We tested the efficacy of three methods (historical nest search, broadcast search, and tree transect search) for detecting presence of the Northern Goshawk (Accipiter gentilis) at occupied nest areas during the 1994 breeding season using only a single visit to a previously known nest area. We used detection rates in a probability model to determine how many visits are required to have confidence in reporting absence of goshawks. The purpose of this study is to understand if the three methods for detecting goshawks are robust enough for managers to rely on them for making land management decisions that may impact goshawk nest areas. Blind tests were conducted throughout the western United States. Results were similar among methods with goshawk presence going undetected at 36-42% of the occupied nest areas after a single visit. These results indicate that a single visit to a nest area is inadequate to provide reliable information on nest area occupation. Our probability of detection model showed that if each detection method is repeated three (historical or tree transect) or four (broadcast) times, goshawk absence can be inferred with a high level of confidence. Conclusions regarding nest area occupation using a single visit sampling method should be made with utmost caution. Classifying a nest area as vacant, when in fact goshawks are present, is a serious concern and leads to spurious conclusions. Land managers making habitat-altering decisions should not rely on a single visit to nest areas to establish the absence of goshawks. Possibilities for improving the detection of nesting goshawks include multiple independent visits using the same method, using a sequence of techniques in combination to yield an improved cumulative probability of detection, or developing a new method yielding a higher probability of detection. The historical nest search obtained the best results, followed by the tree transect and broadcast search.