Clinicians and researchers often ask patients to remember their past pain. They also use patient's reports of relief from pain as evidence of treatment efficacy, assuming that relief represents the difference between pretreatment pain and present pain. We have estimated the accuracy of remembering pain and described the relationship between remembered pain, changes in pain levels and reports of relief during treatment. During a 10-week randomized controlled clinical trial on the effectiveness of oral appliances for the management of chronic myalgia of the jaw muscles, subjects recalled their pretreatment pain and rated their present pain and perceived relief. Multiple regression analysis and repeated measures analyses of variance (ANOVA) were used for data analysis. Memory of the pretreatment pain was inaccurate and the errors in recall got significantly worse with the passage of time (P < 0.001). Accuracy of recall for pretreatment pain depended on the level of pain before treatment (P < 0.001): subjects with low pretreatment pain exaggerated its intensity afterwards, while it was underestimated by those with the highest pretreatment pain. Memory of pretreatment pain was also dependent on the level of pain at the moment of recall (P < 0.001). Ratings of relief increased over time (P < 0.001), and were dependent on both present and remembered pain (Ps < 0.001). However, true changes in pain were not significantly related to relief scores (P = 0.41). Finally, almost all patients reported relief, even those whose pain had increased. These results suggest that reports of perceived relief do not necessarily reflect true changes in pain.