Female preference exerts selection pressure on male traits. This results in a complex interplay of phenotype and genotype. This complexity was examined by an experiment in which females were either mated to males they preferred (S lines) or to males chosen at random from the population (R lines). If genetic correlations between preference and preferred traits (14 song components) resulted from linkage disequilibrium, the correlation would diminish more in the S than the R line. This was not observed, suggesting that the correlation was generated by linkage disequilibrium and pleiotropy. Univariate preference was greatest for time spent calling (CALL), volume (VOL) and chirp rate (CHIRP) but the major contributors in the multivariate function were CALL and CHIRP, the univariate influence of VOL arising from correlations to these traits. Estimation of β for CALL and VOL resulting from female preference showed that both were under strong direct selection. However, CALL did not change over the course of the experiment whereas VOL, CHIRP and other song components did. Simulation of the experiment using the G matrix estimated from generation 11 showed that the lack of change in CALL resulted from indirect genetic effects negating direct effects. CHIRP and VOL were predicted to change as observed. Changes in song components were largely due to indirect effects. This experiment showed that female preference may exert strong selection on traits but how they respond to such selection will depend greatly upon the details of the G matrix.