In this paper we describe how morphological castes can be distinguishe
d using multivariate statistical methods combined with jackknife estim
ators of the allometric coefficients. Data from the polymorphic ant, C
amponotus rufipes, produced two distinct patterns of allometric variat
ion, and thus two morphological castes. Morphometric analysis distingu
ished different allometric patterns within the two castes, with overal
l variability being greater in the major workers. Caste-specific scali
ng variabilities were associated with the relative importance of first
principal component. The static multivariate allometric coefficients
for each of 10 measured characters were different between castes, but
their relative magnitudes within castes were similar. Multivariate sta
tistical analysis of worker polymorphism in ants is a more complete de
scriptor of shape variation than, and provides statistical and concept
ual advantages over, the standard bivariate techniques commonly used.