We tested the prey-base hypothesis to determine whether selection of red squirrel (Tamiasciurus hudsonicus) midden sites (cone caches) by American martens (Martes americana) for resting and denning could be attributed to greater abundance of small-mammal prey. Five years of livetrapping at 180 sampling stations in 2 drainages showed that small mammals, particularly red-backed voles (Clethrionomys gapperi) and shrews (Sorex spp.), were more abundant at midden sites than at non-midden sites. However, logistic regression indicated that middens occurred in spruce– fir (Picea engelmannii - Abies lasiocarpa) stands, being correlated with decreasing lodgepole pine (Pinus contorta) basal area, decreasing distance to water, and increasing canopy cover. Since middens were not randomly distributed, we constructed multiple linear regression models to determine the variability in small-mammal abundance and species richness attributable to structural and landscape variables. Regression models indicated that abundance of small mammals, red-backed voles, and uncommon small mammals could be predicted from structural and landscape variables, but midden presence did not significantly improve these models. Midden presence was a significant but weak predictor of small-mammal species richness. Our data do not support the prey-base hypothesis for explaining martens’ selection of resting and denning sites near red squirrel middens at the scales we tested.