Distinct learning-induced changes in stimulus selectivity and interactions of GABAergic interneuron classes in visual cortex

Distinct learning-induced changes in stimulus selectivity and interactions of GABAergic interneuron classes in visual cortex Khan Poort Chadwick Blot Sahani Mrsic-Flogel Hofer

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.

Distinct learning-induced changes in stimulus selectivity and interactions of GABAergic interneuron classes in visual cortex 2018 Nature Neuroscience 10.1038/s41593-018-0143-z Khan2018-distinct-learning-induced-changes-in-stimulu Khan2018-distinct-learning-induced-changes-in-stimulu.txt, the project's staged copy 1beb26c6fe8d672b4e9c47915efe7865daee11f2 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 how learning changes the stimulus selectivity and the mutual interactions of pyramidal cells and three classes of GABAergic interneuron at once Quotation withheld because this record is not cleared for redistribution. whether an interneuron class responds and changes as one functional unit, and whether learning can make a specific class selective for the behaviorally relevant stimulus Quotation withheld because this record is not cleared for redistribution. PV cells become as selective as pyramidal cells and less tied to the local network, PYR-PV subnetworks emerge, and SOM cells decouple and appear to gate how much any pyramidal cell can change Quotation withheld because this record is not cleared for redistribution. mechanism Quotation withheld because this record is not cleared for redistribution. learning-induced stimulus selectivity in visual cortical interneurons Quotation withheld because this record is not cleared for redistribution. calcium-imaging calcium-imaging Quotation withheld because this record is not cleared for redistribution. mechanisms SOM interneurons gate pyramidal-cell plasticity: dendritic inhibition arriving together with bottom-up drive blocks the pairing of top-down and bottom-up inputs, so strongly SOM-coupled pyramidal cells cannot increase selectivity Quotation withheld because this record is not cleared for redistribution. mechanisms Concurrent SOM-mediated apical dendritic inhibition prevents the association of visual and top-down inputs in pyramidal cells Quotation withheld because this record is not cleared for redistribution. mechanisms Plasticity of excitatory connections onto PV cells makes them integrate local pyramidal input selectively, forming stimulus-specific PYR-PV subnetworks Quotation withheld because this record is not cleared for redistribution. PYR to PV interaction weights become specific for stimulus preference, so PV cells integrate selectively from the pyramidal cells that share their preference rather than from the neighbourhood at large Quotation withheld because this record is not cleared for redistribution. and the increased PV selectivity then feeds back - deleting PV to PYR weights reduces pyramidal selectivity more after learning than before Quotation withheld because this record is not cleared for redistribution. SOM cells inhibit apical dendrites, so a pyramidal cell whose activity is correlated with the SOM population receives dendritic inhibition timed with its bottom-up drive and cannot pair it with top-down input Quotation withheld because this record is not cleared for redistribution. Quotation withheld, and the activity of SOM cells became strongly decorrelated, so they seemed to become uncoupled from the local network Quotation withheld because this record is not cleared for redistribution. The proposed local versus long-range connectivity account of the PV and SOM changes is a hypothesis the study did not test directly. what this limits: the connectivity account of the PV and SOM changes: THESE HYPOTHESES COULD BE TESTED DIRECTLY BY ASSESSING THE LOCAL AND LONG-RANGE CONNECTIVITY ... IN NAIVE AND EXPERT MICE - named as future work Quotation withheld because this record is not cleared for redistribution. The study cannot say how much of the observed single-cell heterogeneity is explained by cell-type diversity within an interneuron class, or how it varies by brain state and region. what this limits: any reading of the single-cell variability: THE CHALLENGE FOR FUTURE STUDIES IS TO ESTABLISH THE EXTENT TO WHICH THE HETEROGENEOUS BEHAVIOR OF SINGLE CELLS IS CAPTURED BY THE DIVERSITY OF CELL TYPES WITHIN AN INTERNEURON CLASS Quotation withheld because this record is not cleared for redistribution. the two accounts of PV selectivity - external drive against selective local input - are separated by a prediction about co-fluctuations rather than by a manipulation what this limits: the mechanism behind the PV selectivity change: PV CELLS MIGHT BECOME MORE STRONGLY INFLUENCED BY EXTERNAL ... INPUTS ... ALTERNATIVELY, INDIVIDUAL PV CELLS MIGHT ACQUIRE MORE SELECTIVE INPUT FROM SUBSETS OF PYR CELLS - two accounts, no measurement between them Quotation withheld because this record is not cleared for redistribution. and the connectivity the model infers has not been measured; the direct test is named as future work duplicate of: limitations[0] · what this limits: the same scope as limitations[0] - the inferred connectivity has not been measured Quotation withheld because this record is not cleared for redistribution. the interneuron class itself is not a reliable label - transgenic and immunohistochemical identification of SOM cells agree only 70% of the time what this limits: every result reported by interneuron class, at a rate that differs by class: transgenic and immunohistochemical identification agree PV 92.6%, SOM 70.1%, VIP 85.4% Quotation withheld because this record is not cleared for redistribution. and the staged copy is an author manuscript in which three reported statistics are still unfilled placeholders what this limits: the model-fit statistics as printed: the staged copy carries unfilled placeholders where three reported values should be Quotation withheld because this record is not cleared for redistribution. the linear dynamical system, used to turn simultaneous recordings into a weight matrix between identified cells Quotation withheld because this record is not cleared for redistribution. d-prime style normalisation - the selectivity index is a response difference divided by the pooled standard deviation Quotation withheld because this record is not cleared for redistribution. post-hoc staining registered to the in vivo imaging plane, so that functional records acquire a molecular label Quotation withheld because this record is not cleared for redistribution. loose-patch recording with simultaneous imaging, imported purely to calibrate the indicator against spikes Quotation withheld because this record is not cleared for redistribution. the behavioural covariates - eye position, pupil size, eye movements, running and licking - were measured and tested against the learning effect on response selectivity eliminates: that the increase in response selectivity after learning is an artefact of the animal's eye position, arousal or movement rather than a change in the cortical representation · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. the same behavioural covariates tested a second time, against the learning effect on NOISE CORRELATIONS rather than on selectivity eliminates: that the change in noise correlations with learning is an artefact of eye position, arousal or movement · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. selectivity selectivity Quotation withheld because this record is not cleared for redistribution. learning learning Quotation withheld because this record is not cleared for redistribution. PV PV Quotation withheld because this record is not cleared for redistribution. noise correlation noise correlation Quotation withheld because this record is not cleared for redistribution. recurrent recurrent Quotation withheld because this record is not cleared for redistribution. model model Quotation withheld because this record is not cleared for redistribution. THE CATEGORY IS ASSUMED TO BE A UNIT: each interneuron class has been suggested to act as a functionally AND THUS COMPUTATIONALLY HOMOGENEOUS unit - and whether learning changes them homogeneously is not clear Quotation withheld because this record is not cleared for redistribution. AND IT IS NOT: responses within each class are HIGHLY HETEROGENEOUS, including increases and decreases in the same class to the same stimulus - broadly different between classes, and heterogeneous within each Quotation withheld because this record is not cleared for redistribution. THE CATEGORY DOES NOT EVEN BOUND THE CORRELATIONS: activity correlations ACROSS classes can be as high as those WITHIN a class, so functional ensembles SPAN multiple cell types Quotation withheld because this record is not cleared for redistribution. AND THE METHOD BUILT ON THE CATEGORY IS DECLARED INAPPROPRIATE: optogenetic manipulation of ENTIRE CELL CLASSES may not be the right approach, and cellular-resolution targeting may be required instead Quotation withheld because this record is not cleared for redistribution. TWO DEFINITIONS OF THE SAME CLASS, AGREEING 70% OF THE TIME: of transgenically labelled cells also confirmed by immunostaining, PV 92.6%, VIP 85.4% and SOM only 70.1% - with the note that this matters for studies that use mouse lines to categorise cells Quotation withheld because this record is not cleared for redistribution. and measured the other way round the same labels agree above 90%, so the disagreement is DIRECTIONAL - which label is the definition decides the population Quotation withheld because this record is not cleared for redistribution. A NUMBER THAT WAS NEVER PRINTED: the manuscript reports the model's fit as R 2 = xx twice, with only the middle value filled in - the claim is made and the evidence for it is a placeholder Quotation withheld because this record is not cleared for redistribution. and again in the constrained fit - the effect is said to be stronger, with P = XX Quotation withheld because this record is not cleared for redistribution. THE MODEL IS BLIND TO CELL TYPE AND THE CLASSES FALL OUT ANYWAY: interaction weights and their changes over learning differ by cell class although the model never saw the labels Quotation withheld because this record is not cleared for redistribution. TWO RIVAL ACCOUNTS STATED WITH THEIR OPPOSITE PREDICTIONS BEFORE THE MEASUREMENT: external drive would make co-fluctuations DECREASE, selective local input would not - and only the first is consistent with what was found Quotation withheld because this record is not cleared for redistribution. AN INSTRUMENT ARTEFACT RULED OUT BY A SECOND INSTRUMENT: loose-patch spikes recorded with simultaneous imaging show a near-linear rate-to-fluorescence relation, and selectivity computed from inferred firing rate matches selectivity from fluorescence Quotation withheld because this record is not cleared for redistribution. with behavior excluded too - eye position, pupil size, eye movements, running and licking cannot account for it, and the change is specific to the task-relevant stimulus and absent in the approach corridor Quotation withheld because this record is not cleared for redistribution. THE PREDICTOR IS MEASURED BEFORE THE THING IT PREDICTS: coupling to the SOM population in the NAIVE animal predicts how much a pyramidal cell will gain in selectivity, and coupling to PV, VIP or PYR populations does not Quotation withheld because this record is not cleared for redistribution. and the proposed mechanism is a gate - high correlation with SOM cells PREVENTS learning-related plasticity, low correlation is permissive Quotation withheld because this record is not cleared for redistribution. A DEBATE SETTLED BY THE FIT: the origins of noise correlations have been debated, and here they are largely accounted for by concurrent functional interactions with simple linear dynamics Quotation withheld because this record is not cleared for redistribution. AND THE GAP IS NAMED: no study yet provides a model-based fit of concurrent activity in multiple identified cell classes that accounts for the influence of the local population on each cell Quotation withheld because this record is not cleared for redistribution. claim_type: mechanism a second reading reaching the same value. Recorded: mechanism — note: a second reading reaching the same value from a different sentence. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. method: a linear dynamical system fit to single-trial activity of all simultaneously imaged cells, with interaction weights deleted or held fixed across learning to ask what each contributes a second reading DIFFERING from the recorded value. Recorded: calcium-imaging — 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: learning-induced sharpening of cortical representations, decomposed into which cell class changes and which interactions carry the change a second reading DIFFERING from the recorded value. Recorded: learning-induced stimulus selectivity in visual cortical interneurons — 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. builds-on → Poort et al., 2015 Quotation withheld because this record is not cleared for redistribution. extends → Kerlin et al., 2010; Bock et al., 2011; Hofer et al., 2011; Scholl et al., 2015 (view that PV cells broadly integrate local pyramidal activity) Quotation withheld because this record is not cleared for redistribution. contradicts → the treatment of a molecularly defined interneuron class as a functionally homogeneous unit Quotation withheld because this record is not cleared for redistribution. contradicts → the view that PV cells necessarily pool broadly and therefore stay untuned Quotation withheld because this record is not cleared for redistribution. supports → the account that noise correlations arise from concurrent functional interactions rather than from shared external noise Quotation withheld because this record is not cleared for redistribution. critiques → optogenetic manipulation of whole cell classes as a way to study circuit mechanisms Quotation withheld because this record is not cleared for redistribution. distinguishes-from → the alternative that PV cells simply acquire selective local input without decoupling, which predicts no drop in co-fluctuations differs in: THE AXIS IS A PREDICTION THAT SEPARATES TWO MECHANISMS, AND ONLY ONE MAKES IT. 'Quotation withheld.' Two accounts of the same learning effect - PV cells driven more by external input, or acquiring more selective local input - are indistinguishable in the selectivity data and separated by ONE measurable consequence. I declined to build this row earlier because I could not find the co-fluctuation clause in the span head; it is in the sentence that carries the marker. Quotation withheld because this record is not cleared for redistribution. predicts → that local and long-range connectivity of PYR, PV and SOM cells will differ between naive and expert mice Quotation withheld because this record is not cleared for redistribution. learning —causes→ selectivity learning selectivity holds when: in PYR, PV and SOM cells; the authors state VIP cells are the exception · direction on the measure: increases · strength of the relation: 1 · note: increased in subsets of PYR, PV and SOM cells BUT NOT VIP - the effect is per-class and per-subset, not general Quotation withheld because this record is not cleared for redistribution. PYR —part-of→ subnetwork PYR subnetwork strength of the relation: 1 · note: PYR to PV interactions reorganized so that stimulus selective PYR-PV subnetworks EMERGED during learning Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says stimulus-selective PYR-PV subnetworks emerged; part-of is loose. Quotation withheld because this record is not cleared for redistribution. SOM —contrasted-with→ VIP SOM VIP strength of the relation: 1 · note: the SOM mean does not move but a SUBPOPULATION does - the class average hides the effect Quotation withheld because this record is not cleared for redistribution. PYR —causes→ PV PYR PV strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but PV cells are 'THOUGHT TO broadly integrate' PYR activity, and their selectivity correlated with nearby PYR cells; causes is hedged. Quotation withheld because this record is not cleared for redistribution. external input —causes→ PV external input PV strength of the relation: 0.9 · note: TWO RIVAL ACCOUNTS WITH OPPOSITE PREDICTIONS STATED BEFORE THE MEASUREMENT: decoupling predicts co-fluctuations DECREASE, selective local input predicts they do not Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but PV cells 'MIGHT become' more influenced by external inputs - one of two hypotheses; causes is hedged. Quotation withheld because this record is not cleared for redistribution. learning —causes→ noise correlation learning noise correlation holds when: in every cell-class pairing EXCEPT SOM-SOM, which the authors report as not significant · direction on the measure: reduces · strength of the relation: 1 · note: decreased within AND between classes, with the single exception of SOM-SOM pairs Quotation withheld because this record is not cleared for redistribution. interaction weight —causes→ noise correlation interaction weight noise correlation strength of the relation: 1 · note: A NUMBER THAT WAS NEVER PRINTED - the fit is reported as R 2 = xx twice over, with only the middle value (-0.18) filled in; the claim is made and its evidence is a placeholder Quotation withheld because this record is not cleared for redistribution. PV —causes→ selectivity PV selectivity direction on the measure: increases · strength of the relation: 1 · note: and it feeds back - deleting PV to PYR weights costs PYR selectivity more after learning, where deleting SOM and PYR weights does not change Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span says PV selectivity boosts PYR selectivity; the endpoint 'selectivity' drops whose. Quotation withheld because this record is not cleared for redistribution. SOM —causes→ plasticity SOM plasticity strength of the relation: 0.9 · note: high correlation with the SOM population PREVENTS learning-related plasticity, low correlation is permissive - dendritic inhibition arriving with the bottom-up drive blocks the pairing Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span suggests SOM cells gate plasticity; causes is loose and the claim is a suggestion. Quotation withheld because this record is not cleared for redistribution. sign test sign test Quotation withheld because this record is not cleared for redistribution. Wilcoxon rank-sum test Wilcoxon rank-sum test 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. correlation correlation Quotation withheld because this record is not cleared for redistribution. t-test t-test Quotation withheld because this record is not cleared for redistribution. bootstrap bootstrap Quotation withheld because this record is not cleared for redistribution. confidence interval confidence interval Quotation withheld because this record is not cleared for redistribution. classifier / decoding accuracy classifier / decoding accuracy Quotation withheld because this record is not cleared for redistribution. regression regression Quotation withheld because this record is not cleared for redistribution. coefficient of determination coefficient of determination Quotation withheld because this record is not cleared for redistribution. sign test sign test Quotation withheld because this record is not cleared for redistribution. Wilcoxon rank-sum test Wilcoxon rank-sum test Quotation withheld because this record is not cleared for redistribution. bootstrap test bootstrap test Quotation withheld because this record is not cleared for redistribution. Learning increased stimulus selectivity in subsets of pyramidal, PV and SOM cells, but left VIP cells unchanged. Quotation withheld because this record is not cleared for redistribution. PV neurons became as selective as PYR cells and less correlated with stimulus preference of the local network Quotation withheld because this record is not cleared for redistribution. Fitting a linear dynamical system to the population activity showed that pyramidal-to-PV interactions reorganized into stimulus-selective PYR-PV subnetworks over learning. Quotation withheld because this record is not cleared for redistribution. After learning, model interaction weights from PYR to PV and between PV cells were stronger for cell pairs sharing a grating preference than for pairs with opposite preference. Quotation withheld because this record is not cleared for redistribution. Learning reduced noise correlations both within and across cell classes, so network firing became more independent. Quotation withheld because this record is not cleared for redistribution. How strongly a pyramidal cell was correlated with the SOM population before learning predicted how much its selectivity would increase. Quotation withheld because this record is not cleared for redistribution. learning increased selectivity in subsets of PYR, PV and SOM cells but not VIP, and PV neurons became AS SELECTIVE AS pyramidal cells while becoming less correlated with the local network Quotation withheld because this record is not cleared for redistribution. PYR to PV interactions reorganized so that stimulus-selective PYR-PV subnetworks emerged, while SOM cells decorrelated from the local network and may gate selectivity changes Quotation withheld because this record is not cleared for redistribution. THE CATEGORY IS ASSUMED TO BE A UNIT: each interneuron class has been suggested to act as a functionally and thus computationally HOMOGENEOUS unit Quotation withheld because this record is not cleared for redistribution. AND IT IS NOT: responses within each class are highly heterogeneous, including increases and decreases to the same stimulus - different between classes, and heterogeneous within each Quotation withheld because this record is not cleared for redistribution. THE CATEGORY DOES NOT BOUND THE CORRELATIONS EITHER: correlations across classes can be as high as those within a class, so functional ensembles SPAN cell types Quotation withheld because this record is not cleared for redistribution. AND THE METHOD BUILT ON THE CATEGORY IS DECLARED INAPPROPRIATE: optogenetic manipulation of entire cell classes may not be the right approach for studying circuit mechanisms Quotation withheld because this record is not cleared for redistribution. TWO DEFINITIONS OF THE SAME CLASS AGREEING 70% OF THE TIME: transgenically labelled cells confirmed by immunostaining were PV 92.6%, VIP 85.4% and SOM only 70.1% Quotation withheld because this record is not cleared for redistribution. and measured the other way round the same labels agree above 90%, so the disagreement is directional Quotation withheld because this record is not cleared for redistribution. A NUMBER THAT WAS NEVER PRINTED: the model fit is reported as R 2 = xx twice, with only the middle value filled in Quotation withheld because this record is not cleared for redistribution. and again in the constrained fit, where the effect is said to be stronger with P = XX Quotation withheld because this record is not cleared for redistribution. THE MODEL IS BLIND TO CELL TYPE and cell-class specific differences in interaction weights appear anyway Quotation withheld because this record is not cleared for redistribution. TWO RIVAL ACCOUNTS WITH OPPOSITE PREDICTIONS STATED BEFORE THE MEASUREMENT: decoupling predicts co-fluctuations decrease, selective local input predicts they do not Quotation withheld because this record is not cleared for redistribution. PV selectivity rose although these cells were expected to pool broadly and inhibit in proportion to average local activity, and the slope on local PYR selectivity fell from 0.2 to 0.05 Quotation withheld because this record is not cleared for redistribution. AN INSTRUMENT ARTEFACT RULED OUT BY A SECOND INSTRUMENT: loose-patch spikes with simultaneous imaging give a near-linear rate-to-fluorescence relation, and selectivity from inferred rate matches selectivity from fluorescence Quotation withheld because this record is not cleared for redistribution. with eye position, pupil size, eye movements, running and licking excluded, and the change absent in the approach corridor Quotation withheld because this record is not cleared for redistribution. learning decreased noise correlations within and between classes - SOM cells decorrelating most, PV cells least Quotation withheld because this record is not cleared for redistribution. the LDS model splits each cell's activity into recurrent input from all other recorded cells, a trial-invariant stimulus-locked input, running speed, and a residual Quotation withheld because this record is not cleared for redistribution. post learning, PYR to PV and PV to PV weights were higher between cells preferring the SAME grating, and deleting weights cost PV selectivity only after learning Quotation withheld because this record is not cleared for redistribution. PV selectivity changes were disrupted only when stimulus drive AND PYR to PV weights were BOTH held fixed across learning Quotation withheld because this record is not cleared for redistribution. and it feeds back - deleting PV to PYR weights cost pyramidal selectivity more after learning, where SOM and PYR weight deletions did not change Quotation withheld because this record is not cleared for redistribution. THE PREDICTOR IS MEASURED BEFORE THE THING IT PREDICTS: coupling to the SOM population in the naive animal predicts the selectivity gain, and coupling to PV, VIP or PYR populations does not Quotation withheld because this record is not cleared for redistribution. with the proposed mechanism a gate - high correlation with SOM cells prevents plasticity, low correlation permits it Quotation withheld because this record is not cleared for redistribution. the authors suggest noise correlations can largely be accounted for by concurrent functional interactions with simple linear dynamics Quotation withheld because this record is not cleared for redistribution. AND THE GAP IS NAMED: no study yet fits concurrent activity in multiple identified cell classes while accounting for the local population's influence on each cell Quotation withheld because this record is not cleared for redistribution. the direct test is named as future work - local and long-range connectivity of PYR, PV and SOM cells in naive against expert mice Quotation withheld because this record is not cleared for redistribution. selectivity index (SI) selectivity index (SI) Quotation withheld because this record is not cleared for redistribution. noise correlations noise correlations Quotation withheld because this record is not cleared for redistribution. behavioral d-prime behavioral d-prime Quotation withheld because this record is not cleared for redistribution. PSTH similarity score PSTH similarity score Quotation withheld because this record is not cleared for redistribution. linear dynamical system (LDS) model linear dynamical system (LDS) model Quotation withheld because this record is not cleared for redistribution. selectivity selectivity Quotation withheld because this record is not cleared for redistribution. noise correlation noise correlation Quotation withheld because this record is not cleared for redistribution. interaction weight interaction weight Quotation withheld because this record is not cleared for redistribution. gate gate Quotation withheld because this record is not cleared for redistribution. cell class cell class Quotation withheld because this record is not cleared for redistribution. selectivity index selectivity index Quotation withheld because this record is not cleared for redistribution. running speed running speed Quotation withheld because this record is not cleared for redistribution. eye position, pupil size, eye movements, running and licking eye position, pupil size, eye movements, running and licking Quotation withheld because this record is not cleared for redistribution. selectivity selectivity Quotation withheld because this record is not cleared for redistribution. noise correlation noise correlation Quotation withheld because this record is not cleared for redistribution. interaction weight interaction weight Quotation withheld because this record is not cleared for redistribution. coupling coupling Quotation withheld because this record is not cleared for redistribution. 0.31 ± 0.32 to 0.41 ± 0.45, P < 10 -7 , N = 1249 — average absolute selectivity of PYR cells pre learning and post learning, with the sign test P and the N Quotation withheld because this record is not cleared for redistribution. 0.25 ± 0.20 to 0.43 ± 0.38, P = 0.002, N = 132 — Mean absolute selectivity of PV interneurons before versus after learning Quotation withheld because this record is not cleared for redistribution. 0.19 ± 0.17 to 0.19 ± 0.16, P = 1.00, N = 175 — VIP cell selectivity, unchanged across learning Quotation withheld because this record is not cleared for redistribution. 1249, 132, 58 and 175 — Numbers of PYR, PV, SOM and VIP cells imaged, across 8 mice Quotation withheld because this record is not cleared for redistribution. slope = 0.2, CI 0.14 to 0.26, R = 0.49 — Pre-learning relationship between a PV cell's selectivity and the mean selectivity of pyramidal cells within 100 µm Quotation withheld because this record is not cleared for redistribution. 0.013 from 0.077 — Average simulated noise correlation with all model interaction weights deleted, versus the observed value Quotation withheld because this record is not cleared for redistribution. 1.00, 0.99, 0.97 — Correlation between selectivity computed from calcium fluorescence and from inferred firing rate, for PV, SOM and VIP cells Quotation withheld because this record is not cleared for redistribution. 92.6% ... 70.1% ... 85.4% — agreement between the transgenic and the immunohistochemical definition of the same interneuron class - and for SOM cells nearly a third of the population depends on which definition is used Quotation withheld because this record is not cleared for redistribution. above 90% — the percentage of overlap between immunostaining and transgenic marker expression, measured in the opposite direction from the table of transgenically labelled cells Quotation withheld because this record is not cleared for redistribution. R 2 = xx — a reported model fit that was never filled in, printed twice in one sentence with only the middle value supplied Quotation withheld because this record is not cleared for redistribution. P = XX — a second unfilled placeholder, this time behind the claim that the effect was stronger Quotation withheld because this record is not cleared for redistribution. 0.31 ± 0.32 ... 0.41 ± 0.45 — mean absolute selectivity of pyramidal cells before and after learning, with standard deviations as large as the means Quotation withheld because this record is not cleared for redistribution. 0.7 — the neuropil subtraction factor, obtained by inspecting the regression slope in a SAMPLE of the dataset rather than fitted per cell 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/s41593-018-0143-z.