Leaf-cutting ants may be optimal foragers.

Kacelnikl proposes a puzzle for optimal foraging theorists, arising from experiments of Races and Nutiez2, who found that leaf-cutting ant (Acromyrmex lundi) workers changed their behavior in response to foragmg-quality information that was only known to scouts. The workers cut and transferred smaller pieces of parafilm when the scouts had been exposed to a 10% sugar solution than when they were exposed to a 1% solution. Optimal foraging theory, according to Kacelnik, predicts no such dependence of load upon resource quality. After considering several possible explanations, Kacelnik concludes that ‘there is no winning functional model, although the rnformabon transfer hypothesis [smaller loads mean more activity, which increases the rate of transfer of information] is in my view the strongest candidate’. I suggest a simpler explanabon, which IS close to standard optimal foraging theory. The experiments demonstrated an inverse relationship between load mass and the speed with which the ants returned to the colony. Given this relationship, the central-place foraging model predicts an optimal load mass. Suppose that in nature ants encounter trees whose leaves have different mean area density (mass per unit area). Highdensity leaves should be cut into smaller pieces than low-density leaves, in order to maintain the optimal load mass. My suggestion. then, is that the workers in Races and Nuriez’s experiment were acting as if the 10% ‘leaves’ had higher area density than the 1% leaves. How could this come about in their laboratory situation? Workers decide on the size piece to cut on the basis of scouts’ information. There are two obvious possible ways in which the scouts’ signals in the experiment could have been interpreted as piece-area instructions. First, ant signals may only represent a single parameter: high, medium or low, to which workers respond by adjusting the area of leaf pieces. Second, and more likely in my view, there may be a consistent correlation in the field between leaf nutrient content and area-density. If so, it is adaptive for workers to respond to signals of hrgher nutrient content by cutting smaller pieces. Kacelnik reminds us that scientific ‘progress is faster and more fun when models and facts clash’. In an evolutionary context it is also important to think about how well or how poorly a given experimental protocol simulates nature. As Mangel and I have pointed outs, behavior has not been selected to optimize fitness under laboratory conditions. Progress in the study of behavioral adap tations will be faster if the limitations imposed by unnatural experimental constraints are kept In mind.