Claims 47
From the paper Rhythmic coordination and ensemble dynamics in the hippocampal-prefrontal network during odor-place associative memory and decision making · claims file · checked 2026-10-21: ready
- fig1b-performance
During recordings the rats performed the odor-place task at a high level, 77.0% correct (mean and s.e.m. over the 8 rats, 1.3%).
animals maintained a high level of performance on the task, indicating accurate decision making based on cued recall of odor-place associations ( Figure 1b , n=8 rats, 77.0% ± 1.3%, mean ± s.e.m.)
Results, Odor-place associative memory and decision-making task, paragraph 2; Figure 1bHow it will be judged
Performance is the proportion of correct trials (Methods, Odor-place association task training), per session (run epochs of a recording day), odor-cued sessions only. The code computes it per epoch and averages (F16/Behavior/cs_performanceScatter.m); the Methods do not say whether trials with sampling under 0.5 s count, and the code keeps them in the performance vector (cs_getNosepokeWindow.m lines 186 to 223). Readings (a) mean over rats of each rat's mean session performance, (b) mean over all sessions, each with and without the short-sampling trials. Reproduced when one reading falls within 77.0 +/- 1.3%.
- fig1c-sampling-duration
Odor sampling lasted 0.82 s on average (s.e.m. across sessions 0.02 s; s.d. across trials 0.25 s) and did not differ between correct and incorrect trials.
The average odor-sampling duration before odor port disengagement and odor offset was 0.82±0.02 s (mean ± s.e.m. across sessions), and this duration was similar between correct and incorrect trials ( Figure 1c ; within-session signed-rank test, p=0.71
Results, Odor-place associative memory and decision-making task, paragraph 3; Figure 1c; Figure 1-figure supplement 1aHow it will be judged
Duration is odor port beam-break offset minus onset (Methods). Per session the mean duration on correct and on incorrect trials; a Wilcoxon signed-rank test across the 38 sessions, as the Results and legend name it. Absence claim, reproduced when that test gives p >= 0.05. The code tests the same per-session means with a rank-sum test (F16/Behavior/cs_nosepokeDuration_2.m line 65), shown as a variant. The effect size (median paired difference) is reported.
- fig1d-turn-direction
On most trials the rat turned away from the odor port toward the side of the reward arm it then chose, on correct and incorrect trials alike, so the choice is made during odor sampling.
Notably, we observed that on a majority of trials (95.4% ± 0.12%), the animals' turn direction away from the odor port matched the direction of the T-maze reward arm that they would ultimately choose on that trial
Results, Odor-place associative memory and decision-making task, paragraph 3; Figure 1dHow it will be judged
Sessions on the truncated mazes are excluded (legend). Turn direction as in F16/Behavior/cs_turnFromNPDirection.m, which the Methods do not describe - the 41 position samples from odor port exit, the sample furthest in x from the odor port position, its sign giving left or right, trials within 40 samples of the epoch end dropped. A binomial test of same against opposite with null probability 0.5 (the code's binopdf(...,0.5)), per outcome. The 95.4% (with a spread of 0.12% as printed) is read as (a) the mean over sessions or (b) the pooled fraction of all trials; the counts give 97.6% pooled. Reproduced when both binomial tests give p < 0.05 with more same than opposite turns and the counts fall within tolerance; the fraction is reported under both readings.
- fig1s1e-sniff-rate
Sniff rate measured by the nasal thermocouple rose slightly during odor sampling (7.1 Hz) compared with time-matched pre-odor periods (6.2 Hz).
There was a small but significant increase in sniff rate during the odor sampling period (within-session increase 7.1±0.39 Hz, mean ± s.e.m.) compared to time matched pre-odor periods (6.2±0.29 Hz, mean ± s.e.m.)
Results, Odor-place associative memory and decision-making task, paragraph 3; Figure 1-figure supplement 1eHow it will be judged
Thermocouple signal of the two implanted animals. The Methods do not describe sniff detection; the code (F16/Behavior/cs_sniffRate.m) takes a pre-odor window of the same length as each trial's sampling ending at odor onset, smooths and downsamples by 25, counts peaks (Chronux findpeaks, whose threshold is not in the repository) and divides by window length, averaged per day, with a signed-rank test across days. Peaks are counted here as local maxima of the smoothed signal; a peak rule with a prominence threshold is shown as a variant. Reproduced when the signed-rank test gives p < 0.05 with odor above pre-odor and both rates within tolerance.
- fig1s1f-air-performance
Performance on the odor-cued task was higher than on the air-cued sessions, where no odor cue was given.
Performance of three animals on the odor-cued task (n = 24 sessions) versus the air-cued task (n = 4 sessions) in which animals were presented with only air as a neutral stimulus
Figure 1-figure supplement 1fHow it will be judged
Per session the fraction correct (correct meaning the well the paper's task scores as rewarded), in the animals that ran an air session; a rank-sum test of odor-cued against air-cued sessions (legend; code F16/Behavior/cs_odorVsAirPerf.m line 98). Reproduced when p < 0.05 with odor-cued performance higher. The paper gives three animals in this legend and four in the Methods; both sets are shown.
- fig2d-beta-power-increase
Beta power rose from the time-matched pre-odor period to the odor sampling period in CA1, PFC and OB.
We observed a strong increase in power in the beta band (20–30 Hz) during this decision-making period compared to a time-matched pre-odor period across all three regions
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 1; Figure 2d topHow it will be judged
Beta power (20 to 30 Hz) per region, per session the mean over trials of odor minus pre-odor power; Wilcoxon signed-rank across sessions. Run under LFP readings A and B of the opening comment. Reproduced when all three regions show an increase with p < 0.05; partially reproduced for one or two regions. The outcome is the one both readings give; when they differ the claim is blocked as methods_underspecified with both shown.
- fig2d-rr-power-no-increase
RR power did not rise after odor onset; it fell slightly in CA1 and did not change in PFC or OB.
power increases were seen during the odor sampling period only in the beta band ( Figure 2d signed-rank tests, n=38 sessions, CA1 p=2.84e-3, PFC p=1.80e-4, OB p=8.3e-6) and not in the RR band
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 2; Figure 2d topHow it will be judged
As fig2d-beta-power-increase, for 7 to 8 Hz. The paper's test is the signed-rank test at 0.05. Reproduced when no region shows a significant increase, CA1 shows a significant decrease and PFC and OB show no significant change; the effect sizes are reported. Run under readings A and B; if they disagree, the claim is blocked as methods_underspecified with both shown.
- fig2e-beta-coherence-increase
Beta coherence between all three region pairs rose from the pre-odor period to odor sampling.
This coherence was specific to the odor-sampling period... it significantly increased compared to time-matched pre-odor periods ( Figure 2e top , signed-rank test, n=38 sessions, CA1-PFC p=2.8e-3, CA1-OB p=5.1e-5, PFC-OB p=3.7e-5)
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 3; Figure 2e topHow it will be judged
Beta (20 to 30 Hz) coherence per region pair, per session the mean over trials of odor minus pre-odor coherence; Wilcoxon signed-rank across sessions, under LFP readings A and B of the opening comment. Reproduced when all three pairs increase with p < 0.05; partially reproduced for one or two pairs. The outcome is the one both readings give; when they differ the claim is blocked as methods_underspecified with both shown.
- fig2e-rr-coherence-increase
RR coherence also rose from the pre-odor period to odor sampling for all three region pairs.
Network wide coherence in the RR band also increased from the pre-odor to the odor-sampling period ( Figure 2e top , signed-rank test, CA1-OB p=6.4e-4, CA1-PFC p=3.9e-3, PFC-OB p=2.4e-4)
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 3; Figure 2e topHow it will be judged
As fig2e-beta-coherence-increase, for 7 to 8 Hz.
- fig2e-beta-coherence-odor-vs-reward
Beta coherence in the last 500 ms of odor sampling was higher than in the last 500 ms before reward well exit, for all three region pairs, so it is not a sign of movement preparation.
We confirmed that beta coherence observed during the decision-making period was significantly higher than coherence at the reward well between all three pairs ( Figure 2e middle , signed-rank test, CA1-PFC p=0.010, CA1-OB p=0.0012, PFC-OB p=0.0024)
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 4; Figure 2e middleHow it will be judged
Windows from the Results - the last 500 ms of odor sampling and the last 500 ms before reward well exit, correct trials (rewarded) with sampling over 0.5 s, as F7/ClaireCrossCoherence.m lines 1040 to 1064 do. Reward well exit is the end of the reward-well beam break; the code takes the next DIO event after the pump turns off (F16/Behavior/cs_getRewardTimes.m line 101), shown as a variant. Coherence under LFP reading B (its pipeline) and reading A's settings on the same windows; per session odor minus reward, signed-rank across sessions. Path A's windows centred on exit (F16/LFPAnalysis/cs_cohOdorVsReward.m, 0.5 s either side) are a variant. Reproduced when all three pairs are higher during odor with p < 0.05; partially for one or two. The outcome is the one both settings give; when they differ the claim is blocked as methods_underspecified with both shown.
- fig2e-rr-coherence-odor-vs-reward
For RR, coherence was higher during odor sampling than at the reward well for CA1-OB and PFC-OB but not for CA1-PFC.
Interestingly, this effect in RR was observed for CA1-OB and PFC-OB, but not between CA1-PFC ( Figure 2e middle , signed-rank test CA1-PFC p=0.569, CA1-OB p=0.0044, PFC-OB p=0.00064)
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 4; Figure 2e middleHow it will be judged
As fig2e-beta-coherence-odor-vs-reward, for 7 to 8 Hz. Reproduced when CA1-OB and PFC-OB are higher during odor with p < 0.05 and CA1-PFC gives p >= 0.05 (the paper's test for the absence); partially when two of the three parts hold.
- fig2d-power-odor-vs-reward
Beta power in all three regions differed between the odor sampling and reward periods, while RR power differed only in OB.
Change in LFP Power from odor-to-reward period (signed-rank tests ... CA1 p=0.0012, PFC p=5.3e-4, OB p=4.4e-4 ... CA1 p=0.11, PFC p=0.062, OB p=4.4e-4)
Figure 2d middle (legend)How it will be judged
Power in the windows of fig2e-beta-coherence-odor-vs-reward, per session odor minus reward, signed-rank across sessions. The legend gives no direction; the figure was not seen, so the direction found is reported. Reproduced when the three beta tests and the OB RR test give p < 0.05 and the CA1 and PFC RR tests give p >= 0.05; partially when four or five of the six hold.
- fig2de-correct-incorrect-no-difference
Neither power nor coherence, in beta or RR, differed between correct and incorrect trials.
There was no significant change in the level of coherence or power in the CA1-PFC-OB network in either frequency band between correct and incorrect trials ( Figure 2e bottom , signed-rank tests; all p's>0.05)
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 7; Figure 2d and 2e bottomHow it will be judged
Twelve comparisons (power in three regions and coherence in three pairs, each band). Correct trials subsampled at random to the number of incorrect trials in each session (legend, "trial-count matched"; F7/ClaireCrossCoherence.m lines 249 to 257). Per session correct minus incorrect. The paper names two tests - signed-rank across sessions (Results; code line 470) and rank-sum on sessions (legend). Absence claim. Reproduced when neither test finds a difference at p < 0.05 in any of the twelve comparisons, not reproduced when both find one in any, blocked as methods_underspecified when they differ; under LFP readings A and B. Effect sizes are reported.
- fig2g-air-beta-coherence
Beta coherence for all three region pairs was lower on air-only trials than on odor-cued correct trials, while RR coherence did not differ.
RR coherence was unchanged between odor- and air-cued sessions, but beta coherence was significantly lower in the absence of odor cues during air-only sessions ( Figure 2g )
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 5; Figure 2gHow it will be judged
One value per trial, the z-scored coherence averaged over the sampling period (legend and code, F16/LFPAnalysis/cs_LFPcoherence_noOdor.m), pooled over sessions; odor trials are correct trials of odor-cued sessions (legend). Bootstrap as the Methods give it - odor trials resampled with replacement to the number of air trials, 1000 times, the mean each time, compared with the observed air mean. The Methods count bootstrap means >= the air mean, which would give a p near 1 for a lower air mean; the code takes the smaller of the two tails. The lower tail is used (the fraction of bootstrap means <= the air mean), and the code's two-tail minimum is reported too. Coherence under LFP reading A (the code's pipeline here) and reading B. Reproduced when all three beta pairs give p < 0.05 with air lower and all three RR pairs give p >= 0.05; partially when four or five of the six hold. The outcome is the one both readings give; when they differ the claim is blocked as methods_underspecified with both shown.
- fig2f-swr-odor-vs-reward
Sharp-wave ripples were rare during odor sampling and frequent at the reward wells.
However, a distinct lack of SWR events at the odor port ( Figure 2f ; signed-rank test, p=1.1e-7) suggests that SWR replay is unlikely to play a direct role in decision making
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 6; Figure 2fHow it will be judged
Detection from the Methods (Sharp-wave ripple detection) - locally referenced CA1 LFP, 150 to 250 Hz, Hilbert envelope above 3 s.d. of its mean for at least 15 ms on at least one CA1 tetrode; tetrodes are the CA1 tetrodes with cells (F16/Ripples/cs_riptettag_anycells.m). The code keeps events at speed under 4 cm/s (cs_rippletimes.m), followed here. Number of SWRs per trial on correct trials (legend). Windows, which the Methods do not fix - (a) the odor sampling period and the reward-well period (reward-well beam break onset to exit); (b) the code's speed-defined stops around the odor port and around reward onset, the latter extended to at least 1 s (cs_rippleRate_NPvsReward.m). Per session the mean count per trial in each window; signed-rank across sessions (legend; the code's rank-sum over trials is a variant). Reproduced when p < 0.05 with fewer SWRs at the odor port under a reading, and the readings agree.
- fig2s1f-rr-beta-coupling
Beta amplitude was coupled to RR phase in CA1, PFC and OB during odor sampling.
During the decision-making period, we also found phase-amplitude coupling between beta and RR rhythms in all three regions
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 2; Figure 2-figure supplement 1fHow it will be judged
Tort modulation index as in the Methods (Cross-frequency coupling) - RR phase in 18 bins of 20 degrees, mean beta amplitude per bin normalized to sum 1, MI from the KL distance to the uniform; null from 500 shuffles of the trial assignment of RR phase series (Methods and legend). Per region, the local tetrode with most cells (as for phase locking; the Methods do not name the tetrode for this analysis), odor sampling periods. The Methods do not say which trials; readings all completed trials and correct trials only. The code restricts samples to high OB beta periods and shuffles amplitudes across samples (F16/LFPAnalysis/cs_cfc.m), a variant. The legend shows one real MI per region against a trial-shuffled null, so MI is pooled over the trials of all sessions; per-session MIs are reported as a variant. Reproduced when the pooled MI exceeds all 500 shuffles in each region; partially when that holds in one or two regions. The outcome is the one both trial readings give; when they differ the claim is blocked as methods_underspecified.
- fig2s2d-phase-offsets
During odor sampling CA1 beta led PFC beta, and CA1 RR led PFC and OB RR, though offsets varied widely across sessions.
During the full odor sampling window, CA1 beta lead PFC (p = 2e-4) and there was no consistent phase difference between OB and PFC or CA1 beta (all p>0.05). Similarly, CA1 RR lead PFC RR (p < 1e-5), as well as OB (p < 1e-5)
Figure 2-figure supplement 2d; Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 7How it will be judged
The paper names no test. From F7/ClairePhaseOffset.m - per trial the circular mean of the instantaneous phase difference over the last 500 ms of sampling (trials of 0.5 to 2 s), the circular mean over trials per session, one tetrode per region (most units), and a test of the session offsets against zero by a sign-flip null of 100,000 draws taking the smaller tail; the code's circ_mtest is a variant. The legend's "full odor sampling window" against the code's last 500 ms gives readings (a) whole sampling period, (b) last 500 ms. The legend also says no consistent offset between OB and CA1 RR, which contradicts CA1 RR leading OB; only the stated leads are tested for RR. Reproduced when the CA1 beta lead over PFC and the two RR leads are significant (p < 0.05) in the stated direction and the beta OB pairs are not. The outcome is the one both window readings give; when they differ the claim is blocked as methods_underspecified.
- fig2s1e-run-theta-shift
After odor port exit, as the rat ran, the dominant LFP frequency in CA1 and PFC shifted slightly from RR up into the theta band.
Following odor port disengagement and the initiation of running down the track, we observed a small shift in the dominant LFP frequency from RR to the theta band in CA1 and PFC
Results, Beta and RR coherence is elevated during odor sampling and decision making, paragraph 1; Figure 2-figure supplement 1eHow it will be judged
Full-maze animals only (legend). The paper gives no test; the code (F16/LFPAnalysis/cs_peakFreqOdorVRun.m with cs_PSD.m) takes power spectra (fpass 0 to 20 Hz, 1 s windows) from 0 to 3 s after odor onset, splits them at 0.8 s into odor and run parts, finds the peak frequency per day and compares odor and run by a signed-rank test. The search range in the code is set by array indices; it is read as the theta band the paper defines, 6 to 12 Hz. Readings for the windows (a) the code's 0 to 0.8 s and 0.8 to 3 s after odor onset, (b) the odor sampling period and run after exit at speed over 3 cm/s. Reproduced when the run peak is higher than the odor peak with signed-rank p < 0.05 in CA1 and in PFC; partially for one region. The outcome is the one both window readings give; when they differ the claim is blocked as methods_underspecified.
- unit-counts
The recordings gave 1,309 CA1 and 717 PFC units, of which 934 and 508 were active, 813 CA1 and 464 PFC active units were putative pyramidal cells (121 and 44 interneurons), 170 and 234 pyramidal cells were active in the odor period, and 138 and 185 of these were task responsive.
we first selected units that fired at least 100 spikes during the run epochs ( Figure 3g–h , 'Active Cells', 934 in CA1, 508 in PFC) and categorized them into pyramidal cells and interneurons based on firing rate and spike width
Results, Single neurons in CA1 and PFC exhibit choice selectivity during decision-making period, paragraph 1; Figure 3g-hHow it will be judged
Selections as defined in the opening comment (active units, cell types, odor period active, task responsive), each under the readings named there; a reading is one choice for each open detail. Reproduced when one reading gives every count within tolerance; partially when the totals and active counts match but some later counts do not. The dandiset's units may already be curated, in which case the totals compare what was deposited.
- fig3-choice-selective-fraction
About a third of task-responsive pyramidal cells were choice selective, 47 of 138 in CA1 and 59 of 185 in PFC.
To determine significance, the SI value for each task responsive pyramidal cell was compared to a distribution of SIs generated by shuffling the odor identities across trials (n=47/138, 34% CA1 selective cells, n=59/185, 31% PFC selective cells)
Results, Single neurons in CA1 and PFC exhibit choice selectivity during decision-making period, paragraph 2; Figure 3How it will be judged
Choice selectivity as defined in the opening comment, in task-responsive pyramidal cells, under each reading of unit-counts. Reproduced when the fractions fall within tolerance under a reading (the counts too when that reading reproduces 138 and 185).
- fig3s1e-nonselective-si
The Results say the correct-incorrect anti-correlation was not significant among task-responsive cells that were not choice selective, while the legend reports a significant anti-correlation (r = -0.34 in both regions).
This pattern was not significant within the population of neurons that were task-responsive but did not meet the selectivity threshold ( Figure 3—figure supplement 1d–e )
Results, Single neurons in CA1 and PFC exhibit choice selectivity during decision-making period, paragraph 3; Figure 3-figure supplement 1eHow it will be judged
As fig3ij-si-anticorrelated, in task-responsive pyramidal cells that are not choice selective. The paper disagrees with itself, so two readings - (a) the Results, no significant anti-correlation (Spearman p >= 0.05); (b) the legend, r near -0.34 (tolerance 0.034) with p < 0.05. Reproduced under whichever reading the result matches, and the inconsistency is recorded; r and p are reported for each region.
- fig3s1c-odor-vs-reward-rate
Firing rates of task-responsive pyramidal cells differed between odor sampling and reward consumption, in CA1 and in PFC.
Firing rates of task-responsive pyramidal neurons during odor sampling and reward consumption (Signed-rank tests ... n = 138 cells, p = 5.1e-13 ... n = 185 cells, p = 0.009)
Figure 3-figure supplement 1cHow it will be judged
No code for this panel was found. Per cell the mean rate over odor sampling periods and over reward periods (reward-well beam break onset to exit) of the same trials; the Methods do not fix the reward window or the trials, so readings are correct trials (rewarded) and all completed trials. Signed-rank test across cells. The direction is shown only in the figure, which was not seen, and is reported. Reproduced when both regions give p < 0.05 under a reading.
- fig4c-interneuron-si
CA1 choice-selective interneurons also had selectivity on incorrect trials anti-correlated with that on correct trials; PFC interneurons, few in number, did not.
For CA1 interneurons, the selectivity response showed anti-correlation during correct versus incorrect trials similar to that of pyramidal cells ( Figure 4c top , r(32) = –0.80, p=3.4e-8)
Results, Single neurons in CA1 and PFC exhibit choice selectivity during decision-making period, paragraph 4; Figure 4cHow it will be judged
Task responsiveness and choice selectivity as for pyramidal cells (Results, "using the same criteria"), in interneurons; correlation and readings as fig3ij-si-anticorrelated. Reproduced when CA1 interneurons give a negative correlation with p < 0.05 and r within tolerance, and PFC interneurons give p >= 0.05.
- fig4d-ccg-lags
Among CA1-PFC pairs with significant cross-correlation peaks during odor sampling, PFC interneurons tended to lead CA1 interneurons and CA1 pyramidal cells, CA1 interneuron-PFC pyramidal pairs were skewed too, and pyramidal-pyramidal pairs showed no skew.
We found a significant skew in CA1-PFC interneuron pairs towards the PFC interneurons leading (sign rank test, x=–0.028, p=3.5e-8) ( Figure 4d ), as well as for CA1 pyramidal cell - PFC interneuron pairs in the same direction (sign rank test, x=–0.013, p=0.0034)
Results, Single neurons in CA1 and PFC exhibit choice selectivity during decision-making period, paragraph 5; Figure 4dHow it will be judged
Method from the Methods (Corrected spike cross-correlogram) - for simultaneously recorded task-responsive CA1-PFC pairs, cross-correlogram from -0.15 to 0.15 s in 2.5 ms bins, one cell's spikes jittered by up to +/-50 ms 1000 times, pointwise z-score against the jitter, the highest peak within +/-100 ms that is significant at p < 0.05 taken as the pair's lag; positive lag is CA1 leading (legend). Open details and readings - spikes from (a) whole sampling periods (Methods) or (b) the last 0.5 s of sampling (code, F7/ClaireCellCellInteractions.m line 1298), correct and incorrect trials as the code uses; pointwise p < 0.05 as one-sided z > 1.645 or two-sided |z| > 1.96. The code's 200 jitters and z > 3 are a variant. Signed-rank test of lags against 0 per pair type. The Results' x is not named; the median and the mean lag are both compared with the expected values, a match under either counting. Reproduced when the two interneuron-involving tests quoted are significant with negative lags, the CA1 interneuron-PFC pyramidal test is significant and the pyramidal-pyramidal test gives p >= 0.05; partially when two or three of these four hold. The outcome is the one all window and threshold readings give; when they differ the claim is blocked as methods_underspecified. Pair counts are reported against 87, 283, 111 and 224.
- fig5c-local-locking
Most CA1 task-responsive cells and a minority of PFC cells were locked to local RR; a modest share of CA1 pyramidal cells and a large share of CA1 interneurons were locked to local beta, while PFC cells were locked to beta no more than chance.
Surprisingly though, while a modest but significant proportion of CA1 pyramidal cells were also coherent to local beta (28/138, 20% binomial test p=4.6e-11) a large fraction of CA1 interneurons were coherent to local beta (36/83, 43%)
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 1; Figure 5cHow it will be judged
Phase locking as in the opening comment, to the local tetrode, task-responsive cells. The interneuron denominators differ between the Results (83 CA1, 28 PFC) and the Figure 5c legend (96 and 36); both are counted and reported. Chance is 0.05, the dotted line of Figure 5-figure supplement 1c and the code's chance line (F7/ClaireBetaAnalysis.m line 622); binomial test, one-sided, of the number locked against n and 0.05. The 1-count tolerance on the smallest counts is a floor, not 10%. Reproduced when the CA1 pyramidal beta proportion exceeds chance with p < 0.05, the PFC beta proportions do not, and the fractions fall within tolerance (counts scaled to the n found when the denominators differ); partially when the significance pattern holds but some fractions do not.
- fig5d-cross-region-beta
More CA1 interneurons were locked to PFC beta than chance, while CA1 pyramidal cells and PFC cells showed no cross-regional beta locking beyond chance.
When we assessed cross-regional spike-beta coherence, we found that there were significantly more CA1 interneurons locked to PFC beta than chance
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 1; Figure 5c-dHow it will be judged
Each task-responsive cell tested against beta on the CA1, PFC and OB tetrodes; locked to a region when the Rayleigh p is below 0.05/3 (Results; F7/ClaireBetaAnalysis.m pcrit = .05/3) and assigned to the region with the smallest p (legend 5d; line 320). Binomial test of each count against chance; the chance probability is not stated, readings 0.0167 and 0.05. The paper gives some OB denominators as 138 where the population is PFC (185); counts are compared, not denominators. Reproduced when the CA1 interneuron count to PFC beta exceeds chance at p < 0.0167 and the CA1 pyramidal and PFC counts do not. The outcome is the one both chance readings give; when they differ the claim is blocked as methods_underspecified with both shown.
- fig5gh-phase-preference
Preferred beta and RR phases were consistent across task-responsive CA1 cells, while in PFC only interneurons had a consistent preferred RR phase.
phase preference across cells was consistent in CA1 for both rhythms ( Figure 5g–h , Rayleigh Z-test, CA1... pyr n=138, beta p=5.4e-5, RR P =5.9e-4, int n=83 beta p=2.62e-4 RR p=5.61e-6)
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 2; Figure 5g-hHow it will be judged
Each task-responsive cell's preferred phase is the mean phase of its spikes on correct trials (Methods, the circular mean under a von Mises model), whether or not it is locked (Results, "regardless of the significance"); local tetrode. Rayleigh test across cells per region, cell type and rhythm. Reproduced when the five significant tests give p < 0.05 and the three PFC tests give p >= 0.05; partially when six or seven of the eight hold.
- fig5-rr-more-than-beta
More cells were phase locked to RR than to beta, and the depth of RR modulation was stronger, in CA1 and in PFC.
even though a higher percentage of cells in both regions were phase locked to RR ( Figure 5c and z-test for proportions ... and depth of this modulation was much stronger for RR compared to beta
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 2; Figure 5c; Figure 5-figure supplement 1aHow it will be judged
Proportions of task-responsive cells locked to RR and to beta compared by a pooled z-test for two proportions (code F16/PhaseLocking/cs_plCellProportions.m, which counts a cell locked to CA1 or PFC LFP; readings local LFP only, and CA1 or PFC LFP as the code does), two-sided and one-sided (the code's ztestprop.m is one-sided). Rayleigh Z of each task-responsive pyramidal cell's spikes for RR against beta by rank-sum (legend). Reproduced when, in each region, RR has the larger proportion with p < 0.05 and larger Rayleigh Z with p < 0.05. The outcome is the one all readings give; when they differ the claim is blocked as methods_underspecified.
- fig5s1de-beta-rr-overlap
The share of cells locked to both beta and RR did not differ from that expected by chance from the share locked to each, in CA1 and PFC, for pyramidal cells and interneurons.
in both CA1 and PFC, the proportion of cells that were phase locked to both beta and RR was no different than chance, given the percentages of cells that were modulated by either rhythm
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 3; Figure 5-figure supplement 1d-eHow it will be judged
Chance probability the product of the fractions locked to beta and to RR (legend; code F16/Spiking/jhb_plotAllProportions.m line 389); binomial test of the number locked to both against n and that probability, two-sided (the legend shows a 95% binomial interval). Absence claim on the paper's test - reproduced when all four give p >= 0.05, not reproduced when any gives p < 0.05; the observed and chance fractions are reported.
- fig5s1fh-selective-and-locked
Choice-selective cells were no more or less likely than chance to be phase locked to beta or to RR, in CA1 and PFC.
Interestingly, we found that the number of cells that met both criteria was no different than chance for either CA1 or PFC ( Figure 5—figure supplement 1f , binomial tests; all p>0.05)
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 4; Figure 5-figure supplement 1f-hHow it will be judged
As fig5s1de-beta-rr-overlap, with chance the product of the fractions choice selective and phase locked among task-responsive cells. "Phase locked" is to any region's rhythm (Results, "within and across regions"); the local-only count is a reading too. Absence claim - reproduced when all six tests give p >= 0.05, not reproduced when any gives p < 0.05.
- fig5e-beta-mvl-correct-incorrect
Beta-locked CA1 pyramidal cells and interneurons and PFC pyramidal cells were more strongly locked to local beta on correct than on incorrect trials.
We found that both pyramidal cells and interneurons in CA1 and pyramidal cells in PFC that were phase-locked to the local beta rhythm exhibited a lower MVL on incorrect trials
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 5; Figure 5eHow it will be judged
Cells - task-responsive cells locked to local beta. Method from the Methods (Phase locking) - correct-trial spikes subsampled to the number of incorrect-trial spikes, the MVL of the subsample bootstrapped 1000 times, the mean bootstrapped MVL compared with the incorrect-trial MVL; signed-rank test across cells. The code resamples trials, not spikes, 500 times (Path A, cs_phaselockBootstrap_v2.m) or subsamples both conditions 200 times and prints a t-test (Path B, F7/ClaireBetaAnalysis.m lines 155 to 172 and 890); variants. Reproduced when the three groups give lower incorrect MVL with p < 0.05; partially for one or two.
- fig5f-rr-mvl-no-difference
Locking to local RR did not differ between correct and incorrect trials.
There was no significant effect for local RR phase locking ( Figure 5f , signed-rank test, no significance found)
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 5; Figure 5fHow it will be judged
As fig5e-beta-mvl-correct-incorrect, for RR-locked cells. Absence claim - reproduced when none of the four groups gives p < 0.05; the median differences are reported.
- fig5s2-ca1-int-pfc-beta-accuracy
CA1 interneurons locked to PFC beta were more strongly locked on correct than on incorrect trials.
Interestingly, cross-region spike-phase coherence showed a similar relationship to decision accuracy for CA1 phase-coherent interneurons ( Figure 5—figure supplement 2 n=20, p=0.0023)
Results, CA1 and PFC cells phase lock to beta and respiratory rhythms during decision making, paragraph 5; Figure 5-figure supplement 2aHow it will be judged
As fig5e-beta-mvl-correct-incorrect, for CA1 interneurons significantly locked to PFC beta; the other cross-regional combinations (legend, all p > 0.05) are reported. Which interneurons form the 20 is not stated - readings locked to PFC beta at p < 0.05, or at p < 0.0167 (as in fig5d-cross-region-beta). Reproduced when correct exceeds incorrect with signed-rank p < 0.05 under a reading.
- fig6d-discrimination-time
CA1 and PFC ensemble trajectories for the two choices separated within the odor sampling period, at similar times (CA1 200 to 800 ms, PFC 400 to 700 ms after odor onset).
Neural discrimination times for CA1 and PFC (n=24 sessions across 8 animals; threshold of 4 task-responsive cells, CA1... 200–800ms; PFC... 400–700ms) were similar ( Figure 6d ; p=0.68, sign-rank test)
Results, Neural ensemble responses in CA1 and PFC during decision making predict the upcoming choice, paragraph 2; Figure 6c-dHow it will be judged
Method from the Methods (PCA) - sessions with at least 4 task-responsive cells per region; spikes in 100 ms bins from 0 to 1 s after odor onset, correct trials (the code, F16/pca/sj_columnVectors_day_npCells.m line 71); first 3 PCs; Euclidean distance between the mean left-bound and right-bound trajectories; null from 1000 shuffles of trial labels (code sj_PCA_AvgDis_Sess_npcells_update2.m line 232); discrimination time is the first bin whose distance exceeds the null's upper bound. Cells are task-responsive pyramidal cells (Results, "We first considered only pyramidal cells that were task-responsive for the ensemble analyses"); the code's pyramidal cells and interneurons (npCells_PyrInt) are a variant. Open details and readings - PCA fitted (c) on the session's trial-averaged bins by neurons matrix (Methods) or (d) per bin on single trials (code lines 175 to 176); the bound the 95th (code) or 97.5th percentile (the upper edge of a 95% interval). The code's window from -0.2 s is a variant. Each reading gives per-session times; reported are the range per region and a signed-rank test of CA1 against PFC over sessions with both. Reproduced when the ranges fall within the expected bounds widened by tolerance and the signed-rank test gives p >= 0.05 (the absence part judged on that test). The outcome is the one all readings give; when they differ the claim is blocked as methods_underspecified.
- fig6c-even-odd-control
A control distance between even and odd trials of the same odor never crossed significance in any session.
A second even-odd shuffle ( Figure 6c black lines) failed to cross significance in any session, suggesting the emergence of a robust discriminatory neural code during decision-making
Results, Neural ensemble responses in CA1 and PFC during decision making predict the upcoming choice, paragraph 2; Figure 6cHow it will be judged
For each session and region of fig6d-discrimination-time, the distance between the mean trajectories of even and odd trials of each odor, averaged over the two odors (legend; code F16/pca/Run_PCA_SJ_copy.m lines 107 to 112), compared with the same null bound bin by bin. The trace the figure code labels even-odd (sj_PCA_AvgDis_Sess_npcells_update2.m line 332) is the first shuffle's distance, a variant. Reproduced when no session crosses the bound in any bin; partially when at most 10% of sessions do.
- fig6e-glm-choice-prediction
A GLM on task-responsive pyramidal ensembles predicted the upcoming choice before odor port exit, from 0.1 s before exit in CA1 and 0.3 s before in PFC.
When we performed this analysis using task-responsive putative pyramidal cells we found that reward choice could be accurately predicted 0.1 s prior to odor port disengagement by CA1 ensembles, and 0.3 s prior by PFC ensembles ( Figure 6e left )
Results, Neural ensemble responses in CA1 and PFC during decision making predict the upcoming choice, paragraph 3; Figure 6e leftHow it will be judged
Method from the Methods (GLM) - task-responsive cells active on at least 10 trials, spike counts per cell as predictors, choice (the well chosen, all completed trials, as the code labels them, cs_predictChoice_GLM.m line 122) as response; binomial GLM, with the Methods' log link or the code's logit (line 142); five-fold cross-validation, accuracy the fraction of test trials correct; null from shuffled choice labels; rank-sum test of real against shuffled accuracy over sessions per window. Tolerance is one 0.1 s window step. The alignment is open, as the Methods use cumulative windows from odor onset (0-0.1 to 0-1.0 s; the plotted code, cs_trialPredictionGLM.m) and the Results speak of time before exit. Readings (a) windows from odor port exit minus 0.6 s to exit plus w, w from -0.5 to 0.2 s in 0.1 s steps (code cs_predictChoice_GLM_v3.m), the time of a window its end; (b) the onset-aligned windows, a window's time before exit read as the mean sampling duration (0.82 s, fig1c) minus its length. The earliest window from which all later windows are significant is reported. Shuffles - 100 per session as the code does (cs_predictChoice_GLM.m line 155), each with its own fold split. The code's t-test is a variant. Reproduced when both regions predict above chance before exit and both earliest times fall within tolerance; partially when only one does. The outcome is the one both alignment readings give; when they differ the claim is blocked as methods_underspecified.
- fig6e-glm-with-interneurons
Adding interneurons made CA1 prediction accurate earlier, 0.2 s before exit, PFC prediction did not change, and resampling pyramidal cells to the same ensemble size gave 0.1 s again.
In this case, we found an improvement in prediction for CA1 ensembles... reward prediction was now accurate starting earlier, at 0.2 s prior to decision execution. Prediction by PFC ensembles remained the same
Results, Neural ensemble responses in CA1 and PFC during decision making predict the upcoming choice, paragraph 3; Figure 6e rightHow it will be judged
As fig6e-glm-choice-prediction with task-responsive pyramidal cells and interneurons. The control resamples pyramidal cells with replacement to the number of pyramidal cells plus interneurons (Results); no code was found for it, and the number of resamples is not stated - read as 100 resamples per session with accuracy averaged, a standard choice. Reproduced when CA1 with interneurons is significant earlier than CA1 pyramidal alone, PFC unchanged and the control as for pyramidal cells, with the times within tolerance. The outcome is the one both alignment readings of fig6e-glm-choice-prediction give; when they differ the claim is blocked as methods_underspecified.
- fig7-spatial-field-counts
Many CA1 and PFC cells had spatial fields on the track, 344 of 585 in CA1 and 159 of 288 in PFC.
A large fraction of cells in CA1 and PFC exhibited spatial activity on the maze (see Methods for details on spatial parameters, n=344 out of 585 CA1 cells; n=159 out of 288 PFC cells had fields on the track)
Results, Representations of choice and space are maintained independently during stem running, paragraph 1; Figure 7bHow it will be judged
Full-maze sessions (Results, 26 sessions from 6 rats; Methods, the five full-maze animals; both sets shown). Cells are pyramidal cells that fire during run (code F7/SummaryObjPlaceInteractions.m lines 349 to 351). Place fields from the Methods (Place field determination) - run epochs at speed over 3 cm/s, each outbound trajectory linearized into 100 bins, rate map smoothed with a Gaussian of 2 bins, peak at least 1 Hz, peak at least 2 s.d. above the mean of a null from circularly shifting spikes within each run, field under 75% of the trajectory. Shuffle count from the code, 500 (F7/CalcLinPlaceField.m); the code's one-tailed p < 0.05 in place of 2 s.d. is a variant. Reproduced when the counts fall within tolerance.
- fig7b-odor-active-more-fields
Odor-period active cells were more likely to have spatial fields than inactive cells, in CA1 (71 of 108 against 273 of 477) and PFC (83 of 128 against 76 of 160).
Cells that were odor period active were more likely to have spatial fields on the track (odor sampling periods were excluded in spatial responses, see Methods ), compared to cells that were not ( Figure 7b , binomial test
Results, Representations of choice and space are maintained independently during stem running, paragraph 2; Figure 7bHow it will be judged
Binomial test of the active cells with fields against n active, with null probability the fraction of inactive cells with fields, one-sided upper (code F7/SummaryObjPlaceInteractions.m lines 432 to 434; the code's P(X > k) is shown beside P(X >= k)). Reproduced when both regions give p < 0.05 with active cells more likely to have fields.
- fig7s1b-odor-active-higher-rate
Odor-period active cells had higher mean firing rates during running than inactive cells, in CA1 and PFC.
Additionally, for both CA1 and PFC neurons, we found odor period active cells had higher firing rates overall ( Figure 7—figure supplement 1b )
Results, Representations of choice and space are maintained independently during stem running, paragraph 2; Figure 7-figure supplement 1bHow it will be judged
Mean firing rate of each cell during running on the maze, excluding the odor sampling periods (legend), at speed over 3 cm/s; rank-sum of active against inactive cells (code lines 485 and 500). Reproduced when both regions give p < 0.05 with active cells higher.
- fig7c-field-properties-no-difference
Place field width, peak rate and sparsity did not differ between odor-period active and inactive cells in either region.
Surprisingly however, we found no difference in field width ( Figure 7c , rank-sum tests ... field peak rate ( Figure 7—figure supplement 1c ), or field sparsity
Results, Representations of choice and space are maintained independently during stem running, paragraph 2; Figure 7c; Figure 7-figure supplement 1c-dHow it will be judged
Each outbound field counted separately (Methods and legend 7c). Width by flood fill from the peak to the bins where rate falls below 25% of peak; peak rate on the smoothed linearized map; sparsity as Skaggs et al. 1993 (Methods). Rank-sum of active against inactive fields. Absence claim - reproduced when all six tests give p >= 0.05, not reproduced when any gives p < 0.05; medians are reported.
- fig7s1e-trajectory-selective-odor-active
Trajectory-selective cells were more likely than non-selective cells to be active in the odor period.
although trajectory-selective cells were significantly more likely than trajectory-nonselective cells to respond in the decision-making period ( Figure 7—figure supplement 1e )
Results, Representations of choice and space are maintained independently during stem running, paragraph 3; Figure 7-figure supplement 1eHow it will be judged
Trajectory selective as in the Methods (Trajectory selectivity) - the correlation of the left and right outbound linearized tuning curves below the 5th percentile of a null from shuffling trajectory identity across runs; the code has no such shuffle (a plain correlation, code line 944) and uses splitter scores instead, so the Methods' rule is used with 500 shuffles (the code's count for its scores, F7/CalcLinTrajectories.m) and the code's SplitterScore at p < 0.05 is a variant. Binomial test as in fig7b-odor-active-more-fields, the share of trajectory-selective cells that are odor active tested against the share among non-selective cells. Reproduced when both regions give p < 0.05 in that direction.
- fig7e-odor-decoder-first-quintile
A decoder trained on odor-period activity decoded the choice from run activity in the first fifth of the track, nearest the odor port, but at chance thereafter.
We found that choice identity could be decoded at the first quintile of track, nearest to the odor port, but decoding accuracy diminished to chance level thereafter ( Figure 7e )
Results, Representations of choice and space are maintained independently during stem running, paragraph 4; Figure 7eHow it will be judged
Memoryless Poisson Bayesian decoder (Methods, Bayesian decoder) built per choice from spikes in the odor sampling period, applied to spikes of each run in each fifth of the linearized outbound trajectory (equal fifths of 123 cm, legend); runs from 0.5 s after odor port exit at speed over 3 cm/s; null from 200 shuffles of choice labels in the training set; rank-sum of real against shuffled accuracy (legend; the code's paired signed-rank over sessions, F7/Claire_Odor_Routedecoder.m line 629, a variant). Cells are task-responsive cells (legend, "task-responsive ensemble activity"); the code's units active on at least 10 trials (line 429) are a variant. The unit of the test is sessions, as in the code. Reproduced when quintile 1 is above chance with p < 0.05 and quintiles 2 to 5 are not; partially when quintile 1 is and one later quintile also is.
- fig7f-spatial-decoder-all-quintiles
Decoders trained on the spatially active ensemble in each fifth of the track decoded the upcoming choice at every position along the trajectory.
In contrast, the animal's upcoming behavioral choice could be accurately decoded at all periods along the trajectory if instead the spatially active ensembles in each respective spatial quintile was used as the training set ( Figure 7f )
Results, Representations of choice and space are maintained independently during stem running, paragraph 4; Figure 7fHow it will be judged
As fig7e-odor-decoder-first-quintile, but trained and tested on run activity in each quintile with five-fold cross-validation over runs (Methods) and 200 label shuffles. "Spatially active" is not defined; readings (a) cells with a place field on either outbound trajectory (fig7-spatial-field-counts), (b) cells firing in run, active on at least 10 runs (code lines 780 and 998 to 1000). Reproduced when all five quintiles are above chance with p < 0.05 in CA1 and in PFC; partially when four of five are in each region. The outcome is the one both readings give; when they differ the claim is blocked as methods_underspecified.
- fig7s1f-ensemble-rate-near-port
In CA1 the firing rate of choice-selective ensembles during outbound runs was highest within 20 cm of the odor port and fell thereafter.
This is further supported by the fact that, at least for CA1, choice-selective ensemble firing rate is highest within 20 cm of the odor port, and decreases thereafter
Results, Representations of choice and space are maintained independently during stem running, paragraph 4; Figure 7-figure supplement 1fHow it will be judged
Mean over choice-selective cells of each cell's linearized outbound rate (code F7/SummaryObjPlaceInteractions.m lines 834 to 873 use odor-selective cells, though the legend says odor period active; both shown); null from circularly shifting each cell's curve by a random number of bins 1000 times (code lines 834 to 873), a position marked when the ensemble rate exceeds the null's 99th percentile (legend). The Results say choice-selective and the legend odor period active, so the cell set is a reading of which subset is meant, a match under either. Reproduced when the CA1 ensemble rate peaks within 20 cm of the odor port and is lower beyond 20 cm than within it; the marked positions are reported.
Pipeline
- Paper
- done (1 of 1)
- Claims
- done
- Claims check
- done (ready)
- Attempt
- todo
- Questions
- none
- Verified
- none