High Memory Load Research Articles

Brain Networks Communicate Through Theta Oscillations to Encode High Load in a Visuospatial Working Memory Task: An EEG Connectivity Study.

The encoding of visuospatial information is the foremost and indispensable step which determines the outcome in a visuospatial working memory (VSWM) task. It is considered to play a crucial role in limiting our ability to attend and process only 3-5 integrated items of information. Despite its importance in determining VSWM performance, the neural mechanisms underlying VSWM encoding have not been clearly differentiated from those involved during VSWM retention, manipulation and/or retrieval. The high temporal resolution of electroencephalography (EEG) and improved spatial resolution with dense array data acquisition makes it an ideal tool to study the dynamics in the functional brain connectivity during a cognitive task. In the present study, the changes in the functional brain connectivity due to memory load during VSWM encoding were studied using 128-channel EEG. Lagged linear coherence (LagR) was computed between 84 regions of interest (ROIs) defined according to the Brodmann areas for seven EEG frequency bands: delta (2-4Hz), theta (4-8Hz), alpha 1 (8-10.5Hz), alpha 2 (10.5-13Hz), beta 1 (13-20Hz), beta 2 (20-30Hz), and gamma (30-45Hz). Interestingly, out of seven EEG frequency bands investigated in the current study, LagR of only theta band varied significantly in 13 brain connections due to memory load during VSWM encoding. LagR of theta band increased significantly at high memory load when compared to low memory load in twelve brain connections with the maximum change observed between right cuneus and right middle temporal gyrus (Cohen's d = 0.836), indicating the integration of brain processes to confront the increase in memory demands. Theta LagR decreased significantly between left postcentral gyrus and right precentral gyrus at high memory load as compared to low memory load, which might have a role for sustaining attention during encoding. Change in the LagR values due to memory load between fusiform gyrus and lingual gyrus in the right hemisphere had a positive correlation (r = 0.464, p = 0.003) with the error rate, signifying the crucial role played by these two regions in predicting the performance. The current study has not only identified the neural connections that are responsible for the formation of working memory traces during VSWM encoding, but also support the notion that encoding is a rate-limiting process underlying our memory capacity limit.

The Long-Term Benefit of Video Modeling Examples for Guided Inquiry

Inquiry-based learning can be considered a critical component of science education in which students can assess their understanding of scientific concepts and scientific reasoning skills while actively constructing new knowledge through different types of activity levels. However, engaging in inquiry activities can be cognitively demanding for students, especially those with low prior knowledge of scientific reasoning skills (reasoning ability). Learning new information when preexisting schemata are absent entails more interacting elements and thus imposes a high working memory load, resulting in lower long-term learning effects. Borrowing knowledge from others via video modeling examples before carrying out an inquiry task provides learners with more working memory capacity to focus on problem-solving strategies and construct useful cognitive schemata for solving subsequent (virtual) inquiry tasks (Kant, Scheiter, and Oschatz, 2017). The goal of the present study (N = 174 6/7th graders) is to investigate the benefits of combining example-based learning with physical, hands-on investigations in inquiry-based learning for acquiring scientific reasoning skills. The study followed a 2 (video modeling example vs. no example) x 2 (guided vs. structured inquiry) x 2 (retention interval: immediate vs. delayed) mixed-factorial design. In addition, the students’ need for cognition (Preckel, 2013), cognitive abilities (Heller and Perleth, 2000), (intrinsic, extraneous and germane) cognitive load (Cierniak, Scheiter, and Gerjets, 2009) and performance success were measured. Although the results of an intermediate test after the first manipulation were higher among students who watched a video modeling example (d =.97), combining video modeling examples with inquiry was not found to benefit performance success. Furthermore, regardless of manipulation, all students achieved equal results on an assessment immediately following the inquiry task. Only in the long run did a video modeling example prove to be advantageous for guided inquiry (ηp2 = .023). A video modeling example turned out to be a crucial prerequisite for the long-term effectiveness of guided inquiry because it helped create stable problem-solving schemata; however, the long-term retention of structured inquiry did not rely on a video modeling example.

Common cortical architectures for phonological working memory identified in individual brains

Phonological working memory is the capacity to briefly maintain and recall representations of sounds important for speech and language and is believed to be critical for language and reading acquisition. Whether phonological working memory is supported by fronto-parietal brain regions associated with short-term memory storage or perisylvian brain structures implicated in speech perception and production is unclear, perhaps due to variability in stimuli, task demands, and individuals. We used fMRI to assess neurophysiological responses while individuals performed two tasks with closely matched stimuli but divergent task demands—nonword repetition and nonword discrimination—at two levels of phonological working memory load. Using analyses designed to address intersubject variability, we found significant neural responses to the critical contrast of high vs. low phonological working memory load in both tasks in a set of regions closely resembling those involved in speech perception and production. Moreover, within those regions, the voxel-wise patterns of load-related activation were highly correlated between the two tasks. These results suggest that brain regions in the temporal and frontal lobes encapsulate the core neurocomputational components of phonological working memory; an architecture that becomes increasingly evident as neural responses are examined in successively finer-grained detail in individual participants.

The Human Intraparietal Sulcus Modulates Task-Evoked Functional Connectivity.

Past studies have demonstrated that flexible interactions between brain regions support a wide range of goal-directed behaviors. However, the neural mechanisms that underlie adaptive communication between brain regions are not well understood. In this study, we combined theta-burst transcranial magnetic stimulation (TMS) and functional magnetic resonance imaging to investigate the sources of top-down biasing signals that influence task-evoked functional connectivity. Subjects viewed sequences of images of faces and buildings and were required to detect repetitions (2-back vs. 1-back) of the attended stimuli category (faces or buildings). We found that functional connectivity between ventral temporal cortex and the primary visual cortex (VC) increased during processing of task-relevant stimuli, especially during higher memory loads. Furthermore, the strength of functional connectivity was greater for correct trials. Increases in task-evoked functional connectivity strength were correlated with increases in activity in multiple frontal, parietal, and subcortical (caudate and thalamus) regions. Finally, we found that TMS to superior intraparietal sulcus (IPS), but not to primary somatosensory cortex, decreased task-specific modulation in connectivity patterns between the primary VC and the parahippocampal place area. These findings demonstrate that the human IPS is a source of top-down biasing signals that modulate task-evoked functional connectivity among task-relevant cortical regions.

Real-time autonomous navigation of an electric wheelchair in large-scale urban area with 3D map

This paper presents the real-time autonomous navigation of an electric wheelchair in a large-scale urban area. Accurate self-pose localization and well-chosen motion control are crucial for application to urban areas, as electric wheelchairs move on paved roads in dynamic environments and travel along sidewalks at a brisk speed. Our system is equipped with a localization module based on a 3D map and a path planning module based on a navigation map. However, the large-scale 3D map causes a high memory load, and the embedded PC can not deal with the map data. In addition, the large-scale navigation map increases the computational cost of path planning, which causes delays in navigation. To achieve real-time navigation independent of map size, we propose a 6-DoF pose localization switching reference 3D map and a two-step path planning framework. We ran tests by using an electric wheelchair on a real street in Tokyo and found that the proposed navigation system achieved autonomous navigation for over 8.8 km in about 133 minutes. The experimental results showed that the memory load was kept constant and the path planning was performed at high frequency, regardless of the size of the map or the distance to the destination.

Post-traumatic stress disorder is associated with altered modulation of prefrontal alpha band oscillations during working memory

ObjectiveTo investigate the relationship between the severity of PTSD symptoms, modulation of alpha band oscillations, and behavioral performance in a working memory task. MethodsMagnetoencephalography data were recorded in 35 participants with combat exposure and various degrees of PTSD symptom severity while they performed a modified Sternberg working memory task: briefly presented sets of two or six letters had to be held in memory and participants indicated whether subsequent probe letters were present or absent from these sets. ResultsPTSD scores were positively correlated with the false positive rate in the high memory load condition. Higher rates of false recognition were associated with negative probes that were seen in recent previous trials (negative probe recency effect) or were physically similar with the list letters. The relative alpha band power in the left middle frontal gyrus was negatively correlated with both PTSD scores and false positive rates. ConclusionsReduced task specific modulation of alpha band oscillations in left middle frontal cortex may reflect alterations in the functions of pattern separation and suppression of memory traces for irrelevant or no longer relevant information in PTSD. SignificanceThe lower amplitude of prefrontal alpha band oscillations may represent an important physiological basis for core PTSD symptoms and can provide a target for interventions to augment response to treatment.

Capture and Control: Working Memory Modulates Attentional Capture by Reward-Related Stimuli

Physically salient but task-irrelevant distractors can capture attention in visual search, but resource-dependent, executive-control processes can help reduce this distraction. However, it is not only physically salient stimuli that grab our attention: Recent research has shown that reward history also influences the likelihood that stimuli will capture attention. Here, we investigated whether resource-dependent control processes modulate the effect of reward on attentional capture, much as for the effect of physical salience. To this end, we used eye tracking with a rewarded visual search task and compared performance under conditions of high and low working memory load. In two experiments, we demonstrated that oculomotor capture by high-reward distractor stimuli is enhanced under high memory load. These results highlight the role of executive-control processes in modulating distraction by reward-related stimuli. Our findings have implications for understanding the neurocognitive processes involved in real-life conditions in which reward-related stimuli may influence behavior, such as addiction.

Inattention and task switching performance: the role of predictability, working memory load and goal neglect

Inattention is a symptom of many clinical disorders including attention deficit hyperactivity disorder (ADHD) and is thought to be primarily related to limitations in working memory. In two studies, we investigated the implications of inattention for task switching performance. In study one, we measured task switching performance using predictable and unpredictable conditions in adults who self-rated inattention and other ADHD-related tendencies. Tasks required proactive control and reactive control, respectively, under both high and low working memory loads. Results revealed that inattentive, but not hyperactive/impulsive traits, predicted switch costs when switching was predictable and working memory load was high. None of the ADHD traits were related to unpredictable switch costs. Study two was designed to: (1) de-confound the role of proactive control and the need to keep track of task order in the predictable task switching paradigm; (2) investigate whether goal neglect, an impairment related to working memory, could explain the relationship between inattention and predictable task switching. Results revealed that neither predictability nor the need to keep track of the task order led to the association between switch costs and inattention, but instead it was the tendency for those high in inattention to neglect preparatory proactive control, especially when reactive control options were available.

Functional Involvement of Human Periaqueductal Gray and Other Midbrain Nuclei in Cognitive Control.

Recent theoretical advances have motivated the hypothesis that the periaqueductal gray (PAG) participates in behaviors that involve changes in the autonomic control of visceromotor activity, including during cognitively demanding tasks. We used ultra-high-field (7 tesla) fMRI to measure human brain activity at 1.1 mm resolution while participants completed a working memory task. Consistent with prior work, participants were less accurate and responded more slowly with increasing memory load-signs of increasing task difficulty. Whole-brain fMRI analysis revealed increased activity in multiple cortical areas with increasing working memory load, including frontal and parietal cortex, dorsal cingulate, supplementary motor area, and anterior insula. Several dopamine-rich midbrain nuclei, such as the substantia nigra and ventral tegmental area, also exhibited load-dependent increases in activation. To investigate PAG involvement during cognitive engagement, we developed an automated method for segmenting and spatially normalizing the PAG. Analyses using cross-validated linear support vector machines showed that the PAG discriminated high versus low working memory load conditions with 95% accuracy in individual subjects based on activity increases in lateral and ventrolateral PAG. Effect sizes in the PAG were comparable in magnitude to those in many of the cortical areas. These findings suggest that cognitive control is not only associated with cortical activity in the frontal and parietal lobes, but also with increased activity in the subcortical PAG and other midbrain regions involved in the regulation of autonomic nervous system function.SIGNIFICANCE STATEMENT Functional neuroimaging in humans has shown that cognitive control engages multiple corticostriatal networks and brainstem nuclei, but theoretical advances suggest that the periaqueductal gray (PAG) should also be engaged during cognitively demanding tasks. Recent advances in ultra-high-field fMRI provided an opportunity to obtain the first evidence that increased activation of intermediate and rostral portions of lateral and ventrolateral PAG columns in humans is modulated by cognitive load. These findings suggest that cognitive control is not solely mediated by activity in the cortex, but that midbrain structures important for autonomic regulation also play a crucial role in higher-order cognition.

Pre-trial and pre-response EEG microstates in schizophrenia: An endophenotypic marker

Cognitive deficits in Schizophrenia interfere with everyday functioning and social functioning. Strong familial associations in schizophrenia might serve to establish cognitive impairments as endophenotypic markers. Therefore, visuo-spatial working memory simulating day-to-day activities at high memory load was assessed in patients with schizophrenia, their first-degree relatives and healthy controls to explore pre-trial and pre-response EEG microstates and their intracranial generators. Twenty-eight patients with schizophrenia, first-degree relatives and matched healthy controls participated in the study. Brain activity during visuo-spatial working memory task was recorded using 128-channel electroencephalography. Pre-trial and pre-response microstate maps of correct and error trials were clustered across groups according to their topography. Microstate map parameters and underlying cortical sources were compared among groups. Pre-trial (correct) microstate Map 1 was significantly different between controls and patients which could qualify it as a state marker with its intracranial generator localized to right inferior frontal gyrus (rIFG). Pre-response (correct) microstate map was significantly different between controls and first-degree relatives which could be considered an endophenotypic marker for schizophrenia. No significant differences were observed for error trials between groups. rIFG which is involved in the execution of multi-component behaviour and selective inhibitory control could distinguish patients with schizophrenia from their first-degree relatives and healthy controls. Further, microstate based biomarkers have the potential to facilitate diagnosis of schizophrenia at a preclinical stage resulting in efficient diagnosis and better prognosis.

Multiple states in visual working memory: Evidence from oculomotor capture by memory-matching distractors.

Visual working memory (VWM) representations interact with attentional guidance, but there is controversy over whether multiple VWM items simultaneously influence attentional guidance. Extant studies relied on continuous variables like response times, which can obscure capture - especially if VWM representations cycle through interactive and non-interactive states. Previous conflicting findings regarding guidance when under high working memory (WM) load may be due to the use of noisier response time measures that mix capture and non-capture trials. Thus, we employed an oculomotor paradigm to characterize discrete attentional capture events under both high and low VWM load. Participants held one or two colors in memory, then executed a saccade to a target disk. On some trials, a distractor (sometimes VWM-matching) appeared simultaneously with the target. Eye movements were more frequently directed to a VWM-matching than a non-matching distractor for both load conditions. However, oculomotor capture by a VWM-matching distractor occurred less frequently under high compared with low load. These results suggest that attention is automatically guided toward items matching only one of two colors held in memory at a time, suggesting that items in VWM may cycle through attention-guiding and not-guiding states when more than one item is held in VWM and the task does not require that multiple items be maintained in an active, attention-guiding state.

Refining memory assessment of elderly people with cognitive impairment: Insights from the short-term memory binding test

Alzheimer's disease (AD) affects temporary memory for bound features more remarkably than for individual features. Such selective impairments manifest from presymptomatic through dementia stages via titration procedures. A recent study suggested that without titration and with high memory load the binding selectivity may disappear in people at risk of AD such as those with Mild Cognitive Impairment (MCI). We compared data from two studies on temporary binding which assessed people with MCI and controls using different memory loads (2 or 3 items). Selective binding impairments were found in MCI, but relative to controls, such selectivity was contingent upon memory load (i.e., present with 2 items). Further analysis with MCI people who tested positive to neuroimaging biomarkers (i.e., hippocampal atrophy) confirmed that this specific binding impairments are a feature of prodromal AD. The temporary binding task has been recently suggested by consensus papers as a potential screening tool for AD. The results presented here inform on task properties that can maximize the reliability of this new assessment tool for the detection of memory impairments in prodromal cases of AD.

Proceedings #36: Electrical Stimulation over Posterior Parietal Cortex Enhances Distractor Filtering and Target Maintenance in Visual Working Memory

Visual working memory (VWM) is known for its severe capacity limit. Distractor filtering in the frontal cortex and target maintenance in the parietal cortex both contribute to the information stored in VWM. The present study used anodal transcranial direct current stimulation (tDCS) to stimulate the prefrontal cortex (PFC) and the posterior parietal cortex (PPC) before participants completing a change detection task with or without distractors to assess the contribution of these two brain areas to distractor filtering and target maintenance, respectively. Our results show that tDCS over PPC selectively enhanced VWM capacity during the high memory load condition with distractors compared with sham and PFC stimulation, accompanied by amplitude increase in frontal distractor-filtering component and parietal delay activity. These results are consistent with a network perspective of WM capacity, suggesting that PPC is a primary hub in networks providing VWM capacity.

Cerebellar contributions to sequence prediction in verbal working memory

Verbal working memory is one of the most studied non-motor functions with robust cerebellar involvement. While the superior cerebellum (lobule VI) has been associated with articulatory control, the inferior cerebellum (lobule VIIIa) has been linked to phonological storage. The present study was aimed to elucidate the differential roles of these regions by investigating whether the cerebellum might contribute to verbal working memory via predictions based on sequence learning/detection. 19 healthy adult subjects completed an fMRI-based Sternberg task which included repeating and novel letter sequences that were phonologically similar or dissimilar. It was hypothesized that learning a repeating sequence of study letters would reduce phonological storage demand and associated right inferior cerebellar activations and that this effect would be modulated by phonological similarity of the study letters. Specifically, while increased phonological storage demand due to high phonological similarity was expected to be reflected in increased right inferior cerebellar activations for similar relative to dissimilar study letters, the reduction in activation for repeating relative to novel sequences was expected to be more profound for phonologically similar than for dissimilar study letters, especially at higher memory load. Results confirmed the typical effects of cognitive load (5 vs. 2 study letters) and phonological similarity in several cerebellar and neocortical brain regions as well as in behavioral data (accuracy and response time). Importantly, activations in superior and inferior cerebellar regions were differentially modulated as a function of similarity and sequence novelty, indicating that particularly lobule VIIIa may contribute to verbal working memory by generating predictions of letter sequences that reduce the likelihood of phonological loop failure before stored items need to be retrieved. The present study is consistent with other investigations that support prediction, which can be based on sequence learning or detection, as an overarching cerebellar function.

A functional promoter variant of the human formimidoyltransferase cyclodeaminase (FTCD) gene is associated with working memory performance in young but not older adults.

The central role of working memory in IQ and the high heritability of working memory performance motivated interest in identifying the specific genes underlying this heritability. The FTCD (formimidoyltransferase cyclodeaminase) gene was identified as a candidate gene for allelic association with working memory in part from genetic mapping studies of mouse Morris water maze performance. The present study tested variants of this gene for effects on a delayed match-to-sample task of a large sample of younger and older participants. The rs914246 variant, but not the rs914245 variant, of the FTCD gene modulated accuracy in the task for younger, but not older, people under high working memory load. The interaction of haplotype × distance × load had a partial eta squared effect size of 0.015. Analysis of simple main effects had partial eta squared effect sizes ranging from 0.012 to 0.040. A reporter gene assay revealed that the C allele of the rs914246 genotype is functional and a main factor regulating FTCD gene expression. This study extends previous work on the genetics of working memory by revealing that a gene in the glutamatergic pathway modulates working memory in young people but not in older people. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Neural processing of working memory in adults with ADHD in a visuospatial change detection task with distractors.

Individuals with Attention-Deficit Hyperactivity Disorder (ADHD) are often characterized by deficits in working memory (WM), which manifest in academic, professional, and mental health difficulties. To better understand the underlying mechanisms of these presumed WM deficits, we compared adults with ADHD to their peers on behavioral and neural indices of WM. We used a visuospatial change detection task with distractors which was designed to assess the brain’s ability to effectively filter out distractors from WM, in addition to testing for effects of WM load. Twenty-seven unmedicated adults with ADHD were compared to 27 matched peers on event-related potential (ERP) measures of WM, i.e., the contralateral delay activity (CDA). Despite severe impairments in everyday life functioning, findings showed no difference in deficits in behavioral tests of working memory for adults with ADHD compared to their peers. Interestingly, there were differences in neural activity between individuals with ADHD and their peers showing that the CDA of individuals with ADHD did not distinguish between high, distractor, and low memory load conditions. These data suggest, in the face of comparable behavioral performance, a difference in neural processing efficiency, wherein the brains of individuals with ADHD may not be as selective in the allocation of neural resources to perform a WM task.

Reaction Time of Erroneous Responses in the Go/No-Go Paradigm.

Reaction time (RT) of erroneous responses in go/no-go tasks tends to be shorter than RT of correct responses. An opposite difference has been reported ( Halperin, Wolf, Greenblatt, & Young, 1991 ) which could be attributed to differences in go trial probability, or to high memory demand. Two experiments aimed here to test these two explanations, a simultaneous matching task with low memory load (Experiment 1), and a sequential matching task with high memory load (Experiment 2). Go trial probability was also manipulated. Short false positive RT was obtained only in the sequential matching task with high go trial probability, while long false positive RT was obtained in the other three conditions. Low go trial probability and high memory load were both found to be sufficient, by themselves, to create long false positives attributable to confusion. Short false positives in the high go trial probability/low memory load condition were attributed to failure of response inhibition.

Effect of sleep deprivation on emotional working memory.

The emotional dysregulation and impaired working memory found after sleep loss can have severe implications for our daily functioning. Considering the intertwined relationship between emotion and cognition in stimuli processing, there could be further implications of sleep deprivation in high‐complex emotional situations. Although studied separately, this interaction between emotion and cognitive processes has been neglected in sleep research. The aim of the present study was to investigate the effect of 1 night of sleep deprivation on emotional working memory. Sixty‐one healthy participants (mean age: 23.4 years) were either sleep deprived for 1 night (n = 30) or had a normal night’s sleep (n = 31). They performed an N‐back task with two levels of working memory load (1‐back and 3‐back) using positive, neutral and negative picture scenes. Sleep deprivation, compared with full night sleep, impaired emotional working memory accuracy, but not reaction times. The sleep‐deprived participants, but not the controls, responded faster to positive than to negative and neutral pictures. The effect of sleep deprivation was similar for both high and low working memory loads. The results showed that although detrimental in terms of accuracy, sleep deprivation did not impair working memory speed. In fact, our findings indicate that positive stimuli may facilitate working memory processing speed after sleep deprivation.

Effects of Working Memory Load and Age on the Comprehension of Passive Sentences

The ability of older adults to comprehend sentences may decline due to the cognitive changes in working memory. Therefore, an increase in working memory demands during sentence comprehension would result in poorer performance among older adults. To test this hypothesis, the present study explored sentence comprehension as a result of manipulations of age and working memory loads using a sentence-picture matching task. 35 older adults and 35 younger adults were required to match Mandarin passive sentences (high working memory load) and active sentences (low working memory load) with pictures. Passive sentences were found to be more difficult than active sentences for all participants. Older adults responded to passive sentences more slowly than younger adults. However, no significant age difference was found in accuracy of responses. Accuracy on passive sentence comprehension was marginally correlated with syntactic complexity effect among older adults. Compared with younger adults, older adults seem to be more disrupted by the increased WM load in passive sentence comprehension, but they can compensate for the decline in the accuracy of comprehension by spending extra time on sentences with high WM load.

The neural mechanisms underlying the maintenance of visual working memory contents

Visual working memory (WM) refers to the maintenance and manipulation of visual information that is not in the visual field. Visual WM provides an essential connection between perception and higher cognitive functions. Previous fMRI studies with classification methods showed that early visual cortex could retain information of visual features maintained in WM. However, we still know little about the neural mechanisms of visual WM during maintenance period in higher visual cortex and the influence of memory load on such mechanisms. Therefore, the present study utilized a modified delayed match-to-sample paradigm to investigate the visual WM performance. Participants were required to maintain one or two target items (face or house) for 11s. At the same time, function magnetic resonance imaging (fMRI) was applied to acquire the neural activities in visual cortical areas. Behavioral results showed that memory load but not stimuli significantly affected the accuracy of memory. Next, the amplitude of the neural activity during maintenance in each area was extracted. In order to reveal the areas sensitive to mnemonic stimuli, we compared the activity induced by face and house WM. The results showed that parahippocampal place area (PPA) and intra-parietal sulcus (IPS) were activated differently to different stimuli. Specifically, the neural response was stronger when house relative to face was held in memory. To further investigate whether other areas also retained visual information in WM, activation pattern analysis with decoding method was applied to the neural response in each area. The results demonstrated decoding performance in face selective areas including superior temporal sulcus (STS), FFA and V4. Taken together, these results indicated that higher visual cortex could retain visual information held in WM. Specifically, PPA and IPS showed general stronger responses to mnemonic houses than mnemonic faces, while FFA, STS, and V4 contained different detailed representations of mnemonic faces and mnemonic houses. These findings were consistent with previous findings concerning visual features, and supported the view that the contents in visual working memory are represented in visual cortex. Notably, we did not find any effects of memory load on the neural activity in PPA and IPS. However, the activity pattern analysis about other areas showed significant effects only in high memory load condition, suggesting that high memory load might refine the representations in face selective areas without boosting the average activation level. In order to find the cortical areas of which responses were significantly modulated by memory load, we further performed a voxel-based whole brain analysis. Specifically, a 2 (stimuli: face, house) * 2 (memory load: 1 item, 2 items) repeated measures ANOVA was performed for each voxel on the neural response during the maintenance. No interaction was observed in each voxel. Interestingly, the main effect for memory load was significant in bilateral putamen. In summary, our findings revealed the crucial role of higher visual cortex in maintaining visual information in WM and the important role of putamen in modulating memory load.