Claims 33
From the paper 10.1073/pnas.2514107123 · claims file · checked 2026-10-21: ready
- table-s1-recordings
The recordings comprise 13 ephys sessions and 6 behaviour-only sessions, one per animal, with the numbers of accepted single units and of trials per animal listed in Table S1, and with the units and OL turn trials of the example sessions given in Fig. 1J-K and Fig. S4.
How it will be judged
Unit counts are single units that passed the acceptance criteria (Table S1 footnote; SI Spike sorting and unit filtering) and must match exactly, as must the OL_L plus OL_R counts of Fig. 1J-K and Fig. S4, which the legends give per turn type (489 266 and 265; 503 305 and 304; 528 95 and 95; 530 402 and 403; 482 350 and 350; 480 185 and 190; 517 270 and 268); the per-type counts are compared as well. Table S1 trials are "Total number of trials including CL straight, randomzied left and right turns" and are matched within 10%. Reproduced when every unit count and every OL count matches and at least 17 of the 19 total trial counts are within tolerance; partially reproduced when the unit counts match for at least 10 of the 13 ephys animals. The notes list each mismatch and the inconsistencies the header names.
- fig1lm-staged-modulation
In the C2 principal barrel column, population spiking across all layers is strongly and synchronously modulated during OL turns, in five consecutive epochs (stable spiking, a sharp peak, suppression, a ramp for contralateral turns, and a return to stable spiking).
Spiking activity across all cortical layers is strongly modulated during navigation decisions and can roughly be divided into five consecutive epochs
Results, Population dynamics in C2 barrel column..., paragraph 2; Fig. 1J-OHow it will be judged
Animal 489, all units, population-averaged spike rate (header) over the 266 OL_L trials and, separately, the 265 OL_R trials. Reproduced when, for OL_L, the rate shows in order a peak between 0 and 0.065 s above the epoch 1 mean (the paper gives only t < 0; readings -1 to 0 s and -0.5 to 0 s, as in fig1n-epoch-rates), a minimum between 0.065 and 0.14 s below the epoch 1 mean, a maximum between 0.12 and 0.3 s above the epoch 1 mean, and a mean rate from 0.5 to 1.8 s that is below that maximum; and, for OL_R, the same epoch 2 peak and epoch 3 suppression. Each "above" or "below" is judged by a two-sided Wilcoxon signed-rank test over units at p < 0.05 on per-unit trial-averaged rates in the two windows (the paper gives no test; the windows are the paper's, the test is this file's). Partially reproduced when the OL_L sequence holds but one OL_R part fails.
- fig1n-epoch-rates
For animal 489, the population rate is about 2 Hz during straight running before the turn, peaks sharply at t = 0.026 s (epoch 2), drops to about 0.7 Hz (epoch 3) for both turn directions, and, for contralateral turns only, ramps to a maximum at about t = 0.25 s (epoch 4); for ipsilateral turns the suppression lasts to the end of the turn.
the stable spiking (t < 0 s, 2 ± 0.1 Hz, mean ± SD) of epoch 1 during CL straight running is disrupted by a sharp peak we refer as epoch 2 (t = 0.026 s) ... Epoch 3 follows, which exhibits strong suppression of spiking activity down to 0.7 ± 0.1 Hz (mean ± SD) for both left and right turns ... While for ipsilateral turns the suppression lasts until the turn end ... contralateral turns produce a strong ramping activity labeled as epoch 4, (t = 0.25 s
Results, Population dynamics in C2 barrel column..., paragraph 2; Fig. 1NHow it will be judged
Animal 489, population-averaged rate in Hz (header), OL_L and OL_R separately. Tolerances are the paper's stated spread for the rates, one 5 ms bin for the epoch 2 peak time (10% of 0.026 s is smaller than the bin), and 10% for the epoch 4 time. The epoch 1 window is not given beyond t < 0; readings (a) -0.5 to 0 s and (b) -1 to 0 s. Epoch 2 peak is the maximum bin between 0 and 0.065 s; epoch 3 minimum is the minimum bin between 0.065 and 0.14 s (both windows from the Results); the epoch 4 maximum is the maximum bin between 0.12 and 0.5 s of OL_L. The rates must hold for both turn types where the paper says so. Smoothing is not named by the SI; the main reading is unsmoothed 5 ms bins, and because a single-bin minimum or maximum depends on noise, a variant with the rates averaged over three bins is shown. Reproduced when all four values match under some epoch 1 reading and the OL_R rate between 0.14 and 1.8 s has no maximum above the epoch 1 mean (suppression to the turn end).
- fig1jk-all-units
All five epochs dominate the activity of every one of the 65 simultaneously recorded units of animal 489, reproducibly across 531 randomized trials with a jitter of about 30 ms.
all five epochs are major features that dominate the activity for all 65 simultaneously recorded units ... highly reproducible across 531 randomized trials with jitter of just 30 ms
Results, Population dynamics in C2 barrel column..., paragraph 2; Fig. 1J-K; Fig. S5How it will be judged
Animal 489, z-scored trial-averaged rate per unit (SI Spike rates, Fig. 1J-K) over OL_L trials. The paper gives no per-unit test and no definition of jitter; both are open. For the units, readings (a) the unit's trial-averaged rate is lower in the epoch 3 window (0.065 to 0.14 s) than in epoch 1 (-1 to 0 s) and higher at its maximum in 0.12 to 0.3 s than in the epoch 3 window, as signs only; (b) the same two differences tested over trials by a two-sided Wilcoxon signed-rank test at p < 0.05. The unit part is reproduced when at least 90% of the 65 units show both under a reading ("all", with a 10% margin), partially reproduced when a majority do. For jitter, readings (a) the SD over trials of the time of the population epoch 2 peak (maximum of the trial's summed spike count, 5 ms bins, 0 to 0.065 s); (b) the SD over trials of the time of the epoch 4 maximum (0.12 to 0.5 s); jitter is reported under both and holds when either is within 10% of 30 ms. The outcome is the unit part; jitter is reported in the notes. When the unit readings disagree the claim is blocked as methods_underspecified.
- fig1n-ramp-slope
For animal 489 the linear slope of the epoch 4 ramp in contralateral (OL_L) trials is 12.25 (12.25 ± 0.88, mean ± SEM, in Fig. 4B; 12.25 ± 0.76 in Fig. 1N).
This is noticeably close to the linear slope of ramping peak 4 of 12.25 ± 0.88 (mean ± SEM)
Results, Ramping neural activity reflects temporal accumulation..., paragraph 2; Fig. 4B bottom; Discussion, paragraph 2; Fig. 1NHow it will be judged
Animal 489, 266 OL_L trials, population-averaged rate (header), linear regression of rate on time over the ramp. The paper does not give the slope's units, the fit window or the normalisation range, so readings are named. Rate units (a) min-max normalised to 0 and 1 (SI Spike rates, "For comparison with DDM results (Fig.4B ...) ... normalized"), per second; (b) Hz per second (Fig. 1N plots the unnormalised rate and its triangle gives the same slope). Fit window (c) 0.12 to 0.17 s (start and end of linear ramping in Fig. 4F, SI Analysis of VarCE); (d) 0.12 to 0.18 s (Fig. 4G arrows, SI); (e) 0.12 s to the epoch 4 maximum (0.25 s in Fig. 1N). Normalisation range, for (a), the minimum and maximum over 0 to 1 s (the Fig. 1N window) and over 0 to 1.8 s; these coincide unless the extremes fall after 1 s, and both are shown. The SEM is the regression's standard error. The tolerance is the paper's SEM. The claim is reproduced when the slope matches under any reading; the notebook shows all.
- fig1o-trial-end-ramp
At the end of the turn, when the contralateral wall moves back out at t = 1.8 s, the contralateral population shows a peak (2'), a longer suppression with ramp onset at t = 2.0 s, and a slower ramp (4') that peaks at t = 2.58 s with a slope of 2.05, about 7 times smaller than the slope of the first ramp.
Epoch 3′ suppression is significantly longer with the onset of ramping activity at t = 2.0 s (0.21 s delay) suggesting regulation of integration onset to task demand. Ramping in epoch 4′ is much slower reaching maximum at t = 2.58 s (0.78 s delay). The fitted linear slope (2.05 ± 0.05, mean ± SEM, black triangle in Fig. 1O ) is 7× smaller than for the first turn at trial start (12.25 ± 0.76, mean ± SEM, triangle Fig. 1N )
Discussion, paragraph 2; Fig. 1O; Results, Stages of decision-making..., paragraph 2How it will be judged
Animal 489, OL_L population-averaged rate (header) over the last 1.4 s of the trial (Fig. 1O, from about 1.6 to 3.0 s; the window's end is not given, readings end at 3.0 s and at 3.2 s). Epoch 2' peak is the maximum bin from 1.8 to 1.9 s (paper 1.84 s, tolerance two bins); ramp onset is not defined by the paper and takes the two readings of fig4b-ddm-matches-ramp applied between the 2' peak and the 4' maximum (tolerance 10% of the 0.21 s delay); the 4' maximum is the maximum bin from 2.0 to 3.0 s (tolerance 10% of the 0.78 s delay). The slope uses the readings of fig1n-ramp-slope for units, with the fit window from the 4' onset to the 4' maximum, the only window the text gives; the tolerance is the paper's SEM. The ratio uses the first-ramp slope under the same reading. Also reported, from the same section of the Results, the time of the OL_R 2' peak (paper 2.15 s). Reproduced when the times and the slope ratio match under one reading; the absolute slope is reported and judged as well.
- fig1-ramp-contralateral-only
The ramps of epochs 4 and 4' appear only in contralateral (OL_L) turns, while the epoch 2 and 2' peaks appear for both turn directions, showing strong lateralization.
While epochs 2 (at trial start) and 2′ (at trial end) are present for both left and right turns, the ramping in epochs 4 and 4′ is exclusively presented only for contralateral left turn trials indicating extraordinarily strong lateralization
Discussion, paragraph 5; Fig. 1J-O; Fig. 5C-DHow it will be judged
For each ephys animal in Fig. 5C-D (480, 489, 503, 528, 482, 517, 530; Table S1), the population-averaged rate (header) in OL_L and OL_R trials. Epoch 4 size is the mean rate from 0.12 to 0.25 s minus the mean rate from 0.065 to 0.12 s (end of suppression), and epoch 4' size the mean rate from 2.0 to 2.58 s minus that from 1.9 to 2.0 s (windows from the times the paper gives for animal 489). Per animal, the OL_L size is compared with the OL_R size by a two-sided Mann-Whitney test over trials of the per-trial window means (the paper gives no test; this is the test of this file). Reproduced when epoch 4 is larger for OL_L in every animal at p < 0.05 and in at least 5 of 7 for epoch 4'; partially reproduced when epoch 4 holds in at least 5 of 7. Epoch 2 and 2' peaks for both directions are reported.
- figs1-running-speed
Mice run at fast speeds, 21 ± 4 cm/s (mean ± SD).
running at fast speeds (21 ± 4 cm/s, mean ± SD
Results, Ethological decision-making..., paragraph 1; SI Fig. S1A; SI Animal preparationHow it will be judged
Speed is the absolute speed of the sphere, from forward and lateral velocities. The paper cites Fig. S1A, the speed histogram of animal 503 in the 481-trial experiment of Fig. 2A, while Table S1 gives animal 503 a mean running speed of 16.1 cm/s; the population it refers to is open. Readings (a) animal 503, mean over samples of the session's absolute speed; (b) the same restricted to running, mean trial speed above 10 cm/s (Table S1 footnote); (c) across all 19 animals, the mean of the per-animal mean running speeds (Table S1 definition); the per-animal values are also compared with Table S1's mean speeds (10% each). The SI (Animal preparation) gives the same 21 ± 4 cm/s for a mouse on habituation day 3, running above 10 cm/s; reading (d), habituation sessions, applies only if the dataset holds them, and the author says whether it does. Reproduced when the mean lies within 4 cm/s of 21 under any reading.
- fig2a-success-rate
In the unrewarded task, with a correct trial defined as ±1 cm lateral clearance at t = 1 s, animal 503 succeeds in 95% of 414 randomized turn trials (14 failed trajectories).
Defining threshold for correct trials as ±1 cm lateral clearance at t = 1 s, the success rate reaches 95% (14 failed red trajectories in Fig. 2A
Results, Behavioral correlates of decision-making..., paragraph 1; Fig. 2AHow it will be judged
Animal 503, OL_L and OL_R trials, lateral position integrated from the sphere's lateral velocity. Open details and readings. The zero of lateral position is not given; readings (e) relative to the position at t = 0; (f) the readings of fig2b-lateral-spread (Fig. 2B puts the t = 0 mean at 0.14 cm, so these differ little). Which 414 trials (animal 503 has 609 OL trials in Fig. S4A and Fig. S7) is not said; readings (a) all OL trials; (b) the first 414 consecutive OL trials ("414 consecutive randomized left and right turn trials"). What "±1 cm clearance" means is not defined; readings (c) displacement at t = 1 s of at least 1 cm in the turn direction (left for OL_L); (d) absolute displacement of at least 1 cm, the deviation rule of Fig. S1B. Reproduced when the success rate is within 10% of 0.95 under any reading; the number of failed trials is reported for each.
- fig2b-lateral-spread
Lateral positions are tightly confined during the straight CL section (0.14 ± 0.55 cm, mean ± SD, at t = 0) and their distribution widens over time after the turn (t = 1 s and 3 s), about twice as wide at 3 s as at 2 s (SD 2.30 against 1.49).
While the straight CL section produces trajectories tightly confined within 0.14 cm ± 0.55 (mean ± SD) ... distributions of lateral distances after the turn widen over time ... (2.30 SD at 3 s trial time vs. 1.49 SD at 2 s time for Fig. 2 B , Top , and SI Appendix , Fig. S6 B , Top )
Results, Behavioral correlates of decision-making..., paragraph 1; Fig. 2B; Discussion, paragraph 2; Fig. S6BHow it will be judged
Animal 503 (Fig. 2B), and animal 489 for Fig. S6B. Lateral distance is not defined beyond Fig. 2B; since the values at t = 0 are not zero, it is not relative to t = 0. Readings (a) the lateral position as the dataset stores it; (b) lateral velocity integrated from the start of the trial's CL section. The SD at 2 s and 3 s are attributed jointly to Fig. 2B top (animal 503, t = 3 s) and Fig. S6B top (animal 489, t = 2 s); readings (c) 2.30 for animal 503 at 3 s and 1.49 for animal 489 at 2 s; (d) both values for animal 503; (e) both for animal 489. Units are taken as cm. The widening holds when the SD rises from t = 0 to 1 s to 3 s. Reproduced when the t = 0 values and the widening hold and the 2 s and 3 s SDs match under one reading; partially reproduced when only the t = 0 values and the widening hold.
- fig2c-psychometric
The success rate rises steeply as the approaching wall comes closer to the snout, from 8 mm (whisker barely touching) to 1 mm (whisker strongly bent).
The resulting success rate as a function of wall distance ( Fig. 2C ) exhibits a strong increase toward shorter distances
Results, Behavioral correlates of decision-making..., paragraph 1; Fig. 2C; Fig. S6E-GHow it will be judged
Behaviour-only psychometric animals of Table S1 (474, 475, 499, 518); the text says n = 7 and the legend n = 5, so any other animal whose session varies the OL wall distance is also used and listed. Per animal and turn direction, success fraction (correct as in fig2a-success-rate, both readings of "±1 cm clearance") at each wall distance with at least 20 trials (Fig. 2C legend). Reproduced when the population average (mean over animals) has a negative Spearman correlation with distance and the average at the two shortest distances exceeds that at the two longest in at least 3 of 4 animals; the paper reports no test and this rule is this file's.
- fig2d-turn-angle
Turns are steeper (larger mean turn angle) when the wall comes closer.
Better threshold clearance is achieved by steeper turn angles ( Fig. 2D ) at closer wall distances
Results, Behavioral correlates of decision-making..., paragraph 1; Fig. 2DHow it will be judged
Same animals and trials as fig2c-psychometric. Turn angle is the run angle of Eq. 1, theta = atan(D_lat/D_for), positive for left turns (SI Behavioral tactile virtual reality rig); turns to the right are sign-flipped. How the "mean turn angle" of a trial is taken is not given; readings (a) the mean of theta over the turn, 0 to 1.8 s; (b) the angle of the displacement from t = 0 to t = 1 s, atan of lateral over forward displacement. Reproduced when the population-average angle has a negative Spearman correlation with wall distance under either reading.
- fig2f-whisking-params
During straight CL running the C2 whisker whisks at 15 ± 2 Hz with an amplitude of 39 ± 5 degrees around a set point of 58 ± 5 degrees; once the wall is close (t > 0.2 s) the whisker angle is clamped at 68 ± 12 degrees.
Whisking during CL straight run (t < 0) is executed at 15 ± 2 Hz (mean ± SD) with an amplitude of just 39 ± 5° (mean ± SD) around a set point of 58 ± 5° (mean ± SD) ... (whisker angle clamped at 68° ± 12 (mean ± SD) for t > 0.2 s
Results, Behavioral correlates of decision-making..., paragraph 2; Fig. 2E-FHow it will be judged
Contralateral (right) C2 whisker angle, 0 degrees perpendicular to the head's AP axis, from the two tracked points nearest the base (SI Whisker tracking), in OL_L trials of the whisker-tracking animal of Fig. 2E-F (573 per Table S1; 589 also has whisker tracking and is shown as a variant). Phase and amplitude from the Hilbert transform (SI Multiparameter GLM); set point is not defined, readings (a) the slowly varying mean of the angle (mean per whisk cycle); (b) the mean angle over t < 0. Frequency is the inverse cycle duration. Window for the CL values is t < 0 within the trial (-1 to 0 s, open); clamped angle is the mean angle for t from 0.2 to 1.8 s. Tolerances are the paper's SD. Reproduced when all four means fall within tolerance under one set-point reading.
- fig2i-ft-rt
The choice period lasts 0.32 s, between the mean first whisker touch of the approaching wall (0.06 ± 0.03 s) and the mean reaction time defined by the stereotypic hind-leg push (0.38 ± 0.07 s).
Corresponding time histograms for FT and RT ( Fig. 2I ) indicate the choice period of 0.32 s between the mean FT (0.06 ± 0.03 s, mean ± SD) and mean RT (0.38 ± 0.07 s, mean ± SD)
Results, Behavioral correlates of decision-making..., paragraph 3; Fig. 2IHow it will be judged
Animal of Fig. 2I. Table S1 assigns Fig. 2I to both 573 ("2EFI") and 480 ("2IL"); readings (g) 573, (h) 480, (i) both pooled; OL_L trials. FT is the first whisker-wall touch after t = 0 to the approaching wall (Fig. 2F; Fig. 3Q); touches of the retracting wall in the first 11 ms (Fig. 3P) are excluded, reading (a), or not, reading (b). RT is the time of the largest peak of the lateral acceleration after t = 0 (SI Fitting of drift diffusion models), in the turn direction (positive for left turns, Fig. 2H). Open details and readings. The signal, (c) the derivative of the run angle theta of Eq. 1, (d) the derivative of the lateral velocity (the SI treats these as the same); the search window, (e) 0 to 1 s, (f) 0 to 1.8 s (the turn). Tolerances are the paper's SD for FT and RT, 10% for the choice period (mean RT minus mean FT). Reproduced when all three match under one combination. RT under the same readings is also reported for every ephys animal, for use by fig4-ddm-fit.
- fig2jk-glm-mixed-tuning
During straight CL running, over 80% of units encode more than one behavioural or sensory variable (whisking kinetics and kinematics, speed, gait), and the wall feedback signal V_wall is among the strongest encoded variables.
In fact, multidimensional tuning is characteristic for the majority of population (over 80% of units, Fig. 2K ) ... Unsurprisingly, among the strongest correlations is the feedback signal V wall
Results, Stages of decision-making..., paragraph 2; Fig. 2J-K; SI Multiparameter GLM regressionHow it will be judged
Animal 579 (Table S1, "2JK"). Poisson GLM with exponential link (SI Eq. 3). The SI gives "13 regressors" and then says V_wall (Eq. 2) was "additionally" included; Fig. S12A shows 13 variables. The SI names whisking phase and amplitude (Hilbert transform of the angle), run angle derivative (lateral acceleration), RH and LH leg push times, forward and absolute speed, run angle and lateral speed; the Results (Fig. 2K) add set angle, curvature and its derivative. These are 12 variables plus V_wall; whether a 13th variable exists, and which, is open, and the fit uses the 12 named plus V_wall. The same 3 s segment of 40 randomly chosen CL trials (SI; the Fig. 2J legend says 55 trials, a reading); spikes in 20 ms bins smoothed with a Savitzky-Golay filter; five-fold cross-validation with ridge regularisation; for each variable, the log likelihood ratio of full against reduced model averaged over the five folds, referred to a chi-squared distribution with 1 degree of freedom. Open details. The Savitzky-Golay window and order, the ridge penalty, the regressor lags or bases, which 3 s segment, and the significance threshold for "encoding" are not given; readings for the threshold are p < 0.05 and p < 0.01, and for the penalty the values a five-fold cross-validated search selects and no penalty. "Among the strongest" is not defined; readings V_wall among the three variables (a) encoded by the most units, (b) with the largest mean log likelihood ratio over units. Reproduced when at least 72% of units encode at least two variables and V_wall is among the strongest under either of these, under all readings; when readings disagree the claim is blocked as methods_underspecified.
- fig2-epoch2-at-first-touch
The epoch 2 peaks appear within 10 ms of the first whisker touch, consistent with the afferent delay from follicle to barrel cortex, which marks them as sensory (mismatch) signals.
In all cases, epoch 2 peaks appear within 10 ms range from FT, which is consistent with the expected afferent delay from whisker follicle to barrel cortex
Results, Stages of decision-making..., paragraph 3; Fig. 1N-O; Fig. 2FHow it will be judged
Needs, in the same session, spikes and whisker touch times. Per ephys session with whisker touch times, the population epoch 2 peak time (maximum bin from 0 to 0.065 s, OL_L and OL_R) is compared with the mean FT of the same trials (FT as in fig2i-ft-rt, readings (a) and (b)), and in a second reading, the peak of the population rate aligned to each trial's FT. Reproduced when the absolute difference is at most 10 ms in every such session under a reading. The header notes that the paper's own numbers (epoch 2 at 0.026 s, FT at 0.06 to 0.07 s) disagree with this; the notes give the difference either way.
- fig2l-running-vs-sitting
The epoch 4 ramp dominates the response only when the animal is actively running; during brief stops with active whisking the epoch 2 peak dominates and the ramp is absent.
peak 4 activity dominates the modulation only when the animal is actively navigating
Results, Stages of decision-making..., paragraph 4; Fig. 2LHow it will be judged
OL_L trials split into sitting, the animal stopped for about 2 s with whisking maintained (Fig. 2L legend, 20 trials), and running (96 trials). The animal is 480 (Table S1, "2IL"); every other ephys animal with at least 10 sitting trials is reported as a variant. Sitting is not defined; readings (a) mean speed below 10 cm/s over -1 to 1 s (Table S1 defines running as above 10 cm/s); (b) mean speed below 10 cm/s over -2 to 0 s. Population-averaged rate (header). Epoch 2 size is the maximum from 0 to 0.065 s and epoch 4 size the maximum from 0.12 to 0.3 s, each minus the epoch 1 mean (-1 to 0 s). Reproduced when, in running trials, epoch 4 exceeds epoch 2, and in sitting trials epoch 2 exceeds epoch 4, in animal 480, under either reading; when readings disagree the claim is blocked as methods_underspecified.
- fig2m-dimensionality-collapse
Population activity is high-dimensional during epoch 1 and collapses to a single dimension after first touch during epochs 2 to 4, staying one-dimensional through the choice period and rising again at decision execution and in epoch 5.
confirms the high-dimensional neural coding during epoch 1 that collapses rapidly to just a single dimension following the FT during epochs 2, 3, and 4 ... the neural population remains one-dimensional ( 46 ) during choice period between FT and RT and starts to increase only at the onset of decision execution and later, in epoch 5
Results, Stages of decision-making..., paragraph 5; Fig. 2M; SI Dimensionality reductionHow it will be judged
Animal 489 (Table S1, "2M"), OL_L trials. As the SI Dimensionality reduction gives it, spikes in 1 ms bins, smoothed with a 2 ms Gaussian kernel, units below 1 Hz removed, cut in 100 ms segments; for each segment, the matrix of trial-averaged rates (units by 1 ms time points) is analysed by PCA and the number of components that explain 90% of the variance is its dimensionality. Open details and readings. The kernel width as (a) SD of 2 ms or (b) FWHM of 2 ms; the 1 Hz rate taken (c) over the whole session or (d) over the OL_L trial window; PCA on time points as samples and units as variables, centred per unit (the standard method of the cited references). Reproduced when, under all readings, dimensionality is 1 in every segment from 0.1 to 0.3 s and greater than 1 in the epoch 1 segments (-1 to 0 s), and rises above 1 in a segment after 0.4 s; when readings disagree the claim is blocked as methods_underspecified.
- figs12-tuning-lost-during-decision
Fitted separately to CL straight running and to OL turn decisions, the GLM shows that most units lose during the decision the significant tuning to sensory and behavioural variables they show during straight running (epoch 1).
When analyzed with reduced GLM (examples in SI Appendix , Fig. S12 B and C ), the majority of units show no significant tuning for both sensory and behavioral variables during decision that they previously exhibited during epoch 1
Results, Stages of decision-making..., paragraph 5; SI Fig. S12B-CHow it will be judged
The GLM, likelihood ratio test and readings of fig2jk-glm-mixed-tuning, fitted separately to CL straight periods and to OL turn decisions, 100 trials per condition with a 2 s segment from each trial (Fig. S12B legend). Open details and readings. The animal; Table S1 lists Fig. S12 under 589, a behaviour-only animal with no units, and 579, the Fig. 2J-K animal, has 55 trials in all, fewer than 100 per condition; readings (a) 579 with all its CL and OL_L trials, (b) each ephys animal with whisker tracking and at least 100 CL and 100 OL_L trials. Where in each trial the 2 s segment lies is not said; readings (c) OL from t = 0 to 2 s and CL the 2 s straight section before an OL trial's t = 0, (d) OL as in (c) and CL the first 2 s of CL straight trials. Turn direction, OL_L trials (contralateral), and OL_R as a variant. Among units encoding at least one variable in the CL fit, the claim holds when more than half encode none of those variables in the OL fit. The paper gives only examples and no count; the majority rule is the text's "majority of units". Reproduced when it holds under the GLM readings in the animal of reading (a) or in most animals of reading (b); when GLM readings disagree the claim is blocked as methods_underspecified.
- fig3-wall-kinematics
At turn onset the wall retracts to its maximum 12 mm distance within a mean of 11 ms, so most (>80%) first touches are on the approaching wall; during the epoch 3 suppression the wall approaches from 8.7 to 4.4 mm, and it reaches its final 2 mm position at 200 ms.
The wall retracts very rapidly ( Fig. 3P ), reaching the maximum 12 mm distance at a mean delay as fast as 11 ms. Therefore, the majority (>80%) of first whisker touch (FT) events ( Fig. 3Q ) correspond to whisker palpating approaching ... when the wall is not retracting but fast approaching from 8.7 mm to 4.4 mm ... at a constant speed until the final position of 2 mm is reached at 200 ms
Results, Could the neural activity modulation be explained by simple sensory drive?, paragraphs 2 and 3; Fig. 3A, 3P-QHow it will be judged
Approaching-wall position in OL_L trials of the session of Fig. 3A and 3P (50 trials; animal 589, which Table S1 lists for "3ABCD, 3G-P"; 489 is shown as a variant). The FT distribution of Fig. 3Q is from 489 per Table S1 ("3EFQR") but "from the same trials as in O" (589) per its legend; readings 489 and 589. Time of maximum distance per trial, then its mean. Wall distances at 0.065 and 0.14 s are the mean trace at the ends of the epoch 3 window the Results give. FT on the approaching wall means FT after the time of maximum distance in that trial. The fraction holds when above 0.72 (">80%" with a 10% margin). Reproduced when all values match in at least one of the two sessions; the wall values are expected in all sessions and are reported for each.
- fig3c-touch-statistics
The whisker touches the wall about 48 ± 4% of the time, in touches of median duration 21 ms; touch probability increases during epoch 4, reaching 90% in epoch 5.
The whisker stays in direct contact with the wall about 48 ± 4% of the time ( Fig. 3C ) with median touch duration of 21 ms ... Whisker touch probability ( Fig. 3C ) increases monotonously during epoch 4, as does mean whisker curvature ( Fig. 3D ), reaching 90% in epoch 5
Results, Could the neural activity modulation be explained by simple sensory drive?, paragraph 3; Fig. 3B-D, 3M-OHow it will be judged
Animal 589 (Table S1, "3ABCD"), the 75 OL_L trials of Fig. 3B-C (which 75 is not said; all OL_L trials with touch annotations are used, and the first 75 as a variant). Touch probability is the fraction of trials in contact per 1 ms frame. The 48% is taken during epochs 2 to 3 (0 to 0.14 s, where the paragraph places it; reading a) and over epochs 1 to 3 (reading b); the median duration over all touches in the trials (reading c) and in epochs 2 to 3 (Fig. 3N; reading d); epoch 5 is 0.3 to 1.8 s. The increase in epoch 4 holds when the touch probability averaged over 0.2 to 0.3 s exceeds that over 0.12 to 0.2 s. Reproduced when all hold under one reading.
- fig3ijkl-whisking-across-epochs
Across epochs 1, 2-3 and 4-5 the whisker palpates the wall at a constant 15 ± 2 Hz, with statistically identical amplitudes, set angles and per-cycle curvature maxima.
The whisker palpates the wall at a constant 15 ± 2 Hz frequency ( Fig. 3I ), with statistically identical amplitudes ( Fig. 3J ), and set angles ( Fig. 3K ) ... producing periodically modulated curvature with statistically identical maximum per each cycle ( Fig. 3L )
Results, Could the neural activity modulation be explained by simple sensory drive?, paragraph 3; Fig. 3I-LHow it will be judged
Animal 589 (Table S1, "3G-P"), 25 OL_L trials (Fig. 3I legend; which 25 is not said, all tracked OL_L trials are used and the first 25 as a variant). Per whisk cycle (Hilbert phase), frequency, amplitude, set angle (cycle mean angle) and maximum curvature, grouped into epoch 1 (t < 0), epochs 2-3 (0 to 0.14 s) and epochs 4-5 (0.14 to 1.8 s). This is a claim of absence judged on the paper's test, two-sample t tests between groups (each pair) at p < 0.05. The samples of the test are not given; readings (a) whisk cycles, (b) per-trial means within each group (n = 25 trials). Reproduced when frequency stays within 15 ± 2 Hz in each group and no amplitude, set angle or curvature comparison is significant; not reproduced when they are, which is what the legends' P < 0.0001 report. The notes give each effect size and say whether the legends' result is what was found.
- fig3-not-sensory-drive
While sensory input is monotonous during the turn (constant whisking, wall contact rising steadily), neural activity is highly nonstationary (peak, suppression, ramp, return), so the modulation is not explained by simple sensory drive.
However, while sensory input is monotonous, in stark contrast to the “simple sensory drive” hypothesis, the neural activity is highly nonstationary
Results, Could the neural activity modulation be explained by simple sensory drive?, paragraph 3; Fig. 3How it will be judged
An interpretation resting on the claims it depends on. Reproduced when fig1lm-staged-modulation is reproduced and, in fig3c-touch-statistics, touch probability does not decrease from 0.065 to 0.3 s (mean per 10 ms bin is non-decreasing in its linear trend), so that the epoch 3 drop and epoch 4 rise are not matched by a drop and rise of contact. Whisking constancy (fig3ijkl-whisking-across-epochs) is reported, not required.
- fig4-ddm-fit
A classical drift-diffusion model fitted to the reaction-time distribution of animal 489 gives a nondecision time of 0.12 s, drift 2.91, bound 0.59 and noise 0.25 (BIC -664); a generalized DDM with collapsing bounds and leak fits better (BIC -705), with normalized drift K/B of 11.12.
Fitting classical DDM ( Materials and Methods ) to experimental distribution of RT ( Fig. 4B ) produces best fit (Bayesian information criterion, BIC of −664, Fig. 4C ) for nondecision time (NDT) of 0.12 s ( Fig. 4D ), drift K = 2.91, bound B = 0.59 ( Fig. 4E ), and noise σ = 0.25 ... that produce a better fit (BIC of −705). Normalized drift K/B that defines the time needed to reach the bound in absence of noise and decay is 11.12
Results, Ramping neural activity reflects temporal accumulation..., paragraph 1; Fig. 4B-E; SI Table S2How it will be judged
Animal 489 (Fig. 4B; Table S2), OL_L trials; which trials is not said, readings (e) the 266 OL_L trials of Fig. 4B bottom, (f) the 520 OL_L trials whose RT histogram is Fig. S7C-D. RT as in fig2i-ft-rt, correct and error trials as in fig2a-success-rate, fitted with PyDDM (SI Fitting of drift diffusion models). Models as the Results describe them. DDM, constant drift K, constant bounds at ±B, noise sigma, nondecision time; full DDM, adding uniform variability of the starting point X0 and of the nondecision time; GDDM, the DDM with exponentially collapsing bounds and a leaky integrator. Open details. Table S2's parameter list could not be read here, and the paper does not say which parameters were free, the fitting loss, the time step or the RT window. Readings (a) sigma free with K, B and NDT (as the reported sigma = 0.25 implies), PyDDM's default likelihood loss; (b) sigma fixed at 0.25 and the rest free. The RT readings of fig2i-ft-rt are carried. The K/B value is taken from the GDDM under reading (c), from the DDM under reading (d) (the header notes the reported DDM values give 4.93). Reproduced when NDT is within 10% of 0.12 s and the GDDM BIC is lower than the DDM BIC under some reading, and each other value is reported against its tolerance; partially reproduced when only one of these two holds.
- fig4b-ddm-matches-ramp
The DDM nondecision time (0.12 s) matches the 0.12 s onset of the epoch 4 ramp, and the normalized drift (11.12) is close to the ramp's slope (12.25), so the ramp is consistent with accumulation of evidence to a bound.
appreciable NDT = 0.12 s from both DDM and full DDM ( Fig. 4D ) fits well to the 0.12 s delay of experimentally observed ramping onset at end of epoch 3 for this example animal
Results, Ramping neural activity reflects temporal accumulation..., paragraphs 1 and 2; Fig. 4BHow it will be judged
Animal 489. Ramp onset is the end of the epoch 3 suppression in the OL_L population-averaged rate; it is not defined further, readings (a) the time of the minimum between 0.065 s and the epoch 4 maximum; (b) the last bin before the epoch 4 maximum at or below the epoch 3 minimum plus 10% of the range between them. Reproduced when the ramp onset and the DDM nondecision time from fig4-ddm-fit both lie within 0.012 s of 0.12 s and the fitted K/B lies within the 10% tolerance of the ramp slope of fig1n-ramp-slope under the normalised reading; partially reproduced when only the timing agrees.
- fig5e-slope-across-animals
Across seven animals, the epoch 4 ramp slope is 13.05 ± 4.59 (mean ± SD).
(other examples in SI Appendix , Fig. S9 and Fig. 5E below, n = 7, 13.05 ± 4.59, mean ± SD)
Results, Ramping neural activity reflects temporal accumulation..., paragraph 2; Fig. 5E; Fig. S9How it will be judged
The seven animals of Fig. 5C-E (480, 489, 503, 528 in C2; 482, 517, 530 in neighbouring barrels; Table S1), OL_L trials, slope per animal under the readings of fig1n-ramp-slope, with the fit window per animal from its own ramp onset (readings of fig4b-ddm-matches-ramp) to (a) its onset plus 0.05 s, (b) plus 0.06 s, (c) its epoch 4 maximum (the windows of 489 carried over by duration; the paper gives no per-animal rule). Fig. 5 uses normalised rates (SI Spike rates), so the normalised reading is the main one. Tolerance is the paper's SD. Reproduced when the mean lies within it under some reading.
- fig4f-varce-example-unit
For an example high-rate unit (489101), the Fano factor is 0.74 ± 0.04 in epochs 1 and 2 and drops to 0.29 at the epoch 4 peak (t = 0.23 s) as total variance saturates, and its VarCE rises quickly at the start of epoch 4 (0.12 to 0.17 s).
results in a Fano factor (Fig.S10A, top, black) of 0.74±0.04 (mean ± s.d.). Total variance saturates midway through a strong increase of mean firing rate during ramping that results in a Fano factor dropping down to 0.29 at the peak of epoch 4 at t=0.23s
Results, Ramping neural activity reflects temporal accumulation..., paragraph 3; Fig. 4F; SI Analysis of VarCE, paragraph 2; Fig. S10AHow it will be judged
Unit 489101 of animal 489, 266 OL_L trials, spike counts in 10 ms bins (SI Analysis of VarCE). Fano factor per bin is variance over mean across trials. The epochs 1 and 2 value is the mean ± SD over bins from -1 to 0.065 s (window open; reading -0.5 to 0.065 s also shown). VarCE = variance minus phi times mean, phi the lowest Fano factor in the trial (SI Eq. 4). Reproduced when the three values match and VarCE averaged over 0.15 to 0.17 s exceeds that over 0.10 to 0.12 s. If the dataset's unit identifiers do not include 489101, the unit is the highest-rate RS unit in L5A or L5B of animal 489 (the paper's description, "an outlier with the highest spiking rate"), and the notes say so.
- fig4g-varce-population
Population VarCE increases linearly between decision onset (0.12 s) and decision termination (0.18 s), whether computed over all units, without outliers, or on residuals, and then declines to a minimum at t = 0.23 s when the decision is executed.
When VarCE is calculated from residuals ( 5 , 51 ) ( Materials and Methods ), it is similar ( Fig. 4G , black) showing linear increase between decision onset ( Fig. 4G , up arrow) and decision termination ( Fig. 4G , down arrow) times. VarCE exhibits a fast decline at decision end, reaching minimum at t = 0.23 when the decision is executed
Results, Ramping neural activity reflects temporal accumulation..., paragraph 3; Fig. 4G-I; SI Analysis of VarCEHow it will be judged
Animal 489, 266 OL_L trials, 10 ms bins, per-unit VarCE with phi the unit's lowest Fano factor that keeps VarCE positive in the choice period (SI Eq. 4). Population VarCE is the "union" of unit VarCEs (SI); readings (a) sum over units, (b) mean over units. Three versions as the SI gives them; all units; without the four outliers of Fig. 4I (the four units with the largest summed VarCE over the choice period, taken as 0.06 to 0.38 s, mean FT to mean RT of fig2i-ft-rt); and on residuals (each unit's per-bin trial mean subtracted from its counts). The paper gives no test of the increase; the regression rule below is this file's. Reproduced when, in all three versions under a reading, the linear regression of VarCE on time over 0.12 to 0.18 s has a positive slope with p < 0.05 and the VarCE minimum between 0.18 and 0.4 s lies within 0.023 s of 0.23 s; partially reproduced when the increase holds in all three but the minimum's time does not.
- fig4j-corce
The CorCE matrix, computed on residuals, shows correlation rising with time along the diagonal and decaying with time lag along a row, a pattern that disappears when bins are shuffled, as expected for a diffusion process.
The calculated ( Materials and Methods ) CorCE matrix ( Fig. 4J ) indeed exhibits characteristic increase of correlation as a function of time and, for given time (e.g., along the top row), shows sequential decay as often observed ( 5 , 51 ) during decision-making. This correlation disappears when bins are randomly shuffled
Results, Ramping neural activity reflects temporal accumulation..., paragraph 4; Fig. 4J; Fig. S10CHow it will be judged
Animal 489, OL_L trials, residual spike counts in 10 ms bins (SI Analysis of VarCE and CorCE, SI Eq. 6), covariance of conditional expectation across bins with VarCE on the diagonal, normalised to correlation. The bins of the matrix are not given; readings (a) 0.12 to 0.18 s (the ramp's VarCE window); (b) 0.06 to 0.38 s (the choice period, mean FT to mean RT). How units are combined is not given; readings (c) residuals summed over units per trial; (d) per-unit matrices averaged. The paper gives no test or shuffle count; the trend measures, the 100 shuffles and the ±0.2 bound below are this file's. Reproduced when, under a reading, the diagonal's linear trend is positive, the top row's correlation decreases with lag (negative Spearman correlation), and in 100 shuffles of bin order within trials (seeded) the top-row trend is gone (shuffled mean Spearman within ±0.2 of zero).
- fig4k-streak-index
Trial-by-trial streak index analysis during ramping shows random local spike count fluctuations with close to zero mean (0.06 ± 1.36, mean ± SD), indicating a gradual, not step-like, transition in each trial.
showed randomness of local spike count fluctuations with close to zero mean (0.06 ± 1.36, mean ± SD, Fig. 4K ), indicated gradual
Results, Ramping neural activity reflects temporal accumulation..., paragraph 4; Fig. 4K; SI Streak Index analysisHow it will be judged
Animal 489, OL_L trials. The SI gives 8 successive 5 ms bins, counts compared with the across-trial median per bin, 1 above and 0 below, and the number of 0-1 switches per trial. Open details and readings. The statistic, (a) the switch count as the SI describes it, which cannot have a mean near zero; (b) the z-scored runs statistic of the cited methods (observed number of runs minus its expectation under randomness, over its SD, Wald-Wolfowitz). Placement of the 40 ms window during the ramp, (c) 0.12 to 0.16 s; (d) 0.14 to 0.18 s. Counts, (e) all units summed per trial ("from all layers combined", SI); (f) per unit and trial ("all units across all trials", Fig. 4K legend). Ties with the median, (g) assigned 0; (h) bins at the median left out. Reproduced when mean and SD match under some reading; the notes say which.
- fig5b-epochs-in-subpopulations
All epochs, including the suppression and ramp of epochs 3 and 4, are present in layer 4 FS and RS units and in the units of every cortical layer and of each cell class, so the modulation is not confined to a subpopulation.
All epochs including epochs 3 and 4 are already present in thalamo-recipient Layer 4 populations of fast spiking (FS, putative inhibitory, Fig. 5 B , Top ) and regular spiking (RS, putative excitatory, Fig. 5 B , Top ) units. They also dominate the activity in subpopulations of all units across different cortical laminae
Discussion, paragraph 3; Fig. 5BHow it will be judged
Animal 489, OL_L trials, normalised population rate (SI Spike rates) of each subpopulation; layers and FS/RS (waveform duration threshold 0.4 ms, SI) as the dataset labels them, and FS/RS recomputed from waveform duration when durations are available. In each subpopulation (L4 FS, L4 RS, each layer, all FS, all RS), the epoch sequence of fig1lm-staged-modulation is tested on that subpopulation's units. Reproduced when the sequence holds in L4 FS and L4 RS and in at least 6 of the 7 others; the counts are reported against the legend's.
- fig5cd-other-animals
Similar synchronous modulation epochs are seen in the other animals recorded in the C2 principal column (n = 4) and in animals recorded in neighbouring barrel columns of the same or a different row (n = 3).
Similar synchronous modulation is observed for other animals (n = 4, Fig. 5C ) with recordings from principal barrel column ... similar characteristic modulation epochs were observed in animals where recordings were done in a neighboring barrel column in the same or even a different row (n = 3, Fig. 5D )
Discussion, paragraphs 3 and 4; Fig. 5C-D; Fig. S4How it will be judged
Animals 480, 489, 503, 528 (C2) and 482 (B2), 517 (C3), 530 (C1), per Table S1 and Fig. S4, OL_L and OL_R trials; the test of fig1lm-staged-modulation applied per animal. Reproduced when it holds in at least 3 of the 4 principal and 2 of the 3 neighbouring animals; partially reproduced when it holds in at least half of each group.
- fig5e-principal-vs-neighbor
The delay of the epoch 2 peak, the delay of ramp onset at the end of epoch 3, and the ramp slope do not differ between recordings in the principal C2 column (n = 4) and in neighbouring columns (n = 3).
delay of epoch 2 ( Fig. 5 E , Left ), start of ramping at end of epoch 3 ( Fig. 5 E , Middle ) and the linear slope of ramping activity in epoch 4 ( Fig. 5 E , Right ) shows no statistically significant difference (two-sample t test, P > 0.05) between principal C2 whisker and neighboring barrel columns
Discussion, paragraph 4; Fig. 5EHow it will be judged
Animals as in fig5cd-other-animals, OL_L trials, normalised population rate. Per animal, epoch 2 delay is the time of the maximum bin from 0 to 0.065 s; ramp onset delay under the readings of fig4b-ddm-matches-ramp; slope under the readings of fig5e-slope-across-animals. A claim of absence on the paper's test. Reproduced when a two-sample t test (equal variances, the default reading of "two-sample t test") of the 4 against the 3 per-animal values gives p > 0.05 for all three measures; not reproduced when any gives p < 0.05. The group means, their difference and Welch's test as a variant are reported. When the measure readings give different outcomes the claim is blocked as methods_underspecified.
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none