Determining the importance of prey taxa in the diets of predacious species is a frequent objective in fisheries research. Various indices of prey importance are in common use, and all give different results because of their emphasis on different aspects of fish diets. We explored these differences by empirically comparing four well-known indices-percent weight (%W), percent occurrence (%O), percent number (%N), and percent index of relative importance (%IRI)-as well as a modified %IRI (%MIRI), as applied to an extensive data set on the diets of six fish species in Spirit Lake, Iowa. Correlations among all indices were positive but were weaker among component indices (%W, %O, and %N) than between the two compound indices (%IRI and %MIRI); correlations among component indices were also weaker than correlations of compound with component indices. Correlation strength of %MIRI with the three component indices varied greatly (%N %O %W), whereas the correlation strength of %IRI with component indices was similar. Importance values based on %W, %MIRI, and %N depend more on prey size than those based on %IRI and %O. The %W and %MIRI emphasized the importance of large prey taxa, whereas %N emphasized small prey in diets; %IRI and %O were similarly unbiased with respect to prey size. The %O yielded substantially higher importance values than all other indices. Thus, for use as a general index of dietary importance, we believe %IRI provides the optimal balancing of frequency of occurrence, numerical abundance, and abundance by weight of taxa in fish diets.