Whamit!

We are as delighted as can be to announce that Shota Momma will be joining our faculty as Associate Professor of Linguistics this Fall!  Prof. Momma is a specialist in psycholinguistics and its interaction with linguistic theory — with a particular focus on the mechanisms of sentence production, an area in which he is a true pioneer. Shota comes to us from the University of Massachusetts, Amherst, where he has been an Assistant Professor since 2019.  He received his PhD in Linguistics from the University of Maryland in 2016, with a dissertation directed by our alum Colin Phillips (PhD 1996), and subsequently completed a postdoctoral fellowship at UC San Diego with Vic Ferreira.  Asked about his thoughts and plans as our newest faculty member, Shota wrote:

 ”I’m deeply honored and excited to be joining the MIT linguistics department, which has been home to so many people I deeply respect, past and present. I’m looking forward to learning from my future colleagues and to building a vibrant intellectual community together. My research aims to understand how people construct sentences in their minds during both comprehension and production, drawing on insights from linguistics and cognitive science more broadly. I’m confident that exciting new research directions will emerge at MIT - something I already began to experience during my visit in 2023.”

Welcome!  We can’t wait for you to join us!

photo source: https://bpb-us-e2.wpmucdn.com/websites.umass.edu/dist/5/13737/files/2020/02/Shota-Momma.jpg

Speaker: Hani Al Naeem (MIT)
Title: On the nature of emphasis spread in Jordanian Arabic
Time: , 5pm - 6:30pm
Location: 32-D831

Abstract: The phenomenon of emphasis spread (ES), a type of tongue root harmony in Arabic, is triggered by emphatics, coronal obstruents with a secondary posterior articulation near the upper pharyngeal wall. The most salient effect of ES is the backing of adjacent low vowels, with notable directionality differences in the extent and magnitude of this effect. While previous works agree that leftward ES is more robust (i.e. has a uniform effect and broader span) than rightward ES, there have been differences in the descriptions of the two patterns of spreading and in the analyses thereof. This work reconsiders the empirical description of ES in Jordanian Arabic (JA) based on data from a production experiment and provides a novel analysis of the phenomenon. The JA data reaffirm that ES uniformly lowers F2 in all leftward low vowels within a stem, while the effect gradually fades out to the right. I argue that this asymmetry reflects two distinct underlying mechanisms, feature harmony and coarticulation. Following Hayes & Londe (2006), feature changing effects are modeled through a distal constraint targeting leftward segments non-locally and a local constraint iterating to a right-adjacent vowel. Once those effects are accounted for, a model of coarticulation that is informed by the locus equation and vowel undershoot (Flemming 2001) is proposed as a basis for the residual coarticulatory rightward effects. I claim that the present analysis provides an explanation of the directional asymmetry in ES and clarifies the nature of the long-distance rightward effects by attributing them to a phonetic mechanism, explicitly modeled.

Speaker: Janet Pierrehumbert (University of Oxford)
Title: LLMs can pass the Turing Test — are they intelligent?
Time: Thursday, May 14, 12:30pm - 2pm
Location: 32-D461

Abstract: In 1950, Turing proposed that if a person can not tell whether they are in conversation with another person or a computer algorithm, we can consider the algorithm to be intelligent. The Turing Test effectively launched the field of AI, and advanced a strong connection between natural language and intelligence. The newest Large Language Models (LLMs) engage in conversations and produce amazingly human-like output. Often, people cannot reliably tell whether they are talking to a chatbot or another person. LLMs seem to pass the Turing Test. Does this mean they are intelligent?

In this talk, I will discuss the nature of the Turing Test and the current level of evidence that LLMs can pass it. I will argue that the Turing Test in its original formulation had a limited conception of intelligence, failing to capture aspects of intelligence that come to the fore in theories of embodied cognition. These aspects of intelligence play a crucial role for humans in acquiring a mental lexicon of meaningful units, and mastering semantic operators (such as markers of temporal, numerical and logical relationships). When probed, LLMs exhibit persistent shortcomings in these areas of language, shortcomings which are systematic consequences of their architecture and training.

Speaker: Daniar Kasenov (NYU)
Title: Salvation by deletion in Russian LBE
Time: Tuesday, May 12, 1:00pm - 2pm
Location: 32-D461

Russian does not allow left branch extraction from NPs that are complements of P. Sluicing alleviates this restriction. I argue that the pattern is best explained by the Cyclic Linearization view of “island repair”: ungrammaticality results from conflicting linearization statements but ellipsis can resolve the conflict.

The talk will cover: 
— Why the pattern must involve salvation by deletion (against the general attitude expressed by Barros et al. 2014)

— Extension to other restrictions on extraction from Russian NPs which are alleviated by sluicing based on the scattered deletion view of left branch extraction (Fanselow, Ćavar 2003; Bondarenko, Davis 2023).

— An account of what we call the Sole Remnant Generalization: the remnant must be the only pronounced item in its clause (observed for preposition drop in other languages too). The model is a mix of Fox & Pesetsky (2003) and Johnson (2020) which allows for scattered deletion and predicts the Sole Remnant Generalization straightforwardly.

The talk is an extension of the material written up in Kalyakin, Kasenov (2025): https://ling.auf.net/lingbuzz/009308

Speaker: Christopher Bader (MIT)
Title: Front Vowels are Palatal: Phonetic and Phonological Evidence
Time: , 5pm - 6:30pm
Location: 32-D831

Abstract: Clements (1991) and Hume (1992) proposed that front vowels are coronal, rather than dorsal (Sagey 1986). But this is the wrong generalization, since it fails to explain the following contrast in Mandarin Chinese: *si, *ʂi, ɕi ‘west’ (西) (Lee-Kim 2014). As this example shows, the Mandarin high front vowel /i/ may only be preceded by a palatal ([-anterior][+distributed]) sibilant. Anterior and retroflex sibilants, which of course are also coronal, may not precede this vowel. Instead of si and ʂi, Mandarin has these consonants followed by the syllabic, non-vocalic coronal sonorants [ɹ ̩ ] and [ɻ ̍ ], sometimes incorrectly referred to as ‘apical vowels’. As I will show, neither anterior nor retroflex consonants pattern with front vowels. But fortition of front vowels can result in palatal consonants and front vowels do pattern with palatal consonants. I will argue that this is because front vowels are palatal. The Mandarin data cited is then the result of AGREE(PAL), AGREE(ANT), and AGREE(RET) being higher ranked than IDENT(Amax), a constraint which preserves vocalic apertures… (continued in attached file)

Speaker: Haoming Li (MIT)
Title: Conditional semantics for permission and weak necessity
Time: Thursday, May 7, 12:30pm - 2pm
Location: 32-D461

Abstract: The classic analyses of permission and weak necessity view them as existential and universal quantification over a modal base with one or two ordering sources. In this talk, I will advance the alternative view that both permission and weak necessity involve an underlying conditional semantics. I will draw analogies between permission and weak necessity on the one hand and conditionals on the other hand across several phenomena, many of which are puzzling for the classic accounts, including free choice/simplification of disjunctive antecedents, Sobel and Reverse Sobel sequences, homogeneity, and performative uses. I will then deliver a concrete implementation based on von Fintel’s (2001) dynamic account of conditionals and Chung’s (2018) work on Korean deontic modals. I will show that the approach naturally captures the parallels mentioned above and derives a stipulation about the nature of the modal base for such modals. If times allows, I will also compare and contrast the present approach with other approaches in the literature to similar empirical puzzles and other work which motivates a conditional semantics for modals, like McHugh (2026), which was previously presented at LingLunch.

Speaker: Edward Flemming (MIT) and Amy Li (MIT)
Title: An optimization-based approach to phonetic grammar
Time: Thursday, April 30, 12:30pm - 2pm
Location: 32-D461

Abstract: A phonetic grammar maps phonological representations onto acoustic or articulatory trajectories. In the tonal domain, this involves mapping tone categories onto f0 trajectories that are aligned to the segmental string (Pierrehumbert 1980). One challenge of modeling a phonetic grammar is to derive variation in the realization of phonological units as a function of variation in the context. To address this challenge, we develop an optimization-based model of phonetic grammar where the constraints on phonetic realization are derived from phonological representations and articulatory limitations. The model is illustrated through a case study of the Mandarin rising tone, focusing on variation in realization due to varying speech rate.

Speaker: Ioannis Katochoritis (MIT)
Title: To Unlock is to (Re)Merge: Locality Domains, Intervention and Minimal Compliance
Time: Tuesday, April 28, 1:00pm - 2pm
Location: 32-D461

Speaker: Jian-Leat Siah (UCLA)
Title: How Language Experience Reshapes the P-Map: The Case of Final Nasalization in Serudung Murut
Time: , 5pm - 6:30pm
Location: 32-D831

Abstract: In this talk, I present my dissertation research testing Steriade (2001/2008)’s P-Map hypothesis, which posits that learners are biased toward alternations that minimize perceptual deviations between input and output forms. In the case of word-final voiced obstruents, the P-Map predicts that final devoicing should be the uniquely preferred repair strategy cross-linguistically. Experimental results from American English listeners in a perceptual AXB task support this prediction. However, results from Serudung Murut (Austronesian) listeners, whose phonological grammar employs (laryngo-)nasalization to avoid voiced stops in word-final position, demonstrate that the perceptual biases posited by the P-Map are not immutable but can be reshaped through exposure to alternations that contradict the P-Map rankings. This synchronic restructuring of the P-Map is especially evident among speakers with advanced proficiency, highlighting the role of language experience in phonological learning.

Speaker: Norvin Richards (MIT)
Title: Long-distance agreement by proxy in Passamaquoddy-Wolastoqey
Time: Tuesday, April 21, 1:00pm - 2pm
Location: 32-D461

Passamaquoddy-Wolastoqey has long-distance agreement: the verb of a matrix clause can (optionally) agree with nominals in a complement clause, even if the complement clause appears to be a full-fledged tensed clause.  I will argue that this long-distance agreement should be treated as an instance of what I’ve called ‘agreement by proxy’.  In this case, the C of the embedded clause has Agreed with the nominals in that clause, and probes in the higher clause, although considerations of locality prevent them from agreeing with any DPs in the embedded clause directly, can agree with the features on embedded C, thereby effectively agreeing with the DPs of the embedded clause.

LEAP Workshop – Cora Lesure and Maya Honda
Title: Communicating Your Research
Time: Tuesdays  5pm – 6pm
Location: 32-D461

We invite you to the LEAP (Linguistics Education and Pedagogy) workshop on Communicating Your Research. Here is the complete schedule of the four-session workshop:

April 14th (last week): Defining the knowledge gap

• What does your audience know, and what do you want them to learn from you?

April 21st (this week): Bridging the knowledge gap

• Given your audience’s prior knowledge, how do you provide essential content in order to facilitate their understanding?

April 28: Evaluating success

• How will you know that understanding has been achieved?

May 12th: Research Slam: Workshop participants will give their presentations!

Workshop sessions take place on Tuesdays from 5-6pm in 32-D461. Note that we will skip May 5th and that the Research Slam will take place during the final week of the semester.

 Come for the snacks! Stay for the workshop!

Do the impressive abilities of neural-network Large Language Models in generating rich, well-formed syntax falsify fundamental principles of generative linguistic theory? The answer I will argue for is: no. But it will be a rather nuanced “no”, trying to identify the proper treatment of generative AI for generative linguistics. Specifically, I will consider these principles:

1. Computability: Generating natural language with rich, human-level syntax requires use of symbolic grammatical rule systems.
2. Explanation: Theoretical explanation in generative linguistics requires built-in discrete symbolic structure.
3. Acquisition: Children’s ability to acquire language requires innate knowledge of grammatical rule systems.
4. Universals: Linguistic universals can only be explained from innate limitations on what languages are learnable.

The quantity of discussion of these questions will decrease sharply from 1–4, the bulk of the presentation focused on 1. The discussion of 1 takes off from Smolensky, Fernandez, Zhou, Opper, Davies & Gao (JAIR 2025; arXiv:2410.17498).

Speaker: Yurika Aonuki (MIT) & William Pacheco (MIT)
Title: Questions and conditionals with disjunction in Gitksan & My Language, My Tools: AI-Assisted Documentation of Kiːwɑ Keres
Time: Thursday, April 16, 12:30pm - 2pm
Location: 32-D461

This week’s LingLunch will consist of two talks.

Talk 1: Questions and conditionals with disjunction in Gitksan (Yurika Aonuki, MIT)
Abstract: A morpheme ji in Gitksan (Rigsby 1986; Hunt 1993; Aonuki 2025) (and cognates in Nisga’a (Tarpent 1987) and Sm’algyax (Sasama 2001; Brown 2022, 2023, 2024, 2026)) introduce conditional antecedents and embedded polar questions. At the same time, ji is also allowed in declarative attitude complements when the attitude holder is uncertain or wrong (Hunt 1993; Gogag et al. in prep; Aonuki 2025); similar patterns are observed in Sm’algyax (Brown 2022, 2023). I will argue that there are reasons to attempt a unified analysis of ji in conditional antecedents and questions on one hand and declarative complements on the other. Against this background, I will provide novel data from disjunction, which shows that ji can introduce each disjunct in conditional antecedents, embedded alternative questions, and polar questions with disjunction. I will propose that ji constrains the set of epistemic possibilities associated with a speech or thought event, via the alternative semantic value of its argument. The proposal highlights a shared component between questions and uncertainty and has implications for an ongoing debate about the semantic relationship between polar and alternative questions (Roelofsen and Van Gool 2010; Pruitt and Roelofsen 2011; Meertens 2021).

*This will be a practice talk for GLOW and SULA-TripleA.

Talk 2: My Language, My Tools: AI-Assisted Documentation of Kiːwɑ Keres (William Pacheco, MIT)
Abstract: Kiːwɑ Wɛkɑ̤ɲi is an underdocumented dialect of Keres — a language isolate spoken in seven Pueblos in New Mexico. I am a native speaker and relatively new to linguistics, and I quickly found that the tools available for documenting my dialect weren’t built with my needs in mind.

Speaker: Giovanni Roversi (MIT) & Jéssica Mendes (Georg-August-Universität Göttingen)
Title: Light-ish attitudes, heavy complements: the view from Äiwoo” (GLOW practice talk, LFRG take-over)
Time: Tuesday, April 14, 1:00pm - 2pm
Location: 32-D461

Abstract: Some attitude predicates in various unrelated languages have been reported to have variable flavor interpretations, covering doxastic (“think”), assertive (“say”), and bouletic (“want”) readings; cf. Navajo (Bogal-Allbritten 2016) and Koryak (Močnik & Abramovitz 2019, Močnik 2025). These have prompted analyses where much of the semantic contribution that distinguishes these readings is taken away from the lexical entry of the attitude verb itself (against the standard Hintikkan view), and rather offloaded to the semantics of the complement clause. Based on novel fieldwork data, we report on another such attitude verb in Äiwoo (Austronesian; Solomon Islands), and we show that in this language there is clearer morphological evidence for this type of analysis: overt elements in the embedded clause determine the reading of the matrix attitude predicate, whose interpretation is derived fully compositionally, and the semantics of these elements is independently verifiable from their use in unembedded contexts. We arrive at a picture where the attitude predicate itself is relatively underspecified, but not fully (against e.g. Bogal-Allbritten). We draw (tentative) cross-linguistic predictions about the range of meanings we expect “variable-flavor” attitudes to cover: these should not extend beyond what can be composed out of independently available elements already in the language (chiefly, relatively underspecified attitudes and already existing modals). Thus, we predict these kind of verbs to receive “think, say, want” interpretations, but not readings like “forget”, “imagine”, “dream”, which cannot be composed out of these simpler more basic atoms.

Note: This week’s Syntax Square will feature a special LFRG take-over: the talk will be fully focused on semantics, so we especially encourage those with an interest in semantics to attend. The presentation will also serve as a practice talk for GLOW, so it’s a great opportunity to hear the work in progress and offer feedback.

LEAP Workshop 4/14 - Cora Lesure and Maya Honda
Title: Communicating Your Research: Defining the Knowledge Gap
Time: Tuesday April 14, 5pm - 6pm
Location: 32-D461

We invite you to the first-ever LEAP (Linguistics Education and Pedagogy) workshop on Communicating Your Research. Across four one-hour sessions, we will explore how the task of communicating key aspects of your linguistics research (eg., your GP or your thesis) can be informed by formal pedagogical theory, and vice versa.

Workshop sessions will take place on Tuesdays from 5-6pm starting this week, in 32-D461. The Research Slam will take place during the final week of the semester on May 12th.

This week, we begin with “Defining the Knowledge Gap”, considering the questions, “What does your audience already know, and what do you want them to learn from you?”

Snacks will be served!

On behalf of LEAP (Linguistics Education and Pedagogy), the department’s outreach effort to K-12 students, Christopher Legerme, Cora Lesure, Vincent Zu (MIT Chem E), and Jacob Kodner (Harvard Linguistics) designed and team-taught a Saturday high school linguistics course at Spring HSSP, an initiative of the student-run MIT ESP (Educational Studies Program). The six-week-long course, An Introduction to Linguistics: The Science of Language, concluded last Saturday. 

Across the six weeks, students examined the linguistics of poetry, as well as phonetics, phonology, morphology, syntax, and machine learning in diverse languages. The teaching team focused on bringing to students’ awareness their unconscious knowledge of language, using the tools of science to formulate and test ideas about the hidden structures of language, and identifying and appreciating similarities as well as differences across diverse languages. The 15 high school students who attended each week can now answer the question that Cora posed to them the first day, What is Linguistics? 

The course grew out of earlier work that the team has done, both individually and collectively, at Spark and Splash, day-long events also organized by MIT ESP, as well as in partnership with teachers for Maya Honda’s Linguistics in K-12 Education seminar.

Speaker: Amy Li (MIT)
Title: The potential effect of phoneme inventory crowding on phonetic variation
Time: Monday April 6, 5pm - 6:30pm
Location: 32-D831

Abstract: Does the crowdedness of a phoneme inventory affect the variability of the realizations of its phonemes? In particular, does a more crowded phoneme inventory reduce the variability of its phonemes? Lavoie (2002) provides some preliminary evidence for such an effect by comparing /k/ in the spontaneous speech of English and Spanish, finding that the percent of phonetically fricative-like or approximant-like realizations is much lower in Spanish (7%), which contrasts /x/ with /k/, than English (21%), which does not. In this presentation, I will discuss my work in progress on the topic, including a new model (and some variants) I’m proposing within Dispersion Theory and my attempts to find evidence for such a cross-linguistic effect on vowel inventories.

Speaker: Joseph Sabbagh (University of Texas Arlington)
Title: The Dynamic Existential in Tagalog
Time: Tuesday, April 7, 1:00pm - 2pm
Location: 32-D461

Abstract: This presentation discusses existential sentences based around a verbal existential predicate (magkaroon) in Tagalog. Such sentences are peculiar because, while they appear to be unaccusative, with a single DP argument projected as an internal argument (=the E(xistential) C(lause) pivot), the EC-pivot surfaces with accusative/dependent case-marked marking. The sole argument of ordinary unaccusative verbs, by contrast, are marked with subject/topic-case. Thus these dynamic existential sentences are problematic for language internal reasons as well as more broadly given proposed universals like Burzio’s Generalization. After demonstrating the the EC-pivot is indeed an internal argument of an apparently unaccusative predicate, I propose that the (verbal) existential predicate is in fact a possessive predicate that projects an external argument. This argument is overt in possessive sentences (which are formed with the same predicate), but covert in existential sentences. The hypothesized covert argument opens up an analysis of the accusative/dependent case marking of the EC-pivot assuming a Dependent Theory of Case. Precedents for such an account from Russian, Icelandic, and Greek are cited; and I conclude with some speculations about the typology of expletives and their availability (overt or covert) across languages.

Abstract:

This talk addresses three cases of anti-singularity in questions, illustrated in (1a)-(1c). They all convey that the speaker expects that the answer will name more than one individual:

(1)   a. Which books did you buy?                            English

        b. Was    hast du   alles gekauft?                    German

             What have you all    bought

             “What all have you bought?”

        c. Quiénes   se       fueron pronto?                  Spanish

             Who-PL REFL left       early

             “Who left early?”

The anti-singularity of (1a) has been explained as arising from competition with its singular version: which book did you buy?, which has a uniqueness presupposition – only one book can be named. This explanation does not extend to the other two cases. There is no uniqueness presupposition in the version of (1b) without alles or in the version of (1c) that has a singular wh quién.

The proper analysis of the anti-singularity observed in (1b)-(1c), I argue, requires us to pivot from consideration of number proper. Drawing inspiration from work on so-called optional plural markers, specifically the Cuzco Quechua morpheme kuna, I show how anti-singularity can be a bi-product of a presupposition of sortability – it should be possible to partition the set denoted by the noun complement along some dimension, such as type, size, color etc.

Speaker: Shigeru Miyagawa (MIT)
Title: Birth of a Language in the Backlands of Brazil
Time: Thursday, April 2, 12:30pm - 2pm
Location: 32-D461

Abstract: It is assumed that in order to acquire a language, children must be exposed to a language during the critical period, which generally lasts until puberty. Here, we report on Cena, an emergent sign language that has developed among a small group of deaf people in an isolated town in the state of Piauí, Brazil. Starting three generations ago, it has developed into a fully functioning communicative system with all characteristics of a typical human language even though Cena developed in a linguistic vacuum. What makes Cena interesting is that we are reasonably certain that Cena had no external input from the national sign language, Libras, or any other language during its formation. Cena challenges the assumption that to acquire the first language, the child must be exposed to a fully developed language. It developed from homesigns to an emergent sign language that is used for all aspects of village life. Cena also lends credence to the interactional model of language acquisition, which considers the interactions between the child and the caregivers to be the crucial element. The nativist model of language acquisition, which assumes a universal system underlying language, also plays a part. Through interaction, what arose is a system with characteristics essential to all human language.

Based on the article by: Anderson Almeida-Silva; Remo Nitschke; Vitor A. Nóbrega; Fernando Valls Yoshida; Shigeru Miyagawa

Cognitive Science (December 2025)
https://onlinelibrary.wiley.com/doi/10.1111/cogs.70159

Title: To Unlock is to (Re)Merge: Locality Domains, Intervention and Minimal Compliance

Speaker: Ioannis Katochoritis (MIT)

Time: March 17, 2026,  1 pm - 2 pm 

Place: 32-D461

Speaker: Eyal Marco (MIT), joint work with Ezer Rasin (Tel-Aviv University)
Title: On the nature of phonological cyclicity: Evidence from Nazarene Arabic
Time: Monday, March 16th, 5pm - 6:30pm
Location: 32-D831

The semantics conference Theoretical Linguistics at Keio was held at Keio University on March 14-16, 2o26. 

The following members of our community presented at the conference:

• Cooper Roberts (3rd year): Part is part (plus pragmatics)
• Viola Schmitt (faculty): Individuation across categories
• Adèle Hénot-Mortier (PhD 2025): “Remind-me” presuppositions with iterated Speech Acts
• Yasutada Sudo (PhD 2012)[UCL], Chris Davis & Tim Jantarungsee: Varieties of sortal restrictions: The case of ingestion verbs
• Christopher Tancredi (PhD 1992)[Keio University]: Epistemic vs. Non-epistemic Modals in Subjective Semantics

The GLOW in Asia 2^nd Workshop for Young Scholars was held at Nanzan University on March 13-15, 2026. Third-year student Cooper Roberts gave a talk titled “Honor omnivorously: The syntax of politeness in Kikai Amami”.  Second-year student Vlad Orlov presented a poster with Daria Belova (Institute of Linguistics RAS/HSE University, Moscow) titled “Detransitivization as agreement with an implicit argument: The case of Tatyshly Udmurt”.

Speaker: Yimei Xiang (Rutgers University)
Title: Function alternations of the Mandarin particle ye: from ‘also’ to ‘even’
Time: Thursday, March 12, 12:30pm - 2pm
Location: 32-D461

Abstract: The Mandarin adverb ye exhibits both a simple additive use (‘also’) and a scalar additive use (‘even’). This alternation is unlikely to be accidental: across genetically and typologically diverse languages, additive markers frequently extend to scalar interpretations. Why does additivity so often give rise to scalarity, rather than to some other meaning component? Mandarin offers a particularly revealing testing ground for this question, as the ‘even’ use of ye overlaps with that of another multifunctional adverb, dou, which displays a distributor–’even’ alternation.

Building on Heim’s (1982, 1983, 1992) model of dynamic semantics, I propose a unified analysis of the two uses of ye. I argue that in both cases ye presupposes Vacuity of Anti-Exclusion (VAE). This condition derives focus-sensitivity and additivity, and—when exclusion is defined in terms of likelihood—scalarity as well. On this view, the scalarity presupposition of ‘even’ ye is not an independent lexical requirement, but a natural consequence of how exclusion is parametrized.

This account clarifies both the parallels and the contrasts between ye and dou in scalar environments. I argue that the ‘also’–’even’ alternation of ye parallels the distributor–’even’ alternation of dou. Both particles can associate with minimizers, a possibility licensed by their shared scalar inference. At the same time, their difference in additivity yields distinct evaluative flavors as well as different distributional patterns in concessive constructions.

Finally, the talk revisits the “independence” requirement of additive expressions. I argue that independence emerges from the interaction between the VAE requirement and constraints on the QUD. This analysis also accounts for cases in which apparent violations of independence do not result in deviance.

Speaker: Jian-Leat Siah (UCLA), joint work with Sam Zukoff and Feng-fan Hsieh
Title: Resolving Reduplicative Opacity in Malay Nasal Spreading: An Argument for Base–Reduplicant Correspondence Theory
Time: Monday, March 9th, 5pm - 6:30pm
Location: 32-D831

Abstract: In Malay (Austronesian), nasality spreads iteratively and rightward from nasal consonants to following vowels and glides, but is blocked by supralaryngeal consonants. In reduplicated forms, Onn (1976) reported overapplication of nasal spreading (e.g., [w̃ãŋĩ-w̃ãŋĩ]): the first syllable of the reduplicant acquires nasality even though there is no local trigger preceding it. This pattern carries significant theoretical implications because only parallelist (McCarthy & Prince 1995) but not serial/derivational theories of reduplication (Inkelas & Zoll 2005; Kiparsky 2010; McCarthy et al. 2012) can account for it. In this talk, we present acoustic data from 30 native speakers of Malay showing that nasal spreading in reduplication displays substantial variation both within and across individuals. In reduplicated words such as /abaŋ-abaŋ/ ‘brothers’, all logically possible combinations of oral and nasal realizations were attested, including underapplication ([abaŋ-abaŋ]), normal application ([abaŋ-ãbaŋ]), unmotivated “pathological” application ([ãbaŋ-abaŋ]), and crucially, overapplication ([ãbaŋ-ãbaŋ]). Of these, overapplication emerged as the most frequent variant, corroborating Onn’s (1976) descriptive observations and providing support to parallelist theories of reduplication. The study further reveals a phonetic correspondence effect, whereby vowels in the reduplicant and base tend to exhibit matching degrees of nasality/orality. To capture these variable and gradient patterns, we develop a constraint-based model within the framework of generative phonetics, in which constraint violations are assessed scalarly rather than categorically (Flemming 2001; Lefkowitz 2017). The model achieves a strong fit to the experimental data, demonstrating how integrating phonetic detail into a formal grammar can shed new light on longstanding questions at the morphology-phonology interface.

Congratulations to our very recent alum Omri Doron (PhD 2025),, who has received a Marie Skłodowska-Curie fellowship!  He will be leading a two-year project titled “Mapping complexity in Language” under the supervision of Pavel Caha at Masaryk University in Brno, Czech Republic,  Omri is currently a visiting lecturer at UMass Amherst.

Speaker: Dean McHugh (University of Edinburgh)
Title: Conditional Modality with Alternatives
Time: Thursday, March 5, 12:30pm - 2pm
Location: 32-D461

Abstract: This talk brings together two ideas. First, that statements under a modal are interpreted as conditional antecedents. ‘Possibly A’ states that if A were true, there would be some case where the relevant ideals are met. Dually, ‘necessarily A’ states that if A were false, there would be no case where the relevant ideals are met. Second, conditional antecedents are interpreted via sets of alternatives, with some items—such as disjunction and ‘any’—introducing multiple alternatives. Combining them returns, in a uniform and automatic way, a solution to three challenges facing the standard theory of modality: free choice inferences, independence inferences, and counterexamples to substitution of logical equivalents.

Speaker: Junshu Jin and Michael Kenstowicz (MIT)
Title: Perception of English Lexical Stress by 2nd-Language Learners
Time: Monday, February 23rd, 5pm - 6:30pm
Location: 32-D831

Abstract: In this presentation we briefly review two earlier studies on the perception of lexical stress contrasts in English by native Mandarin speakers. We then discuss the findings of Jin & Zheng’s (2025) study comparing the perception of English stress by three groups of speakers: native English (control), native Mandarin, and Yanbian Korean. The results indicate that native English speakers and L2 learners weight pitch cues similarly; however, they differed significantly in the weight of vowel quality and duration cues. For L2 learners, Mandarin speakers weighted vowel quality and duration cues more similarly to native English speakers, and they weighted these two cues heavier, compared to Yanbian Korean Chinese. The study provides support for the Language Transfer Theory and tests Cue-Weighting Theory with implications for L2 phonetic teaching and learning. 

Speaker: David Pesetsky (MIT)
Title: A Sparse Theory of Argument Alternations
Time: Thursday, February 19th, 12:30pm - 2pm
Location: 32-D461

Abstract: Argument alternations such as active~passive are common in the languages of the world, with several stable properties. Consequently, we should seek a maximally sparse account of such alternations that does not stipulate their existence by positing alternation-specific entities such as VoiceP or special probes. Ideally, these alternations and their properties should emerge as by-products of the interaction of more fundamental entities and operations. This talk argues for a sparse theory of argument alternations in that vein, in which two components interact to yield their existence and key properties, building on Collins (2005, 2024) and influenced by Newman (2025) and discussion in last Spring’s seminar co-taught with Peter Grishin. Neither is specific to argument alternations; both merely generalize phenomena already known from other domains.

1. Generalized Dependent Case: The flagging of local c-command relations familiar under the rubric dependent case is not limited to nominals. Non-nominal phrases such as VP may both trigger and receive dependent case as well. This provides an account of morphemes specific to one alternant in an alternation, including passive morphology on V and by-phrase morphology, as well as morphemes that emerge and disappear in ditransitive alternations. Their presence correlates with the presence/absence of hyperlocal VP-fronting.
2. Generalized Insatiability: The “Merge XP” feature on a head may apply more than once to the same element so long as some new c-command relation is created (i.e. so long as something changes). This permits hyperlocal complement-to-specifier movement over an intervening element (contra Abels 2003) – offering a simplification and new characterization of the VP-fronting central to Collins’ accounts of passive and other argument alternations. No alternation-specific feature drives this movement. It is just the same Merge feature that added the complement in the first place.

Speaker: Shigeru Miyagawa (MIT), Despina Oikonomou (University of Crete), Onur Özsoy (University of Cologne), Caroline Heycock (University of Edinburgh), Giorgios Vardakis (University of Padova), Rumeysa Bektaş (Tokat University)
Title: Condition C amelioration effects in wh-movement: An interaction between pronominal type and d-linking
Time: Tuesday, February 10th, 1pm - 2pm
Location: 32-D461

Abstract: While wh-movement has been argued to involve obligatory reconstruction, leading to a Condition C effect (Chomsky 1981; Barss 1986; Lebeaux 1988; Heycock 1995; Fox 1999), experimental studies in English and German have shown that, in wh-NP-movement, the Condition C effect is not robust (Adger et al. 2017; Bruening & Al-Khalaf 2019; Stockwell et al. 2021; Salzmann et al. 2023). Stockwell et al. 2021 and Salzmann et al. 2023 suggest that the (strong/weak/null) type of pronominals may be relevant. Along these lines, we investigate anaphora resolution under wh-reconstruction in Italian and Greek, languages which, unlike English, bear both null and overt person pronouns. We present data which reveal a robust correlation of pronouns and Condition C effects: null pronouns resist coreference consistently whereas overt pronouns allow it. Conceiving this contrast is crucial for determining how restrictive the nature of the Condition C as a grammatical phenomenon is. See https://tinyurl.com/4eha8dws for a longer, NELS 56 abstract.

Speaker: Amir Anvari (MIT)
Title: How to be ignorant
Time: Wednesday, February 11th, 1pm - 2pm
Location: 32-D461

Abstract: I discuss two observations that are puzzling for a rather plausible, pragmatic conception of the etiology of ignorance inferences. The first is that certain sentences do not trigger certain ignorance inferences even in highly favorable contexts (Feinmann 2023). The second is that uninformative sentences cannot be used as acceptable vehicles to convey ignorance. I will argue that the latter observation provides a strong motivation for the claim that relevance is closed under speakers’ beliefs (Fox 2016). This claim, in turn, requires adopting Meyer’s (2013) grammatical theory of ignorance computation. I will then show that the two puzzling observations can be addressed using two independently motivated assumptions about exhaustion.

Speaker: Si Berrebi (MIT)
Title: Category mergers are irrecoverable even with robust distributional evidence
Time: Monday, February 9th, 5pm - 6:30pm
Location: 32-D831

Abstract: Can a covert phonological category be learned based on the distribution, without phonetic evidence? Although this idea was debated extensively, it has yet to be tested whether individual speakers have successfully acquired covert categories. I examine a case from Modern Hebrew in which [ħ] and [χ] have undergone a merger in the majority dialect, yet historical alternations triggered by the pharyngeal are preserved in hundreds of words. Speakers of a minority dialect still produce [ħ] and [χ] as distinct, thus allowing for a direct comparison between the status of /ħ/ among speakers who have phonetic evidence for the distinction, and those who received only phonological evidence. Using a new linguistic game paradigm, validated by a small study of non-merged dialect speakers, I show that merged Hebrew speakers generally cannot represent /ħ/ as distinct from [χ] based on its distribution. I’ll discuss tentative conclusions and future directions for this project.

Course announcements in this post:

• Topics in Syntax (24.956)
• Topics in Experimental Phonology (24.967)
• Topics in Semantics (24.979)

• Instructor: Sabine Iatridou, Elise Nerman, David Pesetsky
• Time: Monday, 2pm-5pm
• Room: 32-D461

This class will explore what is known and what is unknown about the internal and external syntax of nominals. We chose the term “nominals” rather than “NP” or “DP” because the headedness of nominals is in fact a topic of continual debate — a remarkable lack of scientific consensus for such a basic question concerning one of the fundamental building blocks of human language. In contrast to the verbal domain, where there is also debate but also significant (justified) consensus, many other fundamental aspects of nominal syntax remain deeply puzzling. What functional elements enter into their maximal extended projections, and how closely (if at all) does their arrangement parallel what is found in the verbal domain? What can nominalized clauses such as gerunds reveal about the parallels and non-parallels between nominals and clauses? Do nouns take complements analogous to complements of verbs, prepositions, and adjectives? Where do relative clauses fit into the picture, and what laws govern the syntax of modification more generally? Is concord within the nominal an instance of Agree or something else? How should we understand constructions seemingly special to the nominal domain such as construct state, polydefiniteness, and others?

Our plan for the beginning of the semester starts as follows, probably one class per topic:

1. The headedness of nominals

2. Gerunds

3. Construct State

— and then we will be off and running.

Course requirements: active attendance and participation, meetings with instructors to develop final paper, final paper

24.967: Topics in Experimental Phonology

• Instructor: Adam Albright & Edward Flemming
• Time: Wednesday, 10am-1pm
• Room: 32-D461

The field of phonology has increasingly looked to experimental results to confirm and extend its understanding of phonological patterns. In this course, we will examine some of the issues involved in deriving experimentally testable predictions from a theory, designing and running an experiment, and interpreting the results.

The class has several goals:

• Consider the relation between linguistic theory, empirical predictions, and experimental results
• Gain practical knowledge in designing and carrying out experiments in the lab and on-line, and performing data analysis using R
• Gain familiarity with some commonly used experimental paradigms, comparing what they can tell us about the linguistic system

The emphasis this year will be on statistical analysis. The course will be organized around the statistical models that are most applicable to linguistic experiments:

• Linear models and linear mixed-effects models
• Generalized linear (mixed) models: logistic/probit regression, ordinal logistic regression, log-linear models
• Factor coding for interpretable statistical analysis
• Possibly: Bayesian linear models

The application of these models will be illustrated through case studies selected based on the interests of the participants. Candidates include: Coarticulation, perceptual similarity, the P-Map Hypothesis, statistics of the lexicon, wug/blick tests and Universal Grammar/learning biases. Experimental paradigms examined are likely to include production, perceptual identification and discrimination, artificial language learning, and acceptability judgments.

Requirements for students taking the course for credit:

• Readings and class participation
• Regular assignments (modest and practical in nature)

• Instructor: Gennaro Chierchia & Danny Fox
• Time: Thursday, 2:30pm-5:30pm
• Room:
  • Harvard: Boylston G 02
  • MIT: 32D-461

Speaker: Amanda Michel (MIT)
Title: A Variable Account of Norwegian Stress
Time: Tuesday, December 8th, 5pm - 6:30pm
Location: 32-D831

Abstract: The stress system of Norwegian has traditionally been argued to be fixed/predictable with a robust set of exceptions. Much of the analysis of Norwegian stress is based on loanwords, as the inventory of native word shapes is limited. In this talk, I will present my ongoing work with my former advisor (Anya Hogoboom, College of William & Mary). We put forth an alternative account in multiple steps. We first seek to explain stress variability diachronically, looking to the placement of stress on the donor word for a given loanword. We then propose a variable model of the stress phonology utilizing MaxEnt and find that speakers are sensitive to the distribution of stress assignment via a nonce word production experiment.

Speaker: Sabine Iatridou (MIT)
Title: Superlatives meet Definiteness in Bulgarian and Greek
Time: Thursday, December 11, 12:30pm - 2pm
Location: 32-D461

Abstract: The aim of this paper (joint work with Artemis Alexiadou and Roumyana Pancheva) is to enrich the debate on the nature of absolute and relative readings of superlatives with data from Bulgarian and Greek and explore possible conclusions from them. We investigate three phenomena exhibited by definite DPs: polydefiniteness, clitic doubling, and clitic left dislocation, and show that the first two preclude relative readings, while the latter allows relative readings in contexts of contrastive topicalization. This leads us to conclude that the absolute-relative distinction in superlatives is a case of a genuine semantic ambiguity, with definiteness playing a central role, but also that there is an important role for context. Our findings also reveal similarities and differences among clitic doubling, clitic left dislocation, and polydefiniteness, in both Bulgarian and Greek.

On November 22rd and 23rd, Hani Al Naeem, Christopher Legerme, Cora Lesure, Vincent Zu (MIT Chemical Engineering postdoctoral associate), and Jacob Kodner (Harvard Linguistics graduate student) taught over 50 ninth through twelfth grade students at Splash, a weekend extravaganza of courses organized by MIT ESP (Educational Studies Program).

Hani and Christopher offered “Sounds in Motion: Exploring the Science of Speech”; Cora offered “Rhyme and Reason: Exploring the Linguistics of Poetry”; Vincent offered “Linguists vs. Machine: Who Had the Telescope?”; and Jacob offered “The Beauty and Complexity of Language: Introduction to Linguistics”. The courses were designed by each instructor and developed and vetted through a collaborative process. Maya Honda observed all of the classes and attests to the great job everyone did sharing their knowledge and passion for linguistics with the Splash students.

Speaker: Juan Cancel (MIT)
Title: Cross-Categorial Syncretisms: Theoretical Predictions and Empirical Observations
Time: Thursday, December 4th, 5pm - 6pm
Location: 32-D769

Abstract: Syncretisms and their generalizations have been the topic of much discussion in the morphological literature for a few years already (ex: Caha 2009, Starke 2017, Zompì 2023, etc). Nonetheless, a kind of syncretism that hasn’t been addressed much in that same literature are syncretisms that span the paradigms of different lexical categories (ex: nouns and verbs). In this presentation, I will be looking at these ‘cross-categorial syncretisms’ in terms of Distributed Morphology (Halle and Marantz 1993), argue about why we should expect them to exist, and make note of what specific features we would expect to behave as such. Finally, I will go over various examples from language families such as Turkic, Inuit-Yupik-Unangan, and Uralic in order to show not only that they indeed exist, but that they seem to behave in ways that comply with our theoretical frameworks as well.

Speaker: Tam Berulava (MIT)
Title: Case-Matching Effects in Long-Distance Wh-Questions in Georgian
Time: Tuesday, December 2nd, 1pm - 2pm
Location: 32-D461

Abstract: I will present an ongoing project on long-distance wh-questions in Georgian, focusing on the interaction between cross-clausal wh-movement and case. Building on recent claims that Georgian lacks true cross-clausal wh-movement and instead uses only proleptic question constructions, I argue that genuine wh-extraction from embedded clauses is in fact available, but its distribution is tightly constrained. In particular, I show that acceptability systematically tracks (i) the case configuration between the matrix subject and the extracted wh-phrase and (ii) the structural size of the embedded clause, giving rise to robust “case matching” effects in long-distance questions. These effects indicate that the extracted wh-phrase is, in some way, visible to the matrix-level case-assigning algorithm, raising the natural question of how—and why—such cross-clausal visibility is possible.

Speaker: Johanna Alstott (MIT)
Title: A cautionary note on word learning paradigms and presupposition triggering
Time: Thursday, December 4, 12:30pm - 2pm
Location: 32-D461

Abstract: Cross-linguistically, predicates with both initial-state and change-of-state components tend to encode them as presupposition and assertion, respectively. Bade et al. (2024) argue on the basis of a series of artificial word learning experiments that this cross-linguistic tendency reflects conceptual biases privileging changes-of-state over initial-states. In their experiments, they gauged how participants encoded the initial-state and change-of-state entailments of a nonce verb wug, and they interpret their results as suggesting that participants generally encoded the initial-state entailment as a presupposition and encoded the change-of-state entailment as the assertion. This finding, they argue, favors their conceptual-bias hypothesis over competing accounts. In this talk (joint work with Athulya Aravind), we further test the validity of Bade et al.’s paradigm via an additional experiment where we both try to replicate their original effect and, in parallel, ascertain whether their results generalize to a nonce initial-state/change-of-state predicate other than the one that they test. We not only fail to extend Bade et al.’s results to our new nonce word but also fail to replicate their original effect: our participants overwhelmingly treated Bade et al.’s wug and our new nonce word as non-presuppositional. A closer look at Bade et al.’s original studies suggests that non-presuppositional construals were common there, too, and we discuss several reasons why this could have been the case. All told, our outlook is pessimistic: adult artificial word-learning tasks do not, in fact, illuminate the mechanisms of presupposition triggering.

Speaker: Iva Kovač (Vienna/UMass Amherst)
Title: Scope in NPI licensing
Time: Wednesday, December 3rd, 12:30pm - 2pm
Location: 32-D461

Abstract: Licensing of weak NPIs like any is subject to at least three scope-related constraints: certain elements, such as every, may not take scope between the NPI and its licenser (Linebarger 1980, 1987), the NPI must scope below its closest licenser (Homer 2020, Barker 2022), and it must be c-commanded by its licenser on the surface (Ladusaw 1979, 1980). In this talk, I bring these three constraints together by drawing a connection between NPI licensing and conditions that apply to scope taking of regular quantifiers (Fox 1995, Mayr & Spector 2010) and their linear order (Bobaljik & Wurmbrand 2012). I propose that NPI licensing is computed incrementally and explore an implementation in terms of Quantifier Raising and an interplay between spell-out domains and interface (LF and PF) principles regulating copy choice. In brief, under certain clearly defined conditions, NPIs like any disambiguate scope relations by marking narrow scope (Barker 2018), but syntactic domains force them to do so locally. If on the right track, this approach provides an argument in favour of an architecture of grammar where PF and LF domains can be distinct and PF has access to LF.

Speaker: Paul Meisenbichler (MIT)
Title: Reference to individuals across worlds and constraints on de re phenomena (Part 2)
Time: Wednesday, November 26th, 1pm - 2pm
Location: 32-D461

Abstract: I will continue our discussion from November 12 about how certain ideas from counterpart theory (CT, see Lewis 1986) could contribute to our understanding of de re/de dicto phenomena. The central tenet of CT is the ontological assumption that individuals exist in only one world. In CT, reference across worlds must therefore be established in an indirect way (i.e. as a relation between an individual in one world and its counterparts in other worlds). In some of the recent literature, it has been suggested that blocking (direct) transworld reference could help us understand some well-known constraints on transparent/opaque readings (e.g. Percus (2013), Sauerland (2014), Cable (2018)). I want to explore these proposals and discuss whether adopting a counterpart ontology is a move worth pursuing.

Speaker: Heidi Durresi (MIT)
Title: Comparing different predictions of learnability on typology
Time: Monday, November 24th, 5pm - 6:30pm
Location: 32-D831

Abstract: Stanton (2016) is not only an argument for learnability shaping typology, but also that the Gradual Learning Algorithm (GLA; Boersma 1997, Magri 2012) is the vehicle for it. In this talk, I will discuss some preliminary ideas on comparing the typological predictions of alternative learning models on pathological stress patterns. Learning strategies discussed include Error-Selective Learning (Tessier 2007) and Expectation Driven Learning (Jarosz 2015).

Speaker: James Cooper Roberts (MIT)
Title: Part is part (plus pragmatics)
Time: Thursday, November 20th, 5pm - 6pm
Location: 32-D769

Abstract: In this work, I argue that the natural language item part (and its translational equivalents) is the same as the mereological notion of PROPER PART, full stop. While this seems trivially true, state-of-the-art semantics often assumes this is not the case; Link (1983) posits two parthood relations for part of an individual and part of a plurality (MATERIAL vs. ATOMIC parts), and those who assume one relation posit additional constraints on the transitivity of parthood (Moltmann, 1997; Wagiel, 2021). The latter is done in service of the observation that parts of atomic individuals seemingly cannot be part of a plurality. For example, a world where only Jerry Seinfeld’s hand is completely covered in paint is not a verifying case for (1).

(1) Part of the New Yorkers are completely covered in paint.

I will argue against the approaches outlined above, opting to instead include parts of atoms in the denotation of plural partitives. The interpretation we get for (1), I propose, is actually the result of pragmatics rather than semantics (Grice, 1975’s MAXIM OF QUANTITY). This position is bolstered by the fact that the “inclusive” reading of a plural partitive becomes available under negation and epistemic uncertainty (cf. plurals, Sauerland et al. 2005).

[This is a practice talk for my upcoming presentation at OASIS 5.]

Speaker: Rotsuprit Saengthong (MIT)
Title: Clause Size Reduction by Projection Feature
Time: Tuesday, November 18th, 1pm - 2pm
Location: 32-D461

Abstract: I will present an ongoing project on clausal complementation. I have observed that when tested with restructuring diagnostics as used in Wurmbrand (2001), clauses containing functional elements (e.g., C and T) behave like full CPs in some environments, but in other environments, they behave as if they are smaller than expected given the presence of those functional elements. Such variation is evident in Thai control constructions, where infinitival complements contain the same morphemes as full CPs, yet lack the defining properties of CPs. For instance, certain clauses include C and T morphemes but behave syntactically like vPs. These facts suggest that the mechanisms of structure building—specifically Merge and Labeling/Projection—may operate differently in such environments. I propose that in Merge (α, β), there may be what I tentatively call a Projection feature, which determines which head is selected for projection. I further argue that clausal reduction arises as a consequence of this operation.

Speaker: Amy Li (MIT)
Title: A phonetic correlate of velar palatalization: shorter front cavity
Time: Monday, November 17th, 5pm - 6:30pm
Location: 32-D831

Abstract: I will start this talk by practicing presenting the poster that I will bring to the ASA (poster abstract below).

Velar palatalization is a common sound change involving a velar stop becoming a palatal affricate or fricative before a front vocoid. To gain insight into its phonetic mechanisms, we test two hypotheses about factors conditioning velar palatalization by comparing similar languages with and without the change: (1) the conditioning vowels are fronter in languages with the change, resulting in fronter closure of the velar through coarticulatory assimilation; or (2) languages with the change have stronger coarticulation of velar stops with following segments, resulting in greater fronting preceding front vocoids. Specifically, we compare two Chinese languages, Mandarin, which underwent velar palatalization in the 16th-17th centuries ({k, kh, x} became {tɕ, tɕh, ɕ} before high front vowels), and Cantonese, which has not undergone the change in the last millennium. Our results support hypothesis (1) but not (2). We find that Mandarin speakers produce [i] and [y] with a higher front cavity resonance (third formant for [i] and second formant for [y]), implying a shorter front cavity. Velar coarticulation, measured by locus equation slope, does not differ significantly between the two languages. This suggests that the phonetic preconditions of velar palatalization lie in an especially front articulation of the conditioning vocoid.

Then, I will discuss two new hypotheses in response to my results: (1) the conditioning vowels have small enough front cavities in languages with the velar palatalization change that that they can be considered coronal, so coarticulation of the velar consonant with this coronal vowel results in the coronal output of the change; and (2) the conditioning vowels have shorter front cavities in languages with the change due to longer tongue constrictions, which facilitate frication given coarticulation with the velar consonant. I will share some of my attempts to test these new hypotheses. Finally, I will present my planned next steps of the project, including making new recordings comparing (some varieties of) Cretan Greek, in which {k, kh, x} became {tɕ, tɕh, ɕ} before [i] and [e], and standard modern Greek.

Speaker: Daniar Kasenov (NYU)
Title: Nonce word wellformedness and abstract URs: the case of Russian yers
Time: Thursday, November 13th, 5pm - 6pm
Location: 32-D769

Abstract: Nonce word studies are part of the toolbox to probe productivity, especially of non-automatic phonological alternations, such as Russian vowel-zero alternations (Russian yers). Existing work (Gouskova, Becker 2013; Becker, Gouskova 2016) shows that Russian speakers extend phonotactic tendencies regarding which words do and do not undergo the alternations to nonce items. Becker and Gouskova argue that the results support Gouskova’s (2012) diacritic-based account of Russian yers against approaches that rely on abstract contrasts between vowels. In this talk, I wish to explore how a proponent of the abstract UR approach might account for Becker and Gouskova’s results without ignoring the experimental results altogether. I present preliminary results that a simple bigram model over URs and SRs might do the trick (cf. Scheer’s 2019 argument that the effects reported by Becker and Gouskova are “lexical”).

Speaker: Paul Meisenbichler (MIT)
Title: Reference to individuals across worlds and constraints on de re phenomena
Time: Wednesday, November 12th, 1pm - 2pm
Location: 32-D461

Abstract: In this LFRG, I want to lead a (mostly informal) discussion on the role that counterpart theory (CT, see Lewis 1986) should play in our approaches to de re/de dicto phenomena. The central tenet of CT is the ontological assumption that individuals exist in only one world. In CT, reference across worlds must therefore be established in an indirect way (i.e. as a relation between an individual in one world and its counterparts in other worlds). In some of the recent literature, it has been suggested that blocking (direct) transworld reference could help us understand some well-known constraints on transparent/opaque readings (e.g. Percus (2013), Sauerland (2014), Cable (2018)). I want to explore these proposals and discuss whether adopting a counterpart ontology is a move worth pursuing.

Speaker: Ogloo Jurkhaichin (MIT)
Title: The Nature of ‘Edge’: Evidence from Cross-clausal A-movement in Mongolian
Time: Thursday, November 6th, 5pm - 6pm
Location: 32-D769

Abstract: Syntactic operations are bounded by phases, in which the edge is typically taken to be the highest specifier (Fox & Pesetsky 2005; Rackowski & Richards 2005; Bošković 2016, a.o.). In this talk, I will argue that the edge need not be only the highest specifier; a lower specifier of the clausal periphery may also act as an escape hatch for further syntactic movement. This is evidenced by the novel observation that Mongolian permits cross-clausal A-movement to escape phases in which the highest specifier is an Ā position (contra Gong 2022, 2023). In particular, given the Ban on Improper Movement, I propose to posit a lower A-specifier that facilitates subsequent cross-clausal A-movement.

Speaker: James Morley (MIT)
Title: An “Only-You” restriction in Chamorro and the problems it poses for the theory of hierarchy effects
Time: Tuesday, November 4th, 1pm - 2pm
Location: 32-D461

Abstract: This talk investigates a person-animacy restriction - henceforth PAR - in Chamorro (Malayo-Polynesian; Austronesian), previously reported in Chung (1998, 2014, 2020) but which has otherwise been subject to little theoretical attention. Chamorro’s PAR prohibits internal arguments from ‘outranking’ external arguments with respect to the hierarchy in (1).

Chamorro-specific person-animacy hierarchy 2nd person > 3rd person animate pronouns > 3rd person animate lexical nouns > Inanimate

Chamorro instantiates what Stegovec (2019, i.a.) calls a *3>2 or “Only-You” person restriction: although 2nd and 3rd person have their distributions constrained by the restriction, 1st person does not. In this talk I make two kinds of argument. First, I argue that this restriction should not be reduced to a language-specific morphological restriction, pace Chung (2014), but should rather be treated as (at least partly) syntactic in nature. The null hypothesis is thus that it should be explained by the same mechanisms postulated elsewhere to capture other syntactic PARs. Second, I argue that this has not been achieved. More specifically, current theories of PARs are either logically incompatible with the Chamorro data, or else incur conceptual or empirical problems when amended to accommodate it. I then sketch some preliminary ideas about how to go about solving this.

Speaker: Thomas Truong (MIT)
Title: Plural superlatives and cumulativity
Time: Wednesday, November 5th, 1pm - 2pm
Location: 32-D461

Abstract: In this talk I will present some ongoing work on the interactions between plurals and superlatives.

To do so, I examine a unique reading of sentences containing plurals and superlatives.

(1) Rafa climbed each of the tallest mountains that his students climbed.

Under one reading of (1), the truth conditions require us to look at each of the students, check which mountains each student climbed, and then take the tallest mountain climbed for each of the students.

I will break down this construction. I claim that this reading is generated as a case of cumulative readings, where the superlative operator is in the scope of the cumulativity operator.

I show that if we assume the ** operator (Krifka 1986, Sternefeld 1998, Beck and Sauerland 2000) to derive cumulativity along with an account of superlatives following Heim (1999), covert movement outside of a relative clause seems to be necessary to derive the correct LF for the relevant interpretation of sentence (1).

Speaker: Gasser Elbanna (Harvard)
Title: A model of speech recognition reproduces behavioral signatures of human speech perception and reveals mechanisms
Time: Monday, November 3rd, 5pm - 6:30pm
Location: 32-D831

Abstract: Humans dexterously extract meaning from variable acoustic signals and can faithfully repeat back novel utterances—hallmarks of spoken communication. Speech perception is thought to subserve these downstream tasks via transforming sound into robust perceptual representations. Yet progress on the nature of these representations and their mechanisms has been limited by the lack of (i) stimulus-computable models that replicate human behavior and (ii) large-scale behavioral benchmarks for comparing model and human speech perception. In this talk, I will present our work on developing candidate artificial neural network models of human speech perception along with new behavioral experiments to compare phonetic judgments in humans and models. Our models reproduce patterns of human responses and confusions alongside recapitulating key behavioral signatures of human speech perception. I will also show how our models enable us to investigate the role of contextual integration and its directionality in speech perception.

Speaker: Cooper Roberts (MIT)
Title: A rational solution to an agreement-interpretation puzzle
Time: Thursday, November 6 12:30pm - 2pm
Location: 32-D461

Abstract: In some Indo-European languages, a fraction partitive (FP) which embeds a plural DP licenses an optional-agreement phenomenon—-in the appropriate syntactic position, an agreeing predicate can copy the features of either the fraction (1b) or the complement (1a). This is puzzling enough if we assume FPs have a DP-within-DP structure (as do Ionin et al., 2006; Benbaji-Elhadad & Wehbe, 2024; a.o.), which under a Locality-governed model of Agree (Chomsky, 1995) would predict that the fraction is the sole target of Agreement. The plot thickens when we observe that the two agreement possibilities yield different interpretations. (1a) is true in a world where, for example, two out of six walls are covered in mold (I call this the COUNT reading). (1b), on the other hand, is true in a world where, given a plurality of walls which have a cumulative surface area of 12m^2, 4m^2 are covered in mold (a MEASURE reading).

(1) [Italian]
a. un terzo delle pareti sono coperti di muffa
‘A third(m.sg) of the walls(f.pl) are covered(f.pl) by mold’
*MEASURE, COUNT
b. un terzo delle pareti `e coperto di muffa
‘A third(m.sg) of the walls(f.pl) is covered(m.sg) by mold’
MEASURE, *COUNT

The goal of this study is to give a theoretical account of the alternation in Italian-like languages while also explaining why some languages in the family lack the equivalent to (1b) (American English). Following the tenet of One Form/One Meaning, I pursue an analysis where measure and count FPs are structurally-distinct. Specifically, I assume that count FPs are the structurally-simpler of the two, bearing a syntax where the complement is actually the head (see Selkirk 1977) and the semantics are s.t. cardinality functions win over other measurement possibilities (Barner & Snedeker, 2005; Bale & Barner, 2009; Wellwood, 2019; Wagiel 2021). To get the measure FP, I posit a special operator TOTAL which takes the bare FP structure and makes two important contributions. First, TOTAL re-merges the fraction via projecting movement (Bhatt, 2002) to make it the new head of the structure. Second, TOTAL changes the “matrix” parameter of evaluation for measure functions to one where cardinality will lose to other forms of measure. Crucial evidence from this proposal comes from Russian, where FPs which include the part-word chast’ lose the agreement-optionality and necessarily have a measure reading. I interpret this item as a realization of TOTAL and conclude that it can be optionally-overt in some languages.

Gĩkũyũ has two types of backcopyingː (1) optional backcopying (i.e., backcopied and non-backcopied variants are in free variation) when prenasalized stop formation (/Nt, Nɾ/ → [ⁿd], /Nʃ/ → [ᶮdʒ], /Nk, Nɣ/ → [ᵑg]) and foot reduplication interact; (2) obligatory backcopying when prenasalized stop formation, foot reduplication, and Meinhof’s law (NC₁…N₂(C) → N₁…N₂(C)) interact.  To account for the obligatory backcopying cases, I propose that the distinguishing factor lies in the onset’s ratio of nasality in the base vs. the reduplicant. On a ratio-of-nasality scale of [t, k, ɾ, ʃ, ɣ] = 0, [ⁿd, ᶮdʒ] = 1, [n] = 2, a Distantial Faithfulness constraint penalizes segments in RED that differ too much from their Base correspondents. Consequently, Base-Reduplicant pairs like [ᶮdʒ]…[ʃ] are acceptable because their nasality distance is ≤1, whereas [n]…[ɾ] is disfavored because the ‘distance’ is 2.

Speaker: Yvette Yi-Chi Wu (Harvard)
Title: Verb classes and affix ordering in Seediq
Time: Thursday, October 30th, 5pm - 6pm
Location: 32-D769

Abstract: This talk looks at verbal morphology in Seediq, with supplementary data from other Formosan languages. I will focus on the ordering of “voice” morphology with respect to derivational and TAM morphology, which allows us to examine the interactions of infixation, reduplication, stress-conditioned suppletion, and more. I show that the actor voice (AV) infix is located in the middle field (below perfective Asp and above Voice), which goes against theories where voice is high (e.g. in T or C). I also attribute apparent allomorphy of AV to regular argument structural morphology in Seediq (cf. Ross 1995, Chen 2020), and discuss the implications this has on verb classes and (anti-)causative structures.

Speaker: Vsevolod Masliukov (MIT)
Title: Participial Complementation in Russian
Time: Tuesday, October 28th, 1pm - 2pm
Location: 32-D461

Abstract: In this talk I will present a syntactic analysis of Russian sentences such as (1), which feature participles marked with the so-called ‘Predicate instrumental case’ (Bailyn 2001). I will argue that these participial clauses are arguments with a PredP (small clause structure, Bowers 1993 et seq.) built on top, whose subject is raised to the structural case licensing position in the matrix clause. I will also compare the properties of these constructions with participles used as adnominal adjuncts, which always require case matching, (2) and finite complements of the same class of verbs (3).

(1) Ona nikogda ne vide-l-a Maš-u plač-ušč-ej. she never NEG see-PST-F.SG M.-ACC cry-PTCP.IPFV-F.SG.INSTR ‘She has never seen Masha crying.’

(2) Ona nikogda ne vide-l-a [plač-ušč-uju Maš-u] . she never NEG see-PST-F.SG cry-PTCP.IPFV-F.SG.ACC M.-ACC Lit.: ‘She has never seen crying Masha.’

(3) Ona nikogda ne vide-l-a, [kak Maša plač-et]. she never NEG see-PST-F.SG COMP M.NOM cry-NPST.3SG ‘She has never seen Masha cry.’ (lit.: ‘how Masha cries’)

Speaker: Bergül Soykan (MIT)
Title: Limitations on meta questions: insights from Turkish
Time: Wednesday, October 29th, 1pm - 2pm
Location: 32-D461

Abstract: In this ongoing work, I present data on the basic patterns of Turkish meta questions (MQs) and show that Turkish appears to allow meta-meta questions as in (1), which Trinh, Fox, and Bassi (2025) predict to be problematic according to their recent NELS talk. They claim that speech act verbs are phase heads, and complements of phase heads are spelled out (á la Chomsky 2001), which implies that there can be only one silent speech act verb in every (audible) sentence. (1)   A1: Ela gel-ecek mi? come-Fut PolQ [IA ask [whether Ela will come]]      B1: Ela nere-ye gel-ecek mi? wh-MQ where-Dat come-Fut PolQ [where1 youB ask [whether Ela will come where1]] A2: Ela nere-ye gel-ecek mi mi? pol-MQ *[whether youB ask [where1 IA ask [whether Ela will come where1]]]

I suggest that the difference between Turkish and English MQs results from 1) their distinct question-formation strategies (e.g., English uses wh-phrases in situ only for MQs, while in-situ-ness is the default form of all questions in Turkish) and 2) from differences in their Spell-Out domains. Given that Turkish always uses wh-in-situ for constituent questions, I propose that C is the phase head in Turkish, not the speech-act verb (or any other v). Also, revising Krifka (2014) and Woods and Vicente (2021), I assume the following structure in (2) for Turkish questions, where Force P indicates whether an utterance has question force [+Q] or not [-Q], and say is the speech act for all utterances that is merged when needed. Thus, I offer a solution to the seeming issue in Turkish MQs along the lines of Trinh, Fox, and Bassi’s (2025).

(2)   A1: Ela gel-ecek mi? come-Fut PolQ [whether1 [Ela will comeF mI1 ForceP[+Q]] Cthat] B1: Ela nere-ye gel-ecek mi? wh-MQ where-Dat come-Fut PolQ [where2 [whether1 [Ela where2 will comeF mI1 ForceP[+Q]] Cthat] sayA] A2: Ela nere-ye gel-ecek mi mi? pol-MQ [whether3 [where1 [whether1 [Ela where1 will comeF mI1 ForceP[+Q]]F mI3 Cthat] sayA]ForceP[+Q]]

Later, I provide evidence from embedded clauses with an overt C head for the interpretation layers of MQs in Turkish. Finally, if time allows, I share cross-linguistic data from various languages, supporting that the Spell Out domain and question-formation strategies are the two parameters that affect MQ layering.

Speaker: Ivy Sichel (UC Santa Cruz)
Title: How resumptive pronouns ameliorate island violations - evidence from Hebrew
Time: Thursday, October 30, 12:30pm - 2pm
Location: 32-D461

Abstract: While some resumptive pronouns (RPs), in some languages, ameliorate island violations (Hebrew, Irish, Lebanese Arabic, among others), not all do (Vata (Koopman 1982), Welsh (Tallerman 1983), Nupe (Kandybowicz 2007)). The variation in this domain is perplexing: what is it that allows RPs to ameliorate island violations in the languages in which they do, and how might this be parametrically constrained? According to the ‘Classic View’ (Chomsky 1977, Borer 1984, McCloskey 1990), the repair is only indirect: RPs are associated with a distinct, non-movement, derivation (1a), in which an island violation is simply not incurred. By now, however, it has become clear that RPs are at least sometimes compatible with movement derivations, even in languages which exhibit island repair (Aoun, Choueiri & Hornstein 2001, Sichel 2014). This raises the possibility that an RP that repairs an island violation is similarly compatible with movement, and that the repair is direct, due to the realization of the gap as an RP within a movement derivation (Perlmutter 1972, Pesetsky 1998). The present study argues for the ‘Classic View’ based on the syntax of RPs in high positions in Hebrew RCs - because in this situation, the RP is not realizing the gap at the foot of the chain. This suggests that languages with island-sensitive RPs lack the non-movement RC structure. Along the way, I will develop a particular view of movement-compatible RPs, according to which RPs directly realize a position in an A-bar movement chain, regulated by Economy of Pronunciation.

Speaker: Filipe Kobayashi (Paris Lodron University of Salzburg)
Title: Syntactic constraints on fake indexicals in relative clauses (joint work with Caroline Gardner, Franziska Keller, Anita Riedl and Susi Wurmbrand)
Time: Thursday, October 23rd, 5pm - 6pm
Location: 32-D769

Abstract: Fake indexicals (i.e., bound occurrences of 1st and 2nd person pronouns) have a much more restricted distribution when they are bound by relative pronouns than when they are directly bound by an indexical pronoun (Kratzer 2009; Charnavel, Meadows and Sportiche 2025). Contra previous claims that the licensing of fake indexicals in these sentences is tied to verbal agreement or subjecthood, we argue for a new generalization: a relative pronoun can bind a fake indexical only if it is the highest argument in its clause. We then propose an account of this generalization grounded on a particular theory of how relative pronouns are able host person features.

Speaker: Ioannis Katochoritis (MIT)
Title: Micro-variation is a conspiracy: Condition C, Case and Wholesale Late Merge in Balinese vs. Malagasy
Time: Tuesday, October 21st, 1pm - 2pm
Location: 32-D461

Abstract: (joint work with Lena Lohninger)

This talk aims to illustrate the idea that microvariation is an epiphenomenonal conspiracy of the derivation. The focus will be two Malayo-Polynesian languages, Balinese and Malagasy, which share the Austronesian voice/pivot system: one (any) argument per clause is promoted to a syntactically and pragmatically salient status by becoming the so-called pivot.

When the internal argument becomes the pivot, the so-called Object Voice forms a non-canonical passive: the internal argument A-moves high to map into the surface subject, while the agent is not demoted to oblique, but remains a syntactically core argument that is licensed via strict right-adjacency to the verb.

However, the two languages differ with respect to Condition C: while Balinese pivot movement obligatorily reconstructs, and thus feeds connectivity, Malagasy pivot movement does not obligatorily reconstruct, and can thus bleed connectivity. To account for this apparent microparameter, we suggest that, while Malagasy allows Wholesale Late Merge (Takahashi & Hulsey 2009) of the pivot’s restrictor to the high landing site, Balinese does not.

We then provide two possible reasons for this: first, Balinese Object Voice still assigns accusative case to its internal argument, while Malagasy has lost this capacity, with the pivot being only licensed with nominative at the landing site; this forces pivots in the former, but not in the latter, to be early merged entirely at their base position. Evidence will come from asymmetries in multiple extraction, resumptive pronouns, quantifier scope and the inventory of Object Voice morphemes. Second, Balinese, but not Malagasy, exhibits DP-internal head movement to derive the suffixal nature of its determiner, which would be completely countercyclic under any tenable formulation of the Extension Condition.

Speaker: Giuseppe Varaschin (Humboldt-Universität zu Berlin)
Title: From descriptions to epithets: A structurally uniform account of definites
Time: Wednesday, October 22nd, 1pm - 2pm
Location: 32-D461

Abstract: Slurs are nominal expressions that contribute meaning on both the truth-conditional (TC) and use-conditional (UC) dimensions, functioning as constraints on utterance contexts (Kaplan 1999; McCready 2010; Gutzmann 2015, i.a.). When such expressions appear as complements of definite determiners, as in the Brazilian Portuguese example in (1), they exhibit a systematic ambiguity (Saab 2022). On one reading — the definite-description reading — the DP refers to the unique individual satisfying the property denoted by the NP (in this case, being rural). On the other — the epithet reading — the descriptive property need not hold; it suffices that the referent have epistemically accessible properties that are contextually inferable from that property (e.g. driving a tractor, having a rural accent, wearing rural clothes). (1) O jeca chegou.   the JECA arrived   ‘The hillbilly arrived.’

Prior analyses derive this split via structural ambiguity inside the DP (Patel-Grosz 2014; Orlando & Saab 2020, i.a.). In this talk I defend a structurally uniform alternative, where the interpretive differences between descriptive and epithet readings follow from two main ingredients: (i) an optional type-shifting operation — encoded by a syntactic EX feature (Gutzmann 2019) and correlated with prosodic deaccenting (Umbach 2002) — that moves NP content from the TC to the UC dimension; and (ii) the presence or absence of a pronominal index associated with the D head (Schwarz 2009; Jenks & Konate 2022). Under the epithet derivation, the NP property is mapped to a contextually determined superset that serves as a constraint on assignments for the index introduced by D; this mapping to a superset (defined by epistemically accessible properties) accounts for the misdescription tolerance characteristic of epithet readings. Because the property is shifted from the TC to the UC dimension where it is a function on contexts, it cannot address relevant QUDs. I show that the same interpretive mechanism also accounts for referential readings of definite descriptions (Donnellan 1966; Gutzmann & McCready 2014, i.a.). Drawing on novel Brazilian Portuguese data involving modification, iteration, binding, and ellipsis, I argue that this analysis is not only conceptually simpler but also more empirically adequate than structural-ambiguity accounts. A key consequence is that DP structure is essentially uniform across definites of all types (slurs, ordinary definites, epithets, and referential pronouns), with variation reducible to fully interpretive lexically-anchored features: whether D supplies a pronominal index, and whether triggers of UC-shifting (correlated with deaccenting) are present on NP nodes.

Speaker: Bingzi Yu (MIT) and Michael Kenstowicz (MIT)
Title: Tones and tone sandhi in Chengdu Mandarin
Time: Monday, October 20th, 5pm - 6:30pm
Location: 32-D831

Abstract: Chengdu Mandarin is a variety of Mandarin Chinese spoken in southwestern China. Like Standard Mandarin, it has four citation tones, but the tone contours and sandhi patterns differ notably. The table below summarizes the etymological correspondences between tones in the two dialects:

In this talk, we will present acoustic description of both the citation tones and tone sandhi in Chengdu Mandarin based on recent recordings and propose a constraint-based account that captures most of the observed sandhi patterns. In addition, we raise some interesting yet unsolved puzzles. Among them, we will focus on the phonological representation of the Chengdu dipping tone (T4 in the table above) and its unusual sandhi behavior.

Speaker: Ido Benbaji-Elhadad (MIT) & Omri Doron (UMass Amherst)
Title: Saving FACE: Fragment answers, copy theory, and radical trace conversion
Time: Tuesday, October 14th, 1pm - 2pm
Location: 32-D461

Abstract: Jacobson (2016) observes that fragment answers to which-questions seem to have different felicity conditions than fully spelled out answers and even ones with VP ellipsis, as demonstrated in (1). Jacobson takes this fact as evidence against an analysis of Fragment Answers as the result of Clausal Ellipsis (FACE, in short; Merchant, 2005), and proposes an analysis that treats them as connected to their preceding questions via a special syntactic relation. But FACE has its merits: (i) It explains why fragments seem sensitive to constraints on movement and morphosyntactically connected to their hypothesized “original position”; and (ii) it unifies fragment answers, VP ellipsis and sluicing under the same recoverability conditions. In this talk, we propose a novel account of Jacobson’s puzzle which does not give up FACE, and instead points at the parallelism constraint on ellipsis as the source of infelicity in (1a). Specifically, assuming trace conversion (Fox, 1999), we propose that the mismatch between the trace left by which Math professor and the one left by Mary prevents them from satisfying parallelism. In doing so, we argue for a modification to the process of trace conversion.

(1) Q: Which Math professor left the party early?
a. #Mary, but she’s not a Math professor.
b. Mary left the party early, but she’s not a Math professor.
c. Mary did, but she’s not a Math professor.

Speaker: Lorenzo Pinton (MIT)
Title: VP-ellipsis and parasitic gaps: observations on temporal modifiers and parallelism
Time: Wednesday, October 15th, 1pm - 2pm
Location: 32-D461

Abstract: As the title suggests, in this LFRG I present ongoing work on PV-ellipsis and Parasitic Gaps. I discuss novel data involving parasitic gaps (henceforth PGs) in VP-ellipsis constructions. In particular, I present a series of minimal pairs and claim that the grammatical contrasts are mainly accounted for by the combination of a local parallelism condition on ellipsis in the style Takahashi and Fox (2005)/Fox and Katzir (2011), and Nissenbaum’s (1998; 2000) theory of PGs. First, the contrast between (1a) and (1b) is introduced to show that adjuncts with a PG cannot appear in the clause containing ellipsis, in contrast with adjuncts with no gap:

(1) a. *This is the song that John listened to, and this is the song that Mary did [listen to] after reading about. b. This is the song that John listened to, and this is the song that Mary did [listen to] after having a coffee .

More data in (2) show that there are actually instances in which a PG-adjunct can appear in the elided clause, but only if another PG-adjunct is present in the antecedent:

(2) a. This is the song that Rosa listened to after reading about, and this is the song that Karl did [listen to] before hearing about. b. *This is the song that Rosa listened to after turning on the radio, and this is the song that Karl did [listen to] before hearing about .

A third - semantic - constrast will also be introduced tin the talk. After considering observations from Hartman (2011) on the size of elided phrases, I will show the need for a silent element that merges in the same position as adjuncts that do not contain a PG. This need can be implemented either with a tense restrictor analysis of temporal modifiers (as for example presented in von Fintel and Heim (2011)) or with a silent pronominal version of temporal adjuncts, as long as this is restricted to non-PG adjuncts.

Speaker: Janos Egressy (UCLA)
Title: Size-sensitive Sequence of Tense in Hungarian
Time: Tuesday, October 7th, 1pm - 2pm
Location: 32-D461

Abstract: In Sequence of Tense (SOT) Analyses (e.g. Ogihara 1995), it has been argued that past-under-past configurations (e.g. John said that Mary was pregnant) have a simultaneous reading if the embedded past head undergoes “deletion” under agreement with the matrix past head: While the deletion does not affect the verb form, the reading becomes John said: ‘Mary is pregnant’. If no such deletion happens, the embedded past tense remains backshifted compared to the matrix past tense, so the reading remains John said: ‘Mary was pregnant’. Previous SOT research has claimed that the availability of this deletion rule is a parameter for entire languages, e.g. English has such a rule but Japanese does not, so the simultaneous reading is never available for Japanese clauses. This paper shows that Hungarian exhibits language-internal SOT-variation in past-under-past configurations: While speech-reporting clauses are Japanese-like, i.e. obligatorily backshifted, non-speech-reporting clauses are English-like, i.e. they have a simultaneous reading. My syntactic analysis proposes a size-difference between the clause types: speech-reporting clauses are structurally larger. Hence, the availability of deletion-under-Agree can be described as a case of Williams Cycle (Williams 2003) or size-dependent opacity in Agree (Keine 2020).

(This talk is a practice talk for NELS.)

Speaker: Yiannis Katochoritis (MIT)
Title: Long-distance agreement on Procrustes’ bed: a revival of spec-head
Time: October 9, 12:30pm - 2pm
Location: 32-D461

Abstract: Long-distance agreement (LDA) refers to the phenomenon where a matrix verb may (optionally) agree in φ-features with an embedded argument of its complement clause, otherwise showing default agreement with the entire complement clause. Spanning a clausal boundary and appearing to involve no displacement of the embedded goal, LDA has been one of the core empirical arguments in favor of the operation Agree, which dissociates agreement from movement.

This talk has a double aim: first, to present a previosuly unnoticed observation; that is, (almost) all geographically diverse LDA languages share fundamental typological properties: they are verb-final, with a (full or split) ergative alignment, and with default SOV order; moreover, LDA-transparent embedded clauses are always either restructuring VPs or nominalized/participial TP/CPs; and once a potential LDA goal somehow overtly fronts to the left of the matrix verb, LDA (usually) becomes obligatory.

Second, I attempt to account for this empirical picture through little bits of Kayne’s (1994) Antisymmetry, Bobaljik’s (2008) m-case hierarchy, and the magic power of pied-piping. Specifically, I will suggest a unified analysis of LDA as local feature-checking in a (covert) spec-head configuration, which may allow for iterative feature-copying under recursive merge. Depending on two parameters, the specifier of the agreeing verbal head is filled either directly by the moving DP goal or by the complement clause that pied-pipes the goal at its edge. Hence, LDA is derived via Merge, not Agree. This will eventually yield a typology of (at least) four types of LDA systems, among which I will (struggle to) intergrate the outlier named Algonquian.

The analysis hinges on two crucial premises: one, LDA often features movement properties/constraints; two, the optionality of LDA is only apparent, as it obligatorily interacts with scope and/or information-structure, reducing to the reflex of independently motivated syntactic processes. Thus, I will distinguish between two types of agreement: narrow-syntactic agreement feeds LF or further syntactic operations, and involves downwards valuation by some interpretable feature, (plausibly) under spec-head. This contrasts with uninterpretable agreement, which is pushed post-syntactically, at early PF, and may operate from a distance, via upwards valuation (though still c-command-based).

Speaker: Zhouyi Sun (MIT)
Title: Magri and Anttila (submitted), “Probabilistic phonology is intrinsically categorical”
Time: Monday, September 29th, 5pm - 6:30pm
Location:32-D831

Abstract: We say that an implicational universal (x, y) → (x̂, ŷ) holds for a (probabilistic) phonological grammar G if the probability of realizing the underlying form x̂ as the surface form ŷ is at least as high as realizing x as y (G(y | x) ≤ G(ŷ | x̂)). I’ll try to present the general technique in Appendix F of the manuscript for exploring principles governing implicational universals in MaxEnt phonology, and then discuss two resulting properties developed in §§4–5, both of which are argued to be problematic.

Speaker: Elise Newman, Cora Lesure, Norvin Richards, Cooper Roberts (MIT), Peter Grishin (Brown University)
Title: What’s new with the Passamaquoddy working group
Time: Thursday, October 2, 12:30pm - 2pm
Location: 32-D461

Abstract: In this talk, we (the Passamaquoddy working group) will each present a part of our research on the Passamaquoddy language, based on our most recent findings from our summer field trip. Topics include: argument-introduction, group-formation, agreement, coordination, and our language revitalization efforts.

Speaker: Yvette Yi-Chi Wu (Harvard)
Title: Morphological decomposition in Austronesian voice: evidence for intermediary movement
Time: Tuesday, September 23rd, 1pm - 2pm
Location: 32-D461

Abstract: This talk takes a morphological starting point to Austronesian voice, and argues in favor of positing several distinct syntactic projections whose various combinations give rise to surface voice alternations (cf. Pearson 2005, Travis 2010). In particular, this includes 1) the functional projection E(vent), responsible for (non-)finite and (ir)realis marking, and 2) functional projections M(iddle)T(opic) (cf. LaCerda 2020) and Appl (cf. Georgala 2012), responsible for argument advancement as a sort of “leapfrogging” movement (Bobaljik 1995). Evidence comes from affix ordering of verbal morphemes, morphological decomposition in indicative and irrealis voice paradigms, and the distribution of voice markers in clausal complementation in Seediq and other Formosan languages. This is ongoing work as part of a dissertation on Austronesian voice in Seediq [Taiwan; Atayalic], and its implications on movement within the extended Voice domain.

Speaker: Juan Cancel (MIT)
Title: A Reanalysis of Syllabic and Rhythmic Gradation in Nganasan As A Single Consonant Gradation Process
Time: Monday, September 22nd, 5pm - 6:30pm
Location: 32-D831

Abstract: The literature on Avam Nganasan (Helimski 1998; Wagner-Nagy 2018) has argued that the language exhibits two kinds of Consonant Gradation: Syllabic Gradation (SG), in which an onset consonant alternates in voicing depending on whether a syllable is open or closed as in ɟi.kə ”mountain” vs. ɟi.gə-ʔ ”mountains”, and Rhythmic Gradation (RG), in which an onset consonant alternates in voicing depending on whether it aligns with the left-edge of a foot or not as in (nɨ-rə)(gɨ) ”woman-like” vs. (ɲotə)-(rəku)”grass-like”.

In contrast to the description above, this analysis will argue in terms of Optimality Theory (Prince and Smolensky 2004) that we should instead reanalyze SG and RG as a single Consonant Gradation process. In particular, it will argue that Consonant Gradation in Nganasan involves a special fortition process that occurs on the onset of a foot-medial open syllable in combination to other cross-linguistically attested lenition processes such as intervocalic and postnasal voicing This reanalysis not only clearly explains why the single consonant alternations in both in SG and RG happen to be the same - they would just be the same instantiation of intervocalic voicing - but it also serves as the ground for reanalyzing in more concrete terms some of the problems involving the homorganic nasal alternations that Vaysman (2009) looked at such as the apparent NT → T → D chain-shift that we see in RG or the cases of counterbleeding that we see in words such as (ɟi.gə)-(tə.nu) ”mountain-LOC.SG”, where we have a voiced onset in an unexpected environment.

Speaker: Elise Newman (MIT)
Title: Some thoughts on clefts and wh-fronting
Time: Tuesday, September 16th, 1pm - 2pm
Location: 32-D461

Abstract: To kick off the new semester of syntax square, I want to explore a question: why do some languages primarily use clefts for question-formation when wh-fronting should be a possibility? I don’t know the answer, but I’m hoping that by making the question more concrete, we can start to see a path towards an eventual solution. I’ll explore questions in two languages, Tagalog and Passamaquoddy, which both have the following in common: they use different strategies for DP-questions than non-DP questions. Tagalog uses a pseudo-cleft strategy for DP questions but a regular fronting strategy for non-DP questions; Passamaquoddy uses different clause type for each one. I’ll revisit some properties of clefts and relatives clauses in each case and show that while the behavior of and restrictions on non-DP wh-phrases are fairly straightforward to understand, the behavior of DPs is less obvious.

Speaker: Yurika Aonuki (MIT)
Title: On the QUD sensitivity of a third reading
Time: Wednesday, September 17th, 1pm - 2pm
Location: 32-D461

Abstract: We provide the first experimental evidence for the hypothesis that transparent readings are sensitive to the QUD (Question Under Discussion) (Schwager (Kaufmann) 2009; Percus 2020; Benbaji-Elhadad 2023; Mayr and Schmitt 2023, 2024), using third readings (Fodor 1970) in attitude reports. In particular, we test Mayr and Schmitt’s (2023, 2024) claim that there is difference in transparent-reading-licensing abilities even among QUDs about the attitude holder’s mental state. Our results have consequences for (replacement) theories of transparent readings and shed light on how a QUD about someone’s mental state can be manipulated.

The 12th TripleA workshop for semantic fieldworkers was held at the Tokyo University of Foreign Studies, Japan on September 10-12, 2025. The following members of our community presented at the conference:

• James Cooper Roberts (3rd year), Yizhen Jiang (3rd year) & Haoming Li (4th year): How to talk about groups in a language without group nouns: The case of Passamaquoddy
• Junri Shimada (PhD 2009)[Meiji Gakuin University] & Abdul-Razak Sulemana (PhD 2021)[University of Ghana]: Embedded tenses in Bùlì
• Tim Jantarungsee, Christopher Davis & Yasutada Sudo (PhD 2012)[UCL]: Selectional restriction in Mlabri ingestion verbs

Speaker: Aljoša Milenković (Harvard) and Kevin Ryan (Harvard)
Title: Stress-weight and stress-tone interaction in South Slavic folk verse
Time: Monday, September 15th, 5pm - 6:30pm
Location: 32-D831

Abstract: We analyze the distribution of stress, tone and syllable weight in the epic decasyllable, a traditional verse form of South Slavic folk poetry. The meter imposes strict distributional constraints on tone and weight in stressed syllables. The stress-modulated nature of tone and weight mapping places the epic decasyllable in the category of hybrid meters of the interactive type (in the sense of Ryan 2017). The contribution is twofold: we demonstrate that hybrid meters can display multiple interactive mappings in parallel (in this case, stress-weight and stress-tone) and offer evidence for stress-tone interaction, previously undocumented in hybrid meters. This is also, to our knowledge, the first demonstration of tone-sensitivity in an Indo-European meter. We provide a MaxEnt model for the epic decasyllable. Building on recent work in MaxEnt metrics (Hayes & Schuh 2019; Henriksson 2022), we discuss different methods for incorporating ordinary language baselines in MaxEnt analyses of metered verse.

Speaker: Eric Baković (UC San Diego)
Title: A computational typology of vowel harmony patterns
Time: , 3:30pm - 5pm
Location: 32-141

Abstract: We propose a typology of vowel harmony patterns using Boolean Monadic Recursive Schemes (BMRS; Bhaskar et al. 2020, Chandlee & Jardine 2021). BMRS describe input-output maps as logical transductions using conditional IF…THEN…ELSE… syntax, and unite computational universals from formal language theory with substantive universals from phonological analysis. We identify how different computational structures (extension) and basic predicate types (intension) systematically derive distinct harmony patterns. Our typology emerges from varying the presence, type, and scope of redundancy conditions within BMRS harmony functions. When potential targets fail to receive or propagate harmonic features, redundancy conditions override harmony in predictable ways by blocking feature reception, halting feature propagation, or preventing both reception and propagation, depending on their type and scope within the function. This proposal builds upon recent work relating phonological generalizations to specific BMRS structures, offering a unified computational account of typological variation in vowel harmony without the representational assumptions of traditional approaches.

Course announcements in this post:

• Topics in Semantics (24.979)
• Topics in Phonology (24.964)

24.979: Topics in Semantics

• Instructor: Amir Anvari & Martin Hackl & Viola Schmit
• Time: Thursday, 2pm-5pm
• Room: 32-D461

The very basic observation that expressions can refer to semantic objects raises two general questions. The first concerns the mapping, i.e., how a particular expression with a given conventional meaning actually gets to pick a particular semantic object (an entity, a function, etc.)? This issue is raised, for example, by pronouns, but also by expressions without overt pro-forms whose values seem to be (partially) dependent on the context. The second question concerns the status of the semantic objects that are being picked out themselves. We often take them to be given (as in standard model theoretic semantics), setting aside e.g. issues concerning how the context might contribute to individuating these objects. The seminar will consist of various explorations and case studies, which we hope will bring these foundational issues into sharper focus. Issues we hope to discuss include:  What are grammatically relevant notions of identity or similarity and distinctness? When do we take two expressions to refer to the same semantic object? When can we use expressions that intuitively relate to identity, like pronouns or operators like same? When do we take two expressions to refer to different objects? And when are those grammatical devices licensed that intuitively seem to relate to distinctness, like different, plurals or numerals?

• Instructor: Adam Albright
• Time: Wednesday, 10am-1pm
• Room: 32-D461

Speaker: Yiannis Katochoritis (MIT)
Title: Long-distance agreement on Procrustes’ bed: a revival of spec-head
Time: , 12:30pm - 2pm
Location: 32-D461

Abstract: Long-distance agreement (LDA) refers to the phenomenon where a matrix verb may (optionally) agree in φ-features with an embedded argument of its complement clause, otherwise showing default agreement with the entire complement clause. Spanning a clausal boundary and appearing to involve no displacement of the embedded goal, LDA has been one of the core empirical arguments in favor of the operation Agree, which dissociates agreement from movement.

This talk has a double aim: first, to present a previosuly unnoticed observation; that is, (almost) all geographically diverse LDA languages share fundamental typological properties: they are verb-final, with a (full or split) ergative alignment, and with default SOV order; moreover, LDA-transparent embedded clauses are always either restructuring VPs or nominalized/participial TP/CPs; and once a potential LDA goal somehow overtly fronts to the left of the matrix verb, LDA (usually) becomes obligatory.

Second, I attempt to account for this empirical picture through little bits of Kayne’s (1994) Antisymmetry, Bobaljik’s (2008) m-case hierarchy, and the magic power of pied-piping. Specifically, I will suggest a unified analysis of LDA as local feature-checking in a (covert) spec-head configuration, which may allow for iterative feature-copying under recursive merge. Depending on two parameters, the specifier of the agreeing verbal head is filled either directly by the moving DP goal or by the complement clause that pied-pipes the goal at its edge. Hence, LDA is derived via Merge, not Agree. This will eventually yield a typology of (at least) four types of LDA systems, among which I will (struggle to) intergrate the outlier named Algonquian.

The analysis hinges on two crucial premises: one, LDA often features movement properties/constraints; two, the optionality of LDA is only apparent, as it obligatorily interacts with scope and/or information-structure, reducing to the reflex of independently motivated syntactic processes. Thus, I will distinguish between two types of agreement: narrow-syntactic agreement feeds LF or further syntactic operations, and involves downwards valuation by some interpretable feature, (plausibly) under spec-head. This contrasts with uninterpretable agreement, which is pushed post-syntactically, at early PF, and may operate from a distance, via upwards valuation (though still c-command-based).

Speaker: Yiannis Katochoritis (MIT) and Magdalena Lohninger (Paris Lodron Universität Salzburg)
Title: Austronesian pivots: topics or subjects? A Malayo-Polynesian typology
Time: Thursday, September 4, 12:30pm - 2pm
Location: 32-D461

Abstract: A long standing debate on the Austronesian voice system is whether pivots–the pragmatically and syntactically most salient clausal argument- are V2-like topics or derived subjects. By investigating micro-variation across the Malayo-Polynesian branch (Tagalog, Balinese, Malagasy, Acehnese, Standard Indonesian), this talk proposes that the syntactic status of pivots varies across languages, though in a unidirectional manner, ranging from pure topicalization to argument-structure alternation. We focus on the A/A’ status of pivots (weak crossover and reconstruction), their subjecthood properties (obligatory control and their impact on licensing of non-pivot agents) and info-structural effects (formal and pragmatic discourse restrictions). Crucially, this variation is not arbitrary but aligns pivots on a continuous cline from topichood to subjecthood, with intermediate stages in-between.

We suggest that this cline results from an ongoing process of topic-to-subject grammaticalization, instantiated as topic-to-subject lowering. We examine four stages of this reanalysis, showing that they stand in an implicational hierarchy, each step incrementally feeding into the next. In stage 1 (Tagalog), the pivot is an obligatory topic, driven by information-structure, but already exhibits some A-movement properties due to the status of Malayo-Polynesian languages as discourse-configurational systems. In stage 2 (Balinese), the pivot acquires a surface-subject status, and its (high) landing site is re-interpreted as a case-assigning position; in Object/Theme Voice, this triggers the voice domain to become a non-canonical passive, which enforces the in-situ (non-demoted) agent to be licensed via strict verbal adjacency. In stage 3 (Malagasy), the voice domain loses its ability to case-license its internal argument with accusative. In the last stage (Acehnese, Standard Indonesian), the topic flavor and discourse restrictions on pivot promotion disappear.

Finally, we address the status of Malayo-Polynesian pivots with respect to the two other typical subject properties, namely, imperatives/hortatives and anaphor binding. Though these diagnostics are sensitive to independent confounds and seem to only target the agent, we show that, once these caveats are dispensed with, they align with obligatory control in targeting the surface subject, whether the agent or not. Relevant case studies on the matter will be the Tagalog non-volitive construction and the several passive-like constructions of Indonesian-type languages.

Whamit! will be on semi-hiatus over the summer, as usual. We will continue to publish breaking MIT Linguistics news as it happens. Weekly posts will resume in the Fall.

Thanks to our editors, contributors, and of course all our readers! See you all in the Fall!