Memory Load Research Articles

Imaging swiFTly: streaming widefield Fourier Transforms for large-scale interferometry

We describe a scalable distributed imaging algorithm framework for next-generation radio telescopes, managing the Fourier transform from apertures to sky (or vice versa) with a focus on minimising memory load, data transfers, and computation. Our algorithm uses smooth window functions to isolate the influence between specific regions of spatial-frequency and image space. This allows the distribution of image data between nodes and the construction of segments of frequency space exactly when and where needed. The developed prototype distributes terabytes of image data across many nodes, while generating visibilities at throughput and accuracy competitive with existing software. Scaling is demonstrated to be better than cubic in problem complexity (for baseline length and field of view), reducing the risk involved in growing radio astronomy processing to large telescopes like the Square Kilometre Array.

Neural mechanisms of Chinese character recognition, updating, and maintenance in the N-back task

Using the N-back task, we investigated how memory load influences the neural activity of the Chinese character cognitive subprocess (recognition, updating, and maintenance) in Mainland Chinese speakers. Twenty-seven participants completed the Chinese character N-back paradigm while having their event-related potentials recorded. The study employed time and frequency domain analyses of EEG data. Results showed that accuracy decreased and response times increased with larger N values. For ERPs, N2pc and P300 amplitudes decreased and SW amplitude increased with larger N values. For time frequency analyses, the desynchronization of alpha oscillations decreased after stimulus onset, but the synchronization of alpha oscillations increased during the maintenance phase. The results suggest that greater memory load is related to a decrease in cognitive resources during updating and an increase in cognitive resources during information maintenance. The results of a behavioral-ERP data structural equation model analysis showed that the ERP indicators in the maintenance phase predicted behavioral performance.

Aerobic exercise improves verbal working memory sub-processes in adolescents: behavioral evidence from an N-back task.

Studies on the effects of aerobic exercise on working memory (WM) have mainly concentrated on the overall effects, yet there is little knowledge on how moderate intensity aerobic exercise impacts the sub-processes of verbal WM (VWM) in adolescents. To address this gap, two experiments were conducted to explore the influence of aerobic exercise on the maintenance and updating sub-processes of VWM. In Experiment 1, a mixed experimental design of 2 (exercise habit: high vs. low) × 3 (memory load: 0-back vs. 1-back vs. 2-back) was used to compare VWM and its sub-processes in 40 adolescents. In Experiment 2, a 2 (group: intervention vs. control) × 3 (time point: pretest vs. 1st post-test vs. 18th post-test) × 3 (memory load: 0-back vs. 1-back vs. 2-back) mixed experimental design was used to investigate the acute and long-term effects of moderate intensity aerobic exercise on VWM and its sub-processes in 24 adolescents with low exercise habits. The results of Experiment 1 showed that VWM performance and its sub-processes in the high exercise habit group were better than those in the low exercise habit group. The results of Experiment 2 showed that the effects of the long-term exercise intervention were superior to those of the acute exercise intervention, and both were superior to the pretest. Meanwhile, it was found that aerobic exercise intervention had a greater effect size on the updating sub-process of VWM. In conclusion, the results indicated that moderate intensity aerobic exercise could enhance the performance of VWM and its sub-processes in adolescents, and long-term intervention showed greater improvement effects compared to acute intervention, especially in the updating sub-process of VWM.

The representation of contextual cue is stimulus-specific yet its expression is flexible.

Contextual cueing refers to the phenomenon in which individuals utilize frequently encountered environmental contexts, comprised of distractors, as cues to expedite a target search. Due to the conflict between the widespread occurrence of contextual cue transfer and the observed impact of changing the identity of distractors on contextual cue learning, the content of contextual cue representations remains contentious. Considering the independent nature of contextual cue learning and expression, our proposition is twofold: (1) Contextual cue representations are stimulus-specific, and (2) their expression is highly flexible. To validate the model, two experiments were conducted. Experiment 1 aimed to confirm the hypothesis that contextual cue representations are stimulus-specific. We manipulated the identity consistency of distractors within repeated scenes during contextual cue learning. Difficulty in contextual cue learning under the identity-changing condition would suggest the necessity of identity within contextual cue representation, indicating the stimulus-specific nature of these representations. Experiment 2 was designed to affirm the conclusion of Experiment 1 and explore the flexibility in the expression of contextual cue representations. This experiment comprised two phases: learning and testing. During the learning phase, participants were exposed to two sets of repeated scenes in different colors under two learning conditions: load and no-load. Working memory load was introduced to interfere with the expression to prevent it from becoming automatic. In the subsequent testing phase, the colors of the two scene sets were interchanged to impede retrieval based on identity. If both load and no-load conditions demonstrate similar levels of contextual cue effects during the testing phase, it implies the flexibility in the expression of contextual cue representations and confirms the conclusion of Experiment 1. In Experiment 1, a notable contextual cue learning effect was observed under the identity-consistent condition (p = 0.001). However, this effect was not evident under the identity-changing condition (p = 0.286). This finding strongly supports the stimulus-specific nature of contextual cue representation. In Experiment 2, the contextual cueing effect appeared but did not show a significant difference between the two conditions (t(23) = 0.02, p = 0.987, BF10 = 0.215), indicating the cognitive system's ability to flexibly redefine retrieval cues. This adaptability aligns with our hypothesis and confirms the high flexibility in the expression process of contextual cue representations and confirms the conclusion of Experiment 1.

Wagers for work: Decomposing the costs of cognitive effort.

Some aspects of cognition are more taxing than others. Accordingly, many people will avoid cognitively demanding tasks in favor of simpler alternatives. Which components of these tasks are costly, and how much, remains unknown. Here, we use a novel task design in which subjects request wages for completing cognitive tasks and a computational modeling procedure that decomposes their wages into the costs driving them. Using working memory as a test case, our approach revealed that gating new information into memory and protecting against interference are costly. Critically, other factors, like memory load, appeared less costly. Other key factors which may drive effort costs, such as error avoidance, had minimal influence on wage requests. Our approach is sensitive to individual differences, and could be used in psychiatric populations to understand the true underlying nature of apparent cognitive deficits.

Human decision making balances reward maximization and policy compression.

Policy compression is a computational framework that describes how capacity-limited agents trade reward for simpler action policies to reduce cognitive cost. In this study, we present behavioral evidence that humans prefer simpler policies, as predicted by a capacity-limited reinforcement learning model. Across a set of tasks, we find that people exploit structure in the relationships between states, actions, and rewards to "compress" their policies. In particular, compressed policies are systematically biased towards actions with high marginal probability, thereby discarding some state information. This bias is greater when there is redundancy in the reward-maximizing action policy across states, and increases with memory load. These results could not be explained qualitatively or quantitatively by models that did not make use of policy compression under a capacity limit. We also confirmed the prediction that time pressure should further reduce policy complexity and increase action bias, based on the hypothesis that actions are selected via time-dependent decoding of a compressed code. These findings contribute to a deeper understanding of how humans adapt their decision-making strategies under cognitive resource constraints.

Role of the prefrontal cortex in musical and verbal short-term memory: A functional near-infrared spectroscopy study

Abstract Auditory short-term memory (STM) is a key process in auditory cognition, with evidence for partly distinct networks subtending musical and verbal STM. The delayed matching-to-sample task (DMST) paradigm has been found suitable for comparing musical and verbal STM and for manipulating memory load. In this study, musical and verbal DMSTs were investigated with measures of activity in frontal areas with functional near-infrared spectroscopy (fNIRS): Experiment 1 compared musical and verbal DMSTs with a low-level perception task (that does not entail encoding, retention, or retrieval of information), to identify frontal regions involved in memory processes. Experiment 2 manipulated memory load for musical and verbal materials to uncover frontal brain regions showing parametric changes in activity with load and their potential differences between musical and verbal materials. A FIR model was used to deconvolute fNIRS signals across successive trials without making assumptions with respect to the shape of the hemodynamic response in a DMST. Results revealed the involvement of the dorso-lateral prefrontal cortex (dlPFC) and inferior frontal gyri (IFG), but not of the superior frontal gyri (SFG) in both experiments, in keeping with previously reported neuroimaging data (including fMRI). Experiment 2 demonstrated a parametric variation of activity with memory load in bilateral IFGs during the maintenance period, with opposite directions for musical and verbal materials. Activity in the IFGs increased with memory load for verbal sound sequences, in keeping with previous results with n-back tasks. The decreased activity with memory load observed with musical sequences is discussed in relation to previous research on auditory STM rehearsal strategies. This study highlights fNIRS as a promising tool for investigating musical and verbal STM not only for typical populations, but also for populations with developmental language disorders associated with functional alterations in auditory STM.

The extrastriate symmetry response is robust to alcohol intoxication.

Visual symmetry activates a network of regions in the extrastriate cortex and generates an event-related potential (ERP) called the sustained posterior negativity (SPN). Previous work has found that the SPN is robust to experimental manipulations of task, spatial attention, and memory load. In the current study, we investigated whether the SPN is also robust to alcohol-induced changes in mental state. A pilot experiment (N = 13) found that alcohol unexpectedly increased SPN amplitude. We followed this unexpected result with two new experiments on separate groups, using an alcohol challenge paradigm. One group completed an Oddball discrimination task (N = 26). Another group completed a Regularity discrimination task (N = 26). In both groups, participants consumed a medium dose of alcohol (0.65 g/kg body weight) and a placebo drink, in separate sessions. Alcohol reduced SPN amplitude in the Oddball task (contrary to the pilot results) but had no effect on SPN amplitude in the Regularity task. In contrast, the N1 wave was consistently dampened by alcohol in all experiments. Exploratory analysis indicated that the inconsistent effect of alcohol on SPN amplitude may be partly explained by individual differences in alcohol use. Alcohol reduced the SPN in light drinkers and increased it in heavier drinkers. Despite remaining questions, the results highlight the automaticity of symmetry processing. Symmetry still produces a large SPN response, even when participants are intoxicated, and even when symmetry is not task relevant.

Alpha event-related decreases during encoding in adults with ADHD – An investigation of sustained attention and working memory processes

BackgroundExecutive functioning deficits are central to established neuropsychological models of ADHD. Oscillatory activity, particularly the alpha rhythm (8–12 Hz) has been associated with cognitive impairments in ADHD. However, most studies to date examined such neural mechanisms underlying executive dysfunction in children and adolescents with ADHD, raising the question of whether and to what extent those ADHD-related working memory impairments are still present in adults. To this end, the current study aimed to investigate the role of alpha event-related decreases (ERD) during working memory processes in adults with and without ADHD. MethodsWe collected electroencephalographic (EEG) data from 85 adults with a lifetime diagnosis of ADHD and 105 controls (aged 32–64), while they performed a continuous performance (CPT) and a spatial delayed response working memory task (SDRT). Time-frequency and independent component analysis (ICA) was used to identify alpha (8–12 Hz) clusters to examine group and condition effects during the temporal profile of sustained attention and working memory processes (encoding, maintenance, retrieval), loads (low and high) and trial type (go and nogo). ResultsIndividuals with ADHD exhibited higher reaction time-variability in SDRT, and slower response times in SDRT and CPT, despite no differences in task accuracy. Although working memory load was associated with stronger alpha ERD in both tasks and both groups (ADHD, controls), we found no consistent evidence for attenuated alpha ERD in adults with ADHD, failing to replicate effects reported in children. In contrast, when looking at the whole sample, the correlations of alpha power during encoding with inattention and hyperactivity-impulsivity symptoms were significant, replicating prior findings in children with ADHD, but suggesting an alternate source for these effects in adults. ConclusionsOur results corroborate the robustness of alpha as a marker of visual attention and suggest that occipital alpha ERD normalizes in adulthood, but with unique contributions of centro-occipital alpha ERD, suggesting a secondary source. This implies that deviations in processes other than previously reported visuospatial cortex engagement may account for the persistent symptoms and cognitive deficits in adults with a history of ADHD.

Where scrollbars are clicked, and why

Scrolling is a widely used mean to interact with visual displays, usually to move content to a certain target location on the display. Understanding how user scroll might identify potentially suboptimal use and allows to infer users’ intentions. In the present study, we examined where users click on a scrollbar depending on the intended scrolling action. In two online experiments, click positions were systematically adapted to the intended scrolling action. Click position selection could not be explained as strict optimization of the distance traveled with the cursor, memory load, or motor-cognitive factors. By contrast, for identical scrolling actions click positions strongly depended on the context and on previous scrolls. The behavior of our participants closely resembled behavior observed for manipulation of other physical devices and suggested a simple heuristic of movement planning. The results have implications for modeling human–computer interaction and may contribute to predicting user behavior.

Inhibition and working memory capacity modulate the mental space-time association.

This research aimed to investigate whether the mental space-time association of temporal concepts could be modulated by the availability of cognitive resources (in terms of working memory and inhibitory control capacities) and to explore whether access to this association could be an automatic process. To achieve this, two experiments were carried out. In Experiment 1, participants had to classify words with future and past meanings. The working memory load (high vs. low) was manipulated and the participants were grouped into quartiles according to their visuospatial working memory capacity (WMC). Temporal concepts were displayed subliminally (immediate masking) and supraliminally (delayed masking). The ANOVA showed a performance pattern consistent with the left-past right-future conceptual scheme, regardless of both the type of masking and the working memory load, except in high WMC participants, in which, interestingly, the space-time association effect was absent. In Experiment 2, participants were asked to respond to the colour of the font of the temporal words, and their attentional control capacity was assessed. The results indicated a timeline effect that was irrespective of the WM load and the type of perceptual processing, but not of the WM capacity or the inhibitory abilities. These findings partially endorse the automatic and implicit access to the mental space-time association and suggest the involvement of the availability of cognitive resources. Individual WMC differences appear to modulate the automatic nature of the effect rather than the processing conditions themselves.

EEG, Pupil Dilations, and Other Physiological Measures of Working Memory Load in the Sternberg Task

EEG, Pupil Dilations, and Other Physiological Measures of Working Memory Load in the Sternberg Task

Control of working memory by phase–amplitude coupling of human hippocampal neurons

Retaining information in working memory is a demanding process that relies on cognitive control to protect memoranda-specific persistent activity from interference1,2. However, how cognitive control regulates working memory storage is unclear. Here we show that interactions of frontal control and hippocampal persistent activity are coordinated by theta–gamma phase–amplitude coupling (TG-PAC). We recorded single neurons in the human medial temporal and frontal lobe while patients maintained multiple items in their working memory. In the hippocampus, TG-PAC was indicative of working memory load and quality. We identified cells that selectively spiked during nonlinear interactions of theta phase and gamma amplitude. The spike timing of these PAC neurons was coordinated with frontal theta activity when cognitive control demand was high. By introducing noise correlations with persistently active neurons in the hippocampus, PAC neurons shaped the geometry of the population code. This led to higher-fidelity representations of working memory content that were associated with improved behaviour. Our results support a multicomponent architecture of working memory1,2, with frontal control managing maintenance of working memory content in storage-related areas3–5. Within this framework, hippocampal TG-PAC integrates cognitive control and working memory storage across brain areas, thereby suggesting a potential mechanism for top-down control over sensory-driven processes.

Learned switch readiness via concurrent activation of task sets: Evidence from task specificity and memory load.

Cognitive flexibility increases when switch demands increase. In task switching experiments, repeated pairing of flexibility-demanding situations with specific contexts leads subjects to become more prepared to adapt to changing task demands in those contexts. One form of such upregulated cognitive flexibility has been demonstrated with a list-wide switch probability (LWSP) effect, where switch costs are smaller in lists with frequent switches than in lists with rare switches. According to a recent proposal, the LWSP effect is supported by a concurrent activation mechanism whereby both task rules are kept available simultaneously in working memory. We conducted four experiments to test two key features in this concurrent activation account of LWSP effects. First, we asked whether the LWSP effects are limited to only the trained tasks, and second, we asked whether concurrent working memory load would reduce the LWSP effects. In Experiment 1, we replicated and extended previous findings that the LWSP manipulation modulates both performance (switch costs) and voluntary switch rates, indicating that context-driven increases in flexibility are generalizable so long as the task-sets remain the same. Results of Experiments 2 and 3 showed that novel tasks do not benefit from the concurrent activation of the two other tasks, suggesting that the LWSP effect is task specific. Experiment 4 showed that holding additional information in working memory reduces the LWSP effect. While these findings support the hypothesis of concurrent activation underlying the increased flexibility in the LWSP effect, caveats remain; additional research is needed to further test this account.

Cognitive offloading is value-based decision making: Modelling cognitive effort and the expected value of memory

How do people decide between maintaining information in short-term memory or offloading it to external reminders? How does this affect subsequent memory? This article presents a simple computational model based on two principles: A) items stored in brain-based memory occupy its limited capacity, generating an opportunity cost; B) reminders incur a small physical-action cost, but capacity is effectively unlimited. These costs are balanced against the value of remembering, which determines the optimal strategy. Simulations reproduce many empirical findings, including: 1) preferential offloading of high-value items; 2) increased offloading at higher memory loads; 3) offloading can cause forgetting of offloaded items (‘Google effect’) but 4) improved memory for other items (‘saving-enhanced memory’); 5) reduced saving-enhanced-memory effect when reminders are unreliable; 6) influence of item-value: people may preferentially offload high-value items and store additional low-value items in brain-based memory; 7) greatest sensitivity to the effort of reminder-setting at intermediate rather than highest/lowest levels of task difficulty; 8) increased offloading in individuals with poorer memory ability. Therefore, value-based decision-making provides a simple unifying account of many cognitive offloading phenomena. These results are consistent with an opportunity-cost model of cognitive effort, which can explain why internal memory feels effortful but reminders do not.

Alpha desynchronization during the filtering initiation phase reflects active processing of distractors

The ability to select task-relevant information and filter out task-irrelevant information is critical to our success in daily goal-directed behavior. Researchers call this ability filtering efficiency and divide it into three cognitive processing stages: detection of distractors, initiation of filtering, and unnecessary storage. Although researchers have conducted more studies on ERP components related to filtration efficiency, there are few studies related to neural oscillations. Alpha oscillation activity is related to the active processing of information and the suppression of distractors. In the current EEG study, we used the change detection task with distracted items to examine whether alpha activity during filtering initiation reflects reactive suppression of distractors by manipulating memory load levels and the presence or absence of distractors. Results showed that, the presence of the distractors caused an increase in the degree of desynchronization of the alpha oscillations, and in the subsequent time, the alpha activity level returned to a level consistent with the absence of interference conditions. Phase synchronization between frontal and posterior brain regions in the upper alpha oscillations found no effects associated with distractors. Based on these results, we believed that the alpha activity during the filtering initiation phase reflected the active processing of distractors, but this may also be due to lower perceptual load of the target items. In addition, we observed a dominance effect of the right hemisphere in both time-frequency results and connectivity results. We speculate that this effect is related to the activation of the right ventral frontoparietal network.

Graph analysis of functional brain networks under varying working memory load: An EEG-based minimum spanning tree approach

Conventional graph theory approach implemented in prior studies to understand brain topology during performance of cognitive tasks is biased and limits comparison of results across studies. This study employed minimum spanning tree (MST) analysis, an unbiased approach, to quantify changes in brain network organization under varying working memory (WM) load. We used 28-channel electroencephalography dataset of 26 subjects performing cognitive (0-, 2- and 3-back) tasks. We computed functional connectivity across channels using weighted phase lag index (wPLI) in multiple frequency bands and constructed MSTs of brain networks under varying WM conditions. Results reveal 0- back had significantly higher parietal alpha wPLI, lower frontal theta wPLI and lower central theta wPLI, than other conditions. In theta band, we observed more star-like topology in 0- back condition than 2- back or 3- back conditions. Our findings suggest sensitivity of these metrics to alterations in brain network topology under varying cognitive demand.

Working memory load modulates anticipatory postural adjustments during step initiation.

Working memory (WM) can influence selective attention. However, the effect of WM load on postural standing tasks has been poorly understood, even though these tasks require attentional resources. The purpose of this study was to examine whether WM load would impact anticipatory postural adjustments (APAs) during step initiation. Sixteen healthy young adults performed stepping tasks alone or concurrently with a WM task in a dual-task design. The stepping tasks involved volitional stepping movements in response to visual stimuli and comprised of simple and choice reaction time tasks and the Flanker task which consisted of congruent and incongruent (INC) conditions. In the dual-task condition, subjects were required to memorize either one or six digits before each stepping trial. Incorrect weight transfer prior to foot-lift, termed APA errors, reaction time (RT), and foot-lift time were measured from the vertical force data. The results showed that APA error rate was significantly higher when memorizing six-digit than one-digit numerals in the INC condition. In addition, RT and foot-lift time were significantly longer in the INC condition compared to the other stepping conditions, while there was no significant effect of WM load on RT or foot-lift time. These findings suggest that high WM load reduces the cognitive resources needed for selective attention and decision making during step initiation.

Instruction Level Parallelism and Memory Synchronization

The simultaneous or parallel execution of a series of instructions within a computer program is known as instruction-level parallelism, or ILP. ILP stands for the average number of instructions executed throughout each stage of this parallel execution, to be more precise. The architecture known as instruction level parallelism (ILP) allows for the execution of several operations in parallel within a single process, each with its own set of resources, including address space, registers, identifiers, state, and program counters. It describes compiler design strategies and processors intended to carry out operations in parallel to increase processor performance, such as memory load and store, integer addition, and float multiplication.

ReSA: Reconfigurable Systolic Array for Multiple Tiny DNN Tensors

Systolic array architecture has significantly accelerated deep neural networks (DNNs). A systolic array comprises multiple processing elements (PEs) that can perform multiply-accumulate (MAC). Traditionally, the systolic array can execute a certain amount of tensor data that matches the size of the systolic array simultaneously at each cycle. However, hyper-parameters of DNN models differ across each layer and result in various tensor sizes in each layer. Mapping these irregular tensors to the systolic array while fully utilizing the entire PEs in a systolic array is challenging. Furthermore, modern DNN systolic accelerators typically employ a single dataflow. However, such a dataflow isn’t optimal for every DNN model. This work proposes ReSA, a reconfigurable dataflow architecture that aims to minimize the execution time of a DNN model by mapping tiny tensors on the spatially partitioned systolic array. Unlike conventional systolic array architectures, the ReSA data path controller enables the execution of the input, weight, and output-stationary dataflow on PEs. ReSA also decomposes the coarse-grain systolic array into multiple small ones to reduce the fragmentation issue on the tensor mapping. Each small systolic sub-array unit relies on our data arbiter to dispatch tensors to each other through the simple interconnected network. Furthermore, ReSA reorders the memory access to overlap the memory load and execution stages to hide the memory latency when tackling tiny tensors. Finally, ReSA splits tensors of each layer into multiple small ones and searches for the best dataflow for each tensor on the host side. Then, ReSA encodes the predefined dataflow in our proposed instruction to notify the systolic array to switch the dataflow correctly. As a result, our optimization on the systolic array architecture achieves a geometric mean speedup of 1.87X over the weight-stationary systolic array architecture across 9 different DNN models.