ABSTRACT
Since the seminal paper by Lande and Arnold (1983), the study of selection in natural populations has been placed in a quantitative framework concordant with evolutionary theory. Recent books by Manley (1985) and Endler (1986) are further evidence of the growing interest in the application of quantitative methods to the study of natural selection. A large number of recent papers (including Arnold, 1986; Koenig and Albano, 1987; Downhower et aI., 1987; Hubbell and Johnson, 1987; Crespi and Bookstein, 1989; Schluter, 1988; Mitchell-aids and Shaw, 1987; Wade, 1987; Wade and Kalisz, 1989) have addressed the statistical advantages and/or disadvantages of the various quantitative approaches to measuring selection. This emphasis on the theoretical methods gives the impression that the quality of an analysis of selection is equated with sophistication in multivariate statistics. It has overshadowed the fact that the analysis of selection is essentially an ecological and not a statistical question. Our purpose is to reexamine recent methods
for measuring selection (Lande and Arnold, 1983; Arnold and Wade, 1984a, 1984b; Mitchell-aIds and Shaw, 1987; Schluter, 1988) particularly with regard to identifying the conditions necessary and sufficient for inferring the agent of selection acting on a trait.
