Previous studies have shown a robust correlation between variability of clinical pain scores and responsiveness to placebo (but not active drug) in pain studies, but explanations for these relationships are lacking. We investigated this further by assessing relationship between the Focused Analgesia Selection Test (FAST); a psychophysical method that quantifies pain reporting variability in response to experimental stimuli, variability of daily clinical pain scores as captured via diary, and response to treatment in the context of a randomized controlled crossover trial of naproxen vs. placebo in knee osteoarthritis. Evoked pain using the Staircase-Evoked Pain Procedure (StEPP®) served as the primary efficacy endpoint. Variability of daily pain scores and the FAST were assessed at baseline. Fifty-five subjects completed the study and were included in the analyses. Our results indicated a statistically significant, moderate linear relationship between variability of clinical and experimental pain reports (r= -0.416, P=0.004); and both also correlated with the placebo response (r= 0.393, P=0.004; r=-0.371, P=0.009 respectively), but only the FAST predicted the treatment difference between naproxen and placebo, as demonstrated both in a regression model (P=0.002, Beta=0.456, t=3.342) and in a Receiver Operator Curve analysis (ROC=0.721). Our results extend previous findings to include a correlation between experimental pain variability and the placebo response, and suggest that experimental pain variability is a better predictor of patients who respond preferentially to drug over placebo. A theoretical model unifying these observations is proposed and practical implications are discussed.