ordinary semantic value, yielding the Hamblin 1973 semantics for questions. Then, the last point aside, the embedded IP in (2) specifies the alternative semantic value with dare as the focus, and what the university is announcing is its ordinary semantic value. This suggests that the wh-phrase in the question in (1) also covertly moves to Spec, FocP. Again, the IP specifies the alternative semantic value with nani ‘what’ as the focus, and what the speaker is asking for is the ordinary semantic value of the sentence. This analysis also fits well with Rizzi and Bocci’s 2017 analysis of wh-movement in Italian. They argue that the landing site of argument wh-phrases in matrix questions is indeed Spec, FocP. I present this analysis in more precise terms in the presentation.
Lydia Grebenyova: Puzzles in Ukrainian Syntax: Implications for Language Structure - [handout]
This talk focuses on a construction in Ukrainian that presents multiple puzzles for the syntactic theory at the same time: the so called -no/-to construction. It is a passive with the structural accusative case remaining on the object. The -no/-to passive will be compared to the canonical passive in Ukrainian and it will be argued that the two constructions are not interchangeable. The -no/-to passive has considerably more aspectual information in it than the canonical passive and it will be shown how this relates to the accusative case assignment. Broader implications for case assignment will be discussed as a result. The accusative object in the -no/-to construction also shows interesting movement behavior, which will be explored alongside definiteness effects. Implications for the nature of scrambling and its optionality will be discussed and a new PF-repair strategy will be explored as a result.
Noam Chomsky: Displacement
Displacement is a ubiquitous and curious feature of language. It became a central topic of inquiry in the early days of the “generative enterprise” 70 years ago, and remains so, still evoking puzzles – at the time, among many puzzles. Reigning structuralist work had registered many achievements, but there were no real puzzles in what was understood to be a “taxonomic science”, with sophisticated methods to yield a compact representation of a corpus. As soon as the first efforts were made to construct generative grammars, it was found that virtually every sentence posed serious puzzles. Accordingly, the immediate task was to construct devices that could at least impose some order on the chaos, so that research could proceed: for displacement, grammatical transformations, which were rich and complex. Over the years, the generative systems were greatly simplified and deepened: for displacement, reaching the limit with Lasnik and Saito’s “Move-alpha”. The minimalist program then sought to place new discoveries and insights on a firm foundation. That has led to rethinking of the nature of structure-building rules, including movement rules, by now the simplest of them. I’d like to discuss some recent ideas about the nature of these generative systems.
Rants
Atakan Ince: Syntactic islands and LLMs
The aim of this study is to show that with a minimal amount of training data designed in terms of 'factorial design', LLMs such as DistilBERT can be fine-tuned successfully. In our case, we trained a classifier that would recognize illicit movement out of syntactic islands in English. With less than 1.2K sentences, the fine-tuned model gets 0.98 F1. This proves that knowledge of language structure is crucial in designing/collecting a minimal amount of data in fine-tuning LLMs, a la Data-centric AI.
Kyle Johnson: Constituency and the cycle - [handout]
The Principle of the Cycle is a way of describing how the flow of information in a phrase marker is monotonically “upwards.” Its effects are difficult to capture in certain contemporary ideas about what phrase markers are. I wish I could do more than rant about this, but a rant is what was requested, and so.
Terje Lohndal: Structured historical approaches to structure
In addition to his massive contributions as a researcher, Howard Lasnik is rightly famous for his gifts as a teacher. Those of us who have had the privilege to be students in a department where he’s been teaching have all been exposed to his particular take on teaching syntax. His approach has always been thoroughly historical and contextual, focusing on the classic and foundational ideas in generative grammar, ‘viewed both on their own terms and from a more modern perspective’ (Lasnik 2000: preface). In this ‘rant’, I will defend this approach to teaching syntax and argue that nothing makes sense except in the light of history.
Jason Merchant: How Howard showed that ellipsis is central to syntax
Lasnik has shown (Lasnik 1995 et multi seq.) that an understanding of ellipsis is central to syntactic theorizing about movement, both head and phrasal. In this rant, I yell about how ellipsis also sheds crucial light on Merge: lexical selection persists under ellipsis in a way that only makes sense if there is syntax inside ellipsis sites, and if Merge operates in its usual way (sensitive to selectional features). Lasnik has lit the way: we just have to follow it!
Keiko Murasugi: Parameterization in Labeling: Evidence from Language Acquisition - [handout]
In Chomsky's minimalist approach to syntax, the operation Merge, a part or UG, accompanies a Labeling algorithm that specifies the nature of the formed object. A finite clause is labeled through the sharing of fai-features between the subject DP and T, and this is the reason a language requires fai-feature agreement. Saito (2016), on the other hand, suggests that suffixal Case markers make it possible to label finite clauses in Japanese, a language that lacks fai-feature agreement. In this short talk, I will discuss the well-known erroneous strings very young children produce such as Root Infinitives (which are caused by the lack the fai-feature agreement), preverbal object sentences in English, and sentences without Case markers in Japanese and Korean. The main question to be addressed is whether those sentences children produce are labeled. I will argue that the labeling of {XP, YP} structure is where variation is found among languages, and children have not acquired the labeling of {XP, YP} structure at the Root-Infinitive stage. This may fit well with Saito's (2016) proposal that the main parametric differences between English and Japanese can be attributed to the difference in the ways the two languages label the {XP, YP} structure.
Zach Stone: Adorning bare grammars
The primary purpose of Keenan and Stabler's bare grammars (BGs) is “to discard those aspects of generative, derivational history which appear just because of our decision to use some particular generative device.” BGs do this by abstracting away from everything but the combinatorics of the grammatical operations. While this is a natural first step in developing a theory of grammars, in most models, the objects manipulated by the grammar are structures, in that they have meaningful properties beyond their identity. With similar goals, the next natural step is to study BGs over structured objects in general. We propose an algebra which assumes that the objects being manipulated are structures and that the operations of the grammar preserve that structure, making few assumptions about what that structure is. In particular, we propose is a BG-like abstraction of graph-rewrite grammars (cf. Ehrig, et al. 2006), with the added goals of being able to analyze the rules, languages, and grammars and how they generate structure in a way that is easy to extend to more complex representations, such as those admitting feature-sharing or geometry.
Posters
Adolfo Ausín: The (lack of) structure of Phrasal Compounds
A long-standing debate in generative grammar has been to determine whether syntax can access the internal structure of words. Phrasal compounds are a good testing ground for alternative theories on the topic due to their dual nature. On one hand, they have the structure of a full sentence. On the other hand, they have the distribution and function typically associated with words. The main empirical facts to be discussed in this poster are the contrasts illustrated in the following sentences:
(1) a. He never paid attention to those Obama-is-a-Muslim rumors. (He ≠ Obama)
b. *He never paid attention to the rumor that Obama is a Muslim. (He = Obama)
(2) a. No woman was carrying Lock-her-up signs. (No woman ≠ her)
b. No woman wants the president to lock her up. (No woman = her)
(3) a. *She always has a Someone-will-fix-my-problems attitude, but she doesn’t know who.
b. She is convinced that someone will fix her problems, but she doesn’t know who.
We believe that the above contrasts provide empirical support for two ideas: 1) some version of the Lexicalist Hypothesis (since the internal structure of phrasal compounds cannot be accessed from outside), and 2) syntactic relations like the ones involved in (1)-(3) must have a global, long-distance nature (since an alternative, pragmatic account would not expect the contrast between syntactic embedding in the b-sentences and morphological embedding in the a-sentences).
Tomohiro Fujii: Choosing among auxiliary fronting transformations
The poster discusses the well-known issue of how the learner can rule out incorrect auxiliary fronting transformations when the input is consistent with them as well as the correct hypothesis. Two ways of evaluating hypotheses are applied to the issue. One is by using a rule simplicity metric, and the other by examining how well each transformation fits certain hypothetical input data. The transformations considered include “Move the first Aux” (H1), “Move the matrix Aux” (H2), and “Move any Aux” (H3). On the one hand, according to the proposed simplicity metric, H3 is simpler than H1, and H1 is simpler than H2. The goodness-of-fit measure, on the other hand, reveals that H2 scores better than H1, and H1 scores better than H3 in particular settings. It is shown that the proportion of data with the second auxiliary fronted affects the results.
Gesoel Mendes: Ellipsis and focus related particles in Brazilian Portuguese and Nupe
In this talk, I argue that the C-element que, following fronted wh-elements and fronted focused elements more generally in Brazilian Portuguese, is realized as Fin, rather than Foc (MENDES & KANDYBOWICZ, 2021; pace MIOTO, 2001; MIOTO & KATO, 2005). I put together three observations from the literature: (i) the appearance of que is contingent on wh/focus fronting; (ii) que introduces a finite clause, and (iii) que disappears under sluicing. We present novel evidence that Nupe’s focus particle is a left- periphery element and that Nupe provides a concrete counterexample to Merchant’s (2001) sluicing-COMP generalization. A comparison between Nupe and Brazilian Portuguese regarding the presence of nonoperator material in sluicing constructions is crucial to establishing sluicing as FinP ellipsis (BALTIN, 2010; ABOH, 2010), instead of TP ellipsis, as standardly assumed, as well as que as a Fin element. I offer an analysis that captures all of the Brazilian Portuguese distributional facts, according to which que is a Fin head with a [FINITE] feature and an uninterpretable [uFOC] feature that must be licensed by Agree with a higher focus head.
Dongwoo Park: Korean specificational pseudoclefts and argument ellipsis
There has been much debate about how null arguments are generated. The pro approach (Hoji 1985, Abe to appear) argues that null arguments are empty pronouns (pro) (cf. Landau to appear). According to the LF-copying approach (Oku 1998, Kim 1999, Saito 2007, Sakamoto 2016), the null argument position is empty in overt syntax, but its antecedent is copied into the empty position at LF. On the other hand, V-stranding VPE approach (Otani and Whitman 1991, Funakoshi 2016) proposes that there is no independent argument ellipsis. Takahashi (2020) proposes a derivational null argument approach by showing that extraction out of null arguments are not impossible. In this presentation, I provide novel Korean specificational pseudocleft data, showing not only that extraction out of null arguments is possible but also rather that it must occur in overt syntax. Furthermore, I argue that argument ellipsis occurs as soon as arguments are merged, and that what is extracted out of null arguments is lexical items that have been deprived of their phonological feature matrices as a result of ellipsis. This is based on Park’s (to appear) proposal that elided elements can participate in further operations followed by ellipsis.
Based on this, I propose that arguments are elided as soon as they are merged. Additionally, in order to account for a seemingly contradictory fact that elements generated inside the embedded CP argument cannot be pronounced outside the ellipsis site when the embedded CP is elided, I adopt Park (to appear). It proses that all elements generated inside the null arguments are deprived of their phonological feature matrices. However, they preserve their formal matrices and can participate in further formal operations.
Round Table
Hamida Demirdache: (Non)covaluation and crossover constraints - [handout]
We can read through Howard’s work the history of Binding Theory, with so many landmark papers, such as ‘Remarks on coreference’, ‘On the necessity of Binding Conditions’, or ‘Weakest Crossover’, to name but a few that I will be discussing. I will try to bring together the conditions under which constraints on ‘coreference’ and bound variable anaphora (Condition C, WCO) can or cannot be obviated and what property of anaphora could contribute to explaining in part apparent immunity to such constraints.
William Idsardi: Comments on Lasnik 1989
Howard's contributions to syntax are legion. Less-well-known are his contributions to phonology and PF. I will review a few publications that deserve to be revived and re-examined in light of current theories of syntax and phonology.
Robert May: Remarks on coreference
Howard Lasnik’s paper “Remarks on Coreference” is a landmark in the research project that explored anaphora, movement, scope and other phenomena related to binding. In my remarks, I will discuss in this historical context the relation of derivation and representation, in particular, the notion of a P-marker, as an equivalence class of derivations (as defined in LSLT), and as structural representations over which global non-derivational properties are defined, in particular c-command, the core theoretical notion of the project. With respect to Lasnik’s paper specifically, I will discuss the notion of non-coreference employed there, and the reactions to it in the work of Evans and Higginbotham, and Fiengo and May.
Maria Polinsky: How many ways to cross over weakly?
In this presentation, I will revisit the paper by Howard Lasnik and Tim Stowell “Weakest crossover” (1991). The paper has stood the test of time and is still relevant for our understanding of operator-variable configurations and resumption. I will discuss the implications of its findings for the current understanding of restrictive relative clauses.
Luigi Rizzi: Howard’s non-coreference in adults, children, and machines
Howard’s basic “non-coreference” paradigm has an extraordinary didactic value in that it clearly illustrates many things at once, in an easily accessible way:
- The unconscious character of mental computations for language,
- The simple and effective experimental paradigm offered by metalinguistic judgments,
- The abstractness and complexity of mental representations and computations.
Because of these rich properties, the non-coreference effect can offer an illuminating path for beginning students through the intricacies of modern formal linguistics. It can also be very useful in discussions with cognitive and brain scientists who do not have a linguistic background, a scientific population which is sometimes reluctant to accept the view that linguistic representations may be more complex than meets the eye.
And, keeping in mind Descartes’ and Turing’s lessons, non-coreference may also represent an interesting test case for machines, remarkably successful in interacting linguistically with humans in applications like Chat-GTP: Can machines calculate non-coreference in a human-like manner?