Examining the relationship between generation constraint and memory.

Self-generated information is often better remembered than read information (the generation effect). Recent research, however, has shown that generating information under fewer experimental constraints (i.e., fewer limitations on what can be generated) can increase the magnitude of the generation effect. This study systematically varied generation constraint to better understand the effects of constraint on memory. Participants encoded associated cue-target word pairs (above-below) on either the left or right side of a computer monitor. At encoding, generation constraint was manipulated by systematically varying the number of letters given to participants to generate the target word (i.e., above-below; option-choic_; bank-mon__; etc.). At retrieval, participants were given either a recognition, cued recall, or free recall test measuring both item (target word) and context memory (location on the computer monitor). Using mixed-effects logistic regression analyses to control for item-selection effects (e.g., participants producing idiosyncratic targets in some conditions relative to others), results indicated that generation constraint significantly influenced item, but not context (location) memory. The relationship between generation constraint and memory performance, however, differed by the type of memory test administered: Recognition data revealed a curvilinear relationship; cued recall showed a negative, linear relationship; and free recall showed no significant relationship. Overall, these findings provide more evidence that generation constraint has a strong yet complex effect on different aspects of memory, and further delineates some boundary conditions of the influence of generation constraint on memory.