= 16) with crown height, loph/lamella count, lamellar frequency, enamel thickness, plicae frequency and plicae amplitude as variables. In each run, the partial correlation of one pair of variables was obtained, with the other variables kept constant . Crown width, as a measure of molar size, was included as a covariate in partial correlation analyses. We therefore used crown height rather than hypsodonty in these analyses as the latter is already corrected for crown width.
Correlations within species are likely to reflect developmental links, but too few specimens preserve all measurements for partial correlation analysis, so product-moment correlation coefficients were obtained for each pairwise combination of variables; this yielded a limited, meaningful set of significant correlations .