Females preferred songs produced by males that solved the novel foraging task
The second Bayesian model (Fig. 2 ) showed that females prefered Solver song and gave information about strength of preference. Population preference for Solver song was 0.594 (95% credible intervals: 0.581, 0.628), and population preference for Non-Solver song was 0.406 (95% credible interval: 0.372, 0.419). ESS was 1674.475. In a Bayesian preference model, confidence that one group is preferred over another increases as overlap in the confidence intervals of both groups declines. In our model there was no overlap between the 95% credible intervals for Solver preference and Non-Solver preference, indicating a substantially higher preference for the song of Solver males, with a 5% or lower probability that Non-Solver song was preferred equally or more than Solver song (Fordyce et al. 2011). For comparison, 95% credible intervals for group conspecific song preference were 0.687, 0.748; and 95% credible intervals for group heterospecific song preference were 0.252, 0.313 (see Fig. 1 ). There was a greater difference between conspecific and heterospecific confidence intervals, but preference for conspecific song was only approximately 11 percentage points higher (0.705) than preference for Solver song (0.594).
These patterns remained consistent when the 6thstimulus set was dropped (Solver preference: 0.581, 95% credible interval: 0.567, 0.617; Non-Solver preference: 0.419, 95% credible interval: 0.383, 0.433).
In Solver/Non-Solver trials, the average number of hops per trial was 704.84 (s.d. = 384.59). Full results from Solver/Non-Solver trials can be found in the Supplementary Materials.