A category-free neural population supports evolving demands during decision-making

A category-free neural population supports evolving demands during decision-making Raposo Kaufman Churchland

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.

A category-free neural population supports evolving demands during decision-making 2014 Nature Neuroscience 10.1038/nn.3865 Raposo2014-a-category-free-neural-population-supports-e Raposo2014-a-category-free-neural-population-supports-e.txt, the project's staged copy be86e5ab19e9c25856578cc33b3e4cfd795ade67 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 multiple behaviors rely on distinct categories of neurons specialized to represent particular variables, or on a single population that is leveraged in different ways Quotation withheld because this record is not cleared for redistribution. whether neural populations are category-free - whether response features are distributed randomly across neurons rather than clustering into types Quotation withheld because this record is not cleared for redistribution. task parameters and temporal response features were distributed randomly across neurons, without evidence of categories, and the network explored different dimensions during decision and movement Quotation withheld because this record is not cleared for redistribution. existence Quotation withheld because this record is not cleared for redistribution. population coding in posterior parietal cortex during multisensory decision-making Quotation withheld because this record is not cleared for redistribution. electrophysiology electrophysiology Quotation withheld because this record is not cleared for redistribution. mechanisms because single neurons are jointly (and randomly) modulated by choice and modality, downstream readouts can weight the same population differently to recover whichever variable is currently needed Quotation withheld because this record is not cleared for redistribution. mechanisms random distribution of parameters across neurons lets an arbitrary subset be linearly combined to estimate a needed parameter, removing the need for pre-patterned connections to downstream targets Quotation withheld because this record is not cleared for redistribution. mechanisms moving population activity into different dimensions at different moments lets PPC gate which signals reach which downstream areas Quotation withheld because this record is not cleared for redistribution. mechanisms PPC inactivation impairs visual choices by lowering the signal-to-noise of the incoming visual evidence, not by changing the animal's decision strategy Quotation withheld because this record is not cleared for redistribution. when parameters are distributed randomly across neurons, an arbitrary group of them can be linearly combined to estimate the parameter needed at a given moment, which obviates the need for precisely pre-patterned connections and means all information is transmitted Quotation withheld because this record is not cleared for redistribution. the average angle between the choice decoder and the modality decoder was 86.5 degrees, ONLY SLIGHTLY LESS THAN A PERFECTLY ORTHOGONAL 90 Quotation withheld because this record is not cleared for redistribution. and the neurons retained their choice preference whether the stimulus was auditory, visual or multisensory, which the paper calls tolerance Quotation withheld because this record is not cleared for redistribution. Categories defined by cell type or connectivity rather than response features were not tested what this limits: the category-free claim, which is relative to the features tested: NEURAL CATEGORIES DEFINED BY OTHER PROPERTIES, SUCH AS CELL TYPE OR CONNECTIVITY, MIGHT REVEAL SPECIALIZATION Quotation withheld because this record is not cleared for redistribution. No clustering test can cover every conceivable way the data might cluster what this limits: the strength of the clustering null: IT IS PROBABLY NOT POSSIBLE TO DETECT ALL CONCEIVABLE WAYS IN WHICH THE DATA MIGHT CLUSTER, though the test is described as reasonably general Quotation withheld because this record is not cleared for redistribution. The choice/modality test leaves categories defined by other features unexamined what this limits: the same scope as limitations[0]: THIS LEAVES OPEN THE POSSIBILITY THAT THERE ARE CATEGORIES DEFINED BY OTHER FEATURES OF THE DATA Quotation withheld because this record is not cleared for redistribution. Decoding was from a small sample of neurons in a noise-limited task what this limits: the decoding accuracy: DECODING WAS PERFORMED ON A SAMPLE OF TENS OF NEURONS OUT OF MANY THOUSANDS, ON A TASK WHERE THE ANIMAL'S PERFORMANCE IS LIKELY NOISE LIMITED Quotation withheld because this record is not cleared for redistribution. Residual visual performance means the inactivation may have been incomplete or other structures may contribute what this limits: the inactivation result: ANIMALS RETAINED SOME ABILITY TO MAKE VISUAL JUDGMENTS DESPITE THE INACTIVATION, SUGGESTING EITHER THAT INACTIVATION WAS INCOMPLETE, OR THAT STRUCTURES IN ADDITION TO PPC SUPPORT THE TASK Quotation withheld because this record is not cleared for redistribution. neural categories defined by other properties, such as cell type or connectivity, might reveal specialization what this limits: categories defined by selectivity to task parameters or by time dependent response features, which is what this test covers Quotation withheld because this record is not cleared for redistribution. PAIRS is not only a test for clustering but more generally for non-uniformity of the distribution of tuning, and it is probably not possible to detect all conceivable ways in which the data might cluster what this limits: the category-free conclusion, whose test is broader than its name: PAIRS IS NOT ONLY A TEST FOR CLUSTERING, BUT ALSO MORE GENERALLY FOR NON-UNIFORMITY OF THE DISTRIBUTION OF TUNING ACROSS NEURONS Quotation withheld because this record is not cleared for redistribution. animals retained some ability to make visual judgments despite the inactivation, suggesting either that inactivation was incomplete, or that structures in addition to PPC support the task what this limits: the same scope as limitations[4], with the effect size: SHALLOWER PSYCHOMETRIC FUNCTIONS ON INACTIVATION DAYS (sigma_saline = 2.64 +/- 0.39; sigma_muscimol = 4.67 +/- 0.67) Quotation withheld because this record is not cleared for redistribution. relatively close neuron pairs were occasionally observed and differed from chance in 1 of 5 cases, so a small fraction of neurons do have pairs in feature space what this limits: the no-clustering claim, per animal: PAIRS INDEX FOR RATS 1-5: -0.011, 0.108, -0.038, 0.011, -0.007; p-values 0.621, 0.001, 0.209, 0.491, 0.857 - one of five deviates Quotation withheld because this record is not cleared for redistribution. auditory responses in PPC, though apparently not necessary for this task, may be invoked by other decision tasks what this limits: the negative inactivation result for audition: AUDITORY RESPONSES IN PPC, THOUGH APPARENTLY NOT NECESSARY FOR THIS TASK, MAY BE INVOKED BY OTHER DECISION TASKS Quotation withheld because this record is not cleared for redistribution. a linear Support Vector Machine, explicitly labelled a standard machine-learning technique, is imported as the decoder for single-trial neural population activity Quotation withheld because this record is not cleared for redistribution. tolerance, and the reading that a linear decoder working under many conditions shows a representation has been untangled Quotation withheld because this record is not cleared for redistribution. a linear support vector machine used as the decoder, and bootstrap aggregation to combine a thousand trained classifiers Quotation withheld because this record is not cleared for redistribution. the prediction that random mixtures let arbitrary groups of neurons be linearly combined, imported as the hypothesis to be tested rather than assumed Quotation withheld because this record is not cleared for redistribution. nearest-neighbor angles compared against randomly oriented vectors, turned into a clustering test for neural response profiles Quotation withheld because this record is not cleared for redistribution. synthetic categories introduced into the population, with noise derived from the real neural data eliminates: the lack of categories being a failure of the analysis to see structure - with faux neurons carrying exactly the same amount of noise as their originals, a strikingly different PAIRS distribution was evident · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. the distribution of decoder weights compared with random weightings eliminates: the decode resting on rare neurons with pure selectivity - the weights were nonzero for all neurons and did not differ from what would be expected by chance · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. auditory decisions measured alongside the impaired visual ones eliminates: reduced motivation or confusion about the stimulus-response contingency, since the sparing of auditory decisions rules both out · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. a second inactivation method that permits visualization of the agent to determine its spread eliminates: the impairment coming from spread into neighboring visual areas - and the rat with more expression posterior to PPC had WEAKER visual impairment · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. the Variance Alignment analysis repeated on visual against multisensory trials within the decision epoch eliminates: noise making two aligned state spaces look misaligned - this positive control gave strong alignment, which is the reassurance that the decision-movement misalignment is real · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. modulation during the stimulus compared against modulation during movement, neuron by neuron eliminates: the misalignment being two separate groups of active neurons - the two were strongly correlated, so it is a population-level phenomenon · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. the excess rate analysis over sliding windows of the stimulus eliminates: snap judgments or evidence leak as the explanation for the impairment - the analysis indicates inactivation reduced the signal-to-noise of incoming evidence instead · outcome: held Quotation withheld because this record is not cleared for redistribution. THE TARGET IS THE CATEGORY ASSUMPTION ITSELF: individual neurons are OFTEN SEEN AS MEMBERS OF HIGHLY SPECIALIZED CATEGORIES, and THE ASSUMPTION OF NEURAL CATEGORIES IS REFLECTED IN MANY EXPERIMENTAL DESIGNS AND ANALYSIS METHODS, even far downstream of early sensory areas Quotation withheld because this record is not cleared for redistribution. AND THE DEVICES THAT CARRY IT ARE NAMED. In sampling: neurons MUST MEET CERTAIN RESPONSE CRITERIA TO BE INCLUDED FOR STUDY, and IMPLICIT IN THIS APPROACH IS THE IDEA THAT THE CELL'S RESPONSE DURING ONE STIMULUS IDENTIFIES IT AS A MEMBER OF THE CATEGORY BEING EXAMINED Quotation withheld because this record is not cleared for redistribution. in analysis: PIE CHARTS, A COMMON WAY OF SUMMARIZING POPULATION DATA, EXPLICITLY ASSIGN NEURONS TO CATEGORIES - and AVERAGING OVER MANY NEURONS LIKEWISE REFLECTS THE ASSUMPTION THAT EACH NEURON IS AN EXEMPLAR OF A CATEGORY, DIFFERENT FROM OTHER CATEGORY MEMBERS MAINLY BECAUSE OF NOISE Quotation withheld because this record is not cleared for redistribution. with the alternative stated as a hypothesis about DISTRIBUTION rather than about types: neurons REFLECT RANDOM COMBINATIONS OF PARAMETERS, leading to populations in which NEURONS' RESPONSES DEFY CATEGORIZATION - and the advantage is that an ARBITRARY GROUP OF THEM CAN BE LINEARLY COMBINED to estimate the parameter needed at a given moment Quotation withheld because this record is not cleared for redistribution. and the earlier test is described by what it could not cover: Quotation withheld A MORE GENERAL TEST IS THUS REQUIRED Quotation withheld because this record is not cleared for redistribution. A NEW STATISTIC NAMED, DEFINED, AND BOUNDED BY ITS OWN AUTHORS: PAIRS, PROJECTION ANGLE INDEX OF RESPONSE SIMILARITY - 1 IF ALL NEURONS HAVE AT LEAST K IDENTICAL PARTNERS, AND 0 IF CLUSTERING IS ONLY AS STRONG AS EXPECTED BY CHANCE Quotation withheld because this record is not cleared for redistribution. and its limits stated in the same paragraph that introduces it: PAIRS IS NOT ONLY A TEST FOR CLUSTERING, BUT ALSO MORE GENERALLY FOR NON-UNIFORMITY of the distribution of tuning, and IT IS PROBABLY NOT POSSIBLE TO DETECT ALL CONCEIVABLE WAYS IN WHICH THE DATA MIGHT CLUSTER Quotation withheld because this record is not cleared for redistribution. A SIGNIFICANT P-VALUE READ AS SUPPORTING THE NULL: no evidence for clear categories in any animal, with p-values 0.236, 0.253, 0.399, 0.004 - and THE ONE SIGNIFICANT P-VALUE INDICATED LESS CLUSTERING THAN EXPECTED BY CHANCE Quotation withheld because this record is not cleared for redistribution. THE NULL VALIDATED BY MANUFACTURING WHAT WAS ABSENT: the lack of categories WAS NOT SIMPLY BECAUSE NEURAL VARIABILITY CAUSED OUR ANALYSIS TO MISS STRUCTURE - when SYNTHETIC CATEGORIES were introduced with noise derived from the real data, A STRIKINGLY DIFFERENT PAIRS DISTRIBUTION WAS EVIDENT Quotation withheld because this record is not cleared for redistribution. and the manufacture is exact rather than approximate - the faux partners were made by RESAMPLING TRIALS WITH REPLACEMENT from real neurons, so THESE FAUX NEURONS HAD EXACTLY THE SAME AMOUNT OF NOISE AS THEIR ORIGINATING NEURONS Quotation withheld because this record is not cleared for redistribution. with the exceptions counted rather than dismissed: relatively close neuron pairs were occasionally observed but DIFFERED FROM CHANCE IN ONLY 1 OF 5 CASES, and THE VAST MAJORITY OF NEURONS EACH REFLECT A UNIQUE COMBINATION OF RESPONSE FEATURES Quotation withheld because this record is not cleared for redistribution. THE SCOPE OF THE NULL, STATED BY THE AUTHORS: neural categories defined by OTHER PROPERTIES, SUCH AS CELL TYPE OR CONNECTIVITY, MIGHT REVEAL SPECIALIZATION - so `PPC is category-free` holds for categories defined by task-parameter selectivity and by time-dependent response features, and for nothing else Quotation withheld because this record is not cleared for redistribution. and the recommendation that follows is procedural, not factual: FUTURE STUDIES CAN TEST DIRECTLY FOR THE EXISTENCE OF CATEGORIES, AND DESIGN APPROPRIATE ANALYSES IF NEURONS ARE SHOWN TO REFLECT RANDOM COMBINATIONS of task parameters Quotation withheld because this record is not cleared for redistribution. THE SAMPLING CHOICE IS ITSELF THE ARGUMENT: WE RECORDED FROM ALL ENCOUNTERED NEURONS AND USED IDENTICAL STIMULI FOR EACH - the traditional approach of selection criteria and customized stimuli IS SUCCESSFUL IN IDENTIFYING NEURONS WITH STRONG CHOICE SIGNALS, BUT MAY LEAVE UNEXAMINED NEURONS WITH SUBTLER CHOICE SIGNALS THAT NONETHELESS SHAPE THE EVOLVING DECISION Quotation withheld because this record is not cleared for redistribution. stated again as a difference from the primate literature - the stimulus WAS NOT OPTIMIZED FOR EACH NEURON'S PREFERENCE and neurons WERE NOT SELECTED BASED ON TUNING PROPERTIES Quotation withheld because this record is not cleared for redistribution. MIXED SELECTIVITY MAKES TWO CONDITIONS INDISTINGUISHABLE IN ONE CELL: rare neurons had PURE choice or PURE modality selectivity, but for the majority THE MIXING SOMETIMES RESULTED IN IDENTICAL FIRING RATES FOR DIFFERENT CONDITIONS Quotation withheld because this record is not cleared for redistribution. and the absence of correlation is explicitly refused as proof: LACK OF CORRELATION DOES NOT CONCLUSIVELY RULE OUT THE EXISTENCE OF FUNCTIONAL CATEGORIES - selectivity for task features could still define categories, if the points formed an X or clusters symmetric about the origin Quotation withheld because this record is not cleared for redistribution. with every member of the population contributing: the neural weights WERE NONZERO FOR ALL NEURONS, and the distribution of weights DID NOT DIFFER FROM THOSE THAT WOULD BE EXPECTED BY CHANCE if neurons reflected random weightings Quotation withheld because this record is not cleared for redistribution. AN INACTIVATION WHOSE SPARED CONDITION IS THE CONTROL: impairment was SPECIFIC TO VISUAL TRIALS with NO CONSISTENT EFFECT ON AUDITORY DECISIONS, and THE SPARING OF AUDITORY DECISIONS REVEALS THAT INACTIVATION DID NOT REDUCE THE ANIMAL'S MOTIVATION, OR INTRODUCE CONFUSION about the stimulus-response contingency Quotation withheld because this record is not cleared for redistribution. and a confound answered by the wrong direction of its own effect: THE MORE POSTERIOR EXPRESSION WAS APPARENTLY NOT THE SOURCE OF THE IMPAIRMENT BECAUSE THE RAT WITH MORE EXPRESSION POSTERIOR TO PPC HAD WEAKER VISUAL IMPAIRMENT Quotation withheld because this record is not cleared for redistribution. with the surprise stated rather than smoothed: our inactivation results WERE SURPRISING IN THAT AUDITORY DECISIONS WERE SPARED DESPITE A CLEAR SIGNATURE OF THE DEVELOPING CHOICE IN PPC NEURONS - a signal that is present, reliable, and not necessary Quotation withheld because this record is not cleared for redistribution. THE NETWORK EXPLORES DIFFERENT DIMENSIONS AT DIFFERENT MOMENTS: dimensions that CAPTURED CONSIDERABLE VARIANCE DURING THE MOVEMENT CAPTURED FAR LESS VARIANCE DURING DECISION FORMATION, arguing AGAINST ALIGNED STATE SPACES Quotation withheld because this record is not cleared for redistribution. and the null is protected by a POSITIVE CONTROL - the same analysis on visual against multisensory trials revealed STRONG ALIGNMENT, which PROVIDES REASSURANCE THAT THE MISALIGNMENT of decision and movement INDICATES A REAL DIFFERENCE IN NEURAL COVARIANCE Quotation withheld because this record is not cleared for redistribution. and it is not two groups of cells taking turns: modulation during the stimulus was STRONGLY CORRELATED with modulation during movement, so THE MISALIGNMENT IS A POPULATION-LEVEL PHENOMENON, NOT DUE TO HAVING SEPARATE GROUPS OF NEURONS Quotation withheld because this record is not cleared for redistribution. visible in single cells as a reversal - one neuron has an ELEVATED response before a leftwards choice and is THEN SUPPRESSED DURING THE LEFTWARDS MOVEMENT used to report that same choice Quotation withheld because this record is not cleared for redistribution. A CONTROL ANIMAL BORROWED FROM ANOTHER STUDY, WITH ITS OWN NULL RECORDED: one animal had stimulating fibers implanted and laser stimulation on half its trials - THE ANIMAL EXPRESSED NO LIGHT-ACTIVATED ION CHANNELS IN ITS BRAIN, AND LASER STIMULATION HAD NO EFFECT ON NEURAL ACTIVITY OR BEHAVIOR Quotation withheld because this record is not cleared for redistribution. and the structure that survives the randomness is named TOLERANCE - although individual neurons reflected RANDOM COMBINATIONS of task components, THE OBSERVED RESPONSES WERE NONETHELESS STRUCTURED, retaining their choice preference across modalities Quotation withheld because this record is not cleared for redistribution. claim_type: null-result a second reading DIFFERING from the recorded value. Recorded: existence — 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: a statistic on the angles between neurons' response vectors and their nearest neighbours, compared against randomly oriented vectors - PAIRS, Projection Angle Index of Response Similarity 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: mixed selectivity - individual neurons modulated by more than one task parameter at once, so that different conditions can give identical firing rates a second reading DIFFERING from the recorded value. Recorded: population coding in posterior parietal cortex during multisensory decision-making — 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 → the widespread assumption of functionally specialised neural categories (refs 1, 5-11, 32) Quotation withheld because this record is not cleared for redistribution. extends → prior work testing for categories defined by response timing (ref 19) Quotation withheld because this record is not cleared for redistribution. builds-on → theoretical work on random mixed representations (refs 12-14) Quotation withheld because this record is not cleared for redistribution. builds-on → the established rat multisensory rate-discrimination task (refs 20, 21) Quotation withheld because this record is not cleared for redistribution. critiques → the assumption of neural categories, which is built into how neurons are sampled and into how populations are summarized differs in: pie charts and averaging, which assign neurons to categories and treat each neuron as an exemplar differing from others mainly because of noise Quotation withheld because this record is not cleared for redistribution. extends → earlier work that argued against categories but tested only for categories defined by response sequence Quotation withheld because this record is not cleared for redistribution. supports → theoretical work predicting that randomly distributed parameters let an arbitrary group of neurons be linearly combined for whatever estimate is needed Quotation withheld because this record is not cleared for redistribution. distinguishes-from → the primate approach of selecting neurons by tuning and optimising the stimulus per neuron, which finds strong choice signals and leaves subtler ones unexamined differs in: THE AXIS IS HOW THE NEURONS AND STIMULI WERE CHOSEN. 'Quotation withheld.' The older method is granted its successes and denied its denominator - it finds what it selects for, so a literature built on it cannot report what fraction of the population is involved. Quotation withheld because this record is not cleared for redistribution. qualifies → its own null, which covers categories defined by task-parameter selectivity and temporal response features and not those defined by cell type or connectivity differs in: THE AXIS IS WHICH PROPERTY DEFINES THE CATEGORY. 'Quotation withheld.' The paper's headline null - no categories - is bounded by the definition of category it used, IN THE SENTENCE AFTER THE CLAIM. Its [4] criticises pie charts and averaging for assigning neurons to categories; here it says which categories its own null covers. Quotation withheld because this record is not cleared for redistribution. supports → an analogy with inferotemporal cortex, where tolerance is read as evidence that representations have been untangled Quotation withheld because this record is not cleared for redistribution. predicts → that future studies test directly for the existence of categories and design their analyses accordingly Quotation withheld because this record is not cleared for redistribution. clustering —contrasted-with→ PAIRS clustering PAIRS strength of the relation: 1 · note: its limits stated where it is introduced - NOT ONLY A TEST FOR CLUSTERING BUT MORE GENERALLY FOR NON-UNIFORMITY, and IT IS PROBABLY NOT POSSIBLE TO DETECT ALL CONCEIVABLE WAYS IN WHICH THE DATA MIGHT CLUSTER Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says PAIRS also tests non-uniformity beyond clustering; contrasted-with is loose. Quotation withheld because this record is not cleared for redistribution. choice —distinct-from→ modality choice modality strength of the relation: 1 · note: MIXED SELECTIVITY MAKES TWO CONDITIONS INDISTINGUISHABLE IN ONE CELL - the mixing SOMETIMES RESULTED IN IDENTICAL FIRING RATES FOR DIFFERENT CONDITIONS Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says pure choice or modality selectivity is rare and most neurons mix both; distinct-from is loose. Quotation withheld because this record is not cleared for redistribution. Mann-Whitney U test Mann-Whitney U test Quotation withheld because this record is not cleared for redistribution. bootstrap bootstrap Quotation withheld because this record is not cleared for redistribution. Monte Carlo simulation Monte Carlo simulation Quotation withheld because this record is not cleared for redistribution. ROC analysis / area under the curve ROC analysis / area under the curve Quotation withheld because this record is not cleared for redistribution. principal component analysis principal component analysis Quotation withheld because this record is not cleared for redistribution. support vector machine classifier support vector machine classifier Quotation withheld because this record is not cleared for redistribution. generalized linear model regression generalized linear model regression Quotation withheld because this record is not cleared for redistribution. sign test sign test Quotation withheld because this record is not cleared for redistribution. kurtosis kurtosis Quotation withheld because this record is not cleared for redistribution. maximum likelihood estimation (psychometric function fitting) maximum likelihood estimation (psychometric function fitting) Quotation withheld because this record is not cleared for redistribution. confidence interval confidence interval Quotation withheld because this record is not cleared for redistribution. bootstrap aggregation (bagging) bootstrap aggregation (bagging) Quotation withheld because this record is not cleared for redistribution. PAIRS PAIRS Quotation withheld because this record is not cleared for redistribution. SVM SVM Quotation withheld because this record is not cleared for redistribution. p-values from Monte Carlo simulations p-values from Monte Carlo simulations Quotation withheld because this record is not cleared for redistribution. PPC INACTIVATION RELIABLY IMPAIRED VISUAL DECISIONS, by Mann-Whitney U test at p < 0.001 pooled across animals Quotation withheld because this record is not cleared for redistribution. Quotation withheld Quotation withheld because this record is not cleared for redistribution. Most PPC neurons carried joint information about both stimulus modality and the animal's choice rather than either one alone. Quotation withheld because this record is not cleared for redistribution. The nearest-neighbor angle distribution of neurons' feature vectors was statistically indistinguishable from that of randomly oriented vectors. Quotation withheld because this record is not cleared for redistribution. A linear decoder read the animal's choice from single-trial population responses significantly above chance in every rat. Quotation withheld because this record is not cleared for redistribution. The same choice decoder did not estimate stimulus modality better than chance. Quotation withheld because this record is not cleared for redistribution. Quotation withheld Quotation withheld because this record is not cleared for redistribution. The choice and modality decoder vectors were nearly orthogonal, meaning the two parameters occupy essentially independent dimensions of population activity. Quotation withheld because this record is not cleared for redistribution. Dimensions that captured most variance during movement captured little variance during decision formation, so the network occupies different covariance patterns in the two epochs. Quotation withheld because this record is not cleared for redistribution. this analysis, BY CONTRAST, REVEALED STRONG ALIGNMENT - index 0.840 at p < 10 to the minus 4, PRESENT IN ALL RATS TESTED Quotation withheld because this record is not cleared for redistribution. Quotation withheld Quotation withheld because this record is not cleared for redistribution. 80.1% of neurons, 218 of 272 units, HAD A MULTISENSORY RESPONSE THAT WAS BETTER PREDICTED BY THE AUDITORY AND VISUAL RESPONSES THAN BY THE MULTISENSORY MEAN, assessed on left-out data Quotation withheld because this record is not cleared for redistribution. THE ASSUMPTION OF NEURAL CATEGORIES IS REFLECTED IN MANY EXPERIMENTAL DESIGNS AND ANALYSIS METHODS, even those focusing on structures far downstream of early sensory areas Quotation withheld because this record is not cleared for redistribution. neurons MUST MEET CERTAIN RESPONSE CRITERIA TO BE INCLUDED FOR STUDY, and IMPLICIT IN THIS APPROACH IS THE IDEA THAT THE CELL'S RESPONSE DURING ONE STIMULUS IDENTIFIES IT AS A MEMBER OF THE CATEGORY BEING EXAMINED Quotation withheld because this record is not cleared for redistribution. PIE CHARTS, A COMMON WAY OF SUMMARIZING POPULATION DATA, EXPLICITLY ASSIGN NEURONS TO CATEGORIES, and AVERAGING OVER MANY NEURONS LIKEWISE REFLECTS THE ASSUMPTION THAT EACH NEURON IS AN EXEMPLAR OF A CATEGORY Quotation withheld because this record is not cleared for redistribution. earlier work ARGUED AGAINST CATEGORIES, BUT ONLY TESTED FOR CATEGORIES DEFINED BY RESPONSE SEQUENCE. A MORE GENERAL TEST IS THUS REQUIRED Quotation withheld because this record is not cleared for redistribution. PAIRS is 1 IF ALL NEURONS HAVE AT LEAST K IDENTICAL PARTNERS, AND 0 IF CLUSTERING IS ONLY AS STRONG AS EXPECTED BY CHANCE Quotation withheld because this record is not cleared for redistribution. PAIRS IS NOT ONLY A TEST FOR CLUSTERING, BUT ALSO MORE GENERALLY FOR NON-UNIFORMITY of the distribution of tuning, and IT IS PROBABLY NOT POSSIBLE TO DETECT ALL CONCEIVABLE WAYS IN WHICH THE DATA MIGHT CLUSTER Quotation withheld because this record is not cleared for redistribution. no evidence for clear categories was present in any animal, and THE ONE SIGNIFICANT P-VALUE INDICATED LESS CLUSTERING THAN EXPECTED BY CHANCE Quotation withheld because this record is not cleared for redistribution. the lack of categories WAS NOT SIMPLY BECAUSE NEURAL VARIABILITY CAUSED OUR ANALYSIS TO MISS STRUCTURE - with SYNTHETIC CATEGORIES introduced, A STRIKINGLY DIFFERENT PAIRS DISTRIBUTION WAS EVIDENT Quotation withheld because this record is not cleared for redistribution. THESE FAUX NEURONS HAD EXACTLY THE SAME AMOUNT OF NOISE AS THEIR ORIGINATING NEURONS Quotation withheld because this record is not cleared for redistribution. close neuron pairs DIFFERED FROM CHANCE IN ONLY 1 OF 5 CASES, and THE VAST MAJORITY OF NEURONS EACH REFLECT A UNIQUE COMBINATION OF RESPONSE FEATURES Quotation withheld because this record is not cleared for redistribution. neural categories defined by OTHER PROPERTIES, SUCH AS CELL TYPE OR CONNECTIVITY, MIGHT REVEAL SPECIALIZATION Quotation withheld because this record is not cleared for redistribution. WE RECORDED FROM ALL ENCOUNTERED NEURONS AND USED IDENTICAL STIMULI FOR EACH, where the traditional approach MAY LEAVE UNEXAMINED NEURONS WITH SUBTLER CHOICE SIGNALS THAT NONETHELESS SHAPE THE EVOLVING DECISION Quotation withheld because this record is not cleared for redistribution. the mixing SOMETIMES RESULTED IN IDENTICAL FIRING RATES FOR DIFFERENT CONDITIONS Quotation withheld because this record is not cleared for redistribution. LACK OF CORRELATION DOES NOT CONCLUSIVELY RULE OUT THE EXISTENCE OF FUNCTIONAL CATEGORIES Quotation withheld because this record is not cleared for redistribution. the neural weights were NONZERO FOR ALL NEURONS and did not differ from those EXPECTED BY CHANCE if neurons reflected random weightings Quotation withheld because this record is not cleared for redistribution. THE SPARING OF AUDITORY DECISIONS REVEALS THAT INACTIVATION DID NOT REDUCE THE ANIMAL'S MOTIVATION, OR INTRODUCE CONFUSION about the stimulus-response contingency Quotation withheld because this record is not cleared for redistribution. THE RAT WITH MORE EXPRESSION POSTERIOR TO PPC HAD WEAKER VISUAL IMPAIRMENT compared to the other rat Quotation withheld because this record is not cleared for redistribution. dimensions that CAPTURED CONSIDERABLE VARIANCE DURING THE MOVEMENT CAPTURED FAR LESS VARIANCE DURING DECISION FORMATION Quotation withheld because this record is not cleared for redistribution. the positive control revealed STRONG ALIGNMENT, which PROVIDES REASSURANCE THAT THE MISALIGNMENT of decision and movement INDICATES A REAL DIFFERENCE IN NEURAL COVARIANCE Quotation withheld because this record is not cleared for redistribution. THE MISALIGNMENT IS A POPULATION-LEVEL PHENOMENON, NOT DUE TO HAVING SEPARATE GROUPS OF NEURONS Quotation withheld because this record is not cleared for redistribution. THE ANIMAL EXPRESSED NO LIGHT-ACTIVATED ION CHANNELS IN ITS BRAIN, AND LASER STIMULATION HAD NO EFFECT ON NEURAL ACTIVITY OR BEHAVIOR Quotation withheld because this record is not cleared for redistribution. Projection Angle Index of Response Similarity Projection Angle Index of Response Similarity Quotation withheld because this record is not cleared for redistribution. Variance Alignment Variance Alignment Quotation withheld because this record is not cleared for redistribution. choice divergence choice divergence Quotation withheld because this record is not cleared for redistribution. choice preference choice preference Quotation withheld because this record is not cleared for redistribution. excess rate excess rate Quotation withheld because this record is not cleared for redistribution. mixed selectivity mixed selectivity Quotation withheld because this record is not cleared for redistribution. decision formation decision formation Quotation withheld because this record is not cleared for redistribution. DREADD (Designer Receptor Exclusively Activated by Designer Drug) DREADD (Designer Receptor Exclusively Activated by Designer Drug) Quotation withheld because this record is not cleared for redistribution. impairment ratio impairment ratio Quotation withheld because this record is not cleared for redistribution. PAIRS PAIRS Quotation withheld because this record is not cleared for redistribution. clustering clustering Quotation withheld because this record is not cleared for redistribution. tolerance tolerance Quotation withheld because this record is not cleared for redistribution. categories categories Quotation withheld because this record is not cleared for redistribution. PPC inactivation PPC inactivation Quotation withheld because this record is not cleared for redistribution. control sessions control sessions Quotation withheld because this record is not cleared for redistribution. selectivity selectivity Quotation withheld because this record is not cleared for redistribution. weights weights Quotation withheld because this record is not cleared for redistribution. variance variance Quotation withheld because this record is not cleared for redistribution. 80.1% (218 of 272 units) — neurons whose multisensory response was better predicted by a linear combination of auditory and visual responses than by the multisensory mean Quotation withheld because this record is not cleared for redistribution. 68.2% — median share of multisensory response variance accounted for by a linear combination of the unisensory responses Quotation withheld because this record is not cleared for redistribution. 33.8% — proportion of neurons with statistically significant modality preference Quotation withheld because this record is not cleared for redistribution. 86.5° — average angle between the choice decoder and the modality decoder, near the orthogonal 90° Quotation withheld because this record is not cleared for redistribution. 68.9%, 61.2%, 59.4%, 80.2%, 70.1% — choice-decoding accuracy in the five rats at 700-800 ms after stimulus onset, outside the training window Quotation withheld because this record is not cleared for redistribution. PAIRS index = −0.052; p = 0.632 — clustering statistic for the example rat's feature vectors, indistinguishable from chance (no categories) Quotation withheld because this record is not cleared for redistribution. index = 0.840, p < 10 −4 — Variance Alignment index for visual versus multisensory decision-epoch activity, the aligned positive control Quotation withheld because this record is not cleared for redistribution. r = 0.709, r = 0.655, r = 0.577, r = 0.666, r = 0.763 — per-rat correlation between each neuron's stimulus-epoch and movement-epoch modulation depth Quotation withheld because this record is not cleared for redistribution. 86.5° — the average angle between the choice decoder and the modality decoder, ONLY SLIGHTLY LESS THAN A PERFECTLY ORTHOGONAL 90 Quotation withheld because this record is not cleared for redistribution. 0.004 — the one significant PAIRS p-value, which INDICATED LESS CLUSTERING THAN EXPECTED BY CHANCE Quotation withheld because this record is not cleared for redistribution. 1 of 5 cases — the number of animals in which close neuron pairs differed from chance at all Quotation withheld because this record is not cleared for redistribution. 3.3 — the signal-to-noise criterion a neuron had to exceed to be included - firing-rate range divided by maximal s.e.m. 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/nn.3865.