3. Results
3.1 The change of skull size from 1953 to 2004
The skull height and width, but not the length, were larger in adult males than in females in S. araneus , as confirmed in the ANCOVA models with sex and year (F 1,110=7.20,P =0.0084; F 1,110=11.25, P =0.0011;F 1,110=1.00, P =0.320; the interaction between the sex and year added to the models was nonsignificant:F 1,109=0.14, P =0.711;F 1,109=0.36, P =0.552;F 1,109=0.37, P =0.543). Nonetheless, as we were mainly interested in relative change and because otherwise sample sizes would have been too small, we pooled the sexes for further analyses.
The skull height and width of juvenile and adult S. araneus , as well as the skull length of juveniles, decreased significantly over the sampling period. In contrast, the relationship between skull length and time was not significant in adults (linear regression models; Table 1; regression equations are in supporting information, Table S1). Characteristically, all these relationships (except skull height in adults) revealed a common pattern when smoothed with LOESS: little change until the early 1970s and decreases thereafter (Figure 4; scatterplots in supporting information, Figure S1). The decrease in skull width of adults from 1970 to 2004 was even sharper than that in the entire time period (F 1,70=11.46,P =0.0012). The decrease in the skull length of adults was significant when restricted to this time period (F 1,70=8.79, P=0.0041 andF 1,69=4.99, P=0.0288 when one outlier in 2004 was removed).