Background: No search strategies have been published that focus on the retrieval of studies on the accuracy of signs and symptoms. Studies of signs and symptoms are an important, albeit small, subset of diagnostic accuracy studies. By facilitating information retrieval, specific and focused search strategies have the potential to enhance both the development of systematic reviews and the practice of evidence-based medicine. Objectives: To develop empirical search strategies in MEDLINE for retrieving studies on the accuracy of signs and symptoms in the diagnosis of disease. Methods: We identified all studies (original and review) that reported data on the sensitivity or specificity of signs or symptoms from within a collection of diagnostic studies compiled by a hand search of 161 journals by the Hedges team in 2000. We tested the sensitivity, specificity, and accuracy of candidate search terms, alone and in combination, using articles on signs and symptoms from the hand searched collection as the gold standard. Results: Sixty (0.1%) of 52,657 articles (total retrieval is from a PubMed search limited to the 161 journals hand searched by the Hedges Team in 2000) reported data on the accuracy of signs or symptoms. The most sensitive single-term, diagnosis[subheading], had a sensitivity of 95% and a specificity of 70%. The most specific single-term (while keeping sensitivity > 50%), specificity[TIAB], had a sensitivity of 57% and a specificity of 98%. Among the multiple-term strategies with a sensitivity of 100%, the most specific strategy had a specificity of 87%. Conclusions: The empirical strategies proposed here can optimize retrieval of studies on the accuracy of signs and symptoms from MEDLINE. Validation of the strategies in an independent dataset is pending.