Cortical object representations

What is the neural code for object shape? This thesis investigates how objects are encoded in the human lateral occipital complex (LOC), a region implicated in object recognition; and tests a possible learning mechanism for these representations. Theoretical and psychophysical evidence suggests that convex curvature may be processed in a privileged fashion by the human visual system. We therefore asked whether LOC shows higher sensitivity to convex than concave shapes. We presented pairs of convex and concave stimuli, which were strictly controlled for low-level properties, and measured neural sensitivity with fMRI adaptation. We found high sensitivity for changes in convex shapes in LOC, but not concave shapes, suggesting that convex contours could be important elements of cortical object representations. Second, we asked whether neural shape representations reflect physical or perceptual stimulus properties. We obtained physical, behavioral, and neural similarity measures for novel shapes, and found that the shape similarities in posterior LOC matched physical shape similarities, whereas those in pFs matched perceptual shape similarities. Further, shape representations were similar across subjects in LO, but variable in pFs. These findings indicate that LOC may contain stimulus-based shape representations in posterior regions, and subjective and observer-specific representations in anterior regions. Finally, we asked how cortical object representations might arise in the first place. We studied this question in the context of category formation, hypothesizing that category formation may be based on occurrence frequency. We presented complex visual stimuli following a bimodal frequency distribution, and asked subjects to categorize them into "male" and "female" without feedback. Subjects categorized one frequency peak as "male" and the other as "female". Frequency-based categorization emerged within 12 trials, and led to a categorical perception effect of higher within- than between-category similarity. Passive viewing of the same frequency distribution did not result in frequency-based classification. Together, these results shed light on the nature of cortical object representations: first, convex shape may play a privileged role in representing shape in LOC; second, subregions of LOC appear to contain stimulus-based vs. observer-specific object representations; and third, object categories may arise through the influence of stimulus frequency.