1983] 
Herbers — Social Organization in Leptothorax 
381 
A startling difference between species was the strong morphologi- 
cal bias underlying polyethism in L. longispinosus but lacking in L. 
anihiguus. Size differentiation according to task was clear in L. lon- 
gispinosus-, the pattern strongly corroborated influences of role and 
caste delineation from the transition matrix (Herbers and Cun- 
ningham 1983). For L. anihiguus, however, there were relatively few 
differences in average worker size among behaviors, and those few 
significant differences were not correlated with roles inferred from 
behavior transitions. Perhaps the absence of morphological correla- 
tion was due to the fact that the range of worker size was narrower 
for L. anihiguus (Figure 3) than L. longispinosus (Herbers and 
Cunningham 1983); a small size range of workers may have pre- 
cluded task specialization by size for L. anihiguus. 
Both species displayed considerable among-colony variation with 
respect to behavior frequency and time budgets. To ascertain the 
relative importance of within- and between-species variation, cluster 
analyses were performed. These techniques involve calculating sim- 
ilarity indices for all possible pairwise comparisons. Then each unit 
(i.e. colony) is placed in a dendrogram based on its similarity to 
every other unit. If behavior data reflect phylogeny, then the three 
L. anihiguus colonies should form one cluster while the four L. 
longispinosus form a second. Moreover, one might expect colonies 
with similar numbers of queens to cluster more closely to each other 
than colonies with different queen numbers. 
The simplest comparisons used the matching coefficient, or 
number of behaviors shared by two colonies relative to the total 
number observed over all (Cole 1980). This similarity index utilizes 
information only on presence or absence of behavior types in the 
ethogram, thereby ignoring relative frequency. Analysis of matching 
coefficients yielded the dendrogram of Figure 5. This simplest clus- 
tering technique produced the satisfying results that L. anihiguus 
colonies were more closely related to each other than to L. longispi- 
nosus nests: the three formed a distinct cluster. Moveover, queen- 
right L. anihiguus colonies were more similar to each other than to 
the queenless nest; this result, however, was simply an artifact of the 
presence of behaviors directed towards queens ( ATQ, RQ, ALQ) in 
queenright but not queenless ethograms. Even so, L. anihiguus col- 
onies did cluster as expected. However, the L. longispinosus nests 
did not. Two (LI -A and Ll-B) clustered closer to L. anihiguus nests 
