Multivariate phenotypic selection on a complex sexual signal

Animal signals are complex, comprising multiple components that receivers may use to inform their decisions. Components may carry information of differing value to receivers, and selection on one component could modulate or reverse selection on another, necessitating a multivariate approach to estimating selection gradients. However, surprisingly few empirical studies have estimated the strength of phenotypic selection on complex signals with appropriate design and adequate power to detect nonlinear selection. We used phonotaxis assays to measure sexual selection on the advertisement signal of Cope's gray tree frog, Hyla chrysoscelis. Female preferences were assessed for five signal components using single- and two-stimulus behavioral assays. Linear, quadratic, and correlational selection gradients were estimated from the single-stimulus data. Significant directional selection is acting on call duration, call rate, pulse rate, and relative amplitude; stabilizing selection is acting on call duration and call rate. Under the two-stimulus paradigm, conclusions were qualitatively different, revealing nonlinear selection on all components except call duration. For individual subjects, the outcomes of single- and two-stimulus trials were frequently discordant, suggesting that the choice of testing paradigm may affect conclusions drawn from experiments.