{ "dataset": { "archive": "dandi", "id": "000673", "label": "DANDI:000673" }, "page": "datasets/dandi/000673/", "generated": "2026-11-05T07:41:23+00:00", "caution": "A failure to reproduce is not a finding that the paper is wrong. It may be a failure of the attempt.", "outcome_counts": { "reproduced": 13, "partially_reproduced": 5, "not_reproduced": 6, "blocked": 33 }, "outcome_counts_with_outside_data": { "reproduced": 0, "partially_reproduced": 0, "not_reproduced": 0, "blocked": 0 }, "outcome_counts_elsewhere": { "reproduced": 0, "partially_reproduced": 0, "not_reproduced": 0, "blocked": 0 }, "outcome_counts_links": { "reproduced": 0, "partially_reproduced": 0, "not_reproduced": 0, "blocked": 0 }, "attempts": [ { "id": "000673-daume-2024-01", "dataset": { "archive": "dandi", "id": "000673", "version": "0.240118.2135", "doi": "10.48324/dandi.000673/0.240118.2135" }, "key": "dandi:000673", "label": "DANDI:000673", "dataset_version": "0.240118.2135", "paper": "10.1038/s41586-024-07309-z", "paper_title": "Control of working memory by phase\u2013amplitude coupling of human hippocampal neurons", "date": "2026-10-31", "kind": "agent", "accountable": "CatalystNeuro", "agent": { "harness": "Claude Code", "model": "claude-opus-5-5", "protocol": "pilot-12" }, "runner": { "platform": "GitHub Actions", "url": "https://github.com/catalystneuro/encore/actions/runs/41183417608" }, "tier": "runs", "deprecated": "", "deprecated_on": null, "entry": "000673-daume-2024", "entry_title": "Phase-amplitude coupling and working memory in human hippocampus (Daume et al. 2024)", "description": "Tests the findings of Daume et al. (2024) on dandiset 000673, which holds 44 sessions of a Sternberg working memory task in 36 epilepsy patients, with 1,454 single neurons and 1,922 microwire LFP channels from the hippocampus, amygdala, pre-SMA, dACC and vmPFC. It tests the cohort, behaviour, category neurons and their maintenance activity, and the LFP theta-gamma phase-amplitude coupling, its load effect and its controls, single-trial PAC against reaction time and category neuron firing, and the selection and firing of PAC neurons. The spike-field coherence, noise correlation and decoding analyses were not run.", "notebooks": [ { "path": "reproduce_daume_2024.ipynb", "title": "Behaviour and single neurons (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024.html" }, { "path": "reproduce_daume_2024_lfp.ipynb", "title": "LFP phase-amplitude coupling (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024_lfp.html" }, { "path": "reproduce_daume_2024_lfp_controls.ipynb", "title": "Hippocampal PAC controls (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024_lfp_controls.html" }, { "path": "reproduce_daume_2024_pac_neurons.ipynb", "title": "PAC neurons (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024_pac_neurons.html" } ], "outcomes": [ { "claim": "cohort-counts", "statement": "The study recorded 36 patients in 44 sessions, with 1,454 single neurons (360 hippocampus, 496 amygdala, 204 pre-SMA, 188 dACC, 206 vmPFC) and 1,922 microwire LFP channels (586, 421, 283, 325 and 307).", "source": "Methods, Patients and Spike sorting; Results, Hippocampal PAC is modulated by WM load, paragraph 1", "quote": "We isolated 360 neurons in the hippocampus, 496 in the amygdala, 204 in the pre-SMA, 188 in the dACC and 206 in the vmPFC. Of the LFP channels, 586 channels were in the hippocampus, 421 in the amygdala, 283 in the pre-SMA, 325 in the dACC and 307 in the vmPFC", "statistics": [], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "44 sessions of 36 patients (21 female, 15 male, age 40.47 +/- 13.76 years), 1,454 units (hippocampus 360, amygdala 496, pre-SMA 204, dACC 188, vmPFC 206) and 1,922 LFP channels (586, 421, 283, 325, 307), every count equal to the paper's. sub-20 has no LFP. Units of a patient recorded in several sessions are counted once per session, as the paper's total does.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig1d-rt-load", "statement": "Patients responded more slowly in load 3 than in load 1 trials (1.46 s against 1.33 s).", "source": "Results, Main, paragraph 3; Figure 1d", "quote": "the participants responded slower (1.46 s versus 1.33 s; t 43 = 6.42, P < 0.001)", "statistics": [ { "test": "two-sided permutation-based t-test, 10,000 permutations", "statistic": "t43 = 6.42", "p": 0.0001, "n": "44 sessions" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "Mean over 44 sessions of the median RT of correct trials, 1.233 s in load 1 and 1.360 s in load 3, both within 10% of 1.33 and 1.46 s; paired permutation t-test t43 = 5.33, p = 0.0001 (paper t43 = 6.42); load 3 slower in 36 of 44 sessions and 29 of 36 patients. With the mean RT per session (variant) load 1 is 1.633 s and load 3 1.596 s (p = 0.66); RTs reach 74.2 s, 69 trials above 10 s.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig1d-accuracy-load", "statement": "Patients were less accurate in load 3 than in load 1 trials (91.60% against 95.71%).", "source": "Results, Main, paragraph 3; Figure 1d", "quote": "less accurately (91.60% versus 95.71%; t 43 = \u22124.45, P < 0.001) in load 3 trials compared with load 1 trials", "statistics": [ { "test": "two-sided permutation-based t-test, 10,000 permutations", "statistic": "t43 = -4.45", "p": 0.0001, "n": "44 sessions" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "95.10% in load 1 and 91.02% in load 3 (paper 95.71 and 91.60%); t43 = -5.24, p = 0.0001 (paper t43 = -4.45); load 3 less accurate in 32 of 44 sessions and 27 of 36 patients.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "behaviour-overall-accuracy", "statement": "Mean accuracy was 93.66 +/- 7.04% (s.d. over sessions), and 78.34 +/- 20.09% of errors were false negatives.", "source": "Results, Main, paragraph 3", "quote": "Mean accuracy was 93.66 \u00b1 7.04% (mean \u00b1 s.d.; 78.34 \u00b1 20.09% of all errors were false negatives)", "statistics": [], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "Accuracy 93.05 +/- 7.22% over 44 sessions (range 74.8 to 100%); false negatives 76.08 +/- 20.11% of errors over the 41 sessions with errors (75.29% of all 425 errors pooled). Both within the paper's s.d.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "category-cell-counts", "statement": "Category neurons were found above chance in the hippocampus (89, 24.72%), amygdala (181, 36.49%) and vmPFC (37, 17.96%), but not in the pre-SMA and dACC.", "source": "Results, Category cells support WM maintenance, paragraph 1; Extended Data Fig. 4a", "quote": "We identified such cells at numbers higher than expected by chance within the hippocampus (89 neurons (24.72%)), amygdala (181 neurons (36.49%)) and vmPFC (37 neurons (17.96%)) but not within the pre-SMA and dACC", "statistics": [ { "test": "right-sided permutation test, 500 label shuffles", "p": 0.002, "corrected": "Bonferroni over five areas (99th percentile)", "n": "hippocampus, amygdala and vmPFC" } ], "kind": "finding", "inferred": "", "outcome": "partially_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "Category neurons 94 of 360 in the hippocampus (26.11%), 181 of 496 in the amygdala (36.49%), 40 of 206 in the vmPFC (19.42%), 18 of 204 in the pre-SMA and 13 of 188 in the dACC, all within 10% of the paper's 89, 181, 37, 18 and 14. Against 500 label shuffles the hippocampus, amygdala and vmPFC exceed the 99th percentile (p = 0.002 each) and the dACC does not (p = 0.088), but the pre-SMA's 18 also exceeds its 99th percentile of 17 (p = 0.010), where the paper reports no excess. With a post hoc threshold of two-sided p < 0.05 the counts are 87, 174, 41, 18 and 14. The pre-SMA sits at the boundary - 18 is one above the 99th percentile and p = 0.010 equals the threshold - so with 500 shuffles neither form of the rule separates it clearly from chance.", "deviations": [ "The shuffle permutes the category labels of all presentations of a session together, one shuffle for all its units; the Methods say only that labels were shuffled across presentations. The permutation null of each test is drawn once per unit and reused for the shuffled labellings (the permutation distribution depends only on the values and group sizes)." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig3b-persistent-activity", "statement": "MTL category neurons stayed above baseline during maintenance in both preferred and non-preferred trials, and fired more when their preferred category was held in memory.", "source": "Results, Category cells support WM maintenance, paragraph 1; Figure 3b", "quote": "Category neurons ( n = 270) remained active (preferred (pref.) versus baseline (bsl), P = 0.0001; non-preferred (non-pref.), P = 0.0001) and retained their selectivity during the maintenance period, with FRs higher for preferred compared with non-preferred categories ( P = 0.001)", "statistics": [ { "about": "preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "270 neurons" }, { "about": "non-preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "270 neurons" }, { "about": "preferred against non-preferred", "test": "two-sided permutation-based t-test", "statistic": "t269 = 2.93", "p": 0.001, "n": "270 neurons" } ], "kind": "finding", "inferred": "", "outcome": "partially_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "275 MTL category neurons (paper 270). Maintenance rates exceed baseline in preferred and non-preferred trials under both readings of O6 (p = 0.0001 each). Preferred exceeds non-preferred in % of baseline (32.0 against 20.5%, t = 2.26, p = 0.0013; paper t269 = 2.93, p = 0.001) but not in Hz (2.81 against 2.76 Hz, t = 1.74, p = 0.072). The criterion asks for all three tests under both readings, so the claim holds in part. In 24 of 34 patients the preferred rate is higher. Under reading (b) alone, adopted for fig3c and fig3d, all three tests hold.", "deviations": [ "Rates are compared in Hz and in % of each neuron's baseline (O6); the paper does not say which its tests use." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf4b-persistent-activity-by-area", "statement": "Persistent and selective maintenance activity of category neurons held in the hippocampus and amygdala separately, but vmPFC category neurons returned to baseline and did not differ between preferred and non-preferred trials.", "source": "Extended Data Fig. 4b legend", "quote": "The activity of category neurons in the vmPFC (right) during the delay period was not significantly larger than that during baseline ... Also, their FRs did not differ significantly between when the preferred and the non-preferred category of a cell was maintained in WM (t(36) = 1.03, p = 0.32, BF 01 = 3.47)", "statistics": [ { "about": "hippocampus, preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "89 neurons" }, { "about": "hippocampus, non-preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "89 neurons" }, { "about": "amygdala, preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "181 neurons" }, { "about": "amygdala, non-preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "181 neurons" }, { "about": "hippocampus, preferred against non-preferred", "test": "two-sided permutation-based t-test", "p": 0.025, "n": "89 neurons" }, { "about": "amygdala, preferred against non-preferred", "test": "two-sided permutation-based t-test", "p": 0.037, "n": "181 neurons" }, { "about": "vmPFC, preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.12, "n": "37 neurons" }, { "about": "vmPFC, non-preferred against baseline", "test": "two-sided permutation-based t-test", "p": 0.26, "n": "37 neurons" }, { "about": "vmPFC, preferred against non-preferred", "test": "two-sided permutation-based t-test", "statistic": "t(36) = 1.03", "p": 0.32, "n": "37 neurons" } ], "kind": "finding", "inferred": "", "outcome": "partially_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "In % of baseline the hippocampus (94 neurons) and amygdala (181) pass all three tests (baseline tests p <= 0.0001; preferred against non-preferred p = 0.019 and 0.039; paper 0.025 and 0.037) and the vmPFC (40) passes none (p = 0.064, 0.24, 0.12). In Hz the vmPFC again passes none, but preferred against non-preferred fails in the hippocampus (p = 0.24) and amygdala (p = 0.18). The criterion, through fig3b, asks for both readings, so the claim holds in part.", "deviations": [ "As fig3b-persistent-activity (O6 readings)." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig3c-category-load", "statement": "In preferred trials, MTL category neurons fired more during maintenance in load 1 than in load 3, while no load effect appeared in non-preferred trials.", "source": "Results, Category cells support WM maintenance, paragraph 2; Figure 3c; Extended Data Fig. 5a", "quote": "The activity of category neurons in the MTL was modulated by load, with FRs higher in load 1 than in load 3 in preferred trials ( t 269 = 2.65, P = 0.004; Fig. 3c ), but not in non-preferred trials ( t 269 = \u22121.46, P = 0.14", "statistics": [ { "about": "preferred trials, load 1 against load 3", "test": "two-sided permutation-based t-test", "statistic": "t269 = 2.65", "p": 0.004, "n": "270 neurons" }, { "about": "non-preferred trials, load 1 against load 3", "test": "two-sided permutation-based t-test", "statistic": "t269 = -1.46", "p": 0.14, "n": "270 neurons" } ], "kind": "finding", "inferred": "", "outcome": "not_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "In % of baseline, the reading adopted, preferred trials do not differ between loads (35.29 against 30.85%, t = 0.47, p = 0.77; paper t269 = 2.65, p = 0.004) and non-preferred trials do not either (p = 0.26). In Hz the claim holds (preferred 2.90 against 2.78 Hz, t = 2.20, p = 0.022; non-preferred p = 0.20). The question of the units is for the paper's authors (finding on O6).", "deviations": [ "Assumed that the tests use the percentage change from baseline. The readings give different outcomes. The Fig. 3 and Fig. 4 legends say \"FRs are shown as the percentage change compared with the baseline\" and print the P values with those panels (Fig. 4f gives \"correct, P = 0.0001; incorrect, P = 0.0001\" against baseline for the plotted values), so the tests are taken to be on the plotted values. The data support this - with at least one incorrect trial, the % of baseline reading gives exactly the paper's n for the frontal areas in Extended Data Fig. 8h (201, 180 and 201; neurons with no baseline spike drop out), and the Hz reading does not (204, 183, 202)." ], "assumed": true, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig3d-category-correct-incorrect", "statement": "MTL category neurons fired more during maintenance in correct than in incorrect trials.", "source": "Results, Category cells support WM maintenance, paragraph 2; Figure 3d", "quote": "FRs were higher in correct as compared to incorrect trials across both load conditions ( t 245 = 2.43, P = 0.02", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t245 = 2.43", "p": 0.02, "n": "246 neurons" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "In % of baseline over all trials, 251 neurons with at least one incorrect trial (paper 246) fire more in correct than incorrect trials (26.56 against 9.27%, t = 2.50, p = 0.0035; paper t245 = 2.43, p = 0.02); with at least two incorrect trials 229 neurons, p = 0.010; three, 209, p = 0.055. In Hz the test fails under every reading (p >= 0.12), and on preferred trials only it fails in both units (p >= 0.10).", "deviations": [ "Assumed the percentage change from baseline (as fig3c-category-load); in Hz the claim is not reproduced. The minimum number of incorrect trials (O7) is not given; 1, 2 and 3 are shown and the first matches the paper's n." ], "assumed": true, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "category-no-rt-effect", "statement": "Category neuron maintenance rates did not differ between fast and slow RT trials in the hippocampus or amygdala.", "source": "Results, Category cells support WM maintenance, paragraph 2", "quote": "There was no significant difference in the FRs between fast and slow RT trials (median split; computed per load condition and then averaged across loads for the preferred category and correct trials only; hippocampus, t 88 = 0.73, P = 0.47; amygdala, t 180 = 1.38, P = 0.18)", "statistics": [ { "about": "hippocampus", "test": "two-sided permutation-based t-test", "statistic": "t88 = 0.73", "p": 0.47, "n": "89 neurons" }, { "about": "amygdala", "test": "two-sided permutation-based t-test", "statistic": "t180 = 1.38", "p": 0.18, "n": "181 neurons" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024.ipynb", "notes": "Correct preferred trials, median split within load. Hippocampus (94 neurons) fast 2.76 against slow 2.77 Hz, p = 0.85 (in % of baseline p = 0.22); amygdala (181) 2.92 against 2.83 Hz, p = 0.098 (p = 0.51). No test reaches p < 0.05.", "deviations": [ "Trials at or below the load's median RT are fast, above it slow (the split of ties is not stated; a standard choice)." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf8fgh-frontal-controls", "statement": "Across all vmPFC neurons, neither within-region SFC nor SFC to hippocampal LFPs differed between loads, and maintenance firing rates of all neurons in the pre-SMA, dACC and vmPFC did not differ between correct and incorrect trials.", "source": "Extended Data Fig. 8f, g, h legends; Results, PAC neurons phase lock to the frontal cortex, paragraph 1", "quote": "We did not find significant differences between the load conditions for ( f ) within-region SFC or ( g ) cross-regional SFC to hippocampal LFPs across all neurons from the vmPFC ... FRs for all neurons recorded in the three frontal areas were not significantly different between correct and incorrect trials during the delay period", "statistics": [], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": "reproduce_daume_2024.ipynb", "notes": "Panels f and g need spike-field coherence, which was not run in this attempt (fig3f-gamma-sfc-preferred). Panel h was run - with at least one incorrect trial, pre-SMA (204 neurons in Hz, 201 in % of baseline; paper 201) correct 2.91 against incorrect 3.07 Hz, p = 0.039, and in % of baseline p = 0.54; dACC (183 and 180; paper 180) p = 0.83 and 0.71; vmPFC (202 and 201; paper 201) p = 0.60 and 0.97. In % of baseline, the reading adopted for the firing rate tests (fig3c-category-load), panel h holds and its n equal the paper's in all three areas; in Hz the pre-SMA differs with one incorrect trial and not with two or three (p = 0.089, 0.11).", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig2a-theta-gamma-pac", "statement": "Across all recorded channels, PAC during maintenance was strongest between theta phase (3-7 Hz) and the amplitude of a low (30-55 Hz) and a high (70-140 Hz) gamma band.", "source": "Results, Hippocampal PAC is modulated by WM load, paragraph 1; Figure 2a", "quote": "PAC was strongest between the phase in the theta range (3\u20137 Hz) and the amplitude in two different gamma-frequency bands ... a lower (30\u201355 Hz) and a higher gamma range (70\u2013140 Hz)", "statistics": [], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Under artefact reading (a), the mean z-scored MI over 1,922 channels (paper 1,917) is largest (2.16) at phase 4 Hz and amplitude 80 Hz; at phase 4 Hz the amplitude profile has local maxima at 50 Hz (within 30-55) and 80 Hz (within 70-140), and at 6 Hz at 40 and 80 Hz. Under reading (b), computed for phase centres 2 to 8 Hz, the mean over 1,918 channels is largest (1.73) at 4 Hz and 80 Hz, with local maxima at 45 and 80 Hz. Both readings meet the criterion. Per area the maximum is at phase 4 Hz in the hippocampus (75 Hz) and amygdala (50 Hz under a, 45 Hz under b) and at 2 Hz in the frontal areas.", "deviations": [ "The visual inspection of the artefact rejection cannot be repeated (O1); both automated readings were run. Under reading (b) the notebook computes phase centres 2 to 8 Hz only, the ones its other claims use.", "Filters follow pop_eegfiltnew's default design (windowed sinc, Hamming) and firfilt's edge padding, reimplemented in Python (sbcat.py); the 200 surrogates draw a random permutation of the trials, as cfc_tort_comodulogram.m does." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig2b-pac-channel-proportions", "statement": "137 of 586 hippocampal channels showed significant theta-high-gamma PAC; the proportions of such channels in the hippocampus, amygdala and vmPFC exceeded the trial-shuffled null, while pre-SMA and dACC had few such channels.", "source": "Results, Hippocampal PAC is modulated by WM load, paragraphs 1 and 2; Figure 2b", "quote": "For theta to high-gamma PAC, 137 out of 586 hippocampal channels showed significant PAC across all correct trials from both load conditions", "statistics": [ { "test": "right-sided permutation test against trial-shuffled surrogates", "p": 0.005, "n": "channels per area; hippocampus, amygdala and vmPFC" } ], "kind": "finding", "inferred": "", "outcome": "partially_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Under artefact reading (a), significant channels are 144 of 586 hippocampal (paper 137), 130 of 421 amygdala (130), 41 of 307 vmPFC (40), 17 of 283 pre-SMA (5) and 14 of 325 dACC (14); all but the pre-SMA are within tolerance. Every area's proportion exceeds the 99th percentile of its surrogate null (O4), p = 0.005 (the smallest possible with 200 surrogates), including the pre-SMA and dACC. Under reading (b) the counts are 125, 112, 27, 12 and 8, with the amygdala, vmPFC, pre-SMA and dACC out of tolerance; the proportions again exceed the null.", "deviations": [ "Artefact rejection without the visual inspection (O1), under two readings of the span it is computed on. The counts are judged under reading (a), adopted for the claims where the readings disagree. On the one channel whose rejected trials the paper's code ships (artif_trials.mat with NWB_samplePAC_SU.m, sub-26 session 1, LFP channel 33), the paper removes 17 trials; reading (a) removes 4, all among them, and reading (b) removes 9, 5 of which the paper kept." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig2c-hippocampus-load", "statement": "In the 137 hippocampal channels with significant PAC, theta-high-gamma PAC was weaker in load 3 than in load 1 trials.", "source": "Results, Hippocampal PAC is modulated by WM load, paragraph 1; Figure 2c (left)", "quote": "This comparison revealed significantly weaker theta\u2013high-gamma PAC in load 3 compared with load 1 trials ( t 136 = \u22124.26, P < 0.001", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t136 = -4.26", "p": 0.0005, "corrected": "FDR over five brain areas", "n": "137 channels" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "In the significant hippocampal channels z-scored theta-high-gamma MI is lower in load 3 under both readings - (a) 144 channels, 4.62 against 3.80, t = -4.16, p = 0.0002, FDR p = 0.0010, load 1 higher in 99 channels and in 18 of 23 patients; (b) 125 channels, t = -3.05, FDR p = 0.014. Paper t136 = -4.26, p < 0.001.", "deviations": [ "Artefact rejection without the visual inspection (O1); 144 or 125 channels where the paper has 137." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig2c-no-load-effect-other-areas", "statement": "Theta-high-gamma PAC did not differ between loads in the amygdala, vmPFC, pre-SMA or dACC channels with significant PAC.", "source": "Results, Hippocampal PAC is modulated by WM load, paragraph 2; Figure 2c (middle and right)", "quote": "theta\u2013high-gamma PAC did not differ significantly between the two load conditions in neither of these two areas (Fig. 2c ; t 129 = 1.43, P = 0.38, Bayes factor (BF) 01 = 1.40; and t 39 = 0.16, P = 0.87, BF 01 = 5.84, respectively)", "statistics": [ { "about": "amygdala", "test": "two-sided permutation-based t-test", "statistic": "t129 = 1.43", "p": 0.38, "corrected": "FDR over five brain areas", "n": "130 channels" }, { "about": "vmPFC", "test": "two-sided permutation-based t-test", "statistic": "t39 = 0.16", "p": 0.87, "corrected": "FDR over five brain areas", "n": "40 channels" }, { "about": "pre-SMA", "test": "two-sided permutation-based t-test", "statistic": "t4 = -0.16", "p": 0.87, "corrected": "FDR over five brain areas", "n": "5 channels" }, { "about": "dACC", "test": "two-sided permutation-based t-test", "statistic": "t13 = -0.82", "p": 0.73, "corrected": "FDR over five brain areas", "n": "14 channels" } ], "kind": "finding", "inferred": "", "outcome": "not_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "With the paper's test and FDR over the five areas, under reading (a) the pre-SMA (17 channels; load 1 1.81, load 3 6.87; FDR p = 0.0017) and the amygdala (130 channels; 3.86 against 4.20, load 3 higher; FDR p = 0.0475) differ between loads; the dACC (FDR p = 0.37) and vmPFC (0.93) do not. Under reading (b) the pre-SMA (12 channels, FDR p = 0.014) and dACC (8 channels, load 1 higher, FDR p = 0.014) differ and the amygdala does not (FDR p = 0.086). The pre-SMA has 17 or 12 significant channels where the paper has 5, and load 1 is higher in only 3 of the 17. The 17 channels come from six patients; four of sub-4 have z values of 19 to 25 in load 3 against 3.5 to 6.8 in load 1, and seven of sub-11 are also higher in load 3. The effect does not rest on a few outliers alone - Wilcoxon p = 0.0011, and without the three channels of largest load 3 excess p = 0.0045 (reading a; under b, p = 0.059).", "deviations": [ "Artefact rejection without the visual inspection (O1). The channels that decide the pre-SMA result are channels that the paper's visual inspection may have removed; the files do not say." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf2b-hippocampus-load-patients", "statement": "Averaging significant PAC channels within each patient, hippocampal PAC was again stronger in load 1 than in load 3, and no other area showed a load difference.", "source": "Extended Data Fig. 2b legend", "quote": "Only in the hippocampus again, PAC was stronger for load 1 as compared to load 3 (n = 23 patients, p = 0.0049), observable in almost each single patient. No significant differences were found between the load conditions in other regions", "statistics": [ { "test": "two-sided permutation-based t-test", "p": 0.0049, "n": "23 patients" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Under artefact reading (a), 23 patients (paper 23) have hippocampal PAC channels; the patient means are higher in load 1 (3.97 against 3.23, t = 2.26, p = 0.0095; paper p = 0.0049; 18 of 23 patients), and no other area differs (amygdala p = 0.10, pre-SMA 0.31, dACC 0.99, vmPFC 0.92). Under reading (b) the hippocampus gives p = 0.11 over 20 patients and no area differs.", "deviations": [ "Assumed artefact reading (a), automated rejection on the -0.5 to 3 s epochs. The two readings give different outcomes. Reading (a) removes 16.0 +/- 12.0 trials per channel (11.4%), below the paper's 20.4 +/- 13.9 (14.6%) that also include the trials removed by visual inspection; reading (b), on whole trials, removes 36.2 +/- 22.8 (25.9%), more than the paper's total. On the one channel whose rejected trials the paper's code ships (artif_trials.mat with NWB_samplePAC_SU.m, sub-26 session 1, LFP channel 33), the paper removes 17 trials; reading (a) removes 4, all among them, and reading (b) removes 9, 5 of which the paper kept." ], "assumed": true, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf2c-low-gamma-no-load-effect", "statement": "Theta-low-gamma (30-55 Hz) PAC did not differ between loads in any region.", "source": "Results, Hippocampal PAC is modulated by WM load, paragraph 2; Extended Data Fig. 2c", "quote": "For PAC involving the low gamma band (30\u201355 Hz), there were no significant differences between the load conditions in any of the regions", "statistics": [], "kind": "finding", "inferred": "", "outcome": "not_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "With channels selected on theta-low-gamma PAC (a - 152, 159, 22, 27, 48 channels; paper 148, 155, 10, 22, 48), the pre-SMA differs between loads (load 3 higher, p = 0.0029; FDR variant p = 0.0145); no other area does (p >= 0.11). Under reading (b) the pre-SMA again differs (13 channels, p = 0.0033) and no other area does.", "deviations": [ "Artefact rejection without the visual inspection (O1)." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "tables1-slow-fast-theta", "statement": "Split into slow (2-5 Hz) and fast (5-9 Hz) theta, PAC differed between loads only in the hippocampus and only with high gamma, most strongly for fast theta (t116 = -3.05, p = 0.0015; slow theta, t147 = -2.31, p = 0.0163).", "source": "Supplementary Table S1; Supplementary Discussion, Related work on role of theta", "quote": "PAC differed significantly only in the hippocampus and only for high gamma power combined with either low or high theta frequency bands, with the strongest differences in the fast theta band", "statistics": [ { "about": "hippocampus, fast theta to high gamma", "test": "two-sided permutation-based t-test, uncorrected", "statistic": "t(116) = -3.05", "p": 0.0015 }, { "about": "hippocampus, slow theta to high gamma", "test": "two-sided permutation-based t-test, uncorrected", "statistic": "t(147) = -2.3141", "p": 0.0163 } ], "kind": "finding", "inferred": "", "outcome": "not_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Under all four combinations of band reading (A phase centres, B direct band filters) and artefact reading, both hippocampal high-gamma tests give p < 0.05 with load 1 higher (fast theta t = -3.44, -3.27, -3.03, -2.50; slow theta t = -2.43, -2.21, -2.46, -2.22), but 8, 5, 6 and 6 of the 20 tests are significant, always including pre-SMA tests with load 3 higher, and under (a, A) also the hippocampal slow theta to low gamma test (t = -3.21). The pre-SMA fast theta to high gamma test, absent from Table S1, has 7 to 21 channels here.", "deviations": [ "Artefact rejection without the visual inspection (O1). For band reading B the amplitude bandwidth is twice the band's centre frequency, by analogy with the comodulogram (not stated)." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig2f-rt-single-trial-pac", "statement": "In the hippocampus, stronger single-trial theta-high-gamma PAC went with faster RTs, while no such relation held in the amygdala or vmPFC.", "source": "Results, Hippocampal PAC is modulated by WM load, paragraph 2; Figure 2f; Supplementary Table S2", "quote": "In the hippocampus, faster RTs were associated with stronger single-trial estimates of TG-PAC", "statistics": [ { "about": "hippocampus, PAC term", "test": "mixed-effects GLM", "statistic": "estimate -49.07, t = -4.01", "p": 6e-05, "n": "137 channels" }, { "about": "amygdala, PAC term", "test": "mixed-effects GLM", "statistic": "estimate -5.15, t = -0.70", "p": 0.48, "n": "130 channels" }, { "about": "vmPFC, PAC term", "test": "mixed-effects GLM", "statistic": "estimate 24.51, t = 1.18", "p": 0.24, "n": "40 channels" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "methods_underspecified", "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Refitted after round 2 with load 3 as the reference level, as Table S2 codes load (its load term is named \"Load1\"), so the PAC term is the slope in load 3 trials. With RT in s under artefact reading (a) (144 hippocampal channels, 16,103 correct clean trials), neither reading of O5 meets the criterion - under O5 (a) the hippocampal term is -53.0, t = -2.43, p = 0.015, but the amygdala term is also negative (p = 0.027); under O5 (b) the hippocampal term has p = 0.099. RT in s does not fit Table S2 - its intercept is 1.70 (O5 a) against the paper's 0.38. With RT z-scored over each session's correct trials the intercept is 0.28 and the Load1 term -0.32 (paper 0.38 and -0.48); under O5 (a) the hippocampal term is -49.39, t = -4.19, p = 2.8e-5 (paper -49.07, t = -4.01, p = 6e-5) but the amygdala term is -19.6, p = 0.015 (paper p = 0.48), and under O5 (b) the hippocampal term is negative (p = 0.029) with amygdala p = 0.20 and vmPFC p = 0.58, meeting the criterion. log10 RT splits the same way (O5 a, amygdala p = 0.036; O5 b, hippocampus p = 0.0045, amygdala 0.22, vmPFC 0.87). With RTs below 10 s (87 trials dropped), both readings meet the criterion (hippocampus p = 2.8e-7 and 0.00046; amygdala 0.097 and 0.097; vmPFC 0.99 and 0.61). Under artefact reading (b) every response scale and both O5 readings meet it (hippocampus p <= 0.031, amygdala >= 0.479, vmPFC >= 0.26). The slope in load 1 trials, which the first coding tested, has p = 0.25 and 0.64 with RT in s. On the scales that fit Table S2 the two readings of O5 disagree, which the criterion makes methods_underspecified; the change from not_reproduced was agreed in the third verification round. Table S2's PAC estimate (-49.07) points to O5 (a), under which the outcome would be not reproduced; this is asked of a reader.", "deviations": [ "The single-trial PAC is not described (O5); both readings of the claims file were run. Mixed model fitted with statsmodels MixedLM (REML), random intercepts for patient and for channel within patient, in place of MATLAB's fitglme. Load coded with load 3 as the reference (Supplementary Table S2, \"Load 1 are tested against load 3 trials\", with a term named \"Load1\").", "Table S2 gives no unit or transform for RT. RT is fitted in s (the claims file's reading), as log10 RT and z-scored over each session's correct trials; the hippocampal trials under reading (a) have RTs up to 56.3 s (87 above 10 s), and a variant drops RTs above 10 s. The paper states no trimming.", "Assumed artefact reading (a), as for edf2b-hippocampus-load-patients; under reading (b) every RT scale and both readings of O5 meet the criterion, which would give reproduced." ], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf3-hippocampus-pac-controls", "statement": "The hippocampal load effect on PAC was not explained by theta or gamma power, theta waveform asymmetry, theta-gamma phase-phase coupling, preferred phase or PAC peak frequency.", "source": "Results, Hippocampal PAC is modulated by WM load, paragraph 2; Extended Data Fig. 3b, c, d, g, h", "quote": "we tested whether the differences in hippocampal PAC between loads could be explained by changes in power, theta waveform shape, cross-frequency phase\u2013phase coupling, preferred phase or PAC peak frequency but did not find evidence for any of those factors", "statistics": [], "kind": "finding", "inferred": "", "outcome": "not_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp_controls.ipynb", "notes": "Under artefact reading (a), over 144 PAC channels, two load tests give p < 0.05 - high-gamma power is higher in load 3 (Hilbert t = 2.09, p = 0.035; Welch t = 2.69, p = 0.006) and the gamma peak frequency is higher in load 3 (93.0 against 86.8 Hz, t = 3.04, p = 0.003; as the coordinate of the single largest bin, p = 0.34). Theta power (p = 0.23, 0.16), rise-decay and peak-trough symmetry (all cycles p = 0.34 and 0.50; burst cycles, present in both loads in only 8 channels, p = 0.17 and 0.79), phase-phase locking (1 to 21, p = 0.90; each centre pair, p = 0.82), the theta peak frequency (p = 0.49; largest bin p = 0.080) and the preferred phase t-test (p = 0.71) do not differ. The V test of the phase differences against 0 gives p < 0.0001 (mean resultant length 0.94, circular mean -1.6 degrees), which shows differences concentrated at 0; the criterion's rule would count it as a test with p < 0.05, which is asked of a reader (questions/000673.yaml). No asymmetry differs from 0.5 (p >= 0.072). With two load tests significant the criterion gives not reproduced. Under artefact reading (b) (125 channels) no load test gives p < 0.05 (high-gamma power p = 0.13 and 0.058, gamma peak frequency p = 0.17).", "deviations": [ "Assumed artefact reading (a), as for edf2b-hippocampus-load-patients; the readings give different outcomes.", "Theta waveform asymmetry from bycycle 1.2 with f_range 3-7 Hz and its default settings on each trial's -0.5 to 3 s epoch, cycles with their peak in 0 to 2.5 s, averaged per load over the trials of the PAC analysis (Methods, PAC - \"We averaged estimates for peak-to-trough as well as rise-to-decay asymmetries across cycles during the maintenance period from the same trials used for our PAC analysis within each load\"). Its burst detection fails under pandas 3 and is replaced by the same default thresholds applied to a writable copy. Whether only burst cycles count is not stated; both are shown. With the default thresholds 99.9% of trial windows have no burst cycle.", "PAC peak frequency (Extended Data Fig. 3h legend - \"recomputed PAC using a finer resolution for phase frequencies (i.e., a step size of 0.5 instead of 2 Hz) and determined the frequency bin for which PAC was maximal for the theta and the gamma band separately\") from raw MI per load, phase centres 3 to 7 Hz in 0.5 Hz steps and amplitude centres 70 to 140 Hz; read as the maximum of the mean over the other axis, with the coordinates of the largest bin as a variant. Raw MI is used as no surrogates are described for this panel.", "Phase-phase coupling under both readings of the criterion; for the second, n is the ratio of each amplitude centre to the phase centre, rounded to an integer." ], "assumed": true, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf3a-kappa-load", "statement": "The distribution of gamma amplitude over theta phase was wider (lower kappa) in load 3 than in load 1 across hippocampal PAC channels.", "source": "Extended Data Fig. 3a legend", "quote": "Across all PAC channels, kappa was significantly lower in load 3 compared to load 1 (n = 137 channels, t(136) = \u22123.7453, p = 0.0001) trials", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t(136) = -3.7453", "p": 0.0001, "n": "137 channels" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Kappa of the theta-high-gamma amplitude distribution is lower in load 3 (a - 144 channels, 0.0360 against 0.0326, t = -4.54, p = 0.0001, 96 channels lower; b - 125 channels, t = -3.11, p = 0.0024). Paper t136 = -3.75, p = 0.0001.", "deviations": [ "Kappa from CircStat's circ_kappa formula on the 18 bin centres weighted by the normalized amplitude, as the criterion says." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf3e-nesting", "statement": "110 of the 137 hippocampal PAC channels (80.29%) showed theta-high-gamma nesting, and PAC remained lower in load 3 than load 1 in those channels.", "source": "Extended Data Fig. 3e legend", "quote": "110 of the 137 significant PAC channels (80.29%) in the hippocampus showed nesting between high gamma and theta. When testing PAC between the load conditions after removing channels that did not show significant nesting, PAC was still significantly lower in load 3 than 1 (n = 110; t(109) = \u22124.10; p = 0.0001)", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t(109) = -4.10", "p": 0.0001, "n": "110 channels" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Aligning the 500 ms windows on the gamma peak in each run of the preferred theta bin, 105 of 144 hippocampal PAC channels (72.92%) are nested under artefact reading (a), within tolerance of 110, and PAC is lower in load 3 in them (4.29 against 3.48, t = -3.61, p = 0.0002; paper t109 = -4.10, p = 0.0001); under (b) 94 of 125 (75.20%), p = 0.0051. Aligning on the maximum of gamma amplitude, as the Methods' words read, only 9 of 144 (6.25%) are nested and the load test fails (p = 0.77). The count depends on the alignment and not on the window - with local maxima counted within 35 to 85 ms the peak alignment gives 104 to 114 nested channels (30 within 25 ms), and the amplitude alignment 4 to 15 for every window.", "deviations": [ "Assumed that the windows are aligned on the peak of the gamma-filtered signal. The Methods say \"the precise timepoint at which the concurrent instantaneous gamma amplitude (70-140 Hz) was maximal\"; aligned on amplitude maxima, gamma cycles of random phase cancel in the average, so few local maxima remain (most channels have 0 or 1), which cannot give the paper's 80%. Aligned on gamma peaks, the average keeps the gamma cycle, as the method of Vaz et al. (ref. 69) needs." ], "assumed": true, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf3f-raw-mi-load", "statement": "Raw (not z-scored) theta-gamma MI was also larger in load 1 than in load 3 in the hippocampal PAC channels.", "source": "Extended Data Fig. 3f legend", "quote": "Raw theta-gamma PAC was significantly larger in load 1 compared to load 3 (n = 137; t(136) = \u22124.0264, p = 0.0001)", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t(136) = -4.0264", "p": 0.0001, "n": "137 channels" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Under artefact reading (a), raw theta-high-gamma MI is larger in load 1 (2.43e-4 against 2.17e-4, t = -4.03, p = 0.0002, 99 of 144 channels; paper t136 = -4.03, p = 0.0001). Under reading (b) p = 0.066 (125 channels).", "deviations": [ "Assumed artefact reading (a), as for edf2b-hippocampus-load-patients; the readings give different outcomes. On the one channel whose rejected trials the paper's code ships (artif_trials.mat with NWB_samplePAC_SU.m, sub-26 session 1, LFP channel 33), the paper removes 17 trials; reading (a) removes 4, all among them, and reading (b) removes 9, 5 of which the paper kept." ], "assumed": true, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "methods-spectral-slope", "statement": "The slope of the log-log power spectrum did not differ between load 1 and load 3 trials in hippocampal channels.", "source": "Methods, LFP preprocessing", "quote": "The slope of log\u2013log power spectra did not differ significantly between load 1 and load 3 trials in hippocampal channels ( n = 586 channels; mean slope \u22121.7526 \u00b1 0.3902 versus \u22121.7517 \u00b1 0.3928, t 585 = \u22120.86, P = 0.39)", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t585 = -0.86", "p": 0.39, "n": "586 channels" } ], "kind": "finding", "inferred": "", "outcome": "not_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Over 586 hippocampal channels, the 2-150 Hz slope is -1.7260 +/- 0.4253 in load 1 and -1.7300 +/- 0.4314 in load 3 (paper -1.7526 and -1.7517, within tolerance), and the paired permutation test finds a difference (t = 2.66, p = 0.0073; artefact reading b, p = 0.0018). Over 30-150 Hz the slopes are -1.838 and -1.842 and the test does not find one (p = 0.12 and 0.28). The difference is 0.004, about 1% of the s.d. across channels.", "deviations": [ "Assumed the 2-150 Hz range. The frequency range and spectral estimate are not given; the two readings give different outcomes. The 2-150 Hz mean slope (-1.726) is closer to the paper's (-1.75) than the 30-150 Hz one (-1.84). Welch spectra (1 s Hann windows, 50% overlap) and the exclusion of 58-62 and 118-122 Hz are standard choices." ], "assumed": true, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig3f-gamma-sfc-preferred", "statement": "Hippocampal category neurons paired with PAC channels were more strongly phase-locked to local high-gamma LFP when their preferred category was held in memory; no such effect held in the amygdala, for non-PAC channels or in the theta band.", "source": "Results, Category cells support WM maintenance, paragraph 3; Figure 3f; Extended Data Fig. 5e", "quote": "In the hippocampus, high-gamma-band SFC was significantly stronger in preferred trials than in non-preferred trials for neuron-to-channel combinations that involved significant PAC channels (cluster P = 0.004; Fig. 3f )", "statistics": [ { "test": "two-sided cluster-based permutation t-test", "p": 0.004, "corrected": "Bonferroni over two MTL areas", "n": "151 neuron-channel combinations" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Not run in this attempt. Spike-field coherence (40 Morlet wavelets, 500 spike subsamples and 500 jittered surrogates per neuron-channel pair and condition, cluster tests with 10,000 permutations) was not written into a notebook, so no timing of it is shown and it is not blocked as too large. The data it needs (spikes, LFP, bundles inferred from location and coordinates) are in the dataset.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig3g-gamma-sfc-anova", "statement": "Gamma (70-140 Hz) SFC of hippocampal category neurons was stronger in preferred than non-preferred trials in both loads, with no main effect of load and no interaction.", "source": "Figure 3g legend; Results, Category cells support WM maintenance, paragraph 3", "quote": "Gamma (70\u2013140 Hz) SFC for hippocampal category neurons was stronger for preferred versus non-preferred trials in both load conditions (main effect preference, F 1,150 = 16.23, P = 0.0001; load 1, P = 0.003; load 3, P = 0.004). No main effect of load ( P = 0.25) or interaction ( P = 0.33) was found", "statistics": [ { "about": "main effect of preference", "test": "permutation-based 2 x 2 ANOVA", "statistic": "F1,150 = 16.23", "p": 0.0001, "n": "151 neuron-channel pairs" }, { "about": "load 1, preferred against non-preferred", "test": "two-sided permutation-based t-test", "statistic": "t150 = 3.14", "p": 0.003 }, { "about": "load 3, preferred against non-preferred", "test": "two-sided permutation-based t-test", "statistic": "t150 = 2.88", "p": 0.004 }, { "about": "main effect of load", "test": "permutation-based 2 x 2 ANOVA", "p": 0.25 }, { "about": "interaction", "test": "permutation-based 2 x 2 ANOVA", "p": 0.33 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig3f-gamma-sfc-preferred, which was not run.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf5d-sfc-high-gamma-power", "statement": "The preferred against non-preferred difference in gamma SFC held only for spikes during high hippocampal gamma power, not during low gamma power.", "source": "Results, Category cells support WM maintenance, paragraph 3; Extended Data Fig. 5d", "quote": "SFC was different between preferred and non-preferred trials only for spikes that occurred during high ( t 150 = 3.06, P = 0.002) but not low high-gamma band power in the hippocampus ( t 150 = 0.26, P = 0.85", "statistics": [ { "about": "high gamma power", "test": "two-sided permutation-based t-test", "statistic": "t150 = 3.06", "p": 0.002, "n": "151 neuron-channel pairs" }, { "about": "low gamma power", "test": "two-sided permutation-based t-test", "statistic": "t150 = 0.26", "p": 0.85, "n": "151 neuron-channel pairs" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig3f-gamma-sfc-preferred, which was not run.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf5c-theta-sfc-null", "statement": "Theta-band (3-7 Hz) SFC of hippocampal category neurons with PAC channels showed no effect of load or preference and no interaction.", "source": "Extended Data Fig. 5c legend; Results, Category cells support WM maintenance, paragraph 3", "quote": "When averaging theta band (3\u20137 Hz) SFC values for hippocampal category neurons paired with significant PAC channels, we did not observe a significant main effect for load or preference nor a significant interaction", "statistics": [], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig3f-gamma-sfc-preferred, which was not run.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf5fg-gamma-sfc-all-channels-patients", "statement": "The gamma SFC preference effect of hippocampal category neurons held across all channels, not only PAC channels, and at the patient level.", "source": "Extended Data Fig. 5f and 5g legends", "quote": "Patient-averaged gamma SFC was significantly higher for preferred trials, suggesting that the effect was not driven by a few channels or patients (n = 19, t(18) = 2.8512, p = 0.005)", "statistics": [ { "about": "all channels, hippocampus", "test": "two-sided cluster-based permutation t-test", "p": 0.007, "corrected": "Bonferroni over two MTL areas" }, { "about": "patient averages", "test": "two-sided permutation-based t-test", "statistic": "t(18) = 2.8512", "p": 0.005, "n": "19 patients" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig3f-gamma-sfc-preferred, which was not run.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig3h-pac-category-firing", "statement": "Trial by trial, single-trial theta-high-gamma PAC was weakly positively related to the spike count of hippocampal category neurons, but not of amygdala category neurons.", "source": "Results, Category cells support WM maintenance, paragraph 4; Figure 3h; Supplementary Table S3", "quote": "In the hippocampus, PAC was weakly but significantly positively correlated with the FR of category neurons", "statistics": [ { "about": "hippocampus, SC term", "test": "mixed-effects GLM", "statistic": "estimate 0.002, t = 2.40", "p": 0.017, "n": "151 neuron-channel combinations" }, { "about": "amygdala, SC term", "test": "mixed-effects GLM", "statistic": "estimate -0.001, t = -0.75", "p": 0.45, "n": "423 neuron-channel combinations" } ], "kind": "finding", "inferred": "", "outcome": "reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_lfp.ipynb", "notes": "Under artefact reading (a), 166 hippocampal pairs of a category neuron and a significant PAC channel of its bundle (52 neurons, 15 patients; paper 151) and 450 amygdala pairs (paper 423). Refitted after round 2 with load 3 as the reference level, as Table S3 codes load, so the SC term is the slope in load 3 trials. The hippocampal SC term is positive with p < 0.05 under every reading - all correct trials, O5 (a) t = 6.34, p = 2.3e-10, O5 (b) 0.0045, t = 3.09, p = 0.0020 (paper 0.002, t = 2.40, p = 0.017); preferred trials, O5 (a) t = 3.09, p = 0.0020, O5 (b) 0.0028, t = 2.16, p = 0.031. The amygdala term has p >= 0.326 under every reading (paper p = 0.45). Under O5 (b) on all trials the other terms are close to Table S3 - intercept -0.057 (paper -0.044), Load1 0.078 (0.076), SC x Load1 -0.0049, p = 0.0002 (-0.004, p = 0.010). The slope in load 1 trials, which the first coding tested, is not positive (all trials O5 b, p = 0.77; preferred O5 b, -0.0045, p = 0.015), and neither is the paper's (0.002 - 0.004). Under artefact reading (b) (145 and 410 pairs) the hippocampal term is positive with p <= 0.014 under three readings and p = 0.056 under the fourth (preferred, O5 b); the amygdala term has p >= 0.32 (one of its fits did not converge). The change from not_reproduced was agreed in the third verification round.", "deviations": [ "Pairs as in the criterion; the claim's depends_on (fig3f-gamma-sfc-preferred) names the SFC pairs, which are the same category neuron and PAC channel pairs, so the claim was run without the SFC. Bundles are the electrodes of a session with the same location and coordinates.", "Mixed model fitted with statsmodels MixedLM (REML), random intercepts for patient and for pair within patient, in place of fitglme, with load 3 as the reference level (Supplementary Table S3, \"Load 1 are tested against load 3 trials\", with a term named \"Load1\"). Artefact reading (a) assumed, as for edf2b-hippocampus-load-patients; under reading (b) one of the four readings has p = 0.056." ], "assumed": true, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "pac-neuron-counts", "statement": "79 of 212 hippocampal neurons (37.29%) and 163 of 358 amygdala neurons (45.53%) qualified as PAC neurons, more than expected by chance.", "source": "Results, Category neurons are not PAC neurons, paragraph 1; Figure 4a, b", "quote": "In the hippocampus, 79 (37.29%) out of 212 available neurons ( P < 0.005; 200 permutations ( Methods ); pre-processing steps removed broadband LFPs for some of the neurons and those were therefore not part of this analysis) qualified as PAC neurons. In the amygdala, 163 (45.53%) out of 358 neurons ( P < 0.005) qualified as PAC neurons", "statistics": [ { "about": "hippocampus", "test": "permutation test, 200 trial re-pairings of spikes and LFP", "p": 0.005, "p_relation": "<", "n": "212 neurons" }, { "about": "amygdala", "test": "permutation test, 200 trial re-pairings of spikes and LFP", "p": 0.005, "p_relation": "<", "n": "358 neurons" } ], "kind": "finding", "inferred": "", "outcome": "partially_reproduced", "blocked_by": null, "notebook": "reproduce_daume_2024_pac_neurons.ipynb", "notes": "With FDR over all combinations of an area and load (O9 a), 8 of 269 hippocampal neurons (2.97%) and 5 of 422 amygdala neurons (1.18%) qualify, against 79 of 212 and 163 of 358; with FDR over each neuron's channels (b), 12 and 24. Every shuffled count is lower than the observed count (shuffle maxima 1 and 1 under a, 5 and 4 under b), p = 0.005 in both areas under both readings, so the shuffle test holds and the counts are far out of tolerance. Without FDR, 1.8% of neuron-channel-load combinations pass both tests at p < 0.01, against 0.1% in the shuffles. 269 and 422 MTL neurons have LFP in their bundle, against the paper's 212 and 358. Without any correction, 41 hippocampal and 66 amygdala neurons have a combination passing both tests at p < 0.01 (shuffles at most 13 and 14), still fewer than the paper's counts after FDR. The paper's plot_PAC_neuron.m fits the three GLMs to the 20 aggregated bin counts of one neuron-channel pair with fixed effects only, as sbcat.py does, so neither a per-trial layout nor a random effect explains the difference.", "deviations": [ "Selection as in plot_PAC_neuron.m (20 theta-gamma bins, three Poisson GLMs, likelihood-ratio tests), fitted by IRLS in Python in place of fitglme (deviances checked against statsmodels). Artefact rejection reading (a) without visual inspection. The FDR family is not stated (O9); both readings were run. Bundles are taken as the electrodes of a session with the same location and coordinates, as the files name no bundle." ], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "pac-category-overlap", "statement": "The overlap of PAC and category neurons (28 of 79 in the hippocampus, 68 of 163 in the amygdala) did not exceed that expected of independent populations.", "source": "Results, Category neurons are not PAC neurons, paragraph 2; Figure 4b", "quote": "In the hippocampus, 28 (35.44%) out of the 79 PAC neurons were both PAC neurons as well as category neurons. In the amygdala, this was the case for 68 (41.72%) out of the 163 PAC neurons (Fig. 4b ). The proportion of category neurons among PAC neurons was not significantly higher than expected by independent subpopulations in any of the two regions (both P > 0.05", "statistics": [ { "test": "test against independent subpopulations", "p": 0.05, "p_relation": ">", "n": "79 hippocampal and 163 amygdala PAC neurons" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": "reproduce_daume_2024_pac_neurons.ipynb", "notes": "Too few PAC neurons under the claim's own selection (8 hippocampal and 5 amygdala, against 79 and 163) for the overlap to describe the paper's populations. For the record, 4 of 8 and 2 of 5 are category neurons; the binomial test gives p = 0.16 and 0.57 and the hypergeometric test p = 0.13 and 0.64.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf7a-pac-neurons-no-category-decoding", "statement": "Category could be decoded from encoding-period rates of category neurons (hippocampus 72.77%, amygdala 88.71%) but not from PAC neurons that are not category neurons (26.06% and 25.86%, chance 20%).", "source": "Results, Category neurons are not PAC neurons, paragraph 2; Extended Data Fig. 7a", "quote": "the decoder was able to differentiate between the picture categories when it was trained on FRs from the category neurons (hippocampus, 72.77%, P = 0.001; amygdala, 88.71%, P = 0.001; Extended Data Fig. 7a ). However, the decoder could not differentiate between the categories when it was trained on FRs from PAC neurons that were not also category neurons in both MTL areas (hippocampus, 26.06%, P = 0.16; amygdala, 25.86%, P = 0.15; chance level = 20%)", "statistics": [ { "about": "hippocampal category neurons", "test": "right-sided permutation test, 1,000 label shuffles", "p": 0.001 }, { "about": "amygdala category neurons", "test": "right-sided permutation test, 1,000 label shuffles", "p": 0.001 }, { "about": "hippocampal PAC-only neurons", "test": "right-sided permutation test, 1,000 label shuffles", "p": 0.16 }, { "about": "amygdala PAC-only neurons", "test": "right-sided permutation test, 1,000 label shuffles", "p": 0.15 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Needs PAC neurons that are not category neurons; the selection gave 4 such neurons in the hippocampus and 3 in the amygdala (pac-neuron-counts), too few to decode from. The decoding of category neurons alone was not run.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig4c-pac-neuron-firing-single-trial-pac", "statement": "Spike counts of hippocampal PAC neurons were positively related to single-trial theta-high-gamma PAC on their selected channel; amygdala PAC neurons showed no relation.", "source": "Results, Properties of PAC neurons, paragraph 1; Figure 4c, d; Supplementary Table S4", "quote": "FRs of PAC neurons were positively correlated with single-trial estimates of TG-PAC in the hippocampus ( c ; n = 79, P = 0.028, mixed-effects GLM), but not in the amygdala ( d ; n = 163, P = 0.98", "statistics": [ { "about": "hippocampus, SC term", "test": "mixed-effects GLM", "statistic": "estimate 0.0032, t = 2.19", "p": 0.028, "n": "79 neurons" }, { "about": "amygdala, SC term", "test": "mixed-effects GLM", "statistic": "estimate -3e-5, t = -0.02", "p": 0.98, "n": "163 neurons" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": "reproduce_daume_2024_pac_neurons.ipynb", "notes": "Too few PAC neurons (8 hippocampal and 5 amygdala pairs). For the record, the hippocampal SC term is not significant (O5 a, p = 0.85; O5 b, p = 0.18) and the amygdala term gives p = 0.0496 and 0.041. These fits code load with load 1 as the reference, so the SC term is the load 1 slope, not the paper's (Table S4 codes load 3 as the reference); they were not refitted, as the claim is blocked on the selection.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig4e-hippocampus-pac-neurons-accuracy", "statement": "Hippocampal PAC neurons fired above baseline during maintenance, more in correct than in incorrect trials, and above baseline in correct trials only.", "source": "Results, Properties of PAC neurons, paragraph 1; Figure 4e", "quote": "Second, their FR was elevated throughout the maintenance period compared with the baseline ( t 78 = 2.43, P = 0.01). Third, FRs varied as a function of accuracy (correct versus incorrect trials ( t 62 = 3.82, P < 0.001", "statistics": [ { "about": "maintenance against baseline", "test": "two-sided permutation-based t-test", "statistic": "t78 = 2.43", "p": 0.01, "n": "79 neurons" }, { "about": "correct against incorrect", "test": "two-sided permutation-based t-test", "statistic": "t62 = 3.82", "p": 0.0001, "n": "63 neurons" }, { "about": "correct against baseline", "test": "two-sided permutation-based t-test", "statistic": "t62 = 2.67", "p": 0.01, "n": "63 neurons" }, { "about": "incorrect against baseline", "test": "two-sided permutation-based t-test", "statistic": "t62 = -0.98", "p": 0.33, "n": "63 neurons" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": "reproduce_daume_2024_pac_neurons.ipynb", "notes": "Too few PAC neurons (8 hippocampal, paper 79). For the record, in Hz maintenance does not exceed baseline (p = 0.51) and correct exceeds incorrect (p = 0.024, 8 neurons).", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig4g-hippocampus-pac-neurons-load", "statement": "Hippocampal PAC neuron rates did not differ between loads but were above baseline in each load; they also did not differ between fast and slow RT trials.", "source": "Results, Properties of PAC neurons, paragraph 1; Figure 4g", "quote": "FRs were not significantly different between the two load conditions (load 3 \u2212 load 1, t 78 = 1.38, P = 0.20), but FRs were elevated compared with the baseline in each condition considered separately (load 1, t 78 = 2.14, P = 0.03; load 3, t 78 = 2.45, P = 0.01", "statistics": [ { "about": "load 3 against load 1", "test": "two-sided permutation-based t-test", "statistic": "t78 = 1.38", "p": 0.2, "n": "79 neurons" }, { "about": "load 1 against baseline", "test": "two-sided permutation-based t-test", "statistic": "t78 = 2.14", "p": 0.03 }, { "about": "load 3 against baseline", "test": "two-sided permutation-based t-test", "statistic": "t78 = 2.45", "p": 0.01 }, { "about": "fast against slow RT", "test": "two-sided permutation-based t-test", "statistic": "t78 = -0.07", "p": 0.94 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": "reproduce_daume_2024_pac_neurons.ipynb", "notes": "Too few PAC neurons (8 hippocampal, paper 79). For the record, in Hz load 3 is lower than load 1 (p = 0.0086) and neither load exceeds baseline (p = 0.14 and 0.16).", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig4-amygdala-pac-neurons", "statement": "Amygdala PAC neurons fired above baseline during maintenance but did not differ between correct and incorrect trials, loads, or fast and slow RT trials.", "source": "Results, Properties of PAC neurons, paragraph 2; Figure 4f, h", "quote": "PAC neurons in the amygdala also showed higher FRs during the maintenance period compared with the baseline ( t 162 = 6.40, P < 0.001), but there were no significant differences between correct and incorrect WM trials ( t 156 = \u22120.77, P = 0.45; Fig. 4f ), loads (load 3 \u2212 load 1, t 162 = 0.26, P = 0.80", "statistics": [ { "about": "maintenance against baseline", "test": "two-sided permutation-based t-test", "statistic": "t162 = 6.40", "p": 0.001, "p_relation": "<", "n": "163 neurons" }, { "about": "correct against incorrect", "test": "two-sided permutation-based t-test", "statistic": "t156 = -0.77", "p": 0.45, "n": "156 neurons" }, { "about": "load 3 against load 1", "test": "two-sided permutation-based t-test", "statistic": "t162 = 0.26", "p": 0.8 }, { "about": "fast against slow RT", "test": "two-sided permutation-based t-test", "statistic": "t162 = 1.87", "p": 0.07 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": "reproduce_daume_2024_pac_neurons.ipynb", "notes": "Too few PAC neurons (5 amygdala, paper 163); no test reaches p < 0.05 with 5 neurons.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf7bc-pac-neuron-local-sfc", "statement": "Local theta and gamma SFC of PAC neurons did not differ between loads in the hippocampus or amygdala, though both exceeded shuffled surrogates.", "source": "Extended Data Fig. 7b, c legends", "quote": "Theta and gamma SFC between PAC neurons and local LFP recordings did not differ as a function of load in ( b ) the hippocampus ... or ( c ) the amygdala ... Theta and gamma SFC, however, were both significantly stronger than shuffled surrogates in both areas (all p = 0.0001)", "statistics": [ { "about": "hippocampus theta", "test": "two-sided permutation-based t-test", "statistic": "t(78) = -1.54", "p": 0.13 }, { "about": "hippocampus gamma", "test": "two-sided permutation-based t-test", "statistic": "t(78) = -1.12", "p": 0.27 }, { "about": "amygdala theta", "test": "two-sided permutation-based t-test", "statistic": "t(162) = -0.71", "p": 0.47 }, { "about": "amygdala gamma", "test": "two-sided permutation-based t-test", "statistic": "t(162) = 0.76", "p": 0.45 }, { "about": "against shuffled surrogates, all", "test": "two-sided permutation-based t-test", "p": 0.0001 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Needs PAC neurons (8 and 5 selected, against 79 and 163) and spike-field coherence, which was not run.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig5b-hippocampus-vmpfc-theta-sfc-load", "statement": "Spikes of hippocampal PAC neurons were more strongly phase-locked to vmPFC theta LFP (2.5-4.3 Hz) in load 3 than in load 1, with no such effect for other frequencies or for pre-SMA and dACC LFPs.", "source": "Results, PAC neurons phase lock to the frontal cortex, paragraph 1; Figure 5b; Extended Data Fig. 8a", "quote": "SFC was significantly stronger in load 3 compared with load 1 between spiking activity of PAC neurons in the hippocampus and theta-band LFPs recorded in the vmPFC (Fig. 5b and Methods ; cluster P < 0.001). We did not observe significant differences for other frequency bands, nor for the other two frontal brain areas", "statistics": [ { "test": "two-sided cluster-based permutation t-test", "p": 0.0001, "corrected": "Bonferroni over two MTL areas, three frontal areas and two cell populations", "n": "175 connections" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Needs hippocampal PAC neurons (8 selected, against 79) and spike-field coherence, which was not run.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig5cd-no-effect-category-amygdala", "statement": "Hippocampal category neurons showed no load difference in SFC to vmPFC LFP, and amygdala PAC neurons showed none to the LFP of any frontal area.", "source": "Figure 5c, d legends; Results, PAC neurons phase lock to the frontal cortex, paragraph 1", "quote": "Category neurons from the hippocampus ( c ; n = 215) or PAC neurons from the amygdala ( d ; n = 767) did not show significant SFC differences between loads relative to the vmPFC LFP", "statistics": [], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig5b-hippocampus-vmpfc-theta-sfc-load, which is blocked.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig5e-random-subsets", "statement": "The theta SFC load difference of hippocampal PAC neurons to vmPFC was larger than for every one of 10,000 random sets of hippocampus-vmPFC connections.", "source": "Figure 5e legend; Results, PAC neurons phase lock to the frontal cortex, paragraph 1", "quote": "Hippocampal PAC cells ( n = 175, cyan line) yielded the strongest long-range theta SFC difference between load 3 and load 1 trials among 10,000 random selections of hippocampus\u2013vmPFC connections ( P = 0.0001, right-sided permutation test)", "statistics": [ { "test": "right-sided permutation test, 10,000 random selections", "p": 0.0001, "n": "175 connections" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig5b-hippocampus-vmpfc-theta-sfc-load, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig5f-theta-sfc-rt", "statement": "Theta SFC between hippocampal PAC neurons and vmPFC was stronger in fast than in slow RT trials; amygdala PAC neurons showed no such effect.", "source": "Results, PAC neurons phase lock to the frontal cortex, paragraph 1; Figure 5f", "quote": "theta SFC was stronger for fast compared with slow RTs for PAC neurons from the hippocampus ( t 166 = 2.10, P = 0.03; Fig. 5f ), but not from the amygdala ( t 705 = 1.40, P = 0.16)", "statistics": [ { "about": "hippocampus", "test": "two-sided permutation-based t-test", "statistic": "t166 = 2.10", "p": 0.03, "n": "167 connections" }, { "about": "amygdala", "test": "two-sided permutation-based t-test", "statistic": "t705 = 1.40", "p": 0.16, "n": "706 connections" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig5b-hippocampus-vmpfc-theta-sfc-load, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf8b-vmpfc-sfc-patients", "statement": "At the patient level, theta SFC between hippocampal PAC neurons and vmPFC was higher in load 3 than in load 1.", "source": "Extended Data Fig. 8b legend", "quote": "At the patient-level, theta cross-regional SFC was significantly higher for load 3 than load 1 trials, suggesting that the effect was not driven by a few channels or patients (n = 20; t(19) = \u22122.8297, p = 0.0071)", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t(19) = -2.8297", "p": 0.0071, "n": "20 patients" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig5b-hippocampus-vmpfc-theta-sfc-load, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf8c-pac-against-category", "statement": "The load modulation of hippocampus-vmPFC theta SFC was stronger for PAC neurons than for category neurons.", "source": "Extended Data Fig. 8c legend", "quote": "the load modulation of cross-regional theta SFC between hippocampal neurons and vmPFC LFPs was significantly stronger for PAC than for category neurons ... t(376.07) = 3.3942, p = 0.0001; unpaired two-sided permutation t -test)", "statistics": [ { "test": "unpaired two-sided permutation t-test", "statistic": "t(376.07) = 3.3942", "p": 0.0001, "n": "175 PAC and 215 category neuron connections" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig5b-hippocampus-vmpfc-theta-sfc-load, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf8de-narrow-spiking", "statement": "The load and RT effects on hippocampus-vmPFC theta SFC held for narrow-spiking PAC neurons only, not broad-spiking ones.", "source": "Extended Data Fig. 8d, e legends", "quote": "we observed a significant difference in theta SFC between load 3 and load 1 only for the narrow-spiking PAC neurons (trough-to-peak time <0.5 ms; n = 91 connections; cluster-p = 0.0001, left). No effect was found for broad-spiking PAC neurons (n = 84)", "statistics": [ { "about": "load, narrow-spiking", "test": "two-sided cluster-based permutation t-test", "p": 0.0001, "n": "91 connections" }, { "about": "RT, narrow-spiking", "test": "two-sided permutation-based t-test", "statistic": "t(90) = 3.02", "p": 0.003, "n": "91 connections" }, { "about": "RT, broad-spiking", "test": "two-sided permutation-based t-test", "statistic": "t(75) = -0.66", "p": 0.52, "n": "76 connections" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig5b-hippocampus-vmpfc-theta-sfc-load and fig5f-theta-sfc-rt, which are blocked. The mean waveforms needed for the trough-to-peak time are in the files (waveform_mean, 256 samples, in 43 of the 44 files; waveforms in all 44).", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6a-noise-correlations", "statement": "Pairs of category and PAC neurons had positive noise correlations on average in both the hippocampus (162 pairs) and amygdala (892 pairs), larger than after shuffling trials within conditions.", "source": "Results, Information-enhancing noise correlations, paragraph 1; Figure 6a; Extended Data Fig. 9a", "quote": "In both the hippocampus (162 pairs; t 161 = 5.26; P < 0.001) and the amygdala (892 pairs; t 891 = 15.51; P < 0.001), pairs of category neurons and PAC neurons had on average positive co-fluctuations of spike counts", "statistics": [ { "about": "hippocampus, against zero", "test": "two-sided permutation-based t-test", "statistic": "t161 = 5.26", "p": 0.0001, "n": "162 pairs" }, { "about": "amygdala, against zero", "test": "two-sided permutation-based t-test", "statistic": "t891 = 15.51", "p": 0.0001, "n": "892 pairs" }, { "about": "hippocampus, against within-condition trial shuffles", "test": "right-sided permutation test, 1,000 shuffles", "p": 0.001, "n": "162 pairs" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Needs pairs of category and PAC neurons; with 8 hippocampal and 5 amygdala PAC neurons (pac-neuron-counts) the pairs cannot approach the paper's 162 and 892.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf9ef-non-category-pac-pairs", "statement": "Noise correlations stayed positive, and above shuffles, for pairs of category neurons with PAC neurons that are not category neurons.", "source": "Extended Data Fig. 9e, f legends", "quote": "Bin-wise correlations among pairs of category neurons and PAC neurons that were not also category neurons were significantly positive on average in hippocampus (n = 101, p = 0.0003) and amygdala (n = 555, p = 0.0001)", "statistics": [ { "about": "hippocampus", "test": "two-sided permutation-based t-test", "p": 0.0003, "n": "101 pairs" }, { "about": "amygdala", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "555 pairs" }, { "about": "hippocampus against shuffles", "test": "right-sided permutation test", "p": 0.0001 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6a-noise-correlations, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6c-adding-pac-neurons", "statement": "In the hippocampus, adding single PAC neurons to the optimized decoding ensemble raised category decoding when noise correlations were intact but not when they were removed.", "source": "Results, Information-enhancing noise correlations, paragraph 2; Figure 6c", "quote": "adding single PAC neurons to the optimized decoding ensemble significantly enhanced category decoding when noise correlations were intact ( t 20 = 3.16, P < 0.001) but not when they were removed ( t 20 = \u22120.12, P = 0.91, BF 01 = 4.37; intact versus removed, t 20 = 3.33, P = 0.003", "statistics": [ { "about": "intact", "test": "two-sided permutation-based t-test", "statistic": "t20 = 3.16", "p": 0.0007, "n": "21 sessions" }, { "about": "removed", "test": "two-sided permutation-based t-test", "statistic": "t20 = -0.12", "p": 0.91, "n": "21 sessions" }, { "about": "intact against removed", "test": "two-sided permutation-based t-test", "statistic": "t20 = 3.33", "p": 0.003, "n": "21 sessions" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Needs hippocampal PAC neurons in at least 21 sessions; 8 were selected (pac-neuron-counts). The greedy decoding was not run.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6d-removing-pac-neurons", "statement": "Removing all PAC neurons from the hippocampal ensembles reduced maximal decoding only with intact noise correlations, while removing as many non-PAC neurons reduced it with intact and removed correlations.", "source": "Results, Information-enhancing noise correlations, paragraph 3; Figure 6d", "quote": "PAC neurons enhanced the decodability of category, but only when noise correlations were intact ( t 21 = 2.94; P = 0.004; FDR corrected for all comparisons; Fig. 6d ). Removing PAC neurons from the ensemble after noise correlations were removed did not significantly change the decoding performance ( t 21 = 1.36, P = 0.20", "statistics": [ { "about": "PAC removed, intact correlations", "test": "two-sided permutation-based t-test", "statistic": "t21 = 2.94", "p": 0.004, "corrected": "FDR over all comparisons", "n": "22 sessions" }, { "about": "PAC removed, removed correlations", "test": "two-sided permutation-based t-test", "statistic": "t21 = 1.36", "p": 0.2, "n": "22 sessions" }, { "about": "non-PAC removed, intact correlations", "test": "two-sided permutation-based t-test", "p": 0.01, "n": "22 sessions" }, { "about": "non-PAC removed, removed correlations", "test": "two-sided permutation-based t-test", "statistic": "t21 = 3.89", "p": 0.0001, "n": "22 sessions" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6c-adding-pac-neurons, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6e-decoding-difference", "statement": "The gain of intact over removed noise correlations in maximal decoding was smaller after removing PAC neurons, but not after removing non-PAC neurons.", "source": "Results, Information-enhancing noise correlations, paragraph 3; Figure 6e", "quote": "the maximal differences between intact and removed noise correlations were larger when PAC neurons were part of the ensembles as compared to being removed ( t 21 = 3.59, P < 0.001; Fig. 6e )", "statistics": [ { "about": "PAC neurons", "test": "two-sided permutation-based t-test", "statistic": "t21 = 3.59", "p": 0.0007, "n": "22 sessions" }, { "about": "non-PAC neurons", "test": "two-sided permutation-based t-test", "statistic": "t21 = -1.77", "p": 0.1, "n": "22 sessions" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6d-removing-pac-neurons, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf9g-non-category-pac-decoding", "statement": "Removing only PAC neurons that are not category neurons also lowered decoding with intact noise correlations only.", "source": "Extended Data Fig. 9g legend", "quote": "Like for all PAC neurons, decoding performance was enhanced by nonCatPAC neurons only when noise correlations were intact (n = 23 sessions, p = 0.0005)", "statistics": [ { "test": "two-sided permutation-based t-test", "p": 0.0005, "n": "23 sessions" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6d-removing-pac-neurons, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf9jk-amygdala-decoding", "statement": "In the amygdala, PAC neurons raised decoding with both intact and removed noise correlations, and removing them lowered decoding in both cases, as for random cells.", "source": "Results, Information-enhancing noise correlations, paragraph 3; Extended Data Fig. 9j, k", "quote": "In the amygdala, adding single PAC neurons (n = 28) to the decoding ensemble did not only enhance decodability when noise correlations were intact (p = 0.0001), but also when removed (p = 0.049", "statistics": [ { "about": "adding, intact", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "28, printed after \"PAC neurons\" (sessions in the hippocampal panel)" }, { "about": "adding, removed", "test": "two-sided permutation-based t-test", "p": 0.049 }, { "about": "adding, intact against removed", "test": "two-sided permutation-based t-test", "p": 0.0009 }, { "about": "removing, intact", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "32 sessions" }, { "about": "removing, removed", "test": "two-sided permutation-based t-test", "p": 0.0001, "n": "32 sessions" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6c-adding-pac-neurons, which is blocked; 5 amygdala PAC neurons were selected.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6g-signal-noise-angle", "statement": "The angle between the signal and noise axes was about 69 degrees with intact noise correlations and became smaller after removing them.", "source": "Results, Information-enhancing noise correlations, paragraph 4; Figure 6g", "quote": "when noise correlations were intact, the signal\u2013noise axis angle was around 69\u00b0 (out of maximally 90\u00b0). After removing noise correlations, this angle became significantly smaller, as hypothesized ( t 31 = 2.77, P = 0.009", "statistics": [ { "test": "two-sided permutation-based t-test", "statistic": "t31 = 2.77", "p": 0.009, "n": "32 sessions" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6c-adding-pac-neurons, which is blocked.", "deviations": [], "assumed": false, "disagreed": true, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6h-projection-variance", "statement": "Removing PAC neurons raised the s.d. of responses projected on the signal axis, and removing noise correlations raised it only while PAC neurons were in the ensemble.", "source": "Results, Information-enhancing noise correlations, paragraph 4; Figure 6h", "quote": "Removing PAC neurons from the ensembles significantly increased the s.d. of the projection values (main effect for ensemble, F 1,18 = 12.55, P = 0.0013; Fig. 6h ). Moreover, the s.d. of the projected values was larger when noise correlations were removed (main effect for noise correlations, F 1,18 = 7.24, P = 0.014). This was only the case if PAC neurons were part of the ensembles ( t 18 = \u22122.66, P = 0.014), and there was no significant difference between intact and removed noise correlations when PAC neurons were removed ( t 18 = \u22121.69, P = 0.11)", "statistics": [ { "about": "main effect of ensemble", "test": "permutation-based F-test (2 x 2 ANOVA)", "statistic": "F1,18 = 12.55", "p": 0.0013, "n": "19 sessions" }, { "about": "main effect of noise correlations", "test": "permutation-based F-test (2 x 2 ANOVA)", "statistic": "F1,18 = 7.24", "p": 0.014, "n": "19 sessions" }, { "about": "intact against removed, with PAC neurons", "test": "two-sided permutation-based t-test", "statistic": "t18 = -2.66", "p": 0.014 }, { "about": "intact against removed, without PAC neurons", "test": "two-sided permutation-based t-test", "statistic": "t18 = -1.69", "p": 0.11 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6g-signal-noise-angle, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6i-noise-correlations-rt", "statement": "Noise correlations of hippocampal PAC-category pairs in preferred trials were stronger in fast than in slow RT trials, positive only in fast trials, and the RT difference held in load 3 but not in load 1.", "source": "Results, Information-enhancing noise correlations, paragraph 5; Figure 6i", "quote": "In the hippocampus, noise correlations were significantly stronger for fast compared with slow RT trials ( t 161 = 2.15, P = 0.028; Fig. 6i (left)", "statistics": [ { "about": "fast against slow", "test": "two-sided permutation-based t-test", "statistic": "t161 = 2.15", "p": 0.028, "n": "162 pairs" }, { "about": "fast against zero", "test": "two-sided permutation-based t-test", "statistic": "t161 = 4.10", "p": 0.001, "p_relation": "<" }, { "about": "slow against zero", "test": "two-sided permutation-based t-test", "statistic": "t161 = 1.94", "p": 0.06 }, { "about": "load 3, fast against slow", "test": "two-sided permutation-based t-test", "statistic": "t161 = 2.60", "p": 0.009 }, { "about": "load 1, fast against slow", "test": "two-sided permutation-based t-test", "statistic": "t161 = 0.96", "p": 0.34 } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6a-noise-correlations, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "edf9hil-noise-correlations-rt-controls", "statement": "The fast against slow difference was absent for non-preferred trials and in the amygdala, and held at the patient level in the hippocampus.", "source": "Extended Data Fig. 9h, i, l legends", "quote": "Correlations for pairs of PAC and category neurons in preferred trials were averaged within each patient and then compared between fast and slow RTs across all patients (n = 16, p = 0.0027)", "statistics": [ { "about": "hippocampus, non-preferred trials", "test": "two-sided permutation-based t-test", "p": 0.9, "n": "162 pairs" }, { "about": "hippocampus, patient level", "test": "two-sided permutation-based t-test", "p": 0.0027, "n": "16 patients" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6i-noise-correlations-rt, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null }, { "claim": "fig6j-specific-to-pac-pairs", "statement": "The fast against slow RT difference in noise correlations was larger for PAC-category pairs than for random pairs of category neurons with non-PAC neurons.", "source": "Results, Information-enhancing noise correlations, paragraph 5; Figure 6j", "quote": "PAC-to-category neuron pairs showed a significantly stronger effect compared with most randomly selected cell pairs ( P = 0.016; Fig. 6j )", "statistics": [ { "test": "right-sided permutation test", "p": 0.016, "n": "162 pairs" } ], "kind": "finding", "inferred": "", "outcome": "blocked", "blocked_by": "other", "notebook": null, "notes": "Depends on fig6i-noise-correlations-rt, which is blocked.", "deviations": [], "assumed": false, "disagreed": false, "data": "dataset", "where": null, "proposed": null } ], "findings": [ { "kind": "format_error", "statement": "Spike times are not sorted in four units of two files. In sub-13 ses-1, units 6, 14 and 15 jump back by up to 1,319 to 1,339 s (units 6 and 15 once, unit 14 three times), as if two spike trains had been concatenated; in sub-22 ses-2, unit 10 steps back 24 times by up to 3.3 ms. The analysis sorts them, which keeps the result of the paper's code (it selects spikes with logical masks).", "found": null, "status": null }, { "kind": "selection_undocumented", "statement": "The files do not mark the LFP trials, channels or regions that the paper's artefact rejection removed, most of it by visual inspection (20.4 +/- 13.9 trials per wire, and whole wires or regions). The automated rules of the Methods remove 11.4% (on the epochs) or 25.9% (on whole trials) of trials, and the remaining channels change several results. The pre-SMA has 17 channels with significant theta-high-gamma PAC where the paper has 5, and they decide three claims of absence. 269 hippocampal and 422 amygdala neurons have LFP in their bundle, where the paper analysed 212 and 358 (\"pre-processing steps removed broadband LFPs for some of the neurons\"). The paper's code ships the rejected trials of one example channel (artif_trials.mat, sub-26 session 1, channel 33, 17 trials), which shows the information exists for the paper's own analysis.", "found": null, "status": null }, { "kind": "metadata_insufficient", "statement": "The files do not identify the microwire bundles that the analyses pair neurons and LFPs within. Every wire is its own ElectrodeGroup, and the location names only area and hemisphere; some locations of a session hold up to 15 wires, and 26 of 285 location groups have more than one set of coordinates. Bundles were inferred from location and coordinates.", "found": null, "status": null }, { "kind": "other", "statement": "The maintenance period lasts 2.53 to 3.20 s in the trials table (probe onset minus maintenance onset), against 2.55 to 2.85 s in the Methods, and the blank between pictures in load 3 lasts 0.011 to 0.341 s, against 17 to 200 ms. RTs reach 74.2 s (69 trials above 10 s).", "found": null, "status": null }, { "kind": "other", "statement": "The paper cites version 0.240118.2135, which this attempt used. In the latest published version, 0.250122.0110, each of the 44 files has a new asset a few hundred kilobytes smaller; the attempt did not compare their contents.", "found": null, "status": null }, { "kind": "methods_unavailable", "statement": "Question for the authors - are the firing-rate tests of Fig. 3b-d and Fig. 4 computed on rates in Hz or on the percentage change from baseline shown in the panels? The two give different outcomes for Fig. 3b (preferred against non-preferred), Fig. 3c and Fig. 3d; this attempt assumed the percentage change.", "found": null, "status": null }, { "kind": "methods_unavailable", "statement": "Question for the authors - in the automated LFP artefact rejection, over what span are the channel's samples z-scored and tested (the -0.5 to 3 s maintenance epochs, or whole trials)? The two give different outcomes for Extended Data Fig. 2b and 3f; this attempt assumed the epochs, which agree with the paper's rejected trials on the one channel its code ships.", "found": null, "status": null }, { "kind": "methods_unavailable", "statement": "Question for the authors - for theta-gamma nesting (Extended Data Fig. 3e), are the 500 ms windows centred on the maximum of gamma amplitude or on a peak of the gamma-filtered signal? Centred on amplitude maxima, 6% of PAC channels are nested here; on gamma peaks, 73%. This attempt assumed gamma peaks.", "found": null, "status": null }, { "kind": "methods_unavailable", "statement": "Question for the authors - over what frequency range, and with what spectral estimate, was the log-log slope compared between loads (Methods, LFP preprocessing)? With Welch spectra, 2 to 150 Hz gives a small but significant difference (p = 0.0073) and 30 to 150 Hz none; this attempt assumed 2 to 150 Hz.", "found": null, "status": null }, { "kind": "methods_unavailable", "statement": "Question for the authors - how is single-trial theta-gamma PAC computed for the mixed-effects GLMs (Fig. 2f, 3h, 4c), and is RT transformed in the model of Table S2? Table S3's negative intercept (-0.044) rules out the raw MI as the response there, while Table S2's PAC estimate (-49.07) fits the raw MI as the predictor with RT z-scored within session (-49.39 here). Table S2 gives no unit for RT, and RT in s gives an intercept of 1.70 against its 0.38. With load 3 as the reference, the hippocampal relation to RT holds on z-scored or log RT, but the amygdala term is also significant under the raw MI (p = 0.015).", "found": null, "status": null }, { "kind": "methods_unavailable", "statement": "Question for the authors - which family is the FDR correction of the PAC neuron selection taken over, and what else differs from plot_PAC_neuron.m? Run as in that function on the dandiset, 8 of 269 hippocampal and 5 of 422 amygdala neurons qualify (12 and 24 with FDR over each neuron's channels), against 79 of 212 and 163 of 358.", "found": null, "status": null } ], "reconciliation": null, "verifications": [ { "date": "2026-10-31", "kind": "agent", "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebooks)", "reviewed": "Round 1. The executed notebooks of the author run with the conclusion cells and the summary removed, the helper module and the claims file, with the paper. The verifier queried the archive metadata and tested helper functions on synthetic inputs. Its report gave its own outcome for each claim; it disagreed with 16 recorded outcomes - fig3c-category-load, and 15 claims the record blocked, 6 of which it judged from results printed in the notebooks and 9 it found testable from the archive.", "agreed": 41, "of": 57, "disagreements": [ { "claim": "edf3-hippocampus-pac-controls", "outcome": "partially_reproduced", "reason": "Under artefact (a), gamma power is higher in load 3 by both estimates (p = 0.035 and 0.006); theta power, phase-phase coupling and the preferred-phase t-test show no difference. Waveform asymmetry, peak frequency and the second phase-phase-coupling ratio were not run. A V test of the load differences against 0 tests concentration around 0, so a significant V test supports no phase shift; the claims file's reading (a), with its rule that no test gives p < 0.05, misreads the test (problem 5)." }, { "claim": "fig3c-category-load", "outcome": "partially_reproduced", "reason": "In Hz - preferred trials p = 0.022 (load 1 higher), non-preferred p = 0.20. In % - preferred p = 0.77. Under O6 (b), which % change with at least 1 incorrect trial supports by reproducing the paper's frontal n exactly (201, 180, 201), fig3c fails (problem 3)." }, { "claim": "pac-category-overlap", "outcome": "partially_reproduced", "reason": "The overlap does not exceed chance (p = 0.13 and 0.57), but the counts (4 and 2) are far out of tolerance and rest on n = 8 and 5. Every claim built on PAC neurons is run at n = 8 and 5 and cannot test the paper (problem 1)." }, { "claim": "fig4c-pac-neuron-firing-single-trial-pac", "outcome": "not_reproduced", "reason": "Hippocampal SC term p = 0.85 and 0.18; amygdala p = 0.050 and 0.041. Only 8 and 5 pairs, so this is not an informative test. Every claim built on PAC neurons is run at n = 8 and 5 and cannot test the paper (problem 1)." }, { "claim": "fig4e-hippocampus-pac-neurons-accuracy", "outcome": "partially_reproduced", "reason": "Correct > incorrect holds (p = 0.02-0.03); maintenance against baseline fails (p = 0.51); n = 8 against 63. Every claim built on PAC neurons is run at n = 8 and 5 and cannot test the paper (problem 1)." }, { "claim": "fig4g-hippocampus-pac-neurons-load", "outcome": "not_reproduced", "reason": "Both baseline tests fail, and load 3 < load 1 is significant (p = 0.009); n = 8. Every claim built on PAC neurons is run at n = 8 and 5 and cannot test the paper (problem 1)." }, { "claim": "fig4-amygdala-pac-neurons", "outcome": "partially_reproduced", "reason": "Baseline test p = 0.12 fails; the other three tests are null as claimed; n = 5. Every claim built on PAC neurons is run at n = 8 and 5 and cannot test the paper (problem 1)." }, { "claim": "fig3h-pac-category-firing", "outcome": "blocked", "reason": "fig3h appears nowhere - it is not in any notebook's table and not covered by the blocked line, yet every input it needs (category neurons x PAC channels) is already computed (problem 2)." }, { "claim": "fig3f-gamma-sfc-preferred", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection." }, { "claim": "fig3g-gamma-sfc-anova", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection." }, { "claim": "edf5d-sfc-high-gamma-power", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection." }, { "claim": "edf5c-theta-sfc-null", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection." }, { "claim": "edf5fg-gamma-sfc-all-channels-patients", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection." }, { "claim": "edf7a-pac-neurons-no-category-decoding", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection. (the category-neuron part, chance 20%)" }, { "claim": "fig5cd-no-effect-category-amygdala", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection. (the category-neuron part)" }, { "claim": "fig6g-signal-noise-angle", "outcome": "blocked", "reason": "Nothing in the archive blocks these; the archive lists all 44 files, with LFP in 43 sessions and units in all, and the notebooks load spikes, trials and LFP for every area. Testable now with no PAC-neuron selection." } ] }, { "date": "2026-11-01", "kind": "agent", "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebooks)", "reviewed": "Round 2. The full executed notebooks of the round-1 revision, with their conclusions and summaries, the helper module, the attempt record and the claims file, with the paper. The verifier agreed with every outcome but eight - fig2f-rt-single-trial-pac and fig3h-pac-category-firing (not established, as the GLMs code load 1 as the reference where Tables S2 and S3 code load 3), and six SFC claims it would block as not run, not as too large. The revision after it changed no outcome; it proposes changes for fig2f-rt-single-trial-pac and fig3h-pac-category-firing.", "agreed": 49, "of": 57, "disagreements": [ { "claim": "edf8fgh-frontal-controls", "outcome": "blocked", "reason": "Blocked, but not as too large. By the record's own timing, the one-subsample surrogate form takes about 45 CPU minutes and would fit; the record says its cluster tests were not written in this run for lack of time. The honest reason is other, not run. No notebook output shows the timings." }, { "claim": "fig2f-rt-single-trial-pac", "outcome": null, "reason": "Not established. The model codes load with load 1 as the reference level, so the PAC term estimates the slope in load 1 trials; Table S2 names its terms Load1 and PAC by Load1, so in the paper load 3 is the reference (or the coding is by effects), and the paper's term is not the load 1 slope. The interaction is significant here (p = 0.0075 under a, O5 a). The sentence that Table S2 gives the estimate in ms is unsupported - Table S2 gives no unit, and its intercept of 0.38 and Load1 term of -0.48 cannot be raw RT in s or ms, so the log10 RT variant (negative, p = 0.0066 under O5 b) cannot be dismissed on this ground. Minor - the printed RT maximum of 16.8 s comes from the last data frame of the loop (reading b, vmPFC); the hippocampal reading (a) set loses 87 trials when RTs above 10 s are dropped." }, { "claim": "fig3f-gamma-sfc-preferred", "outcome": "blocked", "reason": "Blocked, but not as too large. By the record's own timing, the one-subsample surrogate form takes about 45 CPU minutes and would fit; the record says its cluster tests were not written in this run for lack of time. The honest reason is other, not run. No notebook output shows the timings." }, { "claim": "fig3g-gamma-sfc-anova", "outcome": "blocked", "reason": "Blocked, but not as too large. By the record's own timing, the one-subsample surrogate form takes about 45 CPU minutes and would fit; the record says its cluster tests were not written in this run for lack of time. The honest reason is other, not run. No notebook output shows the timings." }, { "claim": "edf5d-sfc-high-gamma-power", "outcome": "blocked", "reason": "Blocked, but not as too large. By the record's own timing, the one-subsample surrogate form takes about 45 CPU minutes and would fit; the record says its cluster tests were not written in this run for lack of time. The honest reason is other, not run. No notebook output shows the timings." }, { "claim": "edf5c-theta-sfc-null", "outcome": "blocked", "reason": "Blocked, but not as too large. By the record's own timing, the one-subsample surrogate form takes about 45 CPU minutes and would fit; the record says its cluster tests were not written in this run for lack of time. The honest reason is other, not run. No notebook output shows the timings." }, { "claim": "edf5fg-gamma-sfc-all-channels-patients", "outcome": "blocked", "reason": "Blocked, but not as too large. By the record's own timing, the one-subsample surrogate form takes about 45 CPU minutes and would fit; the record says its cluster tests were not written in this run for lack of time. The honest reason is other, not run. No notebook output shows the timings." }, { "claim": "fig3h-pac-category-firing", "outcome": null, "reason": "Not established. The model codes load with load 1 as the reference, so the SC term estimates the slope in load 1 trials, while Table S3 names its terms Load1 and SC by Load1. The interaction is significant (p <= 0.0025 under every reading). From Table S3 the paper's own load 1 slope is 0.002 - 0.004 = -0.002, the same sign as the notebook's -0.0045 (preferred trials, O5 b); the sentence calling it the opposite sign to the paper's compares the notebook's load 1 slope with the paper's load 3 (or average) slope. The assumption (reading a) is not decisive, as reading (b) gives the same result." } ] }, { "date": "2026-11-02", "kind": "agent", "by": "Independent agent run by CatalystNeuro (same harness, separate context, did not write the notebooks)", "reviewed": "Round 3. The two changes of outcome proposed by the round-2 revision (fig2f-rt-single-trial-pac and fig3h-pac-category-firing), with the re-executed reproduce_daume_2024_lfp.ipynb (cells 27 and 44). Nothing else was reviewed. The verifier agreed with both.", "agreed": 2, "of": 2, "disagreements": [] } ] } ], "entries": [ { "name": "000673-daume-2024", "type": "reproduction", "title": "Phase-amplitude coupling and working memory in human hippocampus (Daume et al. 2024)", "description": "Tests the findings of Daume et al. (2024) on dandiset 000673, which holds 44 sessions of a Sternberg working memory task in 36 epilepsy patients, with 1,454 single neurons and 1,922 microwire LFP channels from the hippocampus, amygdala, pre-SMA, dACC and vmPFC. It tests the cohort, behaviour, category neurons and their maintenance activity, and the LFP theta-gamma phase-amplitude coupling, its load effect and its controls, single-trial PAC against reaction time and category neuron firing, and the selection and firing of PAC neurons. The spike-field coherence, noise correlation and decoding analyses were not run.", "datasets": [ "dandi:000673" ], "papers": [ "10.1038/s41586-024-07309-z" ], "derived_from": [], "notebooks": [ { "path": "reproduce_daume_2024.ipynb", "title": "Behaviour and single neurons (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024.html" }, { "path": "reproduce_daume_2024_lfp.ipynb", "title": "LFP phase-amplitude coupling (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024_lfp.html" }, { "path": "reproduce_daume_2024_lfp_controls.ipynb", "title": "Hippocampal PAC controls (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024_lfp_controls.html" }, { "path": "reproduce_daume_2024_pac_neurons.ipynb", "title": "PAC neurons (Daume et al. 2024, dandiset 000673)", "executed": "notebooks/000673-daume-2024/reproduce_daume_2024_pac_neurons.html" } ] } ] }