3.1 Variation in Circadian Period among and within Populations
Families with at least 5 replicates were included in analysis of variance of circadian period, with the largest number of replicates within a single family being 18. Both ANOVA and the Welch’s F-test found highly significant differences among and within populations in mean circadian period (Table 1). Among the 30 populations analyzed, mean circadian period varied between 21.91 to 24.92 h, or three hours (Fig. 2). Although most of the populations expressed a circadian period close to a 24 h, many populations had a period that is much shorter; 10 of the 30 populations had a mean period value that was shorter than 22.5 h, and only 2 such populations (LEW, LIB) exhibited spatial autocorrelation. Variation within the populations (among the families within each population) was similar (variance component for the family within population effect 0.687, or 21.3 % of total variation) compared to the variation among the 30 populations (variance component for the population effect 0.787, or 24.4 % of total variation), which defines the structure of the populations. Thus, though the range of family values within many of the populations was large, the populations were structured with slightly higher variation among the populations (Table 2).
Within populations, the difference between the family with the shortestvs . longest circadian period length (range) was 3 h on average. Two populations expressed exceptionally high ranges of >5 h (Happy Jack Trail, HJT, and Crow Creek, CRW, populations; Table 2). HJT and CRW exhibited spatial autocorrelation, such that one cannot exclude the possibility of a founder event and that HJT and CRW are one genetic population; nevertheless, under this hypothesis, at least one population exhibits >5 h range. The wide range of values for the HJT and CRW populations contrasted with another population, Sand Lake, which had a range of values among families of less than an hour (Fig. 3). Neither population mean value nor the number of replicates (both individuals and family lines) had a significant association with the within-population range, indicating the results for circadian range within populations are not a consequence of slight differences in replicate number and potential sampling bias. In sum, the range of variation among families within some populations (> 5 h) is greater than the mean difference among populations (3 h).