Laurel Perkins

University of Maryland Linguistics | perkinsl [at] umd.edu

Research

What we can learn depends on what we already know; a child who can’t count cannot learn arithmetic. Language acquisition, like learning in general, is incremental. At each step of linguistic development, children have access only to a limited set of data about their language, consisting of the portions of adult speech that they are able to parse with their current knowledge of the grammar of their language. My research aims to understand the developmental trajectory of language acquisition: how do children leverage their linguistic knowledge in order to learn from the partial data available to them at each stage in development? I work with Jeffrey Lidz and the Project on Children's Language Learning, with additional collaborators for specific projects below. For a full list of publications and presentations, see my CV.

NON-BASIC CLAUSES IN EARLY SYNTAX ACQUISITION

I'm interested in how children at the very beginning stages of syntactic development come to identify the structure of sentences that do not follow the canonical word order of the language. These so-called "non-basic" clauses, like wh-questions, are frequent in children's input and pose a challenge for clause structure and verb argument structure acquisition. How do children learn that "what" is standing for the object of the verb in a sentence like "What are you holding?", when it does not occur in canonical object position after the verb? And if children do not recognize that "what" is an argument of the clause, how do they avoid faulty inferences about the argument structure of the verb in this sentence?

I use a combination of behavioral methods—preferential looking and the sequential listening preference procedure—to test when English-speaking infants represent the dependency between the wh-phrase and the verb in a wh-question or relative clause. With Mina Hirzel, I'm also looking at what strategies children are using to process wh-questions over the course of development. In addition to these behavioral experiments, I use computational modeling to study how infants might acquire verb argument structure in tandem with the syntax of non-basic clauses. In work with Naomi Feldman, I've investigated how learners might be able to "filter" their input in order to avoid faulty inferences about verb transitivity, before they are able to identify non-basic clauses in their language. We are also modeling how learners might use the signal from an expected but missing argument of a verb to identify when a non-basic clause is present, and thus begin to learn the form of wh-questions, relative clauses, and other non-basic clause types in their language.

•  Perkins, Feldman, & Lidz (under review). “The power of ignoring: Filtering input for argument structure acquisition.” [lingbuzz]

Perkins & Lidz (under review). “Filler-gap dependency comprehension at 15 months: The role of vocabulary.” [lingbuzz]

Perkins, Feldman, & Lidz (2017). “Learning an Input Filter for Argument Structure Acquisition.” Proceedings of the 7th Workshop on Cognitive Modeling and Computational Linguistics.

 

 

TRANSITIVITY AND VERB LEARNING

In joint work with Tyler Knowlton, Angela Xiaoxue He, Mina Hirzel, and Alexander Williams, I'm investigating how infants represent and use basic clause structure in order to draw inferences about aspects of verb meaning. Infants use a verb's distributions in transitive and intransitive clauses to draw inferences about its meaning and argument structure: this is syntatic bootstrapping. We'd like to understand how learners represent transitivity in their input in order to enable these inferences, and what inferences they draw.

Our goal is to distinguish two bootstrapping hypotheses: an expectation that clause arguments and event participants must match in number, vs. an expectation that particular argument positions will link to particular participant roles, e.g. transitive Subject to Agent and Object to Patient. Differentiating these hypotheses requires first establishing how infants view particular events in the world, independent of language. To this end, we've designed a habituation-based task to assess how many participants infants represent when viewing a particular stimulus scene. With this information, we can then ask how they will map a sentence to that scene: can a transitive clause describe a scene that they naturally represent under a 3-participant concept, provided it includes the right participant relations, or must it describe a concept represented with only 2 participants? Answering these questions will help us understand whether children are biased to rely primarily on the number of clause arguments when learning the meanings of verbs, or whether they represent and use more fine-grained information about the grammatical and thematic relations of those arguments.

•  Knowlton, Perkins, Williams, & Lidz (2018). “Getting a grip on infants’ event representations: participant number in TAKE and PICK-UP.” Poster presented at the XXI Biennial International Congress of Infant Studies (ICIS), Philadelphia.

•  Williams, Perkins, He, Björnsdóttir, & Lidz (2017). “A New Test of One-to-One Matching Between Arguments and Participants in Verb Learning.” Poster presented that the 42nd Boston University Conference on Language Development (BUCLD 42), Boston.