Summary: New research shows that decision-related information derived from vision—carried by complex streams of feedforward and feedback signals along visual pathways—is broadcast broadly across the visual system, reaching many neurons including those not directly used to form the decision.
Source: NIH
Researchers at the National Institutes of Health report that when decisions rely on visual input, the brain sends feedback signals widely throughout the visual system. These feedback signals—such as information about an expected object or the choice that was made—reach neurons tuned to features like color, shape, or depth, even when those neurons are not strictly necessary to make the decision.
The study, led by scientists in the National Eye Institute (NEI) and published in Nature Communications, examines how feedback related to decisions is routed back into visual cortex and whether that routing is selective for the neurons that actually contribute to the decision.
“Why and how decision-related information is relayed back into visual processing areas is an open question,” said Hendrikje Nienborg, Ph.D., chief of the NEI Unit on Visual Decision Making and lead author of the study. “Some theories predict that feedback should be selective—affecting only those neurons that are involved in the decision. Our results indicate a different picture: decision-related feedback can be spatially unselective and influences neurons more broadly than expected.”
Feedback signals in the brain serve many roles across sensory and cognitive systems. In visual tasks, feedback often contains information about expectations or attention—for example, where a relevant object is located or which of its features matter—and can boost the activity of neurons that represent that object or feature. Everyday situations illustrate this: while driving through an intersection you may pay more attention to the crosswalk, enhancing neurons tuned to objects in that region; if you’re searching for a child wearing a red jacket, neurons sensitive to the color red may receive extra signal.
Prior work suggested feedback helps the brain focus on hard-to-see features or stabilizes choices during decision-making. Some researchers proposed that decision-related feedback selectively enhances only those neurons directly used for the decision, leaving irrelevant neurons unaffected. But what happens when multiple pieces of information are relevant at once—such as both spatial location and an object feature like depth? Is feedback applied separately to each relevant dimension, and is each of those feedback streams spatially selective?
To investigate, the research team trained macaque monkeys to perform a visual discrimination task in which they judged whether an object at a specific screen location appeared concave or convex, while ignoring stimuli at other locations. During the task, the scientists recorded neural activity in mid-level visual cortex areas that process spatial position and depth cues. The animals performed the task reliably, accurately reporting objects at the relevant location and dismissing irrelevant stimuli elsewhere.
Recordings revealed a nuanced pattern of feedback. Consistent with earlier studies, feedback related to spatial attention was selective for location—the brain enhanced signals for the spatial region the animals attended to—but this location-related feedback did not depend on which choice the animal ultimately made. In contrast, feedback tied to the feature being judged (object depth) correlated with the decision itself but was spatially unselective: depth-related decision signals were broadcast to depth-sensitive neurons across locations, including neurons that could not plausibly have been used to form the decision about the attended location.
“Intuitively, one might expect feedback to be tailored to each task so that only task-relevant neurons receive decision signals,” said Katrina Quinn, the study’s first author and a graduate student in Nienborg’s lab. “Instead, we observed a feedback component that generalizes spatially even when the behavior relies on spatially selective information.”
The findings suggest the brain may use a common feedback mechanism that applies broadly across visual cortex, independent of the spatial specificity demanded by a particular task. Such a mechanism could reflect biological constraints and may help the system generalize across a range of tasks. The results also support a previously proposed link between feature-based attention and decision-related neural activity.
Funding: The research was supported by the NEI Intramural Program, the European Research Council, the German Research Foundation, and the National Science Foundation.
About this visual neuroscience research news
Source: NIH
Contact: Lesley Earl – NIH
Image: Image credit: National Eye Institute
Original Research: Open access.
“Decision-related feedback in visual cortex lacks spatial selectivity” by Katrina R. Quinn, Lenka Seillier, Daniel A. Butts & Hendrikje Nienborg. Nature Communications
Abstract
Decision-related feedback in visual cortex lacks spatial selectivity
Feedback conveys contextual information that supports flexible task performance. Previous empirical and computational work proposed that such feedback targets task-relevant neuronal subpopulations in visual cortex. To test this, the authors combined two tasks that individually produce selective feedback, used feature discrimination at one of two locations, and dissociated decision formation from motor reports while recording in macaque mid-level visual areas.
Although behavior was spatially selective—animals used only task-relevant information—modulation by decision-related feedback was spatially unselective. Population responses showed similar alignments between stimulus and choice regardless of whether the stimulus location was relevant. The results indicate a common feedback mechanism across tasks that operates independently of spatial selectivity demands, which may reflect biological constraints and aid generalization across tasks. The findings also reinforce a hypothesized connection between feature-based attention and decision-related activity.