Pain intensity ratings of 'usual' pain, or pain 'on average', are gaining in popularity since they are arguably a more realistic measure of a patient's pain status than the single snapshot of 'current' pain. An alternative to the 'actual average' of ratings obtained from multiple measures is the single rating of patients' recall of their 'usual' pain over a period of time, usually 1 week. The use of such a scale relies on the assumption that patients can accurately recall their 'usual' pain. Although accuracy of memory for pain has been investigated, most studies have failed to use appropriate statistical analyses for accuracy (validity). In this study, 200 back pain patients completed four daily recordings of pain intensity over 7 days. These were averaged to compute 'actual average' pain intensity. The next day, patients estimated their pain over the week at its 'least', 'worst' and 'on average' ('usual') as well as recording their 'current' pain intensity. Using the Intra-class Correlation Coefficient (ICC) to compute accuracy, the single rating asking patients to estimate their pain 'on average' over the week was found to be an accurate measure of 'actual average' pain intensity (ICC=0.82) and more accurate than 'current' pain (ICC=0.66). Although some composite measures of single ratings gave more accurate estimates of 'actual average' pain, this was not considered sufficient advantage to advocate their use. The results of this study propose the single rating of pain 'on average' as a valid and practical measure of a patient's pain intensity over a period of 1 week.