Are there multiple memory systems? Tests of models of implicit and explicit memory

This article reviews recent work aimed at developing a new framework, based on signal detection theory, for understanding the relationship between explicit (e.g., recognition) and implicit (e.g., priming) memory. Within this framework, different assumptions about sources of memorial evidence can be framed. Application to experimental results provides robust evidence for a single-system model in preference to multiple-systems models. This evidence comes from several sources including studies of the effects of amnesia and ageing on explicit and implicit memory. The framework allows a range of concepts in current memory research, such as familiarity, recollection, fluency, and source memory, to be linked to implicit memory. More generally, this work emphasizes the value of modern computational modelling techniques in the study of learning and memory.     

Are there multiple memory systems? Tests of models of implicit and explicit memory

[Formula: see text] This article reviews recent work aimed at developing a new framework, based on signal detection theory, for understanding the relationship between explicit (e.g., recognition) and implicit (e.g., priming) memory. Within this framework, different assumptions about sources of memorial evidence can be framed. Application to experimental results provides robust evidence for a single-system model in preference to multiple-systems models. This evidence comes from several sources including studies of the effects of amnesia and ageing on explicit and implicit memory. The framework allows a range of concepts in current memory research, such as familiarity, recollection, fluency, and source memory, to be linked to implicit memory. More generally, this work emphasizes the value of modern computational modelling techniques in the study of learning and memory.     

Reflections on working memory: are the two models complementary?

By means of the theoretical modeling of data from Kemps, De Rammelaere, and Desmet (2000, this issue), the working memory theory of Baddeley and the theory of constructive operators of Pascual-Leone are contrasted and compared. It is concluded that although the theory of constructive operators is complementary with working memory theory (for it explains developmental and individual differences that working memory theory cannot explain), the converse is not true; the theory of constructive operators explains all the data without need of working memory theory.     

“Plateau”-related summary statistics are uninformative for comparing working memory models

Performance on visual working memory tasks decreases as more items need to be remembered. Over the past decade, a debate has unfolded between proponents of slot models and slotless models of this phenomenon (Ma, Husain, Bays (Nature Neuroscience 17, 347-356, 2014). Zhang and Luck (Nature 453, (7192), 233-235, 2008) and Anderson, Vogel, and Awh (Attention, Perception, Psychophys 74, (5), 891-910, 2011) noticed that as more items need to be remembered, “memory noise” seems to first increase and then reach a “stable plateau.” They argued that three summary statistics characterizing this plateau are consistent with slot models, but not with slotless models. Here, we assess the validity of their methods. We generated synthetic data both from a leading slot model and from a recent slotless model and quantified model evidence using log Bayes factors. We found that the summary statistics provided at most 0.15 % of the expected model evidence in the raw data. In a model recovery analysis, a total of more than a million trials were required to achieve 99 % correct recovery when models were compared on the basis of summary statistics, whereas fewer than 1,000 trials were sufficient when raw data were used. Therefore, at realistic numbers of trials, plateau-related summary statistics are highly unreliable for model comparison. Applying the same analyses to subject data from Anderson et al. (Attention, Perception, Psychophys 74, (5), 891-910, 2011), we found that the evidence in the summary statistics was at most 0.12 % of the evidence in the raw data and far too weak to warrant any conclusions. The evidence in the raw data, in fact, strongly favored the slotless model. These findings call into question claims about working memory that are based on summary statistics.     

Latent change models of adult cognition: are changes in processing speed and working memory associated with changes in episodic memory?

The authors used 6-year longitudinal data from the Victoria Longitudinal Study (VLS) to investigate individual differences in amount of episodic memory change. Latent change models revealed reliable individual differences in cognitive change. Changes in episodic memory were significantly correlated with changes in other cognitive variables, including speed and working memory. A structural equation model for the latent change scores showed that changes in speed and working memory predicted changes in episodic memory, as expected by processing resource theory. However, these effects were best modeled as being mediated by changes in induction and fact retrieval. Dissociations were detected between cross-sectional ability correlations and longitudinal changes. Shuffling the tasks used to define the Working Memory latent variable altered patterns of change correlations.     

Is external memory memory? Biological memory and extended mind

Clark and Chalmers (1998) claim that an external resource satisfying the following criteria counts as a memory: (1) the agent has constant access to the resource; (2) the information in the resource is directly available; (3) retrieved information is automatically endorsed; (4) information is stored as a consequence of past endorsement. Research on forgetting and metamemory shows that most of these criteria are not satisfied by biological memory, so they are inadequate. More psychologically realistic criteria generate a similar classification of standard putative external memories, but the criteria still do not capture the function of memory. An adequate account of memory function, compatible with its evolution and its roles in prospection and imagination, suggests that external memory performs a function not performed by biological memory systems. External memory is thus not memory. This has implications for: extended mind theorizing, ecological validity of memory research, the causal theory of memory.     

Effect of articulatory suppression on task‐switching performance: Implications for models of working memory

In a series of experiments, we examine some effects of articulatory suppression in task switching. The results from Experiments 1a and 2a showed that switch costs in the articulatory suppression condition were larger than those in the control and tapping conditions when the switching cues were not provided. On the other hand, articulatory suppression did not have any effect on switch costs in Experiments 1b and 2b, where the switching cues were provided. In Experiment 3, using a computer‐assisted experimentation, this pattern of data was replicated in a two‐factor design with articulatory suppression and switching cues factors. The results indicate that a specific component in working memory, the phonological loop, might contribute to the performance in task switching, at least in situations where the external task cues were not available. The data reported here suggest that the phonological loop plays an important role in one of the executive control processes, and challenge the traditional idea that the slave systems are simply governed by the central executive in the working memory.     

Development of working memory: should the Pascual-Leone and the Baddeley and Hitch models be merged?

The data presented by Kemps, De Rammelaere, and Desmet (2000, this issue) appear to have some aspects that fit most readily into our own model (Baddeley & Hitch, 1974), while others appear to support that of Pascual-Leone (1970). We accept that our initial model said little about development and was better able to account for relatively simple memory-based tasks than more complex cognitive activities. More recent elaborations of the model are, however, able to throw new light on the processes underlying cognitive development, offering a better account than that provided by existing neo-Piagetian interpretations. Meanwhile, the addition of a fourth component to the model, namely the episodic buffer, offers a way of dealing with more complex cognitive activities. Given the major differences between our own model and that of Pascual-Leone in basic assumptions, and in theoretical style, we suggest that any attempt to combine the two would be premature.     

The conditional-recency dissociation is confounded with nominal recency: Should unitary models of memory still be devaluated?

The conditional-recency dissociation between immediate and delayed free recall FR; Farrell (Journal of Experimental Psychology: Learning, Memory, and Cognition, 36, 324–347, 2010) has critical implications regarding the prolonged debate between unitary and dual-store models of memory. In immediate FR, when the availability of items is controlled for, the recency of the final list item increases across the first few output positions. No such increase is found in delayed FR, with a trend in the opposite direction. This dissociation challenges temporal context TCM; Howard & Kahana (Journal of Mathematical Psychology, 46, 269–299, 2002) and distinctiveness SIMPLE; Brown, Neath, & Chater (Psychological Review, 114, 539–576, 2007) unitary models of memory and suggests the involvement of a short-term buffer in immediate FR. We show that this dissociation is confounded with the different magnitudes of nominal recency (i.e., the prevalence of the final list item) found in immediate as compared to delayed FR. By reshuffling output orders and comparing the empirical results to those of a null hypothesis of no output-order effect, we controlled for the greater prevalence of the final list item that has been observed in immediate FR. Under this control, we found no evidence for a dissociation in the tendency to recall the final list item across output positions. This finding suggests that the conditional-recency dissociation imposes no new constraint on unitary models of memory. More generally, we demonstrate how biases that influence measures of output-order tendencies (e.g., conditional recency) can be controlled for, thus yielding “purer” measures of these variables.     

Immediate memory for faces: Long- or short-term memory?

Immediate recognition memory span and short-term forgetting for non-verbal stimuli (“unfamiliar faces”) were investigated in normal subjects and amnesic patients. Surnames were used as a verbal control. It was found that normal subjects had a reliable immediate recognition span of one for faces and that there was no decrement in performance in the retention of three faces over a 30-s interval. Amnesic subjects were impaired on both the immediate memory span task and on the short-term forgetting task. This pattern of results contrasts with those obtained using surnames on which the results were comparable to other verbal S.T.M. tasks. It is suggested that performance on these tasks of recognition memory for faces is determined by L.T.M. processes. The implications for the interpretation of S.T.M. are discussed.     

Do explicit memory manipulations affect the memory blocking effect?

The memory blocking effect (MBE) occurs when people are prevented from completing word fragments because they studied orthographically similar words. Across 3 experiments, we investigated how manipulations that influence explicit memory tasks would influence the MBE. Although a significant MBE was observed in all 3 experiments, manipulating depth of processing (Experiment 1), time to complete the fragments (Experiment 2), and awareness of the MBE (Experiment 3) did not change the magnitude of the MBE. We discuss these results in the context of a suppression mechanism involved in retrieval-induced forgetting.     

Are memory self-efficacy and memory performance related? A meta-analysis

The association between memory self-efficacy (MSE) and memory performance is highly documented in the literature. However, previous studies have produced inconsistent results, and there is no consensus on the existence of a significant link between these two variables. In order to evaluate whether or not the effect size of the MSE-memory performance relationship in healthy adults is significant and to test several theory-driven moderators, we conducted a meta-analysis of published and unpublished studies. A random-effects model analysis of data from 107 relevant studies (673 effect sizes) indicated a low but significant weighted mean correlation between MSE and memory performance, r = .15, 95% CI [.13, .17]. In addition, the mean effect size was significantly moderated by the way MSE was assessed. Memory performance was more strongly related to concurrent MSE (perceived current ability to perform a given task) than it was to global MSE (perceived usual memory ability in general). Furthermore, we found marginally larger MSE-memory performance correlations when the memory situations used to assess MSE involved familiar stimuli. No effect of the method used to assess global MSE or domain MSE (memory rating vs. performance predictions) was found. The results also show that the resource demands of the memory tasks have a moderator effect, as the MSE-performance correlation is larger with free-recall and cued-recall tasks than it is with recognition tasks. Limitations (generalization issues, moderators not considered) and implications for future research are discussed.     

What happens when we relearn part of what we previously knew? Predictions and constraints for models of long-term memory

 Part-set relearning studies examine whether relearning a subset of previously learned items impairs or improves memory for other items in memory that are not relearned. Atkins and Murre have examined part-set relearning using multi-layer networks that learn by optimizing performance on a complete set of items. For this paper, four computer models that learn each item additively and separately were tested using the part-set relearning procedure (Hebbian network, CHARM, MINERVA 2, and SAM). Optimization models predict that part-set relearning should improve memory for items not relearned, while additive models make the opposite prediction. This distinction parallels the relative ability of these models to account for interference phenomena. Part-set relearning provides another source of evidence for choosing between optimization and additive models of long-term memory. A new study suggests that the predictions of the additive models are broadly supported. Received: 29 April 1999 / Accepted: 26 July 1999     

The co-witness misinformation effect: memory blends or memory compliance?

For 30 years hundreds of researchers have shown participants videos and slide-sequences of events, presented the participants with misleading information, and found that this misinformation distorted their memories. The purpose of this study was to establish whether those misled participants are reporting a memory blend of the two sources of information or whether they are simply complying with the post-event information. A total of 92 participants were shown one of two versions of six different videos, which included some subtle differences. After having watched each video individually, participants were paired with someone who had seen the other version and they discussed the clips together. They then individually answered questions about the videos, and their responses showed that some of the distorted memories were blends of the original information and the post-event information. The implications of these findings are discussed.     

Does memory affect judgment? Self-generated versus recall memory measures.

Most evidence regarding the independence of memory and judgment comes from studies that used memory measures consisting of Ss' recall of raw input data (recall measures). Such evidence provides the primary support for on-line judgment formation. The results of 2 experiments suggest that self-generated memory measures capture the contents of memory at the time of judgement more effectively than recall measures and, accordingly, are more likely to provide evidence that memory and judgment are related. When directly compared, a self-generated measure provided evidence of a memory-judgment relationship and a recall measure did not. Thus, memory-based judgment formation may be more prevalent than the on-line processing literature suggests.     

A model for recognition memory: REM—retrieving effectively from memory

A new model of recognition memory is reported. This model is placed within, and introduces, a more elaborate theory that is being developed to predict the phenomena of explicit and implicit, and episodic and generic, memory. The recognition model is applied to basic findings, including phenomena that pose problems for extant models: the list-strength effect (e.g., Ratcliff, Clark, & Shiffrin, 1990), the mirror effect (e.g., Glanzer & Adams, 1990), and the normal-ROC slope effect (e.g., Ratcliff, McKoon, & Tindall, 1994). The model assumes storage of separate episodic images for different words, each image consisting of a vector of feature values. Each image is an incomplete and error prone copy of the studied vector. For the simplest case, it is possible to calculate the probability that a test item is "old," and it is assumed that a default "old" response is given if this probability is greater than .5. It is demonstrated that this model and its more complete and realistic versions produce excellent qualitative predictions.     

How many pixels make a memory? Picture memory for small pictures

Torralba (Visual Neuroscience, 26, 123–131, 2009) showed that, if the resolution of images of scenes were reduced to the information present in very small “thumbnail images,” those scenes could still be recognized. The objects in those degraded scenes could be identified, even though it would be impossible to identify them if they were removed from the scene context. Can tiny and/or degraded scenes be remembered, or are they like brief presentations, identified but not remembered. We report that memory for tiny and degraded scenes parallels the recognizability of those scenes. You can remember a scene to approximately the degree to which you can classify it. Interestingly, there is a striking asymmetry in memory when scenes are not the same size on their initial appearance and subsequent test. Memory for a large, full-resolution stimulus can be tested with a small, degraded stimulus. However, memory for a small stimulus is not retrieved when it is tested with a large stimulus.     

Is overgeneral autobiographical memory an isolated memory phenomenon in major depression?

The present study explored the relation between overgeneral autobiographical memory (AM) and other aspects of memory functioning in depression. A total of 26 patients with major depressive disorder completed a set of memory tasks measuring AM specificity (AMT; Williams & Broadbent, 1986), working memory, semantic memory, verbal learning, delayed verbal recall, recognition memory, and source memory. Reduced specificity of AM was related to poor working memory (central executive functioning) and poor source memory. The former finding conforms to the idea that the voluntary retrieval of specific autobiographical memories (AMs) involves central executive processes (e.g., Conway & Pleydell-Pearce, 2000 Conway, M. A. and Pleydell-Pearce, C. W. 2000. The construction of autobiographical memories in the self-memory system. Psychological Review, 107: 261–288. [Crossref], [PubMed], [Web of Science ®] , [Google Scholar]). The latter finding replicates and extends recent findings suggesting that overgeneral AM is part of a broader memory deficit in retrieving the specific details of the context in which information was acquired (Ramponi, Barnard, & Nimmo-Smith, 2004 Raes , F. , Hermans , D. , Williams , J. M. G. , Demyttenaere , K. , Sabbe , B. , Pieters , G . & Eelen , P. ( 2004 ). Reduced specificity of autobiographical memory: A mediator between rumination and ineffective social problem solving in major depression? Journal of affective disorders , 87 , 331 – 335  [Google Scholar]). Furthermore, in line with Ramponi et al. (2004 Raes , F. , Hermans , D. , Williams , J. M. G. , Demyttenaere , K. , Sabbe , B. , Pieters , G . & Eelen , P. ( 2004 ). Reduced specificity of autobiographical memory: A mediator between rumination and ineffective social problem solving in major depression? Journal of affective disorders , 87 , 331 – 335  [Google Scholar]), rumination was found to be related to both overgeneral AM and poor source memory.     

Does overgeneral autobiographical memory result from poor memory for task instructions?

Considerable previous research has shown that retrieval of overgeneral autobiographical memories (OGM) is elevated among individuals suffering from various emotional disorders and those with a history of trauma. Although previous theories suggest that OGM serves the function of regulating acute negative affect, it is also possible that OGM results from difficulties in keeping the instruction set for the Autobiographical Memory Test (AMT) in working memory, or what has been coined "secondary goal neglect" (Dalgleish, 2004). The present study tested whether OGM is associated with poor memory for the task's instruction set, and whether an instruction set reminder would improve memory specificity over repeated trials. Multilevel modelling data-analytic techniques demonstrated a significant relationship between poor recall of instruction set and probability of retrieving OGMs. Providing an instruction set reminder for the AMT relative to a control task's instruction set improved memory specificity immediately afterward.