Working Memory System Research Articles

A digital neuromorphic system for working memory based on spiking neuron-astrocyte network

Among various types of memory, working memory (WM) plays a crucial role in reasoning, decision-making, and behavior regulation. Neuromorphic computing is a well-established engineering approach that offers promising avenues for advancing our understanding of WM processes by mimicking the structure and operation of the human brain using electronic technology. In this work, a digital neuromorphic system is proposed and then implemented in hardware to illustrate the real-time WM process based on the spiking neuron-astrocyte network (SNAN). The implemented SNAN utilizes a bidirectional neuron-astrocyte interaction to realize the WM process, allowing for a more brain-like memory emulation. Various hardware optimization methods, including piecewise linear approximation, double buffering, and time multiplexing are recruited to minimize the area and power consumption and facilitate the implementation of the WM concept on a single field programmable gate array (FPGA) chip. The proposed neuromorphic system is evaluated by testing its capacity for multi-item memory formation, an essential characteristic of human WM. The results show that the time duration between the store and recall phases is a critical parameter for acceptable retrieval performance. Additionally, the results demonstrate that the proposed neuromorphic system for WM is resilient to noise. Finally, the design modularity of the system facilitates easy extension for implementing larger networks and adapting to real-world applications.

The role of variation in phonological and semantic working memory capacities in sentence comprehension: neural evidence from healthy and brain-damaged individuals.

Research on the role of working memory (WM) in language processing has typically focused on WM for phonological information. However, considerable behavioral evidence supports the existence of a separate semantic WM system that plays a greater role in language processing. We review the neural evidence that supports the distinction between phonological and semantic WM capacities and discuss how individual differences in these capacities relate to sentence processing. In terms of neural substrates, findings from multivariate functional MRI for healthy participants and voxel-based lesion-symptom mapping for brain-damaged participants imply that the left supramarginal gyrus supports phonological WM, whereas the left inferior frontal gyrus (LIFG) and angular gyrus support semantic WM. In sentence comprehension, individual variation in semantic but not phonological WM related to performance in resolving semantic information and the LIFG region implicated in semantic WM showed fMRI activation during the resolution of semantic interference. Moreover, variation for brain-damaged participants in the integrity of a fiber tract supporting semantic WM had a greater relation to the processing of complex sentences than did the integrity of fiber tracts supporting phonological WM. Overall, the neural findings provide converging evidence regarding the distinction of these two capacities and the greater contribution of individual differences in semantic than phonological WM capacity to sentence processing.

Relative phase of distributed oscillatory dynamics implements a working memory in a simple brain.

We report the existence of a working memory system in the nematode C. elegans that is employed for deferred action in a sensory-guided decision-making process. We find that the turn direction of discrete reorientations during navigation is under sensory-guided control and relies on a working memory that can persist over an intervening behavioral sequence. This memory system is implemented by the phasic interaction of two distributed oscillatory dynamical components. The interaction of oscillatory neural ensembles may be a conserved primitive of cognition across the animal kingdom.

Exploring the Relationship Between Working Memory Capacity and L2 Oral Fluency

This study investigated the relationship between working memory capacity (WMC) and L2 oral fluency and explored the maintenance of newly presented information in L2 speaking production. When performing speech production tasks, knowledge from long-term memory is at frequent use and hence involves the operation of the episodic buffer, one component of the working memory system, where information is retrieved from long-term memory and stored in chunks as we speak in chunks. In this study, we measured WMC by assessing chunk capacity and chunk size held in and retrieved from the episodic buffer. Chunk capacity was measured by Pair-word Speaking Task; chunk size was measured by Procedure Description Task; and oral fluency was measured by speech rate and mean length of run using an IELTS speaking task. Twenty-nine English-major students participated in this study. The results showed a strong positive correlation between chunk capacity and the two measures of fluency while chunk size negatively correlated with fluency. Data from the recall interviews revealed that participants employed various strategies for the maintenance of the presented information which involved different types of information binding.

Mathematical problem solving in emergent bilingual children: Is growth related to the navigation between two working memory systems?

Mathematical problem solving in emergent bilingual children: Is growth related to the navigation between two working memory systems?

What's in a name: The role of verbalization in reinforcement learning.

(e.g., characters or fractals) and concrete stimuli (e.g., pictures of everyday objects) are used interchangeably in the reinforcement-learning literature. Yet, it is unclear whether the same learning processes underlie learning from these different stimulus types. In two preregistered experiments (N = 50 each), we assessed whether abstract and concrete stimuli yield different reinforcement-learning performance and whether this difference can be explained by verbalization. We argued that concrete stimuli are easier to verbalize than abstract ones, and that people therefore can appeal to the phonological loop, a subcomponent of the working-memory system responsible for storing and rehearsing verbal information, while learning. To test whether this verbalization aids reinforcement-learning performance, we administered a reinforcement-learning task in which participants learned either abstract or concrete stimuli while verbalization was hindered or not. In the first experiment, results showed a more pronounced detrimental effect of hindered verbalization for concrete than abstract stimuli on response times, but not on accuracy. In the second experiment, in which we reduced the response window, results showed the differential effect of hindered verbalization between stimulus types on accuracy, not on response times. These results imply that verbalization aids learning for concrete, but not abstract, stimuli and therefore that different processes underlie learning from these types of stimuli. This emphasizes the importance of carefully considering stimulus types. We discuss these findings in light of generalizability and validity of reinforcement-learning research.

Working memory and inhibitory control deficits in children with ADHD: an experimental evaluation of competing model predictions.

Children with ADHD demonstrate difficulties on many different neuropsychological tests. However, it remains unclear whether this pattern reflects a large number of distinct deficits or a small number of deficit(s) that broadly impact test performance. The current study is among the first experiments to systematically manipulate demands on both working memory and inhibition, with implications for competing conceptual models of ADHD pathogenesis. A clinically evaluated, carefully phenotyped sample of 110 children with ADHD, anxiety disorders, or co-occurring ADHD+anxiety (Mage=10.35, 44 girls; 69% White Not Hispanic/Latino) completed a counterbalanced, double dissociation experiment, with two tasks each per inhibition (low vs. high) x working memory (low vs. high) condition. Bayesian and frequentist models converged in indicating that both manipulations successfully increased demands on their target executive function (BF10>5.33x108, p<.001). Importantly, occupying children's limited capacity working memory system produced slower response times and reduced accuracy on inhibition tasks (BF10>317.42, p<.001, d=0.67-1.53). It also appeared to differentially reduce inhibition (and non-inhibition) accuracy for children with ADHD relative to children with anxiety (BF10=2.03, p=.02, d=0.50). In contrast, there was strong evidence against models that view working memory deficits as secondary outcomes of underlying inhibition deficits in ADHD (BF01=18.52, p=.85). This pattern indicates that working memory broadly affects children's ability to inhibit prepotent tendencies and maintain fast/accurate performance, and may explain the errors that children with ADHD make on inhibition tests. These findings are broadly consistent with models describing working memory as a causal mechanism that gives rise to secondary impairments. In contrast, these findings provide evidence against models that view disinhibition as a cause of working memory difficulties or view working memory as a non-causal correlate or epiphenomenon in ADHD.

Supplemental Material for Mathematical Problem Solving in Emergent Bilingual Children: Is Growth Related to the Navigation Between Two Working Memory Systems?

Supplemental Material for Mathematical Problem Solving in Emergent Bilingual Children: Is Growth Related to the Navigation Between Two Working Memory Systems?

The neurocognitive role of working memory load when Pavlovian motivational control affects instrumental learning.

Research suggests that a fast, capacity-limited working memory (WM) system and a slow, incremental reinforcement learning (RL) system jointly contribute to instrumental learning. Thus, situations that strain WM resources alter instrumental learning: under WM loads, learning becomes slow and incremental, the reliance on computationally efficient learning increases, and action selection becomes more random. It is also suggested that Pavlovian learning influences people's behavior during instrumental learning by providing hard-wired instinctive responses including approach to reward predictors and avoidance of punishment predictors. However, it remains unknown how constraints on WM resources affect instrumental learning under Pavlovian influence. Thus, we conducted a functional magnetic resonance imaging (fMRI) study (N = 49) in which participants completed an instrumental learning task with Pavlovian-instrumental conflict (the orthogonalized go/no-go task) both with and without extra WM load. Behavioral and computational modeling analyses revealed that WM load reduced the learning rate and increased random choice, without affecting Pavlovian bias. Model-based fMRI analysis revealed that WM load strengthened RPE signaling in the striatum. Moreover, under WM load, the striatum showed weakened connectivity with the ventromedial and dorsolateral prefrontal cortex when computing reward expectations. These results suggest that the limitation of cognitive resources by WM load promotes slow and incremental learning through the weakened cooperation between WM and RL; such limitation also makes action selection more random, but it does not directly affect the balance between instrumental and Pavlovian systems.

A mechanistic insight into sources of error of visual working memory in multiple sclerosis

Working memory (WM) is one of the most affected cognitive domains in multiple sclerosis (MS), which is mainly studied by the previously established binary model for information storage (slot model). However, recent observations based on the continuous reproduction paradigms have shown that assuming dynamic allocation of WM resources (resource model) instead of the binary hypothesis will give more accurate predictions in WM assessment. Moreover, continuous reproduction paradigms allow for assessing the distribution of error in recalling information, providing new insights into the organization of the WM system. Hence, by utilizing two continuous reproduction paradigms, memory-guided localization (MGL) and analog recall task with sequential presentation, we investigated WM dysfunction in MS. Our results demonstrated an overall increase in recall error and decreased recall precision in MS. While sequential paradigms were better in distinguishing healthy control from relapsing-remitting MS, MGL were more accurate in discriminating MS subtypes (relapsing-remitting from secondary progressive), providing evidence about the underlying mechanisms of WM deficit in progressive states of the disease. Furthermore, computational modeling of the results from the sequential paradigm determined that imprecision in decoding information and swap error (mistakenly reporting the feature of other presented items) was responsible for WM dysfunction in MS. Overall, this study offered a sensitive measure for assessing WM deficit and provided new insight into the organization of the WM system in MS population.

A mechanistic insight into sources of error of visual working memory in multiple sclerosis.

Working memory (WM) is one of the most affected cognitive domains in multiple sclerosis (MS), which is mainly studied by the previously established binary model for information storage (slot model). However, recent observations based on the continuous reproduction paradigms have shown that assuming dynamic allocation of WM resources (resource model) instead of the binary hypothesis will give more accurate predictions in WM assessment. Moreover, continuous reproduction paradigms allow for assessing the distribution of error in recalling information, providing new insights into the organization of the WM system. Hence, by utilizing two continuous reproduction paradigms, memory-guided localization (MGL) and analog recall task with sequential presentation, we investigated WM dysfunction in MS. Our results demonstrated an overall increase in recall error and decreased recall precision in MS. While sequential paradigms were better in distinguishing healthy control from relapsing-remitting MS, MGL were more accurate in discriminating MS subtypes (relapsing-remitting from secondary progressive), providing evidence about the underlying mechanisms of WM deficit in progressive states of the disease. Furthermore, computational modeling of the results from the sequential paradigm determined that imprecision in decoding information and swap error (mistakenly reporting the feature of other presented items) was responsible for WM dysfunction in MS. Overall, this study offered a sensitive measure for assessing WM deficit and provided new insight into the organization of the WM system in MS population.

What networks in the brain system sustain imagination?

The brain cannot stop elaborating information. While the circuitries implied in processing sensory information, and those involved in programming and producing movements, have been extensively studied and characterized, what circuits elicit and sustain the endogenous activity (which might be referred to as imaginative activity) has not been clarified to a similar extent. The two areas which have been investigated most intensely are visual and motor imagery. Visual imagery mostly involves the same areas as visual processing and has been studied by having the subject face specific visual imagery tasks that are related to the use of the visual sketchpad as a component of the working memory system. Much less is known about spontaneous, free visual imagination, what circuits drive it, how and why. Motor imagery has been studied with several approaches: the neural circuits activated in the brain during performance of a movement have been compared with those involved in visually or kinaesthetically imagining performing the same movement, or in observing another person performing it. Some networks are similarly activated in these situations, although primary motor neurons are only activated during motor execution. Imagining the execution of an action seems unable to activate circuits involved in eliciting accompanying motor adjustments (such as postural adaptations) that are unconsciously (implicitly) associated to the execution of the movement. A more faithful neuronal activation is obtained through kinaesthetic motor imagination-imagining how it feels to perform the movement. Activation of sensory-motor and mirror systems, elicited by observing another person performing a transitive action, can also recruit circuits that sustain implicit motor responses that normally accompany the overt movement. This last aspect has originated the expanding and promising field of action observation therapy (AOT). The fact that the various kinds of motor imagery differentially involve the various brain networks may offer some hints on what neural networks sustain imagery in general, another activity that has an attentive component-recalling a memory, covertly rehearsing a speech, internally replaying a behaviour-and a vague, implicit component that arises from the freely flowing surfacing of internal images, not driven by intentional, conscious control.

Age differences in inhibitory and working memory functioning: limited evidence of system interactions

ABSTRACT Debate persists regarding the nature of age-related deficits in inhibition, and whether inhibitory functioning depends on working memory systems. The current research aimed to measure age-related differences in inhibition and working memory, characterize the relationship between inhibitory functions and working memory performance, and determine how these relationships are affected by age. Toward these ends, we measured performance on a range of established paradigms in 60 young adults (18–30 years) and 60 older adults (60–88 years). Our findings support age-related increases in reflexive inhibition (based on the fixation offset effect and inhibition of return) and age-related decrements in volitional inhibition (based on several paradigms: antisaccade, Stroop, flanker, and Simon). This evidence of stronger reflexive inhibition combined with weaker volitional inhibition suggests that age-related deterioration of cortical structures may allow subcortical structures to operate less controlled. Regarding working memory, older adults had lower backward digit scores and lower forward and backward spatial scores. However, of the 32 analyses (16 in each age group) that tested for dependence of inhibitory functioning on working memory functioning, only one (in young adults) indicated that inhibition performance significantly depended on working memory performance. These results indicate that inhibition and working memory function largely independently in both age groups, and age-related working memory difficulties cannot account for age-related declines in inhibitory control.

The Role of Working Memory in Early Literacy and Numeracy Skills in Kindergarten and First Grade.

The working memory system supports learning processes such as acquiring new information and the development of new skills. Working memory has been found to be related to both early literacy and early numeracy in kindergarten and to linguistic and mathematical academic skills at older ages, but the contribution of each of the memory components at these ages is not yet clear. The purpose of this study is to examine the unique connections among the various systems of WM, early literacy, and early numeracy using various assessment tests of simple WM and complex WM, as well as a variety of tasks in math and language skills administered to the same 250 children in kindergarten and 150 children in first grade. Consistent with the predictions, significant relations among all components of memory and mathematics and language knowledge at both ages were found, although these connections were differential for the different types of tasks and memory systems. The connection of complex WM was stronger in its contribution and more significant in first grade in both mathematics and language domains. Complex WM resources were more important in early literacy at kindergarten age, while simple WM seems to be important in early numeracy. The theoretical and educational implications of these results are discussed accordingly.

The slow rate of working memory consolidationfrom vision is a structural limit.

The speed with which information from vision is transformed into working memory (WM) representations that resist interference from ongoing perception and cognition is the subject of conflicting results. Using distinct paradigms, researchers have arrived at estimates of the consolidation time course ranging from 25 ms to 1 s - a range of more than an order of magnitude. However, comparisons of consolidation duration across very different estimation paradigms rely on the implicit assumption that WM consolidation speed is a stable, structural constraint of the WM system. The extremely large variation in WM consolidation speed estimates across measurement approaches motivated the current work's goal of determining whether consolidation speed truly is a stable structural constraint of WM encoding, or instead might be under strategic control as suggested by some accounts. By manipulating the relative task priority of WM encoding and a subsequent sensorimotor decision in a dual-task paradigm, the current experiments demonstrate that the long duration of WM consolidation does not change as a result of task-specific strategies. These results allow comparison of WM consolidation across estimation approaches, are consistent with recent multi-phase WM consolidation models, and are consistent with consolidation duration being an inflexible structural limit.

The functional unit of neural circuits and its relations to eventual sentience of artificial intelligence systems

The authors discuss a functional unit of neuronal circuits, crucial in both brain structure and artificial intelligence (AI) systems. These units recognize objects denoted by single words or verbal phrases and are responsible for object perception, memorization of image patterns, and retrieval of those patterns as imagination. Functional neuronal units, activated by the working memory system, contribute to problem-solving. The paper highlights the authors' achievements in developing a theory for these functional units, known as 'the equimerec - units', which combine 'the threshold logic unit' and 'the feedback control loop'. The authors emphasize the importance of this theory, especially in the context of highly advanced language-based AI systems. Additionally, the authors note that the functional units' essence relies on backpropagation connections, causing impulse circulation in closed circuits, ultimately leading to the emergence of an electromagnetic field. This phenomenon explains the long-known existence of the human brain's endogenous electromagnetic field. The authors suggest that a similar field likely arises in AI systems. In light of Joe McFadden's "Conscious Electromagnetic Information Field Theory (cemi)", the authors argue that the potential emergence of self-awareness in AI systems deserves due attention.

Cognitive individual differences in the process and product of L2 writing

AbstractThis paper examines how research in second language acquisition has approached the study of cognitive individual differences in the process and product of L2 writing from a theoretical and empirical perspective, paying special attention to the three empirical studies included in this special issue. The paper is divided into three sections. The first section examines the cognitive abilities that have been investigated in L2 writing research, among which working memory stands out as the most widely studied. The second section synthesizes the findings reported by the empirical studies in this issue in relation to the role of working memory in L2 writing behaviors and outcomes. The last section suggests future lines of research that can broaden the current scope of research on writing and cognitive individual differences, mostly centered on the components of the working memory system. This research has important theoretical implications, as little is known about how different cognitive individual differences are implicated in writing, as well as pedagogical implications, as the findings can inform about optimal performance and learning conditions for learners with diverse cognitive ability profiles.

The effect of embedded structures on cognitive load for novice learners during block-based code comprehension

BackgroundCoding has become an integral part of STEM education. However, novice learners face difficulties in processing codes within embedded structures (also termed nested structures). This study aimed to investigate the cognitive mechanism underlying the processing of embedded coding structures based on hierarchical complexity theory, which suggests that more complex hierarchies are involved in embedded versus sequential coding structures. Hierarchical processing is expected to place a great load on the working memory system to maintain, update, and manipulate information. We therefore examined the difference in cognitive load induced by embedded versus sequential structures, and the relations between the difference in cognitive load and working memory capacity.ResultsThe results of Experiment 1 did not fully support our hypotheses, possibly due to the unexpected use of cognitive strategies and the way stimuli were presented. With these factors well controlled, a new paradigm was designed in Experiment 2. Results indicate that the cognitive load, as measured by the accuracy and response times of a code comprehension task, was greater in embedded versus sequential conditions. Additionally, the extra cognitive load induced by embedded coding structures was significantly related to working memory capacity.ConclusionsThe findings of these analyses suggest that processing embedded coding structures exerts great demands on the working memory system to maintain and manipulate hierarchical information. It is therefore important to provide scaffolding strategies to help novice learners process codes across different hierarchical levels within embedded coding structures.

Age-related differences in prefrontal glutamate are associated with increased working memory decay that gives the appearance of learning deficits.

The ability to use past experience to effectively guide decision-making declines in older adulthood. Such declines have been theorized to emerge from either impairments of striatal reinforcement learning systems (RL) or impairments of recurrent networks in prefrontal and parietal cortex that support working memory (WM). Distinguishing between these hypotheses has been challenging because either RL or WM could be used to facilitate successful decision-making in typical laboratory tasks. Here we investigated the neurocomputational correlates of age-related decision-making deficits using an RL-WM task to disentangle these mechanisms, a computational model to quantify them, and magnetic resonance spectroscopy to link them to their molecular bases. Our results reveal that task performance is worse in older age, in a manner best explained by working memory deficits, as might be expected if cortical recurrent networks were unable to sustain persistent activity across multiple trials. Consistent with this, we show that older adults had lower levels of prefrontal glutamate, the excitatory neurotransmitter thought to support persistent activity, compared to younger adults. Individuals with the lowest prefrontal glutamate levels displayed the greatest impairments in working memory after controlling for other anatomical and metabolic factors. Together, our results suggest that lower levels of prefrontal glutamate may contribute to failures of working memory systems and impaired decision-making in older adulthood.

Fluid Intelligence Is (Much) More than Working Memory Capacity: An Experimental Analysis.

Empirical evidence suggests a great positive association between measures of fluid intelligence and working memory capacity, which implied to some researchers that fluid intelligence is little more than working memory. Because this conclusion is mostly based on correlation analysis, a causal relationship between fluid intelligence and working memory has not yet been established. The aim of the present study was therefore to provide an experimental analysis of this relationship. In a first study, 60 participants worked on items of the Advanced Progressive Matrices (APM) while simultaneously engaging in one of four secondary tasks to load specific components of the working memory system. There was a diminishing effect of loading the central executive on the APM performance, which could explain 15% of the variance in the APM score. In a second study, we used the same experimental manipulations but replaced the dependent variable with complex working memory span tasks from three different domains. There was also a diminishing effect of the experimental manipulation on span task performance, which could now explain 40% of the variance. These findings suggest a causal effect of working memory functioning on fluid intelligence test performance, but they also imply that factors other than working memory functioning must contribute to fluid intelligence.