CRCNS:pfc-2

most neurons were generically active for similar standardized distances of 0.27 (62 cm) ± 0.17 (39 cm) (mean ± s.d.) in the maze (as determined by the 50% firing boundaries of the peak firing rate

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 3; Supplementary Figure 2B

Under PSTH reading a (all trials), width i (contiguous run) and cells y (all 627 cells) the width is 0.209 +- 0.154. Readings a-ii-x (0.235 +- 0.158), b-ii-y (0.267 +- 0.171) and c-ii-y (0.239 +- 0.156) also match; in the other eight combinations the mean is within tolerance and the s.d. (0.108 to 0.188) is not. Per rat (c-i) 0.195, 0.198 and 0.143.

How the test differed from the paper’s method

• 1,000 permutations (the claims file's standard choice). Spikes are placed on the maze between the end of the nose-poke and the start of reward consumption, the trial start and end the data description defines (page 2, SessionNP); the files' own spike labels also cover the nose-poke.
• Widths are measured on a grid of 200 points (0.005 of the maze), a standard discretisation.

found only spare evidence that any neurons fired differentially on the basis of previously visited positions ... Note that only a small fraction of neurons show retrospective coding

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 4 (end, page 4); Supplementary Figures 7 and 8

With previous-trial labels 4.3% of EE188 cells (5/117) are selective, against 77.8% with current labels on the same trials; over all sessions 7.3% of layer 2/3 and 8.9% of layer 5 cells, against 66.9% and 65.8%. The paper states no chance level; the test's false-positive rate with shuffled labels is 5.1% of cells at level 0.05, so the previous-trial fractions are close to chance.

How the test differed from the paper’s method

• The trial table lists correct trials only, so where an error trial was left out the previous trial is the previous correct one.
• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells.
• 1,000 permutations (the claims file's standard choice). Spikes are placed on the maze between the end of the nose-poke and the start of reward consumption, the trial start and end the data description defines (page 2, SessionNP); the files' own spike labels also cover the nose-poke.

left and right trajectory-specific neurons (Fig. 1c and Fig. 2) were not clustered together but occupied a large neuronal volume

Results, page 7, paragraph 1; Figure 3c; Discussion, paragraph 2

In EE188 (37 left, 29 right, 15 both, 36 none) the chi-square test of shank against class gives p = 0.577, Cramer's V = 0.23. In the variant over the other sessions, FF130 gives p = 0.047 (V = 0.39) and the other 11 p = 0.105 to 0.942.

How the test differed from the paper’s method

• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells.

The symmetry of these histograms around 0 suggests that, unlike in the hippocampus, this population of neurons does not exhibit a net positive relationship between firing rate and speed.

Supplementary Figure 3C

Over 627 cells the median correlation is -0.018 (45% positive, range -0.52 to 0.64), not above 0; the Wilcoxon test against 0 gives p = 0.00036, a small negative shift, so the distribution is not exactly symmetric around 0. Per rat medians -0.023, -0.007 and 0.031. By CellsInfo type, pyramidal -0.007 (p = 0.065) and interneurons -0.078 (p = 3.2e-05); by the connection-based types (G-b B-a W-a C-a), -0.015 (p = 0.041) and -0.088 (p = 0.065). The criterion tests only a shift above 0; whether a significant shift below 0 counts against the statement's symmetry around zero is asked of a reader (questions/crcns-pfc-2.yaml).

How the test differed from the paper’s method

• Speed is the path length of the LED midpoint (LED 1 alone where LED 2 is not tracked) per 0.5 s bin, in pixels per second.
• The groups are those of CellsInfo, with the connection-based types the claims file names as a variant; the criterion's decision uses all cells.

Neuronal populations were relatively evenly correlated throughout the maze segments with 0.18(41cm) ± 0.087(29cm) (L2/3) and 0.27(62cm) ± 0.13(30cm) (L5) in normalized length

Supplementary Figure 2A

With each cell's PSTH divided by its peak (the assumption), reading a gives 0.239 +- 0.060 for EE188 (layer 2/3) and 0.300 +- 0.078 for the pooled layer 5 sessions, within tolerance; reading b 0.265 and 0.338, also within; the pooled layer 2/3 sessions give 0.243 (a) and 0.268 (b, just outside). With z-scored PSTHs all values match (0.179 to 0.216). With raw rates the correlation stays above 0.35 everywhere and the width is 0.96 to 1.00, so the claim fails. The s.d. over positions (0.026 to 0.086) is below the paper's 0.087 and 0.13; the criterion does not test it.

How the test differed from the paper’s method

• Assumed that each cell's PSTH is normalised before the correlation, by its peak as Figure 2a displays the rates. The paper does not say; with raw rates the high-rate cells dominate and the correlation never falls to the 0.35 of the Supplementary Figure 2A colour scale. z-scoring is shown and gives the same outcome.
• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells. The layer 5 value comes from the pooled FF sessions.
• The width at a position is the contiguous length around the diagonal over which the correlation stays above 0.5, as the claims file reads it.

We recorded a total of 633 well-isolated units from the anterior cingulate area (area 24) and dorsal prelimbic area (area 32) of the medial prefrontal cortex (mPFC)

Results, paragraph 1

627 cells over the 13 working memory sessions (EE 446, FF 146, GG 35), within 10% of 633, but over three of the paper's four rats; 615 with EE188 counted as the 105 cells of EE188_Behavior_PFC.mat. The files list 1,495 clusters (spikeph). By the largest rate over the tenths of the maze, 0 to 10 cells per session are below 1 Hz (10 of 117 in EE188), and 5 of the 12 EE188 cells that EE188_Behavior_PFC.mat leaves out are below 1 Hz, consistent with the data description's 1 Hz rule for Cells without showing it. The match in number does not show that the same units are counted, so the claim holds only in part.

How the test differed from the paper’s method

• The data set holds three of the four rats; its rats EE, FF and GG are taken to be the paper's E, F and G of Supplementary Figure 1 (E and G layer 2/3, F layer 5), as the names match and EE.188, the Figure 2a and 3d session by the data description, is layer 2/3. The count is over three rats, the paper's over four.
• Each session's cells are those of its CellsInfo, a selection of the sorted clusters that the data description describes only as cells firing above 1 Hz in the trials; for EE188 the 117 cells of EE188_example.mat.
• Each cell's largest trial rate is its spikes over time spent in each tenth of the maze on left and right trials, a standard measure used only to compare the files' cells with the data description's 1 Hz rule.

In the recording sessions, the mean correct performance was 91.9%.

Methods, Behavioral task

The trial tables list correct trials only, but the session tar of EE188 holds the .eeg file with the two nose-poke sensor channels (97 right, 98 left), and the position file covers the whole session. A trial is the run of pokes since the last arm-end visit and its choice the arm end reached next. EE188 has 46 trials, 45 correct and 1 error (97.8%, within 91.9 +- 9.19); with a 40 pixel radius the same, with 80 pixels 46 of 48 (95.8%). 41 of the table's 42 correct trials are found as correct trials of the same side; the remaining one (first poke 1133.5 s) lies within a detected correct trial of the same side whose first poke is 3.5 s earlier. The table leaves out 1 detected error and 3 detected correct trials; one of these (2245.1 s) reaches the arm 25.9 s after its first poke, within the 9.0 to 36.1 s of the table's trials, and the other two (822.9 and 2370.7 s) 199.8 and 259.9 s after it, longer than any table trial. Without these two, 43 of 44 (97.7%). The other 12 sessions' sensor channels are in their own tars and were not read, so the claim is tested on one session of 13.

How the test differed from the paper’s method

• One session of 13 (EE188), whose tar is already streamed for EE188_example.mat; each other session's tar is about 2 GB and would have to be streamed for its sensor channels.
• Pokes are the intervals when a sensor channel is below -15,000 (the channels sit near 0 and drop to about -30,000), a standard threshold; the first and last poke of each table trial match the table to within 1.6 ms.
• An arm end is reached within 60 pixels of the median end position of the table's correct trials of that side, a standard choice; 40 and 80 pixels are shown.

the fraction of simultaneously active neurons (10% and 20% in 100-ms windows, layers 2/3 and 5, respectively; Fig. 2c) were relatively constant in all segments of the maze

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 3; Figure 2c; Discussion, paragraph 2

Reading b (all sessions of each layer) gives 12.6% for layer 2/3 (outside 10 +- 1) and 20.3% for layer 5 (within 20 +- 2); reading a gives 11.8% for EE188, the only Figure 2a session in the data set. Every tenth of the maze is within 50% of the mean (layer 2/3 11.2 to 14.5%, layer 5 19.0 to 20.8%). Per session 7.6 to 21.9%.

How the test differed from the paper’s method

• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells.
• The data set holds three of the four rats; its rats EE, FF and GG are taken to be the paper's E, F and G of Supplementary Figure 1 (E and G layer 2/3, F layer 5), as the names match and EE.188, the Figure 2a and 3d session by the data description, is layer 2/3.
• Windows of 100 ms tile each trial from its start to its end; a window's position is the position file's at its centre (4.7% of layer 2/3 windows have none).

Left and right lap trajectories in segments 0 to 0.3 of the central arm overlapped; they began to differ significantly at position 0.3 ... Significant separation of left/right trajectories and head directions emerged at position 0.30 and 0.24, respectively (p<0.01)

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 4; Supplementary Figure 3B

What the paper reports

• permutation test with global bands, left and right labels shuffled, p < 0.01

In EE188 (42 trials) the lateral coordinate first differs significantly between left and right trials (global level 0.01) at 0.28 (reading a), within 0.30 +- 0.03; the distance to the mean path (reading b) at 0.43, the x coordinate at 0.48. Head direction differs from 0.01 under every reading, against 0.24, although left and right trials start 2 pixels apart. The head direction difference lies at the start of the maze (31 to 45 degrees at 0.005 to 0.025, 11 at 0.055, 0 to 8 from 0.105 on), so the rat leaves the nose-poke with its head turned differently on the two trial types. Smoothed, 0.28 for the lateral path. Other sessions give 0.01 to 0.43 for the lateral path; head direction is available only in EE152 (0.17) and EE165 (0.26).

How the test differed from the paper’s method

• Per trial, the mean of each variable in 100 bins of normalised position, gaps filled by linear interpolation; the binning and the absence of smoothing are standard choices, the smoothed version (kernel 0.05) is shown.
• Head direction is the angle from LED 1 to LED 2, compared by the circular difference of the condition means; the paper does not say how it was computed.
• The claims file names the lateral coordinate x. In this position file the stem runs along x (over positions 0 to 0.3 the mean path runs from x 554 to 308 at y 248 to 264) and the arms end at y 134 and 384, so reading a uses y, and x is shown as a variant.

most functionally connected pairs were detected locally and recorded by the same probe shank. The probability of putative connections decreased rapidly as a function of distance between the somata of the recorded pair (Fig. 3f), but connections were detected between neurons up to 1,200 mm apart

Figure 3f; Results, page 6, paragraph 1

Under G-b B-a W-a, 352 connected pairs are on the same shank and 195 on others; the same-shank probability is 6.20% (99% interval 5.41 to 7.07) against 1.24% (0.97 to 1.56) at 200 um, and falls further with distance; connections are found up to 1,400 um, the largest separation on the probe, not 1,200. The 1,400 um value rests on 2 excitatory and 2 inhibitory connections of 504 pairs (0.40% each, interval 0.02 to 1.83%). Every B, W and G combination gives the same pattern and the same largest separation.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• Distance is the shank separation (200 um per shank), as the claims file says.

In addition to a large fraction of putative pyramidal cells (72.2% and 84.7% in layers 2/3 and 5, respectively), many interneurons (58.3% and 83% in layers 2/3 and 5, respectively) ... were also trajectory specific

Results, pages 6 and 7; Figure 3c; Supplementary Figure 11B

Reading b (G-b B-a W-a) gives 69.0% of layer 2/3 pyramidal cells (127/184), 73.3% of layer 5 pyramidal cells (33/45), 59.3% of layer 2/3 interneurons (16/27) and 83.3% of layer 5 interneurons (5/6); three of four are within tolerance, the layer 5 pyramidal value is not (84.7 +- 8.47). B-a W-b gives 75.6% for it, G-a 69.4 to 73.5%; no combination puts all four within tolerance. Reading a (EE188) gives 76.8% and 50.0% for layer 2/3.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells.
• 1,000 permutations (the claims file's standard choice). Spikes are placed on the maze between the end of the nose-poke and the start of reward consumption, the trial start and end the data description defines (page 2, SessionNP); the files' own spike labels also cover the nose-poke.

cell pair 156-136, recorded from different electrode shanks, maintained steady firing rates between maze segments 0.2 and 0.5, yet short-latency coincident spikes significantly in excess of firing rate–controlled coincidences (the ‘expected coincidence rate’) occurred only between maze locations 0.4 and 0.6

Figure 4; Results, Behavioral modulation of monosynaptic interactions, paragraph 2

On left trials the excess coincidences of 156-136 form a single region in mid-maze, from 0.325 to 0.645 by the pointwise band (global p = 0.001, global band broken from 0.372 to 0.622); the end is within tolerance and the start is not (0.075 before 0.4, tolerance 0.04). Between 0.2 and 0.5 the left PSTH of 156 varies by 49% of its maximum and that of 136 by 59%, so 136 is not steady by the criterion's 50%. The cells are on shanks 5 and 4, and 156 to 136 is a connection under the test and in pairs_all. The effect (a single transmission region in mid-maze on left trials) is present, with an earlier start and a less steady target rate.

How the test differed from the paper’s method

• Cell numbers are those of CellsInfo in EE188_example.mat, which the data description gives as the session of Figures 2a and 3d.

Figure 6b shows a putative layer 2/3 interneuron innervated by two pyramidal cells, with ... pyramidal cell 197 showing a depressing effect with spike repetition. In contrast, the first spike of spike trains of cell 186 was ineffective

Figure 6b; Results, pages 7 and 8; Supplementary Figure 14

186-201 facilitates: relative to shuffled labels its later spikes drive 201 more than its first (difference -8.21 against a mean permuted difference of 0.76, P = 0.004, later spikes as any spike within 40 ms; -4.35, P = 0.012, in chained trains with the largest bin). 197-201 has a higher first-spike than later-spike height (difference 4.71 to 8.13), but the permuted differences are as large (mean 4.88 to 7.69) and P is 0.707 to 0.914, so it shows no depression. CellsInfo types 197 and 186 as pyramidal and 201 as interneuron.

How the test differed from the paper’s method

• 100 jitter surrogates and 500 permutations, and the P from both tails of the permutation distribution, as for fig6-depression-facilitation.

viewed as a population, the firing properties of mPFC neurons were quite homogeneous ... individual neurons fired transiently, but, as a whole, the population of neurons fired relatively uniformly over the entire apparatus

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 3; Figure 2a and 2b

In EE188 the peak positions per tenth are 17, 0, 2, 2, 16, 7, 7, 5, 0, 61 of 117 (chi-square against uniform p < 0.001), and the population mean rate per tenth is 0.57 to 2.33 times its mean. The layer 5 sessions of FF show the same pattern (variant). The PSTHs are spike counts per unit of normalised length per trial, as the Methods define them, so they are higher where the rat spends more time; in EE188 it spends 0.86 s per trial in the first tenth and 2.89 s in the last, against 0.47 to 0.71 s in the others (median speed 81 and 22 pixels/s, against 138 to 198).

How the test differed from the paper’s method

• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells.
• The peak position of a cell is that of the larger of its left and right PSTH peaks; the claims file gives Figure 2a's ordering without saying which condition.
• 1,000 permutations (the claims file's standard choice). Spikes are placed on the maze between the end of the nose-poke and the start of reward consumption, the trial start and end the data description defines (page 2, SessionNP); the files' own spike labels also cover the nose-poke.

The fraction of trajectory-selective neurons in the side arms (positions 0.5–1) was almost twice as large in deep (layer 5, 40%) than in superficial (layer 2/3, 22%) neurons

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 4; Figure 2b

Reading a gives 62.3% of layer 5 cells (91/146) and 54.3% of layer 2/3 cells (261/481); reading b 61.6% and 51.8%. Per session 28.6 to 78.1%. With shuffled labels the test flags 5.1% of 627 cells, near its level of 0.05.

How the test differed from the paper’s method

• The data set holds three of the four rats; its rats EE, FF and GG are taken to be the paper's E, F and G of Supplementary Figure 1 (E and G layer 2/3, F layer 5), as the names match and EE.188, the Figure 2a and 3d session by the data description, is layer 2/3.
• Only correct trials are in the data set, so reading T-a (correct trials only) is the one run.
• 1,000 permutations (the claims file's standard choice). Spikes are placed on the maze between the end of the nose-poke and the start of reward consumption, the trial start and end the data description defines (page 2, SessionNP); the files' own spike labels also cover the nose-poke.

a sizable but smaller fraction of neurons in both layer 2/3 and layer 5 (16% and 18%, respectively; Fig. 2b) was also differentially active in the central arm (segments 0–0.3)

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 4; Figure 2b

Reading a gives 31.4% of layer 2/3 cells (151/481) and 26.7% of layer 5 cells (39/146); reading b 28.7% and 22.6%. Per session 8.3 to 57.1%.

How the test differed from the paper’s method

• The data set holds three of the four rats; its rats EE, FF and GG are taken to be the paper's E, F and G of Supplementary Figure 1 (E and G layer 2/3, F layer 5), as the names match and EE.188, the Figure 2a and 3d session by the data description, is layer 2/3.
• Only correct trials are in the data set, so reading T-a (correct trials only) is the one run.
• 1,000 permutations (the claims file's standard choice). Spikes are placed on the maze between the end of the nose-poke and the start of reward consumption, the trial start and end the data description defines (page 2, SessionNP); the files' own spike labels also cover the nose-poke.

Approximately one-quarter of the neurons showed significantly different ... firing rates in response to one of the two odorants ... Although 26% of neurons showed significant odor preference (t-test, p<.05)

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 4; Supplementary Figure 6B

In EE188, with the rate over the sensor-detected pokes, 36 of 117 cells (30.8%, Student) or 34 (29.1%, Welch) differ between left and right trials, against 26 +- 2.6. With the trial table's window from the first poke to the end of the last, which includes the pauses between pokes (median 4.59 s against 2.43 s of poking), 45.3% and 42.7%, and 32.4% over all sessions.

How the test differed from the paper’s method

• The window is the sensor-detected pokes, from channels 97 and 98 of the session's .eeg file, as Supplementary Figure 6A defines it (duration of nose-poking, detected by a photoelectric sensor); the rate is the spikes during the pokes over their summed duration. The trial table's window (first poke to end of the last) is a variant, and the only window for the other sessions.
• Left and right trials are taken as the two odors, as the claims file says.

we ruled out this possibility after examining, and finding no reliable relationship among, the differential firing patterns in the stem area and during nose poking

Results, Medial prefrontal cortical units predict behavioral choice, paragraph 4 (end, page 4); Supplementary Figure 6B

With the odor preference from the sensor-detected pokes, the Spearman correlation of the nose-poke difference with the side-arm difference is 0.237 (p = 0.010), an association with the same sign. The Fisher tests give odds 1.31 (p = 0.549) in the stem and 1.87 (p = 0.161) in the side arms; the stem Spearman correlation is -0.155 (p = 0.095). With Pearson the side-arm correlation is -0.007 (p = 0.94) and the claim would be reproduced as the criterion is written, so the outcome rests on the coefficient. The stem shows an opposite-sign association (Pearson -0.278, p = 0.0024). With the trial table's window (variant) no same-sign association is found (Fisher p 0.191 and 0.267, Spearman -0.184 and 0.134), and the stem Pearson correlation is -0.411. Whether opposite-sign associations count against the claim is asked of a reader (questions/crcns-pfc-2.yaml).

How the test differed from the paper’s method

• Assumed that the correlation of reading b is Spearman's, as the rate differences are heavy tailed and a rank correlation is not driven by a few high-rate cells; the claims file does not name the coefficient. Pearson is shown and gives the other outcome.
• The odor preference is from the sensor-detected pokes (odor-selective-fraction); the trial table's window is a variant.

Of the 62,408 cell pairs (counting each literal pair twice, corresponding to the two directions), 495 (0.79%) had short latency ... 0.55% excitatory connections and 0.24% inhibitory connections (single directions); 0.17% of cell pairs were connected reciprocally

Results, Characterization of mPFC neurons and their connections, paragraph 4; Figure 3e

40,240 directed pairs (three of four rats). Under G-b, 0.94 to 1.32% of pairs are connected, 0.84 to 0.97% excitatory and 0.11 to 0.39% inhibitory over the four B and W combinations; under G-a 1.07 to 1.50%, 0.94 to 1.09% and 0.14 to 0.47%; no combination puts the three directed values within tolerance. Reciprocal pairs are 0.045 to 0.199% of directed pairs under G-b; 0.179% of unordered pairs under G-a B-b W-a is the only reciprocal value within tolerance. G-b B-a W-a finds 41 of the authors' 43 pairs_all connections for EE188, G-a B-a W-a all 43 (W-b 42), B-b 34. With every EE188 train shifted in time the detector flags 0.12 to 0.15% of pairs as excitatory and 0.04 to 0.07% as inhibitory (G-b B-a W-a), against 0.70% and 0.26% unshifted.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• The data set holds three of the four rats; its rats EE, FF and GG are taken to be the paper's E, F and G of Supplementary Figure 1 (E and G layer 2/3, F layer 5), as the names match and EE.188, the Figure 2a and 3d session by the data description, is layer 2/3.
• 1,000 jitter surrogates per session; lags are rounded to the nearest bin centre.

a sizable fraction of the recorded units could be classified as putative pyramidal cells (32.5%) or inhibitory interneurons (12.5%). A large percentage of the postsynaptic targets of the putative excitatory cells (50.7%) were suppressing other neurons ... The ratio of putative principal cells to inhibitory interneurons in the entire population, identified by physiological criteria, was 2.82 ± 0.51 (s.d.)

Results, Characterization of mPFC neurons and their connections, paragraph 4

Under G-b, 35.2 to 36.5% of cells are putative pyramidal cells (C-a; 36.3 to 38.8% with C-b), within tolerance only under B-b, and 2.2 to 6.1% interneurons; 16.9 to 29.2% of the targets of excitation are interneurons; the ratio is 6.28 to 16.57 per session (7 to 9 of 13 sessions have an interneuron). Under G-a 37.8 to 41.8%, 2.2 to 6.4% and 18.2 to 31.8%. The authors' CellsInfo for EE188 gives 28.2% pyramidal cells and 12.8% interneurons.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• Sessions or rats without an interneuron give no ratio and are left out (4 to 6 sessions).

Monosynaptic excitation between putative pyramidal cells was detected in 0.12% of layer 2/3 and 0.27% of layer 5 pairs

Results, Characterization of mPFC neurons and their connections, paragraph 4

Under G-b, pyramidal to pyramidal excitation is in 0.146 to 0.176% of layer 2/3 pairs and 0.109 to 0.219% of layer 5 pairs (reading a), or 0.81 to 0.98% and 1.15 to 1.89% of pyramidal pairs (reading b); under G-a 0.164 to 0.176% and 0.191 to 0.246%. Only the layer 5 value under G-a B-b W-a (0.246%) is within tolerance.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• The data set holds three of the four rats; its rats EE, FF and GG are taken to be the paper's E, F and G of Supplementary Figure 1 (E and G layer 2/3, F layer 5), as the names match and EE.188, the Figure 2a and 3d session by the data description, is layer 2/3.

Figure 3b illustrates significant peaks and troughs of crosscorrelograms of 1172 pairs of layer 2/3 cells (13,572 matrix points) simultaneously recorded in a single session

Figure 3b legend; Results, paragraph 5 of the section; Supplementary Figure 11 legend

In EE188 (117 cells, 13,572 ordered pairs), under G-b, 78 to 81 connections have a pyramidal reference cell and 6 to 18 an interneuron (reading a), from 52 to 56 and 3 to 9 reference cells (reading b), against 48 and 30; under G-a 84 to 86 and 5 to 21. The authors' pairs_all with their CellsInfo types gives 40 and 3. The layer 5 values (73 and 25) cannot be tested.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells.

a large portion of the active neurons (39 of the 117) belonged to a single interconnected circuit ... a large fraction (45 of 142 total) belonged to a single "hub" of network

Figure 3d; Results, page 6, paragraph 2; Supplementary Figure 11C

The largest connected circuit in EE188 holds 82 or 83 (B-a) or 59 (B-b) of 117 cells under G-b, 85 or 86 and 64 under G-a, against 39. The authors' pairs_all graph (43 connections) has a largest component of exactly 39 cells. If Figure 3d draws pairs_all, which the matching count suggests but does not show, it draws a subset of the connections found (G-a B-a W-a finds all 43, G-b B-a W-a 41, B-b 34), which the paper does not describe. The layer 5 value (45 of 142) cannot be tested.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• The layer 5 session of Figure 2a and Supplementary Figure 11 (142 cells) is not in the data set; no FF session has more than 64 cells.

The 1-2% connection probability we detected between the physiologically identified nearby (<200 μm) pyramidal cell-interneuron pairs by the cross-correlation method

Supplementary Note, Potential caveat with the cross-correlation method, paragraph 3

Under G-b, of same-shank pairs of a putative pyramidal cell and interneuron, 38.0 to 63.3% have a pyramidal to interneuron connection (reading a) and 42.9 to 64.6% a connection either way (reading b), over 76 to 163 pairs; under G-a 33.8 to 60.9%. The cell types are defined by the connections themselves, so most such pairs are the connections that typed them. With the authors' CellsInfo types and pairs_all in EE188, 23 of 54 pairs (42.6%).

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.

Across all sessions, out of 343 pairs with significant excitatory monosynaptic connections in all sessions, 67 pairs showed identifiable position dependence in monosynaptic interactions by this measure

Results, Behavioral modulation of monosynaptic interactions, paragraph 1; Figure 4b; Supplementary Figure 12

Of 389 excitatory connections (G-b B-a W-a), 31 to 34 show position dependence under the four readings (8.0 to 8.7%); B-b gives 30 to 36 of 349, G-a 31 to 36 of 387 to 439. Jittering only the presynaptic cell gives 46 (11.8%, trials a) or 53 (13.6%, trials b) of 389. Per rat (B-a, a, i) EE 27 of 306, FF 4 of 78, GG 0 of 5. With the presynaptic spikes shifted by 20 ms, 0 of 39 pairs are significant.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• A coincident event is placed at the presynaptic spike's original position; the jitter moves it by at most 5 ms.

Putative monosynaptic connections that were active selectively in maze segments during left or right turn trajectories (15 of 36 excitatory connections; same set of neurons, and session, as in Fig. 3d)

Figure 5a legend

In EE188, 6 or 7 of 95 excitatory connections (G-b B-a) and 7 to 9 of 81 (G-b B-b) are position dependent, against 15 of 36; under G-a 6 or 7 of 105 and 7 to 10 of 88; of the 43 pairs_all connections, 5 or 6.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.

in pair 197-201, the expected coincidence rate was equally high between maze locations 0 and 0.5 of the left trials, yet significant spike transmission was detected only between maze locations 0.4 and 0.5 (Fig. 5b). In pair 49-52, the expected coincidence rate in the first two segments was equally high on left and on right trajectory trials, yet significant spike transmission was detected only on left trajectory trials. In pair 186-201, significant effects were observed only toward the end of the right arm

Figure 5b; Results, Behavioral modulation of monosynaptic interactions, paragraph 2

197-201, left trials, one region from 0.355 to 0.525 (global band 0.398 to 0.432), so its start misses 0.4 +- 0.04 by 0.005; none on right trials. 49-52 has no region on either side (global p = 0.55 left, 0.80 right). 186-201 has none on left trials (p = 0.065) and one on right trials from 0.715 to 0.975 (global band from 0.752), which begins before 0.75. No pair meets the criterion as written; all three are connections under G-b B-a W-a.

How the test differed from the paper’s method

• Region extents are given at the edges of the 0.005 grid bins.

Applying this approach, we indeed found that in several cases, the conclusion of position-dependent synaptic efficacy remained unaltered after thinning

Results, Behavioral modulation of monosynaptic interactions, paragraph 1; Methods, Controlling for power; Supplementary Figure 13

The thinning reaches its target (each cell's PSTH minimum within 10% of the maximum in the region) in 26% of thinnings under reading a (0.2 to 0.5) and 23% under reading b; it stops early when no spike is left within s of the maximum. Counting a repetition only when both cells were flattened, 0 of 16 pairs keep a significant and a non-significant part in the majority of 10 thinnings under reading a and 1 of 36 under reading b (EE196 8-7, 10 of 10), against at least two. Without the flatness condition 2 pairs hold under reading a (FF130 57-60, 10 of 10; FF130 57-51, 6 of 10), but their thinned PSTHs keep a minimum of 0 to 8% of the maximum.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• The pairs are the position-dependent connections of G-b B-a W-a, trial readings a and dependence i.
• A repetition counts only when both cells' thinned PSTHs reach the Methods' target (minimum within 10% of the maximum); the Methods' thinning stops when no spike is left within [-s, +s] of the maximum, which here happens before the target in most thinnings. The count without this condition is shown.

Among the significant pairs, 32.2% were recorded by different shanks ... we found that approximately equal percentages of neuron pairs showed firing pattern–dependent depressing (12.7%) and facilitating (10.7%) effects (Fig. 6c)

Figure 6c; Results, page 8, paragraph 1

With the P taken from both tails of the permutation distribution, depressing pairs are 4.0 to 5.9% and facilitating pairs 5.1 to 10.6% of the connections over the eight readings under G-b (G-a 4.6 to 6.4% and 5.5 to 9.9%), against 12.7% and 10.7%; facilitation is within tolerance under two readings of each (later spikes as any spike within 40 ms, summed heights), depression under none. 27.3 to 33.3% of the significant pairs are on different shanks. 388 of 389 excitatory connections (529 of 530 in all) are testable. On synthetic trains the test finds a fourfold difference in efficacy in 20 of 20 runs in either direction and flags 3 of 20 runs with equal efficacy; with labels assigned at random to the data's spikes, 16 of 121 pairs (13.2%, 95% interval 7.8 to 20.6%) are significant at P < 0.10, consistent with 10% (binomial p = 0.23). The share of connections significant either way, 10.9 to 14.9% under G-b, is at this random-label rate, so the facilitation within tolerance under two readings is not partial support for the claim.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• 100 jitter surrogates and 500 permutations per pair and reading, fewer than the 1,000 the claims file names, for run time.
• The standardised heights grow with the square root of the number of spikes, so the permutation distribution of the difference is not centred on 0 when the two groups differ in size. The two-sided P is taken from its two tails, as the Methods define a two-sided permutation P (P+- = min(1, 2P+, 2P-)); depression is a difference in the upper tail (first spikes more effective than under shuffled labels), facilitation one in the lower tail. The first version compared absolute values with a null assumed centred on 0.

Coincident discharge of two presynaptic neurons within 5 ms was more effective than the sum of the effects of nonsynchronous spikes (Fig. 7a), and coincidence of three or four spikes resulted in a supralinear effect in various independently tested cell assemblies (Fig. 7b) ... showed only a linear additive effect on the cooperative ability of presynaptic neurons to discharge a postsynaptic partner

Figure 7a and 7b; Results, page 8, paragraph 2; Discussion, paragraph 4

16 satellites under G-b (15 under G-a; 14 in the paper). With 5 ms windows the mean transmission probability after two coincident presynaptic cells is 0.064 against 0.054 after one, below 2 p1 and 1 - (1 - p1)^2 in all 16 satellites; after three cells (2 satellites with enough events) 0.119 against 0.090, below both predictions. No window from 2 to 20 ms gives a mean above the linear predictions; at 2 ms 3 of 15 satellites lie above them.

How the test differed from the paper’s method

• The global band of the jitter test follows the Methods' stepping procedure (G-b): it is stepped inwards, excluding the k largest surrogate maxima and the k smallest minima, for as long as at least 99% of the surrogates lie entirely within both sides. The Methods build the band from the maximum and minimum of each surrogate (Identifying monosynaptic interactions), define a band's global acceptance level as the proportion of data sets lying completely within it, and take the global band as the minimal band with that level (Multiple hypothesis testing and global bands). The band holds a median 99.4% of surrogates (5th percentile 99.0%). The first version's 1% on each side (G-a) is shown for every claim as a variant.
• An event is a presynaptic spike with k cells spiking in the window ending at it and no presynaptic spike in the 4 ms after; single events have no other presynaptic cell within the window before or 4 ms after; values need at least 20 events. These definitions are standard choices; the paper does not give them.