The U.S. Forest Service recently recommended managing forests to increase the abundance of prey for Northern Goshawks (Accipiter gentilis) in the southwestern United States. However, empirical data are needed to indicate whether goshawks select foraging habitat based on prey abundance, on the presence of forest structures that favor their hunting strategy, or both. During the 1993 and 1994 breeding seasons, we indexed prey abundance and measured forest structure at sites used by radio-tagged goshawks and at nearby paired sites that lacked evidence of goshawk use (contrast plots). Goshawks apparently did not select foraging sites based on prey abundance; indeed, abundances of some prey were lower on used than on contrast plots. Goshawks selected foraging sites that had higher canopy closure (P = 0.006), greater tree density (P = 0.001), and greater density of trees >40.6 cm dbh (P 0.0005) than on contrast plots. These results are consistent with the hypothesis that goshawk morphology and behavior are adapted for hunting in moderately dense, mature forests, and that prey availability (as determined by the occurrence of favorable vegetation structure where prey are present above a low threshold) is more important than prey density in habitat selection. Nonetheless, current management recommendations for increasing prey abundance also can create a forest structure more favorable for hunting by goshawks, particularly if stands are managed for canopy closure values above the prescribed minimum level of 40% canopy.