Computational fluid dynamics (CFD) is used increasingly to simulate the distribution of airborne contaminants in enclosed spaces for exposure assessment and control, but the importance of realistic boundary conditions is often not fully appreciated. In a workroom for manufacturing capacitors, full-shift samples for isoamyl acetate (IAA) were collected for 3 days at 16 locations, and velocities were measured at supply grills and at various points near the source. Then, velocity and concentration fields were simulated by 3-dimensional steady-state CFD using 295K tetrahedral cells, the k-ε turbulence model, standard wall function, and convergence criteria of 10(-6) for all scalars. Here, we demonstrate the need to represent boundary conditions accurately, especially emission characteristics at the contaminant source, and to obtain good agreement between observations and CFD results. Emission rates for each day were determined from six concentrations measured in the near field and one upwind using an IAA mass balance. The emission was initially represented as undiluted IAA vapor, but the concentrations estimated using CFD differed greatly from the measured concentrations. A second set of simulations was performed using the same IAA emission rates but a more realistic representation of the source. This yielded good agreement with measured values. Paying particular attention to the region with highest worker exposure potential-within 1.3 m of the source center-the air speed and IAA concentrations estimated by CFD were not significantly different from the measured values (P = 0.92 and P = 0.67, respectively). Thus, careful consideration of source boundary conditions greatly improved agreement with the measured values.