Claims 39
From the paper Retrosplenial cortex enables context-dependent goal-directed sensorimotor transformation · claims file · checked 2026-10-21: ready
- expert-sessions
The expert sessions of the behavioural data come from 41 mice and number 201, with an average of 23 context blocks per session.
Mice were trained in this task until they reached expert performance, completing an average of 23 blocks per session
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 2; Figure 1C legend; Figure 1 figure supplement 1A and 1BHow it will be judged
The Methods (Behavioral task and training curriculum) define a block contrast, for each block but the first and last, as the mean of the absolute differences in whisker hit rate between the block and its two neighbours, and call a session expert when its contrasts are significantly above 0.375. The test is that of the code, the lower bound of a 95% bootstrap interval (scipy.stats.bootstrap, 10,000 resamples, defaults otherwise) of the mean contrast above 0.375 (process main_analysis/figure1_supp_analysis.py:53-187), with early-lick trials dropped. Blocks are the context epochs of the session (the Methods); the code cuts blocks of 20 non-early-lick trials (figure1_supp_analysis.py:72-85), shown as a variant. Blocks per session are counted from the context epochs. Applied to every behavioural session of the dataset, reproduced when the three counts are within tolerance; the overlap with the code's list sessions_Context_sessions_expert.yaml is reported.
- fig1c-auditory-hit-both-contexts
In expert sessions mice lick to the auditory tone in both contexts, with hit rates of 96 ± 4% in W+ and 94 ± 5% in W- (mean ± SD over 41 mice).
In expert sessions, mice showed consistent licking responses to the auditory stimulus in both contexts (Auditory hit rate: W+ = 96 ± 4 %, W- = 94 ± 5 %; n = 41 mice)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 2; Figure 1CHow it will be judged
Lick probability in auditory trials, per mouse as in the opening comment (behaviour), over the expert sessions under either reading of the selection. The tolerances are the paper's stated SDs over mice.
- fig1c-whisker-context-modulation
Licking to the whisker stimulus depends strongly on context, 84 ± 6% in W+ against 32 ± 7% in W- (mean ± SD over mice).
and strong context-modulated licking in response to whisker stimuli (W+ = 84 ± 6 %, W- = 32 ± 7 %) (Figure 1C)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 2; Figure 1CHow it will be judged
As fig1c-auditory-hit-both-contexts, for whisker trials. The tolerances are the paper's stated SDs over mice.
- fig1d-independent-of-noise-identity
The context modulation of whisker licking holds whichever background noise (pink or brown) is associated with each context.
irrespective of which background noise was associated to each context (Figure 1D)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 2; Figure 1DHow it will be judged
The paper reports no test for this panel, and the code computes none (plot utils/figure1D.py), which splits the mouse-level lick probabilities by the noise identity (context_background). The mice are split by the noise assigned to W+ (Methods, each mouse randomly assigned a whisker-rewarded context). Reproduced when, in both groups of mice, the mean whisker lick probability is higher in W+ than in W- and the paired t-test across the mice of the group (the paper's test for Figure 1C) gives p < 0.05; partially reproduced when W+ is higher in both groups but the test passes in one only. The number of mice per group is reported.
- fig1c-catch-higher-wplus
Catch-trial (spontaneous) lick rates are low but slightly higher in W+ blocks, 13 ± 7% against 4 ± 2% in W-.
Lick rates in catch trials were overall low, although slightly higher in W+ blocks (Catch false alarm rate: W+ = 13 ± 7 %, W-: 4 ± 2%; t-value = −9, p-value = 1.6 x 10-10)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 2; Figure 1CHow it will be judged
Per mouse catch-trial lick probability per context, as in the opening comment (behaviour); a two-sided paired t-test of W- against W+ across mice, uncorrected, as plot utils/figure1C.py:78-82. Reproduced when both means are within the stated SDs, W+ is higher and p < 0.05.
- fig1e-context-off-no-discrimination
When the two context noises are replaced by constant white noise, keeping the block reward structure, mice no longer discriminate between the context blocks in their whisker responses.
After removing the explicit contextual information mice were unable to discriminate between context blocks (Figure 1E and Figure 1 – figure supplement 1E)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 2; Figure 1E; Figure 1 figure supplement 1EHow it will be judged
A claim of absence. The sessions are the Context Off days of the 7 mice (process configs sessions_Context_sound_off_only.yaml, the middle of three days in sessions_Context_sound_off.yaml). The paper reports no test for the panel and the code computes none for Figure 1E (plot utils/figure1E.py); for supplement 1E the code runs the Figure 1G test, a paired t-test across mice of the last whisker trial before against the first after the transition, on the Off days (plot utils/figure1_supp.py:85-125), and prints it. That is the only test the paper's material holds, so the absence is judged on it alone (protocol, claims of absence), for each transition direction, two-sided, p < 0.05. Reproduced when both directions give p >= 0.05, not reproduced when either gives p < 0.05 in the direction of the new context. A two-sided paired t-test across the 7 mice of whisker lick probability in W+ against W- blocks on the Off day is shown as a variant. The W+ minus W- difference on the Off day and on the On days is reported.
- fig1g-first-trial-switch
Mice change their whisker response at the first whisker trial after a context change, from 26 ± 8% to 73 ± 13% from W- to W+, and from 87 ± 7% to 57 ± 13% from W+ to W-.
The probability of licking in response to whisker deflection was significantly different comparing the first trial after context change with the last trial of the previous block (transition from W- to W+: pre = 26 ± 8 %, post = 73 ± 13 %, t-value = 21, p-value = 2.8 x 10-23; transition from W+ to W-: pre = 87 ± 7 %, post = 57 ± 13 %, t-value = −14, p-value = 7.3 x 10-17)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 3; Figure 1GHow it will be judged
Whisker trial -1 is the last non-early-lick whisker trial of a block and +1 the first of the next (cicada cicada_tools/behavior/behavior_analysis_utils.py:214-274); per mouse the mean over transitions within a session, then over sessions (Methods, Figure 1G; cicada behavior_analysis_plots.py:787-792); a two-sided paired t-test, post against pre, across mice, per direction (plot utils/figure1FG.py:62-63). The code also saves a Bonferroni value (x2); whether the printed p is corrected is not stated, so both are reported. Reproduced when the four values are within the stated SDs and both tests give p < 0.05 in the stated direction.
- fig1f-gradual-further-adaptation
After the first whisker trial of a new block, whisker responses adapt slightly further over the following whisker trials, the second whisker trial already closer to the new context than the first.
Mice changed their licking behavior as early as the first whisker trial following the context transition with a slight gradual further adaptation over subsequent whisker trials (Figure 1F)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 3; Figure 1F; Results, paragraph 5 (Figure 1L)How it will be judged
The paper reports no test. Mean lick probability over mice at whisker trials +1 to +4 after each transition, built as in fig1g-first-trial-switch (the code's window, 3 trials before to 4 after). Reproduced when, for both directions, trial +2 is closer than trial +1 to the new context (higher after W- to W+, lower after W+ to W-) and trial +4 is at least as close as +2; partially reproduced when this holds in one direction. The paired t-test across mice of +2 against +1 (the paper's test for Figure 1G) is reported per direction.
- fig1h-temporal-integration
The lick probability to the first whisker trial of a block rises with the time elapsed since a transition to W+ and falls with it after a transition to W-, while the last whisker trial before a transition shows no sign of anticipating the switch.
We observed a progressive increase in lick probability in W+ and a corresponding decrease in W−, suggesting temporal integration of contextual information after context switch. However, there was no apparent sign of anticipation of the context switch in the behavior of the mice.
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 3; Figure 1HHow it will be judged
As the Methods (Figure 1H) and the code give it, for each session the last whisker trial of each block and the first of the next (the first of the first block and the last of the last block left out), timed from the transition (negative before it), assigned to 10 s bins, bin -1 being [-10, 0) s and bin +1 [0, 10) s (cicada behavior_analysis_plots.py:1076-1145; plot utils/figure1H.py:107-116); trials pooled over all sessions and mice, the lick probability of a bin being its licked trials over its trials. The code keeps early-lick trials here; the variant without them is shown. The paper reports no test. The rise and fall are judged by the code's fit (plot utils/misc/stats.py:139-212), an exponential A exp(-t/tau) + C over bin -1 (as t = 0) and bins +1 to +5, tau bounded in 0.1 to 50 bins; reproduced for the first part when, for W+, the fitted values increase from bin +1 to bin +5 and, for W-, decrease. The absence of anticipation has no test in the paper; readings (a) the Pearson correlation of bin lick probability with bin time over bins -5 to -1, per direction, p >= 0.05, (b) the lick probability of bin -1 against bins -5 to -2 pooled, Fisher's exact test, p >= 0.05. The second part holds when both readings find no effect, fails when both find one toward the coming context, and is otherwise blocked as methods_underspecified. Reproduced when both parts hold, partially reproduced when one does.
- fig1i-pupil-larger-wplus
Before the whisker stimulus, pupil area is larger in W+ than in W- (0.52 ± 0.13 against 0.50 ± 0.12 mm2), while whisker angle and jaw opening show no general effect of context.
Pupil area was significantly modulated by context with larger values in W+ compared to W- (W+: 0.52 ± 0.13 mm2, W-: 0.50 ± 0.12 mm2, t-value = 3.7, p-value = 0.023)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 4; Figure 1I; Figure 1 figure supplement 1GHow it will be judged
Sessions are the four widefield expert groups with face video (24 mice, 97 sessions in the code, process main_analysis/process_deeplabcut_data.py:266-268). Body parts filtered by likelihood (Methods, Figures 1I-J, 80% whisker, 60% jaw, 50% pupil), a body part dropped for a session when under 70% of its points pass; pupil area as the polygon of its four points, all four above 50% likelihood (Methods, Behavioral kinematics). Whisker trials only; the mean over the 2 s pre-stimulus quiet window per trial, then per session, then per mouse, by context (plot utils/figure1IJ.py:26-147); a two-sided paired t-test across mice, W+ against W-, Bonferroni corrected by 20 as in the code. Reproduced when the pupil means are within the stated SDs, W+ is larger with corrected p < 0.05, and whisker angle and jaw opening give corrected p >= 0.05 (the absence judged on the same test).
- fig1i-posture-before-lick
Independently of context, before trials in which the mouse then licks to the whisker stimulus the jaw rests more open (0.27 ± 0.05 against 0.23 ± 0.04 mm) and the whisker is more posterior (97.3 ± 13 against 98.6 ± 12 degrees).
Whisker angle and jaw opening were modulated by trial outcome with larger values in the baseline preceding lick for jaw opening (Lick: 0.27 ± 0.05 mm, No-Lick-: 0.23 ± 0.04 mm, t-value = 4.6, p-value = 0.0035), and smaller values in the baseline preceding lick for whisker angle (Lick: 97.3 ± 13 °, No-Lick-: 98.6 ± 12 °, t-value = −4, p-value = 0.0103)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 4; Figure 1 figure supplement 1GHow it will be judged
As fig1i-pupil-larger-wplus, with trials split by lick outcome over both contexts (plot utils/figure1IJ.py:171-189). Jaw opening is the vertical distance from the session's 5th percentile of the jaw trace, whisker angle as the Methods (Behavioral kinematics) define it, 180 degrees for a whisker parallel to the body. Reproduced when the four means are within the stated SDs and both tests give corrected p < 0.05 in the stated directions.
- fig1j-jaw-speed-context
Pre-stimulus jaw speed is higher in the rewarded context (3.0 ± 1.0 against 2.5 ± 0.8 mm/s), and also before lick than before no-lick trials, while whisking speed shows no effect of context or lick.
Jaw speed was significantly higher in W+ compared to W- (W+: 3.0 ± 1.0 mm/s, W-: 2.5 ± 0.8 mm/s, t-value = 5.1, p-value = 0.0011) and before Lick compared to No-Lick (Lick: 2.8 ± 0.1 mm/s, No-Lick: 2.6 ± 0.7 mm/s, t-value = −3.7, p-value = 0.0254)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 4; Figure 1J; Figure 1 figure supplement 1HHow it will be judged
As fig1i-pupil-larger-wplus. Speed is the absolute first derivative of the trace times the frame rate (Methods, Figures 1I-J; the code multiplies by 200, plot utils/figure1IJ.py, which holds only for 200 Hz video; the frame rate of each session is used, the fixed 200 is a variant). The lick SD of 0.1 mm/s is far below the others and the t-value's sign is opposite to the stated direction (paper_inconsistency); the tolerance for the lick mean is 10%. The claim stands on the stated directions. Reproduced when the four means are within tolerance, jaw speed is higher in W+ and before lick with corrected p < 0.05, and whisker speed gives corrected p >= 0.05 for both context and lick (the absence judged on the same test).
- fig1k-model-accuracy
A gradient-boosted tree model of the whisker-trial lick choice, from task, history and orofacial features, reaches a cross-validated accuracy of 0.78 on held-out test trials.
Our model reached a cross-validated accuracy of 0.78 on held-out test trials (Figure 1K)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 5; Figure 1KHow it will be judged
As the Methods (Gradient boosted decision tree model) and the code give it. The Methods say "all expert sessions"; the orofacial features exist only for sessions with face video, and the code settles the set. Sessions are the code's 60 sessions of 21 mice (session_context_expert_behaviour_model.yaml), sessions failing the code's 30% missing-value rule on the DLC features skipped, early licks dropped, whisker trials only (process main_analysis/model_context_behaviour.py:34-175); features as in configs/analysis_params/whisker_classification.yaml and explain_behaviour preprocessing/trial_features.py; one session held out (rng seed 20); a stratified 67/33 trial split of the rest (random_state 123); XGBoost tuned by Optuna (TPE, seed 123, 1000 trials, 80/20 inner split, binary log loss, early stopping 100 rounds); balanced accuracy on the 33% test split (explain_behaviour models/behavioural_analysis_GBM.py:80-186). The code's early stopping uses the test split as its evaluation set; this is followed, and the variant with early stopping on the inner validation split is shown. Reproduced when the balanced accuracy is within tolerance; the accuracy on the held-out session is reported.
- fig1k-context-strongest-feature
Among the model's features, context gives the largest drop in accuracy when permuted, followed by whether the previous whisker trial was rewarded, the cumulative number of rewards and the index of the whisker trial in the block.
Among all model features, context information was the one leading to the strongest accuracy drop in the permutation test, followed by whether or not the previous whisker trial was rewarded, the cumulative number of rewards, and the index of the whisker trial of the block (Figure 1K)
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 5; Figure 1KHow it will be judged
Permutation importance as the Methods (Figure 1K) and plot utils/figure1KL.py:512-523 give it, each feature shuffled 10,000 times on the test split, the drop in balanced accuracy recorded. Reproduced when context has the largest mean drop and the next three are, in order, previous whisker rewarded, cumulative water, whisker trial in block; partially reproduced when context is first and the other three are in the top five in any order.
- fig1l-shap-whisker-trial-in-block
The SHAP values of the whisker-trial-in-block feature follow the behaviour, with a single-trial change after a context transition and a slower further component.
we examined the SHAP value of the parameter “whisker trial in block”, finding a pattern consistent with the mouse behavior capturing both single trial adaptation after context transition and a slower component
Results, Rapid behavioral adaptation to sensory stimuli in a changing context, paragraph 5; Figure 1LHow it will be judged
SHAP values by the interventional TreeExplainer with 100 background trials from the training set (Methods, Figure 1L; explain_behaviour models/behavioural_analysis_GBM.py:516-534), over train and test trials; whisker trials 5 to 8 of a block relabelled -4 to -1 (plot utils/figure1KL.py:568). The paper gives no measure. Reproduced when, for both transition directions, the mean SHAP value of whisker trial 1 of the new block differs in sign from that of trial -1 of the previous block, in the direction of the new context, and trial 2 lies further in that direction than trial 1; partially reproduced when only the sign change holds.
- supp1c-wplus-blocks-longer
Whisker-rewarded context blocks last longer than whisker-non-rewarded blocks (196 ± 13 against 185 ± 4 s).
Whisker rewarded context blocks were significantly longer than whisker non-rewarded blocks (W+: 196 ± 13 s, W-: 185 ± 4 s, t-value = 4.3, p-value = 2.0 x 10-4)
Figure 1 figure supplement 1CHow it will be judged
Block duration from the context epochs of the expert sessions, averaged within a session then over sessions per mouse (the legend; process main_analysis/figure1_supp_analysis.py lines 324-351, plot utils/figure1_supp.py:39-67); a two-sided paired t-test across mice, uncorrected. Reproduced when both means are within the stated SDs, W+ longer, p < 0.05.
- supp1d-faster-response-wplus
Spout contact after the stimulus is slightly faster in W+ than in W-, for auditory (370 ± 9 against 400 ± 9 ms) and whisker trials (392 ± 9 against 418 ± 9 ms).
Average spout contact time in response to auditory and whisker trials in W- and W+ showing slightly faster response in W+ compared to W- for both auditory and whisker trials (Auditory: W+: 370 ± 9 ms, W-: 400 ± 9 ms, t-value = −8.7, p-value = 1.8 x 10-10, Whisker: W+: 392 ± 9 ms, W-: 418 ± 9 ms, t-value = −4, p-value = 6.0 x 10-4)
Figure 1 figure supplement 1DHow it will be judged
Lick time minus stimulus onset on lick trials without early lick (cicada behavior_analysis_plots.py:978-1033), averaged per session then per mouse; a two-sided paired t-test across mice per trial type, Bonferroni x2 (plot utils/figure1_supp.py:128-174; the code labels this panel 1E). The paper's ± of 9 ms for every group may be an SEM; the stated value is the tolerance, and the SD and SEM over mice are both reported. Reproduced when the four means are within tolerance and both tests give corrected p < 0.05 with W+ faster; partially reproduced when the directions and tests hold but a mean is outside tolerance.
- fig2c-catch-little-effect
Optogenetic inactivation of cortical sites has little effect on licking in catch trials in either context.
In catch trials, optogenetic inactivation had little effect on licking behavior in either context (Figure 2C)
Results, Distributed cortical areas are involved in task execution, paragraph 1; Figure 2CHow it will be judged
The opto-map rule of the opening comment, catch trials, 7 VGAT-ChR2 mice. A claim of small effect, with no test or threshold in the paper. Reproduced when, under each threshold and null reading, at most 3 of the 39 sites (under 10%) have an effect in either context, and the largest mean |ΔP| over sites is reported; not reproduced when, under every reading, more do; otherwise blocked as methods_underspecified.
- fig2d-auditory-a1-frontal
In auditory trials, inactivating auditory cortex and frontal premotor regions reduces lick probability, similarly in both contexts.
However, in auditory trials, there was a reduction in lick probability when inactivating auditory cortex and frontal premotor regions, which was similar across contexts (Figure 2D)
Results, Distributed cortical areas are involved in task execution, paragraph 1; Figure 2DHow it will be judged
The opto-map rule, auditory trials. Frontal premotor regions are read as the sites wM1, wM2 and ALM (the Results name these "anterior premotor regions" and the paper's section summary lists wM1/2 and ALM). Reproduced when A1 and at least one of wM1, wM2, ALM have a negative effect in both contexts; partially reproduced when this holds in one context only. Each threshold and null reading is run; the outcome is the one they all give, otherwise blocked as methods_underspecified. Similarity across contexts is reported as the W+ and W- mean ΔP at those sites.
- fig2e-whisker-sensory-premotor-reduction
In whisker trials, inactivating wS1, wS2 and anterior premotor regions (wM1/2, ALM) reduces lick probability in both contexts.
Finally, in whisker trials, we observed that inactivating primary and secondary whisker somatosensory cortices (wS1 and wS2), as well as anterior premotor regions led to a reduction of lick probability across both contexts (Figure 2E)
Results, Distributed cortical areas are involved in task execution, paragraph 1; Figure 2E; Results, paragraph 2How it will be judged
The opto-map rule, whisker trials. Reproduced when wS1, wS2 and at least one of wM1, wM2, ALM have a negative effect in both W+ and W-; partially reproduced when they do in W+ only (where the whisker lick rate is high). Each threshold and null reading is run; the outcome is the one they all give, otherwise blocked as methods_underspecified.
- fig2e-rsc-tjs1-increase-wminus
Inactivating tongue-jaw somatosensory cortex (tjS1) or retrosplenial cortex (RSC) increases lick probability in whisker trials specifically in the W- context.
Interestingly, during whisker trials, inactivation of the tongue–jaw primary somatosensory cortex (tjS1) and the retrosplenial cortex (RSC) increased lick probability with respect to the control condition specifically in W− trials, indicating a context-dependent effect of these two regions (Figure 2E, right panel)
Results, Distributed cortical areas are involved in task execution, paragraph 1; Figure 2E rightHow it will be judged
The opto-map rule, whisker trials. RSC is the site (-1.5, 0.5); the Methods give the RSC coordinates as 1.5 mm posterior to bregma and 0.5 mm lateral (Pharmacological inactivation), and the tracing used "the coordinates defined in our optogenetic and imaging experiments" (Results, paragraph 2 of the Figure 3 section). The code's second "RSC" site (-0.5, 0.5) is a variant. Reproduced when tjS1 and RSC have a positive effect in W- and neither has a positive effect in W+; partially reproduced when one of the two does. Each threshold and null reading is run; the outcome is the one they all give, otherwise blocked as methods_underspecified. The mean ΔP of each in both contexts is reported.
- fig2-no-effect-control-mice
In control mice without ChR2, light at the grid sites causes no significant change in lick probability at any site.
In control mice without ChR2 expression, we did not observe any significant changes in lick probability following optogenetic inactivation at any site (Figure 2 – figure supplement 1C&D)
Results, Distributed cortical areas are involved in task execution, paragraph 1; Figure 2 figure supplement 1C and 1DHow it will be judged
A claim of absence, judged on the paper's test, the d' of the opto-map rule for the 4 control mice (process sessions_Context_sessions_opto_control.yaml), with the null by shuffling all trials as the supplement 1D legend says (reading ii). The paper fixes no threshold; under each threshold reading, reproduced when no site in any trial type and context passes it, and not reproduced when one does. When the thresholds disagree, blocked as methods_underspecified. The largest d' and |ΔP| over sites are reported.
- supp2-muscimol-ws1-rsc
Muscimol in wS1 significantly reduces whisker-trial lick probability in W- and W+, while muscimol in RSC gives a non-significant positive trend in whisker-trial licking in W-; pharmacological inactivation is consistent with the optogenetic results.
Muscimol injection in wS1 significantly reduced lick probability in response to whisker trials in W- and W+ (hierarchical bootstrapping, p-value < 1.0 x 10-5) ... Same as E, but for RSC inactivation, which leads to a non-significant positive trend in lick probability for whisker trials in W- (hierarchical bootstrapping, p-value = 0.059)
Figure 2 figure supplement 1E to 1G; Results, Distributed cortical areas are involved in task execution, paragraph 1How it will be judged
As process main_analysis/figure2_supp_stats.py gives it. Per area, trial type and context, the change in lick probability from the baseline day to the injection day, early licks dropped; the statistic is the muscimol change minus the Ringer change; two-level bootstrap (mice with replacement, then trials within mouse), each drug and day resampled independently, seed 42, a two-sided p of twice the smaller tail; the code's default is 10,000 iterations and its docstring names 100,000 for the paper's run, which is used (a p below 1e-5 needs it). Reproduced when wS1 gives a negative change with p < 0.05 in both contexts (p < 1e-5 is reported against the resolution) and RSC in W- gives a positive change with 0.05 <= p < 0.1; partially reproduced when only the wS1 part holds. fpS1 is reported.
- fig2f-latency-frontal-not-rsc
Lick latency increases during inactivation of frontal areas but does not change during inactivation of RSC or tjS1.
While lick latency increased during inactivation of frontal areas, presumably correlated with decreased motor function, there was no change in lick latency associated with inactivation of RSC or tjS1 (Figure 2F and Figure 2 – figure supplement 1H)
Results, Distributed cortical areas are involved in task execution, paragraph 1; Figure 2FHow it will be judged
As the Methods (Figure 2F) and plot utils/figure2F.py give it. Spout contact time minus stimulus onset on lick trials without early lick, auditory trials of both contexts pooled and whisker trials of W+; per mouse and site the mean minus that of control trials, averaged over mice; d' between the site and control distributions over mice, |mean difference| over the root of the mean of the two variances (ddof 0). Thresholds as the opto-map rule readings. Frontal areas are wM1, wM2 and ALM; RSC is (-1.5, 0.5), as fig2e-rsc-tjs1-increase-wminus. Reproduced when at least one frontal site shows an increase passing the threshold in a trial type, and RSC and tjS1 show no change passing it in either trial type; partially reproduced when one of the two parts holds. Each threshold reading is run; the outcome is the one they all give, otherwise blocked as methods_underspecified.
- fig2cd-rsc-no-effect-catch-auditory
Inhibiting RSC does not change performance in catch or auditory trials, so the effect of RSC inactivation needs the context-sensitive whisker stimulus.
Importantly, inhibition of RSC in catch or auditory trials did not lead to any changes in performance, indicating that the presence of the context-sensitive stimulus is necessary for the effect of RSC inactivation to become apparent.
Results, Optogenetic inhibition drives cortical activity modes predictive of behavior, paragraph 1; Figures 2C and 2DHow it will be judged
A claim of absence, judged on the paper's measure for Figures 2C and 2D, the d' of the opto-map rule of the opening comment, 7 VGAT-ChR2 mice, at the RSC site (-1.5, 0.5) as in fig2e-rsc-tjs1-increase-wminus, catch and auditory trials, W+ and W-. The paper fixes no threshold; under each threshold and null reading, reproduced when RSC has no effect of either sign in any of the four trial type and context groups, and not reproduced when it has one. When the readings disagree, blocked as methods_underspecified. The ΔP and d' of RSC in each group are reported, and the site (-0.5, 0.5) is shown as a variant.
- fig3c-jaw-onset
Jaw opening, the first movement, begins 108 ± 24 ms after the sensory stimulus, so the first 120 ms after the stimulus largely precede movement.
We focused our analyses on the first 120 ms after sensory stimuli, largely preceding the onset of jaw opening (108 ± 24 ms) (Figures 3B&C)
Results, Context-dependent flow of activity during sensorimotor transformation, paragraph 1; Figures 3B and 3CHow it will be judged
The jRGECO1a widefield mice, correct trials, jaw opening traces filtered at 60% likelihood and centered on the 250 ms before the stimulus (Methods, Figure 3B). The onset rule of the Methods and the code differ, and the Methods decide (protocol, step 2), on each mouse-average trace the first time after the stimulus at which the d' between jaw opening values and the pre-stimulus 250 ms is >= 1 (Methods, Figure 3C). The code's rule, per trial the first time after the stimulus at which |jaw| over its pre-stimulus SD exceeds 3, averaged per session, then mouse (plot utils/figure3_supp.py:602-674, criterion 3 in figure_files/figure3.py:34), is a variant. Which trial types are pooled is open in the text; the code uses auditory and whisker hits. Reproduced when the mean onset is within tolerance (the ± is the paper's, SEM or CI width in the code).
- fig3e-auditory-similar-across-contexts
Auditory-evoked cortical activity is similar in W+ and W-, beginning in A1, then RSC, then medial secondary motor regions (wM1/2), then orofacial motor cortices (tjS1, ALM).
We observed comparable patterns of cortical activity in auditory trials across W+ and W-contexts, which was dominated by an early increase in fluorescence in the primary auditory cortex (A1), followed by RSC (and nearby regions), and, subsequently, activity propagated anteriorly towards a medial part of secondary motor region (including wM1/2), before transitioning to the orofacial motor cortices (including tjS1 and ALM) (Figures 3D&E)
Results, Context-dependent flow of activity during sensorimotor transformation, paragraph 1; Figure 3E; Figure 3G rightHow it will be judged
The widefield conventions of the opening comment, auditory hits, jRGECO1a mice. The paper gives no measure of similarity or order. Similarity, no ROI reaches |d'| > 2 between W+ and W- from 20 to 120 ms (the paper's Figure 3H threshold and window). Order, readings (a) the time of the peak of the mouse-average trace within 0 to 120 ms, (b) the first time the trace exceeds half that peak. Reproduced when similarity holds and A1 comes before RSC, RSC before wM1 or wM2, and these before tjS1 and ALM, in both contexts; partially reproduced when similarity holds and A1 is first. Each order and baseline reading is run; the outcome is the one they all give, otherwise blocked as methods_underspecified.
- fig3f-whisker-divergence-after-50ms
In whisker trials, wS1 and wS2 activate similarly in both contexts early after the stimulus, and from about 50 ms activity diverges, with more activity in RSC and wM1/2 in W+ than in W-.
Early after stimulus onset, whisker deflection evoked similar activation of primary and secondary whisker somatosensory cortices (wS1 and wS2) in both W+ and W− contexts. However, after ∼50 ms post-stimulus, cortical dynamics began to diverge, with increased activity in RSC and wM1/2 in the W+ context compared to the W-context.
Results, Context-dependent flow of activity during sensorimotor transformation, paragraph 1; Figure 3FHow it will be judged
The widefield conventions, W+ hits against W- misses, jRGECO1a mice. Reproduced when wS1 and wS2 have |d'| <= 2 from 20 to 50 ms, and RSC and wM1 (or wM2) have d' > 2, W+ higher, at some time between 50 and 120 ms; partially reproduced when only one of RSC and wM1/2 does. The d' threshold and window are those of the Methods (Figure 3H). Each baseline reading is run; the outcome is the one they all give, otherwise methods_underspecified.
- fig3h-rsc-first-context-discrimination
RSC is the first cortical area to discriminate the contexts in whisker trials, both in the pixel-wise maps and in the ROI averages, the time of the first d' above 2 between W+ and W-.
This analysis, performed both on pixel-wise activity maps and on the average fluorescence of target regions, revealed RSC as the first area to show significant context discriminability (Figure 3H)
Results, Context-dependent flow of activity during sensorimotor transformation, paragraph 1; Figure 3HHow it will be judged
As the Methods (Figure 3H) and plot utils/figure3H.py give it. Per ROI (wS1, wS2, wM1, RSC, ALM, tjS1, as the code ranks) and per pixel, the first time at which the signed W+ minus W- d' across the 6 jRGECO1a mice exceeds 2, the first 20 ms after the stimulus excluded (the code keeps the 20 ms frame; frames from 30 ms is the Methods' reading, 20 ms the code's, both run). For pixels, the earliest crossing among the pixels within each ROI's 1 mm circle, after the code's Gaussian smoothing of sigma 1 pixel (plot utils/misc/plot_average_widefield_timecourse.py:60-63), and without it as a variant. Reproduced when RSC has the earliest time in both the ROI ranking and the pixel version, under every baseline and every exclusion reading (the outcome is the one they all give, otherwise methods_underspecified); a tie with another ROI at the first time counts as partially reproduced, and so does RSC first in one version only.
- fig3-gcamp-consistent-controls-flat
GCaMP6f mice show whisker-evoked dynamics consistent with jRGECO1a mice, and calcium-insensitive tdTomato and GFP control mice show none of these dynamics.
Importantly, these dynamics were not observed in control mice expressing tdTomato or GFP (Figure 3 – figure supplement 1A&B)
Results, Context-dependent flow of activity during sensorimotor transformation, paragraph 1; Figure 3 figure supplement 1A and 1BHow it will be judged
The widefield conventions per mouse line (GCaMP6f 8 mice, tdTomato 4, GFP 6, from the code's lists). The paper gives no measure. An evoked response at an ROI is read as the Methods' rule for Figure 3C applied to ΔF/F0, a d' >= 1 across mice between the post-stimulus value and the pre-stimulus baseline at some time from 20 to 120 ms. Reproduced when GCaMP6f mice show a whisker-evoked response in wS1 and W+ greater than W- in RSC or wM1/2 (d' > 2, as fig3f-whisker-divergence-after-50ms), and tdTomato and GFP mice show no evoked response in wS1 and no ROI with |d'| > 2 between contexts; partially reproduced when only the controls part or only the GCaMP6f part holds.
- supp3c-rsc-projects-to-wm
Axons of RSC neurons project strongly and directly to frontal motor regions (wM1/2).
finding strong direct connectivity between RSC and frontal motor regions such as secondary motor cortex (wM1/2), consistent with previous anatomical studies
Results, Context-dependent flow of activity during sensorimotor transformation, paragraph 2; Figure 3 figure supplement 1CHow it will be judged
Light-sheet images of cleared brains after AAV-Chronos-GFP injection into RSC, registered to the Allen atlas. Reproduced when GFP-labelled axons are present in M2 at the wM1/2 coordinates in the injected mice the dataset holds. The paper reports no number of mice or quantification, so the claim is judged by presence.
- fig4b-wminus-higher-correlation-correct
Before correct trials, pre-stimulus correlations are higher in W- than W+ among orofacial regions, between whisker motor and orofacial regions, between wS1 and wS2, and between wS2 and wM1/2.
We observed higher correlation values among orofacial regions, and between whisker motor and orofacial regions in W- compared to W+ in the baseline preceding correct trials. Similarly, there was also an increase in correlation between wS1 and wS2 and between wS2 and wM1/2 in W-.
Results, Inter-area functional connectivity is modulated by context and upcoming choice, paragraph 1; Figure 4B; Figure 4 figure supplement 1CHow it will be judged
As the Methods (Seed correlations; Figures 4A&B) and the code give it (process main_analysis/pixel_correlation_analysis.py, pixel_correlation_processing.py; plot utils/figure4A_B.py). jRGECO1a widefield mice; the 2 s before each trial start; seeds wS1, wS2, wM1, RSC, ALM, tjM1, tjS1 (1 mm circles); per trial the zero-lag Pearson correlation of seed and pixel; the null by 1,000 shuffles of block identity; averages over trials per session, context and correct or incorrect, then over sessions per mouse; grid-space values of the masked maps (>= 1.8 SD of the null). The pairs are the code's fixed list (figure4A_B.py:494-495); orofacial pairs ALM-tjM1, ALM-tjS1, tjM1-tjS1; whisker motor to orofacial wM1-ALM, wM1-tjM1; and wS1-wS2, wS2-wM1. The paper's comparison has two forms; readings (a) the Figure 4B line rule, W- minus W+ with paired d' >= 0.8 (Methods, Figures 4A&B), (b) the related t-test across mice with Bonferroni correction (the code corrects over the 7 targets, figure4A_B.py:656-685), p < 0.05. Reproduced when W- is higher for all seven pairs; partially reproduced when it is for at least four, including one of each group the quote names. Both readings are run; the outcome is the one they both give, otherwise blocked as methods_underspecified.
- fig4b-no-context-modulation-incorrect
Before incorrect trials the contextual modulation of correlations is absent, because correlations increase in incorrect W+ trials.
Such contextual modulation of correlation was absent in baseline periods preceding an incorrect trial. This was due to an increase in correlation in incorrect trials in the W+ context
Results, Inter-area functional connectivity is modulated by context and upcoming choice, paragraph 1; Figure 4BHow it will be judged
A claim of absence, on the same pairs and methods as fig4b-wminus-higher-correlation-correct, before incorrect trials. The Figure 4B rule is the paper's test (lines with d' >= 0.8 drawn in colour, the others grey; Methods, Figures 4A&B); the related t-test with Bonferroni correction is the variant. Reproduced when no pair has a W- minus W+ d' >= 0.8 before incorrect trials, and the W+ incorrect minus W+ correct difference is positive for most of the seven pairs; not reproduced when a pair passes the d' rule.
- fig4b-connected-network
The pre-stimulus network of significantly correlated seeds connects whisker sensory regions, tongue-jaw orofacial regions and RSC, with wM1/2 as intermediary node, and pairs such as wS1 and wS2 or ALM and tjM1 are connected in both contexts for correct and incorrect trials.
We found a pattern of significant connectivity between whisker sensory regions, tongue-jaw orofacial regions and RSC with wM1/2 acting as intermediary node in the spontaneous pre-stimulus activity.
Results, Inter-area functional connectivity is modulated by context and upcoming choice, paragraph 1; Figure 4B; Figure 4 figure supplement 1BHow it will be judged
As fig4b-wminus-higher-correlation-correct; a pair is connected when its grand-mean correlation is >= 1.8 SD of the null (Methods, Figures 4A&B; figure4A_B.py:100-101, 469). Reproduced when wS1-wS2 and ALM-tjM1 are connected in all four context and outcome groups and, in each group, wM1 is connected to RSC and to at least one whisker sensory and one tongue-jaw region; partially reproduced when only the first part holds.
- fig4c-inhibition-propagates-forward
Inhibiting wS1 decreases activity widely across cortex, inhibiting wM1/2 decreases activity among motor areas while leaving whisker sensory responses intact, and inhibiting RSC does not increase activity in wM1/2 or tongue-jaw motor regions.
For instance, inhibition of wS1 (which leads to a strong decrease in lick probability) resulted in widespread decrease in cortical activity, whereas inhibiting wM1/2, which also generated strong decreases in lick probability, concentrated the downstream inhibitory effect among motor areas while leaving the whisker sensory responses intact.
Results, Optogenetic inhibition drives cortical activity modes predictive of behavior, paragraph 1; Figure 4CHow it will be judged
The 4 widefield-with-optogenetics mice, whisker trials in W+ (Figure 4C legend), grid ROI ΔF/F0 averaged per session, context and site, then per mouse, then over mice (Methods, Figure 4C), compared with control trials (laser off the brain). The paper gives no measure. Peak ΔF/F0 from 50 to 120 ms (the Figure 3G window) per ROI; a decrease is the site value below control by more than the SD of control over mice. Reproduced when wS1 inhibition decreases a majority of ROIs, wM1 (or wM2) inhibition decreases wM1/2, ALM or tjM1 but not wS1 or wS2, and RSC inhibition does not raise wM1, wM2, ALM, tjM1 or tjS1 above control by more than that SD; partially reproduced when two of the three parts hold.
- fig4d-trajectories-follow-lick
Projected onto PC3, inactivation of wS1, wS2, wM1/2 and ALM drives the trajectory away from the control-lick trajectory toward no-lick, while inhibition of RSC and tjS1 gives trajectories that follow the control-lick trajectory in both contexts.
For regions of interest that decreased task performance (wS1, wS2, wM1/2 and ALM), inactivation strongly deviated the trajectories away from the control-lick trajectory, towards no-lick. Only inhibition of RSC and tjS1, regions which led to increased licking, had trajectories that followed that of the control-lick in both contexts.
Results, Optogenetic inhibition drives cortical activity modes predictive of behavior, paragraph 1; Figure 4DHow it will be judged
As the Methods (Principal component analysis; Figure 4D) and plot utils/figure4C_G.py:434-535 give it. Grid ROI traces of control trials, whisker and catch trials, averaged per mouse then over mice by context, trial type, lick and time, from -150 to 150 ms (the Methods' 300 ms window); StandardScaler then PCA with 15 components, signs set as the code sets them; trials at each stimulated site projected per mouse with all trials pooled. Closeness is measured on PC3 from 0 to 150 ms as the mean squared distance to the control-lick and to the control-no-lick trajectories. Reproduced when, in both contexts, wS1, wS2, wM1/2 and ALM are closer to no-lick and RSC and tjS1 closer to lick; partially reproduced when RSC and tjS1 are closer to lick in both contexts and at least two of the others closer to no-lick.
- fig4e-rsc-tjs1-smaller-angle
The PC3 trajectories under RSC and tjS1 inhibition make smaller angles with the control-lick trajectory than those under inhibition of the other regions, which decreased lick probability.
RSC and tjS1 – areas that increased licking in W- – had smaller angles with respect to the lick trajectories compared to the other regions, which decreased lick probability in all contexts (Figure 4E)
Results, Optogenetic inhibition drives cortical activity modes predictive of behavior, paragraph 1; Figure 4EHow it will be judged
The angle as the Methods (Figures 4E-G) give it, the arccos of the normalised dot product of the per-mouse stimulated-site PC3 trace and the grand-average control-lick PC3 trace, in degrees, over the post-stimulus samples of the window (plot utils/figure4C_G.py:214-237). The code first adds the minimum of the stimulated trace to both traces, which the Methods do not describe; the Methods' angle decides and the code's is a variant. Reproduced when, in each context, the mean angles over mice of RSC and tjS1 are both smaller than those of wS1, wS2, wM1 (or wM2) and ALM; partially reproduced when this holds in one context.
- fig4fg-angle-predicts-lick-change
Across all inactivated sites, in W+ and in W-, the angle between the inhibition and control-lick trajectories relates to the change in lick probability the inhibition evokes.
This relationship held generally true across all inactivated regions in W+ and W-contexts (Figures 4F&G)
Results, Optogenetic inhibition drives cortical activity modes predictive of behavior, paragraph 1; Figures 4F and 4GHow it will be judged
As the Methods (Figures 4E-G) and plot utils/figure4C_G.py:318-341 give it. Per context, a linear regression (scipy linregress, Pearson) over sites of the mean angle over mice against the mean whisker-trial ΔP(Lick) of the same 4 mice from the opto-map rule. Sites are paired by coordinate; the code pairs two tables by sorted position, which may mis-pair them, and is shown as a variant. Reproduced when the correlation is negative (smaller angles with larger ΔP) with p < 0.05 in both contexts; partially reproduced when in one context, or when negative in both without p < 0.05.
- supp4d-photoactivation-control
In mice expressing jRGECO1a without ChR2, the laser gives only a local photoactivation effect and none of the widespread inhibition seen in ChR2 mice.
This control experiment shows a locally restricted jRGECO1a photo-activation effect (jRGECO1a is based on mApple, which shows photoconversion upon blue light stimulation), but no widespread inhibition as observed in mice expressing the ChR2 opsin in inhibitory neurons
Figure 4 figure supplement 1D; Results, Optogenetic inhibition drives cortical activity modes predictive of behavior, paragraph 1How it will be judged
The photoactivation-control mice (2 mice, 6 sessions in the code's list), laser at wS1 and wM1/2, whisker trials, analysed as fig4c-inhibition-propagates-forward. Reproduced when the stimulated ROI rises above control and fewer than half of the other ROIs fall below control by more than the control SD, for both sites.
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none