To establish the value of median nerve compression with wrist flexion as a provocative test for carpal tunnel syndrome (CTS), we performed a prospective study of 64 patients (95 hands) with CTS confirmed by electrodiagnostic studies and 50 normal subjects (96 hands). We recorded results for the common provocative tests (Tinel's percussion test, Phalen's wrist flexion test and the carpal compression test) and the new test which combines wrist flexion with median nerve compression. Using a receiver operator characteristic curve (ROC) technique, we found that the optimal cut-off time for the wrist-flexion and median-nerve compression test was 20 s, giving a sensitivity of 82% and a specificity of 99%. These results were significantly better than for Phalen's wrist flexion test (61% and 83%, respectively) and for the sensitivity of Tinel's test (74%). The positive predictive values of the wrist flexion and median-nerve compression test, which is more important clinically, were 99%, 95% and 81% at population prevalences of 50%, 20% and 5%, respectively. These were significantly better than those of the three other provocative tests at each prevalence. Electrodiagnostic studies have significant false-positive and false-negative rates in CTS, and therefore provocative tests remain important in its diagnosis. We have shown that wrist flexion combined with the median-nerve compression test at 20 s, is significantly better than the other methods, and may thus be clinically useful.