The Neural Correlates of Explicit Categorization
Flanery, Marci A.
:
2005-04-07
Abstract
PSYCHOLOGY
THE NEURAL CORRELATES OF EXPLICIT CATEGORIZATION
MARCI A. FLANERY
Dissertation under the direction of Professor Thomas J. Palmeri
Understanding categorization has been the goal of an extensive amount of research in cognitive psychology. However, current research based on behavioral and modeling methods does not support any single theory of categorization because several competing theories are able to account for many of the same behavioral results. In this dissertation, functional magnetic resonance imaging (FMRI) was used as an additional source of evidence for discriminating alternative theoretical accounts of behavioral differences observed on categorization tasks tested with neuropsychological and normal subject populations. In Experiments 1 and 2 subjects performed two categorization tasks while we collected functional images. The tasks used were variations of prototype-distortion and probabilistic-cue categorization tasks that had previously been used in behavioral neuropsychological studies. Results of these previous studies had suggested that different categorization systems, sensitive to the different category feedback structures (deterministic versus probabilistic) used in the two tasks, supported performance in these tasks. However, similarities and differences in neural activation associated with performance during both the early and late stages of these tasks in our normal subjects suggests that the reported behavioral differences may be due to the use of different strategies that are associated with the type of stimuli used (prototype-distortions versus individual cues) and the type of learning paradigm (feedback versus exposure) rather than the underlying category feedback structures (deterministic versus probabilistic) that were traditionally used in the two tasks. Experiment 3 investigated changes in neural activation associated with a single categorization task over the course of eight learning epochs that were completed in a single training session. Previous behavioral and modeling research had demonstrated a qualitative shift in categorization behavior over learning, but the theoretical explanation for this behavioral shift was debated. Utilizing a priori regions-of-interest, neural activation consistent with a shift in representation from rules to exemplars was found.