Objectives. 1) To compare two measurement models of the BDI in chronic pain sufferers to see which provides the better fit; 2) to assess whether model fit differs for a facial pain sample compared to a sample of pain sufferers attending a multidisciplinary pain clinic; and 3) to establish which affective and somatic sub-scales of the BDI could be used in chronic pain research.

Design. Two groups of chronic pain sufferers, a facial pain group, and a group attending a multidisciplinary pain clinic completed self-report questionnaires on pain (Multidimensional Pain Inventory), depression (BDI), and measures of anxiety and depression-related pain cognitions (the Spielberger State-Trait Anxiety Inventory and the Pain Cognitions Questionnaire). The measurement models of the BDI were tested using LISREL structural equation modelling and their construct validity examined using partial correlation analysis.

Method. A total of 173 people attending a multidisciplinary pain clinic and 157 patients attending a facial pain clinic completed self-report measures of pain and mood prior to their respective clinical consultations.

Results. The model offered by Novy et al. (containing one affective factor ‘Negative-attitude suicide’ and two somatic factors‘Performance difficulty’ and‘Physiological manifestations’) fitted both pain groups better than the model offered by Williams and Richardson (containing one affective factor ‘Self-reproach’, one somatic factor ‘Somatic disturbance’ and one factor with a mixture of both affective and somatic items ‘Sadness about health’). However, when the factors were allowed to correlate in the latter model, both models were broadly equivalent.

Conclusions. The two measurement models adequately fitted data in both pain samples when the factors were allowed to intercorrelate in the Williams and Richardson model. Both the affective scales offered by both models could be used in future research, although the somatic factor offered by the Williams and Richardson model offered much higher levels of internal reliability than either of those offered in the Novy et al. model. The findings are discussed in relation to the issue of depression in chronic pain.