The diet optimization model with nutrient constraints proposed by Pull
iam and a similar model with a toxin constraint by Stephens and Krebs
are not properly formulated. The appropriate version of the model is f
ormulated and solved for the optimal diet. The model reveals that part
ial preference could occur only for one prey type, while all other pre
y types follow the zero-one rule. The model also indicates that the pr
ey selection criterion changes as the intensity of the constraint is c
hanged, so that we cannot rank prey types. The solution of the model f
or the two-prey-type case shows different qualitative predictions from
those of Pulliam and of Stephens and Krebs. (C) 1995 Academic Press,
Inc.