Abstract
HJA has a lesion of the lingual and fusiform gyri. He is unable to integrate local object features into global form and is poor at grouping and segmenting overlapping objects1. In the mean orientation task2 observers must integrate local Gabor elements to judge an arrays overall orientation. This task is probably mediated by V1 or V2, but observers' ability to vary how orientation samples are pooled suggests that something more complex than the filter-rectify-filter model is needed.
Method
HJA indicated whether the mean orientation of arrays of Gabor elements (sf approx 2 cycles/deg, env s.d approx 4 cycles/deg) was to the left, or right of vertical. The orientation of the modulation in each Gabor was selected from a Gaussian distribution with a variable bandwidth. The bandwith's standard deviation was varied from 0 (all aligned) upwards. The density of the array was varied by changing the diameter of the stimulus (3, 6, 12 deg) and the number of elements (4, 64). Mean orientation thresholds for each array bandwidth were fitted with a variance summation model to estimate internal noise and sampling efficiency.
Results
HJA's performance judging the orientation of individual Gabors, arrays of aligned Gabors and arrays with increasing bandwidth was similar to non-lesioned observers, although he appeared to use slightly less samples. Increasing the density of the gabor array affected his performance much less than non-lesioned observers.
Conclusion
HJA appears to be able to integrate low level local information for this task. His performance does not deteriorate with crowded displays, suggesting anomalous lateral interactions. Like non-lesioned observers he is able to change his sampling strategy with the number of display elements. This flexibility must, therefore, be a property of V2 or V1 neurones or a hitherto unrecognised top-down pathway must exist that by-passes the lesion.
Giersch et al (2000) Cognitive Neuroscience 17 p731;
Dakin (2001) JOSA A18 p1016
funded by the Wellcome Trust, UK