tested to determine whether to use a conventional ANOVA (variances 

 assumed equal) or a Brown-Forsythe ANOVA (variances not assumed to 

 be equal) . The ANOVA indicates whether there are significant difference 

 between treatments without indicating which of the four treatments 

 differ from each other. Hence, we used Tukey ' s studentized range 

 test to make pairwise comparisons of each treatment to the others. 



Statistical analyses were limited by small sample size. In 

 some cases there were insufficient degrees of freedom for the 

 test, or there were entire rows or columns in the contingency 

 tables with no values so that expected frequencies could not be 

 computed. We have presented statistical analyses for the four 

 species with the greatest number of individuals and, where 

 appropriate, for all species combined. 



17 



