DANDI:000253

reducing the number of experiments analysed and presented here from 87 to 58

Results, A survey of visually evoked spiking, paragraph 2; Methods, Quality control for the Neuropixels recording session; Methods, Data processing pipeline

000021 and 000022 hold 58 sessions of 58 mice, 30 wild type and 8 Pvalb, 12 Sst and 8 Vip Cre lines crossed with Ai32. The pipeline counts before release (87, 61) are not judged.

How the test differed from the paper’s method

• Tested on dandisets 000021 and 000022, where the paper deposited these data; dandiset 000253 holds a different experiment.

In total, 284 out of 332 probes were mapped to the CCFv3

Methods, Neuropixels recordings, Probes; Methods, Ex vivo imaging quality control

The 58 sessions hold 332 probes; 284 have a positive sum of anterior-posterior CCF coordinates over the units passing the three quality filters, as ExtDataFigure1jk.py counts, and the other 48 have no CCF coordinate on any unit. Reading (a), the 58 released sessions.

Overall, we recorded 682 ± 144 units per experiment, 119 ± 48 units per probe and 56 ± 30 units per visual area ( Extended Data Fig. 1j ), sampling 6.1 ± 1.1 visual areas per experiment

Results, A survey of visually evoked spiking, paragraph 2; Extended Data Fig. 1j, k

Units per experiment 686.2 +/- 144.6, per probe 128.8 +/- 51.1, per area per experiment 56.2 +/- 30.3 (352 experiment-area pairs), areas per experiment 6.07 +/- 1.13. All within tolerance.

the mean visual latency of each area was correlated with its anatomical hierarchy score ( Fig. 3c ; Pearson’s r = 0.95, P = 0.00025)

Results, A functional hierarchy of visual areas, paragraph 4; Fig. 3a to c

What the paper reports

• Pearson correlation, r = 0.95, p = 2.5 × 10⁻⁴ (n = 8 areas)

All flashes, 250 ms window, 13,344 units; area means 66.1 ms (V1) to 75.2 ms (PM), r = 0.959, p = 0.0002. Dark flashes r = 0.904 (p = 0.002), light r = 0.933 (p = 0.0007), so the three readings agree. Variants; 500 ms window r = 0.802, 200 ms r = 0.936, the paper's table r = 0.953.

How the test differed from the paper’s method

• Unit metrics (p_value_rf, area_rf, firing_rate_dg) recomputed from spike times; the paper read them from the AllenSDK.
• The claims file names the flashes as open; Figure3/get_flash_PSTH.py uses every flash presentation (fl.stim_table.index.values), so all flashes is the main reading.

found a systematic increase in receptive field size with anatomical hierarchy score ( Fig. 3d - f ; Pearson’s r = 0.97, P = 8.3 × 10 −5 )

Results, A functional hierarchy of visual areas, paragraph 4; Fig. 3d to f

What the paper reports

• Pearson correlation, r = 0.97, p = 8.3 × 10⁻⁵ (n = 8 areas)

Mean RF area 544 deg2 (LGN) to 905 deg2 (AM), 13,344 units; r = 0.958, p = 0.0002. Without the latency cut r = 0.974. Brain Observatory sessions alone r = 0.957, Functional Connectivity r = 0.913.

The modulation index was highest in the LGN, whereas higher areas showed gradually less phase-dependent modulation ( Fig. 3g - i , Pearson’s r = −0.89, P = 0.003)

Results, A functional hierarchy of visual areas, paragraph 4; Fig. 3g to i

What the paper reports

• Pearson correlation, r = -0.89, p = 0.003 (n = 8 areas)

Mean log10 MI highest in LGN (0.628), lowest in AM (-0.124); r = -0.954, p = 0.0002. The mean of MI gives r = -0.934; each stimulus set alone gives r = -0.911 and -0.963, LGN highest.

How the test differed from the paper’s method

• Assumed - the preferred condition is the drifting grating condition (orientation, temporal frequency, and contrast in the Functional Connectivity set) with the highest mean spike count, blanks excluded. The paper takes MI from the AllenSDK and does not define the preferred condition; this is the AllenSDK's usual definition.
• Welch spectrum with nperseg 256 on the 2 s PSTH of 200 bins, as functions/modulation_index.py.

this mean intrinsic timescale for each area was not correlated with the visual hierarchy ( Extended Data Fig. 9f , g ) (Pearson’s r = −0.24, P = 0.57)

Results, A functional hierarchy of visual areas, paragraph 4; Extended Data Fig. 9f, g

What the paper reports

• Pearson correlation, r = -0.24, p = 0.57 (n = 8 areas)

15,958 units kept (SD < 100, at least 100 spikes); area means 23.6 to 40.1 ms; r = -0.158, p = 0.71. The code's window (1.0 to 2.0 s after onset, 5 to 400 ms) gives r = -0.290, p = 0.49. The test is weak; 8,834 of the 15,958 kept units sit at the 1 ms lower bound of the fit (median 1.00 ms), and without them r = 0.082, p = 0.85.

Higher-order areas had a longer response decay timescale, and therefore maintain stimulus-evoked activity over longer temporal windows, than lower stages ... ( Fig. 3j - l ; Pearson’s r = 0.86, P = 0.007)

Results, A functional hierarchy of visual areas, paragraph 4; Fig. 3j to l

What the paper reports

• Pearson correlation, r = 0.86, p = 0.007 (n = 8 areas, 7,837 units from 58 mice)

Methods window 0 to 250 ms, 5,842 units - r = 0.804, p = 0.016, PM 53.3 and AM 54.6 ms highest. The code's 40 to 290 ms window gives r = 0.876 (7,467 units), the paper's tables r = 0.855.

Under these conditions, TFS for individual areas was not correlated with anatomical hierarchy score (Pearson’s r = 0.57, P = 0.14), but was strongly negatively correlated with their baseline firing rates (Pearson’s r = −0.98, P = 0.00001

Methods, Response latency; Extended Data Fig. 9b, c

What the paper reports

• spontaneous time to first spike against hierarchy score: Pearson correlation, r = 0.57, p = 0.14 (n = 8 areas)
• spontaneous time to first spike against baseline firing rate: Pearson correlation, r = -0.98, p = 10⁻⁵ (n = 8 areas)

Spontaneous latency against hierarchy r = 0.290, p = 0.49 (no correlation, as stated); against the pre-flash rate (reading a) r = -0.991, p = 2e-06; against the overall rate (b) r = -0.987. 156 of 13,344 units have no spontaneous latency.

These metrics are not dependent on overall firing rate, which does not correlate with hierarchy score ( Extended Data Fig. 9i , j )

Results, A functional hierarchy of visual areas, paragraph 5; Extended Data Fig. 9i, j

Mean log10 rate r = -0.327, p = 0.43; mean rate r = -0.363, p = 0.38.

The LGN sits at the bottom of the hierarchy, followed by its major target structure, V1; areas LM, RL, LP and AL reside at intermediate levels, and areas PM and AM occupy the top level of the areas we studied here

Results, A functional hierarchy of visual areas, paragraph 1; Fig. 2a, b; Methods, Anatomical hierarchy analysis

What the paper reports

• RL against LM, bootstrapped hierarchy scores: Wilcoxon paired signed rank-sum test, p = 0.08 (n = 100 bootstrap samples of the connectivity data)

The scores in Figure3/Figure3.py order the areas LGN, V1, LM, RL, LP, AL, PM, AM. The bootstrap test of RL against LM needs the connectivity data of ref. 3 and was not run, so the part of the statement about significant differences is not tested.

the first spike latency for image responses during behaviour was correlated with the anatomical hierarchy score ( Fig. 4e )

Results, A functional hierarchy of visual areas, paragraph 6; Fig. 4e

From change_modulation_data.csv, means from 60.1 ms (LGd) to 77.2 ms (PM); r = 0.917, p = 0.0013.

During active behaviour, CMI was positive for each area ... More importantly, CMI systematically increased along the hierarchy from the LGN to the AM ( Fig. 4f ; Pearson’s r = 0.83, P = 0.011) ... CMI values were larger during active behaviour compared with passive viewing of the same stimulus sequence

Results, A functional hierarchy of visual areas, paragraph 7; Fig. 4f

What the paper reports

• Pearson correlation, r = 0.83, p = 0.011 (n = 8 areas, 4,057 units from 12 mice)

Active CMI r = 0.828, p = 0.011; positive in all eight areas (0.111 to 0.312); above the passive mean in all eight areas (means 0.221 and 0.140), so both readings of part 3 hold.

we found a significant correlation of CMI with hierarchy on hit but not on miss trials ( Extended Data Fig. 9k ; hit trials, Pearson’s r = 0.85, P = 0.007; miss trials, Pearson’s r = 0.51, P = 0.2)

Results, A functional hierarchy of visual areas, paragraph 7; Extended Data Fig. 9k

What the paper reports

• hit trials: Pearson correlation, r = 0.85, p = 0.007 (n = 8 areas)
• miss trials: Pearson correlation, r = 0.51, p = 0.2 (n = 8 areas)

Hit r = 0.853, p = 0.007; miss r = 0.510, p = 0.197.

Other aspects of neural activity during the task ... including the baseline firing rate, response to the pre-change image and the response to the change image ... were not correlated with hierarchy score

Results, A functional hierarchy of visual areas, paragraph 7; Extended Data Fig. 9l to n

Active baseline r = -0.584 (p = 0.13), pre-change r = -0.239 (0.57), change r = 0.107 (0.80). The passive columns are not significant either (p 0.079 to 0.66).

we found a strong increase in trial-wise decoder-behaviour covariation at higher levels of the hierarchy, starting with no correlation at the level of the LGN ( Fig. 4h ; Pearson’s r = 0.88, P = 0.004

Results, A functional hierarchy of visual areas, paragraph 8; Fig. 4g, h

What the paper reports

• Pearson correlation, r = 0.88, p = 0.004 (n = 8 areas)

r = 0.881, p = 0.0038; LGN -0.015 +/- 0.032 (3 sessions) includes zero.

Decoders were separately trained for 20 units within each area and performed significantly better than chance across trials, indicating that change trials could be read out from all 8 areas ... Decoder accuracy for each brain region (mean ± s.e.m., averaged across sessions) is not correlated with the anatomical hierarchy score

Results, A functional hierarchy of visual areas, paragraph 8; Extended Data Fig. 9p and its legend

Mean accuracy 0.688 to 0.835; mean minus s.e.m. above 0.5 and one-sided p < 0.05 in all eight areas (largest p 0.003, LP); accuracy against hierarchy r = 0.566, p = 0.14.

On average, 95% of gaze locations fell within 6.4 ± 2.1° of the mean, with a maximum of 13.6°

Methods, Eye and pupil tracking

50 sessions have gaze mapping. Radial 95% distance from the mean gaze 6.59 +/- 2.18 degrees, maximum 14.29 (reading a); per axis 6.21 +/- 2.13, maximum 13.37 (b). Raw gaze gives 6.99 and 15.96. 54% to 97% of frames per mouse have a finite gaze.

Units recorded in all eight cortical and thalamic visual areas were highly visually responsive, with 60% displaying significant spatial receptive fields within the boundaries of the monitor ... and found them in only 1.4% of units

Results, A survey of visually evoked spiking, paragraph 3; Fig. 1g

26,694 of 43,990 non-noise units of the eight areas (60.7%) have p < 0.01. In CA1, CA3 and DG 415 of 26,289 (1.58%) do pooled, and 1.71% on the mean over 52 mice; both readings are outside 1.4% +/- 0.14%. The criterion names no partial outcome; the 60% holds and the hippocampal fraction is near the paper's but outside its tolerance, so the claim holds in part. Without the 10-spike rule 60.8%, 1.61% and 1.74%; with the units without a p-value left out of the denominator 67.2%, 1.81% and 1.94%. The fraction in the eight areas is 55.4% at p < 0.001 and 62.5% at p < 0.02; in the hippocampus 1.03% at p < 0.005 and 2.58% at p < 0.02.

How the test differed from the paper’s method

• The p-value is computed from the spike times as the Methods describe (the AllenSDK value the figure code reads is not in the NWB files). Shuffles stop once 10 of them reach the observed statistic, which settles p >= 0.01; the threshold decision is the same as with 1,000.
• As the AllenSDK's ReceptiveFieldMapping (github.com/AllenInstitute/AllenSDK, tag v2.0.0, commit 5200c742, minimum_spike_count=10.0), the code the paper names for its stimulus metrics, a unit with fewer than 10 spikes over all Gabor presentations has no p-value and counts as not significant, as Figure1g.py counts it. The AllenSDK resamples presentations with replacement for its null; the Methods' location shuffles are used.

All ids of the three tables (99,180, 23,524 and 99,180 rows) are non-noise units of the dandisets; 664 non-noise units are absent from the layer table. Recomputed timescales agree with timescale_metrics.csv (99.3% of 23,352 units within 1% for the response decay timescale, 99.6% of 21,122 for the intrinsic timescale). Recomputed latencies do not agree with time_to_first_spike.csv; with a 200 ms window 16.1% of 62,483 units are within 1 ms, median absolute difference 4.0 ms, median signed difference -0.5 ms, and the 250 ms, dark, light and 500 ms readings 3.0% to 13.6%. On one session (1,591 units) none of 60 procedures (window ends 100 to 500 ms, all flashes or one half, median or mean) brings more than 14.5% within 1 ms. No script of the repository writes the latency table; Figure3/get_flash_PSTH.py reads it and writes a different file. Layer codes 1, 2 and 3 are also given to CA1, CA2 and CA3 units; with cortical_depth > 0 only VIS* units remain. ed10b and ed10ce use the layer table only, whose ids link fully.

Statistical testing revealed significantly different latencies for all pairs of areas, except for LGN–V1, RL–LP, LP–AL and AM–PM

Results, A functional hierarchy of visual areas, paragraph 4; Extended Data Fig. 9a

LGN-V1, RL-LP, LP-AL and PM-AM are not significant (corrected p 0.44 to 0.90), and RL-AL is not significant either (0.52); 27 of 28 pairs agree. Dark flashes also give one disagreement (LGN-LM); light flashes four.

Statistical testing revealed significantly different receptive field sizes for all pairs of areas, except for LM–RL

Results, A functional hierarchy of visual areas, paragraph 4; Extended Data Fig. 9d

LM-RL is not significant (corrected p = 0.41) as stated; LM-AL (0.38) and RL-AL (0.067) are not significant either. 26 of 28 pairs agree.

Statistical tests revealed significantly different modulation indices for all pairs of areas, except for RL–AL and AM–PM

Results, A functional hierarchy of visual areas, paragraph 4; Extended Data Fig. 9e

PM-AM is not significant (corrected p = 0.12) as stated, but RL-AL is significant (0.0006), and LM-RL (0.63), LM-LP (0.88) and RL-LP (0.54) are not. 4 of 28 pairs disagree.

How the test differed from the paper’s method

• The same assumed preferred condition as fig3i-modulation-index-hierarchy.

Statistical testing revealed significantly distinct response decay timescales for all pairs of areas, except for LM–AL and AM–PM

Results, A functional hierarchy of visual areas, paragraph 4; Extended Data Fig. 9h

With the Methods' window PM-AM is not significant (corrected p = 0.28) as stated, but LM-AL is (0.018), and LGN-LM (0.068), LGN-AL (0.81) and RL-LP (0.064) are not; 4 of 28 disagree. With the code's 40 to 290 ms window only LGN-V1 disagrees, which would be partially reproduced. Which window the paper used is asked of its authors (see findings).

n = 7,837 units from 58 mice ... n = 15,713 units from 58 mice unless otherwise specified

Fig. 3 legend

With the claims' procedure (latency within the Methods' 250 ms, decay timescale over 0 to 250 ms) 13,344 units after the latency cut (reading a) and 24,459 before it (reading b), against 15,713 +/- 1,571, and 5,842 for the decay timescale against 7,837 +/- 784. Variants, not judged; the scripts' windows (200 ms; 40 to 290 ms) give 17,163 and 8,381, the paper's tables 15,849 and 7,798, all within tolerance. The round-1 record adopted the scripts' windows on the reading that get_flash_PSTH.py wrote the paper's latency table; it reads that table, so the claims file's reading is kept. Which windows the paper used is asked of its authors (see findings). Without the firing_rate_dg > 0.1 Hz cut, whose definition is assumed, 13,476 and 25,019; the cut removes 560 units and does not change the outcome.

On average, in comparison to deep layers (5 and 6), superficial layers (2/3 and 4) had an earlier time to first spike (2.59 ms difference, P = 2.7 × 10 −19 , two-sided Wilcoxon rank-sum test), smaller receptive fields (109° difference, P = 1.1 × 10 −33 ), higher modulation index (0.09 MI difference, P = 5.0 × 10 −23 ), and faster response decay timescale (6.6 ms difference, P = 3.8 × 10 −33 )

Extended Data Fig. 10b legend; Results, A functional hierarchy of visual areas, paragraph 5

What the paper reports

• time to first spike, superficial against deep: two-sided Wilcoxon rank-sum test, p = 2.7 × 10⁻¹⁹
• receptive field area, superficial against deep: two-sided Wilcoxon rank-sum test, p = 1.1 × 10⁻³³
• modulation index, superficial against deep: two-sided Wilcoxon rank-sum test, p = 5 × 10⁻²³
• response decay timescale, superficial against deep: two-sided Wilcoxon rank-sum test, p = 3.8 × 10⁻³³

With the claims' procedure (the Methods' windows), deep minus superficial 1.34 ms latency (expected 2.59 +/- 0.26), 105 deg2 RF area (109 +/- 10.9) and 7.71 ms decay timescale (6.6 +/- 0.66), superficial minus deep 0.117 log10 MI (0.09 +/- 0.009); all four have the stated sign with p < 1e-4, and only the RF area is within tolerance, 1 of 4 parts. Raw MI gives 0.587. The code's layer split and the six areas alone change the differences by at most 0.09 ms, 4.3 deg2, 0.010 and 0.13 ms. Variants, not judged; the scripts' windows give 2.80 ms, 113 deg2, 0.127 and 6.39 ms, the paper's tables 2.55 ms, 110 deg2, 0.124 and 6.69 ms, 3 of 4 parts each.

How the test differed from the paper’s method

• The pool of all cortical units with a depth includes VIS, VISmma, VISli and VISmmp units, as the code's commented-out area line implies.

the peak offset distribution between V1 and LM showed a significant positive delay compared to the V1–V1 distribution ( Fig. 2d ; P = 2.6 × 10 −8 , two-sided Wilcoxon rank-sum test), indicating that V1 neurons ... lead LM neurons

Results, A functional hierarchy of visual areas, paragraph 2; Fig. 2d

What the paper reports

• two-sided Wilcoxon rank-sum test, p = 2.6 × 10⁻⁸ (n = V1-LM against V1-V1 peak offsets in one example mouse)

Judged under 2 s and all units; every reading gives the same outcome. In 4 of 15 mice with both distributions the median V1 to LM offset is 4.0 ms with V1-LM above V1-V1 (rank-sum p < 1e-6); 14 of 15 medians are positive and 13 of 15 tests significant with V1-LM higher. The V1-V1 reference counts each pair once, in a random orientation; counted twice it also gives 4 mice, and every other reading at least 4. Offsets are whole ms.

The maximum difference of this trajectory (here between V1 and AM) is defined as the total hierarchy score (THS) of the network, which is 0.89 for the measured functional network in our data ... The resulting DS matrix showed a gradient very similar to the real data, with a THS equal to 0.88

Methods, Network model simulation; Extended Data Fig. 8

2 s, all units, pooled DS; THS = 0.839, within 0.89 +/- 0.089 and between 0.15 and 1.54; V1 highest as source (0.447), AM lowest (-0.392). Without the diagonal 1.006; DS averaged across mice 0.727 (outside the tolerance); the other pooled readings 0.843 to 0.932. Every pooled reading is within the tolerance, so the outcome does not rest on the adopted reading.

Analysis of layer-wise CCG interactions indicated that superficial layers (2/3 and 4) were hierarchically lower compared to deep layers (5, 6) in the same area

Results, A functional hierarchy of visual areas, paragraph 5; Extended Data Fig. 10c to e

Superficial (layer codes 2, 4) to deep (5, 6) DS within an area, 2 s and all units; (a) 0.321, (b) 0.318, from 4,974 pairs; positive in every area (0.11 to 0.56) and under all four readings ((a) 0.220 to 0.321, (b) 0.264 to 0.318). 49 CCG units have no layer in the table.

All between-area distributions are significantly different from the within-area distributions at the 5% confidence level, except for AL–PM

Methods, Cross-correlation analysis, last paragraph; Fig. 2e legend

What the paper reports

• V1-LM against V1-V1: Wilcoxon rank-sum per mouse, combined across mice with Fisher's method, p = 0 (n = 25 mice)
• LM-RL against LM-LM: Wilcoxon rank-sum per mouse, combined across mice with Fisher's method, p = 1.9 × 10⁻⁵ (n = 25 mice)
• RL-AL against RL-RL: Wilcoxon rank-sum per mouse, combined across mice with Fisher's method, p = 2.4 × 10⁻⁵ (n = 25 mice)
• AL-PM against AL-AL: Wilcoxon rank-sum per mouse, combined across mice with Fisher's method, p = 0.081 (n = 25 mice)
• PM-AM against PM-PM: Wilcoxon rank-sum per mouse, combined across mice with Fisher's method, p = 3.2 × 10⁻⁴ (n = 25 mice)

2 s, all units; V1-LM (Fisher p = 3e-148, 15 mice), LM-RL (3.5e-11), RL-AL (5.8e-15) and PM-AM (2.7e-06) differ, as stated, but AL-PM differs too (5.0e-04, 8 mice); 4 of 5. The within-area reference counts each pair once in a random orientation; in file order AL-PM p = 1.3e-04, counted twice 1.1e-04. The three other readings also give 4 of 5 (AL-PM p 2.0e-07 to 0.048), so the outcome does not rest on the adopted reading.

How the test differed from the paper’s method

• A within-area pair enters the reference once, in a random orientation (seeded), since the sign of a within-area offset has no meaning; a standard choice, with file order and both orientations shown.

On average, V1 units lead the activity of units in other areas ( Extended Data Fig. 7c , left column); by contrast, area AM follows other regions

Results, A functional hierarchy of visual areas, paragraph 3; Extended Data Fig. 7c

2 s, all units; the mean median offset is positive from V1 to all five areas (0.68 to 5.22 ms) and into AM from V1, LM, RL and AL (1.92 to 5.22 ms); PM to AM is -1.11 ms. 9 of 10. The Extended Data Fig. 1j readings also give 9 of 10 and 3 s with all units 8 of 10 (V1 to RL -0.20 ms), all partial, so the outcome does not rest on the adopted reading.

The matrix of pairwise directionality scores ( Fig. 2e ) between areas was very similar to the matrix of anatomical hierarchy score differences ( Fig. 2b ) (Pearson’s r = 0.74, P = 1 × 10 −4

Results, A functional hierarchy of visual areas, paragraph 3; Fig. 2e, f

What the paper reports

• Pearson correlation, r = 0.74, p = 10⁻⁴ (n = 21 area pairs (lower triangle and diagonal))

2 s, all units, DS from offsets pooled over mice; r = 0.827, p < 1e-5 over 21 entries, above the tolerance (0.666 to 0.814). The criterion names no partial outcome; the correlation is present and significant but stronger than the paper's, so the claim holds in part. With any one mouse left out r = 0.802 to 0.839, and 3 of the 26 exclusions (mice 820866111, 800250054, 763884096) give r within the tolerance (0.802 to 0.811); the paper used 25 mice and the one left out is not identified, so the outcome also rests on which 25 mice are used. Pooled DS under the other readings r = 0.800 to 0.842; the two Extended Data Fig. 1j readings (0.810, 0.800) are within the tolerance and 3 s with all units (0.842) is not. DS averaged across mice 0.777 to 0.817; zero lags counted positive 0.621 to 0.786. The paper's precomputed matrix gives r = 0.735 with the same anatomy differences.

How the test differed from the paper’s method

• Assumed; no reading reproduces fig2-sharp-peak-pairs, so the claims that use the sharp peaks are judged under all non-noise units and the 2 s window, the reading whose pair count matches the paper's and whose peak count is nearest it. The paper does not state either.
• Within-area pairs enter both ways (offset and its negative), so the diagonal DS is 0; the paper's precomputed diagonal is near 0 (0.008 for V1).
• All 26 Functional Connectivity mice are used, where the paper uses 25; 3 of the 26 single-mouse exclusions put r within the tolerance.

yielding 16,119 pairs of units out of 2,089,890 possible pairs within the cortex ( Extended Data Fig. 7b ; 0.96% ± 0.13% per mouse, n = 25 mice)

Results, A functional hierarchy of visual areas, paragraph 2; Extended Data Fig. 7b

No reading gives the three values within tolerance. Window 2 s and all non-noise units above 2 Hz (10,283 units in 26 mice); 21,962 sharp peaks of 2,116,259 pairs (pairs within tolerance, peaks above 16,119 +/- 1,612), 1.18% per mouse (SD 0.69%, s.e.m. 0.14%; above 0.96 +/- 0.13). The 3 s window gives 23,474 peaks; the Extended Data Fig. 1j filters give 544,044 pairs, far from the paper's 2,089,890. With any one mouse left out, 19,937 to 21,730 peaks and 1.10% to 1.22% per mouse. The round-1 result (17,592 peaks, within tolerance) came from an error in the expected CCG under jitter at negative lags, now fixed and checked against a direct enumeration of spike pairs. The paper's code subtracts one random jitter resample, which adds noise to the flanks; that variant was not run, as the jitter function it imports is not public.

How the test differed from the paper’s method

• 26 Functional Connectivity mice, where the paper uses 25; the one left out is not identified.
• The jitter-corrected CCG subtracts the exact expected CCG under jitter of both trains within 25 ms windows (the Methods' true expected value); the paper's code subtracts one random resample (a variant the claims file names, not run because the jitter module get_ccgjitter imports is in neither repository the paper links).
• Bin 1 ms, from the example in Figure2/compute_ccg.ipynb, as the claims file gives.

this organization was absent during spontaneous activity ( Extended Data Fig. 7f , g )

Results, A functional hierarchy of visual areas, paragraph 3; Extended Data Fig. 7f, g

During the 30 min grey screen, 7,326 sharp peaks of 2,116,259 pairs (94 to 547 per mouse; 21,962 during gratings). With all non-noise units the DS matrix averaged across mice correlates with the hierarchy differences; 2 s segments r = 0.497, p = 0.022; 3 s r = 0.484, p = 0.026; pooled DS r = 0.435 and 0.445 (p = 0.049, 0.043). With the Extended Data Fig. 1j filters it does not (r = 0.142, p = 0.54; r = 0.170, p = 0.46). The outcome rests on the unit set; the one whose pair count matches the paper's is adopted. With any one mouse left out the 2 s correlation stays significant in 26 of 26 (p 0.009 to 0.043) and the 3 s one in 24 of 26 (p up to 0.059). The result is borderline, and it is measured under a reading with 36% more sharp peaks during gratings than the paper's (21,962 against 16,119). The round-1 correlations (r = 0.50 to 0.60 under every reading) were partly an effect of the jitter error.

How the test differed from the paper’s method

• The spontaneous block is cut into consecutive 2 s or 3 s segments and the units are those selected during gratings; the paper does not say how it was done.
• Assumed; the unit set is all non-noise units above 2 Hz, the reading whose pair count matches the paper's (fig2-sharp-peak-pairs). The paper does not state it, and the Extended Data Fig. 1j filters give the opposite outcome.