3.4 Change in the magnitude of Dehnel’s phenomenon from 1955 to 1985
The magnitude of Dehnel’s phenomenon increased over the tested time period (Figure 6a). The skull height of August juvenile and subadult shrews decreased significantly over time (F 1,50=11.83, P =0.0012;F 1,68=19.52, P <0.0001, respectively) but did not change in July juvenile and adult shrews (F 1,149=1.34, P =0.249;F 1,72=0.00, P =0.985; Figure 6b). This is also illustrated in Figure 6a: the standard errors of juvenile and adult skull height overlapped considerably, while the subadult skull height decreased strongly over time (years). There was no significant change in the skull length of August juveniles, subadults or adults (F 1,50=0.00, P =0.978;F 1,68=0.05, P =0.829;F 1,72=0.07, P =0.942, respectively), but the change was significant in July juveniles (F 1,149=5.00, P =0.0269).
In a second approach to analyse the magnitude of Dehnel’s phenomenon, we compared the slopes of the regression lines of skull height over time (years) between juveniles vs. subadults and subadults vs. adults (Figure 6b, scatterplots in Figure S3, see also Figure 4a). A lack of interaction between the year and age (parallel lines) would indicate no change in Dehnel’s phenomenon. We explored this in an ANCOVA model of the skull height with the skull length, year, age, and year × age interaction (Table 4). Year and skull length were highly significant in the comparison of July juveniles with subadults, August juveniles with subadults, and subadults with adults (Table 4). The interaction between year and age was highly significant in two comparisons (Table 4): the decline in skull height from July juveniles to subadults was greater with time (Figures 6b and S3a), and the regrowth in skull height from subadults to adults increased over time (Figures 6b and S3c). In both cases, this was due to the very steep line in subadults. In contrast, the interaction was nonsignificant in the comparison of August juveniles with subadults (Table 4, Figures 6b and S3b). The lines of July and August differed significantly in slope (the interaction between month and year: F 1,198=12.60, P=0.0005 in ANCOVA with year: F 1,198=14.72, P=0.0002; month:F 1,198=12.55, P=0.0005; skull length:F 1,198=27.57, P<0.0001; Figure 6b). The skull width of juveniles (July, August) and adults did not change significantly over time (Table S2). The year × age interaction was not statistically significant in any of the three models for length-corrected skull width (Table S3). When these interactions were removed from the models, age was significant only in the comparison of subadults and adults, and the year was never significant (Table S3).
To discuss the impact of weather on Dehnel’s phenomenon, we additionally tested the regressions of all weather parameters on time (years) in July, August and January between 1955 and 1985. The only significant regression was that of the August mean daily temperature (positive relationship; F 1,29=6.36, P =0.0175). However, the mean daily temperature in January decreased significantly during the first 60% of the tested period (1955 to 1972;F 1,16=8.16, P =0.0114, andF 1,15=19.9, P =0.0005 when one clear outlier was removed; Figure S2c).