Sentence Processing

Point of Contact: Nino Grillo

Collaborators: Andrea Santi, Giusy Turco, Shayne Sloggett, Leah Roberts, Miriam Aguilar

Description: Much work has demonstrated a consistent relationship between the relative rate of production/comprehension processes, and the linguistic complexity of the material involved: longer production latencies and comprehension delays are generally associated with more complex linguistic structures. Our work poses a challenge for this general trend, demonstrating that syntactically nested "complex" structures may actually be produced at a relatively fast rate (Grillo et al. 2018). Moreover, it seems that these faster productions may facilitate comprehension of these structures (Grillo et al. 2019). In this project, we're exploring the relationship between the relative rate at which information is presented and its perceived complexity with the hypothesis that faster rates may serve as a cue to more effectively demarcate syntactic boundaries.

Point of Contact: Nino Grillo

Collaborators: Andrea Santi, Fani Karageorgou, Shayne Sloggett

Description: In the last two decades, cue-based retrieval has arisen as the dominant model of working memory in sentence comprehension (McElree et al. 2003, Lewis & Vasishth 2005; i.m.a.). Much of this work has relied on one highly influential study which suggests that comprehenders encounter difficulty when multiple nouns in memory are good semantic fits for a given verb (Van Dyke & McElree 2006). However, this finding has proven difficult to replicate, with many studies failing to find any such penalty in realtime measures of processing difficulty. In this project, we propose an alternative explanation of Van Dyke and McElree's findings in terms of semantic similarity and surprisal. Specifically, we suggest that their results reflect facilitated integration when a noun and its verb are semantically related or otherwise expected to co-occur (Santi et al. 2020). Crucially, this work doesn't repudiate a role for cue-based memory access, but suggests that it may be more tightly constrained by syntactic factors than previously believed.

Point of Contact: Nino Grillo

Description: Early work in sentence comprehension suggested a strong role for syntactic processing heuristics like Late Closure (Frazier & Rayner 1982): construe incoming information as part of the current phrase rather than positing a new one. However, cross-linguistic investigations quickly called this proposal into question, with some languages (e.g. Dutch, Spanish, Japanese, Korean) seeming to show an early closure bias. This line of work seeks to question this dichotomy by suggesting that so-called "early-closure" languages actually employ alternative syntactic structures that are only surface-similar to English-style attachment ambiguities (Grillo 2012; Grillo & Costa 2014; Grillo et al. 2015; Pozniak et al. 2019). This work highlights the importance of careful scrutiny of the underlying syntax when making psycholinguistic arguments.

Point of Contact: Nino Grillo

Collaborators: Andrea Santi, Artemis Alexiadou, Caterina Paolazzi

Description: Passive sentences are considered more difficult to comprehend than active sentences. Previous online-only studies cast doubt on this generalization. This project compares online and offline processing of passivization and manipulates verb type: state vs. event. Stative passives are temporarily ambiguous between adjectival and verbal interpretations, while eventive passives are always verbal. We show that passives are consistently read faster than actives. This contradicts the claim that passives are generally difficult to parse/interpret. However, this facilitation is compatible with broader expectation/surprisal theories. In contrast, when comprehension questions target the thematic interpretation of the sentence, we find that passives are more errorful, irrespective of verb type. In addition, measures of verbal working memory appear to be correlated with comprehension question performance, but not online comprehension difficulty. This suggests that the initial interpretation of a passive sentence may not itself be difficult, but maintaining or re-accessing that interpretation in memory may be costly (Paolazzi, Grillo, Alexiadou and Santi 2019; Grillo, Alexiadou, Gehrke, Hirsch, Paolazzi and Santi 2018; Paolazzi, Grillo and Santi 2021).

Contact Point: Shayne Sloggett

Collaborators: Amanda Rysling, Nicholas Van Handel

Description: Many prominent models in modern psycholinguistics ascribe a strong role to so-called "good enough" comprehension; the idea that speakers may not always fully interpret the utterances they encounter. For example, Swets et al. (2008) suggest that fully ambiguous sentences are read more quickly than their unambiguous counterparts only when resolution of the ambiguity is not relevant to the current task demands. In contrast, our work demonstrates that there is a task-general ambiguity advantage, such that ambiguous sentences are processed more easily in all task contexts (Sloggett et al. 2020). This suggests that race models of ambiguity resolution may provide a better explanation than strategic underspecification.

Point of Contact: Shayne Sloggett

Collaborators: Amanda Rysling, Adrian Staub

Description: That there is a "focus advantage" is now well established in the psycholinguistics literature. Comprehenders remember focused constituents better, more frequently notice changes in focused positions, and are better at detecting errors that have been focused. However, the source of this advantage remains contentious: is it the product of additional processing of focused material, or increased attention? In this work, we suggest that it may be both. Following Cutler and Fodor (1979), we show that comprehenders attempt to predict the location of focused information (realised as reduced skipping for focused words). However, we also find that comprehenders spend longer reading focused material, suggesting that this increased attention leads to additional processing, rather than facilitation (Sloggett, et al. 2020). Future work will need to consider alternative methods of realising linguistic focus, as well as the role of parafoveal preview.

Point of Contact: Shayne Sloggett

Description: There is an ongoing debate over when, and how grammatical principles are applied in resolving anaphoric reference. One view holds that speakers deploy grammatical information in parallel with domain-general information about potential referents (Jäger et al. 2020, i.a.), while another maintains that grammatical constraints have a privileged status in antecedent re-access (Dillon et al. 2013; Parker & Phillips 2017, i.a.). This project is interested in the interplay of domain-general pressures and sophisticated grammatical knowledge. Specifically, it is suggested that reflexive pronouns may entertain putatively ungrammatical antecedents when those referents constitute the current perspective centre of the utterance. However, they only do so when all available grammatical referents represent an exceptionally poor feature match for the reflexive. If true, this would mean that native English speakers behave like speakers of other languages (e.g. Icelandic, Mandarin, Japanese) when their native grammars cannot provide reasonable interpretations (Sloggett 2017).