Univariate metrics are not adequate to measure avian body size

-In the literature on avian contests and energetics, a single external measure is frequently used to represent overall body size. In an attempt to assess the quality of information available in such external measurements of birds, we measured internal and external elements from museum skeletal plus flat-skin specimens for both sexes of three passerine species. We estimated the "true" overall size of each individual as a factor score computed from the first principal component extracted from a large matrix of skeletal measurements. Bivariate correlations and stepwise regressions indicate that mass or tarsus length, or a principal component factor which combines mass and tarsus length, is the best predictor of overall body size as estimated from bone measurements. Multiple regressions, however, suggest that several external measurements combined often explain only 40-60% of the total variance in overall body size. We suggest that fieldworkers be cautious in their use of single external metrics to represent overall size in small birds. When a single metric for the body size of small passerines is required, fieldworkers should prefer tarsus length or mass to represent overall size. Received 6 October 1988, accepted 18 July 1989. BEHAVIORAL and physiological studies that seek correlations between avian body size and other variables are faced with the problem of how to assess body size. Often the variable of interest is body mass, for example, because of the importance of mass in energetics (metabolic rate, daily caloric demands) or dominance (ability to win escalated contests). Measuring body mass accurately is difficult because mass fluctuates seasonally, daily, or even hourly, depending upon variables such as time since feeding, weather, and activity. Investigators who work with birds in captivity can reduce the effects of these confounding factors by measuring mass at a standardized time of day (often early morning) and by taking repeated measures. Field-workers seldom have these options. Our goals in this study were to determine which (or which combination) of the external measures commonly employed by field biologists best predicts actual body size, and to determine the efficiency of this predictor. We estimated actual body size from a principal component analysis (PCA) of a large number of skeletal measures. The first factor extracted from a PCA of various size measurements has been interpreted by many researchers as an index of overall body size (e.g. Robins and Schnell 1971, Rohwer 1972, Niles 1973, Zink 1982, Schluter 1984, Rising 1987). There are also three theoretical grounds on which to defend our assumption that this internal metric is the best possible measure of "true" body size. First, the internal metric is composed of many different characters, and it should therefore be less prone to variance caused by developmental abnormalities or measurement errors in single characters. Second, the repeatability of bone measurements is greater than the repeatability of most external measures because bone measurements can be made more precisely, and because many external measures vary through time. For example, wing chord and tail length vary both with the individual's age and the degree of feather wear, and beak length varies through time due to rhamphotheca wear. Third, and most importantly, a metric composed of size measurements from many bones summarizes the amount of total attachment surface available for muscle and connective tissue as well as the amount of support structure for internal organs. This metric should be an accurate measure of "structural size" and a biologically sensible predictor of "average massiveness" within populations.