Assessing Gender in Free-ranging Desert Tortoises (Gopherus agassizii) using External Morphology

James L. Boone
SAIC, 1180 Town Center Drive, Las Vegas, NV 89134

ABSTRACT

When using external morphology to determine the gender of tortoises, researchers have long relied on the work of Woodbury and Hardy (1948; Ecological Monographs, 18:145-200). However, these results pertain primarily to large tortoises, and there has always been a need for a quantitative method to determine gender of smaller tortoises.

We applied multivariate statistical techniques to 22 measurements of the external morphology of 105 tortoises to determine whether these measurements could be used to identify male and female tortoises. We determined gender independently of external morphology (serum testosterone levels, x-rays, behavior) for 35 females and 13 males. Of the remaining individuals, those with carapace length (CL) greater than 200 mm (5 females, 33 males) were assigned to a gender category based on morphology that was similar to individuals of known gender. Nineteen individuals (CL < 200 mm) of unknown gender were not assigned to gender a priori.

Three discriminant analyses were performed: only animals of known gender, all larger animals (CL > 192), and all smaller animals (CL < 220). Analysis of the data set containing only animals of known gender correctly classified all but two males (CL = 179, 206) using 11 characters. In the trivial case of discriminating larger animals, all classified correctly. Although it is generally considered easy to classify larger animals based on external morphology, the model required 9 characters to correctly classify all of these individuals. When only smaller animals were considered, all individuals of known gender classified correctly using 5 characters, and when the 19 small individuals of unknown gender were classified with this discriminant function, all animals fell cleanly into two groups.

Only three characters were important for separating males and females in all analyses. In each, a deep plastron concavity (always the largest coefficient) and wide front foot (always a large coefficient) were associated with males. Gular length was important in each case, but a long gular projection was only associated with males for the data sets containing all animals and large animals. For the smaller tortoises, a long gular was associated with females.

Despite the apparent success of these models, when the discriminant scores were plotted against carapace length (a proxy for age), it became clear that males with a CL of 100-190 were probably classified correctly, but all animals smaller than this were classified as females. While it is possible that all of these animals were female, it is more likely that animals with a CL less than 100 included males that had not started to acquire secondary sexual characters. Therefore, it appears possible to use multivariate statistics to determine the gender of tortoises with carapace lengths as short as about 100 mm.