A prospective observational study of cohorts of patients undergoing hip replacement (30), knee replacement (31), and spinal nerve root decompressive surgery (30) were interviewed pre-operatively to identify factors which might correlate with and potentially predict severe post-operative pain and dissatisfaction with analgesic management. The hip patients comprised 33% females and averaged 64 years, while the knee patients were 45% female and older (mean 71 years) and the spinal patients were 43% female and averaged 50 years. The three groups were similar with respect to all other pre-operative variables. Pain intensity was assessed mainly by self-report using the Present Pain Intensity (PPI) and Visual Analogue Scales (VAS) of the McGill Pain Questionnaire. The PPI was preferred by patients and nurses and, as there were no analytical advantages for the VAS, the PPI data are presented. The average post-operative pain during routine management mainly with patient controlled intravenous opiate, was mild to moderate and declined over days 1-5, declined further at discharge but rose slightly 1 month after discharge. The hip replacement patients experienced significantly (P < 0.01) less pain overall than the patients in the other two groups. Nurses' assessments of pain severity from observed behaviour were low and agreed poorly with the patients' self reports. Assessed on Likert Scales (0-6), the patients generally indicated good or excellent pain control, better than expected pain experience, and high levels of satisfaction with analgesic management. Significant (P < or = 0.01) multivariate correlates of severe post-operative pain assessed by logistic regression analysis of 11 variables were female gender, high pre-operative pain severity, and younger age. Significant (P < or = 0.01) multivariate correlates of both worse than expected pain experience and low satisfaction were female gender, high pre-operative pain severity, high anxiety about risks and problems, low expected pain severity, age (younger) and high willingness to report pain. These variables may reasonably be tested in further studies as potential predictors of adverse post-operative pain experience.