MASS/LENGTH RESIDUALS: MEASURES OF BODY CONDITION OR GENERATORS OF SPURIOUS RESULTS?

In studies of animal ecology, it is fashionable to use the residuals from an ordinary least squares (OLS) linear regression of body mass against a linear measure of size (the body size indicator, BSI) as an index of body condition. These residual indices are used to study the relationship between condition and reproductive investment, survivorship, habitat use, and other parameters. I identify a series of key assumptions underlying the use of this method, each of which is likely to be violated in some or all studies. These assumptions are: (1) that the functional relationship between mass and BSI is linear; (2) that condition is independent of BSI length; (3) that BSI length accurately indicates structural size; (4) that there is no correlation between the size of BSI relative to other structural components (i.e., shape) and the parameter against which the residuals are analyzed; (5) that BSI length is strictly independent of mass; and (6) that BSI length is not subjected to error. Violations of these assumptions place the results of some studies in question and explain the poor relationship observed between OLS residuals and more direct measures of condition. I use avian morphometric data to illustrate how OLS methods can easily lead to Type I and Type II errors through violations of assumptions (5) and (6). Significant relationships reported between OLS residual indices and parameters correlating with body size (e.g., size of sexual ornaments or egg size) are at particular risk of being spurious when the correlation coefficient between mass and BSI is low. Residual indices of condition are often likely to be more reliable when calculated with alternative methods such as nonparametric or model II regression. However, whatever the method used to produce them, residual indices are not suitable for studying the heritability of condition.