BACKGROUND: We evaluated the agreement between a questionnaire and an observational checklist for exposure assessment in the setting of an upper-limb work-related musculoskeletal disorders (UWMSD) surveillance program in a population with a high level of physical exposures. METHODS: A surveillance program was implemented in a large shoe factory. Physical exposures were assessed in 1996 by a self-administered questionnaire and by the direct observation of work tasks assessed using a checklist filled out by trained assessors. Items were summed into a "questionnaire" score and an "observational" score. These scores were compared by Pearson's correlation. The association between exposure assessment by each method and UWMSD incidence between 1996 and 1997, defined by a standardized examination, was also studied. RESULTS: Correlation between the "questionnaire" score and the "observational" score was low among the 196 workers (77%) who received both evaluations (rho=0.06, p>0.05). Only exposure assessed by the questionnaire method was significantly associated with high incidence of UWMSD between 1996 and 1997, with good sensitivity (97%) and poor specificity (27%). CONCLUSION: In this surveillance program, self-reported physical exposures assessed by questionnaire and by direct observation did not evaluate same dimensions of high physical exposures. In this sample, exposures assessed by questionnaire identified workers at high risk of incident UWMSD more precisely than exposures identified by direct observation.