Speaker: Melissa Kline (Massachusetts Institute of Technology)

Topic: "Structured Representations of Events unite verb meaning, inference about complex scenes, and early cognitive development"

Reception: All are invited to a reception following the talk

Structured Representations of Events

unite verb meaning, inference about complex scenes,

and early cognitive development.

Melissa Kline

Massachusetts Institute of Technology

Language breaks events into pieces: a sentence like

Jane_agent  danced_manner  across_path  the floor_ground

factors a directed motion event into components that can’t easily be separated in space and time – the dancing, the crossing, the floor, and Jane herself are all "going on" at once. Furthermore these events include not just the objects themselves but the relationships between them, in which particular participants play particular different roles (The floor danced across Jane certainly means something different from Jane danced across the floor). Argument structure, the study of how sentence structure is affected by meaning, has revealed a number of patterns that predict how verbs fit into sentences. But are these patterns a property of the human mind? How do children pick out just the right parts of an event to attend to, and how do they think about these events in order to form sentences and get their meanings across? 

In my work, I use theories of argument structure to motivate empirical work that aims to identify the basic building blocks of meaning that underlie adult language use, language learning, and all of the complex tasks we accomplish as humans by mediating our plans and goals through the channel of language. I will present a series of studies with young children showing how abstract concepts like CAUSE and MANNER support children’s language learning and drive how they attend to different kinds of events. I will discuss extensions of this work to pragmatic phenomena, and implications for semantic theory, early cognitive development, and the architectures that may support processing of sentence-level meaning in the brain. 

The goal of my research program is to create testable models of events cognition that are developmentally plausible, make falsifiable predictions about human cognition, and are consistent with cross-linguistic patterns. The structure and content of human language reflect the computations of our underlying cognitive capacities, filtered by the pressures of communication: understanding how these linguistic abilities are instantiated in the human mind can reveal how we knit together our understanding of individual entities into representations of their interactions in a dynamic and unfolding world.