Sources of interference in item and associative recognition memory

A powerful theoretical framework for exploring recognition memory is the global matchingframework, in which a cue’s memory strength reflects the similarity of the retrieval cuesbeing matched against the contents of memory simultaneously. Contributions at retrievalcan be categorized as matches and mismatches to the item and context cues, including theself match (match on item and context), item noise (match on context, mismatch on item),context noise (match on item, mismatch on context), and background noise (mismatch onitem and context). We present a model that directly parameterizes the matches andmismatches to the item and context cues, which enables estimation of the magnitude ofeach interference contribution (item noise, context noise, and background noise). Themodel was fit within a hierarchical Bayesian framework to ten recognition memory datasetsthat employ manipulations of strength, list length, list strength, word frequency, study-testdelay, and stimulus class in item and associative recognition. Estimates of the modelparameters revealed at most a small contribution of item noise that varies by stimulusclass, with virtually no item noise for single words and scenes. Despite the unpopularity ofbackground noise in recognition memory models, background noise estimates dominated atretrieval across nearly all stimulus classes with the exception of high frequency words,which exhibited equivalent levels of context noise and background noise. These parameterestimates suggest that the majority of interference in recognition memory stems fromexperiences acquired prior to the learning episode.

Global matching models of recognition memory

How does retrieval take place in recognition memory? A number of computational models have been developed that posit that recognition operates by a process of global matching, wherein the cue is compared to each stored memory. These cue-to-memory similarities are then aggregated to produce an index of the global similarity between the cue and the contents of memory which can then be subjected to a decision process. In this chapter, we describe a.) the theoretical rationale and successes of such models, such as their accounts of similarity and list-length effects and generalization to multiple memory tasks, b.) challenges and hurdles they have experienced, including the null list-strength effect, the mirror effect, and the extralist feature effect, and c.) recent developments, such as extensions to predictions about response latency, sources of interference outside of the study list, and integration of more realistic representations.

Testing a Strategy-Disruption Account of the List-Strength Effect

Presenting items multiple times on a study list increases their memorability, a process known as item strengthening. The list-strength effect (LSE) refers to the finding that, compared to unstrengthened (pure) lists, lists for which a subset of the items have been strengthened produce enhanced memory for the strengthened items and depressed memory for the unstrengthened items. Although the LSE is found in free recall ( Tulving & Hastie, 1972 ), it does not occur in recognition ( Ratcliff et al., 1990 ). In free recall, the LSE in mixed lists is attributed to a sampling bias promoting priority recall of strong items and consequent output interference affecting weak items. We suggest that, in recognition, the disruption of this pattern through the randomization of test probes is responsible for the null LSE. We present several pilot experiments consistent with this account; however, the registered experiment, which had more statistical power, did not support this account.

Sources of Interference in Memory across Development

Episodic memory involves remembering not only what happened but also where and when the event happened. This multi-component nature introduces different sources of interference which stem from previous experience. However, it is unclear how different interferences change across development and what may cause the changes. To address these questions, we tested 4-5-year-olds (N = 103), 7-8-year-olds (N = 82), and adults (N = 70) using item and source recognition memory tasks with various manipulations (i.e., list-length, list-strength, and word-frequency), and decomposed sources of interference using a computational model. We find that interference stemming from other items on the study-list rapidly decreases with development, whereas interference from pre-experimental contexts gradually decreases but remains the major source of interference. The model further quantifies these changes indicating that the ability to discriminate items undergoes relatively rapid development, whereas the ability to discriminate contexts undergo protracted development. These results elucidate fundamental aspects of memory development.

The list strength effect in source memory: Data and a global matching model

A critical constraint on models of item recognition comes from the list strength paradigm, in which a proportion of items are strengthened to observe the effect on the non-strengthened items. In item recognition, it has been widely established that increasing list strength does not impair performance, in that performance of a set of items is unaffected by the strength of the other items on the list. However, to date the effects of list strength manipulations have not been measured in the source memory task. We conducted three source memory experiments where items studied in two sources were presented in a pure weak list, where all items were presented once, and a mixed list, where half of the items in both sources were presented four times. Each experiment varied the nature of the testing format. In Experiment 1, in which each study list was only tested on one task (item recognition or source memory), a list strength effect was found in source memory while anull effect was found for item recognition. Experiments 2 and 3 showed robust null list strength effects when either the test phase (Experiment 2) or the analysis (Experiment 3) was restricted to recognized items. An extension of the Osth and Dennis (2015) model was able to account for the results in both tasks in all experiments by assuming that unrecognized items elicit guess responses in the source memory task and that there was low interference among the studied items. The results were also found to be consistent with a variant of the retrieving effectively from memory model (REM; Shiffrin & Steyvers, 1997) that uses ensemble representations.

The list strength effect in source memory: Data and a global matching model

A critical constraint on models of item recognition comes from the list strength paradigm, in which a proportion of items are strengthened to observe the effect on the non-strengthened items. In item recognition, it has been widely established that increasing list strength does not impair performance, in that performance of a set of items is unaffected by the strength of the other items on the list. However, to date the effects of list strength manipulations have not been measured in the source memory task. We conducted three source memory experiments where items studied in two sources were presented in a pure weak list, where all items were presented once, and a mixed list, where half of the items in both sources were presented four times. Each experiment varied the nature of the testing format. In Experiment 1, in which each study list was only tested on one task (item recognition or source memory), a list strength effect was found in source memory while a null effect was found for item recognition. Experiments 2 and 3 showed robust null list strength effects when either the test phase (Experiment 2) or the analysis (Experiment 3) was restricted to recognized items. An extension of the Osth and Dennis (2015) model was able to account for the results in both tasks in all experiments by assuming that unrecognized items elicit guess responses in the source memory task and that there was low interference among the studied items. The results were also found to be consistent with a variant of the retrieving effectively from memory model (REM; Shiffrin & Steyvers, 1997) that uses ensemble representations.