Animal movement paths are the result of a confluence of pattern-generating forces. Numerous hypotheses relate animal relocation patterns to one factor, but interactions among variables have seldom been examined, nor have a cosmopolitan host of such models been allowed to compete on common ground. I tested a set of alternate hypotheses regarding displacement probability and three key variables---distance, resource density, and memory---by fitting 11 models to relocation data collected from five elk ('Cervus elaphus') recently reintroduced to south-central Ontario. Models featuring memory in combination with distance and/or resource density had higher likelihoods than traditional models of distance-based displacement or resource selection. Analysis of the sensitivity of the best-fitting models to changing input values suggested that memory affects the architecture of animal movement paths primarily over short distances, and thus may produce different resource-selection patterns at different spatio-temporal extents.