Incomplete sampling of geographic ranges weakens or reverses the positive relationship between an animal species' geographic range size and its body size

Aim: To study the effects of the sampling regime on body size/range size correlations among species. Background: Body size/range size correlations tend to be positive if geographic sampling includes all or most of the geographic area of each species in a study. However, correlations tend to be negative or non-significant if geographic sampling is only partial. Methods: Using computer bootstrapping, we randomly sampled latitudinal ranges of 1317 species of New World mammals. We sampled along lines of fixed longitude. We studied the effects of three variables: (1) the comprehensiveness of the sample as a proportion of the entire latitudinal range; (2) the number of isolated populations within a species range; (3) the correlation of body size and range size in the computerized data set from which the samples come. The latter is termed 'weak' when we used the empirical body size and range size data. It was termed 'strong' when we replaced the real body sizes with ones assigned to produce a correlation of one with range size. Results: Studies using incomplete ranges yield unpredictable results that might not even get the sign of the relationship correct. The unpredictability arises because incomplete sampling tends to understate the range sizes of species with the largest ranges and miss the smallest ranges entirely. Those unusual range sizes determine the true body size/range size correlation. The unpredictability worsens as more and more locations within a range lack a population, thus increasing the probability that species will be missed by an incomplete sample. The degree of unpredictability caused by incomplete sampling and the number of isolated populations did not differ between the 'weak' and 'strong' data sets.