DANDI:001187 · Encore Science

Claims 29

From the paper 10.1016/j.neuron.2024.09.013 · claims file · checked 2026-10-22: ready

sessions-and-neurons
After excluding two sessions, 46 sessions are analysed, in which 883 single units pass the 0.1 Hz rate filter, 351 in the hippocampus and 532 in the amygdala, of 950 isolated.
In total, we isolated 950 neurons across both areas of the MTL. Neurons with a firing rate lower than 0.1 Hz in either the WM or the LTM tasks were excluded from analysis (67 neurons (7.1%)). Analyses are based on 351 isolated neurons in the hippocampus and 532 in the amygdala
Results, paragraph 3; STAR Methods, Spike sorting
How it will be judged
Counts of sessions and of units by area, before and after the rate filter of U, in the sessions of S. Tolerance zero, as these are counts of the deposited units. Reproduced when all four match; partially reproduced when the counts after the filter match but the isolated count does not (Table S1 sums to 973 recorded units, so the paper itself gives two totals). The Table S1 total is reported beside them.
table-s1-counts
Per session, the numbers of recorded units in the hippocampus and the amygdala are those of Table S1.
How it will be judged
Per session, matched by the session labels of Table S1 to the dataset's subject and session through the patient metadata, the count of units per area in the dataset (before and after the rate filter of U) against the table's recorded count. Reproduced when every one of the 46 sessions matches under one of the two counts; partially reproduced when some do and others do not, the sessions that differ being listed. The category-selective and memory-selective counts per session of the table are compared with those of claims category-neuron-counts and ms-neuron-counts, as a report, not a test.
behaviour-overall-accuracy
Patients answered 93.3% of WM trials and 70.3% of LTM trials correctly (d' 1.19), both above the 50% chance level.
Patients performed well in both parts of the task, with an average accuracy of 93.3 ± 7.6 % (mean ± SD) in the WM task and 70.3 ± 9.8 % (d’ = 1.19 ± 0.62) in the LTM task, respectively (both p < 0.0001 as compared to 50 % chance; permutation-based t-test
Results, paragraph 2
How it will be judged
Per session of S, the proportion of correct WM trials and of correct LTM trials (all 400), and d' as z(hit rate) - z(false-alarm rate) of old and new pictures; means across sessions. The tolerances are the standard errors of the mean (the stated SD over the square root of 46), since the paper's stated spread is the SD across sessions and these are means of the same sessions; the 10% default would be looser. The correction for hit or false-alarm rates of 0 or 1 is not given; none is applied unless a rate is 0 or 1, in which case the 1/(2N) rule, a standard choice, is used and said. Reproduced when the three means are within tolerance and each accuracy exceeds 50% by a one-sample permutation test (P) at p < 0.0001.
fig1c-wm-accuracy-load
WM accuracy is higher in load 1 than in load 3 trials.
In the WM task, patients performed with higher accuracy (Fig. 1c; load 1 – load 3: t(45) = 4.89, p < 0.0001
Results, paragraph 2; Figure 1c
How it will be judged
Per session of S, the proportion correct in load 1 and in load 3 trials (plot_behavior.m), compared by the paired permutation t-test of P. Reproduced when load 1 is higher at p < 0.0001; partially reproduced when it is higher at p < 0.05 only.
fig1d-wm-rt-load
WM responses are faster in load 1 than in load 3 trials.
and faster (Fig. 1d; t(45) = −4.29, p < 0.0001) in load 1 than load 3 trials
Results, paragraph 2; Figure 1d
How it will be judged
Per session of S, the mean response time of load 1 and of load 3 trials, response times above 5 s set to missing as in plot_behavior.m (the Methods name no exclusion; the variant without it is shown), compared by the paired permutation t-test of P. Reproduced when load 1 is faster at p < 0.0001; partially reproduced when it is faster at p < 0.05 only. All trials are used, as in the code; the paper does not restrict to correct trials.
fig1e-ltm-confidence-accuracy
LTM recognition is more accurate in high-confidence than in low-confidence trials.
In the LTM task, retrieval was more accurate (Fig. 1e; t(44) = 8.66, p < 0.0001; one patient did not use confidence ratings and was therefore excluded from all confidence-related analyses)
Results, paragraph 2; Figure 1e
How it will be judged
Per session of S with confidence ratings, LTM accuracy over all trials (old and new) of high and of low confidence (L), compared by the paired permutation t-test of P. Open, the text says "Low confidence was defined as an average of “unsure” and “guessing” responses"; readings (a) accuracy of the unsure and guessing trials pooled, as plot_behavior.m computes it, and (b) the mean of the unsure accuracy and the guessing accuracy. Reproduced when high is more accurate at p < 0.0001; partially reproduced when at p < 0.05 only.
ltm-rt-confidence
LTM responses are faster with high confidence (mean 1.83 s) than with low confidence (2.62 s).
and faster (mean RT high: 1.83 ± 0.57 s; mean RT low: 2.62 ± 0.66 s; t(44) = −11.13, p < 0.0001) when patients indicated high confidence in their responses
Results, paragraph 2
How it will be judged
Per session with confidence ratings, the mean response time of high- and low-confidence LTM trials (L), all trials; means across sessions; the paired permutation t-test of P. Tolerances are the standard errors of the mean (stated SD over the square root of 45), for the reason given in behaviour-overall-accuracy. Open, (a) all response times, as the text names no exclusion, or (b) response times above 5 s set to missing, as plot_behavior.m does for every response time it reads; and low confidence pooled or averaged as in fig1e-ltm-confidence-accuracy. Reproduced when, under one reading, both means are within tolerance and high is faster at p < 0.0001; partially reproduced when high is faster at p < 0.05 but a mean is outside tolerance under every reading.
fig1f-ltm-load
Pictures held in load 1 WM trials are later recognized better than pictures held in load 3 trials.
Items that were encoded in load 1 trials in the previous WM task were remembered better than items encoded during load 3 trials (Fig. 1f; t(45) = 2.42; p = 0.018)
Results, paragraph 2; Figure 1f
How it will be judged
Per session of S, the hit rate (old pictures answered old over all old pictures) for pictures from load 1 and from load 3 trials, from the LTM trials' load field, as plot_behavior.m computes it; the paired permutation t-test of P. Reproduced when load 1 is higher at p < 0.05.
fig1g-ltm-probe
Pictures that were also used as WM probes, and so shown twice, are later recognized better than pictures shown once.
items that were also used as the probe and therefore presented twice were remembered better than items not used as the probe (Fig. 1g; t(45) = 7.25, p < 0.0001)
Results, paragraph 2; Figure 1g
How it will be judged
Per session of S, the hit rate for old pictures shown twice and shown once in the WM task (the LTM trials' num_times_shown field, plot_behavior.m); the paired permutation t-test of P. Reproduced when shown twice is higher at p < 0.0001; partially reproduced when at p < 0.05 only.
category-neuron-counts
104 hippocampal units (29.65%) and 220 amygdala units (41.35%) are category neurons.
In the hippocampus, 104 (29.65 %) neurons qualified as category neurons, and in the amygdala 220 (41.35 %) neurons qualified as category neurons
Results, Category neurons remain persistently active when their preferred category is held in WM, paragraph 1
How it will be judged
Selection C on the units of U in the sessions of S. Reproduced when both counts are within tolerance (10% of each, which also covers the run-to-run variation of the permutation tests that the code warns of).
fig2b-hippocampus-persistent
Hippocampal category neurons are persistently active above baseline during maintenance and fire more when an image of their preferred category is held in WM.
In the hippocampus, category neurons remained persistently active throughout the maintenance period (intercept: beta = 17.86, p = 1.38 × 10−3; mixed-effects GLM) and had significantly higher FRs during trials in which images from the preferred category were maintained in WM (Fig. 2b; preferred category: beta = 12.47, p = 8.82 × 10−7)
Results, Category neurons remain persistently active, paragraph 2; Figure 2b
How it will be judged
Hippocampal category neurons of C; all correct WM trials, both loads; the rate 0 to 2.5 s after maintenance onset, normalized as B; outliers removed as O; the GLM of G with the fixed effect preferred category (true when an encoding picture of the trial is of the neuron's preferred category). Open, (a) one row per trial, as the text describes, or (b) the rows of the code's table (one per encoding picture later tested in the LTM test, the preferred flag per picture); and the baseline B1 or B2. Reproduced when, under the readings, the intercept and the preferred-category effect are positive at p < 0.05 and both betas within tolerance; partially reproduced when both are positive at p < 0.05 but a beta is outside tolerance.
fig2c-hippocampus-correct-load
When the preferred category is held, hippocampal category neurons fire more in correct than incorrect trials and more in load 1 than load 3 trials.
Using only preferred trials and modelling WM accuracy (2 levels, WM correct/incorrect, categorical) as well as load (2 levels, load 1/load 3, categorical) as fixed effects, we further observed that category neurons in the hippocampus had higher FRs in (1) correct than incorrect trials (beta = 16.22, p = 0.046) and (2) in load 1 than in load 3 trials (Fig. 2c; beta = 13.769, p = 0.0023)
Results, Category neurons remain persistently active, paragraph 2; Figure 2c
How it will be judged
Hippocampal category neurons of C; WM trials holding a picture of the neuron's preferred category, correct and incorrect, both loads; maintenance rate (0 to 2.5 s) normalized as B; O; the GLM of G with fixed effects WM accuracy (incorrect as reference) and load (load 3 as reference). Open, (a) main effects only, as the text names two effects, or (b) with their interaction; and B1 or B2. Reproduced when, under the readings, both effects are positive at p < 0.05 and both betas within tolerance; partially reproduced when the load effect holds and the accuracy effect is positive but p >= 0.05 (its reported p is 0.046), or when both hold with a beta outside tolerance.
fig2e-amygdala-persistent
Amygdala category neurons are also persistently active during maintenance and fire more for the preferred category.
In the amygdala, category neurons also remained persistently active throughout the maintenance period across all correct trials (intercept: beta = 9.78, p = 6.76 × 10−4) with FRs higher for preferred than unpreferred categories (Fig. 2e; preferred category: beta = 6.91, p = 9.38 × 10−7)
Results, Category neurons remain persistently active, paragraph 3; Figure 2e
How it will be judged
As fig2b-hippocampus-persistent, with the amygdala category neurons.
fig2f-amygdala-load-not-accuracy
When the preferred category is held, amygdala category neurons fire more in load 1 than load 3 trials, but not more in correct than incorrect trials.
using only preferred trials category neurons in the amygdala showed a significant main effect only for load (Fig. 2f; load 1 – load 3; beta = 12.309; p = 2.46 × 10−7), but the effect for WM accuracy (correct – incorrect: beta = −6.13; p = 0.24) was not significant
Results, Category neurons remain persistently active, paragraph 3; Figure 2f
How it will be judged
The model of fig2c-hippocampus-correct-load, with the amygdala category neurons and the same readings. The accuracy part is a claim of absence, judged on the same GLM at p < 0.05. Reproduced when the load effect is positive at p < 0.05 with its beta within tolerance and the accuracy effect has p >= 0.05 under every reading; partially reproduced when only one part holds. The accuracy beta is reported either way.
fig3a-area-interactions
In a GLM over category neurons of both areas, no main effect of LTM accuracy, confidence or area is significant, nor the LTM accuracy by confidence interaction, but every interaction with area is, so the relation of maintenance activity to later memory differs between hippocampus and amygdala.
We did not observe any significant main effects nor was the interaction between LTM accuracy and confidence significant (Fig. 3a; all p > 0.26). However, all interaction terms including area showed significant modulations (confidence × area: beta = 16.88, p = 0.025; LTM accuracy × area: beta = 59.78, p = 0.0030; confidence × LTM accuracy × area: beta = −28.64, p = 0.0029)
Results, Persistent activity of category neurons in the hippocampus - not the amygdala - predicts LTM formation, paragraph 1; Figure 3a
How it will be judged
As NWB_computeGLM.m, which is the paper's code for this panel. Category neurons of C in both areas; one row per WM picture later shown in the LTM test; maintenance rate 0 to 2.5 s after maintenance onset, normalized as B (mean baseline over all the neuron's rows); then rows kept when the WM trial was correct and the picture is of the preferred category, with a confidence rating; confidence recoded as the Methods say (forgotten sure 1, unsure 2, guessing 3; remembered guessing 1, unsure 2, sure 3; the code maps 4, 5, 6 to 1, 2, 3 and keeps 1, 2, 3, which is the same under L); FR ~ LTMacc * conf * area with the random effects of G, area with amygdala as reference so that the interactions are hippocampus minus amygdala. Outliers O-M for the main result, O-C (the code) shown as a variant. Reproduced when the three interactions with area have p < 0.05 with the reported signs and betas within tolerance, and the main effects and the LTMacc by conf interaction have p >= 0.05; partially reproduced when the three interactions hold in sign and significance with a beta out of tolerance or a main effect significant.
fig3b-hippocampus-predicts-ltm
In the hippocampus, maintenance activity of category neurons for the preferred category predicts later LTM accuracy and confidence, with an interaction between the two, so activity is higher for pictures later remembered than forgotten.
This revealed a significant main effect of confidence (Fig. 3b; beta = 20.30, p = 4.17 × 10−3) and LTM accuracy (beta = 67.90, p = 4.43 × 10−4), as well as a significant interaction between the two terms (beta = −30.80, p = 7.74 × 10−4) in the hippocampus
Results, Persistent activity of category neurons in the hippocampus, paragraph 1; Figure 3b
How it will be judged
The table of fig3a-area-interactions restricted to hippocampal category neurons, FR ~ LTMacc * conf with the random effects of G (the Fig. 3a code with the area term dropped, as the text says the analysis was "repeated ... in each area separately"). Outliers O-M, with O-C as a variant; O applied over both areas together as in Fig. 3a (a) or within the hippocampus (b), the text not saying which. Reproduced when, under the readings, the three effects have p < 0.05 with the reported signs and betas within tolerance; partially reproduced when the three hold in sign and significance with a beta out of tolerance.
fig3b-amygdala-no-prediction
In the amygdala, maintenance activity of category neurons does not predict later LTM accuracy or confidence.
in the hippocampus but not in the amygdala (confidence: beta = 1.54, p = 0.69; LTM accuracy: beta = −2.27, p = 0.82; confidence × LTM accuracy: beta = 1.02, p = 0.83)
Results, Persistent activity of category neurons in the hippocampus, paragraph 1; Figure 3b
How it will be judged
A claim of absence, judged on the paper's test, the model of fig3b-hippocampus-predicts-ltm on amygdala category neurons, with the same readings. Reproduced when none of the three effects has p < 0.05; not reproduced when any does. The betas are reported either way.
fig3c-hippocampus-high-confidence
Across hippocampal category neurons, maintenance firing for the preferred category is higher for pictures later remembered than forgotten with high confidence, but not with low confidence.
In the hippocampus, FRs were higher for remembered than forgotten trials for high confidence (Fig. 3c; t(82) = 2.16, p = 0.028; some neurons were removed due to insufficient data in at least one of the conditions or since they differed ±3 SD from the mean across all neurons and conditions; see Methods), but not low confidence trials (t(68) = −1.38, p = 0.17)
Results, Persistent activity of category neurons in the hippocampus, paragraph 2; Figure 3c
How it will be judged
Hippocampal category neurons of C; the rows of fig3a-area-interactions (correct WM trials, preferred pictures later tested). Per neuron, the mean maintenance rate (0 to 2.5 s) of pictures remembered with high confidence (L, rating 6) and of pictures forgotten with high confidence (rating 1); for low confidence, remembered 4 or 5 against forgotten 2 or 3. Neurons without trials in a condition, or outliers by O-M, are removed; the paired permutation t-test of P. Open, the rate is (a) normalized as B, as the betas of Fig. 3 are, or (b) in spikes per second, the legend not naming the unit of panel c. The low-confidence part is a claim of absence on the same test at p < 0.05. Reproduced when, under both readings, the high-confidence difference is positive at p < 0.05 and the low-confidence one has p >= 0.05; partially reproduced when only the high-confidence part holds. The numbers of neurons tested (83 and 69 in the paper) are reported.
fig3d-amygdala-no-difference
Across amygdala category neurons, maintenance firing does not differ between later remembered and forgotten pictures, with high or low confidence.
In the amygdala, we neither observed a significant effect for high (Fig. 3d; t(158) = 0.84, p = 0.41) nor low confidence trials (t(152) = 1.15, p = 0.26)
Results, Persistent activity of category neurons in the hippocampus, paragraph 2; Figure 3d
How it will be judged
A claim of absence, judged on the test of fig3c-hippocampus-high-confidence applied to amygdala category neurons, with the same readings. Reproduced when neither comparison has p < 0.05; not reproduced when either does. Mean differences are reported.
fig3e-early-delay-cluster
In hippocampal category neurons, firing differs between pictures later remembered and forgotten with high confidence early in the maintenance period, from 0 to 650 ms.
In the hippocampus, FRs during the beginning of the maintenance period (0 – 650 ms) differed significantly between the two conditions (Fig. 3e; cluster-p = 0.0042; cluster-based permutation test
Results, Persistent activity of category neurons in the hippocampus, paragraph 3; Figure 3e
How it will be judged
The high-confidence remembered and forgotten conditions and neurons of fig3c-hippocampus-high-confidence; per neuron, time-resolved rates over the maintenance period 0 to 2.5 s; a cluster-based permutation test as FieldTrip's ft_freqstatistics (Methods, Statistics), paired across neurons (dependent-samples t), 10,000 permutations, alpha 0.025 for each one-sided cluster. Open, (T1) the binning, (a) 200 ms windows stepped by 25 ms, the only PSTH binning the paper gives (Fig. 2a legend), or (b) 50 ms bins, the code's PSTH bin (spikePlotParams in NWB_calcCatSelective_SB.m), each time point at the window centre; (T2) the cluster-forming threshold, not given, taken as FieldTrip's default clusteralpha 0.05, a standard choice that could change the cluster's extent; (T3) the rate normalized as B or in spikes per second, as in fig3c. Reproduced when, under the readings, a cluster with remembered above forgotten has cluster p < 0.025, begins within 65 ms of maintenance onset and ends within tolerance of 650 ms; partially reproduced when such a significant cluster exists but its bounds are outside these.
fig3f-amygdala-no-cluster
In amygdala category neurons, no period of the maintenance period shows a significant difference between later remembered and forgotten pictures.
In the amygdala, we did not observe any significant cluster throughout the entire maintenance period of the WM task (Fig. 3f; cluster-p = 0.16)
Results, Persistent activity of category neurons in the hippocampus, paragraph 3; Figure 3f
How it will be judged
A claim of absence, judged on the test of fig3e-early-delay-cluster applied to amygdala category neurons (the high-confidence conditions of fig3d), with the same readings. Reproduced when no cluster has cluster p < 0.025 on either side; not reproduced when one does. The smallest cluster p is reported.
fig3g-nonpreferred-no-effect
Maintenance activity of category neurons while non-preferred pictures are held does not predict later LTM accuracy or confidence.
None of the main effects nor interactions in either of the mixed-effects GLMs showed any significant relationship between FR and the factors tested (all p > 0.06)
Results, Persistent activity of category neurons in the hippocampus, paragraph 4; Figure 3g
How it will be judged
A claim of absence, judged on the paper's test, the model of fig3a-area-interactions (LTMacc * conf * area, both areas) on the rows of non-preferred pictures in place of preferred ones (Fig. 3g legend). Open, which rows, (a) every non-preferred picture of the code's table (pref false), or (b) only pictures of trials that held no preferred picture. Reproduced when no fixed effect other than the intercept has p < 0.05 under every reading; not reproduced when any does under every reading. The coefficients are reported.
fig3h-encoding-no-effect
Activity of category neurons during the presentation of the first encoding picture, for preferred pictures, does not predict later LTM accuracy or confidence.
or for the visually evoked response of category neurons when images were shown on the screen during encoding (Fig. 3h; encoding 1; 0 – 2s after picture onset; preferred trials only). None of the main effects nor interactions in either of the mixed-effects GLMs showed any significant relationship between FR and the factors tested (all p > 0.06)
Results, Persistent activity of category neurons in the hippocampus, paragraph 4; Figure 3h; STAR Methods, Relating WM maintenance activity to LTM formation, paragraph 2
How it will be judged
A claim of absence, judged on the paper's test, the model of fig3a-area-interactions with the rate 0 to 2 s after the onset of encoding picture 1 in place of the maintenance rate, normalized as B, correct WM trials only (Fig. 3h legend). Open, which trials, (a) those whose first encoding picture is of the preferred category, with the LTM outcome of that picture, or (b) those holding a preferred picture in any encoding slot, with the LTM outcome of the preferred picture and the rate of the picture 1 window. Reproduced when no fixed effect other than the intercept has p < 0.05 under every reading; not reproduced when any does under every reading.
ms-neuron-counts
53 of the 351 hippocampal units (15.10%) are memory-selective, 25 of them (47%) firing more to familiar than novel pictures.
Out of the 351 recorded cells, n = 53 (15.10 %) were MS cells. Of these, 25 (47%) responded more to familiar than novel items, with the remaining responding more to novel than to familiar items.
Results, Activity during WM maintenance is linked to subsequent memory signal, paragraph 2
How it will be judged
Selection M on the hippocampal units of U in the sessions of S. Reproduced when both counts are within tolerance. The amygdala count (68 in Table S1) is reported, not tested.
fig4b-ms-confidence-auc
Hippocampal memory-selective neurons separate old from new pictures better for high-confidence than for low-confidence decisions (higher ROC AUC), so they carry a memory-strength signal.
The response of MS neurons differed significantly more for high- as compared to low-confidence trials (Fig. 4b; t(40) = 2.89, p = 0.0039; 12 neurons excluded due to insufficient data in either one of the conditions, see Methods)
Results, Activity during WM maintenance is linked to subsequent memory signal, paragraph 2; Figure 4b; STAR Methods, Single-neuron AUC analysis
How it will be judged
Hippocampal MS neurons of M. Spike counts 0.2 to 1.2 s after LTM picture onset; per neuron an ROC with 25 thresholds linearly spaced from the minimal to the maximal count observed, AUC by the trapezoid rule; for FS neurons all true negatives against high-confidence true positives (rating 6) and, separately, against low-confidence true positives (4, 5); for NS neurons all true positives against high-confidence (1) and low-confidence (2, 3) true negatives (Methods, Single-neuron AUC analysis). The AUC is oriented so that the neuron's preferred response (old for FS, new for NS) above 0.5 means selectivity, the reading under which the paper's "AUC was significantly higher for high than low confidence" holds across both types. Open, which neurons are kept, (a) at least ten correct novel and ten correct familiar trials, as the Methods say, or (b) at least ten trials in each of the groups compared; O-M then applies. High against low AUC by the paired permutation t-test of P. Reproduced when, under both readings, high exceeds low at p < 0.05; the number of neurons kept (41 in the paper) is reported.
fig4c-overlap-chance
20 hippocampal neurons are both memory-selective and category neurons, an overlap no greater than chance.
Of the 53 MS neurons and the 104 category neurons, 20 neurons (37.8% of MS cells; 19.2 % of category neurons) were part of both groups ... This revealed that the overlap between category neurons and MS cells was not significantly higher than chance (Fig. 4c; p = 0.11)
Results, Activity during WM maintenance is linked to subsequent memory signal, paragraph 3; Figure 4c
How it will be judged
With the hippocampal category and MS sets of C and M, the overlap count; the null as the paper describes, drawing as many hippocampal neurons as there are category neurons, without replacement, 1000 times, and counting how many are MS; p as the fraction of draws with an overlap at least the observed (a standard choice; the seed likewise). The absence part is judged on this test at p < 0.05. Reproduced when the overlap is within tolerance and p >= 0.05; partially reproduced when p >= 0.05 but the overlap is outside tolerance.
fig4d-category-no-memory-signal
During LTM retrieval, hippocampal category neurons code neither familiarity nor confidence for pictures of their preferred category.
None of the effects revealed significant modulations of FRs (Fig. 4d top; confidence: beta = 6.05, p = 0.28; familiarity: beta = 4.49, p = 0.71; confidence × familiarity: beta = −4.32, p = 0.57). To further confirm this result, we directly compared FRs of category neurons between preferred old and new items and did not observe a significant difference (Fig. 4d bottom; t(94) = −0.28, p = 0.78).
Results, Activity during WM maintenance is linked to subsequent memory signal, paragraph 4; Figure 4d
How it will be judged
A claim of absence, judged on the paper's two tests. Hippocampal category neurons of C; correct LTM trials showing a picture of the neuron's preferred category; rate 0.2 to 1.2 s after picture onset; the GLM of G with fixed effects confidence (continuous, 3 levels), familiarity (old or new) and their interaction; and, per neuron, mean rate of preferred old against preferred new pictures by the paired permutation t-test of P, after O-M. Open, (D1) the confidence coding, (a) the coding of the Fig. 3a code (sure new 1 to guessing new 3, guessing old 1 to sure old 3) or (b) certainty, sure 3, unsure 2, guessing 1 for both responses; (D2) the rate, (a) spikes per second in the window, the measure the text names, or (b) normalized as B to the WM baseline, the only baseline the paper defines; (D3) the trials of the t-test, (a) correct LTM trials only, as for the GLM in the sentence before, or (b) all trials of preferred old and new pictures, the sentence on the t-test not restricting them. Reproduced when no GLM effect and not the t-test reach p < 0.05 under every reading; not reproduced when any does under every reading.
fig4e-fs-strong-maintenance
Familiarity-selective hippocampal neurons respond more during LTM retrieval to familiar pictures that were held with strong, compared with weak, maintenance activity of a simultaneously recorded category neuron.
This analysis revealed that FRs of familiarity-selective MS cells during LTM retrieval of items that were previously accompanied by high persistent activity in the WM task were higher than those previously accompanied by low persistent activity (Fig. 4e left; Maintenance FR: beta = 14.63, p = 1.39 × 10−3)
Results, Activity during WM maintenance is linked to subsequent memory signal, paragraphs 5 and 6; Figure 4e; STAR Methods, Interaction between simultaneously recorded category- and memory-selective neurons
How it will be judged
Pairs of a hippocampal MS neuron (M) and a hippocampal category neuron (C) of the same session. Per category neuron, its maintenance rates (0 to 2.5 s) in correct WM trials holding a preferred picture later shown in the LTM test, median-split into strong and weak separately within load 1 and load 3 trials (Methods); a rate equal to the median goes to weak (a standard choice). The MS neuron's rate 0.2 to 1.2 s after onset of those pictures in the LTM test. The GLM of G with the fixed effect maintenance (strong or weak), FS pairs only; and, per pair, the mean MS rate after strong against after weak by the paired permutation t-test of P (O-M applied). Open (the Methods sentence on this GLM lost its variable names), (E1) the random intercept within patient is (a) the MS neuron or (b) the pair; (E2) the LTM trials are (a) all trials of those old pictures or (b) only those answered old; (E3) a neuron that is both MS and category is (a) paired with itself or (b) not; (E4) the rate is (a) spikes per second or (b) normalized as B to the WM baseline. Reproduced when, under the readings, both tests give strong above weak at p < 0.05; partially reproduced when one of them does under every reading. The pair counts and beta are compared with expected values as a report, since E3 changes the counts.
fig4f-ns-no-effect
The response of novelty-selective hippocampal neurons to familiar pictures does not depend on whether the paired category neuron's maintenance activity was strong or weak.
No significant relationship was observed for the activity of novelty-selective MS neurons (Fig. 4f left; Maintenance FR: beta = 0.09, p = 0.98)
Results, Activity during WM maintenance is linked to subsequent memory signal, paragraphs 5 and 6; Figure 4f
How it will be judged
A claim of absence, judged on the paper's two tests of fig4e-fs-strong-maintenance applied to NS pairs, with the same readings. Reproduced when neither test has p < 0.05 under every reading; not reproduced when either does under every reading. The number of NS pairs (122 in the paper) and the beta are reported.

Pipeline

Paper: done (1 of 1)
Claims: done
Claims check: done (ready)
Attempt: todo
Questions: none
Verified: none