MIT CogNet, The Brain Sciences ConnectionFrom the MIT Press, Link to Online Catalog
SPARC Communities
Subscriber : Stanford University Libraries » LOG IN

space

Powered By Google 
Advanced Search

 

Parametric Variation of Structural Similarity in Object Recognition

 Jane Joseph and Gerry Piper
  
 

Abstract:
Three-dimensional structural representations are critical for object recognition, and the degree to which objects overlap in structure within a category determines the efficacy of recognition. Functional neuroimaging studies of object recognition are providing converging evidence that lateral occipital and posterior fusiform regions are involved in processing structural descriptions; however, processing object structure was not directly manipulated in some of these studies. The present study examines parametrically varied structural similarity (SS) of object pairs to establish behavioral and neuroimaging profiles modulated by structural object processing. Eight participants viewed two animal pictures simultaneously. On "yes" trials, the two pictures represented the same animal and on "no" trials the two animals were different. On "no" trials, the animal pairs parametrically varied across four levels of SS, as determined by normative similarity ratings). Mean pair-wise similarity rating were 183, 422, 539 and 736 (on a scale of 1000) for SS1, SS2, SS3 and SS4 objects, respectively. Decision latencies and errors increased as SS level increased, F(3,3) = 10.8, p < .05, F(3,3) = 61.7, p < .01, respectively, such that discriminating "dolphin" from "spider" (low in SS) was much faster and more accurate than discriminating "rhino" from "hippo" (high in SS). Initial fMRI data in individual subjects shows strong modulation of lateral occipital and posterior fusiform regions by SS level, in agreement with previous functional imaging studies.

 
 


© 2010 The MIT Press
MIT Logo