Attention: Neural mechanisms

Talk Session: Sunday, May 18, 2025, 5:15 – 7:15 pm, Talk Room 1

Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions

Talk 1, 5:15 pm

Sharper Spatially-Tuned Neural Activity in Preparatory Overt than in Covert Attention

Damian Koevoet¹, Vicky Voet¹, Edward Awh², Henry M. Jones², Christoph Strauch¹, Stefan Van der Stigchel¹; ¹Utrecht University, ²The University of Chicago

Attention is shifted with or without an accompanying saccade (overtly or covertly, respectively). The neural signatures of overt and covert attention largely overlap, and have even been deemed identical. However, by definition the neural signatures of overt and covert attention must diverge at some point (i.e. saccade initiation), but it remains unclear when and how they diverge. Here, we capitalized on the high temporal resolution of electroencephalography (EEG) in combination with multivariate decoding to investigate when and how overt and covert attention differ neurally. Neural decoding reliably predicted whether overt or covert attention was shifted well before saccade onset (~700ms). We then used an inverted encoding model to compare spatially-tuned neural responses to the attended location between overt and covert shifts. Strikingly, we observed that overt shifts caused sharper spatially-tuned neural responses compared with covert shifts. But why were these spatially-tuned neural responses sharper: did overt attention employ more of the same attention or does imminent saccade execution recruit an additional spatially-tuned process? To address this, we reconstructed spatially-tuned responses when training on only one of the two conditions. We found overt and covert attention to only partly employ similar spatially-tuned responses, arguing against a ‘more-of-the-same attention’ account. Our results instead demonstrate overt attention to recruit an additional spatially-tuned process. We speculate that this additional spatially-tuned process is related to predictive remapping across saccadic eye movements. Together, we demonstrate the neural signatures of overt and covert attention to diverge rapidly because overt attention employs an additional process which sharpens spatially-tuned neural activity.

This project has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement n° 863732).

Talk 2, 5:30 pm

Changes in neural tuning rather than noise structure explain attentional enhancement of population representations in the human visual cortex

Yu-Qi You¹ (), Kendrick Kay², Ru-Yuan Zhang¹; ¹Shanghai Jiao Tong University, ²University of Minnesota

Attention is thought to enhance behavior by refining neural population representations of visual stimuli. Past research has proposed two competing theories to explain these effects: attention either changes the tuning of individual units (i.e., tuning-change theory) or alters trial-by-trial noise correlations between units (i.e., correlation-change theory). However, there is currently no unified framework to accommodate or quantitatively compare these two theories. Leveraging linear Fisher information from computational neuroscience, we first analytically demonstrate that, in addition to tuning and correlation changes, changes in response variability across trials emerge as a third factor that influences population representations. To assess which mechanisms the brain employs, we conducted an fMRI experiment (3T; 2.5mm³) with eight human subjects performing two tasks: attending to digits at center-of-gaze (i.e., a fixation task) or attending to face stimuli randomly presented at one of 16 positions in a 4 × 4 grid (2° spacing) (i.e., a face task). We recorded 80 trials per position for each task and analyzed BOLD responses across regions of the human visual cortex (V1, V2, V3, hV4, OFA, FFA-1, FFA-2). Consistent with previous human imaging studies, compared to the fixation task, the face task systematically altered voxel receptive fields (vRFs) and improved multivariate decoding of face positions. In line with electrophysiological studies, the face task also reduced voxel Fano factors and noise correlations. Most importantly, we developed a neural population geometry approach to quantify the contributions of tuning, noise correlations, and response variability to population representations. Surprisingly, changes in vRFs were the primary driver of improved population representations in both low- and high-level visual areas, while the other two factors had little or even detrimental effects. These findings strongly support the tuning-change theory of visual attention and provide new insights into how attention enhances neural representations to optimize stimulus coding.

Talk 3, 5:45 pm

Pronounced modulation of activity in primate early visual cortex by internal state

Bharath Chandra Talluri¹, Incheol Kang¹, Jacob L. Yates², Daniel A. Butts³, Hendrikje Nienborg¹; ¹Laboratory of Sensorimotor Research, National Eye Institute, National Institutes of Health, Bethesda, MD, USA, ²Herbert Wertheim School of Optometry and Vision Science, University of California, Berkeley, CA, USA, ³Department of Biology and Program in Neuroscience and Cognitive Science, University of Maryland, College Park, MD, USA

Internal states of an organism can have a profound influence on visual processing, exemplified by strong modulation due to non-retinal factors like body movements and arousal reported in several species. Recent work found that spontaneous body movements only minimally modulate early visual processing in primates. However, the degree to which general arousal modulates early visual processing in awake macaques is unknown. Here, we recorded extracellular spiking activity from populations of neurons in early visual areas (V1 & V2) of macaques, while manipulating behavioral states in two conditions that required different levels of task engagement: by rewarding the animal contingent on maintaining central fixation in one set of trials (fixation blocks) or rewarding the animal at unpredictable times while free-viewing with no explicit task (free-viewing blocks). Both conditions contained repeated presentations of full-field flashes, such that the visual input to neuronal receptive fields was invariant to the animal’s gaze position. The blocks were interleaved, and we monitored physiological correlates of arousal, such as pupil size. The stimulus-response amplitude of the neurons showed a pronounced increase (93±5%) in fixation vs free-viewing blocks, and this modulation could not be explained by differences in retinal input, or saccade statistics between blocks. A computational model that had no knowledge of the conditions but allowed for trial-to-trial variations in shared gain and baseline firing across the neuronal population explained the spiking activity better than a model that allowed separate, fixed stimulus-driven responses in each condition. The shared gain and offset signals were different between conditions and correlated with pupil size suggesting that the observed modulations in neuronal activity reflect modulations in the internal state of the animals. Our findings show that modulation of early visual processing in non-human primates by general arousal can be pronounced, in stark contrast to the minimal modulation by body movements.

National Institutes of Health (grant no. R00EY032179); National Science Foundation (NCS-FO 2123568); National Eye Institute Intramural Research Program at the National Institutes of Health (1ZIAEY000570-01)

Talk 4, 6:00 pm

Human pulvinar stimulation engages select cortical pathways

Jordan A. Bilderbeek¹, Nicholas M. Gregg², Maria Guadalupe Yanez-Ramos¹, Harvey Huang³, Morgan N. Montoya¹, Peter Brunner^4,5, Jon T. Willie^4,5, Jamie J. Van Gompel⁶, Gregory A. Worrel², Kai J. Miller⁶, Dora Hermes¹; ¹Department of Physiology and Biomedical Engineering, Mayo Clinic, Rochester MN, ²Department of Neurology, Mayo Clinic, Rochester, MN, ³Medical Scientist Training Program, Mayo Clinic, Rochester, MN, ⁴Department of Neurosurgery, Washington University School of Medicine, St. Louis, MO, ⁵National Center for Adaptive Neurotechnologies, St. Louis, MO, ⁶Department of Neurosurgery, Mayo Clinic, Rochester, MN

The pulvinar is a large associative thalamic nucleus and plays a role in attentional regulation and modulation of visual pathways. Clinical studies in epilepsy often implant electrodes in the pulvinar to treat seizure networks involving the occipital and temporal neocortex. However, the placement of electrodes within the pulvinar is uncertain as anatomical studies have revealed that pulvinar subnuclei are connected to distinct occipital, temporal, and parietal visual areas. Single-pulse electrical stimulation of such pulvinar subnuclei during multi-lead stereotactic EEG (sEEG) provides an opportunity to causally test whether electrical stimulation of pulvinar subnuclei selectively engages distinct anatomical pathways. We delivered biphasic single-pulse electrical stimulation within the medial and lateral pulvinar in nine neurosurgical patients undergoing clinical sEEG for drug-resistant epilepsy evaluation. We analyzed pulse-evoked potentials (PEPs) using a parameterization algorithm that computed the explained variance as a metric of response reliability for each trial. Stimulation in the medial and ventromedial pulvinar elicited reliable PEPs in the temporal neocortex that diminish as stimulation proximity increases toward the lateral pulvinar. Conversely, stimulation of the lateral pulvinar produced reliable striate/prestriate (V1-V2) and extrastriate (V3a/b, hV4, TO1-2, LO1-2, IPS0) PEPs, which diminishes as stimulation proximity increases towards the medial pulvinar. We also found that dorsomedial pulvinar stimulation evokes parietal (IPS1-5) PEPs with relatively limited striate/extrastriate and temporal responses. Altogether, our results highlight that stimulation of specific pulvinar subfields evoke reliable responses in the striate/extrastriate cortex, parietal cortex, and temporal neocortex (Figure 1). Identifying these pulvinar-cortical projection fields is crucial for understanding how the pulvinar modulates neural activity in visual pathways. Moreover, it is a significant step towards the clinical advancement of seizure network-specific deep brain stimulation in epilepsy.

Funding was provided by the National Institute of Mental Health Award Number R01 MH122258. The content is solely the responsibility of the authors and does not represent the official views of the NIH.

Talk 5, 6:15 pm

A common pulvinar-cortical architecture across visual tasks

Xingyu Liu¹ (), Michael J. Arcaro¹; ¹University of Pennsylvania

Through its widespread connections with cortex, the pulvinar nucleus within the primate thalamus plays a crucial role in visual and higher cognitive processing. However, our understanding of how the pulvinar’s internal organization and connectivity with distinct cortical networks support vision remains limited. Using fMRI, we characterized the pulvinar’s functional connectivity “fingerprints” (fFC) — distinct patterns of correlated activity between the pulvinar and specific cortical networks. During task-free, resting state, we identified a topographic organization within the pulvinar characterized by primary axes spanning the entire visual cortical hierarchy as well as non-visual association cortices. To test whether this organization reflects a common architecture for visual processing, we analyzed pulvinar-cortical fFC across diverse visual tasks including naturalistic movie viewing and controlled visual attention paradigms. Remarkably, cortical activation patterns during these tasks could be reconstructed from pulvinar activity using the resting-state fFC patterns, suggesting a stable underlying architecture. This stability was further supported by analyses showing that the topographic organization of pulvinar-cortical fFC was preserved across states. In contrast to the stable pattern of fFC across tasks, the strength of specific pulvinar-cortical connections was flexibly modulated by task demands — naturalistic vision enhanced fFC with visual cortical areas, while attentional tasks strengthened coupling with dorsal attention and cognitive control networks. Moreover, the disruption of this architecture during anesthesia and restoration upon recovery suggests that this stable architecture may serve as a hallmark of consciousness. These findings reveal a fundamental principle of thalamic organization: the pulvinar maintains a stable architectural framework while dynamically adjusting the strength of specific cortical connections based on visual processing demands. This work provides new insights into how the primate visual system achieves both stability and flexibility in visual processing through thalamo-cortical interactions.

This research was supported by National Institute of Mental Health (P50 MH132642)

Talk 6, 6:30 pm

Independent encoding of salience, value, and attention in primate superior colliculus

Matthew Murawski¹, James Herman¹; ¹University of Pittsburgh

Neuronal activity in the primate superior colliculus (SC) is modulated by physical salience, reward value, and attention, but how these signals are integrated remains unknown. One hypothesis is that SC activity reflects a unified “priority map” of the visual field. However, it is unclear how a unified priority map might support distinct roles for SC activity in different contexts: If the same SC neurons are activated by saccade cues and attention cues, how can SC activity evoke a saccade in one context but covert orienting of attention in another context? We hypothesized that distinct sources of modulation cause dissociable rather than unified patterns of SC activation, which would facilitate a context-specific relationship between SC activity and behavior. To test this hypothesis, we recorded 220 SC neurons in a macaque performing two tasks: a spatially cued covert change detection task manipulating goal-directed attention, and a saccade task independently varying reward value and salience. All three factors influenced behavior: salience affected saccadic endpoint error, reward modulated reaction time, and attention cueing affected detection rate. Single-neuron ROC areas for each factor suggested that salience, reward, and attention exert independent influences on SC. At the population level, linear classifiers trained to decode conditions for one factor (e.g., reward) failed to generalize to other factors (e.g., salience), indicating independent population-level encoding. These findings challenge the unified priority map model and suggest SC could support flexible visually guided behaviors by selective routing. This mechanism could explain how SC contributes to overt orienting and covert attention depending on context, advancing our understanding of how the brain flexibly processes visual information to guide behavior.

This work was supported by the Hillman Foundation and the Eye and Ear Foundation of Pittsburgh.

Talk 7, 6:45 pm

Goal-directed visual information processing with glutamatergic excitation and GABAergic inhibition in posterior parietal cortex

Sebastian Frank¹, Sinah Wiborg¹, Antonia Wittmann¹, Nina Beck¹, Markus Becker¹, Zhiyan Wang¹; ¹University of Regensburg

Goal-directed visual information processing involves selecting relevant among irrelevant visual signals. This selection is facilitated if there are separate and stable representations of goal-relevant and goal-irrelevant visual information. Posterior parietal cortex is crucially involved in maintaining such separate representations but it is debated how they are implemented on a neuronal level. The sharpest separation would be achieved by representing goal-relevant information through increased excitatory activity and goal-irrelevant information simultaneously through increased inhibitory activity in different subpopulation of neurons within posterior parietal cortex. If this is the case, increased demands on maintaining separate representations of goal-relevant and goal-irrelevant visual information should be accompanied by a concomitant increase of both excitatory and inhibitory activity. Testing this prediction is difficult with functional magnetic resonance imaging because the contributions of excitatory and inhibitory activity to the hemodynamic response cannot be separated. Here, we measured the concentrations of glutamate, a chief excitatory neurotransmitter, and gamma-aminobutyric acid (GABA), a chief inhibitory neurotransmitter, in the posterior parietal lobe using time-resolved functional magnetic resonance spectroscopy (fMRS). During fMRS participants (n=30) performed a multiple object tracking task with low and high demands on maintaining separate representations of goal-relevant moving targets among goal-irrelevant moving distractors. Tracking trials were 12s long and followed by a jittered inter-trial interval. Changes in glutamate and GABA concentrations with different tracking conditions were measured as a time-series of consecutive 2s-long PRESS and 3s-long MEGA-PRESS fMRS-scans. The results showed greater concentrations of glutamate and GABA in the high than low demand tracking condition in the posterior parietal lobe. No such simultaneous increase of glutamate and GABA concentrations between tracking conditions was found in the occipital lobe. Our results suggest that a simultaneous increase in excitatory and inhibitory activity in posterior parietal cortex is involved in maintaining sharply separated representations of goal-relevant and goal-irrelevant visual information.

Deutsche Forschungsgemeinschaft (DFG): Emmy Noether Grant (Project Number 491290285); Julitta und Richard Müller Stiftung

Talk 8, 7:00 pm

Attentional eye selection affects neural dynamics of binocular rivalry across visual hierarchy

Chuan Hou¹ (), Junxian Rao²; ¹Smith-Kettlewell Eye Research Institute

Attention plays a role in binocular rivalry. However, previous studies have mostly focused on feature-based attention (Mitchell et al., 2004) or directing attention away from the rival stimuli (Zhang et al., 2011). Few studies (Ooi and He, 1999, 2005) have examined whether attentional eye selection affects the dynamics of rivalry, with the underlying neural basis remaining unclear. Here, we investigated how attentional eye selection affects neural dynamics of binocular rivalry across the visual hierarchy by presenting a pair of orthogonal gratings tagged with different temporal frequencies in each eye (horizontal at F1 = 7.5 Hz for the non-dominant eye; vertical at F2 = 6 Hz for the dominant eye) to observers with normal vision. Neural activity during binocular rivalry was recorded using fMRI source-localized high-density EEG to identify regions of interest (ROIs). We correlated neural responses with behavioral reports to examine activity during perceptual dominance and suppression phases and then compared the responses between two conditions: passive viewing vs. paying attention to the horizontal gratings presented to the non-dominant eye, in various ROIs. Our findings showed that neural activity biased toward the stimuli of the attended eye across ROIs. The primary visual cortex (V1) and extrastriate visual areas, including hV4, middle temporal (MT) and lateral occipital cortex (LOC), exhibited a similar attention effect (approximately a 50% increase in response to the attended stimuli compared to the passive viewing). In contrast, V3A, intraparietal sulcus (IPS), temporal pole, and frontal pole showed approximately a 80% attention effect. Our results provided evidence that attentional eye selection can modulate neural dynamics during binocular rivalry across the visual hierarchy, with more pronounced effects observed in the high-level cortices. These findings are consistent with behavioral studies (Ooi and He, 1999, 2005) and also align with the attention literature, demonstrating that neural activity is enhanced by attention.

This research was supported by NIH grant R01EY035346 to C.H.