This is a very interesting cell in layer 4B. Notice how sharp the
response is and that fact that the probability that off-optimal
orientations will drive the cell are driven all the way down to
zero. Also, after the response peaks one can see that the shape
of the probablity distribution looks like a Mexican hat: the angles
that are less effective in evoking a spike are the ones flanking
the optimal, and not the orthogonal one (which is what one would
expect from a simple feed-forward model.