The ﬁrst and the second PCA axes explained, respectively, c. 47% and c. 20% of the variance, for both males and females (Table 1). The ﬁrst axis had high factor loadings for limb and digit variables (0.76 to 0.89). The second axis had high loadings for head length and head width (0.54 to 0.84). Hence, the variation observed for head shape and the size of the extremities was largely independent. Mean population factor scores for the ﬁrst axis increased from south to north, reﬂecting an increase in relative size of limbs and digits (Figs. 3 and 6). Mean factor scores for the second axis did not show any readily interpretable geographic pattern (Fig. 3).
Colour pattern type 2 was the most common one in all examined populations, except for populations 10 and 11 in which type 3 was most frequent (Fig. 1).
Type 3 was present in all populations south of the Douro River. Other pattern types were present in some populations and not in others. Types other than type 2 were absent north of the Douro, with the exception of type 4 that was present, albeit at low frequency, in populations 12, 13 and 20.
Testing causal hypotheses
Scores on the ﬁrst PCA axis were signiﬁcantly associated with geographic distance, group membership (Table 2) and latitudinal distance (Fig. 5). Scores on the second axis were not associated with either of the independent variables (Table 2).
Fig. 5. Regression of mean factor scores of the fi rst Principal Component axis of 18 populations of female Chioglossa lusitanica against distance along a south-north axis.
Classiﬁcation of individuals (both females and males) in group 2, following DA, increased gradually from south to north (Fig. 4). In females, head length and head width were associated with geographic distance and group membership, respectively. Size of the extremities was associated with geographic distance, with the exception of FLL# in males. Digit length and HLL# in males were also associated with group membership. Colour pattern was associated to neither geographic distance or group membership. Morphometric variability was not associated with any of the formulated independent variables (Table 3). Colour pattern variability was associated with hybridity and not with geographic distance or heterozygosity. Additionally, a trend was observed for colour pattern variability to decrease from south to north in group 2 populations (Fig.1).