The job exposure matrix (JEM) has been employed to assign cumulative exposure to workers in many epidemiological studies. In these studies, where quantitative data are available, all workers with the same job title and duration are usually assigned similar cumulative exposures, expressed in mgm(-3)xyears. However, if the job is composed of multiple tasks, each with its own specific exposure profile, then assigning all workers within a job the same mean exposure can lead to misclassification of exposure. This variability of exposure within job titles is one of the major weaknesses of JEMs. A method is presented for reducing the variability in the JEM methodology, which has been called the task exposure matrix (TEM). By summing the cumulative exposures of a worker over all the tasks worked within a job title, it is possible to address the variability of exposure within the job title, and reduce possible exposure misclassification. The construction of a TEM is outlined and its application in the context of a study in the primary aluminium industry is described. The TEM was found to assign significantly different cumulative exposures to the majority of workers in the study, compared with the JEM and the degree of difference in cumulative exposure between the JEM and the TEM varied greatly between contaminants.