CRCNS:fcx-1

We found that the duration of packets was distributed over two orders of magnitude, with a median of 111 seconds

Results, Identification of brain states; Table S1

Median of the 701 packets in SLEEP 111.0 s. Longest over shortest 52.5 (14 to 735 s), less than the two orders of magnitude of the statement, which the criterion only reports.

Whereas nonREM and WAKE firing rates remained highly correlated on a per-cell basis (R = 0.82, p < 1 −10 ), the slope of this relationship significantly deviated from 1 ( Figure 2B ), demonstrating that state transitions exert a differential effect across the firing rate spectrum (95% confidence interval of slope ... 0.66 - 0.71)

Results, Assessment of firing patterns in different brain states; Figure 2B and legend

What the paper reports

• correlation of nonREM and WAKE rates: correlation, R = 0.82, p < 10⁻¹⁰
• slope of nonREM rate on WAKE rate: 95% CI 0.66 to 0.71

Reading (a) slope of log10 nonREM on log10 WAKE 0.684 (0.661 to 0.708), R 0.803; reading (b) 0.687 (0.658 to 0.717), R 0.824. Linear rates give slope 0.45. 82% of cells above the median WAKE rate fire less in nonREM, 55% of those below fire more.

The right tail extended toward higher rates during WAKE and REM compared to nonREM ( Figure 2A ; fraction of >2 Hz neurons

Results, Assessment of firing patterns in different brain states; Figure 2A

What the paper reports

• Chi-Square Test, p < 10⁻¹⁰

Reading (a) 18.8% WAKE, 20.6% REM, 9.3% nonREM; chi-square nonREM against WAKE p = 4.1e-16, against REM p = 4.4e-21. Reading (b) fractions 18.4, 19.5, 9.1% with p = 3.2e-09 and 8.3e-11.

The fraction of principal cells firing in 50 ms overlapping time windows ('population synchrony') showed a skewed distribution. More than 15% of principal neurons rarely fired synchronously with others (4% of windows), while less than 4.2% discharged synchronously in at least half of the windows

Results, Assessment of firing patterns in different brain states; Figure S2F

Reading (a) 62.9% of pE cells fire with another pE cell in at most 4% of windows and 0.00% in at least half; reading (b) 54.2% and 0.00%. Per-cell distributions skewed (2.31, 2.26); population synchrony per window skew 1.28. The criterion's bounds hold with wide margins.

The durations of DOWN states of 'slow oscillations' showed a skewed distribution (median = 0.19s) and so did the UP states (median = 0.47s; Figure S4A )

Results, Assessment of firing patterns in different brain states; Figure S4A; Table S1

Deposited detections, 43230 UP and 86839 DOWN states (the paper's counts exactly), medians 0.473 and 0.188 s, skew 2.56 and 2.15. All start inside scored packets of the cycles.

How the test differed from the paper’s method

• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.

pyramidal neurons fired in reliable spatiotemporal sequences during the early part of the UP state (<200 msec) and these stereotypical sequences remained stable across the entirety of sleep ( Figure 2D ). Additionally, the spike onset times of individual neurons in the sequences were negatively correlated with their firing rates

Results, Assessment of firing patterns in different brain states; Figure 2D and 2E

First and second half latencies correlate across cells, R = 0.733 (halves of each SLEEP, n 932) and 0.660 (halves of the session, n 758). Latency against log10 WAKE rate R = -0.127, p = 7.6e-05; against log10 nonREM rate R = -0.238, p = 6e-14. Per session the WAKE-rate correlation is negative in 20 of 27 sessions.

How the test differed from the paper’s method

• A cell's mean latency in a half is computed when it has at least 5 UP states with a spike within 200 ms (a standard choice).

we found that the arithmetic mean of the population firing rates of pE cells declined over the course of sleep using a test of correlation of spike rate versus time ( Figure 3A-B ; R = - 0.11, p < 10 −3 )

Results, Firing pattern and LFP changes across the course of sleep; Figure 3A and 3B

What the paper reports

• correlation of spike rate versus time, R = -0.11, p < 0.001 (n = 995 cells, 54 sessions, 11 rats)

Normalized mean pE rate against normalized SLEEP time, clock (i), R = -0.273 (a, p = 1.7e-13), -0.154 (b, p = 2.5e-57), -0.732 (c, p = 1.5e-09); clock (ii) R = -0.283, -0.163, -0.796. By rat (reading b) the decline is significant in 8 of 11 rats; Rizzo, Splinter and Templeton show R = -0.017, -0.049, -0.014.

How the test differed from the paper’s method

• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).

Comparison of the population discharge rates in the first and last nonREM packets of SLEEP episodes showed an overall mean rate decrease (1.04±1.20 Hz versus 0.88±0.91 Hz; p < 0.02; Wilcoxon test)

Results, Firing pattern and LFP changes across the course of sleep, sentence citing Vyazovskiy et al., 2009

What the paper reports

• Wilcoxon test, p < 0.02

Reading (a) first 1.027 +- 1.086, last 0.905 +- 0.936 Hz, signed-rank p = 1e-18; reading (b) 1.048 and 0.883 Hz, signed-rank p = 1.9e-20, rank-sum p = 0.0012. All 51 SLEEP periods have at least 2 packets.

we assessed spike rates of the same neurons in the first versus the last nonREM packets of SLEEP and found the slope of this correlation significantly departed from unity (slope 95% confidence interval 0.827-0.880)

Results, Firing pattern and LFP changes across the course of sleep; Figure 3D

What the paper reports

• 95% CI 0.827 to 0.88

Reading (b) slope 0.872 (0.840 to 0.904), within 0.8535 +- 0.0265; reading (a) 0.788 (0.760 to 0.817). Zero-rate cells dropped, 23 (b) and 70 (a). 72% of cells above the median first-packet rate decrease; 49% of those below increase.

The mean spike discharge rate of highest-firing rate sextile groups significantly decreased over sleep whereas the activity of the lowest-firing rate sextile showed a marginal increase ( Figure 3B ; Pearson correlations versus normalized time)

Results, Firing pattern and LFP changes across the course of sleep; Figure 3B, bottom

Top sextile R = -0.333, -0.190, -0.716 (readings a, b, c; p <= 5e-09), bottom sextile R = +0.092, +0.046, +0.311 (p = 0.022, 8.5e-06, 0.028); clock (ii) agrees (top R = -0.344, -0.203, -0.799; bottom +0.090, +0.044, +0.360). M2 reading (b) agrees (top R = -0.291, -0.200, -0.756; bottom +0.167, +0.090, +0.550, p <= 3.6e-05).

How the test differed from the paper’s method

• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

Comparison of firing rates within UP states showed a general decrease across the whole population ( Figure S3C ) and subtraction of UP state firing activity in early sleep from that in late sleep revealed that the firing rates of the fastest sextile group and putative interneurons decreased later in sleep, whereas firing of the slowest sextile increased

Results, Firing pattern and LFP changes across the course of sleep; Figure 3F; Figure S3C

What the paper reports

• within-UP pE rate against normalized SLEEP time (Figure S3C): R = -0.19849, p = 3.3 × 10⁻⁹

Within-UP pE rate against SLEEP time, reading (b) R = -0.087, p = 9.4e-16 (meets p < 1e-8); (a) R = -0.159, p = 5.5e-05; (c) R = -0.568, p = 1.7e-05. Last minus first packet, M2 reading (a), fastest sextile -0.637 Hz (p = 1.2e-07, n 159), pI -0.572 Hz (p = 0.045, n 118), slowest +0.071 Hz (p = 0.27, n 151), which would be partial. M2 reading (b), one entry per cell and cycle, fastest -0.425 Hz (p = 3.7e-07, n 276), slowest +0.123 Hz (p = 0.047, n 262); every part holds under one reading of the claims file. On one unit throughout, M2 reading (b) gives all three parts with one entry per cell and cycle (M1 reading a, fastest -0.425 Hz, p = 3.7e-07; pI -0.672 Hz, p = 0.047, n 182; slowest +0.123 Hz, p = 0.047) and with one entry per cell (M1 reading b, fastest -0.573 Hz, p = 3.3e-07, n 181; pI -0.572 Hz, p = 0.045; slowest +0.141 Hz, p = 0.023, n 189). Under M2 reading (a) the slowest sextile is not significant with either unit (p = 0.48 and 0.27). The pI and slowest-sextile tests lie between 0.023 and 0.047.

How the test differed from the paper’s method

• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).
• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

The incidence of both UP states (R = 0.05, p < 10 −3 ) and DOWN states (R = 0.08; p < 10 −10 ) increased significantly within packets, whereas the duration of UP states decreased (R = - 0.02; p < 0.01) and DOWN states increased (R = 0.07; p < 10 −10 ), changing the UP-DOWN balance (UP/DOWN ratio negative correlation versus time p < 0.05)

Results, Firing rate and LFP changes within nonREM episodes and packets; Figure 4A-B ii

What the paper reports

• UP incidence: R = 0.05, p < 0.001
• DOWN incidence: R = 0.08, p < 10⁻¹⁰
• UP duration: R = -0.02, p < 0.01
• DOWN duration: R = 0.07, p < 10⁻¹⁰
• UP/DOWN ratio: p < 0.05

Reading (b), UP incidence R = 0.042 (p = 1.3e-05), DOWN incidence R = 0.168 (p = 3.5e-68), UP duration R = -0.101 (p = 9.2e-21), DOWN duration R = 0.190 (p = 3.6e-83), UP/DOWN ratio R = -0.156 (durations) and -0.035 (time), p = 8.1e-48 and 0.00051. Reading (c) gives at most 1 of 5. Fraction in neither state R = -0.176 (b), the sign the paper prints. With the Dino ACC sessions left out, the four incidence and duration trends keep their signs (reading b, p <= 1.5e-05).

How the test differed from the paper’s method

• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).
• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.

The incidence of spindles significantly increased both within nonREM episodes ( Figure 4A-B iii ; R = 0.08; p < 10 −10 ) and packets (R = 0.7; p < 10 −10 ). Spindle duration also increased (nonREM episode increase from 0.61 ± 0.33 s to 0.67 ± 0.30 s, R = 0.08; p < 10 −3 ; nonREM packet increase from 0.58 ± 0.25 s to 0.67 ± 0.27 s, R = 0.07, p < 10 −4 ), whereas spindle frequency and amplitude remained unchanged

Results, Firing rate and LFP changes within nonREM episodes and packets; Figure 4A-B iii

What the paper reports

• spindle incidence within episodes: R = 0.08, p < 10⁻¹⁰
• spindle incidence within packets: R = 0.7, p < 10⁻¹⁰
• spindle duration within episodes: R = 0.08, p < 0.001
• spindle duration within packets: R = 0.07, p < 10⁻⁴

Reading (b), incidence within episodes R = 0.083 (p = 1.1e-17), packets R = 0.142 (p = 2.5e-49); duration R = 0.075 (p = 5.4e-06) and 0.091 (p = 3.9e-08). First and last tenth durations 0.597 and 0.650 s (episodes), 0.603 and 0.651 s (packets), all within tolerance. The criterion's reproduced condition (increases and durations) holds. Amplitude does not change (p >= 0.16); frequency does not change within episodes (p = 0.11) but falls within packets (R = -0.043, p = 0.0096), so that part of the statement fails; the criterion's final rule does not judge it. With the Dino ACC sessions left out, incidence and duration still increase (reading b, p <= 2e-06).

How the test differed from the paper’s method

• Deposited spindle detections.
• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).

Spectral comparison of UP states in first third versus last third of packets showed increased sigma power ( Figure 4E ; p < 10 −10 ; Wilcoxon rank-sum test and per-pixel significance; Figure S4B )

Results, Firing rate and LFP changes within nonREM episodes and packets; Figure 4E; Figure S4B and S4D

What the paper reports

• UP state sigma power, first against last third of packets: Wilcoxon rank-sum test, p < 10⁻¹⁰
• DOWN-UP transition-triggered sigma power, late against early packet (Results, Figure S4B): t-test, p < 10⁻⁵

UP states of at least 1 s, late third mean sigma z 0.153 against early 0.009 (n 1708 and 1866), rank-sum p = 3.7e-34; UP states over 500 ms, 0.151 against 0.031, p = 2.5e-76.

How the test differed from the paper’s method

• Sigma power per UP state is the mean of an 8 ms band-power series (log10, z-scored per session), not an UP-onset-triggered spectrogram.
• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.

Firing rates of the fastest-firing sextile groups decreased, whereas those of the slowest-firing sextile groups increased within both nonREM episodes and packets ( Figure 4A, B iv ; R and p values in figure), resulting in a decreasing coefficient of variation of binned population firing rates ( Figure 4 A-B v ; R episode = −0.13, p < 10 −10 ; R packet = - 0.12, p < 10 −10 ). An additional metric, the 15ms spike burst index of pyramidal neurons, also showed a positive slope within both nonREM episodes and packets ( Figure 4 A–B vi ; R episode = 0.09, p < 10 −10 ; R packet = 0.08, p < 10 −10 )

Results, Firing rate and LFP changes within nonREM episodes and packets; Figure 4A-B iv, v and vi

What the paper reports

• CV within episodes: R = -0.13, p < 10⁻¹⁰
• CV within packets: R = -0.12, p < 10⁻¹⁰
• burst index within episodes: R = 0.09, p < 10⁻¹⁰
• burst index within packets: R = 0.08, p < 10⁻¹⁰

Reading (b), top sextile R = -0.139 (episodes), -0.137 (packets); bottom R = +0.029 (p = 0.0048), +0.044 (p = 2.3e-05); CV R = -0.103 (p = 1e-26), -0.139 (p = 5.2e-47); burst ratio within packets R = +0.089 (p = 2.9e-20). Within episodes the burst ratio meets p < 1e-10 with clock (ii) (R = 0.063, p = 8.5e-11) and misses it with clock (i) (R = 0.061, p = 2.6e-10), which would make that correlation partial by M4. The outcome takes the better of the two clocks, here clock (ii), as for every trend.

How the test differed from the paper’s method

• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.
• The outcome depends on the normalization clock, which the paper does not state (Results; Fig. 3B legend); clock (i), the period's clock time, and clock (ii), the cumulative nonREM time, give different outcomes here. As for every trend, the outcome takes the result under one of them, as M4 does for readings; the question is put to the paper's authors.

This resulted in 995 putative pyramidal (pE) cells (mean of 36.9 cells/session ... 10-104) and 126 putative interneurons (pI) (4.67 cells/session ... 0-13)

Results, Assessment of firing patterns in different brain states, and Firing pattern and LFP changes across the course of sleep; Table S1

995 pE and 126 pI cells (mean 36.85 pE per session, range 10 to 104; pI mean 4.67, range 0 to 13), 27 sessions and 11 rats match exactly. The WAKE-SLEEP cycles in _WSRestrictedIntervals.mat number 51, not 54; each session's Wake and Sleep intervals are exactly its WakeSleep pairs (51, 51 and 51 in all), with no unpaired WAKE or SLEEP, so no cycle was lost in parsing. The criterion names no partial outcome; partial is given because five of six counts match, and whether a partial rule applies is asked of a reader.

There were a total of 790 MAs in our dataset (median duration = 15.9±3.3sec; frequency = 0.32±0.12 per minute; 9.1±3.5% of time in SLEEP). Over nonREM episodes the fraction of time occupied by MAs decreased (correlation versus normalized nonREM episode time

Results, Identification of brain states; Table S1

What the paper reports

• MA time fraction against normalized nonREM episode time: correlation, R = -0.85, p < 10⁻¹⁰

434 MA intervals lie inside SLEEP (2130 in the files). Median duration 11.0 s over all MAs (11.5 s over session medians), per-session median incidence 0.184 per minute, per-session median 3.70% of SLEEP, against 790, 15.9 s, 0.32 and 9.1%; none within tolerance. The MA fraction falls over normalized episodes (20 bins, bin means R = -0.956, p = 4.5e-11; 10 bins R = -0.970, p = 3.3e-06), and over SLEEP the bin means are not significant (p = 0.15, 0.098; pooled period-bins p = 0.0057, 0.00061). The criterion names no partial outcome; partial is given because the trends hold and the four values do not. The 345 WakeInterruptions shorter than 40 s are not MAs by the paper's definition, since none lies between two packets (123 between two REM intervals, 80 between REM and a packet). Variant, MA reading (b), counting them as MAs, 779 events, 0.318 per minute (within tolerance), median 13.0 s, 7.85% of SLEEP, episodes R = -0.960 (p = 1.9e-11), SLEEP p = 0.93. With the Dino ACC sessions left out, episodes R = -0.949 (p = 1.7e-10), SLEEP p = 0.089.

How the test differed from the paper’s method

• The episode correlation uses the MA fraction in 10 or 20 normalized bins, both as bin means over episodes and as pooled episode-bins, as the criterion allows; episodes include MA and packet seconds.

Burst index was highest during nonREM (median in nonREM = 11.8%) and lowest during WAKE (4.0 %), with REM (5.5 %) slightly higher than WAKE (p < 0.05; Kruskal-Wallis test)

Results, Assessment of firing patterns in different brain states

What the paper reports

• Kruskal-Wallis test, p < 0.05

Medians (nonREM, WAKE, REM) reading (a) 6.26, 2.43, 3.56%; (b) 4.93, 1.58, 2.33%; (c) 11.47, 4.51, 6.39%. Kruskal-Wallis p from 3.4e-46 to 3.6e-54; REM above WAKE by Dunn (p 1.3e-06 to 1e-05) and rank-sum. The order and the tests hold under every reading; no reading puts all three medians within tolerance.

How the test differed from the paper’s method

• Reading (c), the fraction of spikes with an ISI <= 15 ms before or after them, is added from the Results' wording, 'the fraction of spikes with ISIs below 15 ms' (Results, Assessment of firing patterns in different brain states), beside the claims file's readings of a proportion of ISIs.

a large peak occurred at approximately 8ms inter-spike interval during nonREM, whereas smaller peaks were present at ~125 ms (i.e., theta frequency) during REM and WAKE ( Figure 2C ). ... Time decay of the autocorrelograms of pyramidal neurons was significantly faster during nonREM compared to WAKE and REM ( Figure 2C ; decay fit exponent median nonREM = - 1.35, WAKE = - 0.83, REM = - 0.80; p < 0.05; Kruskal-Wallis test)

Results, Assessment of firing patterns in different brain states; Figure 2C

What the paper reports

• Kruskal-Wallis test, p < 0.05

Average nonREM autocorrelogram peaks at 8.5 ms; WAKE and REM have local maxima at 125 and 126 ms (nonREM also at 125 ms). Median decay, reading (a) WAKE -0.108, nonREM -0.538, REM -0.169; (b) -0.118, -0.336, -0.112; (c) -0.582, -1.623, -0.574. nonREM is the most negative and Kruskal-Wallis p <= 1.7e-95 under all three; no medians within tolerance. Cells with fewer than 100 counts dropped (158 of 995 in WAKE, 131 in nonREM, 284 in REM). The 125 ms bump is small, the average autocorrelogram at 115 to 135 ms being 1.027 (WAKE), 1.026 (REM) and 0.981 (nonREM) times its mean at 80 to 100 and 150 to 170 ms; the local-maximum test also fires at 125 ms in nonREM.

How the test differed from the paper’s method

• Per-cell peaks are found on a 5 ms boxcar-smoothed autocorrelogram in 1 to 50 ms, and cells with fewer than 100 counts are dropped; both are standard choices the paper does not state.

This effect was 'carried over' to subsequent WAKE as shown by the significantly decreased discharge rate of pE neurons in a 5-min post-sleep WAKE compared to a 5-min pre-sleep WAKE epoch (p < 10 −6 ; Wilcoxon test)

Results, Firing pattern and LFP changes across the course of sleep

What the paper reports

• Wilcoxon test, p < 10⁻⁶

Only 23 cycles have a scored WAKE after their SLEEP; in 17 it starts 166 to 10344 s later. pE rates are lower after, signed-rank p = 0.018 (reading a, 703 cell-cycles) and 0.048 (b, 494 cells); rank-sum p = 0.95 and 0.98. M4 gives partial (sign right, p < 0.05, not below 1e-6). Restricted to the 6 cycles whose next WAKE starts within 1 s, signed-rank p = 0.00026. Variant, the paper's wording read as the first 5 min after SLEEP ends (47 cycles), signed-rank p = 3.9e-10 (a) and 2.8e-06 (b); on average 0.18 of those 5 min is scored as a cycle's WAKE (none in 38 cycles). Whether that is the paper's window is asked of a reader; the outcome is the claims file's reading.

How the test differed from the paper’s method

• The data hold scored WAKE only as the WAKE of a WAKE-SLEEP cycle, so the WAKE after a SLEEP is the next cycle's WAKE, often hours later; the variant restricted to an immediate WAKE is shown.

Incidence of spindles (1.94±0.39/min) decreased (R = - 0. 093; p < 10 −5 ), while spindle duration (0.52±0.32 sec), frequency (13.93±0.93 Hz) and amplitude (203.3±87.71 µV) did not change over SLEEP

Results, Firing pattern and LFP changes across the course of sleep; Figure 3C, bottom

What the paper reports

• spindle incidence against normalized SLEEP time: correlation, R = -0.093, p < 10⁻⁵

Deposited spindles (5201 of 5203 start in nonREM of SLEEP); per-session mean incidence 2.852 +- 0.490 per minute, duration 0.616 s, frequency 13.923 Hz, amplitude 208.4 uV. Frequency and amplitude match; incidence matches Table S1 (0.049 Hz, 2.94 per minute) not the Results' 1.94; duration is outside 0.52 +- 0.052. Incidence over SLEEP, reading (b) R = -0.021, p = 0.03 (partial by M4); (a) p = 0.44; (c) p = 0.66; clock (ii) p >= 0.19. Absence of change holds for duration (p >= 0.068, clock i) but frequency changes under (b) (p = 0.011) and amplitude under (a) (p = 0.047). Over all 5201 spindles the duration is 0.614 +- 0.320 s, median 0.524 s, which matches the paper's 0.52 +- 0.32. With the two Dino ACC sessions left out, the incidence decrease is not significant (reading b p = 0.15).

How the test differed from the paper’s method

• Uses the deposited spindle detections (_Spindles.mat), reading (a) of the criterion; a new detection was not run. Amplitude converted from V to uV.
• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).
• The outcome depends on the normalization clock, which the paper does not state (Results; Fig. 3B legend); clock (i), the period's clock time, and clock (ii), the cumulative nonREM time, give different outcomes here. As for every trend, the outcome takes the result under one of them, as M4 does for readings; the question is put to the paper's authors.
• The incidence decrease also depends on the two Dino days that were recorded with two probes and are deposited as two sessions each (ACC and mPFC); without the ACC copies, reading (b) clock (i) gives p = 0.15. Both copies are kept, since the paper's 27 sessions, which the data set matches, include them.

first versus last REM episode, also note significant deviation of slope from 1 (confidence interval ... 0.76-0.83). Right first versus last MA episode per sleep session, confidence interval of slope does not include 1 (0.58- 0.68)

Results, Firing pattern and LFP changes across the course of sleep, last paragraph, and Sleep sub-states induce differential lasting firing rate changes; Figure S3A

What the paper reports

• first against last REM: 95% CI 0.76 to 0.83
• first against last MA: 95% CI 0.58 to 0.68
• WAKE before against WAKE after SLEEP: 95% CI 0.8 to 0.86
• correlation of WAKE before and WAKE after SLEEP rates (Figure S3A legend): Pearson, p < 10⁻¹⁰

Reading (b) REM 0.807 (0.769 to 0.845), MA 0.656 (0.601 to 0.710), WAKE 0.687 (0.627 to 0.746); reading (a) 0.742, 0.530, 0.713. REM and MA within tolerance under (b), WAKE outside under both; all intervals below 1. MA reading (b), short WakeInterruptions as MAs, MA slope 0.618 (0.563 to 0.674, reading b). With the post-sleep window read as the first 5 min after SLEEP, the WAKE slope is 0.793 (0.759 to 0.828, reading b) and 0.787 (reading a), still outside tolerance.

How the test differed from the paper’s method

• WAKE windows as in post-sleep-wake-lower, the next cycle's WAKE, which follows SLEEP immediately in only 6 of 23 cycles.

Comparison of firing rates during 5 minutes of WAKE immediately before SLEEP versus 5 minutes of WAKE immediately after SLEEP showed a significant decrease firing in the top three sextile groups and an increase in the lowest sextile group (all ps < 0.05; stars in Figure 3E )

Results, Sleep sub-states induce differential lasting firing rate changes; Figure 3E

What the paper reports

• one-tailed Wilcoxon, p < 0.05

One-tailed signed-rank per sextile. M2 reading (a), M1 reading (a), sextile 6 p = 0.0018, 5 p = 0.032, 4 p = 0.48, lowest (increase) p = 0.15, 2 of 4 parts; M1 reading (b) 0.00032, 0.073, 0.61, 0.1, 1 of 4. M2 reading (b), per-cycle sextiles, sextile 6 p = 3.2e-06, 5 p = 0.00013, 4 p = 0.23, lowest p = 2e-05, 3 of 4, the criterion's partial outcome. That sextile is ranked on the WAKE that holds the 5 min before SLEEP, so part of the change can be regression to the mean (the session-wide WAKE of reading a holds the window too). Under M2 reading (b) the second-lowest sextile, which the paper does not report as changing, also increases (p = 0.012, against 0.83 under M2 reading a), which fits that concern. Immediate-WAKE variant, 6 cycles, sextile 6 p = 0.00037, 5 p = 0.028, 4 p = 0.11, lowest p = 0.7. With the post-sleep window read as the first 5 min after SLEEP (asked of a reader), 6 p = 9.7e-13, 5 p = 1.7e-06, 4 p = 0.084, lowest p = 0.00096.

How the test differed from the paper’s method

• WAKE windows as in post-sleep-wake-lower.
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

Over nonREM episodes and packets, delta and sigma band power steadily increased ( Trachsel et al., 1988 ), whereas gamma power decreased over episodes but not packets ( Figure 4 A-B i ). ... Mean delta power within whole packets (normalized by nonREM delta power) was positively correlated with packet length (R = 0.31; p < 0.05)

Results, Firing rate and LFP changes within nonREM episodes and packets; Figure 4A-B i

What the paper reports

• packet delta power against packet length: R = 0.31, p < 0.05

Delta and sigma increase within episodes and packets under every reading (packets, reading b, delta R = 0.329, sigma R = 0.306; episodes R = 0.276, 0.209). Gamma within episodes R = -0.220 (a), -0.129 (b), -0.155 (c, p = 0.28) with clock (i), and -0.245, -0.131, -0.121 (p = 0.4) with clock (ii); within packets R = -0.052, p = 6e-08 (b), +0.187, p = 0.19 (c). Packet delta against length R = 0.387, p = 1.8e-26. No single reading gives all six trends; the delta and sigma increases hold under all. Variant, seconds whose 10 s window reaches outside the packet left out, gamma within episodes R = -0.300, p = 0.034 (c) and within packets R = -0.059, p = 3.1e-09 (b), p = 0.14 (c), so under reading (c) all six trends hold with this variant; the claims file does not name it and the outcome is judged on its spectrogram. With the Dino ACC sessions left out, the trends are unchanged (gamma within packets, reading b, R = -0.064, p = 2.2e-10).

How the test differed from the paper’s method

• Band power from a spectrogram of goodeegchannel (10 s Hann windows every 1 s), as the claims file allows.
• Band power is log10-transformed before the per-session z-scoring the criterion names (a standard choice for power, not stated in the paper).
• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).

Per-packet firing rate of the top sextile group was negatively correlated with per-packet LFP metrics, including whole-packet delta power (Z-scored per session; Figure 4C ; R= - 0.28; p < 10 −10 ), DOWN state duration (R = - 0.16, p < 10 −4 ), and DOWN state incidence rate (Z-scored per session; R = - 0.29, p<10 −10 ). A positive correlation was observed between firing rate of the top sextile group and UP state duration (R = 0.14, p < 10 −3 ) and gamma power ( Figure 4D ; R = 0.44; p < 10 −10 ). Finally, a positive correlation was found between normalized firing rates per packet in the bottom sextile group and sigma power in the last 20 sec of the packet (R = 0.25; p < 10 −10 ). Other tested relationship combinations were not significant

Results, Firing rate and LFP changes within nonREM episodes and packets; Figure 4C and 4D

What the paper reports

• top sextile rate against delta power: R = -0.28, p < 10⁻¹⁰
• top sextile rate against DOWN duration: R = -0.16, p < 10⁻⁴
• top sextile rate against DOWN incidence: R = -0.29, p < 10⁻¹⁰
• top sextile rate against UP duration: R = 0.14, p < 0.001
• top sextile rate against gamma power: R = 0.44, p < 10⁻¹⁰
• bottom sextile rate against sigma power in the last 20 s: R = 0.25, p < 10⁻¹⁰

M2 reading (a), top sextile against delta R = -0.200 (p = 1.6e-06), DOWN duration R = -0.387 (p = 1.7e-21), DOWN incidence R = -0.300 (p = 2.9e-13), UP duration R = 0.248 (p = 2.8e-09), gamma R = 0.358 (p = 1.4e-18); bottom sextile against last-20 s sigma R = 0.215 (p = 7.7e-08). Four meet the paper's bounds and two have its sign at p < 0.05; 4 of 6, the criterion's partial outcome. (With the sextiles of the first revision, the mean of per-cycle WAKE rates, 3 of 6 held.) M2 reading (b) also gives 4 of 6, with no sigma correlation (R = 0.062, p = 0.12). Other combinations, bottom sextile against UP duration R = 0.110, p = 0.01, top sextile against late sigma R = -0.093, p = 0.026; the rest p >= 0.23. Sigma from 10 s windows inside the last 20 s, R = 0.199, p = 6.9e-07.

How the test differed from the paper’s method

• Band power is log10-transformed before the per-session z-scoring the criterion names (a standard choice for power, not stated in the paper).
• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

Analysis of rate change versus firing rate for these triplets also showed an inverse correlation between rate and change ( Figure S5C ), as did a similar analysis for REM n - packet - REM n+1 triplets ( Figure S5D )

Results, Sleep sub-states induce differential lasting firing rate changes; Figure S5D

What the paper reports

• R = -0.39537, p = 1.9 × 10⁻⁶

Adopted rule, MA or WakeInterruption allowed between REM n and the packet, 80 triplets in 17 sessions, 660 pE cells (paper about 136). S5D correlation R = -0.097 (p = 0.015, linear WAKE rate, reading b) and R = -0.077 (p = 0.054, log10, reading a); the paper's sign at p < 0.05, short of p < 1e-5, partial by M4. All states adjacent, 25 triplets in 3 sessions, 57 cells, R = 0.019 (p = 0.89) and -0.055 (p = 0.69). Interruptions allowed on both sides, 84 triplets, R = -0.071 (p = 0.071) and -0.103 (p = 0.0089). No rule gives the paper's number of cells.

How the test differed from the paper’s method

• REM n, packet, REM n+1 triplets allow an MA or WakeInterruption between REM n and the packet, the packet and REM n+1 being adjacent (gap at most 2 s). Assumed because the Supplemental Experimental Procedures tolerate an arousal after REM in nonREM-REM-nonREM triplets (claims file, M6), and REM in these files is followed by a WakeInterruption more often than by a packet, so the same tolerance is applied after REM n. The all-adjacent rule (the first revision's) and a rule allowing interruptions on both sides are shown; the question is put to the paper's authors.

WAKE episodes increased the activity of the fastest-firing sextile group of pyramidal cells from nonREM packet n to packet n+1 ( Figure 5B , black). Furthermore, WAKE periods with more movement had a proportionally larger effect ( Figure 5C ; R = 0.43; p < 0.05). WAKE also appeared to exert a rate-decreasing effect on slow firing neurons, although this effect was not significant

Results, Sleep sub-states induce differential lasting firing rate changes; Figure 5B, black; Figure 5C, upper left; Figure S5E

What the paper reports

• movement during WAKE against fastest sextile change: R = 0.43, p < 0.05 (n = 24 nonREM-WAKE-nonREM triplets)
• log rate change against WAKE rate (Figure S5E): R = 0.043591, p = 0.2963

Reading (a) 8 triplets, reading (b) 52 (paper 24). Fastest sextile change +0.092 Hz (p = 0.21) and +0.063 Hz (p = 0.15), cells averaged. Movement correlation, reading (a), raw motion R = 0.436 (p = 0.33, n 7), motion z-scored per session R = 0.869 (p = 0.011, n 7); reading (b) R = 0.124 and 0.248 (p = 0.43, 0.11). M2 reading (b), z-scored R = 0.924 (p = 0.0029, n 7). Slow cells, no significant decrease (lowest sextile p >= 0.12), as in the paper; S5E correlation absent (p >= 0.75), as in the paper. Partial rests on the z-scored motion under reading (a), with 7 triplets. The stored motion's per-session means run from 0.55 to 4076, so the raw values pooled across sessions mostly compare sessions.

How the test differed from the paper’s method

• Triplets use adjacent scored intervals inside a SLEEP with gaps of at most 2 s (a standard choice for intervals scored at 1 s); the paper does not say how adjacency is decided. Assumed, as the reading closest to the Supplemental Experimental Procedures' description of consecutive sub-states.
• The motion signal is z-scored per session before averaging over the WAKE period. Assumed because the data description says the signal is z-scored while the stored values are not (non-negative, per-session means 0.55 to 4076); the raw signal gives no correlation.
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

In contrast to WAKE, MA epochs between packets did not affect fast-firing neurons but significantly increased the rate of slow-firing neurons from packet n to packet n+1 ( Figure 5B , yellow). Again, the differential effect across sextiles was confirmed by a correlation analysis

Results, Sleep sub-states induce differential lasting firing rate changes; Figure 5B, yellow; Figure S5G

What the paper reports

• log rate change against WAKE rate (Figure S5G): R = -0.20668, p = 7.7 × 10⁻¹¹

264 packet-MA-packet triplets. Lowest sextile, one entry per cell and triplet (M1 a), mean +0.0075 Hz, median +0.0017 Hz, 51% positive, signed-rank p = 0.015, t-test p = 0.016; cells averaged +0.007 Hz, p = 0.36; lowest two together p = 0.92. M2 reading (b), lowest sextile +0.012 Hz cells averaged (p = 0.022), per cell and triplet p = 0.00088. Top sextile decreases (-0.140 Hz, p = 5.4e-08), where the paper finds no change. S5G R = -0.174, p = 4.9e-08 (log10), misses p < 1e-10. MA reading (b) gives the same 264 triplets.

How the test differed from the paper’s method

• Triplets use adjacent scored intervals inside a SLEEP with gaps of at most 2 s (a standard choice for intervals scored at 1 s); the paper does not say how adjacency is decided. Assumed, as the reading closest to the Supplemental Experimental Procedures' description of consecutive sub-states.
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

During nonREM, activity of high firing rate neurons decreased, whereas activity of slow firing neurons increased. REM sleep reduced firing rates relatively uniformly of all firing rate groups, whereas microarousals, interspersed between nonREM epochs, increased the discharge rates of slow firing neurons. Accordingly, the net result of sleep is that the firing rate variation of the population is decreased

Discussion, paragraph 1; Figure 6

fig3d-first-last-packet-slope reproduced (0.872, 0.840 to 0.904); fig3e-wake-before-after-sextiles partially reproduced (3 of 4 parts under M2 reading b, 2 under reading a); cv-decreases-over-sleep not reproduced (R = -0.003, p = 0.73). Of the sub-state claims reported beside it, fig5b-ma-between-packets is partially reproduced and fig5b-rem-between-packets is not.

the distribution of per-cell mean firing rates for putative pyramidal neurons was strongly positively skewed ... spanning three orders of magnitude range from 0.05 Hz to 5 Hz ( Figure 2A ). Median rates (±SD) of putative pyramidal cells in each state were: WAKE: 0.76±1.53 Hz, nonREM: 0.69±0.86 Hz, REM: 0.88±1.33 Hz; and for putative interneurons were: WAKE: 5.59±7.25 Hz, nonREM: 4.69±5.62 Hz, REM 4.25±9.43 Hz

Results, Assessment of firing patterns in different brain states; Figure 2A; Figure S2C

Reading (a) medians pE 0.755, 0.706, 0.936 Hz and pI 6.001, 4.565, 5.724 Hz; reading (b) pE 0.784, 0.691, 0.922 and pI 5.632, 4.637, 5.724. Five of six within 10% under both; pI REM is 5.724 against 4.25. pE skew 2.18 to 2.83. The 1st to 99th percentile range covers 0.05 to 5 Hz within 10% in WAKE (0.004 to 7.14 Hz, reading b) and REM (0.008 to 6.03) but not nonREM (0.014 to 4.17; 4.05 under reading a); 5th to 95th percentiles stop below 4.5 Hz in every state. The criterion requires the shape for both the reproduced and the partial outcome, so the outcome is not reproduced. Whether the range applies to each state, when the paper itself says the nonREM tail is shorter, is asked of a reader. The pI REM median is equal under both readings because the middle cell is the same (Dino_061814_mPFC cell 46, 5.706 and 5.743 Hz in its two cycles); reading (a) has no duplicated rows.

How the test differed from the paper’s method

• The range test is applied in each state, as the criterion's 'in each state' reads.

UP states were characterized by an initial surge of gamma power, followed by increased sigma power ( Figure 1E ), corresponding to UP state-related sleep spindles. Firing rates of both pyramidal cells and interneurons decreased over the time course of the UP state, accompanied by a decreasing gain of excitation (excitatory-inhibitory, E-I, ratio) and decreasing population synchrony of pyramidal neurons

Results, Assessment of firing patterns in different brain states; Figure 1E

Per normalized bin over 43230 UP states, pE rate 1.823, 1.359 ... 1.310, 1.619 Hz (last not below the first two; R = -0.293); pI 15.88 to 11.63 Hz (R = -0.983); synchrony 0.079 to 0.071 (last below first two, R = -0.447); E/(E+I) 0.373 to 0.396 (R = +0.470). Sigma power peaks in bin 0 and gamma in bin 3, the reverse of the statement. Two of four measures decrease and the spectral order fails.

How the test differed from the paper’s method

• The criterion's 50 ms bins for synchrony and E/(E+I) are taken as the 10 normalized UP-state bins (47 ms for a median UP state).
• Gamma and sigma power come from an 8 ms band-power series (4th order Butterworth, mean of the squared signal) on goodeegchannel, log10 and z-scored per session; the paper's spectrogram method is not given.
• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.
• The 8 ms series is zero-phase band-passed, which at 9 to 25 Hz spreads power over tens of ms, so the sigma peak in the first bin may carry power from the DOWN-UP edge; the outcome fails on the pE rate and E/(E+I) regardless.

the fastest and slowest firing neurons tended to be positively correlated within groups, but were actually negatively correlated across groups ( Figure 2F )

Results, Assessment of firing patterns in different brain states; Figure 2F

Mean 100 ms count correlations in nonREM, WAKE sextiles, 6-6 0.0328 (p = 2.9e-70), 1-1 0.0062 (p = 2.1e-20), 1-6 0.0064 (p = 2e-12, positive). Correlations rise with rate group across the whole matrix; no negative group pair. nonREM sextiles, the claims file's reading (b) for this figure, give 1-6 0.0095 (p = 8.5e-66).

How the test differed from the paper’s method

• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

the coefficient of firing rate variation also showed a significantly decreasing slope over the course of SLEEP (R = - 0.05; p < 0.01)

Results, Firing pattern and LFP changes across the course of sleep

What the paper reports

• Pearson correlation versus normalized time, R = -0.05, p < 0.01

Reading (b), the criterion's, R = -0.003, p = 0.73 (clock ii R = -0.012, p = 0.2). Variants, one point per packet R = -0.101, p = 0.0073; 50 bins clock (ii) R = -0.483, p = 0.00039; 50 bins clock (i) R = -0.025, p = 0.86.

How the test differed from the paper’s method

• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).

Incidence of both UP states (R = - 0.10; p <10 −6 ) and DOWN states (R = - 0.08; p < 10 −4 ) decreased significantly, while the duration of DOWN states increased (R slope = 0.09; p < 10 −4 ; Figure 3B ) and duration of UP states did not change. In addition, the fraction of time that could be identified as neither UP nor DOWN state ('other') increased significantly over the course of sleep (R = 0.08; p < 10 −2 )

Results, Firing pattern and LFP changes across the course of sleep; Figure 3C

What the paper reports

• UP incidence: R = -0.10, p < 10⁻⁶
• DOWN incidence: R = -0.08, p < 10⁻⁴
• DOWN duration: R = 0.09, p < 10⁻⁴
• fraction neither UP nor DOWN: R = 0.08, p < 0.01

Reading (b), UP incidence R = -0.013 (p = 0.19), DOWN incidence R = -0.031 (p = 0.0015), DOWN duration R = 0.117 (p = 2.4e-32), UP duration R = 0.020 (p = 0.067), other fraction R = -0.007 (p = 0.44). At most 2 of 5 parts hold under any reading and clock. With the Dino ACC sessions left out, reading (b) gives the same pattern (DOWN incidence p = 0.0023, DOWN duration p = 1.5e-30, UP incidence p = 0.23).

How the test differed from the paper’s method

• Trends over normalized time follow M3 and M4 of the claims file; every reading (a, b, c) is run and, for SLEEP and episodes, two normalization clocks, (i) the period's clock time and (ii) the cumulative time of the used state, since the paper says only that SLEEP is time-normalized with nonREM time used (Results, Firing pattern and LFP changes across the course of sleep; Fig. 3B legend). Values are divided by the session mean before pooling (M3).
• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.

Comparison of within-UP state activity in the first versus second third of packets showed a drop in firing rate for the fastest sextile group (P < 10 −8 ; Wilcoxon rank-sum test), an increase in firing rate of putative interneurons (P < 10 −3 ), and a small but significant decrease for the slowest group (P < 0.003; Figure 4B, E )

Results, Firing rate and LFP changes within nonREM episodes and packets; Figure 4E, bottom

What the paper reports

• fastest sextile: Wilcoxon rank-sum test, p < 10⁻⁸
• putative interneurons: Wilcoxon rank-sum test, p < 0.001
• slowest sextile: Wilcoxon rank-sum test, p < 0.003

Per UP state, first against second third, fastest 3.43 to 3.39 Hz (p = 5.8e-05), pI 13.11 to 13.65 Hz (p = 0.00078), slowest 0.29 to 0.31 Hz (p = 0.64, rising); against the last third, fastest p = 0.0069, pI p = 3.1e-05, slowest rising p = 5.3e-07. Per cell (unit ii, one entry per cell, its within-UP rate in each third over all its packets; 166, 126 and 165 cells) no group meets its bound (p >= 0.11). At most 1 of 3 holds under any reading, M2 reading (b) included.

How the test differed from the paper’s method

• Uses the deposited UP/DOWN detections (_UPDOWNINtervals.mat), reading (a) of updown-durations; a new detection was not run, since the LFP reduction kept has no power above 100 Hz.
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.
• Unit (ii) has one entry per cell, its spikes and UP time summed over all its packets in each third (the first revision had one entry per cell and packet).

We found that entry into WAKE corresponded with an increase in firing rate of the highest sextile two groups and a drop in firing rate of the lowest two sextile groups ( Figure S5A ), opposing the effects seen in SLEEP and nonREM. During REM, all sextile groups except the fourth showed an increase in firing rate

Results, Firing rate and LFP changes within nonREM episodes and packets, last paragraph; Figure S5A and S5B

50 bins, sextiles numbered 1 lowest, M2 reading (a). WAKE, sextile 6 R = 0.107 (p = 0.46), 5 R = 0.471 (p = 0.00055), 2 R = 0.603 (p = 3.5e-06, increasing), 1 R = -0.286 (p = 0.044). REM, sextiles 1 to 6 R = -0.245, 0.443, 0.358, 0.394, -0.064, 0.308 (p = 0.086, 0.0013, 0.011, 0.0047, 0.66, 0.029); of the five other than sextile 3, three increase. M2 reading (b) holds less (WAKE sextile 6 p = 0.58; REM sextiles 1, 4, 5, 6 p >= 0.065).

How the test differed from the paper’s method

• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

Firing rates in the top two firing rate sextiles were significantly reduced by the intervening nonREM packet (both p < 0.01; Figure 5B , blue)

Results, Sleep sub-states induce differential lasting firing rate changes; Figure 5B, blue; Figure S5C

What the paper reports

• top two sextiles, MA n+1 against MA n: p < 0.01
• log rate change against WAKE rate (Figure S5C): R = -0.14923, p = 4 × 10⁻⁶

169 MA-packet-MA triplets. Top two sextiles mean change -0.011 and -0.010 Hz (cells averaged), signed-rank and t-test p >= 0.24 under M1 a and b. S5C correlation R = 0.016 (p = 0.65, log10 WAKE rate, 843 cells), R = 0.006 (linear). M2 reading (b), top two p >= 0.12. MA reading (b), short WakeInterruptions as MAs, 234 triplets, top two p >= 0.28, S5C R = -0.054 (p = 0.11).

How the test differed from the paper’s method

• Triplets use adjacent scored intervals inside a SLEEP with gaps of at most 2 s (a standard choice for intervals scored at 1 s); the paper does not say how adjacency is decided. Assumed, as the reading closest to the Supplemental Experimental Procedures' description of consecutive sub-states.
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

we found that sigma power in the last 20 seconds of nonREM packets was correlated with a greater increase in the lowest firing-rate sextile group in the MA n versus MA n+1 analysis ( Figure 5C ; R = 0.181, p < 0.01)

Results, Sleep sub-states induce differential lasting firing rate changes; Figure 5C, lower right

What the paper reports

• R = 0.181, p < 0.01

150 MA-packet-MA triplets with lowest-sextile cells; R = -0.042, p = 0.61. Sigma from 10 s windows inside the last 20 s R = -0.018 (p = 0.83); MA reading (b) R = -0.013 (p = 0.86, 209 triplets); M2 reading (b) R = -0.053 (p = 0.52).

How the test differed from the paper’s method

• Triplets use adjacent scored intervals inside a SLEEP with gaps of at most 2 s (a standard choice for intervals scored at 1 s); the paper does not say how adjacency is decided. Assumed, as the reading closest to the Supplemental Experimental Procedures' description of consecutive sub-states.
• Band power is log10-transformed before the per-session z-scoring the criterion names (a standard choice for power, not stated in the paper).
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

Finally, REM episodes decreased the spike rates from packet n to packet n+1 in all sextile groups ( Figure 5B , red). Longer REM episodes exerted a proportionally larger effect on rate changes from packet n to packet n+1 for all sextile groups ( Figure 5C ; R = - 0.22; p < 10 −2 ; t-test) and REM theta power was significantly correlated with the packet n to packet n+1 rate decrease of the slowest sextile group only (R = - 0.15; p < 0.05). Correlational analysis did not reveal differential effects of REM on neurons across the rate spectrum

Results, Sleep sub-states induce differential lasting firing rate changes; Figure 5B, red; Figure 5C, upper right; Figure S5F

What the paper reports

• REM duration against rate change: t-test, R = -0.22, p < 0.01
• REM theta power against slowest sextile change: R = -0.15, p < 0.05
• log rate change against WAKE rate (Figure S5F): R = -0.038346, p = 0.2372

110 packet-REM-packet triplets. Sextile changes (1 lowest, cells averaged) -0.016, -0.041, -0.012, +0.006, +0.036, +0.164 Hz (signed-rank p 0.11, 0.00056, 0.72, 0.5, 0.23, 0.065). Duration correlation R = -0.019 (p = 0.85). Theta against slowest change R = 0.143 (ratio, p = 0.18), 0.068 (power, p = 0.53). S5F R = 0.068 (p = 0.06, log10), 0.122 (p = 0.00069, linear), where the paper finds none. M2 reading (b), one sextile decreases (the second lowest, p = 0.01).

How the test differed from the paper’s method

• Triplets use adjacent scored intervals inside a SLEEP with gaps of at most 2 s (a standard choice for intervals scored at 1 s); the paper does not say how adjacency is decided. Assumed, as the reading closest to the Supplemental Experimental Procedures' description of consecutive sub-states.
• The second packet is the next packet after the REM episode with any MA or WakeInterruption between them (M6, Supplemental Experimental Procedures).
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

The increase of spike rate from packet n to packet n+1 in the slowest sextile group was positively correlated with the delta power change from packet n to packet n+1 ( Figure 5C )

Results, Sleep sub-states induce differential lasting firing rate changes; Figure 5C, lower left

567 consecutive packet pairs; slowest sextile change against delta change R = 0.003 (p = 0.95); fastest R = -0.285 (p = 2.8e-11, 524 pairs). Delta from windows inside the packets R = -0.008 (p = 0.86); M2 reading (b) R = -0.011 (p = 0.78).

How the test differed from the paper’s method

• Band power is log10-transformed before the per-session z-scoring the criterion names (a standard choice for power, not stated in the paper).
• Sextiles follow M2. Reading (a) ranks each pE cell's rate over all scored WAKE of its session (before round 1's revision, the mean of its per-cycle WAKE rates, which put 65 of 995 cells in another sextile); reading (b) ranks each cell's rate in each cycle's WAKE and uses the cycle that holds the analysed period.

Comparison of log firing rates of individual putative interneurons during the first and last packets of SLEEP. Unlike for pE cells (Figure 2B) the regression line is not different from unity (slope 95% confidence interval ... 0.996-1.005)

Figure S2D legend

What the paper reports

• 95% CI 0.996 to 1.005

pI slope of log10 last on first packet rate, reading (a) 0.836 (0.766 to 0.906, n 192), reading (b) 0.888 (0.802 to 0.973, n 123); neither interval includes 1.