We describe a clinical procedure for assessing knee pain: 10 standardised movements of the knee are made (r active and 6 passive), and the subject's behavioural response to each is scored by the assessor on a 0-3 scale. The total score is the Pain Index of the Knee (PIK). We report investigations of repeatability, inter-assessor agreement, and validity of components of the PIK compared with VAS ratings by the subject. We conclude from these results and from our experience of the PIK in practice that it is useful both clinically and for research purposes. It is a simple and efficient procedure for pain measurement in terms of behavioural response, with good reliability and validity, at least at low and medium levels of pain.