Sexual dimorphism and geographic variation in the morphology of a small southern African tortoise Psammobates oculifer

We studied morphological traits of Psammobates oculifer over its range to evaluate patterns in sexual size dimorphism (SSD) and geographic variation. Females were larger than males for 40 of the 44 characters measured and the growth trajectories of female parameters, scaled to body size (ANCOVA and MANCOVA on carapace length), most often followed juvenile patterns. For males, either the slopes or elevations of morphometric parameters were lower than in females. These divergent growth trajectories resulted in shape differences with female shells being higher and wider than the shells of males. Males matured at a smaller age and size, and had wider hind feet and larger shell openings relative to females; the latter being due to a shorter and narrower plastron, a shorter bridge length, and a bigger anal gap and cranial space. These male characteristics possibly enhance mobility for mate searching and combat, as well as courtship and mating behaviour. Small sample sizes for females in two of the three regions restricted geographic evaluations to males, for which shell shape, front foot width and hind leg length of the eastern group differed from the others. We interpreted these variations as the consequence of differences in the substratum and vegetation openness over the range of the species.

while studying the species' ecology from March 2006 to April 2007 at Benfontein farm (28°53' S; 24°51' E), near Kimberley, South Africa. We counted the number of carapacial, plastral, axillary and inguinal scutes, noted the position of the latter two scutes relative to adjacent plastral and marginal scutes, and recorded the number and strength (none, weak, medium or strong) of serrated marginal scutes. We also recorded the presence or absence of a plastral concavity, rated the concavity as weak or strong, and noted its position relative to abdominal scutes. Furthermore, we noted the presence and intensity (weak, moderate or strong) of a marginal groove (indentation between pleural and marginal scutes) and bridge ridge (edge alongside of shell) and the associated marginal scute numbers. We could not collect a full set of data for each individual because some museum specimens were damaged whereas time constraints during peak activity periods did not allow detailed measurements of all live individuals. In order to evaluate shell colouration, we photographed the carapace and plastron of each tortoise with a Panasonic DMC-FZ30 digital camera. In this species, darkbrown to black pigmented rays overlay a yellow to brown shell base colour and we evaluated base colour gradations and estimated the amount of dark pigmentation in 10% increments for both the carapace and plastron.
Because there were individual variations in the number and symmetry of marginal scutes, we summarised the frequencies as number occurring per individual on each side. We divided marginal scute serration patterns into anterior (1-5) and posterior (6-12) and summarised frequencies for each side. To test for differences among cohorts in frequencies of marginal scute numbers, serration and percentages of carapacial colouration, we used  2 tests or Fisher's exact tests (2x2) where frequencies were low, with a Yates' correction if degrees of freedom equalled one and with mean expected frequencies always equalling six or more (Zar, 2001). Where contingency tables were 2x3, 2x4 or 3x3 and frequencies were low, we used the Freeman-Halton extension to the Fisher's exact test (Freeman and Halton, 1951). We used repeated measures ANOVA (RMA, F statistic) or Friedman's repeated measures ANOVA (FRMA,  2 tests) where data were non-parametric, followed by Student-Newman-Keuls post hoc tests to rank the plastral, pleural, and vertebral scutes formulae for males, females and juveniles. We used paired t tests or, when data were non-parametric, Wilcoxon Signed Rank tests (T statistic), to evaluate differences between carapacial and plastral pigmentation within each cohort and for all tortoises combined. All remaining morphological characteristics (axillary and inguinal scutes and adjoining scutes, plastral concavity occurrences and bridge characteristics) were summarised as percentages among groups.
VassarStats were used to calculate the Freeman-Halton (1951) extension to Fisher's exact test.
We used SigmaStat 2.03 (SPSS Inc., Chicago, U.S.) and Microsoft Excel for all other statistical analyses.

Scute dimorphism
Pleural, vertebral, and plastral scute lengths were compared using formulae based on proportional length (see below). Sexual dimorphism (SDI) for these scutes, however, ranged between 0.117 and 0.277 for the first five vertebral scutes, between 0.171 and 0.283 for the first four pleural scutes, and between 0.111 and 0.331 for plastral scutes, excluding the femoral scute (-0.001; table S1), for which ANOVA results showed no difference between females and males.

Scute counts and scute length formulae
All tortoises had a single nuchal and supracaudal scute with the exception of one female with no nuchal scute and one male with two nuchal scutes (table S2). Most individuals had five vertebral scutes and four pleural scutes on both sides (table S2). Psammobates oculifer usually had either 10 or 11 marginal scutes on a side (table S2), and the pattern was usually symmetrical (table S3). There was no difference among cohorts in the number of individuals with 10 or 11 marginal scutes on each side, or those with asymmetrical versus symmetrical marginals ( 2 tests, P ≥ 0.61; table S3).
All P. oculifer except one female had one axillary scute on each side of the plastron.
The axillary scutes of all P. oculifer were in contact with the humeral and pectoral scutes, and were most often in contact with marginal scutes three and four (table S4). All tortoises had one inguinal scute on each side of the plastron, which was always in contact with the abdominal and femoral scutes. Contact of inguinal with marginal scutes was more variable, but was most often with marginal scutes seven and eight (table S4). Variation in the position of the inguinal scutes did not differ between males and females (P = 0.95), but inguinal position was more variable in adults than it was in juveniles (Fisher's exact tests, P < 0.006; table S4).

Other morphological characteristics
Most P. oculifer had three serrated marginal scutes on each anterior side (table S5) but in juveniles the number of anterior serrations exceeded three more often than in males and females ( 2 1 ≥ 5.68, P ≤ 0.017) with no difference between males and females (P = 0.097; table S5). Males and females usually had three posterior serrations on each side, whereas juveniles most often had four or more posterior serrations. The number of posterior serrations of juveniles exceeded those of males and females ( 2 1 ≥ 5.4, P ≤ 0.02) but serration in males and females did not differ (P = 0.29; table S5). Serrations in all P. oculifer were rated as strong except for four adults where the strength of serrations was rated as medium.
All P. oculifer had a weak bridge ridge, which most often incorporated marginal scutes three to seven or three to eight (table S6). The species lacked a marginal groove.
Females did not have a plastral concavity whereas most males (85%) had. Male concavities occurred on the midline along the length of the abdominals up to the femoral scute (56%) or were limited to the midline at the junction of the abdominal and femoral scutes (44%).
The carapacial background colour of P. oculifer was most often orange-yellow or orange-brown (table S7). The frequency of occurrence of colour categories differed among cohorts ( 2 3 ≥ 11.1, P ≤ 0.011) with juveniles having the palest colour and males tending to have a paler background colour than females (table S7). All tortoises except one female had dark pigment on the carapace (table S8). Juveniles had more dark carapacial pigment than adults had, and male carapaces were darker than those of females were ( 2 3 ≥ 12.31, P ≤ 0.0064). The plastrons of all tortoises had dark pigment (table S8) but plastral pigmentation did not differ among cohorts (P = 0.28). Carapacial pigmentation was marginally greater than plastral pigmentation for females (T 117 = 1135, P = 0.04) and substantially darker for males and juveniles (T ≥ 528, n ≥ 32, P < 0.00001). Supplementary Table S1. Morphometric measurements (see table 1 for descriptions of abbreviations) of Psammobates oculifer cohorts presented as mean and standard deviation with sample size in brackets and minimum (Min) and maximum (Max) values underneath. Measurements were taken from live animals (Benfontein) and wet or dry museum specimens (Iziko and Ditsong museums).
Measurements for SHM and SV are from museum specimens only. Mass is in g, shell volume in cm 3 , with all other measurements in mm. The asterisk indicates additional measurements in cases of abnormal scute numbers. We calculated sexual dimorphism indices (SDIs) as the mean of the larger sex (females) divided by the mean of the smaller sex (males) and subtracted one so that the result was positive when females were larger and negative when males were larger (Lovich and Gibbons 1992