A Euclidean distance (ED) method of wildlife habitat analysis has recently been proposed as an alternative to compositional analysis (CA). We performed simulation analyses to compare performance of ED to that of CA, using data sets with known parameters, where habitat patch size and shape remained the same. We observed extensive misclassification rates for ED but not for CA. For each of the 16 utilization permutations we modeled, of 3 avoided and 2 preferred habitats, results for CA and ED differed. Differences depended on the particular utilization permutations (i.e., juxtaposition of habitats) and did not seem to occur in any clear or predictable pattern. We recommend that ED not be used for future analyses of habitat use or resource selection until or unless these analytical problems can be rectified.