Fear of pain and avoidance are psychological factors of primary importance when assessing chronic musculoskeletal pain, which are often measured with the Fear-Avoidance Beliefs Questionnaire (FABQ). Both two- and three-subscale versions have been described. The aims of this study were: to assess the cognitive traits of musculoskeletal pain patients using a newly validated Greek version of the FABQ, and to further examine the construct validity and responsiveness of the measure. Factor analysis yielded three factors that accounted for 65% of the total variance. Physical activity explained 12.3% of the variance and was identical to the original version, unlike the work subscale which split into two: the FABQ work1 related to "work as cause" (15.2% of the variance) and the FABQ work2 related to "work as prognosis" (37.5% of the variance). Internal consistency was good (0.72-0.90). Test-retest reliability was satisfactory and close to the original version both for individual items and the subscales. Responsiveness of the 3-factor model was satisfactorily assessed as the ability to detect: (A) change in general - (paired t test, effect size); (B) clinically important change (paired t test, standardised effect size), and (C) real change in the concept being measured (ROC analysis). Construct validity of the FABQ was shown through the interaction with anxiety and depression, pain control and responsibility, psychological distress and pain intensity, and criterion-related validity through the association with another fear-avoidance measure (TSK). New aspects of responsiveness and construct validity were demonstrated for the FABQ, using a three-subscale validated Greek version.