P-9 Difference in readout, not encoding, of sensory evidence for continuous decisions dependent on consistency with previous categorical judgment

TY - JOUR

T1 - P-9 Difference in readout, not encoding, of sensory evidence for continuous decisions dependent on consistency with previous categorical judgment

AU - Park, Hame

AU - Arazi, Ayelet

AU - Talluri, Bharath Chandra

AU - Celotto, Marco

AU - Panzeri, Stefano

AU - Stocker, A.

AU - Donner, Tobias

PY - 2023

Y1 - 2023

N2 - Background: Many perceptual decisions are based on the accumulationof protracted streams of sensory evidence. Recent behavioralwork has required participants to produce a continuous estimateof the mean of a stream of visual evidence, combined with a categoricaljudgment in the middle of the stream. In such tasks, human subjectsput stronger weight on subsequent evidence samples that areconsistent compared to inconsistent with that intermittent categoricaljudgment. The neural basis of this cognitive bias is unknown.Objective: We asked whether changes in evidence weightingdepending on consistency with previous categorical judgment stemsfrom (i) encoding of the evidence in visual cortex or (ii) in the readoutof this information for a final continuous estimation report.Methods: During magnetoencephalography (MEG), 34 healthyhuman participants performed an estimation task based on a streamof visual evidence: the angular positions of checkerboard patchesdrawn from a Gaussian distribution (trial varying mean: 1414,fixed s.d.: 20). The task was to report the sample mean (mean of thestream of visual stimuli) by moving a cursor on the screen at the endof the trial. Importantly, after seeing half of the stream, they made acategorical judgment (the mean is on the left vs. right) about thevisual stimuli or viewed a left vs. right cue (75% validity) about themean of the stream. We reconstructed the MEG responses followingsample onsets for areas across the visual cortical hierarchy and usedinformation-theoretic measures to quantify the relationshipbetween the local response pattern and evidence sample and finalestimate. We also quantified a statistical signature of sensory readout:the ‘intersection information’ between evidence sample, neuralresponse, and estimation.Results: At the behavioral level, participants gave stronger weightfor samples consistent compared to inconsistent with the previousintermittent categorical judgment or cue. At the neural level, therewas no difference in the encoding of evidence samples (quantifiedas mutual information) between consistent and inconsistent samplesin any visual cortical area. Critically, the intersection informationin early visual cortex (areas V2-V4, similar but weaker in V1)and downstream areas was larger for evidence samples consistentthan inconsistent with the categorical choice or cue.Conclusion: The consistency between categorical information (ownjudgment or external cue) and subsequent evidence samples doesnot change the encoding of the evidence sample in early sensory cortex,but selectively modulates the readout of these sensory evidencesignals in the downstream processing underlying the behavioralestimation report. This differential sensory readout could be due todifferences in the routing of the sample information to downstreamcircuits involved in integrating the samples into an evolving decisionvariable for the estimation, in the updating of that decision variableby the new sample, or both.

AB - Background: Many perceptual decisions are based on the accumulationof protracted streams of sensory evidence. Recent behavioralwork has required participants to produce a continuous estimateof the mean of a stream of visual evidence, combined with a categoricaljudgment in the middle of the stream. In such tasks, human subjectsput stronger weight on subsequent evidence samples that areconsistent compared to inconsistent with that intermittent categoricaljudgment. The neural basis of this cognitive bias is unknown.Objective: We asked whether changes in evidence weightingdepending on consistency with previous categorical judgment stemsfrom (i) encoding of the evidence in visual cortex or (ii) in the readoutof this information for a final continuous estimation report.Methods: During magnetoencephalography (MEG), 34 healthyhuman participants performed an estimation task based on a streamof visual evidence: the angular positions of checkerboard patchesdrawn from a Gaussian distribution (trial varying mean: 1414,fixed s.d.: 20). The task was to report the sample mean (mean of thestream of visual stimuli) by moving a cursor on the screen at the endof the trial. Importantly, after seeing half of the stream, they made acategorical judgment (the mean is on the left vs. right) about thevisual stimuli or viewed a left vs. right cue (75% validity) about themean of the stream. We reconstructed the MEG responses followingsample onsets for areas across the visual cortical hierarchy and usedinformation-theoretic measures to quantify the relationshipbetween the local response pattern and evidence sample and finalestimate. We also quantified a statistical signature of sensory readout:the ‘intersection information’ between evidence sample, neuralresponse, and estimation.Results: At the behavioral level, participants gave stronger weightfor samples consistent compared to inconsistent with the previousintermittent categorical judgment or cue. At the neural level, therewas no difference in the encoding of evidence samples (quantifiedas mutual information) between consistent and inconsistent samplesin any visual cortical area. Critically, the intersection informationin early visual cortex (areas V2-V4, similar but weaker in V1)and downstream areas was larger for evidence samples consistentthan inconsistent with the categorical choice or cue.Conclusion: The consistency between categorical information (ownjudgment or external cue) and subsequent evidence samples doesnot change the encoding of the evidence sample in early sensory cortex,but selectively modulates the readout of these sensory evidencesignals in the downstream processing underlying the behavioralestimation report. This differential sensory readout could be due todifferences in the routing of the sample information to downstreamcircuits involved in integrating the samples into an evolving decisionvariable for the estimation, in the updating of that decision variableby the new sample, or both.

U2 - 10.1016/j.clinph.2023.02.026

DO - 10.1016/j.clinph.2023.02.026

M3 - Conference abstract in journal

VL - 148

SP - e12-e13

JO - CLIN NEUROPHYSIOL

JF - CLIN NEUROPHYSIOL

SN - 1388-2457

ER -