Statistical Analysis of Character Correlations in Brewer's Blackbirds

Abstract

Power, D. M. (Dept. of Ornithology, Royal Ontario Museum and Dept. of Zoology, University of Toronto, Toronto, Ontario, Canada) 1971. Statistical analysis of character correlations in Brewer's Blackbirds. Syst. Zool., 20:186-203.-Character correlations among 37 skin and skeletal characters within a statistically homogeneous sample of Brewer's blackbirds (Euphagus cyanocephalus) were analyzed by component analysis and average linkage cluster analysis in order to isolate structural and/or functional character sets. Associations were assessed using product-moment correlation coefficients, partial correlations that corrected for a general size factor, and a subset of special partial correlations that corrected for a group factor defining high intercorrelations among limb bone length measurements. With ordinary correlations separate associations were identified for limb bone length, flight feather length, bill length, sternum size, bill width, and limb bone thickness. Some characters were predominantly independent. Partial correlations that corrected for covariation due to the group limb bone length factor indicated special associations among wing bones. An overall size factor is defined as that component which accounts for most of the covariation among all characters. Partial correlations correcting for this factor were analyzed and several size-independent associations emerged. The limb bone length assemblage is in part size-dependent. Size-independent associations include wing bone length, flight feather length, bill length, bill width, and limb bone thickness assemblages, and possibly a partial leg bone length set. Association of three sternal characters appears size-dependent, however a special size-independent association between sternum length and keel height was found. In addition, low-level, size-independent associations are found for pelvis width and bill height characters, as well as between the scapula and coracoid in the pectoral girdle. Results suggest the existence of selective factors that affect the development of functional complex as a whole. Comments are given concerning redundant characters in systematic studies and a temporary solution is offered concerning the problem of adequate sample size in studies of character correlations. [Character correlations; Brewer's blackbirds.] Character correlations within populations are the result of individual ontogenetic processes and natural selection. Perhaps the greatest contributions to the quantitative assessment of character relationships, particularly in terms of size and shape, have been in the study of allometry. Gould (1966), in his review of concepts and methods in t-his field, and in his own work, has shown the contributions that can be made here in understanding ontogeny and phylogeny. In the last several years relatively few papers have dealt in depth with covariation in non-growing systems, but the single underlying generality that can be made seems to be that characters related in structure or function tend to be more highly correlated than are others (e.g., Berg, 1960; Olson and Miller, 1958). As Van Valen (1965) has pointed out, it seems necessary to postulate selection for developmental patterns that affect functional complexes as a whole. For populations of adult, non-domesticated birds, very few attempts have been made to investigate correlations or to isolate structural or functional subsets for more *than just a few morphological characters. Nor have there been attempts to investigate covariation independent of sizeoTr general factors at the population level. The present study is an analysis of correlations among a number of morphological characters from four subsamples (two sexes and two localities, treated separately and in combination) of adult Brewer's blackbirds (Euphagus cyanocephalus; Icteridae), a common species widespread in western and central