MORPHOMETRIC VARIABILITY IN CONTINENTAL AND ATLANTIC ISLAND POPULATIONS OF CHAFFINCHES (FRINGILLA COELEBS)

The magnitudes of phenotypic variances in peripherally isolated populations of common chaffinches (Fringilla coelebs) in the Azores, Madeira, and Canaries relative to their continental source populations in Iberia and Morocco have played a pivotal role in the formulation of competing hypotheses of phenotypic evolution in these isolates. Because Van Valen's niche variation hypothesis and Grant's model of island evolution were based on museum skins prone to measurement error and temporal variation, we re-examined the patterns of phenotypic variability using more precise skeletal measurements from freshly collected specimens. Levene's tests showed that univariate character variances were homogeneous in all island and continental populations, although there was a consistent trend for the magnitude of the variances to be lower for all characters in all Canary island populations. Multivariate Levene's tests, however, revealed significantly reduced total variances in the Hierro and Madeira populations compared to some Azores and continental populations. The Azores and continental populations did not differ in variability, and lower variances in the Canaries were not related to the presence or absence of the congeneric blue chaffinch (F. teydea), contrary to the predictions of Van Valen's niche variation hypothesis. Population variability was not inversely related to differentiation or isolation within the Azores or Canaries archipelagoes, opposite to the association reported by Grant. Our results also differ from both previous studies which reported much larger differences in population variabilities, and this likely reflects the use of heterogeneous samples of museum specimens, less precise external characters, and the use of tests sensitive to sample size. Differentiation among populations has been markedly greater in the Canary islands, implicating founder events and possibly historically stronger directional selection as determinants of this enhanced divergence relative to the Azores. These variance-reducing processes are unlikely to explain current lower levels of phenotypic variability because there has been sufficient time since colonization for replenishment of variability in polygenic characters. Average heterozygosities at putatively neutral allozyme loci are 1.6 times higher in the Azores compared to the Canaries, and support the view that effective population sizes are smaller in the latter archipelago. We argue that reduced variances in the Canary island populations represent lower equilibrium levels maintained by drift and mutation in populations with smaller long-term effective sizes, consistent with Lynch and Hill's neutral model of phenotypic evolution. Although episodes of selection in the past may have been interspersed with long periods of effective neutralism and drift, adaptationist hypotheses invoking a primary role for variance-reducing selection appear to be unwarranted.