Objectives. We sought to assess, among participants in the Gazel cohort, the association between certain occupational biomechanical and psychosocial factors evaluated in 1994–95 and the onset of shoulder pain reported as severe between that period and 2006, as well as its persistence now that most subjects are retired. Methods. All subjects who participated in the ARPEGE study (a subsample of the Gazel cohort) and who completed the Gazel questionnaire in 1994 or 1995 were included in the study. The psychosocial factors, collected in 1995, were derived from a preliminary version of the Karasek questionnaire assessing psychological demand and decision latitude. The biomechanical factor is "work involving arm movement at shoulder height with a load" ("abduction and load"). The association between these factors and the incidence of significant scapular pain (e 4/8 on a numerical scale) between 1994–95 and 2006 or its persistence was studied based on the presence of the biomechanical factor, the psychosocial factor, or both, also taking into account retirement prior to 2001. Results. In 2006, 1,482 men and 1,132 women responded to the questionnaire. None of the psychosocial factors studied were significantly associated with severe shoulder pain among those who did not have it in 1994–95, unlike the occupational biomechanical factor studied in men. Among subjects with significant scapular pain in 1994–95, the only work-related factor among those studied that appeared to be possibly associated with the persistence of significant scapular pain in 2006 in multivariate analysis was retirement before 2001 (which reduces pain, odds ratio = 0.49, 95% CI = 0.23–1.04). Conclusion. In the long term, only occupational biomechanical exposure was found to be associated with the onset of severe shoulder pain in 2006 among men (those exposed only). None of the psychosocial factors studied remained significantly associated in the follow-up, after adjusting for biomechanical and personal factors. (R.A.)