DANDI:000013

Reproducing Hires et al. (2015) from dandiset 000013 45

Notebook reproduce_hires_2015.ipynb

• ✕Not reproducedMaterials and methods, Animals

  The 52 recordings come from 21 mice of four lines, 21 neurons from 7 C57BL/6 mice, 6 from 3 VGAT-ChR2(H134R) mice, 17 from 8 PV-ires-cre mice and 8 from 3 PV-ires-cre x Ai32 mice.

  We report recordings from a total of 21 mice. 52 loose-seal cell-attached recordings were made in the following mice ... 21 recorded neurons from 7 C57BL/6 mice, 6 recorded neurons from 3 VGAT-ChR2(H134R) mice

  Materials and methods, Animals

  52 recordings, but 23 subject identifiers. By genotype, C57BL6 9 mice and 22 neurons (paper 7 and 21), Ai32 x PV-Cre 3 mice and 7 neurons (paper 3 and 8), PV-Cre 8 mice and 17 neurons (as the paper), and 3 mice and 6 neurons with a blank genotype, which match the paper's VGAT-ChR2 counts but carry no label.

• ✕Not reproducedResults, L4 responses are temporally precise, paragraph 1; Figure 1, figure supplement 1E-Gverifier disagreed at first

  Mice whisked in bouts with a mean duration of 261 ms, a peak-to-peak amplitude of 15.7 deg and a frequency of 15.4 Hz, separated by rest.

  Mice whisked in bouts (mean bout duration, 261 ms; peak-to-peak amplitude, 15.7°; frequency, 15.4 Hz) interspersed with periods of rest

  Results, L4 responses are temporally precise, paragraph 1; Figure 1, figure supplement 1E-G

  Measured values are for peak-to-peak as twice the Hilbert amplitude with touch not breaking bouts. Amplitude 15.74 deg and frequency 15.76 Hz match; the mean bout duration is 91.6 ms (69.7 ms when touch breaks bouts) against 261 +- 26.1 ms, so no reading puts all three in tolerance, which the criterion requires; it gives no rule for a partial match. Bouts holding a complete cycle (a variant outside the claims file) last 213.8 to 238.3 ms but then have amplitudes of 33.9 to 35.7 deg. Partially reproduced before the first verification round.

• ✕Not reproducedResults, L4 responses are temporally precise, paragraph 1; Figure 1, figure supplement 1C-Dverifier disagreed at first

  Mice touched the pole 2.33 times on average before responding, and the reaction time from first touch to first lick was 367 +- 234 ms (mean +- s.d.).

  Mice touched the pole multiple times (mean number of touches, 2.33) before reporting perceived object location with licking (mean reaction time 367 ± 234 ms; mean ± s.d.)

  Results, L4 responses are temporally precise, paragraph 1; Figure 1, figure supplement 1C-D

  Touches before the response are 3.00 under reading (a) and 3.51 under (b), outside 2.33 +- 0.233 under both. The reaction time is 190 +- 700 ms over 3093 trials, within the tolerance of 234 ms only because the tolerance is the paper's s.d.; 544 trials have the first lick before the first touch, and with those left out it is 420 +- 411 ms. Partially reproduced before the first verification round.

• ✕Not reproducedResults, L4 responses are temporally precise, paragraph 8; Figure 3A-Cverifier disagreed at first

  For the example neuron, 74.4% of spikes fell in a window 8 to 40 ms after touch onset, which covers 15.3% of exploration time.

  For the neuron illustrated in Figure 3A , 74.4% spikes fall into a time window spanning 8 to 40 ms after touch onset, comprising 15.3% of overall exploration time

  Results, L4 responses are temporally precise, paragraph 8; Figure 3A-C

  Measured values are for the uniform shuffle, the onset before, 100 ms and the transition cutoff. For cell 23 the window runs from 8 ms to 36 to 38 ms under all 8 readings of the transition cutoff, within tolerance. The spike fraction is 78.4 to 79.1% with the closest onset on either side and 83.6 to 84.6% with the onset before; the time fraction is 19.0 to 20.2% under every reading, against 15.3 +- 1.53%, so no reading meets the criterion (as fig3c-touch-spikes-l4c2, all values within tolerance). Partially reproduced before the first verification round.

• ✕Not reproducedResults, L4 responses are temporally precise, paragraph 5

  Spike rates were low in both whisking (0.67 spikes/s) and non-whisking (0.23 spikes/s) periods.

  Firing rates were low in both whisking (0.67 spikes/s) and non-whisking (0.23 spikes/s) periods

  Results, L4 responses are temporally precise, paragraph 5

  Whisking leaves out licking cycles and the 70 ms after each touch, as the claims file defines it. No reading puts both rates in tolerance. The example cell's non-whisking rate is 0.227 spikes/s; its whisking rate 0.777 (2.195 with touch epochs), against 0.67 +- 0.067. The L4 C2 means (1.462 or 3.698), medians (0.376 or 2.270) and pooled rates (1.475 or 3.436) are further off. Without the exclusions (the earlier run's mask, a variant) the example cell's whisking rate is 1.257.

• ✕Not reproducedFigure 2, figure supplement 1A legend

  A touch-response latency could be determined for all 31 L4 C2 cells but for only four of the 10 L4 cells outside C2, for lack of a touch response.

  Latency from touch onset to onset of spiking response for 31 L4 cells in C2 (red), four L4 outside C2 (grey) and eleven L5 near C2 (yellow). Latencies could not be determined for six L4 outside cells and one L5 cell due to lack of touch response

  Figure 2, figure supplement 1A legend

  All 31 L4 C2 cells have a latency under both readings; of the 10 L4 cells outside C2, 7 have one from first touches and 10 from all touches, not 4. L5, not judged, has 10 or 9.

• ✕Not reproducedResults, Decoding touch, paragraph 1; Figure 4B; Figure 4, figure supplement 1

  A naive Bayes decoder reading 200 L4 C2 neurons integrated over 10 ms decodes the time elapsed since touch with high precision, a minimum uncertainty of 0.55 ms (95% confidence) at 10 ms after touch onset.

  a group of 200 neurons in C2, integrating in 10 ms windows after touch, also allowed decoding of elapsed time from touch with high precision (minimum 0.55 ms uncertainty with 95% confidence at 10 ms post-touch onset using a naïve Bayes decoder

  Results, Decoding touch, paragraph 1; Figure 4B; Figure 4, figure supplement 1

  With add-one smoothing the smallest uncertainty 5 to 15 ms after onset, at 8 ms, is an RMS error of 1.09 to 1.40 ms, a 95% half width of 2.0 ms or 1.96 x RMS of 2.14 to 2.74 ms, under both splits and both sets of time classes (c-all, from -21 ms, and c-post, from 0 ms); none is within 0.55 +- 0.055 ms. The smaller pseudocount (0.01), a variant, gives 1.18 to 1.48 ms RMS. The error falls with population size (19.9 ms at 10 neurons, 1.55 ms at 200). What the answer changed, the time classes and the smoothing are now readings of the criterion; both sets of classes were run before, and the outcome is as before.

  How the test differed from the paper’s method

  • The smaller pseudocount of the variant, 0.01, is a value of our choosing; the criterion gives none.
• ✕Not reproducedResults, Decoding touch, paragraph 2; Figure 4C

  Even 1000 L4 neurons give only a rough estimate of whisking phase, a resolution of 24 deg in C2 and 33 deg outside C2 at d' = 1 (76% correct discrimination).

  a large population of neurons (1000) was required to provide even a rough estimate of whisking phase (24° in C2, 33° outside at d′ = 1 performance, equivalent to 76% correct discrimination)

  Results, Decoding touch, paragraph 2; Figure 4C

  Measured values are from Poisson rates with add-one smoothing and random shifts, the first reading that reaches 76% correct (held-out counts never do). On the phase-tuning cycles of the whole trial (median 494 per C2 cell, 0.107 spikes per cycle, 4 of 31 C2 cells with none), held-out cycle counts with add-one smoothing never reach 76% correct in either group. Poisson rates with add-one smoothing reach it in C2 in only 42 and 38 of the 100 runs (random or even shifts), at a median of 102.8 and 98.2 deg over those runs, so in most runs C2 never reaches it; outside C2 reaches it at 90.0 deg (91 and 89 runs). With a floor rate C2 reaches it at 72.0 or 66.0 deg (91 runs each) and outside C2 at 41.2 or 36.0 deg (100 runs). Under every reading C2 is far from 24 +- 2.4 deg and worse than outside, so all readings give the same failing result; the criterion's "else methods_underspecified" is read as its rule for readings that disagree, which they no longer do. With the cycles restricted to the sample epoch, as the earlier run did and the criterion does not, the floor rate put C2 at 12 to 15 deg and the claim was blocked; that restriction, now a variant, gives 60.0 deg in C2 and 47.1 outside with add-one smoothing.

  How the test differed from the paper’s method

  • The floor rate, one spike over all of a cell's training cycles and bins, is a value of our choosing, as the paper gives none. Cells without a phase fit take preferred phase 0. Tests use 50 samples per phase and phase differences in 6 deg steps.
• ◐Partially reproducedTable 1, row L5 Near C2 (n=11)

  For the 11 L5 cells near C2 Table 1 gives, as mean +- s.d. (median), spikes per first touch 1.95 +- 2.96 (0.98), per later touch 1.28 +- 1.88 (0.42), touch-evoked spikes 24.0 +- 18.6% (24.4%), touch and whisking spikes 31.5 +- 16.4% (36.0%), modulation depth 0.18 +- 0.12 (0.13), non-whisking rate 12.0 +- 9.87 (12.8) spikes/s, whisking rate 16.2 +- 15.3 (13.5) spikes/s, onset latency 9.7 +- 5.2 (8) ms and minimum ISI 3.7 +- 2.5 (2.7) ms.

  Not quoted: Numbers from Table 1, row L5 Near C2, with the garbled mean of touch-evoked spikes read from Results, paragraph on Figure 3C; no sentence states them all.

  Measured values are for the first reading of each open detail, as for table1-l4-c2. Under the transition cutoff 11 to 13 of 18 values are within tolerance in every combination of readings, partially reproduced throughout. Rates (median whisking rate 12.82 against 13.5), mean depth (0.182), the means of spikes per touch and the touch-evoked fraction match; the medians of spikes per first and later touch (0.23 and 0.19 against 0.98 and 0.42), the median depth (0.106 against 0.13), the median latency (4 against 8 ms) and the median minimum ISI (1.5 against 2.7 ms) do not. The literal cutoff, a variant, gives 8 to 11.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 2; Materials and methods, Animals and Electrophysiology

  The study holds 52 single-unit recordings, 31 L4 cells in the C2 barrel, 10 L4 cells outside C2 and 11 L5 cells near C2, with 115 +- 60 (mean +- s.d.) trials per neuron.

  We targeted recordings to excitatory neurons in L4 of the principal barrel corresponding to the spared whisker (31 cells in C2; 10 additional outside C2) and L5 near the principal barrel (11 cells), guided by intrinsic signal imaging (number of trials per neuron 115 ± 60

  Results, L4 responses are temporally precise, paragraph 2; Materials and methods, Animals and Electrophysiology

  The electrode location of each file gives 31 L4 C2, 10 L4 surround and 11 L5 C2 cells, one per file, with 115.0 +- 60.6 trials per neuron. The files mark no subset of trials as analysed, so readings (a) and (b) of the trial count are the same.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 1; Figure 1, figure supplement 1B

  Across the 52 sessions the fraction of trials correct was 0.740 +- 0.086 (mean +- s.d.).

  n = 21 mice, 52 sessions; fraction of trials correct, 0.740 ± 0.086; mean ± s.d.

  Results, L4 responses are temporally precise, paragraph 1; Figure 1, figure supplement 1B

  Hits and correct rejections over all trials, 0.741 +- 0.085 over 52 sessions. The files mark no trials as analysed, so readings (a) and (b) are the same.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 1

  The sample epoch, when the pole is presented, lasted 1.54 to 4.05 s, 2.39 s on average.

  during a sample epoch lasting a few seconds (1.54–4.05, mean 2.39 s), a pole appeared in one of two locations on the right side of the head

  Results, L4 responses are temporally precise, paragraph 1

  Under reading (a), all trials pooled, the pole is in reach for 2.394 s on average, 1.543 to 4.049 s. Under reading (b), session means, the mean is 2.403 s but the range 1.930 to 3.125 s is outside tolerance. 617 of 5978 trials are cut by the end of the recording.

  How the test differed from the paper’s method

  • The sample epoch is read from the `pole_available` series, cut at the end of each 4.5 to 5 s recording; the paper's own measure may extend past it.
• ✓ReproducedResults, L4 responses are temporally precise, paragraph 3; Figure 1C-D

  The example L4 C2 neuron of Figure 1 fired at 0.73 spikes/s in the 500 ms before the sample epoch and 1.91 spikes/s during it, a 2.6-fold increase (Wilcoxon rank sum, p = 3.38e-7).

  The neuron had a low baseline firing (0.73 spk/s; 500 ms before start of sample epoch) that increased substantially during the sample epoch (2.6 fold; to 1.91 spk/s; p = 3.38e-7, Wilcoxon rank sum)

  Results, L4 responses are temporally precise, paragraph 3; Figure 1C-D

  What the paper reports

  • Wilcoxon rank sum, p = 3.4 × 10⁻⁷ (n = trials of one neuron)

  Cell 23 (jf35892, 2010-02-19) is the only L4 C2 cell within tolerance on all three, with rank-sum p = 1.36e-7 over 121 trials. Its rates from pooled spikes and time are 0.727 and 1.909 spikes/s, 2.63-fold, the paper's 0.73, 1.91 and 2.6. Cell 17 has no trial with 500 ms before the pole comes into reach and was not evaluated.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 4

  The Fano factor of spike counts over the whole sample epoch was large, 7.51, in the paragraph on the example neuron, which does not say whose value it is.

  FF computed by counting spikes over the entire sample epoch (stimulus presentation), was huge (FF = 7.51)

  Results, L4 responses are temporally precise, paragraph 4

  Reading (a), cell 23, gives 7.51, and reading (c), the mean over 52 cells, 7.89; both match. Reading (b), the mean over 41 L4 cells, gives 6.31.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 4; Table 1

  Each touch evoked on the order of one spike (1.51 spikes per first touch, 0.31 per later touch) with an onset latency of 8 ms.

  Each touch evoked on the order of one spike (1.51 spikes/first touch; 0.31/later touches) with short latency (onset, 8 ms)

  Results, L4 responses are temporally precise, paragraph 4; Table 1

  Under reading (a), the example cell, 1.474 to 1.505 spikes per first touch, 0.302 to 0.307 per later touch and an 8 ms latency under all 8 window readings of the transition cutoff (9 ms from the all-touch PCTH, also within tolerance). Readings (b) and (c) do not match (means 1.16 to 1.28 and 0.62 to 0.69, medians 0.70 to 0.73 and 0.31 to 0.33). The answer on the cutoff leaves the outcome as it was, since the example cell's window is the same under both starts; it moves the population values of (b) and (c), which were 0.53 to 1.28 and 0.28 to 0.68 over both starts.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 5; Figure 1G-H

  During exploration the example neuron's spikes are coupled to whisking phase, with a modulation depth close to 1 and near zero spiking at phase 0.

  Yet during active exploration, the timing of spikes was coupled to the phase of rhythmic whisker movement ( Fee et al., 1997 ; Curtis and Kleinfeld, 2009 ), with a modulation depth close to 1

  Results, L4 responses are temporally precise, paragraph 5; Figure 1G-H

  Depth 1.000 from 14 spikes in the phase-analysis cycles counted in exploration time, the criterion's exploration periods, with no spikes in the five bins centred from -75 to 45 deg, around phase 0. Over the same cycles in the whole trial, as Table 1, the depth is also 1.000, from 92 spikes, with one spike in those five bins (0.19 spikes/s at -15 deg).

• ✓ReproducedTable 1, row L4 Inside C2 (n=31)

  For the 31 L4 C2 cells Table 1 gives, as mean +- s.d. (median), spikes per first touch 1.36 +- 1.32 (0.87), per later touch 0.74 +- 0.87 (0.33), touch-evoked spikes 70.6 +- 20.9% (74.7%), touch and whisking spikes 75.2 +- 19.0% (80.3%), phase modulation depth 0.67 +- 0.32 (0.70), non-whisking rate 1.31 +- 2.88 (0.32) spikes/s, whisking rate 1.39 +- 2.33 (0.33) spikes/s, onset latency 7.8 +- 3.0 (8) ms and minimum ISI 2.9 +- 1.9 (2.3) ms.

  Not quoted: Numbers from Table 1, row L4 Inside C2; no sentence states them all.

  Measured values are for the first reading of each open detail (uniform shuffle, onset before, 100 ms, transition cutoff, phase window +-t_w, latency from first touches, whisking outside touch epochs). Under the transition cutoff, the criterion's after the answer, 15 of 18 values are within tolerance in all 64 combinations of the other readings, at the threshold for reproduced. The three outside are medians, spikes per first touch (0.70 to 0.73 against 0.87), modulation depth (0.859 against 0.70) and whisking rate (0.376 against 0.33); 12 of the 27 cells with a phase fit have depth 1, and with only the 11 cells of at least 20 phase spikes (a variant) the depth median is 0.444. Whisking periods are read as the claims file defines them, outside touch epochs and leaving out licking cycles and the 70 ms after touch; with touch epochs in, a variant, the count is also 15. Modulation depth uses every phase-analysis cycle of the trial. What the answer changed, the earlier run was blocked because the literal cutoff gave 9 to 14 values and the transition one 14 or 15; the literal cutoff is now a variant only (10 to 14). The 100 and 200 ms curves give 15 alike.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 6; Figure 2B

  L4 neurons outside C2 respond much more weakly to touch than L4 neurons in C2, for first and for later touches.

  In L4, but outside of C2, touch responses were much weaker (first touch, p = 1.4e-5; later touches, p = 6.32e-6; Wilcoxon rank sum)

  Results, L4 responses are temporally precise, paragraph 6; Figure 2B

  What the paper reports

  • spikes per first touch, L4 C2 against L4 outside C2: Wilcoxon rank sum, p = 1.4 × 10⁻⁵ (n = 31 and 10 cells)
  • spikes per later touch, L4 C2 against L4 outside C2: Wilcoxon rank sum, p = 6.3 × 10⁻⁶ (n = 31 and 10 cells)

  Reading (a), spikes per touch, gives p < 0.05 for first and later touches with the C2 median higher in all 16 window readings of the transition cutoff (p = 1.0e-4 to 0.0043 and 2.0e-4 to 0.015), and in the 16 of the literal variant. Reading (b), the peak touch rate, gives 148.1 against 25.0 spikes/s (p = 1.63e-5) and 51.3 against 5.6 (p = 1.07e-5). The answer on the cutoff leaves the outcome as it was, as both starts gave it.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 6; Figure 2B-D; Table 1

  L5 neurons near C2 have much higher spike rates than L4 neurons, with more diverse touch responses.

  Layer 5 neurons near C2 had much higher firing rates, with touch responses that were more diverse

  Results, L4 responses are temporally precise, paragraph 6; Figure 2B-D; Table 1

  L5 medians 12.90 (non-whisking) and 12.82 or 11.52 (whisking, without or with touch epochs) spikes/s against 0.32 and 0.38 or 2.27 in L4 C2 and 1.09 and 1.46 or 1.54 in L4 outside C2; all four rank-sum tests p < 0.05 under both readings (largest 0.0092). Diversity is not judged.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 6; Figure 2, figure supplement 1B

  Modulation by whisking phase does not differ between L4 neurons inside and outside C2.

  Modulation by whisking phase was not significantly different between L4 neurons inside and outside the C2 barrel (p = 0.68)

  Results, L4 responses are temporally precise, paragraph 6; Figure 2, figure supplement 1B

  What the paper reports

  • Wilcoxon rank sum, p = 0.68 (n = 31 and 10 cells)

  Rank-sum p = 0.584 between L4 C2 (27 cells with a fit, median 0.859) and L4 outside C2 (10 cells, median 0.548), depth from every phase-analysis cycle of the trial. Four L4 C2 cells have no phase-analysis spikes and no depth. With cells of at least 20 phase spikes only (a variant, 11 and 8 cells) p = 0.934; with the cycles counted only in exploration time (the earlier run's, a variant) p = 0.798.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 6; Figure 2, figure supplement 1B

  L4 neurons, inside and outside C2, are more modulated by whisking phase than L5 neurons.

  but both were significantly more phase modulated than L5 (p = 4.3e-5 and p = 0.012 respectively, Wilcoxon rank sum)

  Results, L4 responses are temporally precise, paragraph 6; Figure 2, figure supplement 1B

  What the paper reports

  • L4 C2 against L5: Wilcoxon rank sum, p = 4.3 × 10⁻⁵ (n = 31 and 11 cells)
  • L4 outside C2 against L5: Wilcoxon rank sum, p = 0.012 (n = 10 and 11 cells)

  L4 C2 against L5 p = 8.05e-5 and L4 outside against L5 p = 0.0092, L4 medians higher (0.859 and 0.548 against 0.106). With cells of at least 20 phase spikes only (a variant) p = 0.0043 and 0.032; with the cycles counted only in exploration time p = 2.9e-4 and 0.0031.

• ✓ReproducedResults, L4 responses are temporally precise, paragraph 6; Figure 2D

  L5 neurons fire modestly but significantly more during whisking than during non-whisking periods.

  whereas L5 showed a modest, but significant increase with whisking (p = 0.019, Wilcoxon signed rank)

  Results, L4 responses are temporally precise, paragraph 6; Figure 2D

  What the paper reports

  • Wilcoxon signed rank, p = 0.019 (n = 11 cells)

  Signed-rank p = 0.0186 without touch epochs and 0.0322 with them, 9 of 11 cells higher when whisking under each, median of the per-cell differences +2.10 and +2.43 spikes/s, positive under both readings, with whisking leaving out licking cycles and the 70 ms after touch. The group medians are 12.8 and 11.5 spikes/s against 12.9, shown only for comparison. Without the exclusions (a variant) p = 0.0244 and 0.0137. What the answer changed, the criterion now names the median of the per-cell differences as the measure of the increase; the earlier run read it as the group medians and recorded partially reproduced.

• ✓ReproducedResults, Low spike count variability, paragraph 2; Figure 5A-C

  Touch response magnitude of L4 C2 neurons is strongly modulated by touch order, whisker velocity before touch and maximum curvature during touch, with mean modulation indices of 0.71 +- 0.25, 0.71 +- 0.27 and 0.72 +- 0.23.

  Touch response magnitude was highly modulated by each of the three variables (mean modulation index ... touch order, 0.71 ± 0.25; velocity, 0.71 ± 0.27; max curvature, 0.72 ± 0.23; mean ± std

  Results, Low spike count variability, paragraph 2; Figure 5A-C

  Measured values are for the uniform shuffle, onset before, 100 ms, transition cutoff, order by ranks, velocity at onset and curvature over the touch. Under every reading with the transition cutoff the means are within tolerance, order 0.666 to 0.692, velocity 0.703 to 0.735 and curvature 0.664 to 0.698. The answer on the cutoff leaves the outcome as it was; the earlier measured values (0.853, 0.857, 0.827) came from the literal cutoff, now a variant (0.744 to 0.894, also within tolerance).

• ✓ReproducedFigure 6, figure supplement 1B legend

  The clustering of touches into five groups finds subsets with significantly less variability in touch properties than random samples of touches.

  The procedure finds subset of points with significantly less behavioral variability than in the full dataset

  Figure 6, figure supplement 1B legend

  28 of 31 cells (0.90) under all four readings; 28 or 29 cells have a group and 8 have all five, as most cells have fewer than 100 touches after the 250 ms interval.

  How the test differed from the paper’s method

  • Pre-touch velocity is taken as the frame before onset, fig5c's "pretouch" reading; the claims file does not repeat which reading applies here.
• ✓ReproducedResults, Low spike count variability, paragraph 4; Figure 6C-D

  Across L4 C2 neurons, the Fano factor following touch drops below one, lying close to the binomial limit (the theoretical minimum) for most cells.

  Across the population of neurons within C2 the FF following touch dropped below one, with the FF lying close to the binomial limit for most cells

  Results, Low spike count variability, paragraph 4; Figure 6C-D

  Mean minimum FF 0.504 to 0.676 and fraction of units close to the binomial limit 0.79 to 1.00 under all 16 combinations.

• ✓ReproducedResults, Low spike count variability, paragraph 7; Figure 6E, condition 1

  Counting spikes over the whole sample epoch, ignoring the millisecond structure of behavior, gives a very large Fano factor of 6.57 +- 5.41 (mean +- s.d. over L4 C2 cells).

  Response variability was very large (FF = 6.57 mean ± 5.41 s.d.) when we ignored the millisecond time scale of behavior

  Results, Low spike count variability, paragraph 7; Figure 6E, condition 1

  6.58 +- 5.40 over the 31 L4 C2 cells.

• ✓ReproducedResults, Low spike count variability, paragraph 7; Figure 6E, condition 2

  Counting spikes in random 38 ms windows, as many per trial as there are touches, gives a Fano factor of 1.59 +- 0.49.

  Integrating activity in time windows as wide as average touch response periods (38 ms) gave a FF of 1.59 ± 0.49

  Results, Low spike count variability, paragraph 7; Figure 6E, condition 2

  1.56 +- 0.47 within the sample epoch, 1.59 +- 0.47 anywhere in the trial.

• ✓ReproducedResults, Low spike count variability, paragraph 7; Figure 6E, condition 3

  Counting spikes in 38 ms windows aligned to touch onset (after a 6 ms latency) gives a Fano factor of 1.37 +- 0.74.

  and FF of 1.37 ± 0.74 ( Figures 6E, 3 ) when aligned to touch onset

  Results, Low spike count variability, paragraph 7; Figure 6E, condition 3

  1.39 +- 0.60 over all touches, 1.34 and 1.36 +- 0.74 over touches after a 250 ms interval.

• ✓ReproducedResults, Low spike count variability, paragraph 7; Figure 6E, condition 4

  A narrower 10 ms window after touch reduces the Fano factor to 0.73 +- 0.19.

  Using a narrower touch window of 10 ms reduced the FF to 0.73 ± 0.19

  Results, Low spike count variability, paragraph 7; Figure 6E, condition 4

  0.85 and 0.86 over all touches, 0.70 to 0.74 over touches after a 250 ms interval, under both readings of the window starts.

• ✓ReproducedResults, Low spike count variability, paragraph 7; Figure 6E, condition 5

  Computed only across similar touches, the 10 ms Fano factor falls further to 0.54 +- 0.27, so that L4 responses to touch are almost noiseless.

  The FF was further reduced to 0.54 ± 0.27 when calculated only across similar touches

  Results, Low spike count variability, paragraph 7; Figure 6E, condition 5

  0.504 to 0.691 under all 16 readings, with the five conditions in the paper's order under each.

• ✓ReproducedDiscussion, paragraph 7; Figure 4

  L4 neurons fire few spikes during whisker movements, on average fewer than 0.15 spikes per whisk cycle, which hinders decoding of whisking phase.

  their low spike rates during whisker movements (mean <0.15 spikes/whisk cycle) hinders efficient decoding of whisking phase from population activity

  Discussion, paragraph 7; Figure 4

  0.107 (L4 C2) and 0.121 (all L4) spikes per cycle as means over cells, 0.116 and 0.124 pooled, over the phase-analysis cycles of the whole trial (17669 and 27769 cycles); 11 cells are above 0.15. Inside the sample epoch only, the earlier run's cycles and a variant here, 0.092 to 0.114.

• ⊘BlockedFigure 6B legend; Results, Low spike count variability, paragraph 5· Methods underspecifiedverifier disagreed at first

  For the example neuron, the 10 ms sliding-window Fano factor is about 1 before and after touch, where the mean spike count is about 0.03, and falls near the peak of the touch response, where the mean count is about 0.7.

  Since the spike rate is very low outside of the touch window, the mean spike count is also very low (∼0.03), implying FF of ∼1 for both the binomial and Poisson models. In contrast, around the peak of the touch response the mean spike count is high (∼0.7)

  Figure 6B legend; Results, Low spike count variability, paragraph 5

  Cell 23 matches under no reading; its mean 10 ms count before touch is 0.000 to 0.008 against about 0.03, its peak mean count 0.15 to 1.36 and its FF at the peak 0.61 to 1.82. With pooled touches no L4 C2 cell matches (62 units); with grouped touches cell 11 matches in 2 of 394 cell and group units (groups of 20 touches, mean counts 0.0305 and 0.0280 before touch and 0.65 and 0.70 at the peak, FF 0.99 and 1.00 before and 0.37 and 0.32 at the peak). The criterion lets another cell decide the claim, and the two readings of the touches then give different outcomes. With about 400 units tried under the grouped reading, a match in two may come by chance.

  How the test differed from the paper’s method

  • The pre-touch windows are those ending at or before onset (starts -50 to -10 ms), a choice the claims file does not fix; the example cell's pre-touch rate is far from the paper's under any window.
• ⊘BlockedTable 1, row L4 Outside C2 (n=10)· Methods underspecifiedverifier disagreed at first

  For the 10 L4 cells outside C2 Table 1 gives, as mean +- s.d. (median), spikes per first touch 0.08 +- 0.19 (0), per later touch 0.02 +- 0.04 (0), touch-evoked spikes 9.2 +- 13.7% (3.6%), touch and whisking spikes 33.0 +- 30.2% (23.8%), modulation depth 0.58 +- 0.36 (0.61), non-whisking rate 2.20 +- 2.91 (1.10) spikes/s, whisking rate 2.41 +- 3.48 (1.31) spikes/s, onset latency 18.5 +- 6.0 (18) ms and minimum ISI 3.3 +- 1.3 (2.8) ms.

  Not quoted: Numbers from Table 1, row L4 Outside C2; no sentence states them all.

  Under the transition cutoff, the criterion's after the answer, 7 to 9 of 17 values are within tolerance, 52 of the 64 combinations of readings below the threshold of 9 (not reproduced) and 12 at it (partially); the readings of the touch window decide which. The medians of spikes per touch, counted among the 17, are never 0; the means of spikes per touch (0.31 and 0.15 under the first reading, against 0.08 and 0.02), the touch-evoked fraction (26.2%, median 23.8%, against 9.2% and 3.6%), the median depth (0.548 against 0.61), the median whisking rate (1.464 against 1.31) and the latency mean (8.1 ms against 18.5) are outside tolerance. Rates, mean depth and minimum ISI match. The literal cutoff, a variant, gives 10 to 11 (partially); the outcome stays blocked because the transition cutoff's own readings straddle the threshold.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 6; Figure 2D· Methods underspecified

  Across L4 neurons, inside and outside C2, spike rate does not change between whisking and non-whisking periods.

  There were no significant changes in firing rate between whisking and non-whisking across the L4 population (L4 inside C2, p = 0.75; L4 outside C2, p = 0.92)

  Results, L4 responses are temporally precise, paragraph 6; Figure 2D

  What the paper reports

  • L4 inside C2: Wilcoxon signed rank, p = 0.75 (n = 31 cells)
  • L4 outside C2: Wilcoxon signed rank, p = 0.92 (n = 10 cells)

  Whisking leaves out licking cycles and the 70 ms after each touch, as the claims file defines it. With touch epochs left out of whisking neither group changes, L4 C2 p = 0.595 (18 of 31 cells higher, median difference +0.074 spikes/s) and L4 outside C2 p = 0.625, reproduced; with touch epochs in, L4 C2 changes (p = 1.28e-7, 29 of 31 higher, +1.18 spikes/s) and L4 outside C2 does not (p = 0.695), partially reproduced. The criterion's two readings give different outcomes. Without the exclusions (the earlier run's mask, a variant) L4 C2 changes under both readings (p = 5.0e-5 and 2.8e-9); the earlier run recorded partially reproduced.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 8; Figure 3C· Methods underspecified

  In L4 C2 neurons 70.6 +- 20.9% (mean +- s.d.) of spikes during exploration are coupled to touch, within an average of 14.9% of exploration time.

  Over the L4 C2 population, 70.6% ± 20.9% (mean, s.d.) of spikes were coupled to touch in an average of 14.9% of exploration time

  Results, L4 responses are temporally precise, paragraph 8; Figure 3C

  Under the transition cutoff, the criterion's after the answer, 66.9 to 75.8% of spikes in 14.6 to 17.6% of exploration time. All 4 readings with the 100 ms curve put both values in tolerance, but only 1 of 4 with the 200 ms curve (the other three cover 16.5 to 17.6% of time), so the readings of the curve's range disagree. What the answers changed, the start of the search is now fixed, and the literal cutoff (38.6 to 64.5% in 8.5 to 13.4%, none within tolerance) is a variant only; the range of the curve, which the paper does not give, is now the criterion's open detail (4), and it alone keeps the claim blocked.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 8; Figure 3C· Methods underspecifiedverifier disagreed at first

  L4 neurons outside C2 have far fewer touch-coupled spikes, 8.4 +- 13.3%, in 3.5% of exploration time.

  L4 neurons outside of C2 had significantly less touch-evoked spikes, 8.4% ± 13.3% in 3.5% of exploration time (p = 6.53e-6, Wilcoxon rank sum)

  Results, L4 responses are temporally precise, paragraph 8; Figure 3C

  What the paper reports

  • touch-coupled fraction, L4 C2 against L4 outside C2: Wilcoxon rank sum, p = 6.5 × 10⁻⁶ (n = 31 and 10 cells)

  Blocked as it depends on fig3c-touch-spikes-l4c2. Every one of its 16 readings with the transition cutoff is outside tolerance on the time fraction (18.8 to 31.7% of spikes in 17.5 to 26.3% of time, against 8.4 +- 13.3% in 3.5%), with the rank-sum test against L4 C2 at p < 0.001. Per cell, under the first reading, five cells (34, 36, 37, 39, 40) have windows of 51 to 86 ms covering 23 to 49% of exploration time; three of them have latencies of 1 to 5 ms. With the answer the literal cutoff (1.5 to 7.4% in 1.2 to 2.6%) is a variant only.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 8; Figure 3C· Methods underspecified

  L5 neurons have fewer touch-coupled spikes than L4 C2 neurons, 24.0 +- 18.6%, in 19.0% of exploration time.

  L5 neurons in C2 also showed significantly less touch-evoked spikes, 24.0% ± 18.6% touch spikes in 19.0% of exploration time (p = 2.30e-5, Wilcoxon rank sum)

  Results, L4 responses are temporally precise, paragraph 8; Figure 3C

  What the paper reports

  • touch-coupled fraction, L4 C2 against L5: Wilcoxon rank sum, p = 2.3 × 10⁻⁵ (n = 31 and 11 cells)

  Depends on fig3c-touch-spikes-l4c2. Under the transition cutoff 4 of the 8 window readings meet the criterion under both latency readings (2 with the 100 ms curve, 2 with the 200 ms one; 20.8 to 30.6% in 15.4 to 22.5% of time over the 8). With the answer the literal cutoff (5.1 to 14.1% in 3.2 to 8.2%, none) is a variant only; the transition cutoff's readings still disagree.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 9; Figure 3D· Methods underspecified

  In L4 C2 neurons 75.2 +- 19.0% of exploration spikes are coupled to touch or to whisking phase, within an average of 22.0% of exploration time.

  Overall, for neurons in C2 75.2 ± 19.0% of spikes were coupled to touch or whisking phase in an average of 22.0% of exploration time

  Results, L4 responses are temporally precise, paragraph 9; Figure 3D

  Depends on fig3c-touch-spikes-l4c2. Under the transition cutoff, which the answer makes the rule for t_w too, 74.9 to 81.4% in 22.5 to 28.5% of time, and 1 of 16 readings meets both tolerances (circular shuffle, onset before, 100 ms curve, total width, 76.4% in 22.5%). The preferred phase now comes from the fit over every phase-analysis cycle of the trial; with the sample-epoch fit of the earlier run 4 of 16 met it. The literal cutoff (41.9 to 67.5% in 10.3 to 17.7%, none) is a variant only.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 9; Figure 3D· Methods underspecifiedverifier disagreed at first

  L4 neurons outside C2 have fewer touch or phase coupled spikes than L4 C2, 33.0 +- 30.2% in 16.9% of exploration time.

  significantly more than outside of C2 (33.0 ± 30.2 in 16.9% of time; p = 2.88e-4)

  Results, L4 responses are temporally precise, paragraph 9; Figure 3D

  What the paper reports

  • touch or phase coupled fraction, L4 C2 against L4 outside C2: Wilcoxon rank sum, p = 2.9 × 10⁻⁴ (n = 31 and 10 cells)

  Depends on fig3d-touch-phase-l4c2. None of the 32 readings with the transition cutoff meets the criterion; the fraction is 42.4 to 65.1% (within tolerance) but the time fraction 28.0 to 42.8% against 16.9 +- 1.69%, with the rank-sum test against L4 C2 at p = 9.0e-5 to 0.0029. None of the literal variant meets it.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 9; Figure 3D· Methods underspecified

  L5 neurons have fewer touch or phase coupled spikes than L4 C2, 31.5 +- 16.4% in 26.1% of exploration time.

  or in L5 (31.5 ± 16.4 in 26.1% of time; p = 2.96e-5, Wilcoxon rank sum)

  Results, L4 responses are temporally precise, paragraph 9; Figure 3D

  What the paper reports

  • touch or phase coupled fraction, L4 C2 against L5: Wilcoxon rank sum, p = 3 × 10⁻⁵ (n = 31 and 11 cells)

  Depends on fig3d-touch-phase-l4c2. 12 of the 32 readings with the transition cutoff meet the criterion, 6 under each latency reading (all 4 with the 100 ms curve and the total-width phase window, 1 with 100 ms and +-t_w, 1 with 200 ms and the total width); over the 32 the fraction is 32.9 to 48.9% in 23.6 to 36.4% of time. None of the literal variant does.

• ⊘BlockedResults, L4 responses are temporally precise, paragraph 9; Figure 3D· Methods underspecified

  Spikes coupled to whisking phase alone are 4.6 +- 8.7% in L4 C2, 23.8 +- 27.9% in L4 outside C2 and 7.6 +- 14.9% in L5.

  Subtraction of the touch spikes from touch and phase spikes reveals that 4.6 ± 8.7% of spikes were purely phase coupled in L4 C2, vs 23.8 ± 27.9% outside of C2 and 7.6 ± 14.9% in L5

  Results, L4 responses are temporally precise, paragraph 9; Figure 3D

  Blocked as it depends on fig3d-touch-phase-l4c2. Its own criterion holds in all 16 readings with the transition cutoff under both latency readings (L4 C2 4.9 to 10.2%, outside 18.6 to 37.1%, L5 8.3 to 21.4%, outside the largest), and in all 16 of the literal variant. With the sample-epoch phase fit of the earlier run it held in 11 of 16.

• ⊘BlockedResults, Decoding touch, paragraph 1; Figure 4A· Methods underspecified

  Pooling the activity of only 15 L4 C2 neurons, integrated over 10 ms, is enough to detect 95% of touches.

  Pooling activity from only fifteen L4 neurons (out of approximately 1600 [ Lefort et al., 2009 ]) in C2 was sufficient to detect 95% of touches (integration time, 10 ms)

  Results, Decoding touch, paragraph 1; Figure 4A

  With 50 repetitions per population size, the population reaching 95% is 14 (fraction detected) or 12 (percent correct) neurons with exploration non-touch windows; 19, 25 and 25 with percent correct and the pre-touch windows (a1), (a2) and (a3); 19, 30 and 25 with the fraction detected and the same windows. Only 14 is within 15 +- 1.5, so the readings disagree. The fraction detected is now the hit rate at the time point of best percent correct, with ties between thresholds broken at the middle one; the earlier run took the highest hit rate over time points, which with ties broken at the lowest threshold counted every sample as a touch and gave 1 to 4 neurons. Breaking ties at the lowest threshold, a variant, changes only the (a2) fraction-detected value (25). One neuron gives 0.618 to 0.642 percent correct. On synthetic neurons whose touch and non-touch counts come from one distribution, one neuron gives a fraction detected of 0.329 at percent correct 0.541 under either tie rule, against 0.987 under the earlier run's rule.

  How the test differed from the paper’s method

  • Reading (1b) leaves out windows within 60 ms before or 100 ms after a touch onset, a standard choice.
  • Ties between thresholds of equal training percent correct are broken at the middle one, a choice the claims file does not fix; the lowest one is shown as a variant.
• ⊘BlockedResults, Decoding touch, paragraph 1; Figure 4A· Methods underspecifiedverifier disagreed at first

  Touch detection by neurons from surrounding barrels, outside C2, is poor.

  Touch detection by neurons from surrounding barrels (>200 µm from the principal whisker) was poor

  Results, Decoding touch, paragraph 1; Figure 4A

  Blocked as it depends on fig4a-touch-detection-c2. Its own criterion now holds under every reading. 15 outside neurons reach 0.522 to 0.541 percent correct (upper bound 0.557 to 0.583) and detect 0.196 to 0.351 of touches (upper bound 0.455 to 0.481), against 0.897 to 0.969 in C2, and no outside population up to 200 reaches C2's level at 15. The earlier run's fraction detected (0.91 to 0.94, upper bound 1.000) came from the threshold rule corrected in fig4a-touch-detection-c2.

• ⊘BlockedResults, Low spike count variability, paragraph 2; Figure 5B-C· Methods underspecifiedverifier disagreed at first

  Across cells, responsiveness to pre-touch velocity and to maximum curvature covary strongly (r = 0.80), and responsiveness to touch order is somewhat less correlated with both (r = 0.70 and 0.65).

  Responsiveness to pretouch velocity and max curvature covaried strongly (pairwise correlation coefficient 0.80, p = 4.3e-8), whereas touch order response was somewhat less correlated with velocity and curvature (correlation coefficients 0.70, 0.65, p = 1.0e-5, p = 8.2e-5)

  Results, Low spike count variability, paragraph 2; Figure 5B-C

  What the paper reports

  • velocity against curvature: correlation, r = 0.80, p = 4.3 × 10⁻⁸
  • touch order against velocity: correlation, r = 0.70, p = 10⁻⁵
  • touch order against curvature: correlation, r = 0.65, p = 8.2 × 10⁻⁵

  Under reading (a), with the transition cutoff, r is 0.618 to 0.788 (velocity, curvature), 0.706 to 0.788 (order, velocity) and 0.588 to 0.641 (order, curvature), and the criterion holds in 16 of 64 combinations of the fig5c readings; under reading (b) it never holds (r 0.27 to 0.51 for velocity and curvature). With the answer the literal cutoff (8 of 64 under (a)) is a variant only; the fig5c readings still give different outcomes.

• ⊘BlockedResults, Low spike count variability, paragraph 6; Figure 6D· Methods underspecified

  The low Fano factor after touch cannot be explained by a Poisson process with a 2.3 ms refractory period and rates matched to each neuron.

  Comparing the simulations with actual spike trains revealed that L4 neuron response precision could not be explained by the refractory period alone

  Results, Low spike count variability, paragraph 6; Figure 6D

  With the rate compensated as the answer says (the free rate of Berry and Meister), the model's mean count is 1.001 to 1.034 of the data's, and its FF at the data minimum 0.71 to 0.79. Under reading (2a) the data FF (0.75 to 0.82) is never below the model's by more than the sum of the s.e.m.; under (2b) it is below in 4 of 6 touch readings (61 of 89, 61 of 100, 59 of 99 units and 22 of 31 pooled cells counted from onsets, p = 0.0003 to 0.035) but not with groups of N points from offsets (52 of 91, p = 0.10) or pooled touches from offsets (17 of 31, p = 0.36). What the answer changed, the uncompensated model, now a variant, has a mean count 0.80 to 0.89 of the data's and a higher FF, and gives (2b) in all 6 readings; compensation lowers the model's FF and removes two of them. The outcome is blocked as before.

  How the test differed from the paper’s method

  • The PSTH used for the free rate starts 10 ms before onset so that the integral over the preceding 2.3 ms is defined at the first window.