Treatments consisted of triplicates in trial 1 and four replicates in trials 2 and 3. Using a tank as the experimental unit, analysis of variance (anova) was used to compare average fish length between treatments at the end of each experiment. Data were checked for normality of residuals (Shapiro–Wilk's W-test) and equality of variance (Levene's test). Data did not need to be transformed before analysis. Fisher's LSD multiple-range test was used to determine differences between treatments at a 5% type I error level (P