Table 1. Results of the MANOVA and post-hoc tests for genital shape.

MANOVA
Predictor				Wilks’ λ		F 7,5	P
Sexual selection (S)				0.082		7.97	0.018
Natural selection (N)				0.118		5.32	0.042
Sexual × Natural selection (S × N)				0.047		14.61	0.005
Centroid size (C)				0.168		3.52	0.092
Univariate ANOVAs
	S		N			S × N
	F 1,11	P	F 1,11		P	F 1,11	P
Warp 1	0.06	0.809	0.36		0.561	4.42	0.064
Warp 2	5.66	0.036	1.52		0.244	18.71	0.001
Warp 3	1.01	0.337	0.01		0.986	6.57	0.026
Warp 4	1.03	0.332	1.36		0.267	0.01	0.981
Warp 5	0.001	0.986	0.001		0.978	1.01	0.337
Warp 6	1.14	0.310	1.55		0.238	0.185	0.675
Warp 7	6.65	0.026	5.08		0.046	0.21	0.658

Mean Relative Warp Scores for each population (i.e. 4 replicate populations/treatment ) were line means. There were significant effects of sexual selection, natural selection and their interaction on the multivariate combination of Relative Warp Scores 1–7. Univariate post-hoc tests showed which Warps generated the multivariate significance (bold P -values). Centroid size was included as a covariate because previous univariate analyses indicated that it was significantly associated with some Relative Warps. However, it was not significant in the MANOVA and hence we do not interpret it further. Removal of centroid size from these analyses increases the strength of selection associations considerably (e.g. the significance level of the sexual selection effect on Warp 7 become P = 0.009) and some marginally non-significant results become significant (SxN for Warp 1 and NS for Warp 4). Nevertheless, to be conservative, we retain centroid size in the model.