CNS 2021 Virtual Meeting | Data Blitz Session Schedule

A Data Blitz is a series of 5-minute talks, each covering just a bite-sized bit of research. It will offer a fast-paced overview of some of the most exciting research presented at this year's poster sessions.

During the abstract submissions process first authors had the option to indicate whether they’d like their abstract to be considered for the Data Blitz Session. Selected abstracts have been scheduled into the Data Blitz session in addition to a poster session. Talks will be given by the abstract first author — a faculty member or student at any level. Accepted presenters will then be asked to prepare a concise, 5-minute video for the session.

Each DataBlitz session includes 15-5 minute talks with an additional 30-minute live Q&A following. We are asking all presenters to be available for the live Q&A directly after your scheduled DataBlitz session.

SESSION

DATE

TIME

LOCATION

CHAIR

Data Blitz Session 1 Saturday, March 13 11:00 - 12:30 pm Axon Room
Data Blitz Session 2 Saturday, March 13 11:00 - 12:30 pm Cerebrum Room
Data Blitz Session 3 Saturday, March 13 11:00 - 12:30 pm Dendrite Room
Data Blitz Session 4 Saturday, March 13 11:00 - 12:30 pm Synapse Room

Data Blitz 2021 Schedule

DATA BLITZ SESSION 1

TALK 1: Evaluation of common brain parcellations used in the a priori identification of primary resting-state networks

Nessa Bryce, Harvard University

Over the past decade there has been a proliferation of work examining the segregation of the human brain into large-scale functional networks. The resulting parcellation schemes are now commonly used as brain atlases for the a priori identification of functional networks. However, the use of these parcellations, particularly in developmental research, hinges on four fundamental assumptions that have not yet been evaluated. In the present study, we examined these four major assumptions. The resting-state functional scans from 113 participants ages 8-18 were used to evaluate eight commonly applied parcellations: Yeo 2011, Power 2011, Glasser 2016, Gordon 2016, Schaefer 2018 (100, 200, 400), as well as a Neurosynth-derived atlas. We found that the networks of interest (default, control, dorsal attention, and salience) were equally well captured in the parcellation-extracted data (Confirming assumption 1), with the exception the Neurosynth parcellation. Furthermore, we found that the primary networks are equally well identified in children's data as in adults (Confirming assumption 2). However, we found significant variability in network connectivity scores across the eight parcellations. Furthermore, there was low internal consistency (reliability) of connectivity scores within subjects across the various parcellations, suggesting that networks named the same across the parcellations (ie. 'default network') may not actually be synonymous (Disconfirming assumption 3). Finally, parcellation selection had a notable impact on results (Disconfirming assumption 4), such that, the association between poverty and default connectivity varied significantly as a function of parcellation, as did the association between age and default connectivity and between inhibition and control connectivity.

 

TALK 2: Dissociable neurophysiological correlates of the fMRI BOLD signal in the hippocampus and neocortex

Paul Hill, University of Arizona

The blood-oxygen-level-dependent (BOLD) signal forms the basis of fMRI but provides only an indirect measure of underlying neural activity. Task-related modulation of BOLD activity is typically equated with changes in gamma-band activity; however, relevant empirical evidence comes largely from the neocortex. We examined neurophysiological correlates of the BOLD signal in the hippocampus, where sparse vascularization and neural coding schemes might produce a more heterogeneous relationship between the respective signals. Fifteen neurosurgical patients implanted with depth electrodes performed a verbal free recall task while local field potentials were recorded simultaneously from hippocampal and neocortical sites. The same patients subsequently performed a similar version of the task during a later fMRI session. Subsequent memory effects (SMEs) were computed for both imaging modalities as patterns of encoding-related brain activity predictive of later free recall. Linear mixed-effects modelling revealed that the relationship between BOLD and high-gamma (70-150 Hz) SMEs was moderated by the lobar location of the recording site: BOLD SMEs positively covaried with high-gamma SMEs in frontal, temporal, and parietal cortex, but demonstrated a negative but nonsignificant relationship in the medial temporal lobe (hippocampus, parahippocampal gyrus). A subsidiary analysis of SMEs recorded from the medial temporal lobe revealed an interaction between region and high-gamma SMEs which was driven by a significant negative relationship between BOLD and high-gamma effects in the hippocampus, and a positive but nonsignificant relationship in the parahippocampal gyrus. These results suggest that the neurophysiological bases of the BOLD signal in the hippocampus differ from those observed in the neocortex.

 

TALK 3: DeepMReye: MR-based camera-less eye tracking using deep neural networks

Markus Frey, Kavli Institute for Systems Neuroscience

Viewing behavior provides a window into many central aspects of human cognition and health, and is an important variable of interest or confound in many fMRI studies. To make eye tracking freely and widely available for MRI research, we developed DeepMReye: a convolutional neural network that decodes gaze position from the MR-signal of the eyeballs. It performs camera-less eye tracking at sub-imaging temporal resolution in held-out participants with little training data and across a broad range of scanning protocols. Critically, it works even in existing datasets and when the eyes are closed. Decoded eye movements explain network-wide brain activity also in regions not associated with oculomotor function. This work emphasizes the importance of eye tracking for the interpretation of fMRI results and provides an open-source software solution that is widely applicable in research and clinical settings.

 

TALK 4: Cortical Volume Heterogeneity in Children with Psychopathology

Randolph Dupont, Vanderbilt University

Symptoms of psychopathology emerge in youth and are associated with abnormalities in brain structure. However, there is considerable heterogeneity in the brain regions implicated. Machine learning tools provide a data-driven approach to parse the neurobiological heterogeneity in psychopathology. The aim of the present study was to determine whether there are neurobiological subtypes of general psychopathology based on brain structure. To this end, we used clinical and cortical volume data from 9,607 9 to 10-year-old children collected in the Adolescent Brain Cognitive Development (ABCD) Study. To discern neurobiological subtypes, we used a semi-supervised machine learning algorithm (HYDRA; heterogeneity through discriminative analysis), which identifies subtypes within a patient group with general psychopathology symptoms in comparison to a control group. Using cortical brain volume as the input, cross-validation methods indicated 2 reliable subtypes of children with general psychopathology (adjusted Rand index = 0.45). Both Subtype 1 and 2 showed more internalizing, attention-deficit/hyperactivity, and conduct problem symptoms, as well as poorer cognitive functioning compared to controls, as expected (p-values <.025, FDR-corrected). Despite these similarities, Subtypes 1 and 2 diverged from each other, with Subtype 1 showing smaller cortical volumes and greater cognitive deficits than Subtype 2 (p-values <.025, FDR-corrected). Interestingly, Subtype 1 also showed greater levels of conduct problems than Subtype 2 (p < .002, FDR-corrected). The results of this study suggest that while psychopathology in general is associated with cognitive deficits, the co-occurrence of smaller brains and greater conduct problems is associated with the worst cognitive functions.

 

TALK 5: Trauma is Associated with Abnormalities in Cortical Thickness in Children

Emily Micciche, Vanderbilt University

Early life stressors such as trauma can impact early development due to the heightened plasticity of the developing brain. While trauma appears to influence brain structure and increase risk for psychopathology, previous research has limitations. Small sample size, case-control designs, and heterogeneous age groups have confounded our understanding of trauma's effects on the maturing brain. The aim of this study is to examine the effects of trauma on cortical thickness and gray matter volume (GMV) during early development using a large sample of 9 to 10-year-old children (N=9,270) from the Adolescent Brain Cognitive Development (ABCD) Study. Structural equation modeling revealed a latent trauma variable that was correlated with decreased cortical thickness in bilateral superior frontal gyri and right caudal middle frontal gyrus (p-values < .001, FDR-corrected) and increased cortical thickness in left isthmus cingulate and left posterior cingulate (p-values ? .027, FDR-corrected). Income and parental education were controlled for using a sensitivity analysis; these results corresponded with the primary analysis. However, GVM was not significantly associated with latent trauma. These findings suggest that trauma may increase vulnerability to abnormal development of cortical thickness in the frontal and cingulate regions.

 

TALK 6: Encoding Specificity, Updating and Musical Expertise in Working Memory for Melodies: An fMRI Study

Chandramalika Basak, University of Texas at Dallas

The main goal of this project was to determine whether past experiences in classical music influences encoding of novel melodies differently than that in novices, and how is it related to context reinstatement and the type of lures, based on neural activations of brain regions. Eighteen young adults with 0-24 years of classical music experience underwent an event-related fMRI task requiring continuous working memory updating. Participants heard a series of 7-note melodies arranged in pairs with 2- and 3-back test items intermingled in an unpredictable manner. Half the test trials were same as the paired trial (nonupdated target; T), and in the other half the melody of one pitch was altered (updated) either to a similar lure (SL) or very different lure (DL). Trials were presented either in the same context or in a different context. In addition, encoding trials were included as a baseline condition in the fMRI analyses. Years of expertise predicted overall accuracy. Although all participants showed significant activations in bilateral primary auditory regions during the task, when T, SL and DL trial types were contrasted with encoding, experts showed increased activation in a visual processing region during the most difficult condition, SL. Novices engaged this region for all trial types. For melodies in different context, expertise was correlated positively with increased activation in precuneus, a default-mode network region, suggesting that encoding specificity of melodic context in experts is likely relying on endogenously driven self-referential processing.

 

TALK 7: Pathological slow-wave activity in post-traumatic amnesia underpins impaired binding of information in working memory

Emma-Jane Mallas, Imperial College London

Associative binding is key to normal memory function and is transiently disrupted during post-traumatic amnesia (PTA) following traumatic brain injury (TBI). The reason for this impairment is unclear but may be caused by impact induced electrophysiological changes disrupting cortical communication. Electrophysiological abnormalities are often seen after TBI, including pathological low frequency delta waves. Here we test the hypothesis that misbinding of information in working memory (WM) is specifically related to the presence of slow-wave abnormalities that disrupt cortical communication. 30 moderate-severe acute TBI patients and 26 healthy controls were tested with a precision WM paradigm that required the association of object and location information. A novel entropy-based measure was calculated to quantify the degree of misbinding observed. Resting state electroencephalography was used to assess the electrophysiological effects of TBI in 17 patients and 21 healthy controls.  Patients in PTA showed increases in slow-wave activity and an increase in delta-alpha ratio. This ratio correlated with WM impairment and specifically with entropy-based measurements of misbinding. Fronto-parietal connectivity was increased in the theta band following all types of head injury and was not correlated with WM impairment. Both WM and electrophysiological abnormalities  normalised at 6 month follow-up, in keeping with a transient increase in slow-wave activity causing PTA and impairing WM function. A pathological shift towards lower frequency oscillations may offer mechanistic insight into WM binding deficits seen in PTA. These results suggest optimal delta to alpha ratio may be critical for successful feature integration in WM.

 

TALK 8: Pathological slowing of neuronal oscillations predicts cognitive decline in patients on the Alzheimer's disease spectrum

Alex Wiesman, Montreal Neurological Institute

Decades of electroencephalography (EEG) and magnetoencephalography (MEG) research has established a pattern of decreased high-frequency (i.e., alpha and beta) and increased low-frequency (i.e., delta and theta) neuronal oscillations in patients with Alzheimer's disease (AD). These changes have been suggested to represent a pathological slowing of cortical oscillatory activity in patients with AD. However, several limitations exist in this literature. First and foremost, limited spatial resolution has made it difficult to determine whether these effects are actually co-occurring in spatially-overlapping cortical regions, which would be implied by a slowing hypothesis. Second, this slowing effect has not been effectively modeled as a continuous metric that lends itself to statistical analysis alongside clinical measures of cognitive function. As such, it is not well established that any such slowing effect is pathological in nature, rather than compensatory. We leverage the high temporal and spatial resolution of source reconstructed MEG to provide evidence of a robust, spatially-resolved oscillatory slowing effect in biomarker-confirmed patients on the AD spectrum. This effect was strongest across a network of middle and medial temporal, inferior parietal, prefrontal, and cerebellar cortices bilaterally. Importantly, a stronger slowing effect in left-lateralized middle temporal and inferior parietal regions significantly predicted both general and domain-specific cognitive impairments in these patients, indicating that this widely-reported effect is indeed pathological in nature. These findings provide important new directions for emerging frequency-targeted clinical interventions, as well as a framework for systematically studying similar spatially-confined oscillatory slowing effects in other patient populations.

 

TALK 9: Temporal regions change their function from sensory processing towards cognition in deaf individuals

Barbara Manini, University College London

Recent research (Cardin et al., 2018; Ding et al., 2015) has found functional reorganisation toward visual working memory in the posterior superior temporal cortex (pSTC) of deaf individuals. The pSTC a region involved in auditory and language processing and multisensory integration in hearing individuals (Kaas, & Hackett, 1999). Executive functions typically recruit frontoparietal regions (D'Esposito, & Grossman, 1996), on the other hand, the low-level processing of the perceptual features is attributed to the sensory cortices (Christophel et al., 2017). Here we ask whether the temporal cortex of deaf individuals is recruited during executive function tasks and whether it is involved in cognitive control or more specific subcomponents of executive processing.  Deaf (N=25) and hearing (N=20) individuals took part in an fMRI experiment. We presented cognitive tasks tapping into four executive functions: working memory, planning, switching, inhibition. We performed an ROI analysis in superior temporal regions: pSTC, Heschl's Gyrus and planum temporale (PT). We found that deaf participants recruited temporal areas during switching, with the level of activity correlating with behavioural performance in the task.  These findings indicate that temporal regions of deaf individuals change their function towards specific cognitive processes, likely to be related to the flexible allocation of attention. Our results do not support a role in cognitive control for the deaf superior temporal cortex.

 

TALK 10: Dynamic perturbation during retention alters human multi-item working memory

Jiaqi Li, Peking University

Temporarily storing a list of items in working memory (WM) for subsequence goal-directed behavior is an important function in many cognitive processes. This ability has been posited to rely on dynamic representations during retention such as neural reactivations or a theta-gamma oscillation code. However, the causal evidence for the role of neural network dynamics in multi-item WM is still lacking. We use a temporal-correlated 'dynamic perturbation' approach to interfere with the underlying representations during retention and manipulate the relative memory strength of memory items. Six human behavior experiments confirm the validity of this manipulation approach. A continuous attractor model with short-term synaptic plasticity (STP) principles reproduces all the experimental results. The model indicates that the modification of multi-item synaptic efficacies accumulates over time and eventually leads to the changes in recall performance. Our results support the causal role of temporal dynamic neural network in mediating multi-item WM and offer a promising, non-invasive method to manipulate WM.

 

TALK 11: Long-term learning transforms prefrontal cortex selectivity during working memory

Jacob Miller, University of California, Berkeley

The neural representations supporting goal-directed behavior may change with learning. However, uncovering how long-term memory influences working memory (WM) is difficult because of inconsistencies across studies, species, and recording techniques. For instance, non-human primate (NHP) electrophysiology research finds that lateral prefrontal cortex (lPFC) circuitry maintains WM representations, while human neuroimaging suggests that WM content is instead stored in sensory cortices. To bridge between these findings, we implemented a unique, longitudinal functional MRI (fMRI) protocol to test the influence of long-term learning on WM processes in each of three human participants. Across three months, each participant was trained on (1) a serial reaction time (SRT) task, wherein complex fractal stimuli were embedded within probabilistic sequences, and (2) a delayed recognition task probing WM for trained or novel stimuli. Participants showed a stimulus learning benefit in the WM task for trained, but not novel, fractals. Neurally, a significant population of voxels increased in delay activity throughout lPFC, paralleling recent findings that spiking activity across NHP lPFC becomes more distributed with training (Qi et al., 2019). Individual voxels in lPFC also showed increased stimulus selectivity over time among the trained fractals. Pattern similarity analyses of WM delay activity demonstrate that across learning, item- and sequence-level representations emerged within lPFC but not in sensory cortices. This work helps reconcile disparate findings across species and scales, showing that with learning, human lPFC develops stimulus-selective properties during WM. Further, this dense sampling of memory substrates establishes novel evidence for long-term memory influences on WM maintenance.

 

TALK 12: Altered Brain Activity in Survivors of Pediatric Acute Lymphoblastic Leukemia with Executive Dysfunction

Kellen Gandy, St. Jude Children's Research Hospital

Functional magnetic resonance imaging (fMRI) and neurocognitive testing were obtained in 138 survivors (44% male; median [min-max] age = 13.5 [8.2-26.5] years; time since diagnosis = 7.6 [5.1-12.5] years) treated on the St. Jude Total 15 protocol. Executive function was assessed with standardized tests of working memory (Digit Span Backwards), fluency (D-KEFS Verbal Fluency) and cognitive flexibility (D-KEFS Trail Making). Impairment was defined as at least one score <10th percentile of age-standardized normative data. fMRI was obtained on a 3T scanner during an N-back working memory task. Functional data were preprocessed (realigned, slice time corrected, normalized and smoothed) and analyzed using Statistical Parametric Mapping with contrasts developed for the 0-back, 1-back vs 2-back, and 2-back conditions which reflect varying degrees of working memory and task load. Of the 138 survivors, 52 (38%) demonstrated impaired executive function on neurocognitive assessments and functional neuroimaging data was compared with between survivors with and without impaired executive function. Survivors with impaired executive function displayed less activation in the left dorsal lateral prefrontal cortex during increased work memory task load (i.e., 2-back vs 1-back), compared to survivors without impaired executive function (p < 0.001, adjusting for age and sex). This pattern suggests that survivors are at risk for disrupted or delayed functional development of frontal lobe, which is associated with impaired executive function.

 

TALK 13: Task network effects are specialized to individuals: lessons from machine learning applied to precision fMRI

Alexis Porter, Northwestern University

Recent precision neuroimaging work has identified unique brain characteristics that correspond to differences in task network effects at the individual level. However, to what extent are task networks specific to a given individual. In this work we examined the role of individual differences and task general modifications by training a machine learning algorithm to identify cognitive states along a single subject's multi-session data. We utilized the Midnight Scan Club, a precision fMRI dataset containing 10 subjects scanned across 10 sessions under 5 different cognitive states (rest, 4 tasks). We trained a classifier to discriminate functional connectivity obtained during task(s) from functional connectivity obtained during rest using a leave one out cross validation scheme. We then assessed the models ability to label the unseen single subject test set from the same subject, or a different subject in the same task. We found that the model was able to perform above chance across all within and between subject analysis, with significantly higher classification accuracy within subject. We then examined the model's ability to generalize features to a different task. We found some evidence for generalizable task modifications but this largely varied across the training task. This work suggests that between-subject approaches are blind to major sources of variation in brain networks across tasks. Moreover, individualized precision approaches have the potential to improve our understanding of how brain networks operate across different states.

 

TALK 14: Theta oscillations shift towards optimal frequency for cognitive control

Mehdi Senoussi, Ghent University

Humans' capacity to pursue goal-oriented behavior despite changes in the environment is uniquely flexible. Cognitive control refers to processes allowing such behavioral adjustments and critically relies on neural oscillations in the theta band (4-7Hz). Theta amplitude has been shown to increase when control is needed, however, it remains largely unknown how theta oscillations support flexible adaptation to task demands. In this study we show that an essential aspect of theta oscillations has been completely overlooked so far: its specific frequency in the 4-7Hz range. We built a novel computational model in which theta oscillations orchestrate control over sensory and action representations, by establishing task-relevant functional networks through synchronization. Critically, we show that the frequency of theta oscillations balances reliable set up of task rules and gating of task-relevant information. Our model additionally predicts that this theta-rhythmic process is observable in behavioral performance, which constitutes, to our knowledge, the first time this phenomenon is predicted from computational simulations. We tested these predictions using a stimulus-response mapping task, and recorded EEG, in 33 participants. We showed that both behavioral performance and mid-frontal theta activity oscillated at slower frequency with increasing task difficulty. Furthermore, we showed that this shift in neural theta predicted trial-by-trial behavioral performance and that the sensitivity of theta frequency to task demands predicted overall task performance across participants. Our study provides a novel computational framework proposing how theta oscillations mechanistically implement cognitive control and adapt to external demands, opening new avenues for research on the regulation of cognitive control.

 

TALK 15: Examining interdependence and independence among processes supporting hierarchical cognitive control

McKinney Pitts, Florida State University

Cognitive control involves using internalized representations to perform non-habitual, goal directed behaviors.  Cognitive control can be fractionated into processes that prepare for the future (temporal control), and those that adapt behavior to present circumstances (contextual control). Neuroimaging paradigms have examined the neural correlates of these processes localizing temporal control to the rostral-most prefrontal cortex (PFC), and contextual control to mid-lateral PFC (Badre & Nee, 2018). An open question regards whether control processes are consistently ordered such that some control processes naturally drive others. On one hand, data supporting the Cascade Model has found a rostral-to-caudal cascade of control indicating a hierarchical apex in the rostral PFC. By contrast, Nee & D'Esposito (2016) found evidence for a more center-focused apex in the mid-lateral PFC. These inconsistencies may reflect differences in the paradigms studied ‚?? data supporting the Cascade Model utilized paradigms wherein contextual control (mid-lateral PFC) depended upon temporal control (rostral PFC), whereas Nee & D'Esposito utilized a paradigm wherein control processes acted independently. To investigate whether dependence among control processes affects hierarchical ordering, we modified the paradigm of Nee & D'Esposito to make contextual control demands depend upon temporal control. Behavioral measures verified significant costs related to both control processes. However, despite dependence among control processes in the paradigm, behavioral costs were additive, indicating independence among the underlying processes. Preliminary fMRI data indicate that despite independence in behavioral signatures of control, activations overlapped contradicting prior data. These patterns suggest that mappings of activations to control processes may require reconceptualization.

DATA BLITZ SESSION 2

TALK 1: Reward prediction error signals in the amygdala selectively guide social learning under uncertainty

Amrita Lamba, Brown University

Learning about others' trustworthiness often serves as an informative cue of social rewards, particularly in highly uncertain situations. While considerable research efforts have focused on nonsocial reward learning, much less is known about the neural computations supporting social reward learning. More specifically, how do humans learn to trust? Here we leverage computational neuroimaging approaches to investigate how individuals learn about information along value and uncertainty dimensions. We then examine how this knowledge is exploited to build a representation of trustworthiness that can adaptively guide social choice. Using the Trust Game (TG) and a matched nonsocial slot machine (SM) task, we examined how swiftly participants (N = 28) learned to make reward-maximizing choices across these contexts. Social partners and slot machines were preprogrammed to drift in their trustworthiness/monetary reward across the task. Compared to nonsocial SMs with matched reward distributions, results reveal that participants were faster to learn and experienced fewer reward prediction errors (RPEs) when interacting with consistently trustworthy or selfish partners, which correlated with activation of the striatum. In contrast, when interacting with social partners who indiscriminately showed trustworthy and untrustworthy behavior (i.e., highly uncertain behavioral profiles), participants were slower to learn and experienced greater RPEs during the TG compared to the SM. RPEs associated with learning about uncertain social partners uniquely scaled with activity in the amygdala, whereas the striatum indexed learning about uncertain SMs. These results suggest that context-selective representation of RPEs in the amygdala and striatum may drive asymmetrical learning profiles across social and nonsocial environments.

 

TALK 2: Inter-Network Neural Connectivity Mediates Intuitive Moral Decision-Making between Younger and Older Adults

Shenyang Huang, Duke University

Older adults (OAs) occupy many positions of power and constitute an increasingly larger share of the population. These demographic shifts underscore the importance of studying age-related changes in decision-making, particularly when it comes to difficult, morally-laden scenarios. However, little is known about age-related differences in moral decision-making and their relationship to the intrinsic network architecture of the brain. In the present study, younger adults (YAs; n = 117, Mage = 22.11) and OAs (n = 82, Mage = 67.54) made decisions on multiple hypothetical moral dilemmas and completed resting-state multi-echo fMRI scans. Relative to YAs, OAs were more likely to endorse deontological decisions and favor an action consistent with a moral principle, but only when the deontological moral choice was intuitive. By contrast, OAs' decision-making did not differ from that of YAs when the utilitarian moral choice was intuitive. Enhanced connections between the posterior medial core of the default network (pmDN) and dorsal attention network , as well as overall reduced segregation of the pmDN from the rest of the brain, partially mediated this increased deontological-intuitive moral decision-making style in OAs. The present study provides novel insight into the differential network connectivity associated with moral reasoning in YAs and OAs, expanding our understanding of the diversity of neurocognitive changes that accompany aging.

 

TALK 3: Scene Construction Ability and White Matter Integrity in Post-Traumatic Stress Disorder

Hannah Marlatte, Rotman Research Institute

People with post-traumatic stress disorder (PTSD) report highly vivid yet disjointed memories that are prone to amnesic gaps. This paradox has been attributed to dysfunction in the hippocampus (HPC), consistent with evidence that the HPC is required for mental construction of coherent scene representations. Methods: Twenty-six trauma survivors, half diagnosed with PTSD, conjured-up then described novel scenes. Scene construction ability was measured on the quality of scenes and their content. Participants underwent structural T1/T2 and diffusion-tensor MRI to quantify gray and white matter integrity. In a multiple factor analysis, we investigated the relationship between scene construction ability, neural integrity and symptom severity. Results: The first component explained 26% of the shared variance. Participants who generated fewer scene details, with less spatial coherence between them, had smaller HPC volumes and had more severe PTSD symptoms. Surprisingly, this was also related to having better right cingulum integrity between cingulate and parahippocampus. The second component explained 18% of the variance. Less vivid scenes-along with deficient spatial coherence-was associated with poorer integrity of long association tracts and tracts connecting the thalamus and cingulate. This was correlated with avoidance symptoms. Conclusions: Our results suggest that different aspects of scene construction are related to PTSD severity and HPC volumes, but also to changes in WM between the cingulate, thalamus and parahippocampus. Spatial integration and detail generation deficits together were related to greater parahippocampal WM integrity. Those who recalled scenes as both highly vivid and spatially integrated had less avoidance and greater subcortical WM integrity.

 

TALK 4: Evaluating EEG-Based Breathing Entrainment Classifiers for Anxiety in Adolescents with Autism

Avirath Sundaresan, The Nueva School

Autism spectrum disorder (ASD) is a set of neurodevelopmental disorders which can often be highly disabling, severely limiting everyday opportunities and overall quality of life. Given the high prevalence of anxiety in adolescents with ASD in conjunction with the associated detrimental life-long consequences, personalized treatment interventions to relieve clinical symptoms and enhance long-term physical and emotional well-being are needed. Concurrently, mindfulness and breathing practices have proven effective at mitigating anxiety and procuring significant health benefits. Although controlled slow breathing bears potential as a low-cost and non-pharmacologic intervention to mitigate anxiety, effectiveness hinges on assessing an individual's current level of anxiety and optimal respiration parameters in real-time.   In recent years, electroencephalogram (EEG)-based brain-computer interfaces (BCIs) have made real-time cognitive monitoring increasingly possible. Among the vast corpus of existing machine learning classifiers we selected conventional BCI methods and compared them with novel deep learning models in order to evaluate the feasibility of an EEG-based BCI for the real-time assessment and mitigation of anxiety through a closed-loop adaptation of respiration entrainment. We trained a total of eleven subject-dependent models - four with conventional BCI methods and seven with deep learning approaches - on the EEG of neurotypical (n=5) and ASD (n=8) participants performing alternating blocks of mental arithmetic stress induction, guided and unguided breathing. To the best of our knowledge, our study is the first to propose a multiclass two-layer LSTM RNN deep learning classifier capable of identifying stress-induced anxious states from ongoing EEG with an overall accuracy of 93.27%.

Supported by: R01DC017991, R01DC016765, R01DC016915, R01MH111419.

 

TALK 5: A data-driven neural map of the affective valence action space.

Keith Bush, University of Arkansas for Medical Sciences

As part of an ongoing control theoretic exploration of affect regulation, we derived a neural map that encodes control actions driving affective valence dynamics. To account for affect regulation processing confounds, we performed conjunction analysis between neural encodings of valence dynamics derived from functional magnetic resonance imaging (fMRI) data recorded from (n=40) age- and sex-matched healthy adult subjects (aged 18-64) during both explicit affect regulation and resting state tasks. Explicit affect regulation task trials (n=30 per subject) were comprised of International Affective Picture Set (IAPS) image stimuli (2 s) succeeded by control steps (8 s) in which subjects volitionally re-experienced the perceived affect of the stimuli while attending to a fixation symbol. Valence was predicted for all task volumes according to previously reported fMRI-derived machine learning models fit separately to each subject using unique IAPS stimuli. We computed temporal derivatives of these valence time-series and then fit (and cross-validated) similar fMRI-based predictive models to these derivatives. We validated these models by demonstrating that numerical integration of the predicted derivatives produced significantly less error (compared to the observed time-series) than integration of random derivatives drawn from the same distribution (p<.05). Following Haufe-transformation of the decoding models, we computed group-level voxel-wise t-statistics of the encodings followed by conjunction analysis and clustering (p<.01 threshold; p<.05 cluster-size correction). The resulting neural map includes regions associated with affective reactivity and regulation including posterior cingulate cortex, precuneus, right posterior medial temporal gyrus, right dorsolateral prefrontal cortex, and bilateral superior parietal lobules.

 

TALK 6: Amygdala reactivity as neural correlate of negative memory bias

Fleur Duyser, Radboudumc

Enhanced memory for negative self-relevant information is one of the main cognitive symptoms of depression. Recent evidence suggests that this self-referent negative memory bias is also present in other psychiatric disorders. Although the neural underpinnings of this bias are still largely unclear, previous findings point towards increased amygdala reactivity as a potential neural correlate. In this study, we therefore aimed to assess the association between amygdala reactivity and self-referent negative memory bias as well as its two components: negative endorsement, which reflects negative self-referential processing, and negative recall, which reflects the preferential memory for negative information. We examined this in a naturalistic psychiatric patient sample (N = 125) of different (comorbid) stress-related and neurodevelopmental psychiatric disorders. We used the self-referent encoding task to assess negative endorsement, negative recall, and their composite score: self-referent negative memory bias. Participants performed an fMRI emotion processing task to measure amygdala reactivity. Associations were assessed using linear regression models. Surprisingly, we found a significant negative relationship between amygdala reactivity and self-referent negative memory bias in a subsample of patients (n = 58) who actually had this bias. There was no association between amygdala reactivity and negative endorsement, but increased amygdala reactivity was related to more negative recall. These findings indicate that endorsement and recall represent intrinsically different processes with different underlying mechanisms. To conclude, our study shows that amygdala reactivity might be a neural correlate of negative memory bias, not only in depression, but in a wide range of different psychiatric disorders.

 

TALK 7: Uncovering the association between individual variations in emotional awareness and brain emotional reactivity by IS-RSA

Yun-Jie Wu, National Taiwan University

Abnormal emotional reactivity is one of the most prominent features among several mental illnesses. Numerous studies have found a great amount of individual variations in the brain emotional reactivity and attempted to link these variations to particular psychopathology or personality traits. In our study, we were interested in testing whether individual variability in the emotional reactivity represented in the multivariate pattern was associated with individual variability in traits linked to the emotional awareness. Fifty-three participants were recruited in this study and underwent one run of an emotional reactivity task presented in the block design with positive-, negative- and neutral-emotion conditions. We also used a widely used scale, the Mindfulness Awareness Attention Scale, which measures individual variations in the awareness of present experience to measure individual differences in emotional awareness. We then used intersubject representation similarity analysis (IS-RSA) to test whether intersubject similarity in the brain spatial representations to three different emotional conditions was associated with similarity in emotional awareness. Our results showed that individual variations in emotional awareness were associated with variations in the brain representation within the prefrontal executive control network when participants viewed positive and negative images, whereas no such association was found when they viewed neutral images. Our findings demonstrated that by using IS-RSA, researchers can have new understandings of emotional awareness and its relationship to emotional reactivity, which is essential for individuals to pursue a stable emotional state and mental well-being.

 

TALK 8: Locus coeruleus-related insula activation supports saliency processing across the adult lifespan

Martin Dahl, Max Planck Institute for Human Development

The quick and reliable processing of salient information is crucial for goal-directed behavior, yet with advancing age, deficits emerge in attention. Decrements in attention have been associated with altered noradrenergic activity in animals, however methodological challenges have long impeded in-vivo human research. We measured pupil dilation, a non-invasive marker of noradrenergic neuromodulation, while younger (N = 39) and older adults (N = 38) completed an arousal-modulated oddball task in the scanner. Arousing stimuli elicited a pronounced dilation of the pupil and activation of the dorsal attention and saliency networks relative to perceptually matched control stimuli. Larger pupil dilations were related to a greater activation in the anterior insula, indicating an influence of noradrenergic neuromodulation. Task-based functional connectivity analyses showed a close interconnection of the insula and other parts of the saliency network such as the thalamus and anterior cingulate cortex. Crucially, connectivity analyses further revealed an association between the anterior insula and the locus coeruleus, the main source of cortical norepinephrine. Structural magnetic resonance imaging (MRI) corroborated these conclusions by demonstrating a positive relation between locus coeruleus MR contrast ratios, a proxy for noradrenergic cell density, and anterior insula activation. Finally, multilevel modeling confirmed the behavioral relevance of moment-to-moment fluctuations in locus coeruleus-related saliency signals, particularly in aging. Overall, our results suggest a prominent role of noradrenergic neuromodulation in saliency processing and attention across the lifespan.

 

TALK 9: Too little, too late, and in the wrong place: the Alpha band does not reflect an active mechanism of selective attention

Soren Andersen, University of Aberdeen

Selective attention focuses visual processing on relevant stimuli in order to allow for adaptive behaviour despite an abundance of distracting information. It has been proposed that increases in alpha band (8?12 Hz) amplitude reflect an active mechanism for distractor suppression. If this were the case, increases in alpha band amplitude should be succeeded by a decrease in distractor processing. Surprisingly, this connection has not been tested directly; specifically, studies that have investigated changes in alpha band after attention-directing cues have not directly assessed the neuronal processing of distractors. We concurrently recorded alpha activity and steady-state visual evoked potentials (SSVEPs) to assess the processing of target and distractor stimuli. Participants covertly shifted attention to one of two letter streams (left or right) to detect infrequent target letters 'X' while ignoring the other stream. In line with previous findings, alpha band amplitudes contralateral to the unattended location increased compared to a pre-cue baseline. However, there was no suppression of SSVEP amplitudes elicited by unattended stimuli, while there was a pronounced enhancement of SSVEPs elicited by attended stimuli. Furthermore, and crucially, changes in alpha band amplitude during attention shifts did not precede those in SSVEPs and hit rates in both experiments, indicating that changes in alpha band amplitudes are likely to be a consequence of attention shifts rather than the other way around. We conclude that these findings contradict the notion that alpha band activity reflects mechanisms that have a causal role in the allocation of selective attention.

 

TALK 10: Heart states are associated with altered perception of brief temporal durations

Saeedah Sadeghi, Cornell University

Early models of time perception theorized a role of the autonomic nervous system in perceived duration. Pacemaker-accumulator models suggest that the rate of the internal pacemaker depends on bodily markers of arousal, which may also shape our perception of time through interoceptive processes. Here we examined the relationship between heart and time perception for durations shorter than half a single heartbeat. Our stimuli were too brief that they resembled spikes causing perturbations in the heartbeat sequence signal. Individuals (N=45) made binary short/long choices in a temporal bisection of brief tones in the range of 80-188 ms synchronized to heart beats. Post-stimulus heart rate deceleration was associated with a temporal discrimination bias, reflecting a physiological marker of phasic attentionial orienting toward the stimulus. Given the same objective duration, temporal perceptions were subjectively longer vs shorter when associated with greater vs. lesser heart rate deceleration.  Higher pre-stimulus heart rate, shorter response time, and shorter duration of cardiac orienting reflex were all intercorrelated and each significantly predicted higher temporal discrimination sensitivity. These results suggest that pre-stimulus heart rate, response time, duration of orienting reflex, and temporal sensitivity, all share one underlying source which reflects fluctuating states of attentional preparedness. In conclusion, our results show that heart rate based physiological biomarkers of attention relate to time perception, tracking temporal bias and temporal sensitivity, through stimulus dependent and independent mechanisms, respectively.

 

TALK 11: Attentional modulation in early visual cortex: a combined re-analysis of steady-state visual evoked potential studies

Nika Adamian, University of Aberdeen

Steady-state visual evoked potentials (SSVEPs) are a particularly powerful tool for investigating selective attention. The SSVEP is a continuous oscillatory response of the visual cortex that has the same fundamental frequency as the driving stimulus and whose amplitude is increased with attention to the driving stimulus. When multiple stimuli flickering at different frequencies are presented concurrently, each one of them will drive an SSVEP at its respective frequency, thereby allowing for the assessment of the allocation of attention to each element in a multi-stimulus display. Here we combined the data of eight published SSVEP studies in which participants (n=139 in total) attended to flickering random dot stimuli based on their defining features (e.g. location, color, luminance, or orientation) or feature-conjunctions. The reanalysis first established that in all the studies attention reliably enhanced amplitudes and shortened latencies of SSVEPs, with colour-based attention providing the strongest effect. Next we investigated the modulation of SSVEP amplitudes in a subset of studies where two different features were attended concurrently. While most models of feature-based attention assume that multiple features are combined additively, our results suggest that neuronal enhancement provided by concurrent attention is better described by multiplicative integration. Finally, we used the combined dataset to demonstrate that the increase in SSVEP amplitudes cannot be explained by the synchronization of single-trial phases. Contrary to the prediction of the phase locking account, the variance of complex Fourier coefficients increases with attention, which is more consistent with boosting of largely phase-locked signal with non-phase-locked noise.

 

TALK 12: Visual processing speed is linked to functional connectivity between right frontoparietal and visual networks

Adriana Ruiz-Rizzo, LMU Munich, General and Experimental Psychology

Visual information processing requires an efficient visual attention system. The neural theory of visual attention (TVA) proposes that visual processing speed depends on the coordinated activity between frontoparietal and occipital brain areas. Previous research has shown that the coordinated activity between (i.e., functional connectivity, 'inter-FC') cingulo-opercular (COn) and right-frontoparietal (RFPn) networks is linked to visual processing speed. However, how inter-FC of COn and RFPn with visual networks links to visual processing speed has not been directly addressed yet. Forty-eight healthy adult participants (27 females) underwent resting-state (rs-)fMRI and performed a whole-report psychophysical task. To obtain inter-FC, we analyzed the entire frequency range available in our rs-fMRI data (i.e., 0.01-0.4 Hz) to avoid discarding neural information. Following previous approaches, we analyzed the data across frequency bins (Hz): Slow-5 (0.01-0.027), Slow-4 (0.027-0.073), Slow-3 (0.073-0.198), and Slow-2 (0.198-0.4). We used the mathematical TVA framework to estimate an individual, latent-level visual processing speed parameter. We found that visual processing speed was negatively associated with inter-FC between RFPn and visual networks in Slow-5 and Slow-2, with no corresponding significant association for inter-FC between COn and visual networks. These results provide the first empirical evidence that links inter-FC between RFPn and visual networks with the visual processing speed parameter. These findings suggest that direct connectivity between occipital and right frontoparietal, but not frontoinsular, regions support visual processing speed.

 

TALK 13: Neural and behavioral signatures reveal the temporal trajectory of auditory distraction

Troby Lui, University of L¸beck

Attentional sampling operates in a rhythmic manner. However, it is less well studied whether distractor suppression is modulated across time as well. In the present behavioral and electroencephalography (EEG) study, we probed the temporal trajectory of distraction in auditory attentional filtering in a sample of N = 30 human participants. In a pitch comparison task, we systematically varied the onset time of a task-irrelevant 25-Hz modulated tone sequence, which was presented in-between two to-be-compared target tones. Two metrics of distraction were utilized. First, on the level of behavior, perceptual sensitivity (d') in pitch comparison inversely relates to the degree of distraction, with decreasing sensitivity signifying increasing distraction. Second, on the neural level, distractor-evoked low frequency (1-8 Hz) inter-trial phase coherence (ITPC) is thought to quantify distractor encoding. Behavioral results evidenced a robust distractor effect, in which perceptual sensitivity was lower in trials with a distractor, compared to trials without a distractor (mean difference in d' = 0.597, SE = 0.074). Distractor onset time modulated both perceptual sensitivity and ITPC at frequencies below 8 Hz. Importantly, a significant relation between perceptual sensitivity and ITPC across distractor onset times suggests that neural processing of distraction relates to behavioral sensitivity. These results have important implications for the temporal dynamics of the auditory attention filter. In particular, the present study supports the hypothesis that the vulnerability to distraction is not constant, but instead modulated across time.

 

TALK 14: The Influence of ADHD Symptoms and Gamification on Attention-based Performance

Jessica Schachtner, University of California, San Francisco

Literature suggests that youth with ADHD perform worse on traditional attention tasks, but equivalently or better on gamified attention tasks and entertainment video games, than neurotypicals. However, such effects in adults across the lifespan are still unknown. Here, we explored whether gamification influences adult attention-based performance, and how ADHD symptoms moderate these effects. Via mTurk, 94 participants aged 21-71 played ungamified and gamified versions of a Continuous Performance Task (CPT) measuring sustained attention and impulsivity. During the tasks, participants responded to target stimuli and withheld responses to non-target stimuli. In the ungamified version, stimuli were white squares. In the gamified version, target stimuli were fish and non-target stimuli were 'junk'. Other gamification elements included a robust storyline, background sounds, and audiovisual feedback. Participants also filled out the Adult ADHD Self-Report Scale Symptom Checklist.  There was a positive correlation between ADHD symptoms and gaming benefits on response time variability (RTV) and D prime (D') performance for the impulsivity condition (p=0.007 and p=0.077, respectively), suggesting that gamification may be more advantageous for those with more ADHD symptoms. There was also a negative correlation between age and gaming benefits on RTV and D' performance for sustained (p=0.004 and p=0.023, respectively) and impulsivity (p=0.010 and p=0.007, respectively) conditions, suggesting that gamification may be more advantageous for younger populations. These findings not only validate previous ADHD literature on gamification effects, but also expand upon understanding these influences across the lifespan. Future work will examine the different components of gamification that may be driving these effects.

 

TALK 15: Alpha Power over Right/Mid-Frontal Brain Regions Support the Generation of Remote Associations in Higher Creative Indivi

Yoed Kenett, Technion - Israel Institute of Technology

Little is known about the neural mechanisms underlying the spontaneous generation of creative ideas. We aimed to investigate the role of alpha oscillations during the production of spontaneous remote associations, and how they vary in relation to individual differences in creative ability. Participants were presented with a stimulus word and were asked to produce as many associative responses as possible in 2 minutes to a set of cue words, while having their EEG recorded. Participants also underwent a battery of creativity tests, which were used to divide them into lower and higher creativity groups. To estimate the semantic proximity of the word streams that participants produced, we used forward flow (www.forwardflow.org) which employs co-occurrence statistics of words in textual corpora to compute the semantic distance between consecutive associative responses. At the behavioral level, higher creative participants generated significantly more responses compared to lower creative participants. While higher creative participants did not generate more remote associations than lower creative participants, forward flow was positively correlated with general creativity. At the neural level, the generation of semantically distant concepts was associated with higher alpha frequency activation over right and mid frontal areas, but only for higher creative participants. This activation has been previously linked to increased cognitive control and inhibition mechanisms of the prefrontal cortex. Thus, we suggest that enhanced alpha oscillations at right/mid-frontal areas relate to the spontaneous generation of semantically remote concepts, uniquely in higher creative individuals.

DATA BLITZ SESSION 3

TALK 1: Structural connectome and lesion-based predictors of reading in subacute left-hemisphere stroke.

Olga Boukrina, Center for Stroke Rehabilitation Research

Reading relies on a distributed network of brain areas subserving word-form recognition (orthography), letter-to-sound mapping (phonology), and auditory and visual word-form to meaning mapping (semantics).  Stroke can directly impact these areas or result in disconnections among multiple areas not directly affected by a lesion. In the aphasia literature, there is a lack of consensus on the prognostic value of the structural connectome in predicting post-stroke impairments compared to lesion volume and location data. While some studies show that lesion and connectome-based predictors of language deficits are comparable in accuracy, others report no advantage for connectivity-based prediction over lesion data alone on a wide range of language tests. Using multimodal neuroimaging data from 37 left-stroke patients undergoing acute rehabilitation, we examine the contribution of structural lesions and white matter (WM) connectivity to reading impairments defined by orthography, phonology, and semantics competence. Voxelwise lesion analysis controlling for lesion volume identified supramarginal, inferior temporal and fusiform gyri, and insular white matter as areas of damage most associated with phonological impairments. Lateral occipital, middle frontal, parahippocampal cortex, precuneus, putamen, and the splenium were associated with semantic impairments. Fusiform, supramarginal, insular cortex, and white matter underlying the inferior frontal cortex were linked with orthographic impairments. Our findings re-emphasize the importance of major white matter (WM) tracts in reading through WM integrity-behavior associations for orthography (inferior fronto-occipital fasciculi (IFOF)), phonology (superior and inferior longitudinal fasciculi (SLF, ILF), and semantics (SLF, IFOF, cingulum, corpus callosum). Disruption of this connectivity is associated with specific dimensions of reading.

 

TALK 2: Non-literal language processing is jointly supported by the language and Theory of Mind networks

Miriam Hauptman, New York University

Going beyond the literal meaning of utterances is key to communicative success. However, the mechanisms that support non-literal inferences remain debated. We use a meta-analytic approach to evaluate the contribution of linguistic, social-cognitive, and executive mechanisms to non-literal interpretation. We identified 73 fMRI experiments (n=1,358 participants) from 2000-2019 that contrasted non-literal language processing with a literal control condition, spanning ten phenomena (e.g., sarcasm, metaphor). Applying the activation likelihood estimation approach to the 792 activation peaks yielded six clusters located in the frontal cortex bilaterally, and in the left temporal cortex. We then evaluated the locations of these clusters against probabilistic functional atlases (cf. macro-anatomy, as typically done) for three distinct candidate brain networks-the core language-selective network (Fedorenko et al.,2011), the Theory of Mind (ToM) network (Saxe & Kanwisher,2003), and the domain-general Multiple-Demand (MD) network (Duncan,2010). These atlases were created by overlaying individual activation maps of participants who performed 'localizer' tasks targeting each network (n=220 for language; n=119 for ToM; n=197 for MD). Our analysis revealed that 3 cluster peaks fell within the language network, 2 peaks fell within the ToM network, and 1 peak overlapped with both the language and the ToM network and could not be unambiguously assigned to either. The results suggest that non-literal processing is supported by both i) mechanisms that process literal linguistic meaning, and ii) mechanisms that support general social inference. They thus undermine a strong divide between literal and non-literal aspects of language and challenge the claim that non-literal processing requires additional executive resources.

 

TALK 3: Natural Language Processing Approaches to the Classification of Primary Progressive Aphasia

Neguine Rezail, Massachusetts General Hospital, Harvard Medical School

Primary Progressive Aphasia (PPA) is a clinical neurodegenerative syndrome characterized by abnormalities in language with initial relative sparing of other cognitive processes. The non-fluent variant (nfvPPA) is characterized by effortful speech and agrammatism; the logopenic variant (lvPPA) by difficulties in sentence repetition and lexical retrieval; the semantic variant (svPPA) by deficits in object naming and word comprehension. While widely used, this classification system has been criticized. Here we sought to provide a theory-neutral approach to the specification of PPA using state-of-the-art methods from natural language processing.  Language data was obtained from 78 patients with PPA (28 nfvPPA, 26 lvPPA, 24 svPPA) describing a picture. A transformer model, RoBERTa, was used to measure similarities in language features. Patients were clustered using the IVIS dimensionality reduction and OPTICS clustering algorithms. Regions of shared cortical atrophy were identified using logistic regression and recursive feature selection. Seven variants of PPA were identified with 81% agreement with the classic classification system. Individuals in Clusters 1 and 2 (mainly svPPA) exhibited deficits in nouns/verbs access with atrophy in the left temporal pole and inferior and middle temporal gyri. Those in Clusters 3 and 4 (predominantly lvPPA) exhibited difficulties in subject-verb agreement, demonstratives and tense and shared atrophy in the left supramarginal and angular gyri. Individuals in Clusters 5-7 (mainly nfvPPA) exhibited speech disfluency and reduced clausal complexity with atrophy in the left caudal middle frontal, pars opercularis, and pars triangularis gyri.  Artificial intelligence can advance our understanding of the behavioral and neuroanatomical characteristics of PPA.

 

TALK 4: Modeling pronoun resolution in the brain

Jixing Li, New York University Abu Dhabi

The current study investigates the neural mechanisms of pronoun resolution using computational models that lay out specified and carefully thought-out steps to achieve pronoun resolution. We selected three symbolic models that formalize three influential theories on pronoun resolution: The syntax-based Hobbs model implements the classic Binding Theory in formal linguistics; the discourse-based Centering model implements the Centering Theory that views pronominalization as a means to achieve discourse coherence; the memory-based ACT-R model conforms to the salience account for pronoun resolution, and selects the most highly-activated entity in the working memory as the antecedent of the pronoun. We also included one data-oriented deep neural network model that learns the statistical patterns behind coreferential entities from a labeled dataset. To compare model predictions with brain activities, we recorded BOLD signals while both English and Chinese participants listened to a 100-minute audiobook of 'The Little Prince' in the fMRI scanner. We also collected source-localized MEG data while English speakers listened to a 12-minute audio excerpt from the YouTube channel 'SciShow Kids'. We applied both multivariate RSA and univariate GLM analyses to compare the four models' relatedness to the fMRI and MEG data time-locked at each third person pronoun in the narratives. Our combined results suggest that the memory-based ACT-R model best explains the neural signatures for third person pronoun processing, primarily localized at the left middle temporal gyrus (LMTG) at around 300-400 ms after the onset of the pronoun. We propose a domain-general mechanism for pronoun resolution that resembles memory retrieval.

 

TALK 5: Correlated brain indexes of semantic prediction and prediction error: brain localization and category specificity

Luigi Grisoni, Freie Universit‰t, Brain Language Laboratory

Accurate predictions allow to understand a message easily and quickly, whereas unpredictable utterances require more processing. Consistently, previous evidences have shown a linear relationship between anticipatory signals occurring before predictable stimuli (Prediction Potential) and post-stimulus responses (Mismatch Negativity, MMN). However, since MMN paradigms are not ecological, as they are characterized by a highly redundant stimulus presentation, it remains to be investigated whether similar mechanisms also occur in situations closer to everyday experiences. We here demonstrate the interplay between prediction and perception during sentence comprehension. Sentence fragments constraining the expectation of a specific word induced anticipatory brain activity before the expected input; this slow-wave potential was absent in case of weak expectations. That this anticipatory slow wave was related to predictive processing was further demonstrated by correlations between this signal and both subjective reports of certainty about upcoming words and objective corpus-based measures, thus confirming this response as a semantic prediction potential (SPP). Furthermore, an inverse correlation between the SPP and the following N400 brain response suggested the interpretation of the N400 as a prediction error response. The sources underlying the pre- (SPP) and post-stimulus (N400) responses were located in inferior prefrontal and posterior temporal cortices, respectively. In addition, category-specific clusters of activation in modality-preferential visual and motor brain areas for animal- and tool-related words, respectively, indicated that both measures reflected aspects of sentence meaning. Overall, these data reveal that the N400, has an antecedent, the SPP which may determine the N400 dynamics.

 

TALK 6: Generalizable predictive modeling of semantic processing ability from functional brain connectivity

Danting Meng, South China Normal University

Semantic processing (SP) is a critical ability to humans for representing and manipulating meaningful information. Neuroimaging studies of SP typically collapse data from many subjects, but both neural organization and behavioral performances from semantic tasks vary between individuals. It is not yet understood whether and how the neural variabilities contribute to the individual differences in SP. Here we aim to identify the neural signatures underlying SP variabilities by analyzing individual functional connectivity (FC) patterns based on a large-sample Human Connectome Project (HCP) dataset (N = 873) and rigorous predictive modeling. We used a two-stage prediction approach to build an internal cross-validated model with least squared regression and test the model generalizability with unseen data from different sub-populations, task contexts, and independent out-sample datasets. FC patterns extracted with a semantic brain template were significantly predictive of individual SP scores summarized from two semantic tasks (p = .0001, permutation test). This cross-validated prediction model can be used to predict unseen HCP data (p range .0002 to .0651). The model generalizability was better with FCs in language tasks than other task contexts and better for females than males. Moreover, the model constructed from the HCP dataset can be generalized to two independent cohorts that participated in different semantic tasks. FCs connecting to the Perisylvian network show reliable contributions to the predictive modeling and the model generalization. These findings contribute to understanding the neural basis of individual differences in SP, which potentially facilitate advancements in personalized education and intervention for persons with SP deficits.

 

TALK 7: Right Hemisphere Syntactic Processing in Demanding Conditions-An Event-Related potential study

Yi Chun Ko, National Taiwan University

The right hemisphere (RH) has been found to possess left hemisphere (LH) equivalent syntactic processing ability as indexed by the P600s to some grammatical errors. RH P600s, however, are sometimes associated with lower behavioral accuracy for grammaticality judgment (e.g., in moderate L2 learners or healthy older adults), and such association could reflect a detrimental effect for syntactic processing (due to potentially conflicting outputs across hemispheres) or compensatory attempts from the RH under more challenging processing conditions. To clarify, we recorded Event-Related Potentials (ERPs) from 67 young right-handers while they monaurally learned and judged the grammaticality of 3-element strings generated based on predetermined artificial grammar rules- non-adjacent dependencies between the 1st and last elements. We additionally manipulated the difficulty level of learning these dependencies by varying the size of the set from which the intervening 2nd elements were drawn, creating salient (i.e., easier to learn) and less salient (more difficult to learn) dependencies. To adjudicate between the two possible accounts for the RH P600, we focused on successful learners only and contrasted ERPs over the last 2 blocks to include data only after high proficiency was attained. Despite comparably high behavioral performance, we found a LH-only P600 effect in the easier condition (accuracy=0.98%, N=20), but bilateral P600 effects in the more challenging condition (accuracy=0.98%, N=20). These results thus disfavor the hindrance account of the RH P600 responses, and instead support the compensatory role of the RH P600 in challenging syntactic processing/learning tasks.

 

TALK 8: ERP correlates of semantic and syntactic processing in foreign language learners through a non-immersive environment

Yi-Suan Huang, National Central University

Previous neurophysiological studies show that syntactic violation in one's native language elicits a pronounced P600 component. In contrast, syntactic violation in a foreign, usually second, language (L2) elicits a N400 component in beginning learners, which might shift into a P600 effect as the proficiency increases, regardless of the age of acquisition. However, most of the previous findings were from college students who experience L2 in an immersive environment, while the majority of people in the real world only encounter L2 in a suboptimal setting. To determine whether relatively limited exposure to and usage of a foreign language would result in native-like sensitivity to syntactic violation, we recorded event-related potentials (ERPs) elicited by semantic anomaly and subject-verb (syntactic) disagreement in English sentences from Taiwan college students who have studied English as a school subject (i.e., up to a couple hours per day) for more than 10 years. Despite that these participants all obtained middle to high scores in conventional English tests, the neurophysiological results only revealed a robust N400 effect associated with semantic anomaly, while no ERP component was associated with syntactic violation. Interestingly, although the N400 effect was not significant in the condition of syntactic violation at the group level, its magnitude correlated significantly with individual participants' behavioral accuracy of detecting subject-verb disagreement. These findings suggest that L2 learners might rely on similar mechanisms to perform semantic and syntactic processing even after years of learning in a suboptimal environment, which is insufficient for the development of native-like syntactic processing.

 

TALK 9: Multisensory Integration across Vision, Hearing, and Somatosensation in Typical and Autism Spectrum Development

Patrick Dwyer, University of California, Davis

Prior research suggests multisensory integration (MSI) is reduced in autism; however, most prior studies of MSI in autism have examined only one combination of modalities (typically audiovisual). The present study used onset reaction times (RTs, strain-gauge measured) and 125-channel EEG to examine MSI across different modality combinations. 36 autistic and 19 typically-developing adolescents aged 11-14 participated. Auditory (A) stimuli were 65 dB SPL 20 ms noise bursts, somatosensory (S) stimuli were mechanical taps to the right index finger, and visual (V) stimuli were 85 cd/m2 20 ms flashes (~920 stimuli, ISI 1-2.25 s). EEG data were processed using second-order blind source identification. Event-related potentials (ERPs) were converted to current source density. Maximum-based permutation tests were used to examine RT race model (RM) violation as evidence of MSI-related RT facilitation, and cluster-based permutation tests were used to compare summed unimodal and multimodal waveforms as an index of MSI in ERPs. MSI RT facilitation and ERP difference waves were compared across groups using independent-samples versions of these tests. The ASV RM was violated only in typically-developing participants. Significant AS, AV, and ASV MSI was observed in ERPs from both groups. SV MSI was observed in ERPs from typically-developing participants only. AS MSI effects began as early as ~45 ms. Groups differed in both AV RM violation and, from 84-153 ms, AV ERP MSI. These data suggest broadly intact MSI in autism, but they also provide both behavioural and neural evidence of reductions in AV MSI, a modality combination of particular real-world importance.

 

TALK 10: Time-resolved connectivity reveals the 'how' and 'when' of brain networks reconfiguration during face processing

Antonio Maffei, University of Padova

Recent advances in the study of human brain networks suggest that efficient cognitive operations depend on dynamic changes in large-scale connectivity. In this study we used face processing as a probe to shed light into these dynamics, considering that it is relies on a set of well-studied brain regions, whose activity has been well detailed in terms of its timing. By modeling cortical connectivity from MEG recordings during the presentation of face and scrambled images, we show that the whole-brain network topology becomes more efficient and complex in response to a face than a scrambled image. Interestingly, this coherent topological changes occur in an early time-window with a peak at ~170 ms, consistently with the timing of the N170 event-related potential, which is a typical cortical signature of face processing. We also observed that the core and the extended systems of the face processing network become topologically closer, in a dynamic readjustment of connectivity weights that maximize the efficiency of their communication. Furthermore, using time-resolved decoding we observed that face identity can be distinguished very early on from the functional connectivity. Altogether, these results represent a crucial advancement for understanding of the dynamic reshaping of cortical connectivity that supports cognitive processing of complex visual stimuli, and provide critical insights on the dynamic subtending face processing.

 

TALK 11: Contextual expectations shape cortical reinstatement of sensory representations

Alex Clarke, University of Cambridge

When making a turn at a familiar intersection, we know what will come into view. Such perceptual expectations are derived from knowledge of the context, however it's currently unclear how memory systems use contextual knowledge to reactivate sensory details in cortex. To address this, human participants learned the spatial layout of animals in two cross maze contexts. During fMRI, participants navigated between animals to reach a target, and in the process saw a predictable sequence of five animal images. Importantly, identical sequences could be seen in both contexts. In order to isolate activity patterns related to item predictions, rather than bottom-up inputs, a quarter of navigation trials ended early, instead presenting a blank screen. Employing multivariate pattern similarity analysis, our data revealed that activity patterns in early visual areas showed greater similarity when seeing the same item compared to different items, and critically, activity patterns when seeing an item were related to activity patterns when an item was expected, but omitted from the sequence. Testing regions related to contextual processing, these item expectation effects were also seen in the precuneus. We next tested how contextually driven expectations were related to the hippocampus, finding that activity patterns in the hippocampal body at one point in the sequence were related to patterns in early visual cortex and the precunues at a later point in the sequence. Together, our results reveal how hippocampal representations might reactivate sensory and contextual details of expected items, providing mechanistic insight into the nature of perceptual expectations.

 

TALK 12: Shared and distinct neural representations of agentive actions and inanimate events

Seda Akbiyik, Harvard University

Action recognition relies on a network of prefrontal, parietal and middle temporal brain regions. A hierarchy of action representation in these regions is observed from specific perceptual and kinematic features such as movements of body parts or modality-specific sensory representations to more stimulus-general, conceptual aspects. However, the functional profiles of these regions remain unclear. Specifically, it is typically assumed that the 'action recognition system' refers to a network of regions that supports recognizing actions of humans (e.g., a person jumping over a box). Yet, inanimate entities can also be involved in motion events that are structurally similar to human-agent actions (e.g., a ball bouncing over a box). Here, we used fMRI-based MVPA to test which components of the action recognition system encode action representations that are specific to goal-directed actions of human agents or more general representations that also define structurally similar non-agentive motion events. During fMRI, participants observed structurally similar human-agent actions and inanimate events. Cross decoding revealed that large parts of prefrontal, parietal and middle temporal cortices carry similar representational profiles in encoding meaningfully different human-agent actions or non-agentive motion events. Furthermore, a subregion in right superior temporal sulcus could better distinguish human-agent actions compared to non-agentive actions. These findings imply that action representations that are encoded in these frontoparietal and middle temporal regions cannot be limited to sensorimotor features specific to human-agent actions. However, the action recognition system also contains components that are distinctly associated with human-agent action processing.

 

TALK 13: Occipital alpha desynchronization during visual category learning as an index of Learned Categorical Perception

Fernanda Perez-Gay Juarez, McGill University

Learning to categorize depends on detecting the features that are relevant for category membership.  Learned Categorical Perception (Learned CP), occurs when stimuli in the same category are perceived as more similar (compression) and stimuli in different categories as more different (separation) as the result of learning new categories. In this experiment, we trained 91 subjects to categorize visual fish stimuli by trial and error with corrective feedback. We compared pairwise dissimilarity judgments before and after training and recorded EEG throughout the training. About 40 subjects succeeded in learning the categories ('learners': criterion 80% accuracy) while the rest did not ('non-learners'). Learners showed significant between-category separation and within-category compression in pairwise dissimilarity judgments after training compared to before (Learned CP). A neural network model of category learning indicates that, when detecting features relevant for categorization, the stimulus dimensional space 'shrinks' from n dimensions (or features) to the k dimensions that are relevant for categorization, generating a 'feature-filter' that makes relevant features 'pop-out' effortlessly. An analysis of the EEG Event Related Spectral Perturbation following the presentation of each stimuli revealed a significant decrease in occipital alpha power (8-12 Hz) in a time window of 100-250 ms after having learned, compared to the ERSP of the trials before learning. This effect was absent in Non-Learners, when comparing the second half to the first half of trials. Alpha desynchronization has been previously related to top-down inhibition of task-irrelevant structures, and our results suggest it could be an index of the 'feature-filter' behind Learned

 

TALK 14: Neural entrainment underlies beat-based, but not pattern-based temporal expectations in rhythm

Fleur Bouwer, University of Amsterdam

The brain continuously forms expectations about the timing of incoming information to optimize sensory processing. Temporal expectations can be based on different types of environmental structure, and can be studied exceptionally well in musical rhythm, which often contains both a regular beat (eliciting 'beat-based' expectations), and predictable patterns of absolute temporal intervals (eliciting 'pattern-based' expectations). While beat-based expectations are thought to result from entrainment of low-frequency cortical oscillations to rhythmic input, it is unclear whether entrainment also underlies pattern-based expectations. Here, we examined behavioral responses and EEG activity in silent periods following rhythmic sound sequences that allowed for beat-based or pattern-based expectations, or had random timing. In Experiment 1 (32 participants), we obtained ratings of how well probe tones presented at various times in the silence fitted the previous rhythm. While beat-based expectations affected fitness ratings for multiple beat-cycles, the effects of pattern-based expectations subsided after the first tone omission in the silence. In Experiment 2 (27 participants), using EEG and a frequency tagging approach, we found enhanced power at the beat frequency for beat-based sequences not only during listening, but also during the silence, confirming a theoretical prediction of entrainment models. For pattern-based sequences, we observed enhanced power at pattern-related frequencies only during listening, but not during the silence. Finally, we show that multivariate pattern decoding may be a useful time-resolved alternative to spectral analyses in probing temporal expectations with EEG. Taken together, our findings suggest that neural entrainment may underlie beat-based, but not pattern-based temporal expectations.

 

TALK 15: Spaceflight induces longitudinally reversible and sustained changes in brain functional connectivity

Athena Demertzi, University of Liege

In continuation of manned space missions, there is a need for a better understanding of the brain's coping abilities in extreme environments and how such changes develop in the long-term. We here aimed at characterizing fMRI functional connectivity changes after spaceflight to the International Space Station. Resting state fMRI data (300 scans) were acquired from 15 Roscosmos cosmonauts before (pre-flight) and after (post-flight) space missions.  Eleven cosmonauts were further scanned at 6-month follow-up. Hypothesis-free voxel-wise exploration of connectivity changes at the whole-brain level utilized the intrinsic connectivity contrast (ICC), which reflects the degree to which each voxel is connected to other voxels. Hypothesis-driven modelling of pre-, post- and follow-up sessions tested connectivity changes which were sustained [-1;0.5;0.5 two-tailed] and reversed [-0.5;1;-0.5 two-tailed] across time. A threshold of p<0.005 uncorrected at the voxel-level was applied, with a cluster-level threshold of p<0.05 corrected for multiple comparisons using the family-wise error rate. We found increased and sustained ICC connectivity changes in the right angular gyrus (k=104,t(10)=7.13,p=0.012), next to decreased and sustained changes in the posterior cingulate cortex (k=123,t(10)=6.62,p=0.016) and the thalamus (k=117,t(10)=8.92,p=0.021). Post-flight ICC decreases which reversed in the long-term were found in the anterior insula bilaterally (left hemisphere: k=169,t(10)=7.26,p=0.003; right hemisphere: k=140,t(10)=6.73,p=0.011). Although neuropsychological data are required to explore if a causal link with behavioural changes exists, our findings show that spaceflight impacts brain function and point to cerebral plasticity as a means to cope with such profound exposure to extreme environments.

DATA BLITZ SESSION 4

TALK 1: Probabilistic decision making in humans and recurrent neural networks

Nuttida Rungratsameetaweemana, Salk Institute for Biological Studies; US Army Research Lab

In nature, sensory inputs are often highly structured, and statistical regularities of these signals can be extracted to form expectation about future sensorimotor associations, thereby facilitating optimal behavior. To date, the circuit mechanisms that underlie these probabilistic computations are not well understood. Through a recurrent neural network (RNN) model and human psychophysics, the present study investigates circuit mechanisms for processing probabilistic structures of sensory signals to guide behavior. We first constructed and trained a biophysically constrained RNN model to perform a probabilistic decision making task similar to paradigms designed for humans. Specifically, the training environment was probabilistic such that one stimulus was more probable than the others. We show that both humans and the RNN model successfully extract information about stimulus probability and integrate this knowledge into their decisions and task strategy in a new environment. Specifically, performance of both humans and the RNN model varied with the degree to which the stimulus probability of the new environment matched the formed expectation. In both cases, this expectation effect was more prominent when the strength of sensory evidence was low, suggesting that like humans, our RNNs placed more emphasis on prior expectation (top-down signals) when the available sensory information (bottom-up signals) was limited, thereby optimizing task performance. Finally, by dissecting the trained RNN model, we demonstrate how competitive inhibition and recurrent excitation form the basis for neural circuitry optimized to perform probabilistic information processing.

 

TALK 2: Category knowledge facilitates value-based decisions

Jonathan Shum, Zuckerman Institute, Columbia University

We effortlessly learn the value of visual cues in the environment through trial and error and organize these cues into categories. However, questions remain about how and whether learned value is applied to new cues. Here, we address this question. Participants (N=100) first learned that sets of paintings belonged to specific galleries (three galleries with six paintings each). Paintings within a gallery were drawn from the same artist and differed in abstraction level. Participants then learned the value of individual paintings: value was determined according to gallery membership and abstraction level within a gallery. Then, to test generalization of categorical value, participants had to make decisions between pairs of new paintings from the same artists, without receiving any feedback, giving them an opportunity to use learned category value to drive decisions (six new paintings for each gallery). Results reveal that participants generalized category value knowledge to novel paintings. Through training, participants learned to associate each painting with the associated gallery (92.75 ± 1.30%, p < 0.001) and were more likely to choose a painting from a higher-valued gallery (64.92 ± 1.07%) than a lower-valued gallery (33.61 ± 1.06%, p < 0.001). Critically, when shown novel paintings, participants chose paintings based on their inferred gallery membership: paintings associated with a higher-valued gallery (76.96 ± 1.49%) were chosen significantly more than those from a lower-valued gallery (22.70 ± 1.48%, p < 0.001). These results provide evidence for the effects of category knowledge on value-based decisions, suggesting that category knowledge guides motivated behavior.

 

TALK 3: Behavioral and neural substrates of learning attentional rules

Caroline Jahn, Princeton Neuroscience Institute

We must constantly adapt the rules we use to guide our attention. To understand how the brain learns attentional rules, we designed a novel task that required monkeys to learn which color is the most rewarded at a given time (the current rule). However, just as in real life, the monkey was never explicitly told the rule. Instead, they had to learn it through trial and error by choosing a color, receiving feedback (amount of reward), and then updating their internal rule. After the monkeys reached a behavioral criterion, the rule changed. This change was not cued but could be inferred based on reward feedback. Behavioral modeling found monkeys used rewards to learn the rules. After the rule changed, animals adopted one of two strategies. If the change was small, reflected in a small reward prediction error, the animals continuously updated their rule. However, for large changes, monkeys 'reset' their belief about the rule and re-learned the rule from scratch. To understand the neural correlates of learning new rules, we recorded from the prefrontal and parietal cortex. Preliminary results suggest that prefrontal cortex neurons encode the rule, developing stable rule representations early in learning and maintaining them throughout the block of trials. This representation was lost immediately after the rule switch, reflecting the reset seen in behavior. In contrast, parietal cortex encoded the rule only once the rule was well-known and behavioral accuracy was high. Together, our results provide insight into the behavioral and neural basis of learning new rules.

 

TALK 4: Predicting age in 2- to 12-month-old infants using resting state EEG measures

Carol Wilkinson, Boston Children's Hospital

Alterations in early brain development are thought to underlie various neurodevelopmental disorders whose behavioral manifestations often do not present until later in childhood. MRI and EEG have been used to predict chronological age in adults and children, which can serve as a measure of brain maturity and development. Here, we use longitudinal resting state EEG data (n=543) collected from three different studies within the same lab at various ages between 2- and 12-months-old. Individual power spectra were parameterized into aperiodic and periodic components. Using 47 features, including those characterizing aperiodic and periodic power in canonical frequency bands, we trained several machine learning algorithms to predict brain maturity as defined by chronological age. Each model used a 70-30 train-test ratio with proportional representation from each sub-age group within the total 2- to 12-month range. We employed recursive feature elimination with cross-validation, which ranks and tunes our features before the train/test pipeline. Our best predictions of chronological age were from random forest regressors (RMSE=64.3 days, R-squared=0.52), AdaBoost (RMSE=66.3 days, R-squared=0.50), and gradient boosting regressors (RMSE=61.3 days, R-squared=0.57). Notably, mean error when using a second test set of EEGs from infants later identified to have developmental delays (n=120) worsened by 7-10 days, suggesting that infants with later delays have alterations in EEG measures of brain maturation.  Future directions include expanding our feature set to include functional connectivity and nonlinear EEG measures in order to further improve accuracy and facilitate early identification of developmental delays.

 

TALK 5: The role of the dorsomedial prefrontal cortex in guiding emotional memories

Jaclyn Ford, Boston College

Emotional enhancements on episodic memory are typically characterized by increased likelihood of retrieving emotional relative to neutral events. As such, models of episodic emotional memory typically focus on ways in which interactions between the amygdala and other medial temporal-lobe regions may support these enhancements. However, among remembered emotional events, memories can be distinguished by their affective tone and framing. We propose that the dorsomedial prefrontal cortex (dmPFC)-a region that cross-cuts socio-affective and cognitive domains and is implicated in the control of actions, emotions, and social interactions-plays a key role in this aspect of emotional memory. Here, we present cumulating evidence from our lab that the dmPFC supports the abstraction of meaning from events and the control of emotional memories during encoding, consolidation, and retrieval. At each phase, the dmPFC participates in the integration of affective and cognitive components of memories, setting up networks and framings that either emphasize, or de-emphasize, emotional content. We argue that incorporating the dmPFC into models of episodic emotional memory can allow researchers to better understand the affective tone with which experiences are brought to memory and how this can vary across individuals.

 

TALK 6: Warnings prior to misinformation exposure modulate encoding processes and improve subsequent memory performance

Jessica Karanian

We recently demonstrated that warning participants about the threat of misinformation either before or after exposure to the misinformation reduced memory errors and biased sensory reinstatement during a later memory test. These results revealed that warnings can protect memory from misinformation by biasing reconstructive processes at the time of memory retrieval. Using fMRI, the present study investigated whether warning participants prior to misinformation exposure also influences the initial encoding of misinformation into memory ‚?? for example, by encouraging shallower processing or extra scrutiny of the post-event information. After watching a crime video and answering questions about it, participants were presented with an auditory retelling of the crime that included some misleading information. Immediately before listening to this narrative, some participants were warned about the reliability of its source (warned), while other participants were not warned (unwarned). Warned participants displayed greater activity in prefrontal regions (BAs 44, 6, 10) while listening to the post-event narrative compared to unwarned participants, and the magnitude of activity in these frontal regions positively correlated with subsequent memory performance during the final memory test. Furthermore, prefrontal activity positively correlated with activity in the auditory cortex during the post-event information, and that magnitude of activity in auditory cortex predicted memory accuracy on the final memory test. Together, these results suggest that warnings influence the encoding of misinformation into memory and improve memory accuracy by encouraging enhanced scrutiny of potentially misleading post-event information.

 

TALK 7: Orbitofrontal cortex lesions disrupt hippocampal connectivity in humans

Ian Ballard, University of California, Berkeley

The orbitofrontal cortex (OFC) has a unique pattern of connectivity with temporal lobe and the mesolimbic dopaminergic system that has been hypothesized to support its role in decision making. Previous work from our lab showed that patients with lesions to the OFC make more impulsive choices in intertemporal decisions, but the precise role of OFC remains unclear. One possibility, based on anatomical connectivity, is that the OFC integrates reward information with hippocampal-dependent prospection in order to support farsighted decision making. We hypothesized that damage to the OFC influences behavior in part by disrupting a network that links reward and mnemonic information. We examined resting state data from ten patients with lesions to the orbitofrontal cortex and compared them to a control group of patients with lesions to visual cortex, as well as age-matched healthy controls. We found reduced connectivity between the hippocampus and reward-related areas, including the ventral midbrain and caudate, in OFC patients relative to controls. Moreover, patients with weaker hippocampal-midbrain connectivity made more impulsive decisions, suggesting that patients with more severe network disruption are more impaired in their decision making. Additional analyses show evidence for disrupted integration between canonical anterior and posterior hippocampal networks in OFC patients. Together, these results suggest that the OFC provides an important regulatory role over a network that integrates mnemonic and reward information to support goal-directed behavior.

 

TALK 8: The role of the default mode network and control network in support of curiosity-related memory enhancements

Charlotte Murphy, Cardiff University

Initial findings highlight the pertinent role of the hippocampus-dopaminergic circuit during the elicitation of curiosity. However, the whole-brain mechanisms underpinning curiosity states and their effects on memory remain elusive. The default mode network (DMN) and fronto-parietal network (FPN) are potential candidates due to their role in memory integration and control processes. We hypothesised that DMN and FPN should distinguish when participants are in a high- compared to low-curiosity condition and be recruited more heavily for trials that are subsequently remembered. The current study utilised functional magnetic resonance imaging whilst participants completed a trivia question task; trivia questions with varying levels of curiosity were presented, followed by the associated trivia answer. Memory for trivia answers was tested after a short delay. We adopted a network-based parcellation of the brain into 17 well-established neural networks to examine the relationship between neural activity in each network, curiosity state (high vs. low), timepoint (trivia question vs. trivia answer) and memory performance (hits vs. misses). We found that both FPN and DMN are recruited more during high-curiosity conditions. This activation was most pronounced during the presentation of the trivia answer and predicted curiosity-related memory enhancements. Furthermore, during the presentation of the trivia answer, functional connectivity analyses revealed that curiosity-enhanced memory was supported by increased functional connectivity between dopaminergic regions and regions of both FPN and DMN. Taken together, our results provide a neuromodulatory mechanism from which dopaminergic input that trigger states of curiosity communicate to higher-order cortical regions to facilitate curiosity-enhanced memory.

 

TALK 9: Stimulation of distinct parietal locations differentiates frontal vs hippocampal network involvement in episodic memory

Shruti Dave, Northwestern University

The hippocampus and inferior frontal cortex are thought to have distinct yet interactive roles in episodic memory formation, although mechanisms for their interaction remain unclear. Furthermore, these regions belong to distinct large-scale networks that are thought to include non-overlapping territories of parietal cortex, although this parietal network distinction has not been directly tested. To address these issues, we applied noninvasive theta-burst stimulation to two locations of parietal cortex: one putatively belonging to the hippocampal network and one to the frontal network. These stimulation conditions were performed in different experimental sessions before a memory encoding task and we measured the effects of stimulation on fMRI metrics of network interactions and on memory performance. Replicating previous experiments, fMRI activity associated with successful memory formation was identified within left inferior frontal gyrus. Stimulation of a parietal location of the hippocampal network decoupled the hippocampal network from the frontal areas implicated in memory formation and reduced memory performance. These effects were observed relative to stimulation of the nearby parietal location of the frontal network and relative to a control stimulation condition. This provides direct evidence that interactions of inferior frontal cortex with the distributed hippocampal network support memory formation and indicates that adjacent locations of lateral parietal cortex have functionally distinct effects on these interactions, confirming the segregation of hippocampal versus frontal network locations within lateral parietal cortex.

 

TALK 10: Tuned to Learn: Anticipatory hippocampal pattern stability during mesolimbic activation predicts memory formation

Jia-Hou Poh, Duke University

The motivation to learn can enhance the formation of new memories through the anticipatory engagement of the hippocampus and mesolimbic midbrain. In the current work, we examine how anticipatory representational state in the hippocampus is influenced by the engagement of the midbrain ventral tegmental area (VTA) during motivated learning in a state of curiosity. Human subjects underwent fMRI while viewing trivia questions eliciting different levels of curiosity, and were shown the associated answer after a variable time interval. Following the fMRI session, a surprise memory test was conducted where participants were presented with the trivia questions and were required to recall the associated answers. Consistent with prior work, memory performance was better for high curiosity than low curiosity trivia questions, and the presentation of high curiosity trivia questions was associated with greater anticipatory activation in the midbrain VTA. Using a novel multivariate approach to characterise representational state during information anticipation, we showed that the centrality of hippocampal representation (the distance to centroid in N-dimensional representational space) was modulated by curiosity, associated with VTA activation, and was also predictive of subsequent recall. Specifically, trial-by-trial variability in hippocampal representational centrality was positively associated with VTA activity, and higher centrality was associated with a greater likelihood of subsequent recall. Critically, hippocampal representational centrality accounted for the association between VTA activity and subsequent recall. We propose that activity in the VTA supports memory formation by creating a stable 'learning-state' in the hippocampus.

 

TALK 11: MTL subregions differentially track repetition and recency across large timescales

Mason Price, University of Oregon

It is well established that the medial temporal lobe (MTL) is essential for successful episodic memory formation and retrieval. Previous neuroimaging studies have revealed greater BOLD activation for novel items relative to repeated presentations of a stimulus (repetition attenuation) in the hippocampus and surrounding MTL cortex. At the same time, work in behavioral neuroscience has also identified 'time cells' within subregions of the MTL, neuronal ensembles whose firing rates differ as a function of time elapsed. Recent human neuroimaging studies have provided further evidence that MTL subregions might be sensitive to temporal information. Taken together, these findings suggest that different MTL subregions might encode specific features of item repetitions. To investigate this question, we collected  whole-brain fMRI responses at high field strength (7T; 1.8-mm resolution) while item repetitions (naturalistic images) were dispersed across intervals spanning seconds to nearly 9 months. Voxels sensitive to novelty were identified and activity was averaged within a range of MTL subregions. Further sensitivity to item repetition and recency were probed. Linear mixed effect models suggested perirhinal, subiculum, DG, and PHC tracked both repetition and recency, in line with what might be expected by attenuation. However, ERC and CA regions were better fit by a reduced model including recency alone, suggesting these areas potentially track time between repetitions rather than more basic aspects of item repetition. These effects provide insight into how the MTL might support fine-grained distinctions between the passage of time and item repetition.

 

TALK 12: Interval-scale sequence memories in the hippocampus generalize temporal structure across sequences

Jacob Bellmund, Max Planck Institute for Human Cognitive and Brain Sciences

Sequences of events shape episodic memory. The hippocampal-entorhinal region forms cognitive maps of learned relations such as the temporal relationships of events in a sequence. However, it is unclear whether representations of sequence structure reflect the order of events, elapsing time, or whether they are flexibly scaled to experimentally-defined temporal reference frames. Furthermore, how the formation of specific sequence memories relates to the extraction of regularities from sequences with shared temporal structure remains elusive. Here, we combined a sequence learning task with fMRI. Participants encountered four event sequences and inferred the times of individual events based on infrequent umaskings of a hidden clock. Importantly, we manipulated the clock's speed between sequences to partially dissociate event times from sequence order and objectively elapsing time. Memory tests revealed that participants successfully inferred the times of individual events relative to the hidden clock. After learning, multi-voxel pattern similarity in the anterior hippocampus correlated with temporal distances between event pairs. These sequence representations reflected interval-scale temporal relations relative to the hidden clock beyond event order and elapsed time. Our findings further revealed that temporal relations shape representational similarity in the hippocampus differently for event pairs from the same or from two different sequences. In contrast, the anterior-lateral entorhinal cortex employed a common representational format within and across sequences. Our data suggest that hippocampal sequence representations can be scaled to a continuous external reference frame. Consistent with the cognitive mapping of structural knowledge, our findings demonstrate that temporal structure organizes memories across multiple sequences.

 

TALK 13: The human hippocampus guides visual sampling based on the recent past to optimize learning

James Kragel, Northwestern University

Although the human hippocampus is essential for long-term memory, controversial findings indicate that the hippocampus may support learning through short-term computations that coordinate effective behaviors during learning. We tested the counterintuitive hypothesis that the hippocampus contributes to long-term memory by supporting remarkably short-term processing, as reflected in the temporal sequence of eye movements made during the encoding of natural scenes. While viewing scenes for the first time, participants generated specific patterns of eye movements that reflected a shift from stimulus-driven to memory-driven viewing and signaled effective spatiotemporal memory formation. Theta oscillations from hippocampal depth electrodes signaled when this viewing pattern was about to occur, suggesting short-term retrieval directed visual exploration. Moreover, effective viewing patterns were preceded by shifts towards top-down influence of hippocampal theta on activity within cortical networks that support visual perception and visuospatial attention. The hippocampus thus supports short-term memory processing that coordinates perception, attention, and behavior in support of successful spatiotemporal learning. These findings motivate the reinterpretation of long-term memory deficits as reflecting the loss of the organizing influence of the hippocampus on learning.

 

TALK 14: Dense-sampling reveals flexible network architectures supporting adaptive decision strategies during recognition memory

Tyler Santander, University of California, Santa Barbara

The placement and flexible adaptation of a decision criterion during recognition memory is a stable cognitive trait, unique to individuals and independent of other higher-level cognitive functions. Mass-univariate analyses of task-related brain activity have revealed differential patterns of frontoparietal engagement associated with an individual's criterion placement. However, the functional network architectures in support of these behaviors remain elusive. Here we employ a dense-sampling framework across three studies in an effort to robustly identify variable network configurations that predict adaptive decision-making in memory judgments. In Study 1, a naive subject was scanned for 30 consecutive days: recognition memory tests incentivized a conservative decision criterion by imposing monetary penalties on false alarms. In Study 2, the same subject was scanned again using the same paradigm one year later. Finally, in Study 3, a subject prescreened for reliably-adaptive behavior was scanned across 16 sessions under a 4x4 manipulation of memory evidence and criterion. Behavioral performance (discriminability and criterion) in Study 1 linearly increased over time: sparse Bayesian machine learning and graph network analyses revealed patterns of whole-brain connectivity and large-scale network topologies that significantly predicted variation in decision strategy (i.e. setting a more appropriately-conservative criterion). However, these effects were not evident one year later in Study 2. In Study 3, we identified flexible architectures in control, attention, and default mode networks as the participant adapted to more nuanced levels of memory evidence and criterion placement. Together, these findings provide the first evidence of functional network dynamics while shifting decision strategies during recognition memory.

 

TALK 15: Greater neural differentiation in the ventral visual cortex is associated with youthful memory in superaging

Yuta Katsumi, Northeastern University

Superagers are older adults who maintain youthful memory despite advanced age. Previous neuroimaging studies showed that superagers exhibit greater structural and intrinsic functional network brain integrity, which contribute to their youthful memory performance. However, no studies to date have examined brain activity as superagers learn and remember novel information. Here, we analyzed functional magnetic resonance imaging data collected from 41 young (24.5 ± 3.6 years) and 40 older adults (66.9 ± 5.5 years) while they performed a paired-associate visual recognition memory task. Superaging was defined as youthful performance (males: 13, females: 14) on the long delay free recall measure of the California Verbal Learning Test. We performed a representational similarity analysis to assess the fidelity of neural representations as participants encoded and later retrieved from memory a series of word stimuli paired with a face or a scene image. As predicted, superagers-similar to young adults-exhibited more distinct neural representations in the fusiform and the parahippocampal gyrus while viewing visual stimuli belonging to different categories (i.e., greater neural differentiation) and more similar representations between encoding and subsequent retrieval (i.e., greater neural reinstatement), compared with typical older adults. Greater neural differentiation and reinstatement were associated with superior item and associative recognition memory performance in older adults. Given that the fidelity of cortical sensory processing depends on neural plasticity and is possible to improve with training, the mechanisms identified in the current study could be translated into biomarkers for trials of potential interventions to promote successful aging.

CNS2021-LogoVM-rev

MARCH 13–16 • 2021

Latest from Twitter