Caution is essential, however, in interpreting the results of our
analyses. The results for the crural index are a good case-in-point,
where the population structure-only model is the best fit despite
a substantial apparent effect of latitude in the combined model
(Table 2). Because latitude is distributed in a way that can sometimes
be very similar to the distribution of a neutrally evolving trait,it is possible that selection and random genetic drift might be highly confounded in some cases. This confounding of terms in the model may account for this result rather than a genuine lack of effect of latitude. As there is such high evolutionary variance for phenotypic characteristics (Rogers and Harpending, 1983; Lynch,1990), we do not expect it to happen all of the time, but it is a concern that we all should carry forward into future investigations.