Alytes logo

Frog morphometrics: a cautionary tale

Article published in volume 18 fascicle 3, of 01/2001, pages 153 - 177



Scant attention has been paid to measurement error in frog morphometric studies. We study both interobserver effects of measurement on the same specimens of Vanzolinius discodactylus (Anura, Leptodactylidae) and intraobserver effect of repeated measurements on a single V. discodactylus specimen. Interobserver measurements differ statistically and result in different biological interpretations in some cases. Evidence is provided that log transformation of raw data is often unnecessary. Allometric transformation of measurement variables to remove size effect requires parallel regression slops of variable against size. This requirement is not met with the V. discodactylus data, nor is it likely to be met when several variables are used in a morphometric study. We recommend: assume measurement differences between sexes in frogs and analyze data separately by sex; consider and select the most appropriate statistical model options for data analyses; avoid pseudoprecise measurements; do not rush to logarithmic transformation; remeasure at least one individual frog 20 times to provide an assessment of measurement error in data interpretation; be conservative in drawing biological inferences from morphometric analyses, basing interpretations and conclusions only on very robust effect size estimates and differences.