Pigeon subjects received training of a diffuse stimulus as either a conditioned inhibitor or a conditioned facilitator of a keylight signal for food. The ability of these diffuse stimuli to modulate responding was then assessed with two other keylights that were undergoing discrimination reversal. At a point where responding was equivalent for the two targets, the modulators had a greater impact on the target undergoing extinction than on the target undergoing acquisition. These results have implications for the nature of modulation and extinction.Considerable recent interest has focused on the ability of one Pavlovian stimulus to modulate responding to another. Several Pavlovian procedures appear to generate stimuli that do not simply signal the unconditioned stimulus (US), but instead set the occasion for the relationship between another stimulus and the US. Perhaps the two most frequently studied instances are conditioned inhibition (feature negative) and conditioned facilitation (feature positive). In a conditioned inhibition procedure, one stimulus (A) is followed by the US except when it is accompanied by another stimulus (X). The result ofsuch a procedure is that A evokes a response but X inhibits that response to A. In a conditioned facilitation procedure, the converse contingencies are arranged so that A is followed by the US only when it is accompanied by X. Under some circumstances, such a procedure results in responding to A only in the presence of X.A major focus for studies of such modulation has been the degree to which X will transfer its modulation from A, the original training target, to another target stimulus, B. Successful transfer has been taken as indicating that X has some general property, perhaps extending to various B stimuli paired with the same US as that with which A has been paired (e.g., Rescorla, 1985). Failure of transfer has been taken as indicating that X acts more specifically on A, perhaps by affecting A's associations (e.g., Holland, 1983).Because of the diagnostic potential of transfer, there have been a number of investigations of the nature of the target stimuli to which X might transfer. Studies of modulatory transfer have been conducted on both facilitative and inhibitory learning; however, most of this work has concentrated on the ability ofa facilitatory X to improve