EXAMINING SELECTION ON THE MULTIVARIATE PHENOTYPE: PLANT RESISTANCE TO HERBIVORES

Cost-benefit theory posits that stabilizing selection, produced by a trade-off between associated costs and benefits, often maintains phenotypic traits at intermediate equilibrium values. Measurement of selection on one type of trait, resistance to herbivory, should provide evidence to test this prediction. However, most plants host more than one species of herbivore, and resistance to various herbivores may be phenotypically correlated. Consequently, selection must be measured on a multivariate phenotype, which may produce a very complex selection gradient. Canonical analysis of the matrix of quadratic coefficients of the phenotypic selection gradient is presented as a method to simplify the interpretation of the estimated multidimensional adaptive surface. Because this analytical method finds the canonical axes of second-order effects on the selection surface, it shows where nonlinear, and possibly stabilizing, selection is strongest. The resulting index traits may generate hypotheses regarding underlying resistance traits to which more than one type of herbivore responds.