Estimates regarding population parameters are often based on data from surveys. To ensure that such estimates are as accurate as possible, it is important to know the detectability resulting from the particular survey method used. We used radiotelemetry to measure detectability of ornate box turtles (Terrapene ornata), using visual-encounter surveys in a sand prairie in northwestern Illinois, USA. We found that the overall detection probability of visual-encounter surveys was 0.03, and our high frequency of nondetection was due to a failure to detect visible turtles rather than turtles being underground or hidden in dense vegetation. Despite the substantial population density at our study site, visual-encounter surveys failed to detect box turtles on most visits, which resulted in a prohibitively high number of surveys that would be required to accurately estimate population size or to infer absence of the species from a site. Our method of using radiotelemetry to measure detection probability of a survey method could be easily applied to other small, cryptic, or rare species. However, our low detection probability and high frequency of nondetections recommend against use of visual-encounter surveys alone in estimating population parameters for ornate box turtles. © 2011 The Wildlife Society.