The results from the current study should be considered with respect to a number of limitations. We applied a normalizing transformation to our alcohol data prior to model fitting. This has the potential to eliminate some sorts of the interaction effects that might have been detected in the untransformed data. A related issue is that the moderation effects may be scale-specific. That is, the moderation may not be present when religiosity is scored in a different way. We tested this possibility by re-analyzing the data using summed scores of religiosity rather than the mean. We found that, similar to the results described above, there was a significant moderating effect of religiosity on genetic and environmental effects in adolescence but not young adulthood, and that these moderations went in the same direction as those previously reported. Consequently, these results add confidence to the findings discussed above. Furthermore, as described in the method section, our model does not test for the moderation of those genetic effects on alcohol problems shared in common with religiosity; the interaction effects we report here are on
