Discriminant analysis of Lepidopteran prey characteristics and their effects on the outcome of bird-feeding trials

Discriminant analysis was used to analyse the results of 348 bird-feeding trials conducted from 1982 to 1985 in Leverett, Massachusetts. Four size classes, seven appearance categories, and five larval host types, based on 163 species of moths and butterflies used as prey in two or more trials, were selected as predictor variables to discriminate between prey taken and not taken by birds. Discriminant analysis of individual feeding trials correctly classified 97.5 percent of prey taken or not-taken and ranked the predictor variables according to their relative importance in determining prey acceptability. Characteristics most acceptable to birds were: (1) large size, (2) bark-like appearance, (3) warning colouration, (4) woody generalist, (5) dead-leaf-like appearance, (6) woody specialist, and (7) medium size. Characteristics least acceptable to birds were: (1) small size, (2) mimetic appearance, (3) butterfly appearance, (4) herbaceous specialist food type, (5) black-and-white appearance, (6) extra large size, and (7) overall generalist feeder. A summary of the analyses includes a discriminant function based on lepidopteran characteristics that can be used to predict the prey acceptability of species not used in this study. A multiple regression analysis of prey taken revealed that size alone and larval host type combined with other prey characteristics were the most important variables in determining the selection of prey regardless of their abundance in the trials.