The backward detection masking of 10‐ms tonal targets by a 150‐ms tonal mask was contrasted with the backward recognition masking of the same tones by the same mask. The target‐mask interval required for 75% correct performance was about eight times as great in the recognition condition as in the detection condition. Furthermore, a generalized improvement in performance occurred over the initial course of training in both the detection and recognition conditions. In a second experiment it was found that a remote mask produced greater backward recognition masking but less backward detection masking. That these differences were observed with the same subjects, at the same level of training, and with identical stimuli, indicates that procedural differences alone cannot account for the differences between backward detection and recognition masking.