The principles of recency and contiguity are two cornerstones of the theoretical and empirical analysis of human memory. Recency has been alternatively explained by mechanisms of decay, displacement, and retroactive interference. Another account of recency is based on the idea of variable context (Estes, 1955; Mensink & Raaijmakers, 1989). Such notions are typically cast in terms of a randomly fluctuating population of elements reflective of subtle changes in the environment or in the subjects ’ mental state. This random context view has recently been incorporated into distributed and neural network memory models (Murdock, 1997; Murdock, Smith, & Bai, 2001). Here we propose an alternative model. Rather than being driven by random fluctuations, this formulation, the temporal context model (TCM), uses retrieval of prior contextual states to drive contextual drift. In TCM, retrieved context is an inherently asymmetric retrieval cue. This allows the model to provide a principled explanation of the widespread advantage for forward recalls in free and serial recall. Modeling data from single-trial free recall, we demonstrate that TCM can simultaneously explain recency and

this article may be addressed to either Michael Kahana or Robert Sekuler, Volen National Center for Complex Systems, MS 013, Brandeis University, Waltham, MA 02254-9110. E-mail may be sent to kahana @brandeis.edu or sekuler@brandeis.edu plex multidimensional stimulus spaces (Nosofsky, 1992; Maddox & Ashby, 1996; Ashby & Perrin, 1988), with decision rules that can predict performance in a variety of classification paradigms (Nosofsky & Palmeri, 1998; Nosofsky & Alfonso-Reese, 1999; Maddox & Ashby, 1996). Although models of classification and models of visual discrimination share many assumptions about stimulus representation and subjects' decision rules, models of classification have been primarily developed to explain subjects' classification of combinations of simple geometric forms, whereas models of discrimination have been developed to explain subjects ' discrimination of elemental visual stimuli, including sinusoidal luminance gratings. Because such stimuli can be combined to synthesize more complex images such as textures and natural scenes, they represent a natural test-bed for assessing theories' power and generalizability

Despite a century of research, the mechanisms underlying short-term or working memory for serial order remain uncertain. Recent theoretical models have converged on a particular account, based on transient associations between independent item and context representations. In the present article, the authors present an alternative model, according to which sequence information is encoded through sustained patterns of activation within a recurrent neural network architecture. As demonstrated through a series of computer simulations, the model provides a parsimonious account for numerous benchmark characteristics of immediate serial recall, including data that have been considered to preclude the application of recurrent neural networks in this domain. Unlike most competing accounts, the model deals naturally with findings concerning the role of background knowledge in serial recall and makes contact with relevant neuroscientific data. Furthermore, the model gives rise to numerous testable predictions that differentiate it from competing theories. Taken together, the results presented indicate that recurrent neural networks may offer a useful framework for understanding short-term memory for serial order.

studies have directly examined whether order of study itself influences retrieval efficacy. In contrast, many dozens of studies have examined this question in paired-associate learning, asking whether memory for simple pairs exhibits a forward asymmetry effect (i.e., better forward recall than backward recall). Surprisingly, such asymmetries are exceedingly hard to detect in pairedassociate tasks, with many studies producing nearly identical levels of forward and backward recall (see Ekstrand, The authors acknowledge support from National Institutes of Health grant MH55687. We are grateful to Kelly Addis for assisting in data collection and for helpful discussions on the analyses of Experiment 2. We also thank Marc Howard, Franklin Zaromb and Nelson Cowan for their helpful comments on an earlier version of this manuscript. Correspondence concerning this article should be addressed to Michael Kahana, Volen National Center for Complex Systems, MS 013, Brandeis University, Waltha

The classic literature presents two opposing views of association: The independent association hypothesis sees associations as unidirectional links between stored items. The associative symmetry hypothesis sees associations as newly formed patterns combining elements of each stored item. Arguing for the symmetry hypothesis, S. Asch & S. Ebenholtz (1962) note that many experiments nd equivalent forward and backward recall of paired associates. Arguing for independent associations, G. Wolford (1971) cite ndings of asymmetric retrieval. A reexamination of the literature reveals that symmetric recall is often observed, but that certain conditions reliably elicit asymmetries. Although robust ndings of asymmetry appear fatal for mathematical memory models that assume symmetric associative mechanisms, careful analyses reveal that these models can account for both material-based and temporal asymmetries, just as models with separate forward and backward associative mechanisms c...

This article presents an attractor neural network model of paired-associate learning and uses a model-based analysis of experimental data to Neural Computation 13, 2075--2092 (2001) c 2001 Massachusetts Institute of Technology shed light on some basic unresolved questions concerning the nature of associations in human memory. The paired-associate learning task is one of the standard assays of human episodic memory. Typically subjects are presented with randomly paired items (e.g., words, letter strings, pictures) and asked to remember each A-B pair for a subsequent memory test. At test, the A items are presented as cues, and subjects attempt to recall the appropriate B items

This paper considers the relation between item recognition and cued recall -- two standard measures of episodic memory. Going beyond measures of performance on each task, we examine the degree to which correlations between successful recognition and successful recall of a single studied episode reflect the commonality of memory processes underlying the two tasks. Specifically, we consider whether four computational memory models (local and global match versions of both matrix and convolution-correlation models) can account for the relatively invariant correlation (# 0.5) between successive recognition and recall tests. Whereas basic versions of each model cannot account for the correlation, versions that take into account variability in goodness-of-encoding and in response criteria, as well as output encoding, are able to account for the level of dependency between tasks. These elaborated models also succeeded in fitting data from two new experiments that manipulated the level of variability in goodness-of-encoding across conditions. This model-

Abstract not found

This paper examines the temporal associative processes governing memory retrieval in free recall by using a decomposition technique that elucidates the distinct contributions of recency and contiguity (Kahana, 1996; Howard & Kahana, 1999). Recency refers to the pervasive finding that the retrievability of previously experienced stimuli or events diminishes with the passage of time and the concomitant presence of interfering activity. Temporal contiguity of events also influences the ease of memory retrieval. If two events, A and B, are experienced in temporal proximity, information about A facilitates retrieval of B, and vice-versa

Associations in episodic memory are formed between items presented close together in time. The temporal context model (TCM) hypothesizes that this contiguity effect is a consequence of shared temporal contexts rather than temporal proximity per se. Using double function lists of paired associates, which include chains of pairs (e.g. A-B, B-C), we examined associations between items that were not presented close together in time but were presented in similar temporal contexts. For instance A and C do not appear together, but both occur in the context of B. Although within-pair associations (e.g. A-B) were asymmetric, across-pair associations (e.g. A-C) showed no evidence for asymmetry. We attempted to describe these transitive associations using two models. One was a heteroassociative model in which the A-C associations resulted from mediated chaining as a result of “stepping through ” the links in the chain. Although this heteroassociative model and TCM make identical predictions regarding simple contiguity effects, the heteroassociative model had great difficulty accounting for the form of transitive associations between items. TCM provided an excellent fit to the data. These data raise the surprising possiblity that episodic contiguity effects do not reflect direct associations between items but rather a process of binding, encoding and retrieval of a gradually-changing