Selective Visual Attention, Alive and Well!

Symposium: Friday, May 16, 2025, 10:30 am – 12:30 pm, Talk Room 1

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

Organizers: Marisa Carrasco¹, Miguel Eckstein², Anna C Nobre³; ¹New York University, ²University of California Santa Barbara, ³Yale University
Presenters: Marisa Carrasco, Sam Ling, Farran Briggs, Miguel Eckstein, Anna C Nobre

Visual attention is critical for perception, enabling prioritized and selective information processing in human and non-human primates. This symposium encompasses a broad range of studies on covert visual attention across spatial [Carrasco, Eckstein, Briggs], feature [Ling, Briggs], and temporal [Nobre] domains. Studies employ diverse methods and levels of analysis, including psychophysics [Carrasco, Briggs, Nobre], multi-unit recording [Briggs], MEG/EEG [Nobre], neurostimulation [transcranial-magnetic stimulation, TMS, Carrasco], and computational modeling using Convolutional Neural Networks [CNNs; Eckstein]. Central and common to these studies is the careful definition, operationalization, and manipulation of attention, enabling us to isolate and characterize the roles and mechanisms of different types of attention. Together, these methods, level of analyses, and approaches advance our understanding of visual attention at both behavioral and neural levels. The symposium will begin with an overview of the central questions addressed by each of the five speakers, including the types of attention studied, levels of analysis employed, the strengths and limitations of the respective protocols and methods, and the theoretical contributions. Each presentation will be followed by a brief Q&A session. Presentations: Marisa Carrasco discusses how endogenous (voluntary) and exogenous (involuntary) covert spatial attention shape human perception by modulating visual processing in distinct ways. Her methodological approach isolates the perceptual effects and neural correlates of each attention type. Sam Ling explores how feature-based attention alters population responses within the human visual cortex (V1-V3), altering the contrast response and spatial-frequency tuning for attended items. Farran Briggs investigates neuronal mechanisms of spatial and feature attention, using neurophysiological measurements at the neuronal and circuit scale in non-human primates performing discrimination tasks that involve shifting spatial and feature attention. Miguel Eckstein examines how feedforward Convolutional Neural Networks (CNNs), despite lacking built-in attention mechanisms, exhibit behavioral signatures of human-like covert spatial attention. Moreover, they show some "neuronal unit" responses analogous to documented physiological signatures of attention, as well as other unit responses not yet observed in neurophysiological studies. Kia Nobre presents recent MEG and EEG studies showing how goals, temporal expectations and attention modulate sensory and motor processing stages, enabling the prioritization of items in perceptual and working-memory tasks in humans. The symposium will conclude with a Panel discussion summarizing the contributions of each talk through concise answers to the posed questions and facilitating an open dialogue with the audience, enhancing our collective understanding of visual attention.

Talk 1

How spatial attention shapes perception

Marisa Carrasco¹, Antonio Fernández¹, Laura Dugué^2,3, Nina Hanning^1,4; ¹New York University, ²Université Paris Cité, ³CNRS, ⁴Humboldt-Universität zu Berlin

We investigated how endogenous (voluntary) and exogenous (involuntary) covert spatial attention differentially modulate visual perception, at behavioral and neural levels. To understand the effects of endogenous and exogenous attention, we conducted head-to-head comparisons using psychophysical, computational, neuroimaging (fMRI), and neurostimulation (TMS) methods. In each empirical study, we kept participants, stimuli, and discrimination tasks constant, varying only the attentional cue manipulation. This approach allowed us to isolate the distinct effects of each type of attention. First, I highlight findings from psychophysical experiments showing how endogenous and exogenous covert spatial attention differentially affect performance in tasks mediated by basic visual dimensions (e.g., contrast sensitivity) and their featural representations (using reverse correlation methods). Next, I present findings from fMRI studies, revealing distinct neural correlates of endogenous and exogenous attention. For example, in a pre-cue condition, endogenous attention increasingly modulated stimulus-evoked activity along the visuocortical hierarchy (V1 to LO1), with largest effects in higher cortical areas; however, this modulation remained constant in a post-cue condition. In contrast, exogenous attention produced consistent modulation across the visual hierarchy in a pre-cue condition but showed no modulation in a post-cue condition. Lastly, I discuss how we used TMS experiments to uncover that brief disruption of V1/V2 activity eliminates exogenous attention effects, whereas disruption of the (human homologue of) the right frontal eye fields (rFEF+) significantly reduces endogenous attention effects. Together, these studies reveal how endogenous and exogenous attention shape perception by selectively modulating the processing of basic visual dimensions across different brain regions.

Talk 2

How feature-based attention alters representations in human visual cortex

Sam Ling¹, Joshua Foster¹, Feiyi Wang¹, Luis Ramirez¹; ¹Boston University

I will describe human neuroimaging research that examines the modulatory effects of feature-based attention on two fundamental building blocks in vision: contrast response and spatial frequency tuning. To assess the influence of attention on the gain of the contrast response function (CRF), we measured BOLD activity evoked by a probe stimulus in one visual hemifield while participants attended to the probe orientation (attended condition), or to an orthogonal orientation (unattended condition), in the other hemifield. To measure CRFs in visual areas V1-V3, we parametrically varied the contrast of the probe stimulus. In all three areas, feature-based attention increased contrast gain, improving sensitivity by shifting CRFs toward lower contrasts. In extrastriate cortex, we also found an increase in response gain that was greatest in the parafovea. In a second study, we used fMRI to measure population spatial-frequency tuning (pSFT), which allowed us to examine how feature-based attention alters the response profiles of neural populations in early visual cortex. Participants were cued to attend to one of two spatially competing letter streams, each defined by low or high frequency content. Concurrently, we measured pSFT in a task-irrelevant hemifield to examine how the known spatial spread of feature-based attention influenced the SF tuning properties of neurons sampled within a voxel. We discovered that attention elicited attractive shifts in SF preference, towards the attended SF. Taken together, these two studies reveal reliable signatures of feature-based attention on the gain and tuning preferences of population responses within human visual cortex.

Talk 3

Cortical and thalamic contributions to spatial and feature attention

Farran Briggs^1,2, Alesandra Martin¹, Sabrina Mai¹, Shraddha Shah¹; ¹University of Rochester, ²National Eye Institute

Much is known about how shifting attention between locations in visual space (spatial attention) or across different stimulus features (feature attention) alters the activity of visual neurons. However, the mechanisms guiding these effects at the scale of individual neurons, neuronal circuits, or neuronal populations remain a mystery. We compare attention effects across early visual system structures and across spatial and feature attention tasks as a way to infer similarities and differences in underlying mechanisms at the neuronal and circuit scale. We trained non-human primates (NHPs) to perform feature discrimination tasks involving shifting spatial attention and switching attention between multiple stimulus features (contrast, orientation, and color). Neuronal recordings were made with multi-electrode arrays implanted in the visual thalamus or primary and secondary visual cortex. Measurements of neuronal feature tuning were made independent of the attention task. Our observations suggest attention effects depend critically on areal organization, measurement scale, neuroanatomical connections, and neuronal feature selectivity. First, attentional modulation of multi-unit activity is quite robust in visual cortex but absent in visual thalamus. Second, attentional modulation of correlated variability among neurons in visual cortex depends on anatomical connectivity: attention strongly reduces correlated variability among neurons connected via excitatory synapses. Third, attentional modulation of cortical neurons depends on the match between their feature selectivity and the task-relevant feature, even when spatial attention is held constant. These findings show that attention effects in NHPs depend critically on factors that span anatomical scales and feature encoding dimensions, necessitating exploration across scales and task protocols.

Talk 4

Convolutional Neural Network Uncovers Emergent Neurally-Plausible Mechanisms Mediating Human-like Covert Attention

Miguel Eckstein¹, Sudhanshu Srivastava¹; ¹University of California Santa Barbara

Covert spatial attention allows the brain to select different regions of the visual world without concurrent eye movements. Cues predictive of a target location orient covert attention and improve perceptual performance. In most computational models, researchers explicitly incorporate an attentional mechanism that alters processing at the attended location (gain, noise reduction, divisive normalization, biased competition, Bayesian priors). Here we assess the emergent neuronal mechanisms of feedforward Convolutional Neural Networks (CNNs), with no built-in attention mechanism, which shows behavioral signatures of human-like covert attention. We use a system-neuroscience-inspired approach to analyze 1.8M computational units (“neurons”) of CNNs trained on a spatial cueing task. Consistent with neurophysiology, we show early layers with retinotopic units separately tuned to the target or cue, and later layers with units with joint tuning and increased influence of the cue on target responses. We find cue-inhibitory and location-opponent units unreported by neurophysiologists. The cue influences the mean neuronal unit responses and changes target sensitivity through three mechanisms: cue-weighted summation and opponency across locations, and interaction with the thresholding Rectified Linear Unit (ReLU). CNN computational stages mirror a Bayesian ideal observer (BIO), but with more gradual transitions, and also include computations (the opponency and ReLU interaction) distinct from the BIO and not discussed in the physiology literature. Together, the findings establish a likely system-wide characterization of the brain computations that mediate the behavioral signatures of covert attention and provide a theoretical framework for neurophysiologists and vision scientists.

Talk 5

Moments of attention

Anna C Nobre¹; ¹Yale University

Visual selective attention functions – that anticipate, select, prioritize, and prepare contents – are infrastructural and core to guiding adaptive perception and behavior. Decades of psychophysical and neuroscientific research across levels of analysis have advanced our understanding of how internal signals related to experience and goals flexibly and proactively modulate processing according to receptive field properties of relevant anticipated stimuli. The next challenge is to situate attention functions within the dynamic flux of natural experience. Active agents move through changing environments, with evolving goals and expectations. Research on selective temporal attention investigates how the brain focuses on specific moments of predictable or relevant events to guide perception, action, or mnemonic access. From the diverse and multilayered sensory stream, the brain extracts various types of temporal structures to tune the processing of events according to their timing. I will discuss recent studies in our group using MEG and EEG that show how goals and temporal expectations modulate sensory and motor stages of processing to prioritize items in perceptual and working-memory tasks in humans. Modulatory functions can proceed in tandem and independently. The findings highlight the robust impact and flexibility of temporal attention, implicating short-term and long-term sources of predictions that interact with various receptive-field properties to enhance perception and memory through multiple mechanisms. Time is pressing for the fourth dimension of attention research.