A linear analogue for rating pain with 10, 15 and 20 cm lines is significantly less variable than a 5 cm line (mean error of 15 cm line is 0-19%, 95% confidence limits for the group +/- 2% and an inood correlation between repeated ratins of a recalled pain distant in time. The variance of the rating is significantly less than the repeated rating of a random mark. The linear analogue rating of a constant pain stimulus is reproducible and changes in rating are likely to be real changes of opinion. Pethidine 150 mg intramuscularly had no significant effect, tested 30 minutes after the administration, on the accuracy or reproducibility of the analogue rating. A linear analogue seems a suitable method of recording the patient's opion of a severe pain such as that of labour.