The prefrontal cortex has long been thought to subserve both working memory (the holding of information online for processing) and executive functions (deciding how to manipulate working memory and perform processing). Although many computational models of working memory have been developed, the mechanistic basis of executive function remains elusive, often amounting to a homunculus. This article presents an attempt to deconstruct this homunculus through powerful learning mechanisms that allow a computational model of the prefrontal cortex to control both itself and other brain areas in a strategic, task-appropriate manner. These learning mechanisms are based on subcortical structures in the midbrain, basal ganglia, and amygdala, which together form an actor-critic architecture. The critic system learns which prefrontal representations are task relevant and trains the actor, which in turn provides a dynamic gating mechanism for controlling working memory updating. Computationally, the learning mechanism is designed to simultaneously solve the temporal and structural credit assignment problems. The model’s performance compares favorably with standard backpropagation-based temporal learning mechanisms on the challenging 1-2-AX working memory task and other benchmark working memory tasks.

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

A computational model of human memory for serial order is described (OSCillator-based Associative Recall [OSCAR]). In the model, successive list items become associated to successive states of a dynamic learning-context signal. Retrieval involves reinstatement of the learning context, successive states of which cue successive recalls. The model provides an integrated account of both item memory and order memory and allows the hierarchical representation of temporal order information. The model accounts for a wide range of serial order memory data, including differential item and order memory, transposition gradients, item similarity effects, the effects of item lag and separation in judgments of relative and absolute recency, probed serial recall data, distinctiveness effects, grouping effects at various temporal resolutions, longer term memory for serial order, list length effects, and the effects of vocabulary size on serial recall. The serial ordering of behavior is central to much, perhaps most, of human cognition (e.g., Lashley, 1951). Studies of memory for serial order have provided rich data on the psychological repre-sentation of serial order information and therefore offer a signifi-cant challenge to any model of serially ordered behavior. In this

The current state of A.D. Baddeley and G.J. Hitch’s (1974) multicomponent working memory model is reviewed. The phonological and visuospatial subsystems have been extensively investigated, leading both to challenges over interpretation of individual phenomena and to more detailed attempts to model the processes underlying the subsystems. Analysis of the controlling central executive has proved more challenging, leading to a proposed clarification in which the executive is assumed to be a limited The term working memory appears to have been first proposed by Miller, Galanter, and Pribram (1960) in their classic book Plans and the Structure of Behavior. The term has subsequently been used in computational modeling approaches (Newell & Simon, 1972) and in animal learning studies, in which the participant animals are required to hold information across a number of trials within the same day (Olton, 1979). Finally, within cognitive psychology, the term has been adopted to cover the system or systems involved in the temporary maintenance and manipulation of information. Atkinson and Shiffrin (1968) applied the term to a unitary short-term store, in contrast to the proposal of Baddeley and Hitch (1974), who used it to refer to a system comprising multiple components. They emphasized the functional importance of this system, as opposed to its simple storage capacity. It is this latter concept of a multicomponent working memory that forms the focus of the discussion that follows. I myself have been using the concept for over 25 years; does it still work? Before addressing this issue, it is perhaps appropriate to consider what are the criteria for working. The multicomponent model of working memory was proposed as a theoretical framework whose function was to give

A model of memory retrieval is described. The model embodies 4 main claims: (a) temporal memory— traces of items are represented in memory partly in terms of their temporal distance from the present; (b) scale-similarity—similar mechanisms govern retrieval from memory over many different timescales; (c) local distinctiveness—performance on a range of memory tasks is determined by interference from near psychological neighbors; and (d) interference-based forgetting—all memory loss is due to interference and not trace decay. The model is applied to data on free recall and serial recall. The account emphasizes qualitative similarity in the retrieval principles involved in memory performance at all timescales, contrary to models that emphasize distinctions between short-term and long-term memory.

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.

The medial temporal lobe (MTL) has been studied extensively at all levels of analysis, yet its function remains unclear. Theory regarding the cognitive function of the MTL has centered along 3 themes. Different authors have emphasized the role of the MTL in episodic recall, spatial navigation, or relational memory. Starting with the temporal context model (M.W. Howard and M. J. Kahana, 2002), a distributed memory model that has been applied to benchmark data from episodic recall tasks, the authors propose that the entorhinal cortex supports a gradually changing representation of temporal context and the hippocampus proper enables retrieval of these contextual states. Simulation studies show this hypothesis explains the firing of place cells in the entorhinal cortex and the behavioral effects of hippocampal lesion in relational memory tasks. These results constitute a first step towards a unified computational theory of MTL function that integrates neurophysiological, neuropsychological and cognitive findings.

In the single-store model of memory, the enhanced recall for the last items in a free-recall task (i.e., the recency effect) is understood to reflect a general property of memory rather than a separate short-term store. This interpretation is supported by the finding of a long-term recency effect under conditions that eliminate the contribution from the short-term store. In this article, evidence is reviewed showing that recency effects in the short and long terms have different properties, and it is suggested that 2 memory components are needed to account for the recency effects: an episodic contextual system with changing context and an activation-based short-term memory buffer that drives the encoding of item–context associations. A neurocomputational model based on these 2 components is shown to account for previously observed dissociations and to make novel predictions, which are confirmed in a set of experiments.

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

Recency effects (REs) have been well established in memory and probability learning paradigms but have received little attention in category learning research. Extant categorization models predict REs to be unaffected by learning, whereas a functional interpretation of REs, suggested by results in other domains, predicts that people are able to learn sequential dependencies and incorporate this information into their responses. These contrasting predictions were tested in 2 experiments involving a classification task in which outcome sequences were autocorrelated. Experiment 1 showed that reliance on recent outcomes adapts to the structure of the task, in contrast to models ’ predictions. Experiment 2 provided constraints on how sequential information is learned and suggested possible extensions to current models to account for this learning. Recency effects (REs) are a robust phenomenon in cognitive psychology. REs are said to occur whenever more recent experiences are better remembered or are more influential in judgments about present situations. For example, in research on verbal working memory, REs are arguably among the most fundamental established phenomena, most commonly seen as increased performance