Abstract
It is shown that, for technical reasons, the additivity of variance criterion employed by Lloyd (1988) to define a unit of selection is, in almost all models of selection, inconsistent with the possibility that genes are sometimes not the unit of selection. A case when the latter view is particularly attractive is that of heterosis, and the additivity criterion is inadequate in even such an extreme case. The connection between that criterion and the so-called "fundamental theorem of natural selection" is briefly explored. Skepticism is expressed about the value of measures such as variance in efforts to resolve any of the disputes about the "units of selection.".