If you want to count sheep, count the legs and divide by four. This piece of faux folk-wisdom has, as is usual in such cases, a grain of truth. The human eye finds it easier to distinguish elongated objects if the axes of the object are separated and perpendicular (or so I believe). It is easier to count the candles mounted on a cake than the same candles arranged in a line. This – | | | | | | | – is easier to count than this – _ _ _ _ _ _ _ , I feel. (I’ve used Google to see if I can find evidence and came across this, which seems to align with what I am saying, though I’ve not accessed the paper).
If I am correct it is easier to, say, count the horns on a herd of cattle and divide by two than count their backs. It occurs to me that an optical-electrical counting device might have issues in this regard too, since a leg might stand out from the background, and produce a short pulse in the sensor, but a whole cow might take a while and its colours would blend into the next cow. Of course, one could always use higher technology to resolve the issue with respect to cow counting, (RFIDs in ear tags would be an obvious solution), but it doesn’t solve the wider issue.
Maybe the reason that the counting device and the eye/brain find it easier to distinguish objects orientated (roughly) perpendicular to their (roughly) linear arrangement is similar. If they are (roughly) aligned in the same direction as their linear arrangement they may, possibly, overlap, and this can confuse sensor and/or eye. Was that one, two, or three objects that passed the sensor? It’s easy if they are perpendicular, but harder if they are aligned.
I’m pretty much reduced to saying the same thing in different words, but I hope that what I am trying to get at is clear. It may or may not be relevant that humans and higher primates tend to stand more or less vertically, so one individual is more easily distinguished from others than an individual cow is from the herd.