Claims 40
From the paper Distributed coding of choice, action and engagement across the mouse brain · claims file · checked 2026-10-21: ready
- fig1c-easy-performance
On high-contrast single-stimulus trials mice made 1.7 ± 2.5% incorrect choices and 10.1 ± 8.3% Misses (mean ± s.d. over 39 sessions of 10 mice).
Their choices were most accurate for high-contrast single stimuli (i.e. when the other stimulus was absent; 1.7 ± 2.5% incorrect choices, i.e. turns in the wrong direction; 10.1 ± 8.3% Misses, i.e. failures to turn; mean ± s.d., n = 39 sessions, 10 mice)
Results, Brain-wide recording in visual behavior, paragraph 1; Fig. 1c,dHow it will be judged
Trials with the high contrast on one side and zero on the other. Per session, the percentage of such trials with a turn to the wrong side, and with no turn (the task's recorded response, NoGo); the mean over the 39 sessions. Tolerances are the paper's s.d. over sessions. The paper does not say which trials enter. Readings (a) all trials of the session; (b) trials after dropping disengaged trials as in definition T. Reproduced when both values are within tolerance under one reading; the s.d. over sessions is reported as well.
- fig1c-low-contrast-performance
With low-contrast single stimuli mice made 5.1 ± 6.4% incorrect choices and 29.8 ± 19.8% Misses.
with low-contrast single stimuli (5.1 ± 6.4% incorrect choices; 29.8 ± 19.8% Misses)
Results, Brain-wide recording in visual behavior, paragraph 1How it will be judged
As fig1c-easy-performance, for trials with the low contrast on one side and zero on the other, under the same readings.
- fig1c-competing-performance
With competing stimuli of similar but unequal contrast (high against medium, medium against low) mice made 20.0 ± 7.9% incorrect choices and 13.9 ± 11.7% Misses.
or with competing stimuli of similar but unequal contrast (20.0 ± 7.9% incorrect choices; 13.9 ± 11.7% Misses, on trials with high vs. medium or medium vs. low contrast)
Results, Brain-wide recording in visual behavior, paragraph 1How it will be judged
As fig1c-easy-performance, for trials with high against medium or medium against low contrast (either side), an incorrect choice being a turn towards the lower contrast, under the same readings.
- fig1d-reaction-time-difficulty
Reaction times were longer in the harder conditions (low-contrast single stimuli and competing stimuli of similar contrast) than for high-contrast single stimuli.
As expected, in these more challenging cases reaction times were longer ... multi-way ANOVA
Results, Brain-wide recording in visual behavior, paragraph 1; Fig. 1dHow it will be judged
Reaction time is the movement onset (definition T, findWheelMoves3) minus stimulus onset on Left and Right choice trials. The paper names neither the factors of its multi-way ANOVA nor the trials. Readings of the factors (F1) condition class (high-contrast single, low-contrast single, competing similar contrasts) and session; (F2) condition class and subject; (F3) contrast and presence of a competing stimulus (the variables of the Fig. 1d legend) and session, the condition effect being that of either factor; readings of the trials (a) all Left and Right choice trials; (b) those of definition T (onset 125 to 400 ms). Reproduced when, under one reading, the condition factor has p < 1e-4 and the mean reaction time of the high-contrast single class is the shortest of the three.
- ed1q-wheel-choice-decoding
Choice decoded from instantaneous wheel velocity is at chance (47.7%) 50 ms before detected movement onset and near perfect (94.6%) 50 ms after it.
The decoder performed essentially at chance 50 ms prior to movement onset (47.7%) compared to near perfect performance 50 ms after movement onset (94.6%)
Methods, Training protocol, paragraph 3; Extended Data Fig. 1qHow it will be judged
Velocity is the wheel position at t minus that at t - 10 ms (Methods), at times relative to the detected onset of Left and Right choice movements. The paper does not name the decoder. Readings (a) the sign of the velocity, a zero velocity counting as a guess (50%); (b) a logistic regression on velocity with 10-fold cross-validation over trials, per session. Accuracy is the mean over the 39 sessions (Extended Data Fig. 1q, s.d. across sessions). Reproduced when both values are within tolerance under one reading.
- early-response-fraction
On trials that ended in a Left or Right choice, mice turned the wheel within 400 ms of stimulus onset on 64.9 ± 14.0% (s.d.) of trials, nearly always (96.6 ± 3.4%) in the direction of their final choice.
turning the wheel within 400 ms of stimulus appearance (64.9 ± 14.0% s.d., n=39 sessions), nearly always in the same direction as their final choice (96.6 ± 3.4%)
Methods, Training protocol, paragraph 1How it will be judged
Per session, of trials whose recorded response is Left or Right, the percentage with a turn within 0.4 s of stimulus onset, and of those the percentage whose turn direction equals the recorded response; mean over the 39 sessions. The paper does not say what counts as "turning the wheel" here. Readings (a) the onset of a turn reaching 90 d.v.a. (the Left and Right Choice definition of the next sentence of the Methods); (b) any movement onset detected by findWheelMoves3, its direction that of the movement. Tolerances are the paper's s.d. over sessions. Reproduced when both are within tolerance under one reading.
- passive-no-movement
In the passive replay mice hardly ever turned the wheel; 94.1% ± 0.6% of passive trials with high-contrast stimuli had no movement.
In these passive trials the mice hardly ever turned the wheel (94.1% ± 0.6% of trials with high contrast stimuli had no movement)
Results, Propagation of activity during the task, paragraph 3How it will be judged
Passive replay presentations of a visual stimulus at the high contrast (on either side). The paper does not give the window or the spread. Readings of the window (a) -0.05 to 0.4 s from stimulus onset, the NoGo window of definition T; (b) 0 to 0.4 s; of the value (i) the mean over sessions; (ii) the pooled fraction. No movement is no detected wheel movement onset and no movement in progress in the window. The tolerance is the paper's stated spread. Reproduced when one reading is within tolerance.
- recording-counts
Across 39 sessions in 10 mice with 92 probe insertions, 29,134 neurons were recorded (747 ± 38 per session, mean ± s.e.), 22,458 of them localized to one of 42 brain regions; the behaviour comprised 9,538 trials.
Across all sessions we recorded from 29,134 neurons (n=747 ± 38 neurons per session, mean ± s.e.), of which 22,458 were localizable to one of 42 brain regions
Results, Brain-wide recording in visual behavior, paragraphs 2 and 3; Fig. 1c legendHow it will be judged
Counts of sessions, subjects and probes are exact. Neurons are units curated as single units. Readings of "neurons" (a) all such units; (b) those also with a trial firing rate of at least 0.1 Hz. Readings of "localizable" (i) neurons whose channel lies in one of the 42 regions of Extended Data Table 1; (ii) neurons of definition N. Readings of the trials (a) all trials; (b) trials kept after definition T. Reproduced when every count is within tolerance under one reading of each; partially reproduced when the exact counts hold and one other count does not.
- edtable1-region-counts
Extended Data Table 1 gives, for each of the 42 regions, the number of recordings, mice, task-responsive neurons and neurons (for example MOs 19 recordings, 9 mice, 993 of 1,534 neurons responsive; MRN 11, 7, 722 of 857).
How it will be judged
For each region of the table, the number of sessions with neurons of the region (definition N), the number of mice, the number of neurons and the number active during the task (definition N). Reproduced when the recordings and mice match exactly and the neuron counts are within 10% in at least 38 of the 42 regions; partially reproduced in 30 to 37 regions. The regions that differ are listed.
- supptable2-sessions
The 39 sessions are those of Supplementary Table 2, each with its subject, date, age and recorded areas (for example Cori 2016-12-14, 11 weeks, ACA, CA3, DG, LS, MOs, SUB, VISp), from the 10 subjects of Supplementary Table 1 with their sex and genotype.
How it will be judged
Each session of the table matched by subject and date. Reproduced when all 39 sessions are present, their subjects' sex and genotype match Supplementary Table 1, and the set of areas with neurons in each session matches the table in at least 35 sessions; partially reproduced when all sessions are present but the areas match in fewer.
- task-modulated-fraction
60.0% of neurons (13,466) were significantly modulated during the task; of these 74.3% consistently increased and 20.2% consistently decreased their activity.
A sizeable fraction of all neurons (60.0%, 13,466 neurons) had significant modulation of firing rate during the task. Most of these neurons (74.3%) consistently increased their activity during the task, but a sizeable minority (20.2%) consistently decreased
Results, Propagation of activity during the task, paragraph 1; Supplementary Figs. 1 and 2How it will be judged
Active during the task is any of the six tests of the Methods below 0.05/6, on trials T of the neurons localized to the 42 regions with a rate of at least 0.1 Hz (13,466 of 22,458 is 60.0%, and Extended Data Table 1 sums to these). The direction of a test against baseline is the sign of the code's effectSizes (kernelFitBasic.m); for the signrank tests 1, 2, 3 and 5 the count of positive minus negative trial differences, for the ranksum test 6 the rank sum of the post-reward rates minus its null value. Supplementary Fig. 1b counts a decrease as a higher baseline rate. The paper does not define "consistently". Readings (a) every test against baseline (1, 2, 3, 5, 6) with p below 0.05/6 has the same sign, test 4 being left out; (b) as (a) over the tests with p below 0.05. Reproduced when all four values are within tolerance under one reading.
- suppfig2-rate-amplitude
Statistically responsive neurons have higher baseline firing rates than other neurons (d' = 0.35 ± 0.04), while their spike amplitudes differ much less (d' = 0.10 ± 0.02).
Significant cells tend to have higher baseline firing rates ... while differences in spike amplitude are much smaller
Supplementary Fig. 2 legendHow it will be judged
Values read from the supplement's Word file (the "±" is lost in the text). Baseline is the mean rate -0.2 to 0 s before stimulus onset; amplitude the mean spike waveform amplitude. The legend shows the cells per test (b to g) and gives one d' each; the paper does not say over what the ± is taken. Readings (a) significant on any of the six tests against the rest, the ± being a bootstrap s.e. over neurons; (b) the mean ± s.e. of the six per-test d' values. d' is the difference of means over the root mean of the two variances, a standard choice; the paper does not define it. Reproduced when both are within tolerance under one reading.
- fig2de-activity-order
When the mouse chose a contralateral stimulus, activity arose first in visual cortex and superficial superior colliculus and then spread to most regions; when it chose an ipsilateral stimulus, VIS and SCs were among the last areas to respond.
When a mouse successfully selected a visual stimulus contralateral to the recorded hemisphere, activity emerged first in classical visual regions such as visual cortex (VIS) and superficial superior colliculus (SCs), and soon spread to most of the remaining recorded regions ... VIS and SCs were now amongst the last areas to respond
Results, Propagation of activity during the task, paragraph 2; Fig. 2d,eHow it will be judged
Correct contralateral (Right choice on a right stimulus alone) and correct ipsilateral trials (T, matched in contrast for d as in definition M). Per region the mean over responsive neurons of the normalized rate of the Methods (5 ms bins, causal half-Gaussian 0.02 s, trial mean, causal half-Gaussian 0.03 s, minus the baseline -0.02 to 0 s, divided by baseline + 0.5 sp/s). Responsive is open; readings (a) the display selection of the same Methods paragraph, a ranksum p < 0.0001 in at least three consecutive 0.1 s windows stepped by 0.005 s between task trials with stimuli and movements and those without both; (b) active during the task (definition N); (c) the Figs. 3 to 5 set. The paper defines no latency. Readings (L1) the first time after stimulus onset the region's trace reaches half its maximum in 0 to 0.4 s; (L2) the time of that maximum. Reproduced when, under one reading, on contralateral trials VISp (and the VIS group mean) and SCs fall in the earliest quarter of regions, and on ipsilateral trials both fall in the latest third. The quarter and the third are this file's reading of "first" and "amongst the last".
- fig2fg-miss-restricted
On Miss trials with a contralateral stimulus, activity appeared in a visual pathway of visual areas, basal ganglia and several midbrain structures but did not spread to the whole brain; on Miss trials with an ipsilateral stimulus there was essentially no activity.
activity was found in a “visual pathway” consisting of classical visual areas, basal ganglia, and several midbrain structures ... but failed to propagate globally. When the mouse missed an ipsilateral stimulus, essentially no activity was seen in the recorded hemisphere
Results, Propagation of activity during the task, paragraph 2; Fig. 2f,gHow it will be judged
Fraction of neurons per region significantly activated (definition R, z > 4; readings Z1, Z2) on contrast-matched (M) correct contralateral Go trials and contralateral Miss trials, and on ipsilateral Miss trials. The visual pathway set is the VIS areas, LGd, LP, LD, SCs, CP, GPe, SNr, SCm, MRN, APN and ZI (the areas Fig. 3d and the Results name for vision); the rest are other regions. Reproduced when, under one reading, (i) on contralateral Miss trials the mean fraction over visual pathway regions exceeds that over other regions, (ii) the mean fraction over other regions is at least twice as large on contralateral Go as on contralateral Miss trials, and (iii) the pooled fraction on ipsilateral Miss trials is below half that on contralateral Miss trials. The thresholds are this file's reading of "failed to propagate globally" and "essentially no activity".
- fig3d-vision-localized
Neurons encoding the contralateral visual stimulus were found mainly in classical visual areas, common in visual cortex, thalamus and SCs, and occasionally in frontal cortex (MOs, ACA, PL), basal ganglia (CP, GPe, SNr) and midbrain nuclei (SCm, MRN, APN, ZI).
Neurons encoding vision (i.e. requiring Vision kernels) were found in a pathway comprising primarily classical visual areas ... They were common in visual cortex (VIS) and thalamus, and superficial superior colliculus (SCs), but also were found occasionally in other structures such as frontal cortex (MOs, ACA, PL), basal ganglia (CP, GPe, SNr), and several midbrain nuclei (SCm, MRN, APN, ZI
Results, Globally distributed action coding, paragraph 3; Fig. 3d,fHow it will be judged
Kernel test K for the Contralateral Vision kernels, on neurons N (Figs. 3 to 5 set) and trials T; fraction per region of neurons passing analysis criteria, regions with fewer than 5 such neurons left out (Fig. 3d legend). Reproduced when more than half of all neurons requiring Contralateral Vision lie in the VIS areas, visual thalamus (LGd, LP, LD) and SCs; the regions with the five highest fractions are among these; and each of MOs, ACA, PL, CP, GPe, SNr, SCm, MRN, APN and ZI that is analysed has at least one such neuron. The half and the five regions are this file's reading of "primarily" and "common". Partially reproduced when the first two hold only.
- fig3e-action-global
Neurons encoding action (requiring the Action kernel) were found in all recorded regions, and most did not require a Choice kernel.
In contrast, neurons encoding action (requiring an Action kernel) were spread throughout all recorded regions ... The great majority of neurons encoding action did not require an additional Choice kernel
Results, Globally distributed action coding, paragraph 3; Fig. 3e,gHow it will be judged
Kernel test K for the Action and the Choice kernels, as in fig3d-vision-localized. Reproduced when every analysed region (at least 5 neurons) has Action-requiring neurons and more than half of Action-requiring neurons do not require the Choice kernel; partially reproduced when at least 90% of regions have them.
- ed4f-focality
The Focality index was 0.079 for choice, 0.069 for contralateral vision and 0.040 for action, with choice and vision each significantly more focal than action.
This focality index was 0.079 for choice, 0.069 for visual kernels and 0.040 for action kernels; the differences between Choice and Move, as well as Contralateral Vision and Action, were statistically significant (p<0.05; bias-corrected bootstrap)
Extended Data Fig. 4f legend; Methods, Focality IndexHow it will be judged
F = sum of p_a squared over the square of the sum of p_a, p_a the fraction of neurons in region a requiring the kernel (kernel test K; Methods), over the 42 regions (the legend's floor of 0.0238 is 1/42). Intervals are 95% bootstrap intervals, normal approximation with bootstrap bias and s.e. (MATLAB bootci "norm"); 2,000 resamples, a standard choice. The paper does not say what is resampled or how the difference is tested. Readings (B1) neurons resampled within each region; (B2) sessions resampled; (S1) the bootstrap interval of the difference (same resamples) excludes 0; (S2) the two intervals do not overlap. Reproduced when the three values are within tolerance and both differences are significant under one reading.
- ed5h-choice-shuffle-fpr
With Left and Right choice labels shuffled, 14 neurons (0.33%) passed the Choice kernel test, the false-positive rate of the test.
A small number (14, 0.33%) of neurons are false positives in this shuffle test
Extended Data Fig. 5h legend; Methods, Estimation of false positive rateHow it will be judged
Each Left or Right choice trial gets the choice of another Left or Right trial drawn without replacement, and the whole kernel test K for Choice is rerun from the start (Methods). The fraction is of neurons entering the nested test (full-model VE at least 2%; 14 of about 4,240). The tolerance is two Poisson s.d. of a count of 14 (7.5 neurons), since 10% (1.4 neurons) is below the spread of one shuffle. Reproduced when the fraction is within tolerance.
- fig4b-choice-regions
Choice-selective neurons were rare and found in frontal cortex (MOs, PL, MOp), basal ganglia (CP, SNr), higher-order thalamus, SCm, MRN and ZI, but not in VISp; 54 were in the midbrain (MRN, SCm, SNr, ZI) and 48 in the forebrain (MOs, PL, MOp, CP).
These neurons were rare, and were found in frontal cortex (MOs, PL, and MOp), basal ganglia (CP, SNr), higher-order thalamus, motor-related superior colliculus (SCm), as well as two subcortical nuclei which unexpectedly contained neurons selective for choice (MRN and ZI ... but choice-selective neurons were not found in VISp
Results, Forebrain and midbrain choice coding, paragraphs 1 and 3; Fig. 4bHow it will be judged
Kernel test K for the Choice kernel, as in fig3d-vision-localized. The paper does not define higher-order thalamus; this file reads it as the thalamic regions other than the first-order relays LGd, VPL, VPM and MG. Reproduced when both counts are within tolerance, VISp has no choice-selective neuron, each of MOs, PL, MOp, CP, SNr, SCm, MRN, ZI and one higher-order thalamic region has at least one, and the overall fraction of analysed neurons that are choice-selective is below 5% (this file's reading of "rare"). Partially reproduced when the regions hold but a count does not. The VISp count is reported with the count the 0.33% false-positive rate predicts.
- fig4c-choice-timing
The time course of choice decoding did not differ between frontal cortex, striatum and midbrain (two-way ANOVA on area and time, no effect of area and no interaction).
the time course of choice decoding could not be statistically distinguished between choice-selective regions in frontal cortex, striatum, and midbrain (two-way ANOVA on brain region and time, interaction p > 0.05
Results, Forebrain and midbrain choice coding, paragraph 2; Fig. 4c; Methods, Population decodingHow it will be judged
Residual rates from the model without the Choice kernel (kernel test K, code determineResidualsWithout "moveLR"), Left against Right choice trials; an L1-regularized logistic regression on the residuals -0.05 to 0 s from movement onset of training trials; test trials' residuals at each time projected on its weights; the score is the mean projection on Left minus Right test trials (Methods). One population per area per recording; each score is divided by its area's mean across recordings at choice time (movement onset). A claim of absence, judged on the paper's test, a two-way ANOVA with factors time and area group (frontal, midbrain, striatum), at alpha 0.05. Time has 51 levels (50 d.f.). Open details, each a reading the notebook runs; (a) times -0.25 to 0 s from movement onset in 5 ms steps (the Choice kernel support and bin) or (b) -0.2 to 0.05 s; folds (a) 5-fold or (b) 10-fold over trials; L1 strength (a) chosen by inner cross-validation or (b) the library default. Reproduced when the interaction and area effects both have p > 0.05; not reproduced when either has p < 0.05. The F values and the effect sizes are reported.
- fig4d-jpecc-vis-leads
In jPECC, visual cortex activity leads frontal cortex activity by about 40 ms in the period about 200 ms before movement, and visual cortex leads frontal cortex and midbrain at about 100 ms after stimulus onset.
population activity in visual cortex predicts that in frontal cortex following a ~40 ms lag, but only in the period ~200ms prior to movement
Fig. 4d legend; Extended Data Fig. 8d legend; Methods, jPECC analysisHow it will be judged
jPECC as in the Methods and jPECC.m. Recordings with neurons of both groups (VIS, Frontal, Midbrain); spike counts in 10 ms bins, causal half-Gaussian of s.d. 25 ms, divided by baseline + 1 sp/s; per group and time bin PCA to 10 dimensions (fewer when either group has fewer neurons, the same number for both, jPECC.m); 10-fold over trials; CCA on training PCs with L2 regularization lambda 0.5 (rows lambda times identity appended, as jPECC.m); Pearson r of test projections on the first canonical pair, for each pair of time bins. Asymmetry index at t, the mean r along the slice from (t, t) to 50 ms of lead of the first group minus the mean along 50 ms of lead of the second; across recordings a t-test against 0 at each t, alpha 0.05 (the paper names no correction). Open; the baseline window, readings (a) -0.2 to 0 s and (b) -0.1 to 0 s before stimulus onset; the trials, readings (a) all trials T for stimulus alignment and Left and Right choice trials for movement alignment, (b) Left and Right choice trials for both. Reproduced when, under one reading, VIS leads Frontal at time points between -0.2 and -0.05 s from movement onset and VIS leads Frontal and Midbrain near 0.1 s after stimulus onset.
- fig4d-jpecc-no-lag-frontal-midbrain
Population activity in midbrain and frontal cortex shows no consistent lead or lag.
it revealed no lag for activity correlations between frontal and midbrain areas
Results, Forebrain and midbrain choice coding, paragraph 2; Fig. 4d; Extended Data Fig. 8How it will be judged
A claim of absence, judged on the paper's test, the t-test across recordings of the Frontal-Midbrain asymmetry index against 0 at each time point (as fig4d-jpecc-vis-leads, same readings of the open details), alpha 0.05 uncorrected as the paper gives it. Reproduced when no time point from -0.25 to 0 s from movement onset is significant; not reproduced when any is. The mean index and the count of significant points are reported, with a Bonferroni variant.
- fig4f-contralateral-preference
Nearly all midbrain choice-selective neurons (53/54, 98%) preferred contralateral choices, while 19/48 (40%) forebrain ones preferred ipsilateral choices, significantly more than in the midbrain.
In midbrain (MRN, SCm, SNr, and ZI) nearly all choice-selective neurons (53/54, 98%) preferred contralateral choices ... with a sizeable proportion preferring ipsilateral choices (19/48, 40%, significantly more than in the midbrain
Results, Forebrain and midbrain choice coding, paragraph 3; Fig. 4e-gHow it will be judged
Choice-selective neurons of fig4b-choice-regions. Preference is the choice with the higher model activity at movement onset relative to baseline (Fig. 4f,g legend, "estimated from the kernel model"). Open is the model activity. Readings (a) the Action kernel plus or minus the Choice kernel at lag 0 (kernels on baseline-subtracted rates); (b) the full model's prediction at movement onset averaged over Left and over Right choice trials. Reproduced when both percentages are within tolerance and a two-sided Fisher's exact test of the 2 by 2 table has p < 1e-5, under one reading.
- fig4fg-opposed-activity
Many midbrain choice-selective neurons (29/54, 54%) increased before one choice and were suppressed below baseline before the other, against 10/48 (21%) in the forebrain.
Many midbrain choice-selective neurons, moreover, exhibited directionally-opposed activity ... (10/48, 21% suppressed for non-preferred choice, significantly less than in the midbrain
Results, Forebrain and midbrain choice coding, paragraph 3; Fig. 4f,gHow it will be judged
As fig4f-contralateral-preference, a neuron being suppressed when its model activity for the non-preferred choice at movement onset is below baseline (below 0 after the baseline subtraction), under readings (a) and (b). Reproduced when both are within tolerance and the two-sided Fisher's exact test has p < 1e-3, under one reading.
- ed7b-detect-probability-widespread
Neurons with significant detect probability (Go against NoGo for the same stimulus) around movement onset are found widely across regions, replicating the global action signal.
Percentage of neurons in each area with significant Detect Probability measured around movement onset, replicating the finding from Fig 2d,e and Fig 3e that non-selective action signals are distributed widely
Extended Data Fig. 7b legend; Methods, ccCP and Detect Probability analysisHow it will be judged
ccCP algorithm of the Methods and choiceProbShuf.m, Go (Left or Right) against NoGo trials (T) within each of the 16 stimulus conditions, 2,000 within-condition label shuffles; a neuron is significant when its rank among shuffles puts it below 0.025 or above 0.975 (chance 5%, the dashed line of the legend). Sessions with fewer than 10 includable trials left out. NoGo trials take a movement time drawn from the Go trials (definition T). The window "around movement onset" is open; readings (a) -0.1 to 0.05 s (the pre-movement window of test 3); (b) -0.05 to 0 s (the choice decoding window). Regions are marked when the 95% binomial interval of the fraction (Clopper-Pearson, a standard choice) excludes 5%. Reproduced when at least three quarters of regions are marked above chance under one reading.
- ed7c-choice-probability-restricted
Neurons with significant choice probability around movement onset are restricted to frontal cortex, basal ganglia, midbrain and certain thalamic nuclei; six sessions were excluded for too few trials, holding 20.8% of the MOs choice-selective neurons of the kernel analysis.
this analysis broadly replicates the finding from Fig 4b that choice selective neurons, around the time of movement onset, are restricted to frontal cortex, basal ganglia, midbrain, and certain thalamic nuclei
Extended Data Fig. 7c legend; Methods, ccCP and Detect Probability analysisHow it will be judged
As ed7b-detect-probability-widespread, for Left against Right choice trials within stimulus conditions; only conditions with both choices contribute. Reproduced when, under one reading, no VIS area and no hippocampal region (CA1, CA3, DG, SUB, POST) is marked above chance, at least one frontal, one basal ganglia (CP, GPe, SNr, ACB, LS) and one midbrain region is, and six sessions are excluded. Partially reproduced when the regional pattern holds and the session count does not.
- ed5g-ipsilateral-vision-rare
Ipsilateral visual stimuli are hardly encoded, unlike contralateral ones.
since the same argument should apply to Ipsilateral Vision correlates, and we found very little of such correlates
Methods, Determining individual neuron selectivity, paragraph 2; Extended Data Fig. 5gHow it will be judged
Population decoding of the Methods (as fig4c-choice-timing) of contralateral against no contralateral stimulus, and ipsilateral against no ipsilateral stimulus, on residuals from the model without the tested Vision kernels, logistic regression on 0.05 to 0.15 s after stimulus onset; mean score per area. Same open readings as fig4c-choice-timing. Reproduced when the largest area mean for ipsilateral decoding is below a quarter of the largest for contralateral decoding under every reading. The counts of neurons requiring each kernel (test K) are shown as a variant.
- fig3c-variance-explained
The kernels often explained a small fraction of cross-validated variance, even for neurons whose mean rates they predicted well (50.2% for the Fig. 3a neuron, 13.6% for the Fig. 3b neuron).
The fraction of cross-validated variance the kernels explained was frequently small ... (50.2% for the neuron in Fig. 3a and 13.6% for the neuron in Fig. 3b
Results, Globally distributed action coding, paragraph 2; Fig. 3cHow it will be judged
Full-model cross-validated VE (kernel test K). Reproduced when the median VE over all analysed neurons is below 10% (this file's reading of "frequently small") and, if the two example neurons can be identified (the paper gives only their regions and rasters, Figs. 2a,b and 3a,b), their VE is within tolerance. When they cannot be identified the claim is judged on the median alone and the examples are reported as not identifiable.
- fig5b-miss-more-than-passive
More neurons were significantly activated by contralateral visual stimuli on Miss trials in the task than on passive trials with the same contrasts, and passive responses were weaker than those on Miss trials.
more neurons were significantly activated by visual stimuli during the task (Miss trials), than in the passive trials
Results, Distributed coding of task engagement, paragraph 1; Fig. 5b; Fig. 2hHow it will be judged
Contrast-matched (M) Miss and passive trials; per region the fraction of neurons significantly activated (definition R, z > 4; readings Z1, Z2). Neurons are open; readings (a) neurons N of the Figs. 3 to 5 set; (b) all neurons of the 42 regions with a rate of at least 0.1 Hz. Passive trials with a wheel movement in -0.05 to 0.4 s are kept (the Methods drop none). Reproduced when, under one reading, the pooled fraction is larger on Miss than on passive trials, the excess (Miss minus passive) is positive in more than half the regions, and 34 sessions are kept.
- fig5c-prestimulus-task-vs-passive
Before the stimulus, neocortex and sensory thalamus were less active in the task than in the passive context, while basal ganglia and other subcortical choice areas were more active; VISp was lower and CP higher on Miss trials than on passive trials.
While neocortex and sensory thalamus showed a net decrease in pre-stimulus activity during task context, other regions ... showed a consistent increase
Results, Distributed coding of task engagement, paragraph 1; Fig. 5a,c; Extended Data Fig. 9a; Methods, Engagement Index and pre-stimulus analysesHow it will be judged
Per region a nested ANOVA (Methods) on each neuron's mean rate 250 to 50 ms before stimulus onset, log10(x + 1 sp/s), trials with wheel movement in that window dropped. The factors are open, as the Methods and the Extended Data Fig. 9a legend disagree; readings (a) condition, session and neuron nested in session (Methods); (b) condition, session and subject (legend). Task trials are trials T; passive trials are open, readings (a) all passive replay events; (b) passive visual stimulus presentations. A region differs when the condition p < 0.0012 (0.05 over 42 regions). VISp and CP are compared as in Fig. 5a, the mean pre-stimulus rate over neurons on contrast-matched (M) Miss trials against passive trials, windows as the readings of ed6b-prestimulus-order; the paper gives no test for them. Reproduced when, under one reading, every differing neocortical region is lower in the task, every differing non-neocortical region other than visual thalamus (LGd, LP, LD) is higher, VISp is lower on Miss than on passive trials, and CP is higher.
- ed6b-prestimulus-order
Pre-stimulus firing in midbrain, basal ganglia and hippocampal regions is ordered Passive < Miss < Go, while in neocortical areas it is higher in Passive than in Miss and Go.
the pre-stimulus firing rate of midbrain, basal ganglia, and hippocampal regions is in the order Passive < Miss < Go; whereas pre-stimulus activity in neocortical areas instead is arranged as Passive > Miss and Go
Extended Data Fig. 6b legendHow it will be judged
Per region the mean over neurons (N, Figs. 3 to 5 set) of the mean pre-stimulus rate on contrast-matched (M) Go, Miss and passive trials. The window is open; readings (a) -0.2 to 0 s (the engagement index window); (b) -0.25 to -0.05 s (the nested ANOVA window). Groups from utils/brainRegionGroups.m (midbrain MRN, SCm, SCs, APN, PAG; basal ganglia CP, GPe, SNr, ACB, LS; hippocampus POST, SUB, DG, CA1, CA3); neocortex as defined above. Reproduced when, under one reading, more than half the midbrain, basal ganglia and hippocampal regions show Passive < Miss < Go and more than half the neocortical regions show Passive above both Miss and Go.
- fig5d-go-miss-matches-task-passive
Regions whose pre-stimulus rate differed between task and passive contexts showed similar differences between Go and Miss trials.
Areas that showed differences in pre-stimulus firing rate between task and passive contexts also showed similar differences between Go and Miss trials
Results, Distributed coding of task engagement, paragraph 2; Fig. 5dHow it will be judged
Per region the mean over neurons of task minus passive and of Go minus Miss pre-stimulus rates (as ed6b-prestimulus-order, same readings), contrast-matched. The paper gives no test for Fig. 5d; the test used is the Pearson correlation across regions, the measure the paper reports for the companion comparison of Extended Data Fig. 7f. Reproduced when r > 0 with p < 0.05 under one reading.
- ed7f-detect-probability-engagement
The pre-stimulus detect probability of each area correlated with its task minus passive difference (r = 0.48, p = 0.001), as the Go minus Miss difference does.
The pre-stimulus Detect Probability was correlated with the Engagement Index (i.e. Task-Passive difference) similarly to the Go-Miss difference (with r = 0.48, p=0.001; not shown)
Extended Data Fig. 7f legendHow it will be judged
Per neuron the pre-stimulus DP (ccCP of Go against NoGo within stimulus conditions, minus 0.5); per area the mean over neurons; Pearson r across areas with the area's mean task minus passive pre-stimulus rate (as fig5d-go-miss-matches-task-passive). Windows as in ed6b-prestimulus-order, readings (a) and (b). Reproduced when r is within tolerance and p < 0.05 under one reading.
- fig5f-engagement-index
An engagement index, the projection of pre-stimulus population activity onto the task minus passive difference vector, was higher on Go than on Miss trials consistently across recordings (mean difference 8.42 a.u., paired t-test p < 1e-4).
This engagement index differed between Go and Miss trials consistently across recordings ... paired t-test
Results, Distributed coding of task engagement, paragraph 2; Fig. 5f; Methods, Engagement Index and pre-stimulus analysesHow it will be judged
Neurons of the regions that differ in fig5c-prestimulus-task-vs-passive, x_n the mean rate -0.2 to 0 s before stimulus onset on all task trials minus that on all passive trials, normalized to unit L2 norm per session; the index of trial i is x times the trial's pre-stimulus rate vector; per session the mean on Go minus the mean on Miss trials, with N trials per contralateral contrast, N the smaller of the Go and NoGo counts (Methods). Open; readings of task trials (a) trials T, (b) all task trials; of the contrast conditions (a) contralateral contrast > 0 (as definition M), (b) all four contralateral contrast levels, zero included. The tolerance of the neuron count is the paper's s.e. Reproduced when the paired t-test across sessions has p < 1e-4 with Go higher and the mean is within tolerance, under one reading; partially reproduced when the test holds and the mean does not.
- fig5e-engagement-predicts-go
The probability of a Go response rises with the engagement index on single trials.
it was possible to predict whether an animal would respond to the stimulus on a given trial within the task context, by projecting pre-stimulus population activity onto a weight vector given by the difference of pre-stimulus activity in passive and task contexts
Results, Distributed coding of task engagement, paragraph 2; Fig. 5eHow it will be judged
Engagement index as in fig5f-engagement-index, z-scored; contrast-matched Go and Miss trials pooled over sessions; logistic regression of Go on the index (Fig. 5e legend). Open is the z-scoring, readings (a) within session, (b) over all trials. Reproduced when the slope is positive and its 95% interval excludes 0 under one reading.
- ed9i-engagement-beyond-arousal
The engagement index improves the prediction of Go responses beyond inter-trial interval, previous reward, pupil area and motion energy (deviance test, p = 1.5e-8).
The null hypothesis that engagement index had no additional effect on Go probability over all other variables was rejected using a deviance test (p=1.5e-8)
Extended Data Fig. 9i legend; Methods, GLM prediction of P(Go) from pre-stimulus variablesHow it will be judged
Binomial GLM of Go against Miss on inter-trial interval, previous reward (0 or 1), pre-stimulus pupil area and motion energy (z-scored) and their squares, against the same model plus the engagement index, contrast-matched trials (Methods). Sessions with usable pupil video only (34 of 39 in the Methods). Open; the pre-stimulus window, readings (a) -0.2 to 0 s, (b) -0.25 to -0.05 s; pooling, readings (a) all sessions in one GLM, (b) one GLM per session with the deviances summed. Reproduced when the deviance test has p < 0.05 under one reading; p is reported against 1.5e-8.
- ed9cf-engagement-controls
The Go against Miss difference in engagement index persists on trials after rewards, after regressing out previous reward, and on trials after short and after long inter-trial intervals.
c-f, Same plot after restricting to contrast-matched trials following rewards ... The effect persists in each case
Extended Data Fig. 9c-f legendHow it will be judged
The index of fig5f-engagement-index on four trial sets; trials after a rewarded trial; all trials with each previous-reward group's mean subtracted; trials after short and after long inter-trial intervals. Open; the interval split, readings (a) median per session, (b) median over all trials; the test (Extended Data Fig. 9b says t-test), readings (a) paired t-test across sessions, (b) two-sample t-test over pooled Go and Miss trials. Reproduced when all four sets give Go higher with p < 0.05 under one reading.
- ed9k-engagement-vector-specific
The engagement vector correlates better with the Go minus Miss vector than with the vectors of other behavioural variables, though not significantly for previous reward (p = 0.096).
In each case we find that the correlation is greater for the Go-Miss vector (not reaching significance at alpha=0.05 level, however, for the comparison with the previous reward vector, p=0.096)
Extended Data Fig. 9k legendHow it will be judged
Per recording, the Pearson correlation over neurons between the engagement vector (task minus passive pre-stimulus rate) and the Go minus Miss vector, and between the engagement vector and each other variable's vector. Open; each variable's vector, readings (a) the difference of pre-stimulus rates between trials above and below the variable's median (for previous reward, rewarded minus not); (b) each neuron's regression slope on the variable; the test across recordings, readings (a) paired t-test, (b) Wilcoxon signed-rank. Variables are pupil area, motion energy, inter-trial interval and previous reward. Reproduced when, under one reading, the Go-Miss correlation is larger on average for all four and the difference has p < 0.05 for pupil, motion energy and inter-trial interval.
- suppfig3-nonresponsive-neurons
The engagement index differs between Go and Miss trials even when computed from neurons that are otherwise not responsive in the task.
All neurons show a significant difference in Engagement Index between Go and Miss trials
Results, Distributed coding of task engagement, paragraph 1; Supplementary Fig. 3 legendHow it will be judged
As fig5f-engagement-index, computed separately from all neurons, from task-responsive neurons (the Figs. 3 to 5 set) and from the others of the same regions. The p-values are only in the figure. Reproduced when the paired t-test gives Go higher with p < 0.05 for each set.
- histology-alignment-error
Probe tracks were localized with a lateral error of 39.3 um (median distance of DiI spots from the fitted track) and a longitudinal error of 88.7 um (median absolute deviation of held-out landmarks), with 15.1 ± 6.9 DiI spots per probe and a scaling of 1.08 ± 0.02.
We quantified the lateral localization error by the median distance of the DiI spots from this common trajectory (39.3 μm)
Methods, Anatomical targeting and probe localization, paragraphs 2 and 3How it will be judged
Needs the DiI spot coordinates in the atlas and the electrophysiological landmarks per probe (Methods); the trajectory by total least squares (Extended Data Fig. 2b), the longitudinal map by linear regression of landmarks, leave-one-landmark-out errors. Reproduced when all four values are within tolerance (the tolerances of the last two are the paper's s.d.).
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none