For the threshold of interest there is no doubt a population threshold
around which there is variation due to individual differences. For
traits influenced by multiple genes and environmental sources of
variation (like perception), this will always be a reasonable
description. Try a nested model of some kind, maybe "nlogit"?
