Varieties of working memory as seen in biology and in connectionist/control architectures

Biological and computational concepts that underlie the nature working memory are briefly reviewed. The conceptualization of working memory has changed dramatically in the last 30 years. Current biological work has monitored several aspects of memory, including activation decay, sustained activation, long-term connection change, and differential structures for episodic (hippocampal formation) and procedural learning. Current connectionist modeling has identified factors including multiple-region-based processing, control processing as well as data storage, tradeoffs between fast- and slow-connection-change learning effects, and the speeding of acquisition via multiple levels of learning. The need to relate the biological, behavioral, and computational constraints into models of working memory is discussed. Finally, conceptualizations of working memory must acknowledge the need for human learning systems to be robust enough to operate in a dynamic world.     

Working memory retention systems: a state of activated long-term memory

High temporal resolution event-related brain potential and electroencephalographic coherence studies of the neural substrate of short-term storage in working memory indicate that the sustained coactivation of both prefrontal cortex and the posterior cortical systems that participate in the initial perception and comprehension of the retained information are involved in its storage. These studies further show that short-term storage mechanisms involve an increase in neural synchrony between prefrontal cortex and posterior cortex and the enhanced activation of long-term memory representations of material held in short-term memory. This activation begins during the encoding/comprehension phase and evidently is prolonged into the retention phase by attentional drive from prefrontal cortex control systems. A parsimonious interpretation of these findings is that the long-term memory systems associated with the posterior cortical processors provide the necessary representational basis for working memory, with the property of short-term memory decay being primarily due to the posterior system. In this view, there is no reason to posit specialized neural systems whose functions are limited to those of short-term storage buffers. Prefrontal cortex provides the attentional pointer system for maintaining activation in the appropriate posterior processing systems. Short-term memory capacity and phenomena such as displacement of information in short-term memory are determined by limitations on the number of pointers that can be sustained by the prefrontal control systems.     

Hippocampal and neocortical contributions to memory: advances in the complementary learning systems framework

The complementary learning systems framework provides a simple set of principles, derived from converging biological, psychological and computational constraints, for understanding the differential contributions of the neocortex and hippocampus to learning and memory. The central principles are that the neocortex has a low learning rate and uses overlapping distributed representations to extract the general statistical structure of the environment, whereas the hippocampus learns rapidly using separated representations to encode the details of specific events while minimizing interference. In recent years, we have instantiated these principles in working computational models, and have used these models to address human and animal learning and memory findings, across a wide range of domains and paradigms. Here, we review a few representative applications of our models, focusing on two domains: recognition memory and animal learning in the fear-conditioning paradigm. In both domains, the models have generated novel predictions that have been tested and confirmed.     

The effects of refreshing and elaboration on working memory performance,and their contributions to long-term memory formation

Refreshing and elaboration are cognitive processes assumed to underlie verbal working-memory maintenance and assumed to support long-term memory formation. Whereas refreshing refers to the attentional focussing on representations, elaboration refers to linking representations in working memory into existing semantic networks. We measured the impact of instructed refreshing and elaboration on working and long-term memory separately, and investigated to what extent both processes are distinct in their contributions to working as well as long-term memory. Compared with a no-processing baseline, immediate memory was improved by repeating the items, but not by refreshing them. There was no credible effect of elaboration on working memory, except when items were repeated at the same time. Long-term memory benefited from elaboration, but not from refreshing the words. The results replicate the long-term memory benefit for elaboration, but do not support its beneficial role for working memory. Further, refreshing preserves immediate memory, but does not improve it beyond the level achieved without any processing.     

Modeling individual differences in working memory performance: a source activation account

Working memory resources are needed for processing and maintenance of information during cognitive tasks. Many models have been developed to capture the effects of limited working memory resources on performance. However, most of these models do not account for the finding that different individuals show different sensitivities to working memory demands, and none of the models predicts individual subjects' patterns of performance. We propose a computational model that accounts for differences in working memory capacity in terms of a quantity called source activation , which is used to maintain goal-relevant information in an available state. We apply this model to capture the working memory effects of individual subjects at a fine level of detail across two experiments. This, we argue, strengthens the interpretation of source activation as working memory capacity.     

Linking working memory and long-term memory: a computational model of the learning of new words

The nonword repetition (NWR) test has been shown to be a good predictor of children's vocabulary size. NWR performance has been explained using phonological working memory, which is seen as a critical component in the learning of new words. However, no detailed specification of the link between phonological working memory and long-term memory (LTM) has been proposed. In this paper, we present a computational model of children's vocabulary acquisition (EPAM-VOC) that specifies how phonological working memory and LTM interact. The model learns phoneme sequences, which are stored in LTM and mediate how much information can be held in working memory. The model's behaviour is compared with that of children in a new study of NWR, conducted in order to ensure the same nonword stimuli and methodology across ages. EPAM-VOC shows a pattern of results similar to that of children: performance is better for shorter nonwords and for wordlike nonwords, and performance improves with age. EPAM-VOC also simulates the superior performance for single consonant nonwords over clustered consonant nonwords found in previous NWR studies. EPAM-VOC provides a simple and elegant computational account of some of the key processes involved in the learning of new words: it specifies how phonological working memory and LTM interact; makes testable predictions; and suggests that developmental changes in NWR performance may reflect differences in the amount of information that has been encoded in LTM rather than developmental changes in working memory capacity.     

Models of visuospatial and verbal memory across the adult life span

The authors investigated the distinctiveness and interrelationships among visuospatial and verbal memory processes in short-term, working, and long-term memories in 345 adults. Beginning in the 20s, a continuous, regular decline occurs for processing-intensive tasks (e.g., speed of processing, working memory, and long-term memory), whereas verbal knowledge increases across the life span. There is little differentiation in the cognitive architecture of memory across the life span. Visuospatial and verbal working memory are distinct but highly interrelated systems with domain-specific short-term memory subsystems. In contrast to recent neuroimaging data, there is little evidence for dedifferentiation of function at the behavioral level in old compared with young adults. The authors conclude that efforts to connect behavioral and brain data yield a more complete understanding of the aging mind.     

Working memory, long-term memory, and medial temporal lobe function

Early studies of memory-impaired patients with medial temporal lobe (MTL) damage led to the view that the hippocampus and related MTL structures are involved in the formation of long-term memory and that immediate memory and working memory are independent of these structures. This traditional idea has recently been revisited. Impaired performance in patients with MTL lesions on tasks with short retention intervals, or no retention interval, and neuroimaging findings with similar tasks have been interpreted to mean that the MTL is sometimes needed for working memory and possibly even for visual perception itself. We present a reappraisal of this interpretation. Our main conclusion is that, if the material to be learned exceeds working memory capacity, if the material is difficult to rehearse, or if attention is diverted, performance depends on long-term memory even when the retention interval is brief. This fundamental notion is better captured by the terms subspan memory and supraspan memory than by the terms short-term memory and long-term memory. We propose methods for determining when performance on short-delay tasks must depend on long-term (supraspan) memory and suggest that MTL lesions impair performance only when immediate memory and working memory are insufficient to support performance. In neuroimaging studies, MTL activity during encoding is influenced by the memory load and correlates positively with long-term retention of the material that was presented. The most parsimonious and consistent interpretation of all the data is that subspan memoranda are supported by immediate memory and working memory and are independent of the MTL.     

A source activation theory of working memory: cross-task prediction of performance in ACT-R

Over the decades, computational models of human cognition have advanced from programs that produce output similar to that of human problem solvers to systems that mimic both the products and processes of human performance. In this paper, we describe a model that achieves the next step in this progression: predicting individual participants’ performance across multiple tasks after estimating a single individual difference parameter. We demonstrate this capability in the context of a model of working memory, where the individual difference parameter for each participant represents working memory capacity. Specifically, our model is able to make zero-parameter predictions of individual participants’ performance on a second task after separately fitting performance on a preliminary task. We argue that this level of predictive ability offers an important test of the theory underlying our model.     

Distinguishing short-term memory from working memory

The aim of the present research was to determine whether short-term memory and working memory could be distinguished. In two studies, 7- to 13-year-olds (N = 155, N = 132) were administered tasks thought to assess short-term memory as well as tasks thought to assess working memory. Both exploratory and confirmatory factor analyses distinguished short-term memory tasks from working memory tasks. In addition, performance on working memory tasks was related to word decoding skill but performance on short-term memory tasks was not. Finally, performance on both short-term memory and working memory tasks were associated with age-related increases in processing speed. Results are discussed in relation to models of short-term and working memory.     

Dynamic search and working memory in social recall

What are the mechanisms underlying search in social memory (e.g., remembering the people one knows)? Do the search mechanisms involve dynamic local-to-global transitions similar to semantic search, and are these transitions governed by the general control of attention, associated with working memory span? To find out, we asked participants to recall individuals from their personal social networks and measured each participant's working memory capacity. Additionally, participants provided social-category and contact-frequency information about the recalled individuals as well as information about the social proximity among the recalled individuals. On the basis of these data, we tested various computational models of memory search regarding their ability to account for the patterns in which participants recalled from social memory. Although recall patterns showed clustering based on social categories, models assuming dynamic transitions between representations cued by social proximity and frequency information predicted participants' recall patterns best-no additional explanatory power was gained from social-category information. Moreover, individual differences in the time between transitions were positively correlated with differences in working memory capacity. These results highlight the role of social proximity in structuring social memory and elucidate the role of working memory for maintaining search criteria during search within that structure.     

Temporal clustering and sequencing in short-term memory and episodic memory

A model of short-term memory and episodic memory is presented, with the core assumptions that (a) people parse their continuous experience into episodic clusters and (b) items are clustered together in memory as episodes by binding information within an episode to a common temporal context. Along with the additional assumption that information within a cluster is serially ordered, the model accounts for a number of phenomena from short-term memory (with a focus on serial recall) and episodic memory (with a focus on free recall). The model also accounts for the effects of aging on serial and free recall, apparent temporal isolation effects in short- and long-term memory, and the relation between individual differences in working memory and episodic memory performance.     

Similarity effects in visual working memory

Perceptual similarity is an important property of multiple stimuli. Its computation supports a wide range of cognitive functions, including reasoning, categorization, and memory recognition. It is important, therefore, to determine why previous research has found conflicting effects of inter-item similarity on visual working memory. Studies reporting a similarity advantage have used simple stimuli whose similarity varied along a featural continuum. Studies reporting a similarity disadvantage have used complex stimuli from either a single or multiple categories. To elucidate stimulus conditions for similarity effects in visual working memory, we tested memory for complex stimuli (faces) whose similarity varied along a morph continuum. Participants encoded 3 morphs generated from a single face identity in the similar condition, or 3 morphs generated from different face identities in the dissimilar condition. After a brief delay, a test face appeared at one of the encoding locations for participants to make a same/different judgment. Two experiments showed that similarity enhanced memory accuracy without changing the response criterion. These findings support previous computational models that incorporate featural variance as a component of working memory load. They delineate limitations of models that emphasize cortical resources or response decisions.     

Visual search is slowed when visuospatial working memory is occupied

Visual working memory plays a central role in most models of visual search. However, a recent study showed that search efficiency was not impaired when working memory was filled to capacity by a concurrent object memory task (Woodman, Vogel, & Luck, 2001). Objects and locations may be stored in separate working memory subsystems, and it is plausible that visual search relies on the spatial subsystem, but not on the object subsystem. In the present study, we sought to determine whether maintaining spatial information in visual working memory impairs the efficiency of a concurrent visual search task. Visual search efficiency and spatial memory accuracy were both impaired when the search and the memory tasks were performed concurrently, as compared with when the tasks were performed separately. These findings suggest that common mechanisms are used to process information during difficult visual search tasks and to maintain spatial information in working memory.     

On the division of working memory and long-term memory and their relation to intelligence: A latent variable approach

The present study examined the extent to which working (WM) and long-term memory (LTM) reflect the same, related, or completely different constructs and how they relate to other cognitive ability constructs. Participants performed various WM, recall, recognition, general fluid (gF) and general crystallized intelligence (gC) measures. Confirmatory factor analyses suggested that the memory measures could be grouped into three separate yet correlated factors (WM, recall, and recognition) and that these factors were strongly related to gF, but were related less so with gC. Furthermore, it was found that the common variance from the three memory factors could be accounted for by a higher-order memory factor which was strongly related to gF, but less so with gC. Finally, structural equation modeling suggested that both the variance common to the WM tasks and the variance common to all the memory tasks accounted for a unique variance in gF. These results are interpreted within an embedded process model of memory and suggest that WM and LTM tasks measure both shared and unique processes, which are important for intelligence.     

Word length and phonological similarity effects in simple, complex, and delayed serial recall tasks: implications for working memory

Some current models of working memory argue that a passive short-term store is not involved in more dynamic working memory tasks. Other models argue that standard short-term memory and working memory tasks rely on common storage facilities. We examine these issues by exploring two signature effects of passive short-term storage in simple span, complex span, and Brown-Peterson tasks. The finding that all three tasks show word length and phonological similarity effects suggests that common processes or storage mechanisms are involved in all tasks. The implications for models of working memory are discussed.     

Word length and phonological similarity effects in simple,complex, and delayed serial recall tasks: Implications for working memory

Some current models of working memory argue that a passive short-term store is not involved in more dynamic working memory tasks. Other models argue that standard short-term memory and working memory tasks rely on common storage facilities. We examine these issues by exploring two signature effects of passive short-term storage in simple span, complex span, and Brown-Peterson tasks. The finding that all three tasks show word length and phonological similarity effects suggests that common processes or storage mechanisms are involved in all tasks. The implications for models of working memory are discussed.     

Poorer aerobic fitness relates to reduced integrity of multiple memory systems

Epidemiological investigations have revealed increases in the prevalence of sedentary behaviors in industrialized societies. However, the implications of those lifestyle choices and related cardiorespiratory fitness levels for memory function are not well-understood. To determine the extent to which cardiorespiratory fitness relates to the integrity of multiple memory systems, a cross-sectional sample of young adults were tested over the course of 3 days in areas related to implicit memory, working memory, long-term memory, and aerobic fitness. Findings revealed an association between aerobic fitness and memory function such that individuals with lower cardiorespiratory fitness exhibited poorer implicit memory performance and poorer long-term memory retention. These data indicate that cardiorespiratory fitness may be important for the optimal function of neural networks underlying these memory systems.     

Working memory(s).

Working memory is variously defined as a set of linked and interacting information processing components that maintain information in a short-term store (or retrieve information into that store) for the purpose of the active manipulation of the stored items. The purpose of the this Special Issue is to present data relevant to the question of the functional organization of working memory. In this Introduction we review the two models of working memory and suggest that some of the similarities may be more apparent than real. We further suggest that the two models describe different systems that are specialized for different kinds of stimuli and for different kinds of information processing.