196 
Psyche 
[Vol. 95 
dure, which is corrected for sample size bias. Since all of the poly- 
morphic loci detected in this study had 2 alleles, the method is 
essentially identical to that of Pamilo (1984, Crozier et al. 1984, 
Pamilo and Rosengren 1984), when groups are weighted by the 
number of individuals. Weighting of groups by size appears prefer- 
able in the case of V. pergandei and A. versicolor foundress associa- 
tions, which varied from 2 to 15 queens in these samples. 
Standard errors (S.E.) for the relatedness estimates were obtained 
by a jacknife procedure over groups (Sokal and Rohlf 1981, Pamilo 
1984, Crozier et al. 1984). Simulation studies have shown that S.E. 
estimated by this method tend to be overly conservative and can be 
unreliable when allele frequencies are highly unequal (Crozier et al. 
1984, Wilkinson McCracken 1985). Because of this, use of these S.E. 
in formal statistical hypothesis testing is not justified with the pres- 
ent data. In the case of V. pergandei foundress associations, a more 
robust estimate of r is possible by combining estimates across the 4 
informative loci (Wilkinson and McCracken 1985). For this pooled 
estimate we used the weighted means of F st and Fit across loci (Long 
1986). 
Results 
Veromessor pergandei. Relatedness within V. pergandei foun- 
dress associations does not differ from 0 (Table 1; mean estimate 
across loci = 0.033, Table 2). Allele frequencies of V. pergandei are 
similar between subsites (Table 1); therefore, we treat them as a 
single population. Pooling subsites would inflate estimated related- 
ness if subsites differed. Pgm and Mdh-l allele frequencies are 
highly unequal, which limits their usefulness as genetic markers for 
relatedness; they are most informative in combination with Est-1 
and Est-2 alleles. None of the loci appear linked. 
Acromyrmex versicolor. Genetic relatedness among A. versico- 
lor co-foundresses is no greater than that expected from randomly 
associating queens (Table 2, 3). Negative value of the estimate 
(-0.125) is likely a statistical artifact resulting from unequal Pgm 
allele frequencies and relatively small numbers of queens in each 
foundress association (mean = 3.8) (Crozier et al. 1984, Wilkinson 
and McCracken 1985) rather than an indication that queens avoid 
kin (Hamilton 1972). The small standard error associated with the 
estimate (0.028) indicates little variation in relatedness among 
groups, consistent with purely random mixing of genotypes. 
