The myth of diagnostic arthroscopy still remains with us. Trainees going through this unit have been known to list patients with a knee problem for arthroscopy without making a clear clinical diagnosis.To improve this situation, a prospective consecutive series of 130 arthroscopies took place. Each patient was clinically assessed preoperatively and a diagnosis was committed to paper. This was compared with the operative findings. The diagnostic accuracy, and the ability to exclude normal knees from the operating theatre, were analysed.An overall 78% diagnostic accuracy rate was achieved. Only 7.7% of all arthroscopies were found to be normal. These figures compare favourably to previous studies.The clinical accuracy was directly correlated with the clinical experience of the surgeon involved in each case.Strictly controlling the indications for arthroscopy with a protocol reduces the number of unnecessary procedures and leads to an improvement in diagnostic accuracy.