BACKGROUND: Nerve conduction studies (NCS) are useful tools for diagnosing carpal tunnel syndrome (CTS). Establishing the normal values is the first step required for utilizing NCS for diagnosis. Previous epidemiological studies demonstrated the presence of fairly large number of false-positive subjects regarding NCS among control population, which has not been properly considered in past studies. This study proposed a new method to address this issue. METHODS: Non-diabetic 144 CTS patients were retrospectively enrolled using clinically defined inclusion criteria. Controls consisted of 73 age-matched volunteers without hand symptoms. Six NCS parameters were evaluated including peak-latency difference by the thumb method (thumbdif) and that by the ring-finger method (ringdif). The Youden index of the receiver operator characteristic curve was used both to judge the sensitivity of a parameter and to identify false-positive cases that were thought to have subclinical median neuropathy at the wrist. The linear function of six parameters was constructed, and the coefficient for each parameter was variously changed. RESULTS: When the Youden index took on the maximum value, seven control subjects (10%) were identified as false-positive and were excluded from the calculation of normal values. The most sensitive parameter before exclusion was thumbdif, whereas ringdif became the most sensitive after exclusion. The cut-off value for ringdif was 1.15 ms before exclusion, but was 0.37 ms after exclusion. CONCLUSION: This method can be widely applied to solve the statistical problem when the gold standard is lacking, and the outside reference standard is not completely reliable.