An issue in the assessment of toxic effects on neurobehavioural function in occupational or environmental settings is the presence and relative contribution of confounding variables. Factors such as education level and alcohol use are well-known to exert effects on the performance of neurobehavioural tests. In a prospective study of a cohort of vehicle spray painters, measures of neurobehavioural function, demographic characteristics, daily habits including alcohol-use and general health status were taken from the onset of their exposure on four occasions for three full years. This study provided the opportunity to assess the relative contribution of confounding variables and the study factor, solvent exposure, on neurobehavioural function in a study design which limits the heterogeneity of the study sample. In such a design the effects of confounding would expected to be reduced. Multiple linear regression analysis was used with three confounding variables, education level, alcohol use and occupational experience as well as solvent exposure as predictor variables of neurobehavioural test performance. Although the level of solvent exposure was considerably less than the composite threshold limit value for the solvent mixtures encountered, the results indicated that psychomotor performance deteriorated with increasing solvent exposure, but only on the hand steadiness test. The confounding variables showed even greater influences on performance. Education level affected performance on both psychomotor and cognitive tests in the expected direction. Occupational experience also influenced test performance but only for psychomotor tests. With increasing time in the trade, spray painters showed significantly superior performance on reaction time, hand steadiness and visual pursuit tests. Since training in spray painting focuses on psychomotor coordination, this result would be expected. Alcohol use was associated with mixed effects on neurobehavioural function. Painters who reported using greater amounts of alcohol showed significantly poorer performance on the visual test, critical flicker fusion, but better performance on the short term memory and learning measures of the paired associates test so supporting the findings of previous studies of facilitating effects of lower levels of alcohol on memory. These findings reinforce the importance of separating effects due to confounders from effects due to toxic exposures in studies of neurobehavioural function. They emphasise the dilemma of test selection, choosing tests that are sensitive enough to detect effects due to toxic exposure but which are not affected by confounding variables.