To improve accuracy in the diagnosis of carpal tunnel syndrome (CTS) by comparing cross-sectional area (CSA) measurements of the median nerve obtained at the level of the carpal tunnel (CSAc) with those obtained more proximally (CSAp), at the level of the pronator quadratus muscle. The study protocol was approved by the institutional review board, and all subjects gave written informed consent. One hundred wrists of 68 consecutive patients with CTS (16 men, 52 women; mean age, 57.9 years; range, 25-85 years) and 93 wrists of 58 healthy volunteers (16 male, 42 female; mean age, 55.1 years; range, 17-85 years) were examined with ultrasonography (US). Electrodiagnostic test results confirmed the diagnosis of CTS in all 68 patients. The US examiner was blinded to these test results. The CSA of the median nerve was measured at the carpal tunnel and proximal levels, and the difference between CSAc and CSAp (Delta CSA) was calculated for each wrist. The mean CSAc in healthy volunteers (9.0 mm(2)) was smaller than that in patients (16.8 mm(2), P < .01). The mean Delta CSA was smaller in asymptomatic wrists (0.25 mm(2)) than in CTS-affected wrists (7.4 mm(2), P < .01). Receiver operating characteristic analysis revealed a diagnostic advantage to using the Delta CSA rather than the CSAc (P = .036). Use of a Delta CSA threshold of 2 mm(2) yielded the greatest sensitivity (99%) and specificity (100%) for the diagnosis of CTS. Receiver operating characteristic analysis revealed improved accuracy in the diagnosis of CTS determined with the Delta CSA compared with the accuracy of the diagnosis determined with the CSAc.