Power Function Forgetting Curves as an Emergent Property of Biologically Plausible Neural Network Models

Empirical forgetting curve data have been shown to follow a power function. In contrast, many connectionist models predict either an exponential decay or flat forgetting curves. This paper simulates power function forgetting curves in a Hopfield network modified to incorporate the more biologically realistic assumptions of bounded weights and a distribution of learning rates. The modified model produces power function forgetting curves. The bounded weights introduce exponential decay for individual weights, and a power function forgetting curve when summing exponential decays with different learning rates. Because these assumptions are biologically reasonable, power function forgetting curves may be an emergent property of biological networks. The results fit empirical data and indicate that forgetting curves restrict possible implementation of models of memory.     

A model for stochastic drift in memory strength to account for judgments of learning

Previous research has shown that judgments of learning (JOLs) made immediately after encoding have a low correlation with actual cued-recall performance, whereas the correlation is high for delayed judgments. In this article, the authors propose a formal theory describing the stochastic drift of memory strength over the retention interval to account for the delayed-JOL effect. This is done by first decomposing the aggregated memory strength into exponential functions with slow and fast memory traces. The mean aggregated memory strength shows power-function forgetting curves. The drift of the memory strength is large for immediate JOLs (causing a low predictability) and weak for delayed JOLs (causing a high predictability). Consistent with empirical data, the model makes a novel prediction of JOL asymmetry, or that immediate weak JOLs are more predictive of future performance than are immediate strong JOLs. The JOL distributions for immediate and delayed JOLs are also accounted for.     

Connectionist models of recognition memory: constraints imposed by learning and forgetting functions

Multilayer connectionist models of memory based on the encoder model using the backpropagation learning rule are evaluated. The models are applied to standard recognition memory procedures in which items are studied sequentially and then tested for retention. Sequential learning in these models leads to 2 major problems. First, well-learned information is forgotten rapidly as new information is learned. Second, discrimination between studied items and new items either decreases or is nonmonotonic as a function of learning. To address these problems, manipulations of the network within the multilayer model and several variants of the multilayer model were examined, including a model with prelearned memory and a context model, but none solved the problems. The problems discussed provide limitations on connectionist models applied to human memory and in tasks where information to be learned is not all available during learning.     

A power-law model of psychological memory strength in short- and long-term recognition

A classic law of cognition is that forgetting curves are closely approximated by power functions. This law describes relations between different empirical dependent variables and the retention interval, and the precise form of the functional relation depends on the scale used to measure each variable. In the research reported here, we conducted a recognition task involving both short- and long-term probes. We discovered that formal memory-strength parameters from an exemplar-recognition model closely followed a power function of the lag between studied items and a test probe. The model accounted for rich sets of response time (RT) data at both individual-subject and individual-lag levels. Because memory strengths were derived from model fits to choices and RTs from individual trials, the psychological power law was independent of the scale used to summarize the forgetting functions. Alternative models that assumed different functional relations or posited a separate fixed-strength working memory store fared considerably worse than the power-law model did in predicting the data.     

Neuropsychological dissociations between priming and recognition: a single-system connectionist account

A key claim of current theoretical analyses of the memory impairments associated with amnesia is that certain distinct forms of learning and memory are spared. A compelling example is that amnesic patients and controls are indistinguishable in repetition priming but amnesic patients are impaired at recognizing the study items. The authors show that this pattern of results is predicted by a single-system connectionist model of learning in which amnesia is simulated by a reduced learning rate. They also demonstrate that the model can reproduce the converse pattern in which priming but not recognition is impaired if the input is assumed to be additionally degraded in a priming test. The authors conclude that dissociations between priming and recognition do not require functionally or neurally distinct memory systems.     

The power law as an emergent property

Recent work has shown that the power function, a ubiquitous characteristic of learning, memory, and sensation, can emerge from the arithmetic averaging of exponential curves. In the present study, the forgetting process was simulated via computer to determine whether power curves can result from the averaging of other types of component curves. Each of several simulations contained 100 memory traces that were made to decay at different rates. The resulting component curves were then arithmetically averaged to produce an aggregate curve for each simulation. The simulations varied with respect to the forms of the component curves: exponential, range-limited linear, range-limited logarithmic, or power. The goodness of the aggregate curve's fit to a power function relative to other functions increased as the amount of intercomponent slope variability increased, irrespective of component-curve type. Thus, the power law's ubiquity may reflect the pervasiveness of slope variability across component functions. Moreover, power-curve emergence may constitute a methodological artifact, an explanatory construct, or both, depending on the locus of the effect.     

Item and Associative Interactions in Short-term Memory: Multiple Memory Systems?

Three interactions of item and associative information are discussed: differential forgetting (item recognition declines over lag ranges where pair recognition does not; Hockley, 1991); differential emphasis (instructions to attend to items degrade later pair recognition but instructions to attend to pairs does not degrade later item recognition; Hockley & Cristi, 1996a), and multiple presentations of pairs or items from pairs has little effect on judgements of frequency of items and pairs (Hockley & Cristi, 1996b). Taken as a whole, these results seem to be inconsistent with a separate memory-systems view of memory. TODAM, a global-matching model, has always assumed a single memory system for the storage and retrieval of item and associative information and, with the addition of context and mediators, it can generate the whole pattern of data.     

Comparing decay rates for accurate and false memories in the DRM paradigm

Although previous studies have consistently reported different forgetting rates for true and false memory when tested with recall, studies comparing the rates of decay for true and false recognition have reported inconsistent results. The present study attempted to clarify this inconsistency by comparing forgetting rates for true and false recognition in addition to addressing methodological differences among the previous studies. Recognition of list items and critical lures was assessed at six delays (2, 5, 10, 15, and 20 min, and 2 days). A comparison of power function slopes across item type provided evidence of slope differences such that false recognition decayed more quickly than true recognition, which is inconsistent with predictions from activation/ source-monitoring and fuzzy tracetheories.     

Dissociating familiarity from recollection in human recognition memory: Different rates of forgetting over short retention intervals

Two functionally distinct forms of recognition memory have been identified in human and nonhuman species—the ability to recollect qualitative information about previous events, and the ability to differentiate between familiar and novel stimuli. Separate dual-process theories have been developed in the animal and human literatures to account for these findings. However, it is not clear whether these theories describe the same two underlying memory processes. On the basis of animal studies of medial temporal lobe function, familiarity is expected to exhibit disproportionately fast forgetting compared with recollection over short retention intervals. We tested this prediction in healthy human subjects by examining recognition forgetting rates across a range of 8–32 intervening items and found significant forgetting in the accuracy of familiarity-based discriminations and no evidence of forgetting in the accuracy of recollection-based discriminations. In agreement with the results from nonhuman species, the present results indicate that item familiarity decreases more rapidly than recollection over short retention intervals.     

材料性质对有意遗忘影响的实验研究

以大学生为被试 ,采用字表方式 ,探讨了词语和图片两种材料下的有意遗忘。结果发现 :①被试在词语和图片两种材料下的有意遗忘没有显著差异 ;②对于词语材料 ,被试既对指示记忆的项目进行了更加精细的编码 ,又对指示遗忘的项目产生了抑制 ;③对于图片材料 ,被试只对指示遗忘的项目产生了提取抑制 ,而没有对指示记忆的项目进行有意复述。     

On the differential nature of implicit and explicit memory

In this article, we report two experiments that provide further evidence concerning the differential nature of implicit and explicit memory. In Experiment 1, subjects first undertook a sentenceverification task. While carrying out this task, half of the subjects were also required to carry out a secondary processing task involving tone monitoring. Twenty-four hours later, the subjects’ memory for target items in the sentence-verification task was tested explicitly by means of a recognition task and implicitly by examining the extent to which the items primed fragment completion. Recognition performance was significantly impaired by the imposition of secondary processing demands during the original learning phase. In contrast, fragment completion was completely unaffected by this additional processing, even though substantial priming was observed. In Experiment 2, we examined whether priming in fragment completion is influenced by the nature of repetition during initial learning. Subjects studied a list of target items that were each repeated twice. Halfthe items were repeatedimmediately (lag 0) and halfwere repeated after six intervening items (lag 6). Memory for the itemswas assessed by recognition and by priming in fragment completion. Recognition was affected by lag, with lag 6 items being recognized better than lag 0 items. However, although significant priming was obtained, the extent of this priming was uninfluenced by lag. These data indicate two additional dimensions along which implicit and explicit memory differ and, furthermore, they support recent conceptualizations of processing differences underlying these two forms of memory.     

Recognition Practice Results in a Generalizable Skill in Older Adults: Decreased Intrusion Errors to Novel Objects Belonging to Practiced Categories

Accessing memories is often accompanied by both positive and negative consequences. For example, practice recognizing some visual images held in memory can improve memory for the practiced images and hurt memory for related images (i.e., recognition‐induced forgetting). However, visual stimuli have been shown to improve memory for older adults by decreasing false memories. This suggests that older adults may be immune to recognition‐induced forgetting and that recognition practice may decrease susceptibility to intrusion errors. We first tested the hypothesis that older adults are immune to recognition‐induced forgetting. We found older adults exhibit recognition‐induced forgetting. Next, we tested the hypothesis that recognition practice decreases older adult's rates of intrusion errors. We found lower intrusion errors for novel objects from practiced categories. This represents a generalizable learning effect; practice recognizing a target object (e.g., your pill bottle) improves the rejection of new lures (e.g., identifying the pill bottle that is not yours).Copyright © 2016 John Wiley & Sons, Ltd.     

The role of spatial location in remembering and forgetting peripheral words.

This study examined whether spatial location mediates intentional forgetting of peripherally presented words. Using an item-method directed forgetting paradigm, words were presented in peripheral locations at study. A recognition test presented all words at either the same or a different location relative to study. Results showed that while recognition of Remember words was unaffected by test location, when Forget words were presented in the same location at test as at study, recognition accuracy was significantly greater than when presented in a different location. Experiment 2 showed that the speed to localize a previously studied word was faster when it was presented in the same rather than a different study-test location but that the magnitude of this spatial priming was unaffected by memory instruction. We suggest that the location of peripherally presented words is represented in memory and can aid the retrieval of poorly encoded words.     

情绪对连续事件定向遗忘的影响

本研究采用项目法定向遗忘范式以中性和负性连续事件的视频为实验材料, 通过2个实验考察了情绪对连续事件定向遗忘的影响, 并进一步探讨了情绪对细节记忆和要义记忆的定向遗忘的影响。结果发现, 情绪可以消除细节记忆的定向遗忘效应, 而仅有要义记忆的定向遗忘效应则主要受到事件连续性而非情绪的影响。     

Forgetting rates in modality memory for young, mid-life, and older women

A mixed-modality (visual and auditory) continuous recognition task, followed immediately by a final recognition test, was administered to young (18-23 years), mid-life (38-50 years), and older (60-74 years) women. Subjects gave recognition responses for both the words and their presentation modality. Although older adults remembered less information about input mode than did the two younger groups, the age decrement was not the result of faster forgetting of such information by the elderly. When a ceiling effect at the initial lag was taken into account, forgetting rates for both words and input mode were comparable across the adult life span.     

Examining the lag effect under incidental encoding: Contributions of semantic priming and reminding

Memory is better when repeated learning events are spaced than when they are massed (spacing effect), as well as when material is processed semantically than when it is processed graphemically (levels-of-processing effect). Examination of the relationship between levels of processing and spacing for both deeply and shallowly encoded items has shown a spacing effect for items processed deeply, but not shallowly. A semantic priming account of spacing was proposed to explain the interaction between levels of processing and spacing on memory. The current study manipulated levels of processing and the amount of spacing (lag) that occurred between repetitions of items that were incidentally encoded. Results from Experiments 1A and 1B revealed lag effects in test performance when items were deeply and shallowly encoded. Although these findings are inconsistent with a semantic priming account, they can be interpreted within a reminding account, which is explored in Experiment 2. Results from the second experiment indicate that bringing reminding under conscious control benefited items that were presented at a long lag but not at a shorter lag. Together, this study provides evidence that is difficult to accommodate with a semantic priming account of spacing and instead provides additional support for a reminding account suggesting that automatic and controlled processes may both underlie the reminding process.     

Effects of retention intervals on receiver operating characteristics in artificial grammar learning

Current theories of memory suggest that recognition is composed of separate processes of familiarity and recollection (e.g. [Yonelinas, A. P. (2002). The nature of recollection and familiarity: a review of 30 years of research. Journal of Memory and Language, 46, 441-517]). A key feature of these two processes is that they decay, or are forgotten at different rates. The dual-process model has also been useful in understanding artificial grammar learning. We obtained evidence for recollection and familiarity in artificial grammar learning by analyses of receiver operating characteristics (ROC). Furthermore we found that these were dissociated by retention intervals of 14 days. The slope of the zROC curves deviated reliably from 1 immediately after study and increased towards 1 suggesting that recollection contributed to recognition decisions but declined over the 14-day period leaving familiarity as the only basis for recognition. These data show similar patterns to those observed in word-recognition [Gardiner, J. M., & Java, R. I. (1991). Forgetting in recognition memory with and without recollective experience. Memory &Cognition, 19, 617-623; Tunney, R. J. (submitted for publication). Changes in the subjective experience of recognition over time suggest independent processes] and confirm the view that recollection and familiarity are implicated in artificial grammar learning. Moreover, the data confirm the finding that recollection and familiarity-based memory show different patterns of forgetting.     

Buffered forgetting: When targets and distractors are both forgotten

In three experiments, we investigated prior findings that, following some memory tasks, essentially flat d' or forced-choice retention curves are produced. These curves have been interpreted as indicating that forgetting is not present over the intervals examined; however, we propose in this article that forgetting is actually present whenever hit rates and false alarm rates are both declining, despite the result being a flat retention curve. We demonstrate that such curves can be produced using a pair recognition procedure, a plurality discrimination task, and a verbal discrimination task. For all of these tasks, we provide either new evidence or refer to evidence already in the literature that tends to contradict alternative explanations. Then we show how the failure to consider both signal strength and noise has led to distortions in theoretical thinking about forgetting.     

自我参照对情绪性记忆定向遗忘的影响

关于自我参照效应的研究大多从信息编码角度考察自我参照的加工优势, 但作为记忆过程的另一重要方面, 自我参照对遗忘影响的研究较少。遗忘自身经历的情绪性记忆是日常生活中的常见现象。研究采用项目法定向遗忘范式(item-method directed forgetting)考察了自我参照对情绪性记忆定向遗忘的影响。实验一和实验二的结果都表明参照方式会影响定向遗忘：自我参照条件下的材料会发生定向遗忘; 而他人参照条件下的材料不会发生定向遗忘, 两种参照方式下材料间区辨性的不同是导致这一现象的原因。此外, 实验二采用回忆测试发现：自我参照对不同效价的情绪性记忆定向遗忘产生了不同影响, 而自我提升(self-enhancement)动机在其中起着重要的调节作用。