The 60-item GHQ was validated in a community population by comparison with the CIS. The GHQ failed to identify nearly half of the psychiatric ‘cases' in this population. Those missed were similar to those detected except for greater chronicity of illness and more frequent social and interpersonal problems. The GHQ appears to be unsuitable as a screening instrument for mental illness in the community and the possible reasons are discussed. Principal components analysis resulted in a 15-item GHQ factor which, when used with Likert scoring, resulted in considerable improvement and failed to identify only 4 per cent of ‘cases'. It is suggested that this may provide a more satisfactory screening instrument.