Evolution in the House Sparrow. IV. Replicate Studies in Phenetic Covariation

Abstract

Johnston, R. F. (Museum of Natural History and Dept. of Systematics and Ecology, University of Kansas, Lawrence, Kansas 66044) 1973. Evolution in the House Sparrow. IV. Replicate Studies in Phenetic Covariation. Syst. Zool. 22:219-226.-Covariant character sets extracted from house sparrow specimens from all sectors of North America by principal component analysis were obtained twice again from specimens taken on an eastwest transect of North America and from specimens generally from western Europe. Uniformity in character covariation is thus evident in ancestral-descendent populations separated by more than 100 generations and in contemporary populations separated by distinct climatic environments. Such covariation is assumed to be a consequence of natural selection, and to represent the genetic ground more immediately than do single characters. [Passer domesticus; phenetics; covariation; evolution; principal component analysis.] Concern by systematists over adequacy of character variables used in evolutionary or taxonomic studies is always justified. In the past decade we have seen debate not merely over what makes up an adequate character and what it should formally be called, but also on the format and significance of intercharacter correlation and the meaning this may have to the biology of organisms. Intercharacter correlation has been called character or phenetic covariation, and there has been some effort to link this concept with the earlier one of genic coadaptation (Sokal and Rinkel, 1963; Johnston and Selander, 1971; Gould and Johnston, 1972) and thus to identify the selective interface of the organism with its environment. A number of studies have been concerned with character covariation at least in part because techniques of factor analysis and principal component analysis were used in the assessments of character variables (Thomas, 1968; Gould, 1969; Atchley, 1971), and such analyses reduce character space and in so doing identify covariant character sets. It is worth noting that univariate analyses do not provide usable character sets, even though putative covariation can be inferred from serially repeated coincidences of clinal variation, so that the long history of character analysis has little to say about covariation. Yet, the significance of phenetic covariation in representing the consequences of natural selection is such that any study in evolutionary morphology that lacks assessment of covariation is evidently incomplete. One reason for past indirection in handling covariation as an evolutionary phenomenon is the knowledge that character variation can be caused by variation in either genetic or environmental factors, or by their interactions. A field naturalist rarely knows which set of factors is paramount for a set of variables, and he ordinarily does not formally study inheritance of characters in his organisms. He usually infers that character variables are modified by natural selection, a classical inference based on analogy to studies in which the genetic and environmental components of characters have been identified. In some instances the inference is unsatisfying, and this is true in considerations of control of geographic variation in the house sparrow Passer domesticus in North America. Subsequent to 1852 house sparrows have developed complex patterns of size and color variation in North America, and there is no difficulty in assigning inferential adaptive significance to such variation because it is ordered geographically and on independent environmental variables (Selander and Johnston, 1967; Johnston, 1966; Johnston and Selander, 1964, 1971, 1973). But, precisely because of the rapid ad-