Measuring the Effect of Natural Selection on Phenotypic Interaction Systems

This paper presents a general model, based on Sewall Wright's method of path analysis, for measuring the effect of selection on phenotypes. The path model analyzes the covariance among measured phenotypic variables (characters) in terms of estimated, unmeasured variables (factors), which account for character covariance in causal biological terms. Factors are recognized by using data about the correlations among characters and information about causes of such correlation. Causal information about the interaction between phenotypes and environments is used for the construction of a path-diagram hypothesis, relating characters and factors to fitness, which is specific to each set of data. In the computation of selection coefficients by the path method, we adjust only for explicable aspects of covariance matrices (factors), rather than using multiple regression to adjust for the full covariance matrix. The sign, magnitude, and pattern of the selection coefficients derived from the initial path diagram are used in conjunction with causal information to refine the diagram through the measurement of additional characters and through experimentation designed to reveal selection mechanisms. One of the primary purposes of the path method is to determine which aspects of the phenotype are subject to selection.