A common procedure in paleohydraulic analysis is to use some measure of the size of particles in a deposit to estimate flow velocity of the depositing current, assuming that the selected size reflects the competence of the flow. For deposition from turbidity currents, a seemingly logical selection for determination of the suspension competence is the median size, because just prior to deposition all of the particles in the mixture must have been in suspension. Many papers in the literature use grain size of turbidites to estimate flow strength in this way. Komar (1985) discovered, however, that use of the median size can lead to serious underestimation of the flow velocity as determined independently from stability fields for bed forms. This discrepancy can be explained if deposition from turbidity currents is controlled mainly by decreasing capacity of the decelerating flow to carry particles of all sizes, including those that a clear current of the same speed would be fully competent to suspend. An important consequence is that suspension criteria (e.g., u (sub *) > settling velocity) cannot be used to reliably estimate the flow velocity of a turbidity current from its deposit. The hypothesis that deposition is controlled mainly by decrease of capacity is supported by the results of a computer simulation of turbidite deposition. The "deposit" predicted by the simulation differs from poorly sorted natural turbidites due to settling lags and lateral grading in the bodies of natural currents.