Learning and representation of relational categories

Relation-based category learning is based on very different principles than feature-based category learning. It has been shown that relational categories are learned by a process akin to structured intersection discovery, which is formally powerful than feature-based associative learning, but which fails catastrophically with probabilistic category structures. This research provided consistent evidence that relational concepts are qualitatively different from featural concepts, and they are also learned in a qualitatively different manner. Experiment 1 showed that relational category learning with probabilistic structures can be improved by comparing systematic pairs of exemplars, where shared relations between the exemplars can be abstracted. Experiment 2 showed that comparing the exemplars to the prototype can improve learners’ ability to learn probabilistic relational categories in terms of prototype-plus-exception rules. Experiment 3 and 4 examined further the distinction between feature-and relation-based category learning using a dual task methodology. Experiment 3 revealed that featural category learning was more impaired by a visuospatial dual task than by a verbal dual task, whereas relational category learning was more impaired by the verbal dual task. Experiment 4 examined how the dual task that involves more relational information interacts with feature-and relation-based category learning. The results showed that there was no reliable difference between two category learning. Taken together, Experiment 3 and 4 results suggest that in contrast to featural category learning, which may involve mainly non-verbal mechanisms, relational category learning appears to place greater demands on more explicit and attention-demanding verbal or verbally-related learning mechanisms. The findings presented in this dissertation contribute to the growing body of theoretical and empirical results suggesting that relational thought is a qualitatively different thing than the kinds of thinking and learning afforded by feature-based representations of the world.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Psychology

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2014

Author dc:creator

• Jung, Wookyoung

Contributors dc:contributor

• Hummel, John E.
• Ross, Brian H.
• Dell, Gary S.
• Morrison, Robert G.
• Cimpian, Andrei

• Relational category learning
• Featural category learning
• Category learning
• Schema induction
• Relational invariants
• Dual task
• Category learning algorithms

Statement dc:rights

• Copyright 2013 Wookyoung Jung

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/46904

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/46904