Does Morphology Predict Ecology? Hypothesis Testing within a Freshwater Stream Fish Assemblage

To test the hypothesis that morphometric data serve as an index of assemblage structure or niche width, 17 fish species (across 8 genera and five families) from the Roanoke River (VA, USA) were evaluated on the following characters: (a) species-occurrence across 12 microhabitat measures; (b) percent occurrence of 13 food items in gut analyses; and (c) 34 head, fin, and body measurements. Ecological data were arcsine transformed, while morphological data were converted first into common logarithms, then into canonical variate scores and scores from a sheared principal components analysis. A taxonomic matrix based on accepted phylogenetic relationships amongst species of the assemblage was also constructed. Microhabitat, trophic, morphological, and taxonomic matrices were then compared against one another using pair-wise Mantel tests, matrix correlation analyses, and three-way Mantel tests. In analyses across families, two-way Mantel tests revealed that taxonomy and body shape (based on sheared principal component scores) were each significantly related to one another, and to the trophic matrix. Yet, three-way Mantel comparisons indicated that body shapes of constituent species were significantly correlated only with taxonomy, and not with diet. To circumvent the confounding effects of taxonomy, the largest family in the study (Cyprinidae, 8 species represented) was then evaluated separately. The same analyses were performed as before, using trophic, microhabitat, and morphological matrices adjusted for numbers of species. In these analyses, morphology (represented by canonical variate and sheared principal component scores) was significantly related to the microhabitat matrix, and to no others. Previous field studies have indicated that cyprinids segregate along habitat rather than trophic dimensions. Ecomorphological analyses thus appear to be a valid assay for structure within fish assemblages only when limited to within-family comparisons (where the effects of history appear less significant). When many different families are involved in the analysis, historical influences predominate.