CRCNS:ia-1

The resulting firing rates had a mean of 113 Hz (SD = 16 Hz), and they did not show any significant variation in the different stimulus conditions, as assessed by a one-way ANOVA ( p = 0.58)

Materials and Methods, Electrophysiology and stimulus design, paragraph 4

What the paper reports

• one-way ANOVA across stimulus conditions, p = 0.58

Over the 89 sessions of selection (c) the mean rate is 107.1 Hz (SD 14.5 Hz); one-way ANOVA across the seven protocols F = 0.49, p = 0.814. Over the 115 of selection (a) 110.2 Hz (SD 16.7), F = 0.49, p = 0.818; with whole-trial rates and over all seven-protocol sessions p is 0.471 to 0.809. No population is significant.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it. The outcome is the same under every population shown, so it does not rest on this assumption.

The probability of generating isolated spikes is minimal for large amplitude fluctuations and cutoff frequencies around 100 Hz.

Figure 5; Results, Stimulus characteristics modulate burst probability, paragraph 10

Under the main reading over selection (c) P(n = 1) is 0.816, 0.749, 0.837, 0.980 and 1.000 at 25, 100, 200, 400 and 800 Hz, lowest at 100 Hz, and 0.953, 0.837 and 0.668 at 3, 6 and 12 dB, lowest at 12 dB; the same with bursting sessions only, over selection (a) and with the plateau-excluded peak. P(n = 2) and P(n = 3) are highest at 100 Hz and 12 dB over all kept sessions. With minimum reading b the lowest P(n = 1) moves to 25 Hz.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Increasing the standard deviation of the amplitude distribution (from Figure 3 A to D to G) results in correlation functions that exhibit progressively sharper peaks ... Decreasing the typical time scale of the stimulus fluctuations (going right from Figure 3 B to F) leads from multi-modal (Figure 3 B) to single-peaked (Figures 3 C,D) to increasingly shallower and broader correlation functions (Figures 3 E,F).

Figure 3; Results, Stimulus characteristics modulate burst probability, paragraph 3

The cell 03-04-23-ad to -aj of singlecells.txt was recorded with all seven protocols (six reference sessions). The first peak of C_s/rbar^2 above its baseline is 0.50 at 3 dB, 1.14 to 2.35 in the six 6 dB sessions and 5.06 at 12 dB (reading a, sharper with SD); its width does not order (reading b). C_s has 4 to 14 peaks at 25 Hz, one at 200 Hz, and one at 100 Hz at prominence 0.2 (two at 0.05 and 0.1); 3 dB, 400 and 800 Hz have no separable first peak. The model panels of Figure 3 are not tested (no thresholds).

How the test differed from the paper’s method

• The cell is identified from singlecells.txt as the only one with all seven protocols; the paper does not name it.
• Peaks are counted on C_s/rbar^2 smoothed over 1 ms as local maxima with a prominence above a threshold the paper does not give; 0.05, 0.1 and 0.2 are shown and the outcome rests on 0.2 for the single peak at 100 Hz.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

The population mean of the temporal average of this estimated error was 3.4% (SD 1.5%) of the total span of C s (τ) (that is, the difference between the maximum and the minimum)

Materials and Methods, Burst identification, paragraph 6

With sigma the SD of the product terms and N their number (reading a), over the whole lag range (reading b) and without tau = 0, the error averages 2.63% (SD 1.39%) of the span of C_s over the 89 sessions of selection (c) and 2.69% over the 115 of selection (a); the third reading of sigma and N gives 2.61% and 2.67%. With tau = 0 in the span all readings give 0.06 to 0.16% or, with N the number of trials, 6.3 to 7.9%; over 0 to 1.25 / cutoff without tau = 0 the mean is unstable (30 to 4500%).

How the test differed from the paper’s method

• The span of C_s leaves out tau = 0, where C_s holds the self term (two orders of magnitude above the rest); the paper does not say. Assumed because the claims file's own detail (1) excludes the zero-lag self term from the peak.
• A third reading of sigma and N (sigma the SD over trials of each trial's C_s, N the number of trials), from paragraph 6 ("C_s can be interpreted as an average"), is shown beside the claims file's two.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.

For the cell shown in Figure 6 , the mean value of c s 2 ( t , τ ) averaged over all t ∈ [200, 990 msec] and τ ∈ [0, 10 msec] is 2.94 times larger than the corresponding mean of c b 2 ( t , τ ) . The population average of this ratio on all bursting sessions is 2.89 (SD 1.49).

Results, Quantitative description of the information transmitted by bursts, paragraph 2; Figure 6

Over the 64 bursting sessions of selection (c) the mean square Pearson coefficient is larger for spikes than for bursts in every session. Without tau = 0 the mean ratio is 1.99 (SD 0.92) with 0.1 ms bins and 2.17 (SD 1.86) with 0.4 ms bins, within 2.89 +/- 1.49; with tau = 0, 1.29 and 1.21, outside. Over the 84 of selection (a), 1.99 and 2.13. All bursting sessions have bursts of n > 1.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

At the population level, in 80% of the bursting sessions the amount of jitter was roughly independent from n (the best linear fit had a slope that was not significantly different from 0) ... The mean slope was −0.03 msec (SD 0.2 msec).

Results, Quantitative description of the information transmitted by bursts, paragraph 6; Figure 7E and 7F

Over the bursting sessions of selection (c), events read as onsets within the first-peak domain around an event centre (reading b) give a mean slope of 0.005 ms (SD 0.052) and 86.9% of the 61 sessions with a test not significant by the two-sided t-test, within 80 +/- 8. Events between histogram troughs (reading a) give -0.001 ms and 88.9% (just outside); with significance read as one standard error 32.8% and 44.4%. Over selection (a), 87.7% and 91.6%. Median jitter of 1-bursts 0.389 ms (b) and 0.058 ms (a). Three sessions have fewer than three n values with a jitter and no test.

How the test differed from the paper’s method

• How events are formed is open (Rokem et al., 2006 is cited for the definition). Reading b is implemented with event centres taken as the 0.1 ms onset-histogram bins from the highest down and a window of the limiting ISI centred on each; reading a fragments events (median jitter 0.058 ms, below the paper's 0.2 ms error bar).
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Moreover, the mean delay between stimulus upstroke and burst generation decreases systematically with n . This implies that stimulus deflections that are either high or wide tend to produce prompt responses, with high- n bursts.

Results, Qualitative description of the information transmitted by bursts, paragraph 3; Figure 8D; Discussion, Implications for the neural code, paragraph 4

With the central maximum searched before the onset, mean tau_n is 5.80, 4.23, 3.32 and 2.67 ms for n = 1 to 4 over the 53 sessions of selection (c) (reading b). No tau_n is negative, by construction, as the search ends at 0 ms; in 13 of the 53 sessions 17 values sit on that edge (1, 5 and 11 for n = 2, 3, 4), and the mean tau_4 includes them. Without the values on the edge the means are 5.80, 4.32, 3.67 and 3.37 ms; over the 40 sessions with none on the edge 5.77, 4.33, 3.77 and 3.40 ms; of the 48 sessions with tau_1 to tau_3 off the edge, 32 have tau_1 > tau_2 > tau_3. 51 of 53 have tau_4 < tau_1 (Wilcoxon p = 4.5e-9) but only 20 of 53 decrease at every step (reading a, not a majority). Searched over the whole window, as in the first round, 5.80, 4.22, 3.14 and -0.50 ms, with a negative tau_n in 33 sessions. With at least 20 or 50 bursts of each n the means fall from 6.01 to 2.12 and from 5.88 to 2.05 ms; over selection (a) from 5.75 to 2.46 ms.

How the test differed from the paper’s method

• The central maximum of an nBTA, and so tau_n, is searched from -25 to 0 ms, before the burst onset, as the Results call tau_n the delay between stimulus upstroke and burst generation (Qualitative description, paragraph 3); the first round searched -25 to +15 ms, which is shown as a variant.
• Sessions need at least two bursts of each n = 1 to 4, the least that gives a variance; the paper says only that eight sessions had too few bursts (Results, Qualitative description, paragraph 6). Assumed; the results with at least 20 and 50 bursts of each n are shown.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

For 98% of the bursting cells and for all n -values, σ n (τ) was smaller than the standard deviation σ(τ) of the stimuli preceding all spikes (prior to any classification). The population average of the ratio of the minimum value of σ(τ) to the n -average of the minimum values of σ(τ) was 1.62 (SD 0.56).

Results, Qualitative description of the information transmitted by bursts, paragraph 5

In 98.1% of the 53 sessions of selection (c) the minimum of sigma_n is below that of sigma for all n = 1 to 4 (reading a); the ratio of minima averages 1.596 (SD 0.280). At every tau (reading b) 1.9%. A size-matched control (sigma from random subsets of the spike-triggered segments of each n's size, 20 draws) has a minimum about 5% below that of all segments; min sigma_n is below the size-matched minimum for every n in 94.3% of sessions, with a ratio of 1.517, so the result is not an effect of sample size. With at least 20 or 50 bursts of each n, 97.4% and 1.562, 96.4% and 1.571. Ratios per n 1.27, 1.45, 1.62 and, for n = 4, 23.5, inflated by sessions with few 4-bursts (2.38 with at least 20).

How the test differed from the paper’s method

• Sessions need at least two bursts of each n = 1 to 4, the least that gives a variance; the paper says only that eight sessions had too few bursts (Results, Qualitative description, paragraph 6). Assumed; the results with at least 20 and 50 bursts of each n are shown.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.
• A control not in the paper: sigma from random subsets of the spike-triggered segments of the size of each n's, 20 draws per n.

A large fraction of the cells show a significant difference, for both real and simulated data. Hence, also at the population level, the n BTAs differ significantly from the convolved 1BTAs. As n increases, the number of sessions with significant differences diminishes.

Results, Qualitative description of the information transmitted by bursts, paragraphs 7 and 8; Figure 9A and 9B

Over the bursting sessions of selection (c) the ANOVA finds the nBTA different from the summed 1BTAs in 100% of sessions for n = 2, 3 and 4 paired (64, 63 and 53 sessions), and 100, 100 and 98.1% grouped, so the fraction does not grow; the share with over 70% of the interval significant falls from 93.8% to 84.1% and 52.8%. With the STA the ANOVA fractions are 100, 100 and 98.1% and the 70% share falls from 100% to 37.7%. The average of the shifted copies and selection (a) give the same picture. The ANOVA, which treats every segment and frequency bin as independent, rejects in nearly every session; the criterion specifies it. The model sessions are not tested.

How the test differed from the paper’s method

• The central maximum of an nBTA, and so tau_n, is searched from -25 to 0 ms, before the burst onset, as the Results call tau_n the delay between stimulus upstroke and burst generation (Qualitative description, paragraph 3); the first round searched -25 to +15 ms, which is shown as a variant.
• Eq. 15 sums 1BTAs of the stimulus in dB; the sum is taken after subtracting the session's mean stimulus, as the paper does for Figure 9C, since a sum of n absolute levels cannot compare with an nBTA. The average of the shifted copies is shown as a variant.
• The paired ANOVA (reading a) tests the mean difference between each burst's segment and its own f_n and its variation across frequency bins.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Figure 9 C demonstrates that also at the population level, high- n bursts are associated with either higher or wider stimulus deflections (or both).

Results, Qualitative description of the information transmitted by bursts, paragraph 9; Figure 9C

Selection (c) holds 7 bursting protocol-5 sessions with bursts of n = 1 to 3, from 7 distinct cells (rows of singlecells.txt; 03-04-23-aa and 03-04-23-ah are different cells), the paper's seven cells. Their population-average central peak is 3.13, 7.03 and 8.46 dB above the mean stimulus for n = 1, 2, 3, with half-height widths 5.00, 5.50 and 7.35 ms (12.55, 13.60, 17.80 ms between minima) and SD across sessions 0.54, 0.83 and 1.51 dB at the peak. Both searches for the maximum give these values; the 10 sessions of selection (a) give 3.22, 7.11 and 8.60 dB.

How the test differed from the paper’s method

• The central maximum of an nBTA, and so tau_n, is searched from -25 to 0 ms, before the burst onset, as the Results call tau_n the delay between stimulus upstroke and burst generation (Qualitative description, paragraph 3); the first round searched -25 to +15 ms, which is shown as a variant.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• 7 protocol-5 sessions of selection (c), as the paper has seven cells; 10 over selection (a).
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

More specifically, a 50 Hz increase in firing rate shifts the limiting ISI by <0.4 msec, which is comparable to its estimated error bar. The average intra-burst spike count n , in turn, shows an increase of <25%.

Results, Stimulus characteristics modulate burst probability, paragraph 9

No cell was recorded at two target rates (headers hold 100 and 90 Hz), so reading a cannot be formed. Under reading b over selection (c) a 50 Hz higher rate goes with a limiting ISI shorter by 0.337 ms (p = 0.44, 64 sessions) and an average n higher by 15.7% (p = 0.132, 89 sessions); within protocol 1, 0.057 ms and 18.9%. Over selection (a), 0.189 ms and 13.6%. These are differences between cells.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Thirty-seven receptor cells were recorded, from 23 animals. Each cell was tested with two or more stimuli, resulting in 132 data sets in total ... There were 86 out of 132 data sets that fulfilled these two conditions ... N repetitions of each stimulus were presented, with N ranging between 98 and 503 (average 172)

Materials and Methods, Electrophysiology and stimulus design, paragraphs 1, 4 and 5

The 116 sessions of the seven protocols come from 37 cells (rows of singlecells.txt) recorded on 23 dates; the 122 Gaussian AM sessions of any cutoff are within 10% of the paper's 132 (all 150 Data1 sessions are not). Under the criterion's readings the SD < 35 Hz condition keeps 115 of the 116 (113 with a 100 Hz target, reading b; 116 with whole-trial rates), against 86; over the 115 trials range from 69 (outside 98 +/- 9.8) to 503 and average 160.3, and the mean trial-to-trial SD of the rate is 6.48 Hz (within 6 +/- 0.6). Selection (c), which also leaves out the 9 sessions under 98 trials and the off-line sorted sessions, keeps 89 sessions with 98 to 503 trials (mean 168.7) and a mean SD of 6.00 Hz, within tolerance, but they come from 29 cells on 18 dates, outside 37 +/- 3.7 and 23 +/- 2.3 (the 37 cells and 23 dates hold over all seven-protocol sessions); it is not a reading of the criterion, which defines the selection by the rate conditions, and its fewest trials are 98 by construction, so it does not decide this outcome. The trial blocks were checked: late spikes (up to 2.76 s) occur only in off-line sorted sessions, in the pause, no block restarts its times or holds twice the usual spike count, and the data set's firing_rates file gives the same trial counts in 136 of 138 files.

How the test differed from the paper’s method

• Data1 does not record the animal; the 23 recording dates stand in for the 23 animals, assuming one animal per date.
• Sessions recorded are read as the 122 Gaussian AM sessions; the paper's 132 is not otherwise identifiable.

All sessions had significantly positive slopes. Figure 7 B shows the distribution of slopes throughout the population. The average slope across the 59 bursting cells was 1.5 bits (SD 0.7 bits).

Results, Quantitative description of the information transmitted by bursts, paragraph 5; Figure 7A and 7B

Over the 64 bursting sessions of selection (c), within 59 +/- 5.9, the mean slope of I_n(1) on n is 2.22 bits (SD 0.87) with 0.1 ms bins, just outside 1.5 +/- 0.7, and 1.91 (SD 0.75) with 0.4 ms bins in bits per event (reading i), 1.78 and 1.53 as Eq. 10 prints it (reading ii), within. All 64 slopes are positive, 63 significant by one standard error and 62 by the t-test, so not every session. Over the 84 of selection (a): 2.14, 1.85, 1.71 and 1.48, 83 and 82 significant. A baseline with each trial's burst onsets shifted at random gives a mean slope of 2.59 bits (0.1 ms bins) and 2.23 (0.4 ms), above the data's; the data's slope exceeds its shifted copy in 3 of 64 sessions (9 with 0.4 ms bins). A shifted train has lost its stimulus locking, so its plug-in bias is at least that of the data; the comparison shows that the estimator's bias for rare events is of the order of the rise of I_n(1) with n, not that the whole rise is bias. The criterion fixes the formula, so the outcome rests on it.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.
• A baseline not in the paper: I_n(1) and its slope recomputed with each trial's burst onsets shifted by a random offset, wrapped within 200 to 1000 ms, ten shifts per session.

This result was also observed in all other recorded cells ... the frequency of occurrence always decreased significantly with n . The population data in Figure 7 D had an average slope of −2.6, with SD of 0.7.

Results, Quantitative description of the information transmitted by bursts, paragraph 6; Figure 7C and 7D

Over the 64 bursting sessions of selection (c) the mean slope of log2 occurrence on n is -2.82 (SD 1.12), within -2.6 +/- 0.7 (natural log -1.96, log10 -0.85). All 64 slopes are negative; 63 are significant by one standard error and 61 by the t-test, so not every session, but above 90%. Over selection (a): -2.71, 83 and 81 of 84.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

The information measures obtained by our method and by the direct method were highly correlated ( R = 0.95, using all sessions). The population average obtained with the direct method is 222 ± 69 bits/s. With our method, instead, this average was 191 ± 72 bits/s. In all cases but one, the information obtained with the direct method was higher than the one obtained with our method, the average difference being 31 ± 16 bits/s.

Results, Quantitative description of the information transmitted by bursts, paragraph 10; Discussion, Comparison with other neural codes, paragraph 3

With the burst code in 0.4 ms bins over the 89 kept sessions of selection (c), R = 0.967, direct (8 bins) 198.4 bits/s, burst code 178.1, mean difference 20.3 (SD 21.2), all within tolerance; 10.2% lost, outside 14 +/- 1.4; the direct method is higher in 74 of 89 (83%), not all but one. Over the bursting sessions R = 0.951, 220.7 against 199.8, difference 21.0, higher in 52 of 64. Over selection (a) R = 0.965, 204.3 against 183.6, 10.1% lost, higher in 94 of 115. Extrapolated direct rates differ by under 1 bit/s. With 0.1 ms bins the burst code exceeds the direct method on average (220.1 against 198.4).

How the test differed from the paper’s method

• NSB entropies computed by this entry's implementation (ia1.nsb_entropy), not the authors' code, which is not public.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

All recorded and simulated bursting cells exhibited significantly different n -BTAs, for n ranging between 1 and 4 ... out of the 58 sessions where bursting cells were found, 50 exhibit n BTAs whose central peaks were significantly different from one another

Results, Qualitative description of the information transmitted by bursts, paragraphs 3 and 6; Figure 8D

Of the 64 bursting sessions of selection (c) (just outside 58 +/- 5.8), 53 have at least two bursts of each n = 1 to 4 and all 53 pass the ANOVA, with all frequency bins and with bins up to the cutoff. With the central maximum searched before the onset, 38 have marked points in every central peak (reading a), outside 50 +/- 5, and 5 have most of every central peak marked (reading b); over the whole window 42 and 7. With at least 20 bursts of each n 33 of 39, with at least 50, 25 of 28. Over selection (a), 54 (60 over the whole window) of 71. The ANOVA part holds; the time-domain count does not.

How the test differed from the paper’s method

• The central maximum of an nBTA, and so tau_n, is searched from -25 to 0 ms, before the burst onset, as the Results call tau_n the delay between stimulus upstroke and burst generation (Qualitative description, paragraph 3); the first round searched -25 to +15 ms, which is shown as a variant.
• Sessions need at least two bursts of each n = 1 to 4, the least that gives a variance; the paper says only that eight sessions had too few bursts (Results, Qualitative description, paragraph 6). Assumed; the results with at least 20 and 50 bursts of each n are shown.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Not quoted: The numbers are read from Table 1; no sentence states them. The legend calls the row recorded sessions; the counts sum to 86, so they are read as the kept sessions.

Kept sessions per protocol are 60, 9, 8, 10, 11, 8 and 9 (reading a, 200 to 1000 ms) against 43, 6, 7, 7, 8, 7 and 8; none within 10% under any reading of the criterion. Selection (c), which also leaves out sessions under 98 trials and off-line sorted sessions, gives 47, 6, 7, 7, 8, 7 and 7, six of seven within tolerance, but was chosen in part from these counts. Leaving out only the off-line sessions gives 3 of 7, only the sessions under 98 trials 1 of 7, and the six 450 ms pause sessions 2 of 7. The data set holds 55 experiments pairing the reference with each other protocol (8 to 11 per protocol).

The fraction of bursting sessions, the percentage of isolated spikes (1-bursts), and the maximum n -value depend strongly on the standard deviation and cutoff frequency of the stimulus

Table 1, row Sessions with n > 1; Results, Stimulus characteristics modulate burst probability, paragraph 8

Under the main burst reading the sessions of selection (c) with bursts of n > 1 are 40/47, 1/6, 7/7, 6/7, 7/8, 3/7 and 0/7 against 40/43, 0/6, 7/7, 6/7, 7/8, 3/7 and 0/8: six of seven counts equal the paper's, and one 3 dB session bursts where the paper has none. The plateau-excluded first maximum (peak c) gives 37/47 and the same other counts. Over the 115 sessions of selection (a) at most 2 of 7 are within tolerance under any combination (52/60, 2/9, 8/8, 9/10, 10/11, 3/8, 0/9 under the main reading), the first round's result. Peak reading a gives 5 of 47 bursting reference sessions, and N the number of trials none.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it. For this claim selection (c) was chosen in part from the expected counts, so it is a variant and the outcome follows the selections the criterion supports, (a) and (b), under which at most 2 of 7 counts are within tolerance under any burst reading.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Notice, however, that in all cases, isolated spikes are more frequent than any other burst of n > 1.

Table 1, rows % Isolated spikes and Most probable n; Results, Stimulus characteristics modulate burst probability, paragraph 8

Under the main reading over selection (c) the pooled percentages of isolated spikes are 80.7, 71.7, 66.1, 78.0, 71.3, 95.0 and 100%; 4 of 7 within tolerance (protocols 1, 3, 6, 7), at most 5 under any combination and 4 over selection (a). n = 1 is the most probable n in every protocol under the main reading. The criterion requires every value and gives no partial level; 4 of 7 is below the five of seven at which the other Table 1 criteria call a result partial, so the first round's partially_reproduced is changed to not_reproduced, as the verifier gave it.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it. The outcome is the same under every population shown, so it does not rest on this assumption.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Not quoted: The values are read from Table 1; the sentence of paragraph 8 states the dependence on the stimulus but not the values.

Under the main reading over selection (c) highest n 7, 9, 8, 14, 7, 5, 1, average n 1.23, 1.39, 1.49, 1.90, 1.39, 1.05, 1.00, SD of n 0.51 (p1), 0.82 (p3), 2.22 (p4), 0.71 (p5), 0.23 (p6); 8 of 19 within tolerance, at most 9 under any combination of readings and either selection (7 over selection (a) under the main reading). The longest bursts are at 25 Hz and 800 Hz has none, as in the paper; 3 dB has bursts of up to 9 spikes in its one bursting session.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it. The outcome is the same under every population shown, so it does not rest on this assumption.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

I ′ n strongly decreases with n (Figure 7 G). Similar results were obtained in all other recorded sessions (Figure 7 H), with an average slope of −11 bits/s (SD 16 bits/s) ... The population average of this fraction among all bursting cells was 47%.

Results, Quantitative description of the information transmitted by bursts, paragraph 7; Figure 7G and 7H

The information rate falls with n in all 64 bursting sessions of selection (c); mean slope -46.5 bits/s (SD 39.0) with 0.1 ms bins and -38.1 (SD 33.1) with 0.4 ms bins, outside -11 +/- 16; bursts with n > 1 carry 35.5% and 37.7%, outside 47 +/- 4.7. Total rates average 241.4 and 199.8 bits/s. With each trial's burst onsets shifted at random the same formula gives 106.5 and 37.5 bits/s, slopes of -16.1 and -3.9 bits/s and fractions of 46.8% and 63.2%, so much of these rates is estimator bias. Over selection (a): -44.7 and -36.6 bits/s, 36.3% and 38.6%.

How the test differed from the paper’s method

• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.
• A baseline not in the paper: I_n(1) and its slope recomputed with each trial's burst onsets shifted by a random offset, wrapped within 200 to 1000 ms, ten shifts per session.

The constancy of the location of the minima also holds at the population level ... Its average among all bursting cells is 0.4 msec (SD 0.7 msec) considering 1 ≤ n ≤ 4.

Results, Qualitative description of the information transmitted by bursts, paragraph 4; Figure 8E

The minimum of sigma_n lies near -6 ms for every n (medians -6.85, -6.3, -4.8, -6.6 ms) but over the 53 sessions of selection (c) the SD of its position across n = 1 to 4 averages 2.37 ms (SD 2.44) in window a and 2.84 ms in window b (2.05 and 2.46 ms with the population SD), outside 0.4 +/- 0.7; times the cutoff 0.35 to 0.48, against 0.05. With at least 20 bursts of each n 2.14 ms (39 sessions), with at least 50, 1.74 ms (28 sessions; 0.19 times the cutoff), and 1.46 ms over the 38 such sessions of selection (a).

How the test differed from the paper’s method

• Sessions need at least two bursts of each n = 1 to 4, the least that gives a variance; the paper says only that eight sessions had too few bursts (Results, Qualitative description, paragraph 6). Assumed; the results with at least 20 and 50 bursts of each n are shown.
• The SD across n is the sample SD (N - 1, MATLAB's default); the population SD is shown too.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

Whereas fairly low excursions produce either no response (dotted line) or an isolated spike (black, solid line), large deflections are associated with doublets (dark grey), triplets (grey), or bursts with four spikes (light grey line). All cells showing a bursting behavior exhibited this phenomenon.

Results, Qualitative description of the information transmitted by bursts, paragraph 10; Figure 10

Over the 64 bursting sessions of selection (c), with tau_n from the maximum before the onset, the h at which P(n | h) peaks increases from n = 0 to the largest n (up to 4) in 7 sessions (10.9%) and the P-weighted mean h in 15 (23.4%), under both widths T (median 7.40 and 3.05 ms), against all bursting cells; tau_n over the whole window gives 7 and 16, selection (a) 8 and 19 to 21 of 84. Restricted to n = 0 to 3 (outside the criterion) 34 and 36 to 37 of 64 hold. The two T readings match different numbers of maxima in 58 of 64 sessions (median 8316 against 8173), but the peak flag changes in none and the mean flag in 2, one each way, which is why the counts are equal.

How the test differed from the paper’s method

• The central maximum of an nBTA, and so tau_n, is searched from -25 to 0 ms, before the burst onset, as the Results call tau_n the delay between stimulus upstroke and burst generation (Qualitative description, paragraph 3); the first round searched -25 to +15 ms, which is shown as a variant.
• A stimulus maximum whose closest burst was assigned to a nearer maximum counts as no response; the Methods do not say whether it is matched to another burst.
• The kept sessions are the 89 of the 116 seven-protocol sessions with a trial-to-trial rate SD below 35 Hz, at least 98 trials and spikes detected on-line (selection c), not the paper's 86, whose selection is not stated. Assumed because the Methods give 98 to 503 trials for the paper's sessions, and because with the off-line sorted sessions also left out the kept sessions per protocol (47, 6, 7, 7, 8, 7, 7) and the bursting sessions per protocol come close to Table 1; the off-line exclusion has no passage in the paper behind it. The 115 sessions of selection (a), the first round's population, are shown beside it.
• Burst identification takes the first peak as the largest value of C_s below 5 ms (peak reading b), the first local minimum after it that differs significantly from it (minimum reading a), and the error bar with sigma the SD of the product terms of Eq. 1 and N their number (reading a). Assumed because the first local maximum (reading a) falls on edge effects of the refractory plateau, where C_s has no spike pairs, which the Methods say the criteria reject (Burst identification, paragraph 5), and leaves 11 of 89 sessions bursting; started after the plateau (peak reading c, added after the first verification) it leaves 61, near the 64 of the main reading; with N the number of trials no session bursts; the main reading gives an error bar of 2.6% of the span of C_s, near the paper's 3.4% (paragraph 6). The notebook shows every combination.

The mean threshold across the population was 58 dB (SD 14 dB)

Materials and Methods, Electrophysiology and stimulus design, paragraph 2

Data1 holds spike times, cell.dat (stimulus names, played intensities, cutoffs, same-cell links) and the stimulus files; no threshold, best-frequency or tuning measurement. The data descriptor says the best frequency was determined from f-I curves but does not deposit them for Data1.