Date of Award

Spring 1-1-2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

First Advisor

Matt Jones

Second Advisor

Michael Saddoris

Third Advisor

Alice Healy

Fourth Advisor

Tim Curran

Fifth Advisor

Michael Mozer

Abstract

Relational concepts pervade daily life, as people are regularly required to comprehend, articulate, and reason about relational ideas and scenarios. Critically, these processes might be altered by how such concepts are represented. The dominant theories of relational learning have been built on the assumption that relational concepts are represented compositionally, based on the relationships among a concept’s components. However, these theories have typically neglected the possibility that a concept’s components can be consolidated or chunked into a unitized concept, producing a representation that is devoid of the concept’s component parts. The distinction between compositional and unitary representations of relational concepts is a natural consequence of structure-mapping theory, but its psychological implications have not been explored. This paper reports 7 studies that examine how people represent relational concepts and how such representations affect relational learning. The general take away from these studies is that people do indeed appear to be capable of representing relational concepts in two fundamentally different ways, unitarily and compositionally. Furthermore, unitary representations seem to lead to better relational learning than compositional representations, especially for inference-based tasks. However, the data suggest that there might be various factors that interact with how representation affects relational learning (e.g., individual differences in representation, type of task, type of comparison). The conclusion that follows from these studies is that unitary representations might incur less cognitive load than structural alignment of compositional representations, and thus may be the default for everyday relational reasoning.

Share

COinS