Circularity in linear programming models of optimal diet.

The linear programming model of optimal diet for herbivores has been criticized for being biologically unrealistic, for being too successful given statistical realities, and for being circular. I try to clarify the issue of circularity. Circularity arises if constraint lines are estimated from average values for governing parameters, when the assumed constraints are not effective. This may occur (1) under benign season conditions when consumers do not maximize their food intake, because of costs associated with food processing and storage, or (2) when an unidentified constraint limits intake. To evaluate hypotheses about the factors controlling diet composition, it must be shown that (1) consumers respond appropriately to variation in the parameters controlling constraint settings, and (2) the assumed constraints are close to their maximum (or minimum) settings.