Behavioral origin of sound-evoked activity in mouse visual cortex

Behavioral origin of sound-evoked activity in mouse visual cortex Bimbard Sit Lebedeva Reddy Harris Carandini

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Behavioral origin of sound-evoked activity in mouse visual cortex 2023 Nature Neuroscience 10.1038/s41593-022-01227-x Bimbard2023-behavioral-origin-of-sound-evoked-activity-i Bimbard2023-behavioral-origin-of-sound-evoked-activity-i.txt, the project's staged copy 334d0500edd66a70653b4e3b1d1af3fec8c8ded7 not-redistributable neuroscience

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them whether the activity that sounds evoke in visual cortex is an auditory signal at all, or a low-dimensional signal of internal state and movement Quotation withheld because this record is not cleared for redistribution. whether that activity has the attributes of a behavioral signal - low dimension, a footprint beyond the cortex, independence from auditory input, and predictability from movement Quotation withheld because this record is not cleared for redistribution. the sound-evoked activity is one-dimensional, identical in the hippocampal formation, unchanged by cutting the auditory-to-visual fibers, and fully predicted by the small movements the sounds elicit Quotation withheld because this record is not cleared for redistribution. mechanism Quotation withheld because this record is not cleared for redistribution. sound-evoked activity in visual cortex Quotation withheld because this record is not cleared for redistribution. electrophysiology electrophysiology Quotation withheld because this record is not cleared for redistribution. mechanisms Sounds change the animal's internal state, which simultaneously drives uninstructed body movements and a widespread low-dimensional modulation of cortical (and hippocampal) activity; the apparent multisensory response is that state signal, not an auditory input. Quotation withheld because this record is not cleared for redistribution. mechanisms Sound-evoked V1 activity is generated by sound-triggered changes in internal and behavioral state rather than by auditory-cortical projections. Quotation withheld because this record is not cleared for redistribution. sounds elicit small uninstructed movements, mainly whisker twitches, which are stereotyped across trials and mice and different for each sound Quotation withheld because this record is not cleared for redistribution. and the movement precedes the neural activity by a few tens of milliseconds, even with no stimulus present Quotation withheld because this record is not cleared for redistribution. the subspace of activity predictable from movement during spontaneous behavior overlaps the sound-related subspace far more than the video-related one Quotation withheld because this record is not cleared for redistribution. and the paper stops short of a direct cause - cortical activity and body movements may both arise from changes in the internal state instead Quotation withheld because this record is not cleared for redistribution. The results do not rule out genuine auditory signals reaching visual cortex from auditory cortex. what this limits: the negative conclusion about auditory input: OUR FINDINGS DO NOT EXCLUDE THE POSSIBILITY OF GENUINE AUDITORY SIGNALS INHERITED BY VISUAL CORTEX FROM AUDITORY CORTEX Quotation withheld because this record is not cleared for redistribution. Recordings sampled mainly deep layers, not layer 2/3 where earlier conflicting results were obtained. what this limits: any comparison with the earlier conflicting layer 2/3 results: OUR V1 RECORDINGS WERE BIASED TOWARD LAYERS 4-6 Quotation withheld because this record is not cleared for redistribution. Auditory projections might affect only a small or low-firing subset of V1 neurons that these recordings missed. what this limits: the same negative conclusion, from the sampling side: IT IS ALSO POSSIBLE THAT AUDITORY PROJECTIONS AFFECT ONLY A MINOR FRACTION OF V1 NEURONS, OR THAT THEY AFFECT NEURONS THAT DO NOT FIRE AT HIGH RATES, AND THAT WE MISSED THESE NEURONS IN OUR RECORDINGS Quotation withheld because this record is not cleared for redistribution. Only facial movements were measured and only with linear analyses; other body parts or richer models could predict the neural activity better. what this limits: the claim that facial movement predicts the activity well: IT IS POSSIBLE THAT MOVEMENTS OF OTHER BODY PARTS, OR MORE COMPLEX ANALYSES, WOULD PROVIDE EVEN BETTER PREDICTIONS OF THE NEURAL ACTIVITY ELICITED BY SOUNDS Quotation withheld because this record is not cleared for redistribution. The behavioral-origin conclusion is scoped to these experiments. what this limits: the behavioural-origin conclusion, bounded by the paper itself: THESE RESULTS SUGGEST THAT, AT LEAST IN OUR EXPERIMENTS, THE SOUND-EVOKED ACTIVITY HAD A BEHAVIORAL ORIGIN Quotation withheld because this record is not cleared for redistribution. genuine auditory signals are not excluded - the projections exist and may carry them for other stimuli or contexts, or affect neurons the recordings missed what this limits: the same scope as limitations[0], with the anatomy: PROJECTIONS FROM AUDITORY TO VISUAL CORTEX EXIST AND MAY CARRY AUDITORY SIGNALS FOR OTHER stimuli or contexts Quotation withheld because this record is not cleared for redistribution. only facial movements were measured and the analyses are linear, so other body parts or better analyses could predict even more what this limits: the prediction result: ALL THE MOVEMENTS WE MEASURED ARE IN THE FACE, AND OUR ANALYSES ARE LINEAR Quotation withheld because this record is not cleared for redistribution. the disagreement with the study that introduced the same cut cannot be settled here, and three method differences are offered instead what this limits: the disagreement with the transectomy study: THE PREVIOUS STUDY WAS CONDUCTED INTRACELLULARLY AND MOSTLY IN LAYERS 2/3, against extracellular recording in layers 4-6 here Quotation withheld because this record is not cleared for redistribution. and the disagreement with the whisking study is likewise attributed to recording method, layer, and an analysis that used a hard threshold where this one estimated a fraction what this limits: the comparison with the whisking study, which agrees in part: IN AGREEMENT WITH OUR RESULTS, THIS STUDY FOUND THAT WHISKING EXPLAINS A FRACTION OF THOSE NEURONS' SOUND-EVOKED ACTIVITY Quotation withheld because this record is not cleared for redistribution. the low dimension and brain-wide footprint of behavioral signals, used here as a diagnostic signature Quotation withheld because this record is not cleared for redistribution. the marginalisation that separates video-related from sound-related components Quotation withheld because this record is not cleared for redistribution. cross-validated principal component analysis, and the contrast with a high-dimensional image code Quotation withheld because this record is not cleared for redistribution. a published fiber atlas, used to estimate what fraction of the projection the cut actually removed Quotation withheld because this record is not cleared for redistribution. a ridge regression of the auditory PC1 on the motion PCs, run alongside the cross-correlogram eliminates: the AUTOCORRELATION of the two signals as the source of an apparently large cross-correlation - two slow signals correlate at long lags whether or not they are related · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. movement → neural activity CORRELATION WITH A STATED LEAD-LAG, and no type for it. Auditory PC1 activity correlated with movement during spontaneous behaviour (r = 0.29 +/- 0.03 s.e.), and movement PRECEDED neural activity by a few tens of milliseconds. Recording this as 'causes' in either direction overstates it; recording it as nothing loses a measured result with its direction. Replaces the retracted concept_edges[32], which pointed the other way. Quotation withheld because this record is not cleared for redistribution. internal state → behavior CO-ORDINATE, NOT CAUSAL. The paper names them as a pair every time - 'low-dimensional changes in internal state and behavior', 'behavioral and state signals have profound effects on sensory areas', 'these behavioral and state signals are low-dimensional and largely orthogonal'. Both are offered as joint explanations of the multisensory signals, and movement is an instance of the behavioural half. No edge type can record that two things occupy the same role. Quotation withheld because this record is not cleared for redistribution. body movements → multisensory THE PAPER'S ACTUAL CLAIM, WHICH THE RETRACTED EDGE INVERTED: stereotyped body movements are elicited by sounds, images and odours, and those movements EXPLAIN neural activity previously interpreted as multisensory. In their own experiments video-evoked responses in hippocampal formation 'could be largely explained by video-evoked body movements'. Quotation withheld because this record is not cleared for redistribution. sound-evoked activity → hippocampal formation THE ACTIVITY RESEMBLES, NOT THE REGIONS: sound-evoked activity in V1 is stereotyped across neurons and mice, independent of auditory-cortex projections, and Quotation withheld - which receives little direct auditory input. The paper's account is that both reflect small stereotyped body movements, so the resemblance is evidence of a COMMON CAUSE and not of any identity between V1 and the hippocampal formation. Quotation withheld because this record is not cleared for redistribution. AN APPARENTLY MULTISENSORY SIGNAL EXPLAINED AWAY BY behavior: neural activity that is apparently multisensory MAY SIMPLY ARISE from low-dimensional signals associated with internal state and behavior Quotation withheld because this record is not cleared for redistribution. and the account being replaced is stated first, with its evidence - suppression by inhibiting auditory cortex, mimicry by stimulating auditory fibers Quotation withheld because this record is not cleared for redistribution. THE HYPOTHESIS IS TURNED INTO FOUR ADVANCE PREDICTIONS: low dimension, a broad footprint beyond the cortex, independence from auditory input, and predictability from behavior Quotation withheld because this record is not cleared for redistribution. THREE MODELS, AND THE RANKING IS THE FINDING: one from sound identity alone, one from movement alone, one from both - and the behavioral model performs JUST AS WELL as the full model Quotation withheld because this record is not cleared for redistribution. with the auditory model's built-in advantage disclosed - it is equivalent to a test-retest prediction and would fit perfectly with infinite trials, so it performs well REGARDLESS of the origin of the activity Quotation withheld because this record is not cleared for redistribution. THE MANIPULATION WAS CHOSEN FOR WHAT IT DOES NOT DISTURB: silencing or stimulating auditory cortex would interfere with auditory processing and could affect the sound-evoked behavior, so the fibres were cut instead Quotation withheld because this record is not cleared for redistribution. and the control is INSIDE the animal - bilateral recording, the uncut side, the same mouse and the same behavior Quotation withheld because this record is not cleared for redistribution. THREE METHOD DIFFERENCES ENUMERATED WHERE THE RESULT CONTRADICTS THE STUDY THAT INTRODUCED THE SAME CUT: intracellular in layers 2/3 against extracellular in layers 4-6, hours against days after surgery, and anesthetized against awake Quotation withheld because this record is not cleared for redistribution. AND A THRESHOLD AGAINST A FRACTION: the earlier whisking study used a hard threshold to call a response auditory or movement-related, where this one estimated the fraction explained Quotation withheld because this record is not cleared for redistribution. THE DIMENSIONALITY OF THE RESPONSE IS USED TO CLASSIFY WHAT KIND OF SIGNAL IT IS: four significant components for sounds against seventy-nine for videos, and zero for their interaction Quotation withheld because this record is not cleared for redistribution. with a between-animal correlation reported against its own within-animal ceiling - 0.34 across mice against a test-retest of 0.44 within them Quotation withheld because this record is not cleared for redistribution. A SELECTION STEP CHECKED AGAINST THE RESULT IT COULD HAVE PRODUCED: the reliability of auditory responses did not correlate with the spike-contamination score, and if anything went the other way Quotation withheld because this record is not cleared for redistribution. A WHOLE LITERATURE'S CONTROLS DECLARED INADEQUATE: distinguishing auditory signals from state and behavior requires careful and systematic controls, WHICH ARE RARELY PERFORMED in passively listening mice Quotation withheld because this record is not cleared for redistribution. and the correlation is not claimed as causation - cortical activity and body movements may BOTH arise from changes in the internal state Quotation withheld because this record is not cleared for redistribution. with the same suspicion extended outward - stereotyped movements are elicited by images and odors too, so other multisensory findings may have the same origin Quotation withheld because this record is not cleared for redistribution. claim_type: critique a second reading DIFFERING from the recorded value. Recorded: mechanism — note: a second reading, verbatim-verified, disagreeing with the value on the record. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. method: three encoding models fitted to the same responses - one from sound identity alone, one from measured movement alone, one from both - with the ranking of their predictions as the result a second reading DIFFERING from the recorded value. Recorded: electrophysiology — note: a second reading, verbatim-verified, disagreeing with the value on the record. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. phenomenon: activity evoked by sounds in visual cortex, previously read as multisensory integration a second reading DIFFERING from the recorded value. Recorded: sound-evoked activity in visual cortex — note: a second reading, verbatim-verified, disagreeing with the value on the record. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. contradicts → studies ascribing sound-evoked V1 activity to direct input from auditory cortex Quotation withheld because this record is not cleared for redistribution. contradicts → the original study that introduced the auditory-to-visual transectomy Quotation withheld because this record is not cleared for redistribution. replicates → the study of whisking contributions to sound-evoked activity in layer 1 V1 neurons Quotation withheld because this record is not cleared for redistribution. reinterprets → sound-evoked activity in visual cortex, as a signal of movement and internal state rather than an auditory input Quotation withheld because this record is not cleared for redistribution. contradicts → the study that introduced the transectomy and found the activity abolished, with three method differences named differs in: the recording mode and the layer: THE PREVIOUS STUDY WAS CONDUCTED INTRACELLULARLY AND MOSTLY IN LAYERS 2/3 (WHERE SOUNDS HYPERPOLARIZED CELLS, UNLIKE IN OTHER LAYERS WHERE SOUNDS INCREASED SPIKING), WHEREAS WE RECORDED EXTRACELLULARLY MAINLY IN LAYERS 4-6 Quotation withheld because this record is not cleared for redistribution. qualifies → an earlier report that whisking explains only a fraction of sound-evoked activity, by pointing at its hard threshold differs in: the recording technique: THIS DISCREPANCY COULD BE DUE TO DIFFERENCES IN RECORDING METHODS (TWO-PHOTON IMAGING VERSUS ELECTROPHYSIOLOGY), on a point where the two studies otherwise agree Quotation withheld because this record is not cleared for redistribution. confirms → the reported low dimension and brain-wide footprint of movement-related activity Quotation withheld because this record is not cleared for redistribution. generalises → the behavioural explanation to other multisensory findings, since images and odors also elicit stereotyped movements Quotation withheld because this record is not cleared for redistribution. critiques → the controls used in passively listening mice, as rarely systematic enough to separate auditory from behavioral signals differs in: which movements were controlled for: SOME STUDIES HAVE CONTROLLED FOR EYE MOVEMENTS OR FOR OVERT BEHAVIORS SUCH AS LICKING. HOWEVER, EVEN THESE STUDIES MAY HAVE OVERLOOKED THE TYPES OF MOVEMENT THAT WE OBSERVED Quotation withheld because this record is not cleared for redistribution. multisensory —distinct-from→ internal state multisensory internal state strength of the relation: 1 · note: AN APPARENTLY MULTISENSORY SIGNAL EXPLAINED AWAY: it MAY SIMPLY ARISE from low-dimensional signals associated with internal state and behavior Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span offers internal state as an alternative explanation of multisensory signals; distinct-from is loose. Quotation withheld because this record is not cleared for redistribution. sound —causes→ V1 sound V1 strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says V1 is influenced by sounds; causes 'V1' names a region, not an effect. Quotation withheld because this record is not cleared for redistribution. low-dimensional —contrasted-with→ high-dimensional low-dimensional high-dimensional strength of the relation: 1 Quotation withheld because this record is not cleared for redistribution. behavior —causes→ multisensory behavior multisensory strength of the relation: 1 · note: the alternative offered explicitly as an alternative EXPLANATION, before any of its own data Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span offers behaviour as an alternative explanation; causes is loose. Quotation withheld because this record is not cleared for redistribution. video —contrasted-with→ sound video sound strength of the relation: 1 · note: separated by a marginalisation borrowed from factorial analysis of variance, each averaged over the other and referred to the grand mean Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span decomposes responses into visual and auditory components; contrasted-with is loose. Quotation withheld because this record is not cleared for redistribution. cvPCA —operationalised-as→ PC1 cvPCA PC1 strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but PC1 is the first component FOUND BY cvPCA; operationalised-as is loose. Quotation withheld because this record is not cleared for redistribution. intracellular —contrasted-with→ extracellular intracellular extracellular strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span notes the earlier study was intracellular; the contrast is implied. Quotation withheld because this record is not cleared for redistribution. whisker —is-a→ movement whisker movement strength of the relation: 1 · note: subtle whisker twitches, quantified as the first component of facial motion energy Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span images whisker and body movements; is-a is loose. Quotation withheld because this record is not cleared for redistribution. auditory model —contrasted-with→ behavioral model auditory model behavioral model differs in: THE AXIS IS WHETHER THE MODEL CAN DISCRIMINATE THE ORIGIN OF THE ACTIVITY. 'THE FIRST WAS A purely auditory model where the time course of neural activity depends only on sound identity. This model is EQUIVALENT TO A TEST-RETEST PREDICTION, so it will perform well regardless of the origin of sound-evoked activity.' A model that cannot fail is not evidence, and the authors say so about their own first model before using the others. Same shape as a row of Desimone1996 in this same batch. · strength of the relation: 1 · note: AND THE AUDITORY MODEL'S BUILT-IN ADVANTAGE IS DISCLOSED: it is equivalent to a test-retest prediction and would fit perfectly with infinite trials, so it performs well REGARDLESS of the origin of the activity Quotation withheld because this record is not cleared for redistribution. pupil —part-of→ behavioral model pupil behavioral model strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but pupil is one input to the behavioural model; part-of is loose. Quotation withheld because this record is not cleared for redistribution. sound —distinct-from→ behavior sound behavior differs in: THE AXIS IS NECESSITY VERSUS SUFFICIENCY FOR PREDICTING THE ACTIVITY. 'Quotation withheld' The two are separated not by what they are but by their standing in one prediction, and the paper's title claim rests entirely on that asymmetry. · strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says the sounds were unnecessary and the movements sufficient to predict V1 activity; distinct-from is loose. Quotation withheld because this record is not cleared for redistribution. movement —distinct-from→ internal state movement internal state strength of the relation: 0.8 · note: THE CORRELATION IS NOT CLAIMED AS CAUSATION: cortical activity and body movements may BOTH arise from changes in internal state, which also covers the anesthetized case where movement is impossible Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says movement and cortical activity may both arise from internal state; distinct-from is loose. Quotation withheld because this record is not cleared for redistribution. V1 —same-as→ visual cortex V1 visual cortex strength of the relation: 1 · note: used interchangeably throughout, with the recording sites pooled under the second name Quotation withheld because this record is not cleared for redistribution. Wilcoxon signed-rank test Wilcoxon signed-rank test Quotation withheld because this record is not cleared for redistribution. principal component analysis (cross-validated) principal component analysis (cross-validated) Quotation withheld because this record is not cleared for redistribution. ridge regression ridge regression Quotation withheld because this record is not cleared for redistribution. reduced-rank regression reduced-rank regression Quotation withheld because this record is not cleared for redistribution. cross-validation cross-validation Quotation withheld because this record is not cleared for redistribution. ANOVA ANOVA Quotation withheld because this record is not cleared for redistribution. permutation/shuffling test permutation/shuffling test Quotation withheld because this record is not cleared for redistribution. correlation correlation Quotation withheld because this record is not cleared for redistribution. Fisher z-transformation Fisher z-transformation Quotation withheld because this record is not cleared for redistribution. F-test F-test Quotation withheld because this record is not cleared for redistribution. Wilcoxon right-tailed signed rank test Wilcoxon right-tailed signed rank test Quotation withheld because this record is not cleared for redistribution. reduced-rank regression reduced-rank regression Quotation withheld because this record is not cleared for redistribution. cross-validated Principal Component Analysis cross-validated Principal Component Analysis Quotation withheld because this record is not cleared for redistribution. shuffling method shuffling method Quotation withheld because this record is not cleared for redistribution. The population activity that sounds evoke in mouse V1 is so stereotyped across neurons that it is essentially one-dimensional. Quotation withheld because this record is not cleared for redistribution. Sound-evoked activity in the hippocampal formation is remarkably similar to that in visual cortex. Quotation withheld because this record is not cleared for redistribution. The sound-evoked activity in visual cortex in these experiments cannot be explained by direct inputs from auditory cortex. Quotation withheld because this record is not cleared for redistribution. A model using only sound-elicited body movements predicted sound-evoked V1 activity as well as a model that also had access to sound identity. Quotation withheld because this record is not cleared for redistribution. The sound-elicited movements lead the neural activity in time, consistent with movement (or the state driving it) preceding the cortical signal. Quotation withheld because this record is not cleared for redistribution. Sounds evoked small but highly reproducible uninstructed body movements that differed between sounds. Quotation withheld because this record is not cleared for redistribution. The population response showed no significant components for interactions between videos and sounds. Quotation withheld because this record is not cleared for redistribution. AN APPARENTLY MULTISENSORY SIGNAL EXPLAINED AWAY BY behavior: it MAY SIMPLY ARISE from low-dimensional signals of internal state and behavior Quotation withheld because this record is not cleared for redistribution. with the account being replaced stated first, including its evidence from silencing and from stimulating auditory fibers Quotation withheld because this record is not cleared for redistribution. the alternative is offered explicitly as an alternative explanation, before any of the paper's own data Quotation withheld because this record is not cleared for redistribution. THE HYPOTHESIS IS TURNED INTO FOUR ADVANCE PREDICTIONS: low dimension, a footprint beyond cortex, independence from auditory input, predictability from behavior Quotation withheld because this record is not cleared for redistribution. and all four are then reported as met, in the order they were stated Quotation withheld because this record is not cleared for redistribution. video and sound components are separated by a marginalisation borrowed from factorial analysis of variance Quotation withheld because this record is not cleared for redistribution. sound identity is decodable from single trials at 18% against a chance of one in twelve, and video identity at 95% Quotation withheld because this record is not cleared for redistribution. the sound-evoked activity is essentially ONE-DIMENSIONAL, one component carrying 55% of the sound-related variance Quotation withheld because this record is not cleared for redistribution. A BETWEEN-ANIMAL CORRELATION REPORTED AGAINST ITS OWN WITHIN-ANIMAL CEILING: 0.34 across mice against a test-retest of 0.44 within them Quotation withheld because this record is not cleared for redistribution. while the video-evoked activity is larger and much higher dimensional Quotation withheld because this record is not cleared for redistribution. the same probe reached the hippocampal formation, which receives little input from auditory cortex or thalamus Quotation withheld because this record is not cleared for redistribution. and the sound-evoked activity there is hardly distinguishable from the one in visual cortex, r = 0.82 Quotation withheld because this record is not cleared for redistribution. with the prediction stated before the result - IF the activity comes from auditory cortex it SHOULD BE DRASTICALLY REDUCED on the cut side Quotation withheld because this record is not cleared for redistribution. THE CENTRAL NULL: identical to the opposite hemisphere at r = 0.9 and to intact controls, with cross-correlation peaking at zero delay Quotation withheld because this record is not cleared for redistribution. THE MANIPULATION WAS CHOSEN FOR WHAT IT DOES NOT DISTURB: silencing or stimulating would interfere with auditory processing and could change the behavior Quotation withheld because this record is not cleared for redistribution. THREE METHOD DIFFERENCES against the study that introduced the transectomy: the previous study was conducted intracellularly and mostly in layers 2/3 where sounds hyperpolarized cells, performed recordings hours after the transectomy, and anesthetized the mice Quotation withheld because this record is not cleared for redistribution. sounds evoke small uninstructed movements, mainly whisker twitches, stereotyped across trials and mice Quotation withheld because this record is not cleared for redistribution. and sound identity decodes from the face at 16%, not significantly different from the 18% from the neurons Quotation withheld because this record is not cleared for redistribution. the two time courses correlate at r = 0.75, and the mice that moved more consistently carried more sound information in V1 Quotation withheld because this record is not cleared for redistribution. the coupling survives with no stimulus at all, and the MOVEMENT PRECEDES the activity by a few tens of milliseconds Quotation withheld because this record is not cleared for redistribution. the movement subspace, defined during the no-stimulus period, explains 75% of the sound-related variance and only 35% of the video-related Quotation withheld because this record is not cleared for redistribution. AND THE AUDITORY MODEL'S BUILT-IN ADVANTAGE IS DISCLOSED: it is a test-retest prediction and would fit perfectly with infinite trials Quotation withheld because this record is not cleared for redistribution. THREE MODELS AND THE RANKING IS THE FINDING: the behavioral model performs just as well as the full model, so the sounds themselves were unnecessary Quotation withheld because this record is not cleared for redistribution. THE CONTROL THAT MAKES THE CLAIM SPECIFIC: a purely visual model explains the video-evoked activity where the behavioral one does not Quotation withheld because this record is not cleared for redistribution. and the trial-to-trial variation is captured too, which the auditory model cannot do BY DEFINITION Quotation withheld because this record is not cleared for redistribution. the conclusion is stated with its scope - AT LEAST IN OUR EXPERIMENTS, the activity had a behavioral origin Quotation withheld because this record is not cleared for redistribution. AND THE NULL IS BOUNDED THREE WAYS: other stimuli, other contexts, or low-firing neurons the recordings missed Quotation withheld because this record is not cleared for redistribution. A WHOLE LITERATURE'S CONTROLS DECLARED INADEQUATE: the necessary controls ARE RARELY PERFORMED in passively listening mice Quotation withheld because this record is not cleared for redistribution. A THRESHOLD AGAINST A FRACTION: the earlier whisking study used a hard threshold where this one estimated the fraction explained Quotation withheld because this record is not cleared for redistribution. THE CORRELATION IS NOT CLAIMED AS CAUSATION: activity and movement may BOTH arise from changes in internal state Quotation withheld because this record is not cleared for redistribution. and the suspicion is extended outward - images and odors elicit stereotyped movements too Quotation withheld because this record is not cleared for redistribution. with the limitation running in the direction that would STRENGTHEN the claim - only facial movements, only linear analyses Quotation withheld because this record is not cleared for redistribution. A SELECTION STEP CHECKED AGAINST THE RESULT IT COULD HAVE PRODUCED: the contamination score did not correlate with response reliability, and if anything went the other way Quotation withheld because this record is not cleared for redistribution. THE DIMENSIONALITY CLASSIFIES THE SIGNAL: 79 significant components for videos, 4 for sounds, ZERO for their interaction Quotation withheld because this record is not cleared for redistribution. with components found on one half of the data and their reliability measured on the other Quotation withheld because this record is not cleared for redistribution. the extent of the cut was estimated by overlaying tractography from the Allen connectivity atlas on the reconstructed cut Quotation withheld because this record is not cleared for redistribution. and the mice were habituated to the rig but NOT to the specific stimuli before the experiment Quotation withheld because this record is not cleared for redistribution. with the recording areas pooled under one name after checking that the exact location did not affect the results Quotation withheld because this record is not cleared for redistribution. auditory PC1 auditory PC1 Quotation withheld because this record is not cleared for redistribution. dimensionality dimensionality Quotation withheld because this record is not cleared for redistribution. behavioral subspace behavioral subspace Quotation withheld because this record is not cleared for redistribution. body motion variables body motion variables Quotation withheld because this record is not cleared for redistribution. cvPCA cvPCA Quotation withheld because this record is not cleared for redistribution. subspace subspace Quotation withheld because this record is not cleared for redistribution. auditory model auditory model Quotation withheld because this record is not cleared for redistribution. behavioral model behavioral model Quotation withheld because this record is not cleared for redistribution. dimension dimension Quotation withheld because this record is not cleared for redistribution. transectomy transectomy Quotation withheld because this record is not cleared for redistribution. binned firing rate binned firing rate Quotation withheld because this record is not cleared for redistribution. body movements body movements Quotation withheld because this record is not cleared for redistribution. decoding accuracy decoding accuracy Quotation withheld because this record is not cleared for redistribution. dimension dimension Quotation withheld because this record is not cleared for redistribution. variance variance Quotation withheld because this record is not cleared for redistribution. accuracy accuracy Quotation withheld because this record is not cleared for redistribution. correlation correlation Quotation withheld because this record is not cleared for redistribution. 55% — share of the cross-validated sound-related variance in V1 explained by the first principal component of sound-related activity Quotation withheld because this record is not cleared for redistribution. 18 ± 2% — single-trial accuracy of decoding which sound was played from marginalized V1 population activity (chance = 1/12) Quotation withheld because this record is not cleared for redistribution. 16 ± 2% — accuracy of decoding sound identity from facial motion components alone, statistically indistinguishable from decoding it from V1 activity Quotation withheld because this record is not cleared for redistribution. r = 0.82 — correlation between the auditory PC1 time courses measured in visual cortex and in the hippocampal formation Quotation withheld because this record is not cleared for redistribution. 75 ± 3% — fraction of sound-related variance explained by the first four components of the movement-related subspace defined during stimulus-free periods Quotation withheld because this record is not cleared for redistribution. 79 — average number of significant cvPCA components for video-related activity, versus four for sound-related activity Quotation withheld because this record is not cleared for redistribution. four — average number of significant cvPCA dimensions carrying sound-related variance in V1 Quotation withheld because this record is not cleared for redistribution. >3.6 — estimated factor by which the transectomy reduced auditory-cortical input to the visual areas on the cut side relative to the intact side Quotation withheld because this record is not cleared for redistribution. 640 — number of single cells recorded in primary visual cortex across eight mice (plus 233 in hippocampal formation) Quotation withheld because this record is not cleared for redistribution. 0.39 — correlation between trial-by-trial fluctuations of V1 auditory PC1 and their cross-validated prediction from movement Quotation withheld because this record is not cleared for redistribution. 0.34 — the correlation of the sound-evoked component across different mice, reported against its own within-mouse test-retest ceiling of 0.44 Quotation withheld because this record is not cleared for redistribution. 18 ± 2% — sound decoding accuracy from V1 neurons against a chance of one in twelve, and against 16% from the face alone Quotation withheld because this record is not cleared for redistribution. r = 0.82 — the correlation between the sound-evoked components of visual cortex and hippocampal formation - two regions with very different auditory input Quotation withheld because this record is not cleared for redistribution. 75 ± 3% — the sound-related variance explained by a subspace defined entirely from movement during the no-stimulus period, against 35% of the video-related variance Quotation withheld because this record is not cleared for redistribution. zero significant components — the audiovisual interaction term - a reported null that removes the simplest form of multisensory integration Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1038/s41593-022-01227-x.