Several variants of a learning model for forced‐choice detection experiments (Atkinson and Kinchla, 1965) may be produced by making various reasonable assumptions regarding which events are effective in producing response bias changes. Atkinson and Kinchla assumed the bias changed according to a single parameter stochastic learning mechanism; such changes occurring only when no signal was detected. An alternative formulation uses two learning parameters and postulates that the bias changes on every trial, but at different rates according to whether or not the signal is detected. This two‐parameter model was applied to the data of Atkinson and Kinchla, and numerical estimates of the parameters were obtained which confirm the conjecture that the bias changes principally during non‐detection trials. In addition, the parameter estimates indicate that the relative effectiveness of information feedback is determined by its relative frequency of occurrence, an inference which was not possible from the single‐paramete