In this study, we used a combination of geometric morphometric and evolutionary genetics methods for the inference of possible mechanisms of evolutionary divergence. A sensitivity analysis for the constant-heritability rate test results regarding variation in genetic and demographic parameters was performed, in order to assess the relative influence of uncertainty of parameter estimation on the robustness of test results. As an application, we present a study on body shape variation among populations of the poeciliine fish Poecilia vivipara inhabiting lagoons of the quaternary plains in northern Rio de Janeiro State, Brazil. The sensitivity analysis showed that, in general, the most important parameters are heritability, effective population size and number of generations since divergence. For this specific example, using a conservatively wide range of parameters, the neutral model of genetic drift could not be accepted as a sole cause for the observed magnitude of morphological divergence among populations. A mechanism of directional selection is suggested as the main cause of variation among populations in different habitats and lagoons. The implications of parameter estimation and biological assumptions and consequences are discussed.