The time-averaged alongshore bottom stress is an important component of nearshore circulation models. In a widely accepted formulation the bottom stress is proportional to ⟨|u&vec;|v⟩, the time average of the product of the instantaneous velocity magnitude |u&vec;| and the instantaneous alongshore velocity component v. Both mean and fluctuating (owing to random, directionally spread waves) velocities contribute to ⟨|u&vec;|v⟩. Direct estimation of ⟨|u&vec;|v⟩ requires a more detailed specification of the velocity field than is usually available, so the term ⟨|u&vec;|v⟩ is parameterized. Here direct estimates of ⟨|u&vec;|v⟩ based on time series of near-bottom currents observed between the shoreline and 8-m water depth are used to test the accuracy of ⟨|u&vec;|v⟩ parameterizations. Common ⟨|u&vec;|v⟩ parameterizations that are linear in the mean alongshore current significantly underestimate ⟨|u&vec;|v⟩ for moderately strong alongshore currents, resulting in overestimation of a drag coefficient determined by fitting modeled (with a linearized bottom stress) to observed alongshore currents. A parameterization based on a joint-Gaussian velocity field with the observed velocity statistics gives excellent overall agreement with the directly estimated ⟨|u&vec;|v⟩ and allows analytic investigation of the statistical properties of the velocity field that govern ⟨|u&vec;|v⟩. Except for the weakest flows, ⟨|u&vec;|v⟩ depends strongly on the mean alongshore current and the total velocity variance but depends only weakly on the mean wave angle, wave directional spread, and mean cross-shore current. Several other nonlinear parameterizations of ⟨|u&vec;|v⟩ are shown to be more accurate than the linear parameterizations and are adequate for many modeling purposes. ¿ 2000 American Geophysical Union