Minding the gap?
Mechanisms underlying resumption in English

Glossa, 2019 [Link]

Resumptive pronouns in English are argued to be ungrammatical. If they are ungrammatical, why do they sound so good? The classic explanation is that they facilitate processing of difficult filler-gap dependencies. However, there's also good reason to believe that filler-gap dependency processing is sensitive to grammatical constraints, i.e., comprehenders are good at avoiding ungrammatical filler-gap dependencies (Phillips 2006). How can both of these be true? I argue that memory constraints may result in the comprehender forgetting that they are seeking a resolution to a dependency, and resumptive pronouns help stitch together a coherent interpretation of the sentence in this circumstance. More broadly, this implies that processing filler-gap dependencies involves fine-grained predictive mechanisms of syntactic/semantic structure.

Linguistic representations and memory architectures:
The devil is in the details

Brain and Behavioral Sciences, 2016 [Link]
With Shota Momma & Colin Phillips

Theory construction in psycholinguistics is hard. We have to stay within the confines of the limitations of the memory system, and we also have to be responsible to the structural properties of language. It also has to be powerful enough to make predictions about how sentence processing unfolds that are clear and testable. In this commentary on Christian and Chater's paper, we argue that trying to account for grammatical phenomena in sentence comprehension and production with the foundational principle of lossy, quick encoding is unlikely to be the best way to start.

Locality and word order in active dependency formation in Bangla

Frontiers in Psychology, 2016 [Link]
With Mashrur Imtiaz, Shirsho Dasgupta, Sikder M. Murshed, Mina Dan, & Colin Phillips

It's a well-established fact in psycholinguistics that comprehenders eagerly seek to resolve filler-gap dependencies. But what drives this preference? Do we search for the first possible syntactic position available, or do we opt for the one that's closest in the structure? By looking at languages with flexible word order, we can isolate these factors from one another. In this study, we argue that speakers of Bengali – a language in which embedded clauses may precede or follow the verb – seek to resolve filler-gap dependencies with the first position in the sentence. This is true whether the first position is an embedded clause or not. However, we demonstrate that this is not the whole story, and that there still is a role for distance as measured by embedding depth in determining Bengali speakers’ processing strategies.

Computerized analysis of speech and language to identify psycholinguistic correlates of frontotemporal lobar degenaration

Cognitive and Behavioral Neurology, 2010 [Link]
With Serguei Pakhomov, Glenn E. Smith, Yara Feliciano, Neill Graff-Radford, Richard Caselli, & David S. Knopman

This is the first paper out of RAship with Serguei Pakhomov working on quantifying changes in language production in patients with cognitive degeneration. Alzheimer's Disease and frontotemporal lobar degeneration affect language production, but quantifying this is challenging. However, if it's possible to automatically detect these changes, then this raises the possibility of using language to diagnose dementias. Here, we quantify a number of psycholinguistic features of frontotemporal lobar degeneration (FTLD) patients describing the ‘Cookie Theft’ picture, and highlight which variables might be associated with which variants of FTLD.

The real-time status of semantic exceptions to the adjunct island constraint

Proceedings of the 54th Chicago Linguistic Society, 2018 [Draft]
With Annika Kohrt & Trey Sorensen

Extraction from adjunct clauses is ungrammatical, except when it isn't. The syntactic/semantic criteria that determines whether an adjunct is an island appear to be complex. We argue that comprehenders use a ‘bottom-up’ approach, rather than the typical ‘top-down’ predictive mechanism associated with filler-gap dependencies.

Embedding, covert movement, and intervention in Kathmandu Newari

Proceedings of the 54th Chicago Linguistic Society, 2018 [Link]
With Borui Zhang

This paper is a progress report on Borui Zhang's research in Newari complementation and question formation, which resulted from my Field Methods course at University of Minnesota, in which she applied several tests for covert movement and intervention for wh-operators in different embedding contexts. This was the beginning of my interest in the languages of Nepal, as well!

Plural shifted indexicals are plural: Evidence from Amharic

Proceedings of the 44th Annual Conference on African Linguistics 2015 [Draft]
With Chris LaTerza, Morgan Rood, Ruth Kramer, and Jen Johnson

In some languages, indexical expressions (‘I’, 'you’, ‘now’) “shift” in embedded clauses that describe the content of someone else's speech or thoughts. There are several theories of these phenomena that are difficult to distinguish, but here we report on new data from plural indexicals to help shed light on this issue.

Testing the real-time status of covert movement of wh-operators and QPs in English

Proceedings of the 54th Chicago Linguistic Society, 2018 [Draft]
With Austin Kraft and Jonathon Coltz

Processing of wh-in-situ phenomena in Mandarin Chinese appears to rely on cue-based retrieval mechanisms launched after the wh-operator is encountered, presumably to find its scope position. But, quantifier expressions also scope higher than their surface position. In this paper, we compare the processing profile of QPs and wh-operators, and argue that the latter alone use a cue-based retrieval strategy.

Comparative Psychosyntax

Ph.D. Thesis, University of Maryland 2015 [Link]

Languages are different. This presents challenges to theories of syntax, sentence processing, and child language acquisition, and these challenges are connected. How do adults who speak different languages process their languages, given the fact that they're different? What kinds of learning strategies might an equipotent-language-learning child use to grow into the right kind of language-processor given their linguistic input? I have some thoughts.

New puzzles for shifting indexicals: An Amharic case study

Proceedings of the 44th Annual Conference on African Linguistics 2015 [Link]
With Chris LaTerza, Morgan Rood, Ruth Kramer, and Jen Johnson

In some languages, indexical expressions (‘I’, 'you’, ‘now’) “shift” in embedded clauses that describe the content of someone else's speech or thoughts. One language that has been critical in theorizing about this phenomena is Amharic. Here, we present on some judgments and informal acceptability experiments on how Amharic speakers interpret plural pronouns in shifting contexts.

Disentangling semantic prediction and association in processing filler-gap dependencies: An MEG study in English

[Draft]
With Liina Pylkkänen

How does the brain use semantics and world knowledge to predict what words it'll see next? Normally, the brain can tell when a word is unexpected very quickly – 400ms or less. If a word is unexpected because it generates an implausible semantics, sometimes this can be slower, >600ms. Here we show a ‘sluggish’ response to implausible semantics for NP-verb relations (‘Do you think mops clean fastiidously?’). But, we failed to find any evidence of sensitivity to implausibility if there's a wh-phrase (‘Which mops do you maids clean fastidiously?’). We tie this into theories of how long-distance dependencies are processed generally.

Split ergative agreement in Khaleeji Bastaki

[Draft]
With Alia Albastaki, Benjamin Lang, & Alec Marantz

Some languages use ergative/absolutive cases in one aspect, and nominative/accusative in another aspect. In a subset of these languages, the verb changes which argument it agrees with depending on the case alignment. Khaleeji Bastaki – a Southern Indo-Iranian language spoken in the United Arab Emirates – exhibits such an agreement pattern, but without the morphological case.

Limits on semantic prediction in the processing of extraction from adjunct clauses

[Draft]
With Annika Kohrt, Peter O’Neill, & Trey Sorensen

Studies on the processing of filler-gap dependencies show that we use top-down language knowledge rapidly to constrain our expectations about the rest of the sentence. For instance, if you hear ‘What coffee did you…?’, you're more likely to expect a verb like ‘drink’ rather than ‘eat’. This is also reflected in our ability to avoid construing filler-gap dependencies in island contexts. Following a line of research in semantics that explores why some predicates allow extraction from adjuncts but others do not. Can comprehenders guess that, upon hearing ‘What coffee did you…?’ a resolution like ‘arrive drinking’ is acceptable but ‘work drinking’ isn't? Our findings suggest no!

Retrieval precedes evaluation?: An MEG study on implausible gaps in English

[Poster]
With Liina Pylkkänen

There's a lot of interest in how the brain generates predictions for upcoming material, and how specific that information is. Filler-gap dependencies are an interesting test-case for this, since a fronted wh-object can front-load a lot of cues about the identity of the verb. In this study, we show that the brain distinguishes plausible and implausible thematic relations for sentences without fronted wh-objects, but not for sentences with them. We suggest that this is because comprehenders use semantic cues like lexical association to determine the intended syntactic analysis of a filler-gap dependency, not the other way around.

Morphological recomposition and the concrete/abstract distinction in Bangla

[Poster]
With Swarnendu Moitra & Linnaea Stockall

How do we comprehend morphologically complex words? We've learned a great deal about how the brain ‘decomposes’ complex words into their constituent morphemes, and then rebuilds the meaning of the word. One productive line of work shows that in morphologically complex verbs, syntactic category selectional restrictions are easier and quicker to implement in this ‘recomposition’ stage than semantic selectional restrictions. In a study on Bangla complex nouns, however, we fail to replicate this result. This may teach us something about the differences between how semantic information in the nominal and the verbal domain.