Kelsey Sasaki

Discourse Processing

Long-distance pronominal anaphora
In discourse, one or more clauses may separate an anaphor and its antecedent. Very little is known about how these long-distance pronominal anaphora are processed. Previous findings on pronominal anaphora resolution between adjacent clauses support the broad hypothesis that any factor that modulates antecedent prominence can modulate long-distance pronoun resolution, such that greater antecedent prominence leads to lower processing costs. Meanwhile, some formal discourse theories—notably, Segmented Discourse Representation Theory (SDRT; Asher & Lascarides 2003)—posit that pronoun resolution is determined by what they call the Right Frontier Constraint (RFC; Polanyi 1988), which I argue is an essentially grammatical constraint on discourse structure-building. I further argue that SDRT does not straightforwardly predict that any factor besides the RFC should modulate intersentential pronoun resolution. Through a combination of stops-making-sense task and Maze task studies, I have found evidence suggesting that, as SDRT predicts, RFC-violating pronominal anaphora may be harder to process than their RFC-observing counterparts. However, contra SDRT’s predictions, I also found that prominence-modulating factors—here, lexical semantics and animacy—may modulate that effect. However, it remains to be seen whether prominence-modulating factors can fully account for long-distance pronominal anaphora processing; my ongoing work—in collaboration with Pranav Anand and Amanda Rysling—is testing for effects of more such factors.

Sasaki, K., Anand, P., & Rysling, A. 2024. Animacy and long-distance pronominal anaphora in discourse: Evidence from the Maze. Poster at Architectures and Mechanisms for Language Processing 30. Edinburgh, UK, September 6. [PDF]

Sasaki, K., Anand, P., & Rysling, A. 2023. Animacy and event structure modulate long-distance pronominal anaphora in discourse. Poster at the 36th Annual Conference on Human Sentence Processing. Pittsburgh, USA, March 10. [PDF]

Clause-internal coherence
Most work on coherence relations focuses on cross-sentential contexts, or at least cross-clausal ones (Hoek et al. 2021). However, it has also been proposed that CIC is also possible---that is, a discourse interpretation may arise from a single clause (Hobbs 2010). For instance, a causal coherence inference may be drawn from the monoclausal, A drenched child got hit by a big water balloon---namely, a child was drenched because she got hit by a big water balloon. In joint work with Daniel Altshuler and Hannah Rohde on English, we have found offline evidence suggesting that both adjectives and nouns can give rise to CIC inferences, which in turn suggests that proposition-like content can be extracted from adjectival and nominal elements. Our ongoing work is exploring the implications of this for formal semantics and pragmatics; in particular, we are developing an account that adopts Schwarzschild’s (2022) Pure Event Semantics, in which adjectives and nouns are treated as one-place predicates of eventualities. We are also testing whether the online processing of CIC parallels that of cross-clausal coherence.

Sasaki, K., Rohde, H., & Altshuler, D. 2024. Clause-internal causal inferences: Evidence from nouns. Poster at Sinn und Bedeutung 29. Noto, Italy, September 17. [PDF]

Sasaki, K. & Altshuler, D. 2023. Clause-internal coherence: A look at deverbal adjectives. In Proceedings of Sinn und Bedeutung (Vol. 27). [PDF]

Sasaki, K. & Altshuler, D. 2022. Clause-internal coherence, deverbal adjectives, and presupposition resolution. In Proceedings of 23rd Amsterdam Colloquium, 437-443. [PDF]

Exclamatives and exclamative-embedding predicates

In joint work with Tom Roberts, we are investigating the semantics of exclamatives using corpus studies and acceptability judgment studies on English. There is no consensus on what type of semantic object exclamatives denote. In many languages, their form resembles that of interrogatives, which has motivated analyses of exclamatives as special questions (Grimshaw 1979). However, the interpretive function of exclamatives is to express surprise, often at the degree to which a certain property holds, which has motivated analyses of exclamatives as predicates of degrees (Rett 2008) and as facts (Ginzburg & Sag 2000). We focus on embedded wh-exclamatives, as in Annie knows what a cute hippo Moo Deng is, in order to separate the semantics of exclamatives from their conventional discourse effects (following Karttunen 1977 on interrogatives). We provide corpus and experimental evidence for a novel empirical generalization: a predicate V can take wh-exclamative complements iff V is wh-responsive, and vice versa. (A predicate is wh-responsive iff it can take declarative complements and wh-interrogative complements.) We argue that this points to an analysis of exclamatives as questions that presuppose a particular answer (Zanuttini & Portner 2003), and supports the view that responsive predicates s-select for questions (Uegaki 2016). In our ongoing work, we are testing whether the proposed empirical generalization holds cross-linguistically and seeking ways to account for the unembeddability of other kinds of English exclamative (e.g., Wow, is she a cute hippo!).

Roberts, T. & Sasaki, K. 2024. Oh, what predicates embed exclamatives! Talk at MECORE Closing Workshop. Konstanz, Germany, June 19. [Email for slides]

Fieldwork

Relative clause processing in Santiago Laxopa Zapotec
Santiago Laxopa Zapotec (SLZ; Oto-Manguean), an indigenous language of Oaxaca, Mexico. As a graduate student at UC Santa Cruz, I was involved in a range of collaborative projects between UCSC linguists and members of the Zapotec community. Collectively known as the Zapotec Language Project, these activities ranged from the field psycholinguistics project described here, to contributing to an online Zapotec dictionary, to organizing free public language-learning experiences as part of the Nido de Lenguas initiative.

The field psycholinguistics project is a collaboration with native speaker Fe Silva-Robles, Steven Foley, Jed Pizarro-Guevara, Maziar Toosarvandani, and Matt Wagers. Much previous work on relative clause parsing has shown a subject relative clause (SRC) advantage: SRCs like the fox that is biting the dog are easier to parse than object relative clauses (ORCs) like the fox that the dog is biting. It is an open question whether the SRC advantage is a universal or language-specific parsing preference. SLZ provides a strong test for the SRC advantage because it has strict VSO ordering and lacks case and verbal agreement—its RCs are ambiguous unless they contain a resumptive pronoun. Further, there is evidence that the SRC advantage in English, for instance, is modulated by animacy, in this case an animate/inanimate distinction (Traxler et al. 2002). SLZ distinguishes four grammatical animacy categories—inanimate, animal, human, and elder (human)—which map onto three notional animacy categories—inanimate, animal, and human. Our investigation explores the roles of grammatical and notional animacy in RC parsing and tests for the SRC advantage in SLZ, via a series of eye-tracking and picture-matching studies.

Sasaki, K., Foley, S., Pizarro-Guevara, J.S., Silva-Robles, F., Toosarvandani, M., & Wagers, M. Manuscript. Evidence for a universal parsing principle in Santiago Laxopa Zapotec. [PDF]

Predicate-initiality in Hawai'i Creole
Hawai'i Creole is one of my family languages (though not one I speak natively)—my curiosity about the language, as well as its stigmatization, was one of the key factors that led me to linguistics as an undergraduate. I went on to conduct fieldwork focusing on a predicate initial construction in the language, which has default SVO word order. Not just any predicate can participate in this construction: for instance, one can say, Ono, dat! ('That's delicious!'), but not *Running, her! (Intended: 'She's running (hard)!'). This construction also seems to be expressive—I hypothesize that it is the exclamative construction (or at least one of the exclamative constructions) in Hawai'i Creole. I hope to return to this work before too long; if you have any insights, please get in touch!