Claims 29
From the paper Naturalistic behavior and self-generated neural activity predictive of self-correction · claims file · checked 2026-11-05: ready
- fig1c-learning
All rats learned the spatial sequence task over a period of 5 to 19 days, the percentage of consecutive trials conforming to the correct sequence transitions rising over the days analysed.
All rats learned the task over a period of 5-19 days (Figure 1B,C, Figure S1E)
Results, Rat spatial sequence task and change-of-mind behavior, paragraph 4; Figure 1CHow it will be judged
As in the Figure 1C legend and behavior.correctPairSessions, the measure per session is the fraction of transitions between adjacent trials (outer arm on trial t to outer arm on trial t+1) that are transitions of the rat's correct sequence, averaged over the day's sessions weighted by trial count, sessions with fewer than 20 trials left out. The paper gives no test. Reproduced when, for every rat, the value on the day of peak performance is above the value on the first analysed day and above 25% (chance for a random choice among four arms, the dashed line in Figure 1C), and a Spearman correlation of the daily value with day is positive; partially reproduced when this holds for three or four rats.
- fig1e-trial-counts
The dataset analysed holds 2370, 3554, 4628, 3779 and 2087 trials for rats M, E, J, K and L, of which 107, 225, 138, 315 and 309 are COM trials, hundreds of COM trials per subject.
In total, we identified hundreds of COM trials per subject (Figure 1E inset
Figure 1E inset; Results, Rat spatial sequence task and change-of-mind behavior, paragraph 5How it will be judged
Counts read from the Figure 1E inset table. A COM trial, per the Methods (Change-of-mind trial detection and Change of mind time detection), is a trial in which the head enters an outer arm other than the final one beyond the threshold and reverses (moving in the negative direction along the arm for more than 100 ms after binary closing of the 1D position). The code also counts the larger of the by-arm and by-time COM counts and requires at least 0.8 s of position data in the second before the reversal (behavior.get_com_num_animal, findProportion); followed here. Open, named readings for the threshold, (a) 10 cm into the arm, (b) 0.1 of the arm's length; and for the trials counted, (i) every trial of the analysed days, (ii) trials of sessions with at least 20 trials. Reproduced when, under one reading, at least 8 of the 10 counts are within tolerance and every rat has at least 100 COM trials; partially reproduced for 5 to 7 counts.
- fig1d-initial-switch-bias
Early in learning every rat was biased to side-switching transitions, with a ratio of switch-side to same-side transition probabilities greater than 2 (the ratio for a random choice).
Initially animals were biased strongly to switch-side transitions (Figure 1D; ratios were greater than 2 for all animals
Results, Rat spatial sequence task and change-of-mind behavior, paragraph 4; Figure 1DHow it will be judged
Per rat and day, P(switch-side)/P(same-side) from the outer-arm transitions between adjacent trials (behavior.find_behavior_transitions, return_neighbor_nonneighor; arms 1-2 and 3-4 are same-side), sessions weighted by trial count. "Initially" is open; readings (a) the first analysed day, (b) the mean of the first two analysed days. Reproduced when, under one reading, the ratio exceeds 2 in all five rats (a one-sided binomial test of 5 of 5 against 0.5 gives p = 0.031).
- fig1d-switch-bias-declines
As learning progressed the switch-side to same-side ratio fell to about 1.5, with each day in the maze multiplying the bias by an estimated 0.92.
As learning progressed, this ratio decreased to ~1.5 (mixed generalized logistic effect, mean multiplicative estimate of number of days spent in the maze in reducing bias is 0.92, p < 1e-4)
Results, Rat spatial sequence task and change-of-mind behavior, paragraph 4; Figure 1DHow it will be judged
The model is not described in the Methods. Following the paper's other GLMs (Methods, Generalized linear models), a logistic regression over transitions between adjacent trials, the response 1 for switch-side and 0 for same-side, on the day index with one indicator per rat, the coefficient exponentiated. Readings for the day index, (a) days from peak performance, (b) days from the first analysed day (the same slope); for the unit, (i) single transitions, (ii) daily counts as binomial responses (the same fit). The final ratio is read as (a) the mean over rats of the ratio on the day of peak performance, (b) the ratio pooled over all rats' transitions on that day. Reproduced when the day effect is below 1 with p < 1e-4 and within tolerance, and one reading of the final ratio is within tolerance; partially reproduced when the effect is below 1 with p < 0.05 but a value is outside tolerance.
- figs1f-multiple-com
In most COM trials the rats changed their minds once (entering two choice arms), but trials with several COM events (more than two arms entered) occurred in every rat.
In most COM trials, rats changed their minds once (entered 2 choice arms), though trials with multiple COM events (>2 arms entered in a trial) were also seen in all rats
Results, Rat spatial sequence task and change-of-mind behavior, paragraph 5; Figure S1FHow it will be judged
COM trials as in fig1e-trial-counts; the number of COM events per trial is the code's larger of the by-arm and by-time counts. Reproduced when in every rat more than half of COM trials have one COM event and at least one trial has two or more.
- fig1g-com-corrective
COM trials were corrective; the final choice of a COM trial was rewarded more often than the initial choice would have been, in each rat.
the final decision in COM trials resulted in higher reward rates than would have been the case had the animal continued with their initial choice
Results, Rat spatial sequence task and change-of-mind behavior, paragraph 6; Figure 1GHow it will be judged
Per the Figure 1G legend, the daily probability that the initial choice would have been rewarded (it would have been when the initial arm is the next arm of the sequence after the last rewarded arm) and the daily probability that the final choice was rewarded, over COM trials; one value per day. Per rat a two-sided Wilcoxon rank-sum test of the final against the initial daily values. Reproduced when the final values are higher and p < 0.05 in all five rats; partially reproduced for three or four.
- fig1g-com-matches-run-through
After correction, the reward rate on COM trials did not differ from that of temporally nearby run-through trials.
These corrections brought the average performance on COM trials up to the level seen on run-through trials
Results, Rat spatial sequence task and change-of-mind behavior, paragraph 6; Figure 1GHow it will be judged
A claim of absence. Daily reward probability of the final choice on COM trials against that of nearby run-through trials (for each COM trial one trial drawn at random from t-3 to t+3 that is not a COM trial, Methods, Identification of nearby trials), one value per day. The paper names no test for this comparison; the test of the same panel, a two-sided Wilcoxon rank-sum test per rat at p < 0.05, is used. As the random draw can change the result and the paper does not say how it was handled, the test is repeated over 100 draws and judged on each rat's median p-value over the draws, a choice the paper does not make; the share of draws with p < 0.05 is shown. Reproduced when no rat shows a significant difference; not reproduced when any does. The difference of medians is reported per rat either way.
- fig1h-final-choice-less-switching
The initial choice on COM trials was heavily biased to side-switching, while the final choice was much less so, with a switch-side to same-side ratio close to 1.
The initial choice on COM trials was heavily biased to be a side-switch, while the final choice was much less likely to be a side-switch with a ratio of switching to same-side transitions close to 1
Results, Rat spatial sequence task and change-of-mind behavior, paragraph 7; Figure 1HHow it will be judged
As in changeOfMind_byTransition.info2matrix, per COM trial the transition from the previous trial's outer arm to the initial COM arm and to the final arm (first COM only, first trial of a session skipped), counted per day, and the daily ratio P(switch-side)/P(same-side) for initial and final choices. Per rat a two-sided paired t-test over days of initial against final ratio, and two-sided sign tests of the daily ratios against 2. Reproduced when the initial ratio exceeds the final with p < 0.01 in at least four rats, the initial ratio is above 2 by the sign test (p < 0.05) in at least four rats and the final ratio below 2 (p < 0.05) in at least four rats, as the legend reports for rats E, J, K and L; partially reproduced when the paired t-test holds in at least three rats. Days with no same-side transition give an undefined ratio and are dropped, a detail the paper does not settle.
- fig1h-final-ratio-near-one
The final choice of COM trials did not differ from a switch-side to same-side ratio of 1 in any rat.
between final ratio and 1 yield insignificant difference in all rats
Figure 1H legendHow it will be judged
A claim of absence, judged on the paper's test, a two-sided sign test per rat of the daily final-choice ratios against 1 at p < 0.05. Reproduced when no rat differs; not reproduced when any does. The median ratio is reported per rat.
- fig3b-extended-before-stopping
Before stopping, COM trials showed more extended hippocampal representations within the arm than outbound run-through trials over matched arm regions.
We found that there were more extended representations preceding stopping on COM trials than run-through trials (Figure 3B, ranksum test, p < 0.01 for all rats)
Results, Hippocampal representations of alternatives are only expressed after stopping, paragraph 4; Figure 3BHow it will be judged
Per the Figure 3B legend and Methods (Change of mind triggered decoded location), per trial the maximum absolute difference between the decoded MAP position and the head position within the COM arm, at head speed of at least 4 cm/s, from the causal posterior of the state-space decoder; for each COM trial one nearby run-through outbound trial, its distance taken only over an arm range drawn at random from the ranges the COM trials covered (figure4c.return_distance_day). Moments with less than 0.2 of the posterior in the COM arm are left out only in the Figure 3A plots; the readings are (a) all moments, (b) moments with at least 0.2 posterior in the arm. The pre-stopping window is not stated; readings (i) from arm entry to the stopping time, (ii) the 5 s before stopping (the code's control parameter is named "5seconds"). Per rat a rank-sum test, COM greater. Reproduced when, under one reading, p < 0.01 with COM greater in all five rats; partially reproduced for three or four.
- fig3b-distance-predicts-com
Longer decoded distances before stopping predicted that a trial would have an upcoming COM event, with an odds fold change of 1.06 per unit distance.
decoded distances were predictive of whether the trial would have an upcoming COM event (Mixed logistic regression, mean multiplicative effect (95% CI) ... 1.06 (1.05, 1.08), p < 1e-20
Results, Hippocampal representations of alternatives are only expressed after stopping, paragraph 4How it will be judged
The tolerance is the wider side of the 95% CI. Logistic regression of COM trial (1) against matched run-through trial (0) on the per-trial maximum distance of fig3b-extended-before-stopping, with one indicator per rat (figure4c.make_GLM_xy_version2 and do_GLM_version2, which also take a second covariate). Readings, (a) distance alone, (b) distance and the mean speed before stopping (figure4c.make_distance_plot computes this speed for the same trials), distance in cm. Reproduced when, under one reading, the fold change is within tolerance with p < 1e-20; partially reproduced when it is above 1 with p < 0.05.
- fig3c-distant-after-stopping
Representations far from the animal were much more prevalent after stopping than before; before stopping they rarely reached the end of the current arm or other arms.
These more distant representations were much more prevalent following stopping (Figure 3A (b, c), 3C, paired t-test, p < 0.001 for all rats
Results, Hippocampal representations of alternatives are only expressed after stopping, paragraphs 4 and 5; Figure 3CHow it will be judged
As in figure4c.maze_distance, per COM event the maximum 2D maze distance (along the track graph, through the center when the decode is in another arm) between head and decoded MAP position, before and after the stopping time, while the head is within the first 80 cm of the COM arm and at head speed of at least 4 cm/s; events lacking either part dropped. Per rat a two-sided paired t-test of post against pre. Reproduced when post exceeds pre with p < 0.001 in all five rats; partially reproduced for three or four.
- fig3d-distant-vs-inbound
After stopping on COM trials, representations reached more distant locations than when animals were at the same locations without a COM, returning from the outer well to the home arm.
We also found that post-COM representations were more likely to represent distant locations than when animals were in the same location without a COM
Results, Hippocampal representations of alternatives are only expressed after stopping, paragraph 5; Figure 3DHow it will be judged
Post-stopping maximum maze distance of fig3c-distant-after-stopping against the same measure on inbound run-through trials over arm ranges drawn from the COM ranges (Figure 3D legend). The code uses a one-sided rank-sum test, COM greater (figure4c.plot_2D_distance_distribution_post); the two-sided test is shown as a variant. Reproduced when COM is greater with p < 0.0001 in at least four rats; partially reproduced when p < 0.05 in at least four.
- figs6-remote-during-forward-running
Remote content also occurred during normal forward running after the animal had turned around, so the jumps are not due to backing up.
we also observed remote content during normal forward ... running after the animal had turned around but had yet to exit the arm
Results, Hippocampal representations of alternatives are only expressed after stopping, paragraph 6; Figure S6How it will be judged
Remote events detected as in fig4e-next-choice; each assigned to backing out (moving in the negative direction along the arm, as in COM time detection) or forward running (moving toward the center after turning, still in the arm). The paper gives no test. Reproduced when every rat has remote events during forward running; the proportion during backing out and the two event rates (Figure S6) are reported.
- figs7b-no-swr
The non-local activity during COM occurred during theta, not during sharp-wave ripples; COM periods had a lower ripple rate, while trial ends, rewarded or not, had higher rates.
occurred exclusively ... during the hippocampal theta rhythm, which is generally associated with movement
Results, Hippocampal representations of alternatives are only expressed after stopping, paragraph 2; Figure S7BHow it will be judged
Tolerances are the half-widths of the 95% CIs. SWRs detected as in the Methods (SWR detection), only at speed below 4 cm/s. Per trial the ripple rate during the COM event and at the end of a trial, end-of-trial samples drawn to match the number of COM trials (Figure S7B legend); OLS of the rate on indicators for COM, unrewarded end and rewarded end with one indicator per rat. Open, the window, (a) the 2 s around COM time used in Figure S7A and from the outer poke to 2 s after, (b) from stopping to arm exit and from outer poke to leaving the well; and the reference level of the indicators, (i) a constant and per-rat indicators with the three conditions as given, (ii) the rate rescaled (A.U.) to z-scores within rat. If every row is one of the three conditions, the three condition indicators and the rat indicators are collinear; reading (iii) is the fit with all of them by pseudo-inverse, the default of statsmodels OLS, the package the Methods name for their GLMs. Reproduced when, under one reading, the COM coefficient is negative and the two end-of-trial coefficients positive, each with p < 0.05; partially reproduced when the COM rate is below both end-of-trial rates in at least four rats.
- figs3-local-extended-at-reward
Local extended content after stopping over-represents the bin next to the animal and the reward well end of the arm.
This jump to the reward area of the arm is also seen in local extended content (Figure S3)
Results, Representations of remote alternatives during COM are concentrated in reward regions, paragraph 2; Figure S3BHow it will be judged
Tolerances from the wider side of the 95% CIs. The p for the animal bin is printed "p < 10e-24", recorded as 1e-23. Per Figure S3, the MAP of decodes outside the GP confidence interval in the COM arm (posterior in the arm above 0.2), binned in 10 bins, 5 in front of and 5 behind the animal, distances normalized; log(proportion) fit by OLS on bin indicators with one indicator per rat, coefficients exponentiated. Open, the reference bin, (a) the middle bins, (b) the mean of all bins (sum-to-zero coding). Reproduced when, under one reading, both fold changes are above 1 with p < 0.001 and within tolerance; partially reproduced when both are above 1 with p < 0.05.
- figs11b-theta-cycles-correct
The presence of more representations of alternatives, counted in theta cycles, was associated with a correct next choice, a fold change of 1.06 per theta cycle.
Mean estimate, 95% CI, and p value of the fold change effect of number of theta cycle on correct rate is 1.06, (1.01,1.11), p < 0.015
Figure S11B(d); Results, Hippocampal representations of alternatives predict correct choices, paragraph 2How it will be judged
Local extended events as in the Methods (Identification of trials with large decode to body position distance); per day a GP mean and a GP deviation fit (sklearn, RBF kernel of length 5 cm, alpha the squared standard deviation of the day's decode to animal distance) to 1000 points subsampled from about as many run-through trials as COM trials, at head speed above 4 cm/s; MAP outside mean plus or minus 6 times the deviation for more than 30 ms makes an event. Remote events as in fig4e-next-choice. Theta cycles from MUA-derived theta (Methods, Theta LFP, for Figure S11). Per COM event the number of theta cycles holding content; logistic regression of the next choice's correctness on that number with one indicator per rat. The legend of S11B(d) gives one estimate and does not say which content is counted; readings, (a) cycles with either kind (panel c), (b) cycles with local extended and cycles with remote content as two predictors. Tolerance from the wider side of the CI. Reproduced when the fold change is within tolerance and p < 0.05; partially reproduced when above 1 with p < 0.1. The GP subsample and the choice of run-through trials are random; the fit is repeated over 10 seeds and the median reported.
- fig4a-local-extended-correct
COM events with local extended content were followed by a correct choice more often than those without, a pooled fold change of 1.61 in correct rate.
point estimate of fold change on correct rate due to is (a) local extended content is 1.61, 95% CI ... (1.22~2.11), p<0.001
Figure 4A(a); Results, Hippocampal representations of alternatives predict correct choices, paragraph 2How it will be judged
Tolerance of the fold change from the wider side of the CI. Per the Methods (Generalized linear models to measure the effect of representations of alternatives on correctness), event by event, the next choice is the arm entered for more than 5 cm after the COM event and correct when that choice is rewarded; logistic regression of correctness on the presence of local extended content with one indicator per rat, exponentiated. The code (extra_correctness.figure4_correctness) instead takes one entry per COM trial and the reward of the trial's final choice; shown as a variant. Per rat a two-sided Fisher exact test. Reproduced when the pooled fold change is within tolerance with p < 0.001, at least 8 of the 10 counts are within tolerance, and rat L has p < 0.05 by the Fisher test; partially reproduced when the fold change is above 1 with p < 0.05.
- fig4a-remote-correct
COM events with remote content were followed by a correct choice more often than those without, a pooled fold change of 1.50 in correct rate.
(b) remote content is 1.50, 95% CI ... (1.18~1.90), p<0.001
Figure 4A(b)How it will be judged
As fig4a-local-extended-correct with remote content (events as in fig4e-next-choice) as the predictor, and rat J the rat that should reach p < 0.05 by the Fisher test. The same estimate and CI are printed for Figure 4B (either kind of content), which may be a copying slip in the paper; each is tested as printed.
- fig4b-any-alternative-correct
Pooled over all rats, the presence of any representation of alternatives (local extended or remote) on a COM event was associated with a correct next choice, a fold change of 1.50.
We found that the presence of local extended or remote arm was, on average, associated with ... higher likelihoods of correct subsequent choices across animals
Results, Hippocampal representations of alternatives predict correct choices, paragraph 2; Figure 4BHow it will be judged
As fig4a-local-extended-correct with the predictor the presence of either kind of content (Methods, Generalized linear models, Figure 4B). The per-trial reading of the code is a variant. Reproduced when the fold change is within tolerance with p < 0.001; partially reproduced when it is above 1 with p < 0.05.
- figs11a-more-representations-over-days
Across days, COM events engaged an increasing rate of remote representations in all five animals.
COM events engaged an increasing rate of remote content across days in all five animals
Results, Hippocampal representations of alternatives predict correct choices, paragraph 1; Figure S11AHow it will be judged
Tolerance is the CI half-width. The Results speak of the number of representations and the figure axis of the proportion of COM trials with remote content; readings, (a) the daily proportion of COM trials with remote content, (b) the daily mean number of remote events per COM event. OLS of the daily value on the day index with one indicator per rat. Reproduced when, under one reading, the day effect is within tolerance and p < 0.005, and the daily value rises with day (a positive OLS slope per rat, a direction check for the paper's "in all five animals", for which it gives no per-rat test) in all five rats; partially reproduced when the pooled effect is positive with p < 0.05.
- fig4c-home-with-other-content
COM representations often combined the home arm with local extended or remote arm content; events with home representation more often had local extended content and had more remote outer arms.
decoded representations on COM trials often involved a combination of the home arm, where the animal would typically start each trial, ... and either local-extended or remote arm representations
Results, Hippocampal representations of alternatives predict correct choices, paragraph 2; Figure 4CHow it will be judged
Per COM event, whether it has remote content in the home arm, whether it has local extended content, and the number of distinct remote outer arms. Per rat one-sided Fisher tests, (a) on the 2 by 2 table of home against local extended; (b) for the number of remote outer arms the paper does not say how the counts enter a Fisher test; readings, (i) a 2 by 2 table of home against at least one remote outer arm, (ii) home against two or more remote outer arms, (iii) a Fisher-Freeman-Halton test on the 2 by 4 table of counts 0 to 3. Reproduced when, under one reading, all five rats show the effect in the stated direction for both (a) and (b) (binomial p = 0.031), at least three rats reach p < 0.01 in (a) and at least three reach p < 0.05 in (b), as the legend reports; partially reproduced when one of (a) and (b) holds.
- fig4d-one-remote-arm
In COM events with remote content, most often only one of the three other arms was represented.
Most often, only one of the three other arms was represented (Figure 4D)
Results, Hippocampal representations of alternatives predict correct choices, paragraph 3; Figure 4DHow it will be judged
Per COM event with remote outer-arm content, the number of distinct remote outer arms. Reproduced when one is the most frequent count in all five rats; partially reproduced in three or four.
- fig4e-next-choice
Remote arm representations during COM events predicted the immediate next choice (fold change 1.96), but not the last rewarded arm, the last arm chosen or the next correct arm.
We found that remote arm representations were predictive of the immediate next choice (Figure 4E, mixed logistic regression for all rats data, fold change (95% CI), and p-value due to rewarded past, immediate past, correct future, and next choice are 1.07 (0.78,1.50), p < 0.7; 1.16 (0.85,1.59), p < 0.35; 1.31 (0.96, 1.80), p < 0.1; 1.96 (1.45, 2.65), p < 0.0001)
Results, Hippocampal representations of alternatives predict correct choices, paragraph 4; Figure 4EHow it will be judged
Tolerances from the wider side of each CI. Remote events per the Methods (Identification of remote theta events) at head speed above 4 cm/s, MAP in a choice arm other than the current one for more than 20 ms, posterior of the continuous state above 0.5 for more than 20 ms, the decode in one arm segment throughout, the summed posterior in that arm above 0.2 (code parameters dur_0.02_sum_0.2). GLM per the Methods (Generalized linear models to measure the effect of arm category), event by event, four rows per event (arms 1 to 4), response whether the arm holds remote content, binary predictors for last rewarded arm, last arm entered (more than 5 cm before), next arm entered (more than 5 cm after) and next correct arm, one indicator per rat; exponentiated. The sequence used to define the next correct arm is read per rat from the reward log (see the opening comment). Reproduced when the next-choice fold change is within tolerance with p < 0.0001 and the other three have p > 0.05; partially reproduced when the next-choice effect is above 1 with p < 0.05. The event counts are compared with the n printed (they differ from the remote counts of Figure 4A, which the paper does not explain).
- figs11c-next-choice-unthresholded
With no thresholds for remote content, the next choice was still represented longer, by 6.15 ms, than other arm categories.
Linear effect on addition time that a choice is represented (in milliseconds) due to immediate next choice is 6.15, 95% confidence interval ... (2.38~9.93), p value < 0.002
Figure S11C; Results, Hippocampal representations of alternatives predict correct choices, paragraph 4How it will be judged
Tolerance is the CI half-width. Per the Figure S11C legend, all COM events, every 2 ms bin whose MAP is in an outer arm other than the COM arm counted as remote; per event and arm the total time the MAP is in that arm; OLS of that time on the four category predictors of fig4e-next-choice with one indicator per rat. Open, which bins of the event, (a) from stopping to arm exit, (b) the whole time in the COM arm, and whether a speed limit of 4 cm/s applies, (i) yes, (ii) no. Reproduced when, under one reading, the next-choice effect is within tolerance and p < 0.002; partially reproduced when positive with p < 0.05.
- fig5a-remote-at-arm-ends
Remote representations after stopping were concentrated at the ends of the arms, where reward is delivered; the arm end was the bin contributing most for every rat.
The results indicated a strong bias toward representations of the ends of the arms in individual epochs and across epochs for all animals
Results, Representations of remote alternatives during COM are concentrated in reward regions, paragraph 2; Figure 5AHow it will be judged
Tolerance is the CI half-width. Per Figure 5A, the MAP of every 2 ms bin of remote events placed in 5 equal bins of each of the five arms (home and arms 1 to 4, center to arm end). Per rat, as figure4d.GLM_by_animal, OLS of log(proportion + 0.0001) over the 25 arm and bin cells on bin indicators with a constant; the arm-end bin coefficient. Open, the normalization of the proportion, (a) within each arm, (b) over all of the rat's remote bins; and the pooled model (not in the code), OLS on bin indicators with one indicator per rat over all rats' cells. Reproduced when, under one reading, the arm-end coefficient is the largest bin coefficient in all five rats and positive with p < 0.05 in at least four, and the pooled effect is within tolerance with p < 0.0001; partially reproduced when the arm end is the largest bin in at least four rats.
- fig5b-remote-behind
Most remote representations lay at large angles from the head direction, behind the animal; the proportions of events at angles above 110 degrees were 0.67, 0.55 and 0.54 for rats M, J and K.
We found that most of the remote representations were located at large angular deviations from the head corresponding to locations behind the animal
Results, Representations of remote alternatives during COM are concentrated in reward regions, paragraphs 3 and 4; Figure 5BHow it will be judged
Head direction from the extended Kalman filter combining overhead tracking with the headstage gyroscope and accelerometer (Methods, Head direction tracking; trodestrack package). The angle between the head direction and the vector from the animal to the 2D maze location of the remote content. The paper says three rats and shows M, J and K, while the Methods list inertial sensors for M, J, K and L; the claim is tested on M, J and K. Open, the angle per event, (a) the mean over the event's 2 ms bins (circular mean), (b) the angle at the event's onset, (c) every bin counted. Reproduced when, under one reading, all three proportions are within tolerance; partially reproduced when all three exceed 0.5.
- figs8-late-theta-phase
Remote and local extended events were more likely to occur during the late theta phase.
Events were more likely to occur during late theta phase
Figure S8 legendHow it will be judged
Theta phase from MUA (Figure S8 legend; Methods, Theta LFP, peaks of MUA smoothed with a 40 ms Gaussian and found at least 100 ms apart as phase 0 and 2 pi, phase linearly interpolated). Per the Methods (Content / Theta Phase relationship), for each event the window from 60 ms before to 60 ms after onset, while the animal is in a choice arm, in 4 ms bins, each bin's theta phase and its content (for remote events the summed posterior outside the animal's outer arm, for local extended events the decode to animal distance). How the figure turns these into "more likely to occur" is open; readings, (a) the phase at event onset, (b) the content-weighted distribution of bin phases over the window. The paper gives no test; a Rayleigh test (weighted in reading b) is used, and late phase is read as a mean phase between pi and 2 pi. Reproduced when, under one reading, the phases of both kinds of event are non-uniform (p < 0.05) with a mean phase between pi and 2 pi in at least four rats.
- figs10-theta-power-and-cycle-length
After controlling for movement speed, theta cycles with local extended or remote content had lower theta power and longer cycles.
the existence of local extended content or remote content is significantly associated with lower theta power and longer theta cycles
Figure S10D and S10EHow it will be judged
Tolerances are the CI half-widths. Per theta cycle while the animal is in the outer arm during a COM event, z-scored theta power and cycle length, regressed (OLS) on pre-stopping, speed, local extended content and remote content, with one indicator per rat (Figure S10 legend). Theta from the corpus callosum LFP (Methods, Theta LFP, which reserve MUA-derived theta for Figures S9 and S11 only; the S8 legend names S8 and S11); MUA-derived theta is shown as a variant. Open, the z-score, (i) within session, (ii) within rat. Reproduced when, under one reading, the four effects have the stated signs with p < 0.0001 and are within tolerance; partially reproduced when the signs hold with p < 0.05.
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none