Postures and muscle activity in the upper body were recorded from 50 academics office workers during 2 hours of normal work, categorised by observation into computer work (CW) and three non-computer (NC) tasks (NC seated work, NC standing/walking work and breaks). NC tasks differed significantly in exposures from CW, with standing/walking NC tasks representing the largest contrasts for most of the exposure variables. For the majority of workers, exposure variability was larger in their present job than in CW alone, as measured by the job variability ratio (JVR), i.e. the ratio between min–min variabilities in the job and in CW. Calculations of JVRs for simulated jobs containing different proportions of CW showed that variability could, indeed, be increased by redistributing available tasks, but that substantial increases could only be achieved by introducing more vigorous tasks in the job, in casu illustrated by cleaning.