Date of Award

Spring 1-1-2010

Document Type

Thesis

Degree Name

Master of Arts (MA)

Department

Psychology & Neuroscience

First Advisor

Randall O'Reilly

Second Advisor

Tim Curran

Third Advisor

Matt Jones

Abstract

Although the rich bidirectional architecture of the ventral visual stream has been documented for some time, relatively little work has gone into understanding its function in object recognition. Recently, computational modeling work has suggested that the computations performed within a bidirectional architecture (recurrent processing) could be beneficial to object recognition by clean- ing up ambiguity in input signals, thus providing robustness to degradations like occlusion that underspecify the visual stimulus. The research described here tests this claim by using visual masking to disrupt this recurrent processing. In one experiment and a series of accompanying modeling simulations, it is shown that there is a significant interaction between a mask and oc- clusion such that performance on an object categorization task is differentially impaired when a moderately occluded stimulus is masked compared to a relatively unoccluded one. Furthermore, the modeling simulations provide a mechanistic explanation of how recurrent processing resolves ambiguity as well as how masking interrupts this process. Together, the results of the experiment and modeling simulations suggest that object recognition is a dynamic process characterized by interactions between adjacent areas along the ventral visual stream.

Share

COinS