This is another cell in layer 5 which is very sharply tuned.
Notice, that the profiles have a very clear Mexican-hat shape:
the less effective orientations in driving the cell are at
+/- 45deg from the preferred orientation which is around 90deg.
This kind of responses are difficult to replicate with a simple
feedforward models, for which the worse orientation is the one
orthogonal to the preferred one. This is clearly not the
case here... Actually, the cell shows a local maxima at the
orthogonal orientation (see also cells in layer 6).