peter grishin

I'm a postdoctoral associate at the Department of Linguistics and Philosophy at MIT (the linguistics part); I got a PhD in linguistics at MIT in 2023 and a BA in linguistics at the University of Cambridge in 2018.  Currently, I'm thinking about the theory and typology of agreement systems, issues in clause size and cross-clausal phenomena, the Passamaquoddy modal system, the semantics and pragmatics of future talk, and the interaction between evidentiality and the dynamics of assertion. I also do fieldwork, and am currently working on Passamaquoddy (Eastern Algonquian, Maine), looking at various morphosyntactic and semantic topics. I have also worked on ellipsis and case phenomena.

My dissertation examines issues in the morphosyntax of CP in Passamaquoddy, looking at how the presence/absence of CP contributes to the syntactic properties of different clause types, as well as the unusual phi agreement behavior C displays.

I was born and raised in Dallas, Texas. Outside of linguistics, I enjoy cooking, eating, board games, tabletop RPGs, social deception games, and petting cats. In a past life I was an active amateur violinist.

my cv (updated 1/2024)                                                                              email me at


clause size and cross-clausal phenomena

I am interested in issues of clause size and restructuring in clausal embedding as well as coordination. I investigated these issues in Passamaquoddy in my dissertation, looking at the distribution of three distinct clause types/inflectional paradigms—the independent, the conjunct, and the subordinative—examining what kinds of syntactic processes are available within and across these clause boundaries. The independent prototypically appears in matrix clause declaratives and polar questions, as well as under the verbs itom ‘say’ and litahasu ‘think’; the conjunct appears in various kinds of "more finite" embedded clauses as well as in most (but not all) clauses containing Ā dependencies; and the subordinative appears in the complement of verbs that crosslinguistically tend to select for smaller, "less finite" complements, as well as in asymmetric coordinations. I propose that independent and conjunct clauses are both CP-sized, with independent clauses being larger than conjunct clauses, and I argue that subordinative clauses are bare TPs, without a CP layer. 

Aspects of this work have appeared in several venues. There is a particular kind of long distance agreement construction involving subordinative complements to the verbs 'pawatomon ‘want’ and 'kisehtun ‘make’ which I argue involves genuine raising to object, potentially covertly (with consequences for the right theoretical account of covert movement), and shows a strikingly restricted locality profile compared to other cases of long distance agreement in Passamaquoddy (with consequences for the right analysis of the inverse construction in Algonquian). I propose that the properties of this construction, especially in comparison to other kinds of embedding, naturally arise from subordinative clauses being structurally reduced. 


I have been working on a number of typological and theoretical issues in agreement, in large part seeking to understand the unusual behavior of the peripheral suffix in Algonquian, an agreement suffix in C which shows third person preference and displays interesting variation in its locality properties across the family. I am working to understand how these agreement patterns fit within a larger theory and typology of agreement. My work on agreement touches on theories of phi features, how to derive seemingly nonlocal cases of phi agreement, what the morphology does when it needs to spell out multiply-valued heads, and how to derive person portmanteaux.

The peripheral suffix agrees omnivorously with third persons across the family, and I argue that the only feasible analysis of this fact is to accept that there is a distinct representation of third person that can be targeted in the narrow syntax. A paper making this argument and exploring the ramifications of this conclusion for theories of phi features has appeared in Glossa, and a slightly expanded version of that paper appears as Chapter 9 of my PhD dissertation.

In some Algonquian languages, like Passamaquoddy, C prefers agreeing with the lowest clausemate third person argument after A movement. This is a striking case of lowest preference that seems to fly in the face of standard locality principles governing agreement. I argue that various existing analyses of lowest preference fail to capture all of the properties of the Passamaquoddy pattern, and instead propose that Passamaquoddy C agrees with all accessible third persons, only exponing the last agreement relationship it participates in—drawing a parallel with instances of multiple case assignment where only the last case assigned to a nominal is spelled out. I argue that this proposal makes a number of interesting and correct typological predictions. A version of this paper appears as Chapter 10 of my PhD dissertation.

With Amy Rose Deal (Berkeley), we are exploring more broadly how the morphology can deal with instances of multiple valuation, both in the domain of agreement as well as case. We want to understand what variation exists, how best to analyze it theoretically, and what consequences this has for the theory of the syntax-morphology interface.

With Will Oxford (MIT/Manitoba), we are examining another agreement slot in the Algonquian verb (the central suffix in the conjunct mode), in particular seeking to understand variation in the behavior of person portmanteaux and the consequences for the analyses of person portmanteaux theoretically. Across the family, we find variation in whether person portmanteaux can be blocked in the presence of morphemes deriving from the Proto-Algonquian negative/irrealis markers *-w and *-(h)k, as well as variation in when person portmanteaux can co-occur with Fissioned-off plural markers. We believe that fully capturing the attested variation across the family (as well as variation within the same language) requires at least three possible paths to portmanteau formation: spelling out a multiply-valued head, local contextual allomorphy, and non-local contextual allomorphy. The empirical survey for this project has been presented at the 54th Algonquian Conference, and we presented the theoretical aspect of this work at WSCLA 26.


I have looked at the differences between the English future operators will and be going to, and their entailment patterns when embedded under present and past tense. Strikingly, though present tense will p and is going to p both seem to entail p, in the past their entailment properties diverge: would p (under the future-in-the-past reading) still entails p, but was going to p does not (an observation going back at least to Binnick 1971). I argue that these entailment patterns shouldn't be analyzed as a compositional semantic interaction between future operators and tense, but rather as arising from the pragmatics and epistemology of future talk and thought—the ways in which we think and talk about the future. I argue that will and be going to have rather boring denotations, ones in which will is veridical and be going to is nonveridical, and that their truth conditions collapse when their temporal perspective is the utterance time, due to certain properties about how we reason and talk about the (actual) future. A key observation driving this perspective is that, when you get present tense but past perspective (e.g. in the historical present), will and be going to again diverge in veridicality, just like their past-marked counterparts.


I've been thinking about the semantics and pragmatics of a certain class of evidential markers that can be denied with belief claims, but not with knowledge claims or plain assertions: EVID p, but I believe ¬p, but not #EVID p, but I know ¬p or #EVID p, but ¬p. I call this property partial deniability, and these operators partially deniable evidentials. Azerbaijani imiş (composed of the copula i- and the perfect -mIş), English apparently (for certain speakers), and Dutch schijnen (Koring 2013) seem to have this property. Partial deniability is a novel typological option that hasn't been explored in the literature, which generally assumes that evidentials are either uniformly deniable or uniformly non-deniable.

I am currently exploring the hypothesis that we can derive the properties of (certain kinds of) partially deniable evidentials by analyzing them as assertions (in that they propose to add their prejacent to the common ground) that can be made without full belief in the prejacent. This point of view has interesting and striking ramifications for phenomena like Moore's paradox and the way we think about norms of assertion.


There are various paths to unpronounced material—standardly, in the spirit of Hankamer and Sag (1976), we distinguish between (null) proforms/deep anaphora and ellipsis/surface anaphora. However, there is another logical possibility: null anaphors with the semantics of bound pronouns (Chao 1987, Schwarz 2000). 

I argue that this possibility is instantiated in an understudied kind of clausal ellipsis I call scrapping, where the clausal complement of a clause-embedding verb goes missing in certain syntactic contexts, like degree phrases, relative clauses, and temporal adjuncts. I show that scrapping exhibits behaviors that are difficult to account for under standard theories of ellipsis licensing (e.g. Merchant 2001, Aelbrecht 2010), such as global structural requirements like the gap needing to be contained within its antecedent and the gap needing to be c-commanded by its antecedent at LF, as well as behaviors that distinguish it from a null proform/deep anaphor, like the obligatory movement of an operator out of the gap. I argue that these properties tell us that the gap hosts a clausal anaphor that must be bound by a c-commanding clausal antecedent.

I'm currently extending this work to the case of conditional scrapping, which is where the complement of a clause-embedding verb goes missing in the antecedent of a conditional: you can sit here if you want Δ. I argue that conditional scrapping gives us further insight into how scrapping is licensed and the syntactic category of the scrap/null clausal anaphor. Additionally, conditional scrapping has striking consequences for the syntactic analysis of conditional antecedents: it shows us that we must allow structures where conditional antecedents are first-merged as restrictors to modals, even in the case of biscuit (speech act) conditionals.


There are various environments where DPs might plausibly receive case multiple times—however, most languages only allow case to be exponed once on a single DP. What do languages do when a DP receives case from multiple sources, and what does this tell us about the (morpho)syntax of case?

With my coauthor Tamisha L. Tan (Nanyang Technological University), I have thought about this issue in the domain of free relatives. Free relatives in many languages are well-known to impose a matching effect on the wh item: the case and category requirements imposed on the wh item by the embedded and matrix clause must be identical in order for a free relative to be grammatical. However, there are languages that allow mismatches in certain contexts—in cases of mismatch, how do languages pick which case to realize? 

We argue that we can understand the standard matching effect as a reflex of a DP getting case multiple times, resulting in that DP hosting multiple conflicting feature values which are unable to be spelled out at once (a similar idea is pursued in Coon and Keine 2019 in the domain of PCC and various other hierarchy effects). We propose that languages can allow mismatches in free relatives if they have certain properties that allow them to "get out of" having conflicting feature values in a given syntactic configuration.

previous work

For my BA thesis, I examined clausal complementation and hyperraising in Zulu. I argued for two main points: i) clauses embedded under the complementizer ukuthi host a null expletive in their left periphery; and ii) raising to a preverbal position is triggered by a composite [φ] and [Top] probe. I showed that these two components correctly predict certain agreement facts in the domain of hyperraising and clausal extraposition.

view from outside the Waponahki Museum & Resource Center, Sipayik (Pleasant Point), ME