In a series of trials, ward nurses independently classified the disability and distress states of medical and surgical patients. Criteria for inter-rater reliability, in terms of correlation and absence of bias, were met after each rater had taken part in a training session which gave immediate feedback about performance.