Testing the primary and convergent retrieval model of recall: Recall practice produces faster recall success but also faster recall failure

Asymmetric Weights and Retrieval Practice in an Autoassociative Neural Network Model of Paired-Associate Learning

Rizzuto and Kahana (2001) applied an autoassociative Hopfield network to a paired-associate word learning experiment in which (1) participants studied word pairs (e.g., ABSENCE-HOLLOW), (2) were tested in one direction (ABSENCE-?) on a first test, and (3) were tested in the same direction again or in the reverse direction (?-HOLLOW) on a second test. The model contained a correlation parameter to capture the dependence between forward versus backward learning between the two words of a word pair, revealing correlation values close to 1.0 for all participants, consistent with neural network models that use the same weight for communication in both directions between nodes. We addressed several limitations of the model simulations and proposed two new models incorporating retrieval practice learning (e.g., the effect of the first test on the second) that fit the accuracy data more effectively, revealing substantially lower correlation values (average of .45 across participants, with zero correlation for some participants). In addition, we analyzed recall latencies, finding that second test recall was faster in the same direction after a correct first test. Only a model with stochastic retrieval practice learning predicted this effect. In conclusion, recall accuracy and recall latency suggest asymmetric learning, particularly in light of retrieval practice effects.

Item repetition and retrieval processes in cued recall: Analysis of recall-latency distributions

The SAM (search of associative memory) model provides a unified account of accuracy effects, assuming that retrieval is a cue-dependent two-stage process of sampling and recovery, which depends on the strength of items relative to all others and on that item associated with the sampling trace, respectively. On the other hand, the relative strength model uniquely provides latency predictions, assuming that recall latency is determined solely by relative strength (similar to the sampling rule in SAM): Latency should remain unchanged for strong and weak items in pure lists, but will be shorter for strong items than for weak items in mixed lists. To test the predictions, the present study examined accuracy and latency distributions, which were fit with the ex-Gaussian, using item repetition as a means of strengthening. Massed versus spaced repetitions were used where repetitions were either cue-target pairs or cue alone. When repetitions were spaced in mixed lists, accuracy and latency both increased with cue-target repetitions, relative to cue-only repetitions, and slow recall for cue-target repetitions was due to initially nonretrievable items. However, even after successful recall on a pretest, cue-target repetitions led to an increase in latency in pure lists. These findings are difficult to reconcile with relative-strength explanations of latency. They indeed suggest that (1) separate traces are created for each repetition, (2) memory traces are updated if the item is retrieved (otherwise, new traces are stored), and (3) recovery plays a role in latency, which are discussed with the distinction between sampling and recovery of SAM.