A within-patient change in pain score after treatment is statistically 'reliable' when it exceeds the smallest detectable difference (SDD). The aims of the present study were to: (i) determine SDDs for VAS-scores of pain intensity, for sufficiently long test-retest intervals to include most biological fluctuations, (ii) examine whether SDD is invariant to baseline score, and (iii) discuss the value of reliable change (RC) for detecting clinically important difference (CID) or as a possible indicator of successful treatment. SDDs were determined using duplicate data from 118 patients with myogenous Temporomandibular disorders: (1) VAS-scores of pain intensity from the masticatory system in a pre-treatment diary, and (2) VAS-scores of pain intensity from the hand (cold-pressor test). RC was determined in VAS-scores from a pre- and post-treatment questionnaire. The long-term SDD was 49mm. A regression analysis on duplicate VAS-scores showed that SDD was largely invariant to the baseline level. Because RC (change>SDD) exceeded CID, it might serve as an indicator of successful treatment. However, only 17% of the patients showed RC after treatment, mainly because the baseline was smaller than SDD in 67% of the patients thus making detection of any treatment effect impossible. For patients with possible detection (33%), the frequency of RC was 51%. If the detection threshold would be avoided by provoking pain in patients with a low baseline, a long-term RC in VAS-scores might occur in about half of all myogenous TMD patients and might then serve as an indicator of cases of treatment success.