Divergent life histories among populations of the fish Brachyraphis rhabdophora: detecting putative agents of selection by candidate model analysis

Studies of natural selection in the wild almost always begin by examining patterns of association between phenotypic adaptations and environmental factors thought to shape evolutionary change. Unfortunately, many studies pay little attention to the effects of model selection on the evolutionary inferences drawn from such correlative data. In this study, I employed a candidate model analysis to examine four potential causes of life-history evolution in the livebearing fish Brachyrhaphis rhabdophora. Combining factor analysis with path analysis, I constructed a nested set of 17 models that represent the hypothetical effects of four putative selective agents (mortality, density, resource availability, and habitat stability) on life-history evolution in this species. These models represent both direct and indirect effects of selection on the life history. Using the Akaike Information Criterion to distinguish among models, I found that simple models that contained only single selective agents most parsimoniously explained life-history divergence among 27 B. rhabdophora populations. Surprisingly, the four putative selective agents could not be distinguished, suggesting that the selective environment could be composed of a single selective agent confounded with other environmental factors, or could be composed of a suite of environmental factors that act in concert to shape the life history. In addition, comparisons among more complex models indicated that direct effects of selective agents appear to have primacy over combinations of indirect selective interactions in explaining intraspecific variation in B. rhabdophora life histories.