234 
Psyche 
[Vol. 88 
items; to examine their defensive behavior, small Solenopsis invicta 
workers were introduced into their container. 
The two morphological castes were easily distinguished on the 
basis of mandible thickness. In order to record division of labor 
among individuals of different sizes, yet similar proportions — so 
critical in weakly allometric species such as Daceton — the minor 
workers were arbitrarily divided into small and medium size classes, 
also distinguishable by eye. For convenience, these subcastes will be 
referred to as “minors” and “medias”, as in Wilson 1978. By-eye 
assignment of caste to preserved specimens, subsequently measured, 
produced the following definitions of size classes and castes: minors, 
head width less than 1.12 mm; medias, head width between 1.13 and 
1.64 mm; majors, head width greater than 1.65 mm. After some 
initial die-off, the colony contained fifty-two adults for the duration 
of the study: one queen, thirty minors, fifteen medias and six 
majors. 
Results 
O. versicolor is in fact polyphagous. Live flightless Drosophila 
were readily accepted, and young were successfully raised on this 
diet. The ants also accepted Drosophila larvae, and, not surprisingly, 
collembolans. (Alternative foods were not offered simultaneously to 
test preferences; however, most collembolan specialist species would 
not touch other prey even if starving.) The same colony had been fed 
mealworm and cockroach fragments, various diptera and honey- 
water in Australia (B. Holldobler, pers. comm.). 
The ethogram or behavioral catalogue of workers and queen is 
presented in table 1, which gives both numbers of individual acts 
performed and the relative frequencies of acts in the total repertory 
of each caste. The colony repertory consisted of twenty-seven 
categories of behavior. (Worker regurgitation with the queen was 
added as a twenty-eighth because it was seen twice during prelimi- 
nary observations, though never during the study.) The observed 
minor and media repertories both contained twenty-seven behavior 
categories; the observed major repertory contained twenty-four. 
Using the Fagen-Goldman statistical method, the estimated total 
repertory size for minors — the observed repertory plus an estimate 
of the number of categories not observed — was calculated to be 
twenty-nine, with a 95% confidence interval of (27,32) acts. The 
