Clinical voice testing forms part of the routine assessment of children's hearing. However, its role in paediatric practice has not been defined. In particular, the sensitivity and specificity of the test in detecting the presence of a hearing impairment is not known. The value of the test as a screen was assessed by comparison of the results of free-field clinical voice testing with pure-tone audiometric thresholds in 141 children aged 5-12 years. In order to pass the screen, a child had to respond to a whispered voice 2 feet (60 cm) from the test ear. The mean pure-tone average (average at 0.5, 1.0 and 2.0 kHz) of those ears that passed was 14 dB HL compared with 31 dB HL for those that failed (P less than 0.001). The test was 80% sensitive (95% specific) in detecting a hearing impairment greater than or equal to 25 dB HL, and 89% sensitive (90% specific) in detecting a hearing impairment greater than or equal to 30 dB HL. Therefore, clinical voice tests will fail to detect a significant proportion of children with a hearing impairment and cannot be advocated as a screening method. The test would appear to be of limited value in a specialist situation where facilities for pure-tone audiometry exist.