Research

Syntax-semantics interface

I investigate what looks like a proleptic construction in Modern Greek, where an attitude verb takes an accusative object, followed by a complement clause. The latter has an obligatorily co-referring with the accusative object pronoun. This ressembles prolepsis in many ways, except that the accusative object can be read opaquely. The puzzle is that the accusative object, which syntactically is not part of the embedded clause, can nonetheless be read in the scope of the attitude verb. I outline two different proposals, one solving the puzzle with the machinery of semantic lowering, and one arguing for a covert CAUSE verb in between the attitude verb and the complement clause. I show that the accusative object is the subject of the covert causative, and that under this analysis the semantic interpretations are completely expected given syntactic scope.

Tense semantics & Fieldwork

My M.Sc. thesis, Embedded Tense: Insights from Modern Greek, was co-supervised by Philippe Schlenker and Amir Anvari. It was conducted at the Institut Jean Nicod (IJN) laboratory of the Centre National de Recherche Scientifique (CNRS) and funded by the ERC grant Sources of Meaning (PI: Philippe Shclenker). You can find the abstract below:

Abstract

Embedded tense in Modern Greek (MG) displays an unexpected ‘optionality’: both present and past tenses can be used under a past tense attitude verb to convey a simultaneous reading. We claim that MG has a mixed tense system, being able to delete the embedded past like English and shift the embedded present like Russian. Are these two the only routes to the simultaneous reading in MG? We claim that sometimes there is a third one, namely interpreting the embedded past with respect to the time of the utterance. Based on a cross-linguistic investigation of the availability of simultaneous readings in languages without a deletion rule, we provide evidence that there is variation, both across languages and across speakers. We provide an analysis using a pragmatic Prefer De Se principle and a syntactic Prefer Local Binding parameter. The first states that de se readings are preferred whenever possible, be they obtained via de se Logical Forms or via de re ones with temporal descriptions that happen to be de se. Prefer Local Binding expresses a preference for locally bound temporal variables, therefore giving rise to a back-shifted reading of past-under-past in the absence of a deletion rule. Based on data from ellipsis, we argue that in MG this parameter is inactive, and thus MG has a third route to the simultaneous reading. Finally, we introduce the ‘then’-present puzzle, namely the observation that ‘then’ is incompatible with the shifted present. We extend Ogihara & Sharvit’s (2012) and Vostrikova’s (2018) paradigm for Hebrew and Russian, arguing that the puzzle holds not only for present-under-past but also for present-under-future environments cross-linguistically, both under attitude verbs and in relative clauses. Furthermore, we provide novel data, and conclude that the puzzle also holds in a mixed tense language like MG. Finally, we show that ‘then’ is compatible with the present in other environments and we argue against competition-based accounts, leaving the puzzle open.


In Spring 2023, I started doing fieldwork on Colloquial Jakartan Indonesian, where I investigate the use of `want' mau as a future marker. I argue that mau is the dispositional `will', and I identify a puzzle with negation. Cross-linguistically future morphemes are often diachronically derived from desire verbs, a change which is synchronically attested in colloquial Indonesian.

Anaphora & Co-speech gestures

As part of the Meaning & Modality Lab at Harvard, Kathryn Davidson and I experimentally investigate whether co-speech gestures can license complement anaphora (e.g. Few students came to class. They stayed home) with upward monotone quantifiers (e.g. most, some). This project is motivated by the behaviour of non-default loci in ASL, which license complement anaphora with upward monotone quantifiers (Schlenker 2012). We find that iconic co-speech gestures can have the same effect, suggesting that the ASL facts are modality-independent and that there is a common cognitive mechanism interpreting iconicity both in loci and in gesture. We suggest that iconic co-speech gestures trigger iconic inferences of existence, just like iconic loci in sign language (Kuhn 2020).

Effects of iconicity and monotonicity on licensing complement anaphora (with Kathryn Davidson) [slides]

Definiteness in clausal subjects

Katya Morgunova and I show that, at least for some speakers, the D-layer (definite article to) in clausal subjects in Greek is optional, and can only be preferred/dis-preferred in certain environments. We argue that its distribution is governed by semantic and pragmatic considerations rather than syntactic ones, contrary to previous claims in the literature. We propose that the D-layer in clausal subjects introduces a presupposition that the proposition it modifies is consistent with the beliefs of the speaker.

Exclusivity of disjunction 

I participated as a research assistant at the Logic across languages: expressing and interpreting connectives cross-linguistically (CrossConn) project of Andreea Cristina Nicolae at the Leibniz-Centre of General Linguistics (ZAS). The aim of the project is to compare natural language connectives cross-linguistically with boolean/logical ones. We designed and run an experiment to test the extent to which various disjunctive particles give rise to exclusive interpretations. The experiment was translated and ran in English, French, Modern Greek, Russian, Romanian and Turkish.