Temporal and spatial variabilities of flow distribution significantly influence solute transport in soils. This laboratory study was designed to investigate the effects of temporal variation in flow velocity on pesticide transport in soils. Two pesticides, metribuzin (weakly adsorbed) and atrazine (moderately adsorbed), were chosen along with the following two soils: Cecil (<2 mm) and Sharkey (2--4 mm). Several tritium pulses were introduced into packed soil columns (15 or 30 cm in length) under different flow velocities to obtain velocity-dependent dispersion coefficients (D). Subsequently, several atrazine and metribuzin pulses were introduced under conditions of constant and variable velocities. For each experiment, changes in flow velocity were stepwise using a piston flow pump and were carried out during pulse application and leaching. For constant and variable flow velocity experiments, approximately similar pulse volumes and average flow velocities were maintained. Values of D versus pore water velocity (v) from tritium breakthrough curves (BTCs) were well described using a linear equation for both soils. Identical BTCs for metribuzin were observed under conditions of constant or variable flow velocities in the Cecil soil column. However, metribuzin transport in the Sharkey soil was significantly influenced by velocity variations. Atrazine transport in the Sharkey soil was also significantly affected by variations in flow velocity. We further examined the error when an average rather than actual velocity distribution was used in BTC representation. The resulting experimental BTCs (concentration versus velocity-averaged pore volume) exhibited early arrival and the appearance of multiple peaks. Moreover, predictions of such BTCs based on the convective-dispersive equation were not successful. We concluded that actual water velocity distributions should be used in BTC representation, and, whenever possible, the use of an average velocity should be avoided. ¿ American Geophysical Union 1996