EXPLORING FITNESS SURFACES

We present a nonparametric method to estimate the form of multivariate selection on a suite of quantitative traits. Its advantages are threefold. First, the procedure is flexible and does not force estimates of the surface to conform to a specific mathematical shape. The need for a flexible method is illustrated by an example using quadratic regression. Second, estimates of multidimensional surfaces can be visualized in two or three dimensions. This simplification is accomplished by making cross sections of the surface in the few most interesting directions. Finally, the method is designed to handle survival and other nonnormal fitness components. We apply the procedure to two data sets. In song sparrows, the survival surface is approximated by a ridge favoring an allometric relation between body mass and wing length. Survival in human infants rises steeply with increasing birth mass and maternal gestation period to a broad flat dome. Our results emphasize the ubiquity of ''correlational'' selection and illustrate how traits jointly determine fitness.