Aims: To test a measurement model based on clinicians' assessments of patient data that allows simple and confident clinical validation of any statistical or numerical technique designed to separate patients improving with treatment from those who are not, particularly for pain that shows large daily variation. Methods: Diaries using daily visual analog scales (VAS) of pain intensity were obtained from 39 patients treated for chronic temporomandibular disorders. Three experienced clinicians visually assessed 39 VAS/time graphs. Criteria indicating improvement (general trend, height and apparent frequency of graph spikes) evolved over 3 assessments. The third assessment defined improvers visually. Numeric analyses considered the difference between first and last months of treatment for mean, area under the curve (AUC), and maximum VAS scores. Thresholds of 40%, 50%, or 60% pain reduction defined improvement numerically. Aggregate sensitivity and specificity was compared with visual definition to find the optimal threshold. Results: Patients were defined visually as improvers, nonimprovers, and borderline cases. Interexaminer reliability for identifying improvers was good (k = 0.79). Mean VAS and AUC were highly correlated (r = 0.999). The optimal threshold of mean and maximum VAS relative to visual definition was 50% pain reduction. Cases defined as improvers by both mean and maximum agreed best with the visual definition (sensitivity 90%, specificity 84%). Conclusion: Visual assessment of VAS demonstrates distinct pain/time patterns that can validate numeric definition of complex pain recovery. No single numeric method can be guaranteed to give a clinically valid outcome. Keywords: chronic pain, pain diaries, pain measurement, stabilization splint, temporomandibular disorders, treatment outcomes, visual analog scales