Figure 8. The surround may amplify spatial attention under natural viewing conditions.
(A) Original image. (B) Model neurons tiled the image. Each pixel contained one model neuron with its receptive-field centered on that pixel. An example cRFs (solid red) and surround (red dashed) for one neuron are shown. The radius of each neuron's surround was approximately five times larger than the radius of its cRF. The model neurons computed local contrast within the excitatory and suppressive component of their receptive field. The response maps show each neuron's response: neurons near high-contrast regions responded most as indicated by the luminance of the pixels. (C) Original image scaled according to the response map in (B). (D) Attention was directed to the left eye. Attention weighed the excitatory and suppressive inputs with its Gaussian kernel, resulting in stronger responses of the neurons with receptive fields near the attended location relative to neurons with receptive fields outside the locus of attention. (E) Original image scaled according to the response map in (D), illustrating the way attention changes the visual representation. (F) Attention modulation of each neuron, defined as (responseAtt - response) / (responseAtt + response). Here, response is the response map without attention as in (B), while responseAtt is the response map with attention as in (D). Upper panels (B–F) are based on model neurons with a suppressive surround. Lower panels (B–F) are based on model neurons without a suppressive surround, but with the same amount of suppression inside the cRF as the neurons with a suppressive surround.