Retrieval practice via corrective feedback: is learning better for targets in an expected or surprising sense?

Adult participants learned homographic cue words and weakly associated targets. Each target was in the dominant (expected) sense of the cue (e.g., habit - daily) or the subordinate (surprising) sense (e.g., habit - nun). Attempting to guess the target before reading it produced better target retention than did simply reading the cue and target without guessing. Replicating recent studies, recall accuracy was also higher for expected than surprising targets, whether the cue + target reading time was fixed (Experiment 1) or under participants' control (Experiment 2). A new result was that this advantage was larger in the guess than the read condition. In Experiment 3, all targets were in the dominant sense of the cue, and prime phrases activated the dominant or subordinate sense before the target was either guessed or presented. Experiment 3 thus disentangled guess-target congruence from target sense. When the analysis was restricted to trials with a guess consistent with the prime, subordinate primes (incongruent with the targets) produced substantially lower target recall accuracy. This result suggests that guess-target congruence aids learning, and that the results of Experiment 1 and 2 were not due to pre-existing differences in the characteristics of dominant and subordinate targets.