The generation effect: Support for a two-factor theory.

When a response word bearing an orthographic, acoustic, or semantic relation to a stimulus word is generated rather than read, later recall is enhanced. Such have been attributed to the activation or strengthening of response-specific features in memory and to the activation or strengthening of the relation between a stimulus and response. This series of experiments yields evidence suggesting that both mechanisms are involved. The pattern of interactions in the size of the generation effect across type of recall test (cued or free) cannot be accommodated by any one-factor theory. The results of these experiments also suggest that within-subjects manipulations of read and generate study conditions inflate the apparent size of the effect of generation on a given pair by confounding such pair-specific effects with certain whole-list effects, such as differential attention and output interference.