FIGURE 3. Phenotypic variance partitioning and niche breadth for each dispersal treatment.

a, Bars show the mean quantity of phenotypic variance (s.e.m; n = 3) attributable to effects of G, E and G  E. The absolute quantities of G, E and G  E do not differ between dispersal treatments (one-way ANOVA, n = 12, P > 0.05). b, Diamond symbols show functional diversity measured as a percentage of total phenotypic variance attributable to inconsistency. Functional diversity peaks at an intermediate dispersal level as judged by quadratic regression (quadratic test: F_2,9 = 9.84, P = 0.05, n = 12; quadratic parameter t = -3.16, P = 0.012) and multiple means comparison (Tukey's test: q = 3.2, = 0.05, grouped means are labelled 'a' for 0%, 1% and 100%; and 'b' for 1%, 10% and 100%). c, Diamond symbols show the average genotypic niche breadth for each dispersal treatment (s.e.m; n = 48), calculated as the number of substrates giving positive growth scores for each genotype. Average niche breadth increases exponentially with dispersal rate (F_1,3 = 420, P = 0.0024, r² = 0.99) and there are no differences in mean niche breadth among replicate lines nested within treatments (F_3,8 = 1.43, P = 0.19).