Not all adjectives have the same semantic representation. Something can be 'bigger' or 'fuller', but not 'more wooden' or 'more extinct'. Likewise, something can be 'completely full' but not 'completely big'. In this line of work, I investigate psycholinguistic evidence for these semantic differences in lexical representation, in both child and adult participants, and how participants' responses in these experiments can inform our understanding of the interaction of semantics and pragmatics in this area.

I also turn an eye to the acquisition process and ask what surface-level patterns a learner could use to resolve the mapping problem and whether these cues actually perform this function in language learning experiments with novel words. I have been especially interested in extending the 'syntactic bootstrapping' approach to word learning in a way that takes account of the link between surface-level distributional patterns and underlying semantic representaitons as well as lexical and collocation frequencies to determine the extent to which certain patterns are informative to the language learner. I have targeted in particular the role of adverbial modification, and have investigated their role in novel adjective learning, and more recently in novel verb learning.

Collaborators: Chris Kennedy, Jeff Lidz, Sudha Arunachalam

Go to my publications on this topic.

Number words are notoriously challenging for children to acquire, and it is not until children are around three years of age that they consistently map these words onto their correct representations and demonstrate knowledge of the domain of number. This topic has gained a lot of attention, and for good reason. The ability to count and use arithmetic is one of the defining characteristics of humans. I am interested in the language component of this learning process. Because number words appear in the same places in language as other words like adjectives or quantifiers, it is possible that part of the challenge for language learners involves teasing apart these lexical representations. How is this done? Can systematic differences in the syntax inform children about systematic differences in the semantics? Can this syntax-semantics mapping account for some cross-linguistic differences?

At the same time, I am interested in what young children know about the interpretations of sentences containing number words. Often these sentences are ambiguous, and the only way to appreciate this ambiguity is to have a grammar that can generate the representations needed to capture the right meanings. For example,Two boys pushed a car could mean that the two boys worked together collectively to push the car, of that each pushed their own car, or that each pushed the same car at different times. These different interpretations may be accounted for by the scopal interaction of the subject and object, by the presence or position of a distributivity operator, by appealing to events, or by the interaction of these features with others in the linguistic system.

On the other hand, these sentences might not be ambiguous, and the availability of one or the other interpretation may be constrained by structural or lexical information, such as active/passive voice, the presence of an overt distributivity operator such as 'each', or the presence of 'together', Experiments on these topics are complemented by corpus data and responses from adult participants that indicate that the language adults use to describe events may serve to highlight the competing interpretations, and that adults' interpretation of these sentences calls on a variety of semantic, pragmatic, syntactic, and real-world knowledge.

Collaborators: Julien Musolino, Rochel Gelman, Roger Schwarzschild

Go to my publications on this topic.

In order to understand a sentence such as 'John read every book that Mary did', your grammar has to allow you to resolve the instance of verb phrase (VP) ellipsis in did. But there's a catch. If you simply insert the entire verb phrase in that spot, you will copy in did, If you can move the phrase every book that Mary did out of the VP first (by Quantifier Raising), you can arrive at a representation that says (roughly), 'For every book that Mary did read, John read those books'. I have been interested in whether preschoolers have the grammatical capacity to interpret these--and other more complex cases--correctly, and what their responses say about both the grammar and language processing, for both children and adults. This work demonstrates that children have the Quantifier Raising operation in their grammar, and entertains the possibility that this operation may not be bounded by Tense. It also suggests that the activation of different structural components in the incremental processing of these sentences can give rise to differences in which interpretations are accessed.

Collaborator: Jeff Lidz

Go to my publications on this topic.

In my dissertation, I investigated what children know about differences among the semantic representations of gradable adjectives (GAs) and how they could have arrived at this knowledge. Experimental results demonstrate that children as young as age three distinguish between relative (e.g., big, long), maximum standard absolute (e.g., full, straight), and minimum standard absolute (e.g., spotted, bumpy) GAs in the way that the standard of comparison is set and how it interacts with the discourse context. A corpus analysis demonstrates that there are statistically significant patterns of adverbial modification among these GAs and others like them. These findings--combined with results from an experiment showing that 30-month-olds use adverbial modifiers they are not necessarily producing to assign an interpretation to novel adjectives--suggests that infants perform a probabilistic analysis of the input to learn about abstract differences within this category.

The experimental work presented in my dissertation was published in the 'adjectives and adverbs' section above, and some of the ideas in the last chapter gave rise to current work in the 'number words semantics' section above.

Thesis committee: Jeff Lidz (chair), Chris Kennedy, Stefan Kaufmann, Sandy Waxman

Download a pdf of my dissertation.

In more recent work, I have been investigating the relationship between sound and meaning, specifically the role of prosody in conveying information and the interaction between prosody and pragmatics. Early work in graduate school with Janet Pierrehumbert on a project investigating the role of context in stress shift was enough to get me hooked on this topic. My recent work involves a range of topics and collaborators, all united by an interest in probing the ways in which the prosodic and acoustic delivery of an utterance can be informative about information structure and meaning in context.

In one line of work, my collaborators and I have followed up on a claim that wh-in-situ questions in French are licensed by an intonation morpheme, requiring a rising intonational contour on par with yes/no est-ce que questions. In a production study with native French speakers, we found that most speakers did produce sentences as predicted, but that the rising contour was mitigated by information structure. Our results further called into question the precise nature of the intonation morpheme--whether it is obligatorily realized as a H% or whether it is underspecified and has a choice of docking sites.

In a second line of work, we have asked whether children modulate features of their speech when communicating in different speech contexts. For example, do they speak differently when they are helping someone learn a new word? Our complementary production and perception studies have demonstrated that clear speech in preschoolers results in features such as longer vowel duration and higher intensity, and that these characteristics are perceptible. Our findings demonstrate, then, that there are aspects of speaker-hearer interaction that chidren as young as 3 are aware of, and that they are sensitive enough to these aspects to alter their speech accordingly.

Licensing of French wh-in-situ interrogatives (Collaborators: Viviane Déprez, Shigeto Kawahara)

Production of clear speech by preschoolers (Collaborator: Shigeto Kawahara)

Prosody in ambiguity resolution