Estimating exposure to contaminants emitted into workroom air is essential for worker protection. Although contaminant concentrations are often not spatially uniform within workrooms, many methods for estimating exposure do not adequately account for this variability. Here the impact of temperature differences within a room on spatial contaminant distribution was studied. Tracer gas (99.5% propylene) concentrations were monitored automatically at 144 sampling points with a photoionization detector. One wall was chosen to represent a building's external wall and was heated or cooled to simulate summer or winter conditions. Experiments were preformed at two flow rates (5.5 and 3.3 m(3) min(-1)) and six thermal conditions (isothermal, three summer conditions and two winter conditions). For 5.5 m(3) min(-1) and all thermal conditions, the coefficient of variation (CV) ranged from 0.34 to 0.45 and the normalized average concentrations were similar. For 3.3 m(3) min(-1), winter conditions produced greater spatial variability of concentration (CV = 0.72 and 1.10) than isothermal or summer conditions (CV range = 0.29-0.34). Tests simulating winter conditions suggest that the resulting stable temperature structure inhibited the dilution of the tracer and enhanced its segregation in the lower portion of the room, especially for the lower flow rate (3.3 m(3) min(-1)). Therefore, not explicitly addressing thermal effect in exposure modeling may impact the estimated accuracy and precision when used for rooms that are non-isothermal and not well mixed. These findings also have implications for air monitoring. Dispersion patterns for different thermal conditions were found to be substantially different, even when the mean concentrations were nearly the same. Thus, monitoring data from a single season should not be taken as representative of the entire year, when summer and winter conditions create temperature gradients in a room.