Unmasking latent inhibition in humans.

Presentations of a to-be-conditioned stimulus (CS) on its own impairs subsequent learning when that CS is paired with an unconditioned stimulus (US). Evidence for this latent inhibition (LI) effect in humans is said to require a "masking task" that diverts attention from the CS during preexposure. We present three experiments that demonstrate LI in humans without masking. Subjects performed a computerised task, making speeded responses to an imperative cue (the US) presented within a continuous stream of stimuli. During preexposure, a to-be-CS was presented 20 times among other stimuli, but excluding the US. Instructions ensured subjects actively monitored all stimuli at this time. This was immediately followed by the training phase, which included the US, the preexposed CS, and a novel CS. Both CSs were reliably followed by the US, but these associations were incidental to the instructed task. Nonetheless, some subjects learned the CS-US associations, responding faster when the US followed a CS than when it was unsignalled. All three experiments also found evidence for LI, in that subjects learned the novel CS-US association sooner than the preexposed CS-US association. We conclude that humans can show LI even when actively attending to the CS during preexposure.