Reader Comments

Post a new comment on this article

what about aspects of receptive fields not in the model?

Posted by sydneyunivisphys on 08 Jun 2007 at 23:07 GMT

The model models retinal/LGN cells as single mechanism (‘centre-only’) and isotropic. Most of the experimental literature (eg. Rodieck, Thibos and Levick, Leventhal and Shou) points to a) surrounds, and b) radial anisotropy (receptive fields are elongated along the axis from the area centralis), in receptive fields in cat retina/LGN. Would the model survive or be improved with the addition of those elements?

RE: what about aspects of receptive fields not in the model?

dario replied to sydneyunivisphys on 11 Jun 2007 at 20:42 GMT

Thank you for the comment.

(a) In related work, I have specifically modeled the surrounds, and they did not change the basic results reported here. So, I believe the results will remain essentially unchanged.

(b) As you correctly point out, I did not consider radial anisotropy. I can offer the following conjectures.

1. One would expect that when a receptive field gets input from pairs of nearby ON/OFF-center cells, the anisotropy will not contribute much. This is because the axis of the ON/OFF subregions will dominate the slight elongation of the inputs in terms of the preferred orientation of the cell.

2. Instead, the radial anisotropy of RGC receptive fields may be felt in regions dominated by input of only one RF sign. Thus, regions that are now defined as 'unoriented' in the model, would display a broad tuning with radial preference. Interestingly, this would predict that broad tuned cells will all tend to have the same (radial) orientation preference. It would be interesting to test this prediction.