Different genetic identity or distance measures are compared that cons
ider allelic variation within and between populations. Particularily w
e analyse those suggested by Nei (I-S, D-S), Rogers (D-R), Reynolds, W
eir and Cockerham (D-theta), Nei, Tajima and Tateno (D-A), Tomiuk and
Loeschcke (I-TL, D-TL) and Goldstein et al. ((delta mu)(2)). The simul
ations focus on the influence of non-equilibrium conditions on the sta
bility of these measures. The degree of homozygosity of an ancestral p
opulation before it splits into two sister populations is most importa
nt for the stability of the different estimates of genetic identity. I
f populations are not close to their equilibrium homozygosity, a consi
derable bias can occur and, thereby, provide very misleading estimates
of the time span since divergence. The I-TL-measure based on estimate
s of ancestral alleles is more robust than other measures of genetic i
dentity, especially for large population sizes and high mutation rates
. For the infinite allele model, the analysis shows that more precise
estimates of the frequency of ancestral alleles can greatly improve th
e reliability of the estimate of genetic identity in the case of I-TL
For the stepwise mutation model, the TL-measure combines the attribute
s of the D-A- and (delta mu)(2)-measures. The TL-measure is efficient
for the construction of the correct tree topology of related populatio
ns as well as for the estimation of the branch length when protein or
microsatellite data are analysed.