Decision-making engages a recurrent loop between auditory cortex and prefrontal cortex, not a one-way street
Classical models of cortical processing describe a hierarchical system: sensory information flows forward from primary sensory areas to higher-order association regions, where decisions are computed. A new study published July 8 in Nature Communications by Franco Giarrocco and Bruno B. Averbeck at the National Institute of Mental Health suggests this picture is incomplete.
By simultaneously recording 1,199 neurons from the primary auditory cortex (A1) and the dorsolateral prefrontal cortex (dlPFC) of macaques performing a spatial auditory decision-making task, the researchers found two fundamentally different patterns of information flow, one unidirectional, the other a dynamic recurrent loop.
Two kinds of information, two kinds of flow
The macaques were cued to a spatial location and had to detect a target sound at that location, making a behavioral choice. During the task, sensory information (the auditory cue) and the monkey’s eventual choice (which sound it decided was the target) were encoded by neurons in both A1 and dlPFC, but the timing and direction of that encoding differed dramatically.
For sensory information, what the monkey heard, the flow was classical and straightforward. Information peaked in A1 approximately 130 milliseconds after the cue and traveled unidirectionally forward to dlPFC. This is the feedforward sweep predicted by hierarchical models.
For decision-related information, which sound the monkey chose, the pattern was entirely different. Choice-related signals peaked approximately 340 milliseconds after the cue and engaged a bidirectional recurrent loop: decision-related information first appeared in dlPFC, then traveled backward to A1, then returned to dlPFC. The flow was not a one-way relay but a back-and-forth exchange between sensory and cognitive regions.
What this means
The finding challenges the traditional view of primary sensory areas as passive relay stations that simply process raw input and pass it up the hierarchy. Here, A1 is actively engaged in the decision-making process itself, receiving choice-related signals from prefrontal cortex and sending processed information back.
“The cortex does not simply pass data up a ladder,” the pattern suggests. “It talks in rounds, mixing raw sound with plans and expectations.”
The dorsolateral prefrontal cortex is known to be critical for working memory, rule-based behavior, and decision-making. That it would send choice-related signals back to primary sensory cortex indicates that the brain’s executive regions do not compute decisions in isolation and then broadcast the result, they consult, in real time, the sensory regions that originally received the input.
Methodological strength
The study used simultaneous laminar multi-contact probes (“V-Trodes”) inserted across cortical layers in both A1 and dlPFC, a technically demanding setup that allowed the team to compare information timing and directional flow directly between the two regions, rather than inferring connectivity from separate recordings. Analyses included population trajectory analysis via PCA, SVM decoding, cross-area prediction, and directed-information flow analysis.
The work was supported by the Intramural Research Program of the NIMH (grant ZIA MH002928) and represents the efforts of the Laboratory of Neuropsychology at the NIH.
Sources:
1. Giarrocco, F. & Averbeck, B.B. “Neuronal dynamics, timing, and flow of sensory and choice-related information in auditory-prefrontal circuitry.” Nature Communications (2026). DOI: 10.1038/s41467-026-75349-2
2. Laboratory of Neuropsychology, National Institute of Mental Health, NIH, Bethesda, MD.