A critical review of methods of studying fish feeding based on analysis of stomach contents: application to elasmobranch fishes

Using real data sets of elasmobranch fishes as examples, this paper presents a critical review of selected methods and statistical approaches used in fish feeding studies and makes recommendations on the application of such methodology. The percent index of relative importance is proposed as a standardized measure in dietary analyses, and a three-dimensional graphical representation of the diet is introduced. Multiway contingency table (log-linear) analysis is recommended to test for dietary variations. Caution is advised when using rank correlation to study dietary overlap and parametric tests when stomach content data do not satisfy parametric assumptions. Sampling gear type, experimental design, and statistical tests can affect results on diel feeding chronology, and stomach content weights do not suffice to interpret diel feeding chronology. On the basis of sampling requirements and model assumptions, the Diana and Olson-Mullen methods appear to be the most appropriate approaches for estimating daily ration in sharks. Use of resampling techniques is highly desirable because they provide a measure of the error in daily ration estimates. Using several criteria to evaluate the best-fitting model of gastric evacuation in fishes is also strongly advocated. Overall, increased consolidation of methods and analyses is recommended to facilitate comparative studies.