Storage and retrieval of serial-order information

In this paper I will first review some seminal work by Conrad on the storage and retrieval of serial-order information which is still very relevant today. Then I will discuss the TODAM (theory of distributed associative memory) approach to serial-order effects. I will compare the three TODAM serial-order models (the chaining model, the chunking model and the power-set model; Murdock, 1995) but concentrate on the power set model. Its original problems can be solved, but a revised and augmented version has some new problems. This paper is more of a progress report than a finished product, so the reader should be prepared to follow the twists and turns of the argument.     

Storage and retrieval of serial‐order information

In this paper I will first review some seminal work by Conrad on the storage and retrieval of serial‐order information which is still very relevant today. Then I will discuss the TODAM (theory of distributed associative memory) approach to serial‐order effects. I will compare the three TODAM serial‐order models (the chaining model, the chunking model and the power‐set model; Murdock, 1995) but concentrate on the power set model. Its original problems can be solved, but a revised and augmented version has some new problems. This paper is more of a progress report than a finished product, so the reader should be prepared to follow the twists and turns of the argument.     

Judgments of Frequency and Recency in a Distributed Memory Model

TODAM is a theory of distributed associative memory based on the convolution-correlation formalism of A. Borsellino and T. Poggio (1973, Kybernetik, 122, 113-122), and TODAM2 is a revised version which includes context, auto-associations for binding, a dual basis for item information, and mediators for associative information. It can explain some complex interactions (differential forgetting and differential attention) between item and associative information. In this paper we derive the basic expressions for the memory-probe dot product for the 2x2 cases (item and pair study crossed with item and pair probe), and then we apply these expressions to judgments of frequency (JOF) and judgments of recency (JOR). We report an experiment which tests both JOF and JOR for single items with post-cuing to control for encoding strategies and suggest an attenuation factor for repetition to improve the fits. With attenuation, TODAM2 can fit the JOR data, but the JOF fits, while not too bad, consistently predict too much dependence between item and associative information. Copyright 2001 Academic Press.     

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.     

Associative isolation: Unifying associative and list memory

Rather than treating paired associate and serial learning as involving the acquisition of distinct types of information [e.g. Murdock (1974). Human memory: Theory and data. New York: Wiley] I propose an Isolation Principle which treats the two as ends of a continuum. According to this principle, consecutive pairs of items are relatively isolated from other studied items in paired associates learning, but not isolated in serial list learning. The consequence is that variability that dominates forward and backward probed recall is highly correlated in pairs but less so, due to differential interference, in lists. This can explain an important dissociation: whereas forward and backward probes of pairs are nearly perfectly correlated, the correlation is only moderate for serial lists. I demonstrate this in a chaining model by varying item-to-item associative strengths and in a positional coding model by varying the similarity structure of item positions. This enables a range of models to account for data on pairs and lists, as well as potential intermediate or hybrid paradigms, within a single theoretical framework.     

Unchained Memory: Error Patterns Rule out Chaining Models of Immediate Serial Recall

Many models of serial recall assume a chaining mechanism whereby each item associatively evokes the next in sequence. Chaining predicts that, when sequences comprise alternating confusable and non-confusable items, confusable items should increase the probability of errors in recall of following non-confusable items. Two experiments using visual presentation and one using vocalized presentation test this prediction and demonstrate that: (1) more errors occur in recall of confusable than alternated non-confusable items, revealing a 'sawtooth' in serial position curves; (2) the presence of confusable items often has no influence on recall of the non-confusable items; and (3) the confusability of items does not affect the type of errors that follow them. These results are inconsistent with the chaining hypothesis. Further analysis of errors shows that most transpositions occur over short distances (the locality constraint), confusable items tend to interchange (the similarity constraint), and repeated responses are rare and far apart (the repetition constraint). The complete pattern of errors presents problems for most current models of serial recall, whether or not they employ chaining. An alternative model is described that is consistent with these constraints and that simulates the detailed pattern of errors observed.     

Time course of item and associative information: implications for global memory models

The time course of availability of associative and item information was examined by using a response signal procedure. Associative information discriminates between a studied pair of words and a pair with words from two different studied pairs. Item information is sufficient to discriminate between a studied pair and a pair not studied. In two experiments, discriminations that require associative information are delayed relative to those based on item information. Two additional experiments discount alternative explanations in terms of the time to encode the test items or task strategies. Examination of the global memory models of Gillund and Shiffrin (1984), Hintzman (1988), and Murdock (1982) shows that the models treat item and associative information inseparably. Modifications to these models which can produce separate contributions for item and associative information do not predict any difference in their availability. Two possible mechanisms for the delayed availability of associative information are considered: the involvement of recall in recognition and the time required to form a compound cue.     

Item and associative recognition with precuing and postcuing

M. Duncan & B. B. Murdock (2000) compared precued and postcued item recognition and serial recall showing precued-postcued differences for item recognition but not for serial recall. Precuing and postcuing refer to 2 conditions in which the instructions as to the type of recall test following the presentation of short lists of items is given before or after the list presentation. This methodology was extended here to a paired-associate task. In 2 experiments, short lists of paired associates were presented followed by single-item, old-new, or intact-rearranged pair recognition tests; test type was precued or postcued. A fast or slow presentation rate was used to discourage or encourage mediators. TODAM2 (a theory of distributed associative memory) predicts that there should be little or no cuing differences regardless of whether subjects use mediators to remember the pairs. As predicted the recognition data were essentially identical for the precued and postcued conditions.     

Positional and temporal clustering in serial order memory

The well-known finding that responses in serial recall tend to be clustered around the position of the target item has bolstered positional-coding theories of serial order memory. In the present study, we show that this effect is confounded with another well-known finding—that responses in serial recall tend to also be clustered around the position of the prior recall (temporal clustering). The confound can be alleviated by conditioning each analysis on the positional accuracy of the previously recalled item. The revised analyses show that temporal clustering is much more prevalent in serial recall than is positional clustering. A simple associative chaining model with asymmetric neighboring, remote associations, and a primacy gradient can account for these effects. Using the same parameter values, the model produces reasonable serial position curves and captures the changes in item and order information across study-test trials. In contrast, a prominent positional coding model cannot account for the pattern of clustering uncovered by the new analyses.     

Associative and serial-order information: Different modes of operation?

Two versions of a distributed-memory model, one for associative information and one for serial-order information, leave open the question of whether there are two different modes of operation or only one. An analysis of these two versions showed that differential predictions emerge when one considers the role of item information. To test these predictions, a standard item-recognition task (Sternberg, 1966) was embedded in a paired-associate probe-recognition task, a serial-order probe-recognition task, or both. The item-recognition serial-position curves were quite different in the paired-associate and the serial-order conditions, and followed the paired-associate condition when the two were combined. In general, the results were consistent with the predictions of the distributed-memory models, and supported the idea of different modes of operation rather than a single common mode of operation.     

Order information and retrieval distinctiveness: recall of common versus bizarre material

The order-encoding hypothesis (E. L. DeLosh & M. A. McDaniel, 1996) assumes that serial-order information contributes to the retrieval of list items and that serial-order encoding is better for common items than bizarre items. In line with this account, Experiment 1 revealed better free recall and serial-order memory for common than for bizarre items in pure lists, and Experiment 2 showed that recall for bizarre items increased and the recall advantage of common items was eliminated when serial-order encoding for bizarre items was increased to the level of common items. However, inconsistent with a second assumption that bizarre-item advantages in mixed lists reflect better individual-item encoding for bizarre items, Experiments 3 and 4 showed that the bizarreness effect in mixed lists is eliminated when alternative retrieval strategies are encouraged. This set of findings is better explained by the differential-retrieval-process framework, which proposes that contextual factors (e.g., list composition) influence the extent to which various types of information are used at retrieval, with the bizarreness advantage in mixed lists dependent on a distinctiveness-based retrieval process.     

Serial recall of colors: Two models of memory for serial order applied to continuous visual stimuli

This study investigated the effects of serial position and temporal distinctiveness on serial recall of simple visual stimuli. Participants observed lists of five colors presented at varying, unpredictably ordered interitem intervals, and their task was to reproduce the colors in their order of presentation by selecting colors on a continuous-response scale. To control for the possibility of verbal labeling, articulatory suppression was required in one of two experimental sessions. The predictions were derived through simulation from two computational models of serial recall: SIMPLE represents the class of temporal-distinctiveness models, whereas SOB-CS represents event-based models. According to temporal-distinctiveness models, items that are temporally isolated within a list are recalled more accurately than items that are temporally crowded. In contrast, event-based models assume that the time intervals between items do not affect recall performance per se, although free time following an item can improve memory for that item because of extended time for the encoding. The experimental and the simulated data were fit to an interference measurement model to measure the tendency to confuse items with other items nearby on the list—the locality constraint—in people as well as in the models. The continuous-reproduction performance showed a pronounced primacy effect with no recency, as well as some evidence for transpositions obeying the locality constraint. Though not entirely conclusive, this evidence favors event-based models over a role for temporal distinctiveness. There was also a strong detrimental effect of articulatory suppression, suggesting that verbal codes can be used to support serial-order memory of simple visual stimuli.     

The recall of missing items

In two experiments, we studied the recall of missing items. Short lists of common words were presented once and were followed immediately by a random permutation of all but one of the presented items. The task of the subject was to recall the missing item--that is, the item present in the study set but missing from the probe set. Experiment 1 replicated the high accuracy with five-item lists originally reported by Yntema and Trask (1963) and showed that the latencies were quite short (about 750 msec). Experiment 2 varied list length unpredictably and showed that accuracy was a function of both list length (four, five, or six items) and serial position. Latency was again quite short but was essentially independent of list length and serial position. It was possible to simulate most of the effects with the power set model with no free parameters (i.e., parameters that varied with the experimental manipulations). The results seemed to be more consistent with a direct access model (the power set model of TODAM; Murdock, 1995) than with a simple search or serial-scanning model.     

Encoding location and serial order in auditory working memory: evidence for separable processes

In this study, we investigated the interactions between temporal and spatial information in auditory working memory. In two experiments, participants were presented with sequences of sounds originating from different locations in space and were then asked to recall either their position or their serial order. In Experiment 1, attention during encoding was manipulated by contrasting 'pure' blocks (i.e., location-only or serial-order-only trials) to 'mixed' blocks (i.e., different percentages of spatial and serial-order trials). In Experiment 2, 'pure' blocks were contrasted to blocks in which spatial and serial-order trials were intermixed with a third task requiring a semantic categorization of sounds. Results from both experiments showed that, whereas serial-order recall is linearly affected by the simultaneous encoding of a concurrent feature, the recall of position is mostly unaffected by concurrent feature encoding. Contrastingly, overall performance level was lower for spatial recall than serial recall. We concluded that serial order and location of items appear to be independently encoded in auditory working memory. Serial order is easier to recall, but strongly affected by the processing of concurrent item dimensions, while item location is more difficult to recall, but relatively automatic, as shown by its strong resistance to interfering dimensions in encoding.     

Sequential dependencies in recall of sequences: Filling in the blanks

Sequential dependencies can provide valuable information about the processes supporting memory, particularly memory for serial order. Earlier analyses have suggested that anticipation errors—reporting items ahead of their correct position in the sequence—tend to be followed by recall of the displaced item, consistent with primacy gradient models of serial recall. However, a more recent analysis instead suggests that anticipation errors are followed by further anticipation errors, consistent with chaining models. We report analyses of 21 conditions from published serial recall data sets, in which we observed a systematic pattern whereby anticipations tended to be followed by the “filling in” of displaced items. We note that cases where a different pattern held tended to apply to recall of longer lists under serial learning conditions or to conditions where participants were free to skip over items. Although the different patterns that can be observed might imply a dissociation (e.g., between short- and long-term memory), we show that these different patterns are naturally predicted by Farrell’s (Psychological Review 119:223–271, 2012) model of short-term and episodic memory and relate to whether or not spontaneously formed groups of items can be skipped over during recall.     

Theoretical correlations and measured correlations: relating recognition and recall in four distributed memory models

This article addresses the relation between item recognition and associative (cued) recall. Going beyond measures of performance on each task, the analysis focuses on the degree to which the contingency between successful recognition and successful recall of a studied item reflects the commonality of memory processes underlying the recognition and recall tasks. Specifically, 4 classes of distributed memory models are assessed for their ability to account for the relatively invariant correlation (approximately .5) between successive recognition and recall. Basic versions of each model either under- or overpredict the intertask correlation. Introducing variability in goodness-of-encoding and response criteria, as well as output encoding, enabled all 4 models to reproduce the moderate intertask correlation and the increase in correlation observed in 2 mixed-list experiments. This model-based analysis provides a general theoretical framework for interpreting contingencies between successive memory tests.     

Deciding about decision models of remember and know judgments: a reply to Murdock (2006)

B. B. Murdock (2006) has interpreted remember-know data within a decision space defined by item and associative information, the fundamental variables in his general recognition memory model TODAM (B. B. Murdock, 1982). He has related parameters of this extended model to stimulus characteristics for several classic remember-know data sets. The authors show that this accomplishment is shared by both one- and two-dimensional signal-detection-based remember-know models (J. C. Dunn, 2004; C. M. Rotello, N. A. Macmillan, & J. A. Reeder, 2004; J. T. Wixted & V. Stretch, 2004), which can be represented in this same decision space and can be related to stimulus characteristics with similar success. Murdock claims that his model, unlike its competitors, is a process model; however, the process aspects of TODAM are not used in his application, and the decision aspects are identical to a previously proposed model. Murdock's claim that one competing model (STREAK; C. M. Rotello et al., 2004) is not fully specified is shown to be false. The new model is not superior to existing ones, but comparisons among the models to date are not definitive. The authors describe several strategies that might be applied to distinguish among them.     

Associative retrieval processes in free recall

I present a new method for analyzing associative processes in free recall. While previous research has emphasized the prominence of semantic organization, the present method illustrates the importance of association by contiguity. This is done by examining conditional response probabilities in the output sequence. For a given item recalled, I examine the probability and latency that it follows an item from a nearby or distant input position. These conditional probabilities and latencies, plotted as a function of the lag between studied items, reveal several regularities about output order in free recall. First, subjects tend to recall items more often and more rapidly from adjacent input positions than from remote input positions. Second, subjects are about twice as likely to recall adjacent pairs in the forward than in the backward direction and are significantly faster in doing so. These effects are observed at all positions in the output sequence. The asymmetry effect is theoretically significant because, in cued recall, nearly symmetric retrieval is found at all serial positions (Kahana, 1995; Murdock, 1962). An attempt is made to fit the search of associative memory model (Raaijmakers & Shiffrin, 1980, 1981) with and without symmetric interitem associations to these data. Other models of free recall are also discussed.     

The revelation effect for item and associative recognition: Familiarity versus recollection

The revelation effect occurs when items on a recognition test are more likely to be judged as being old if they are preceded by a cognitive task that involves the processing of similar types of stimuli. This effect was examined for item (single-word) and associative (word-pair) recognition. We found, in Experiments 1 and 2, a revelation effect for item, but not for associative recognition under normal study conditions. A revelation effect for both item and associative recognition was observed in Experiments 3 and 4 when study time was extremely brief, thus limiting the encoding of information that would support recall or recollection. In Experiment 5, we demonstrated that the revelation effect for item recognition is eliminated when item recognition decisions are made in the context of a study item. The results show that the revelation task influenced recognition decisions based on familiarity, but not decisions that involved recall or recollection.     

Recall cues interfere with retrieval from visuospatial working memory

Visuospatial working memory allows us to hold multiple visual objects over short delays. It is typically tested by presenting an array of objects, then after a delay showing a ‘probe’ indicating which memory item to recall or reproduce by adjusting a target feature. However, recent studies demonstrate that information at the time of probe can disrupt recall. Here, in three experiments we test whether traditional memory probes, which contain features that compete with the feature to be recalled, may themselves interfere with performance. We asked participants to report the direction of one of the several coloured arrows in memory, based on its colour. First, we demonstrate that recall is better when the probe is initially just a coloured dot, rather than a coloured arrow which has to be adjusted to match orientation memory, consistent with interference from features of the probe itself. Second, this interference is present even when a mask follows the memory array, suggesting that the interference does not work by degrading immediate or iconic memory. Finally, when items are shown sequentially, the first and last items are invulnerable to probe interference. Our findings support recent theories of associative recall, in which probes reactivate features in WM, retrieving information by pattern completion.