DANDI:000959

In our main experiments (Eyes stimulation; n = 15 per area from monkey L and n = 12 per area from monkey T)

STAR Methods, Closed-loop microstimulation protocol, paragraph 1; Figure 1B legend

Days holding kept stimulation and sham trials are 15 per area for monkey L, 12 per area for monkey T (Eyes) and 5 per area for monkey T (Mouth). Each day holds one region.

The total number of microstimulations (and shams) received per day was comparable across the three stimulated regions and comparable between the two animals both in the social gaze and non-social control conditions (Fig. 1E, live social gaze condition, all p > 0.90

Results, opening section, paragraph 2; Figure 1E legend

What the paper reports

• Wilcoxon rank sum, two-sided, p > 0.9, FDR-corrected (n = 15 days per area for monkey L and 12 for monkey T)

Over the 12 rank-sum tests for stimulations and the 12 for shams, the smallest FDR-corrected p is 0.908 and 0.942, so all exceed 0.90 as reported (smallest uncorrected 0.509 and 0.187). Kept trials per day average 116 to 125 per type and area.

How the test differed from the paper’s method

• Benjamini-Hochberg correction from statsmodels in place of dsp3.fdr, which the claims file names; both implement the same procedure.

Fig. S1D, non-social control condition, all p > 0.10

Results, opening section, paragraph 2; Figure S1D legend

What the paper reports

• Wilcoxon rank sum, two-sided, p > 0.1, FDR-corrected

No FDR-corrected p is below 0.05 (smallest 0.108 for stimulations, 0.082 for shams). The paper's "all p > 0.10" does not hold for two sham tests between the monkeys (OFC and dmPFC, corrected 0.082, uncorrected 0.014 and 0.010); monkey T has more control trials per day (71 to 77) than monkey L (52 to 57).

The total number of fixations on partner's Eyes per day was significantly higher than fixations on non-eye Face for all three stimulated brain regions (Fig. S1E, top, all p < 10−4, Wilcoxon signed rank, two-sided, FDR-corrected)

Results, opening section, paragraph 2; Figure S1E

What the paper reports

• Wilcoxon signed rank, two-sided, p < 10⁻⁴, FDR-corrected (n = 27 days per area)

Under both day counts (a, total; b, mean per run) and both event sets (M1-only and mutual, M1-only), Eyes fixations exceed non-eye Face fixations for every area, corrected p from 1.1e-07 to 3.1e-05, all below 1e-4. Median day totals are 382 to 408 Eyes and 134 to 179 non-eye Face fixations.

How the test differed from the paper’s method

• The events counted are not fixed by the claims file. M1's fixations in the ROI including the M1 side of mutual events are the main set, following the Methods' "total number of gaze fixations ... of the stimulated monkey" (STAR Methods, Frequency of microstimulations and regions of interest); M1-only events are shown and give the same outcome.
• The non-eye Face is the files' m1_neface table, which includes the Mouth, with mutual_face and mutual_mouth for mutual events (mutual_face is the face outside the Eyes and the Mouth; the notebook shows that m1_wholeface equals m1_eye plus m1_neface and that mutual_wholeface equals mutual_eye plus mutual_face plus mutual_mouth).

Specifically, microstimulations of OFC led to more clustered subsequent gaze fixations around the partner monkey (Fig. 2A

Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 2; Figure 2A

The day-averaged OFC stimulation minus sham density summed over the whole Face squares is 7.47 (z-scored) and 0.0093 (by total), and over the Eyes squares 3.71 and 0.0046; it is positive in each monkey under each normalisation. dmPFC (whole Face -2.56 and -0.0027) and ACCg (-2.55 and -0.0048) are negative. 10,771 of 53,542 post-gaze fixations fall outside the 40-degree grid.

How the test differed from the paper’s method

• Fixations are the Methods reading (M1-only and mutual events). The grid is centred on each run's Eyes ROI centre, and the ROI squares are those whose centres lie in the ROI rectangle, relative to the Eyes centre, of every social run of the day, a standard choice; the three recalibrated days, which have two ROI sets, are dmPFC and ACCg days. Pixels are converted with 27 cm at 50 cm over 768 pixels (claims file, opening comment).

microstimulations of OFC significantly decreased the average distance of one's own gaze positions in space relative to partner's Eyes during the post-gaze epoch, compared to shams (Fig. 2C, p < 0.001, Wilcoxon signed rank, two-sided

Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 3; Figure 2C legend

What the paper reports

• OFC, both monkeys combined: Wilcoxon signed rank, two-sided, p < 0.001 (n = 27 OFC days)
• OFC, monkey L: Wilcoxon signed rank, two-sided, p = 0.008 (n = 15 OFC days)
• OFC, monkey T: Wilcoxon signed rank, two-sided, p = 0.002 (n = 12 OFC days)

Methods reading, median stimulation minus sham day difference -17.9 pixels (-0.70 degrees), p = 0.00021 over 27 days; monkey L p = 0.0067 (paper 0.008), monkey T p = 0.016 (paper 0.002), both negative. Code reading, p = 0.00013, 0.0067 and 0.0034. 1,270 of 19,474 social trials have no fixation in the window and are left out of the day means.

How the test differed from the paper’s method

• The mutual events added in the Methods reading are the M1 side of mutual_wholeface, which holds every mutual event (mutual_eye, mutual_face and mutual_mouth are subsets of it, shown in the notebook); mutual events are defined only within the face, and m1_all holds no M1 side of a mutual event.
• Each run's own ROIs are used, since three days were recalibrated between runs (08142019, 12102019, 12132019).

By contrast, we did not observe such stimulation effect on social gaze distance for dmPFC (Fig. 2C, p = 0.361) or ACCg (Fig. 2C, p = 0.374)

Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 3; Figure 2C

What the paper reports

• dmPFC: Wilcoxon signed rank, two-sided, p = 0.361 (n = 27 dmPFC days)
• ACCg: Wilcoxon signed rank, two-sided, p = 0.374 (n = 27 ACCg days)

dmPFC median day difference -6.5 pixels, p = 0.21; ACCg -4.9 pixels, p = 0.39 (paper 0.361 and 0.374). The code reading gives 0.29 and 0.29.

Notably, the observed stimulation effect of OFC was more pronounced for gaze fixations in the contralateral visual field of the stimulated brain hemisphere (Fig. 2D–E; contralateral: p < 0.001; significant statistics for individual animals are reported in the figure legend; ipsilateral: p = 0.068; contralateral vs. ipsilateral: p = 0.026, Wilcoxon signed rank, two-sided)

Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 3; Figure 2E legend

What the paper reports

• contralateral, stimulation against sham, both monkeys: Wilcoxon signed rank, two-sided, p < 0.001 (n = 27 OFC days)
• contralateral, monkey L: Wilcoxon signed rank, two-sided, p = 0.015
• contralateral, monkey T: Wilcoxon signed rank, two-sided, p = 0.012
• ipsilateral, stimulation against sham, both monkeys: Wilcoxon signed rank, two-sided, p = 0.068
• contralateral against ipsilateral effect, both monkeys: Wilcoxon signed rank, two-sided, p = 0.026
• contralateral against ipsilateral effect, monkey L: Wilcoxon signed rank, two-sided, p = 0.073
• contralateral against ipsilateral effect, monkey T: Wilcoxon signed rank, two-sided, p = 0.204

Contralateral median effect -21.2 pixels, p = 0.00048; ipsilateral -12.3 pixels, p = 0.058; contralateral against ipsilateral median -17.7 pixels, p = 0.046 (paper below 0.001, 0.068, 0.026). Per monkey, contralateral p = 0.018 (L) and 0.012 (T) (paper 0.015, 0.012); hemifield difference p = 0.073 (L) and 0.30 (T) (paper 0.073, 0.204). The code reading gives 0.00038, 0.055 and 0.026. All 27 days have fixations in both hemifields.

How the test differed from the paper’s method

• The hemifield is computed from the centre of `roi_m2_face` as the claims file says; the files' own laterality_label agrees for 99.98% of M1 events.

Again, no such effect was observed in either hemifield for dmPFC (Fig. 2E, contralateral: p = 0.501; ipsilateral: p = 0.149; contralateral vs. ipsilateral: p = 0.230) or ACCg (Fig. 2E, p = 0.442; p = 0.517; p = 0.564)

Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 3; Figure 2E

What the paper reports

• dmPFC contralateral: Wilcoxon signed rank, two-sided, p = 0.501
• dmPFC ipsilateral: Wilcoxon signed rank, two-sided, p = 0.149
• dmPFC contralateral against ipsilateral: Wilcoxon signed rank, two-sided, p = 0.23
• ACCg contralateral: Wilcoxon signed rank, two-sided, p = 0.442
• ACCg ipsilateral: Wilcoxon signed rank, two-sided, p = 0.517
• ACCg contralateral against ipsilateral: Wilcoxon signed rank, two-sided, p = 0.564

dmPFC contralateral p = 0.48, ipsilateral 0.10, difference 0.22; ACCg 0.53, 0.53, 0.82 (paper 0.501, 0.149, 0.230 and 0.442, 0.517, 0.564). The code reading gives no p below 0.05 either (smallest 0.11).

Crucially, these stimulation effects of OFC were exclusively observed in the live social gaze condition ... and not in the non-social gaze control condition using the RDM stimulus with no behavioral meaning to monkeys (Fig. 2F, OFC

Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 4; Figure 2F

What the paper reports

• OFC: Wilcoxon signed rank, two-sided, p = 0.118
• dmPFC: Wilcoxon signed rank, two-sided, p = 0.719
• ACCg: Wilcoxon signed rank, two-sided, p = 0.302

OFC median day difference -13.4 pixels, p = 0.22; dmPFC 1.7 pixels, p = 0.64; ACCg -5.7 pixels, p = 0.50 (paper 0.118, 0.719, 0.302). By monkey, dmPFC in monkey L gives p = 0.048 (distance larger after stimulation), outside the claim's test.

How the test differed from the paper’s method

• The RDM centre is the centre of each Dot run's `roi_m1_eyes`; the Dot runs hold M1 events only, so the fixations are m1_all.

The absence of stimulation effect for the RDM stimulus was also found when gaze fixation locations were split by hemifield for OFC (Fig. 2G–H; contralateral: p = 0.097; ipsilateral: p = 0.249; contralateral vs. ipsilateral: p = 0.442), dmPFC (p = 0.374; p = 0.517; p = 0.773) or ACCg (p = 0.532; p = 0.943; p = 0.171)

Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 4; Figure 2G-H

What the paper reports

• OFC contralateral: Wilcoxon signed rank, two-sided, p = 0.097
• OFC ipsilateral: Wilcoxon signed rank, two-sided, p = 0.249
• OFC contralateral against ipsilateral: Wilcoxon signed rank, two-sided, p = 0.442
• dmPFC contralateral: Wilcoxon signed rank, two-sided, p = 0.374
• dmPFC ipsilateral: Wilcoxon signed rank, two-sided, p = 0.517
• dmPFC contralateral against ipsilateral: Wilcoxon signed rank, two-sided, p = 0.773
• ACCg contralateral: Wilcoxon signed rank, two-sided, p = 0.532
• ACCg ipsilateral: Wilcoxon signed rank, two-sided, p = 0.943
• ACCg contralateral against ipsilateral: Wilcoxon signed rank, two-sided, p = 0.171

None of the nine tests over both monkeys gives p below 0.05; the smallest is OFC contralateral, 0.16 (paper 0.097). By monkey, dmPFC contralateral in monkey L gives p = 0.035, outside the claim's tests.

How the test differed from the paper’s method

• The Dot runs do not hold `roi_m2_face`; the hemifield reference is taken from the same day's social runs, where it never varies within a day.

In fact, the effect was present and lasted longer beyond 1.5 sec in one of the stimulated monkeys (within 2 sec after trial onset

STAR Methods, Gaze distance analyses, paragraph 1

What the paper reports

• within 2 s, both monkeys: Wilcoxon signed rank, two-sided, p = 0.003
• within 2 s, monkey L: Wilcoxon signed rank, two-sided, p = 0.008
• within 2 s, monkey T: Wilcoxon signed rank, two-sided, p = 0.204
• within 3 s, both monkeys: Wilcoxon signed rank, two-sided, p = 0.004
• within 3 s, monkey L: Wilcoxon signed rank, two-sided, p = 0.007
• within 3 s, monkey T: Wilcoxon signed rank, two-sided, p = 0.204

Within 2 s, both monkeys p = 0.0030, L 0.012, T 0.23; within 3 s, 0.0032, 0.0067 and 0.20 (paper 0.003, 0.008, 0.204 and 0.004, 0.007, 0.204), stimulation below sham in each. The code reading gives the same pattern (2 s 0.0022, 0.010, 0.18; 3 s 0.0025, 0.0067, 0.15).

Following OFC microstimulations, unlike more clustered subsequent gaze fixations around another social agent, we did not observe any change in the total number of fixations within partner's Eyes (within 1.5 sec

STAR Methods, Gaze distance analyses, paragraph 2

What the paper reports

• Eyes, 1.5 s, each monkey: Wilcoxon signed rank, two-sided, p > 0.18
• Eyes, 2 s: Wilcoxon signed rank, two-sided, p > 0.24
• Eyes, 3 s: Wilcoxon signed rank, two-sided, p > 0.3
• whole Face, 1.5 s: Wilcoxon signed rank, two-sided, p > 0.12
• whole Face, 2 s: Wilcoxon signed rank, two-sided, p > 0.2
• whole Face, 3 s: Wilcoxon signed rank, two-sided, p > 0.22

Under the code's reading (a), none of the 12 per-monkey tests (Eyes and whole Face, 1.5, 2 and 3 s) gives p below 0.05; the smallest is 0.12. Both monkeys combined, outside the claim, give p = 0.025 for whole Face fixations within 1.5 s (0.044 more per trial after stimulation). Under reading (b), M1-only and mutual events, monkey L gives p = 0.026 for the whole Face within 1.5 s, 1 of 12 tests.

On the day level, none of the three brain regions had any microstimulation effect on social gaze distance (OFC

STAR Methods, Gaze distance analyses (Mouth stimulation); Results, Closed-loop microstimulations of OFC facilitate momentary social attention in the spatial dimension, paragraph 4

What the paper reports

• social, OFC: Wilcoxon signed rank, two-sided, p = 1 (n = 5 days from monkey T)
• social, dmPFC: Wilcoxon signed rank, two-sided, p = 1
• social, ACCg: Wilcoxon signed rank, two-sided, p = 0.125
• social, each hemifield, all areas: Wilcoxon signed rank, two-sided, p > 0.18
• non-social, OFC: Wilcoxon signed rank, two-sided, p = 1
• non-social, dmPFC: Wilcoxon signed rank, two-sided, p = 0.188
• non-social, ACCg: Wilcoxon signed rank, two-sided, p = 0.188
• non-social, each hemifield, all areas: Wilcoxon signed rank, two-sided, p > 0.06

None of the 27 tests on the Mouth days gives p below 0.05. Social OFC, dmPFC and ACCg give p = 1, 1 and 0.19 (paper 1, 1, 0.125), non-social 1, 0.125 and 0.19 (paper 1, 0.188, 0.188); the hemifield tests give at least 0.0625. With 5 days per area the smallest two-sided exact signed-rank p is 0.0625, so the paper's test cannot find an effect at 0.05 on these days.

How the test differed from the paper’s method

• The Mouth centre is the centre of each run's `roi_m1_mouth`, and the non-social centre that of the Dot run's `roi_m1_mouth` (the RDM at the Mouth location); the files do not say whether the Mouth rectangle is the one used online.

Microstimulations of OFC decreased this inter-looking interval (Fig. 3B, p = 0.035, Wilcoxon signed rank, two-sided; see Fig. S3A for an alternative visualization). However, we did not observe such stimulation effect for dmPFC (Fig. 3B, p = 0.792) or ACCg (Fig. 3B, p = 0.291)

Results, Microstimulations of OFC also promote momentary social attention in the temporal dimension, Inter-looking interval, paragraph 1; Figure 3B legend

What the paper reports

• OFC, both monkeys: Wilcoxon signed rank, two-sided, p = 0.035 (n = 27 OFC days)
• OFC, monkey L: Wilcoxon signed rank, two-sided, p = 0.188
• OFC, monkey T: Wilcoxon signed rank, two-sided, p = 0.092
• dmPFC: Wilcoxon signed rank, two-sided, p = 0.792
• ACCg: Wilcoxon signed rank, two-sided, p = 0.291

OFC median day difference -0.069 s, p = 0.044 (paper 0.035); dmPFC p = 0.77, ACCg 0.34 (paper 0.792, 0.291); monkey L 0.25 and T 0.11 (paper 0.188, 0.092). The outcome is near the threshold and depends on two details. 91 trials have a whole Face event that starts at the trial sample itself; the criterion's "starts after it" leaves them out, and counting them as latency 0 gives OFC p = 0.058. M1-only events give OFC p = 0.0022 (T 0.0024).

How the test differed from the paper’s method

• An event "after" the trial must start later than the trial's own sample (criterion, "starts after it"); 91 events that start on the trial sample, likely the triggering fixation as the offline fixation detection saw it, are not counted. Whether the paper's code counts them is not known; this is asked of the paper's authors in the findings.

Further, this reduction of inter-looking interval from OFC microstimulations was specific to social attention as no effect was observed in the non-social gaze condition

Results, Microstimulations of OFC also promote momentary social attention in the temporal dimension, Inter-looking interval, paragraph 1

What the paper reports

• OFC: Wilcoxon signed rank, two-sided, p = 0.773
• dmPFC: Wilcoxon signed rank, two-sided, p = 0.08
• ACCg: Wilcoxon signed rank, two-sided, p = 0.943

OFC p = 0.70, dmPFC 0.13, ACCg 0.84 (paper 0.773, 0.080, 0.943). By monkey, dmPFC in monkey T gives p = 0.034, outside the claim's test. 7,736 of 10,148 control trials have an interval.

How the test differed from the paper’s method

• The whole Face in the Dot runs is the files' m1_wholeface table of those runs.

the stimulated monkey did not look back at the partner's whole Face during the examined time window on 41% of microstimulation and sham trials per day on average

Results, Microstimulations of OFC also promote momentary social attention in the temporal dimension, Inter-looking interval, paragraph 1

Reading (a), over the 81 Eyes days, 41.3% (expected 41 plus or minus 4.1); reading (b), over the 27 OFC days, 39.9%. Pooled over trials, 40.5%.

Nevertheless, when we combined all trials for each stimulated region, we still observed similar results to the day-level analysis above (Fig. S3B, OFC

Results, Microstimulations of OFC also promote momentary social attention in the temporal dimension, Inter-looking interval, paragraph 1; Figure S3B legend

What the paper reports

• OFC, both monkeys: Wilcoxon rank sum, two-sided, p = 0.01
• OFC, monkey L: Wilcoxon rank sum, two-sided, p = 0.026
• OFC, monkey T: Wilcoxon rank sum, two-sided, p = 0.143
• dmPFC: Wilcoxon rank sum, two-sided, p = 0.301
• ACCg: Wilcoxon rank sum, two-sided, p = 0.602

OFC median 1.19 s after stimulation against 1.36 s after sham, p = 0.017 (paper 0.010); dmPFC p = 0.39, ACCg 0.62 (paper 0.301, 0.602); OFC monkey L 0.049 and T 0.14 (paper 0.026, 0.143).

On the day level, microstimulations did not seem to greatly reduce such reciprocation latency (Fig. S3C–D, OFC

Results, Reciprocation latency, paragraph 1; Figure S3C legend

What the paper reports

• OFC: Wilcoxon signed rank, two-sided, p = 0.13
• dmPFC: Wilcoxon signed rank, two-sided, p = 0.701
• ACCg: Wilcoxon signed rank, two-sided, p = 0.4

OFC p = 0.11, dmPFC 0.56, ACCg 0.32 (paper 0.130, 0.701, 0.400). Trials with no M2-then-M1 sequence are 86.4% per day over the 81 days (reading a; expected 86 plus or minus 8.6) and 85.1% over the OFC days. 2,689 of 19,474 trials have a latency; every day has one in each trial type.

How the test differed from the paper’s method

• A mutual event enters both lists, at M1's and at M2's own start times; ties keep the M1 event first, a standard choice.

When combining all trials for each stimulated region, we observed that microstimulations of OFC decreased reciprocation latency (Fig. 3D, p = 0.011, Wilcoxon rank sum, two-sided). We again did not observe such stimulation effect for dmPFC (Fig. 3D, p = 0.777) or ACCg (Fig. 3D, p = 0.368)

Results, Reciprocation latency, paragraph 1; Figure 3D legend

What the paper reports

• OFC, both monkeys: Wilcoxon rank sum, two-sided, p = 0.011
• OFC, monkey L: Wilcoxon rank sum, two-sided, p = 0.074
• OFC, monkey T: Wilcoxon rank sum, two-sided, p = 0.079
• dmPFC: Wilcoxon rank sum, two-sided, p = 0.777
• ACCg: Wilcoxon rank sum, two-sided, p = 0.368

OFC median 0.84 s after stimulation against 1.01 s after sham, p = 0.023 (paper 0.011); dmPFC p = 0.72, ACCg 0.36 (paper 0.777, 0.368); OFC monkey L 0.15 and T 0.078 (paper 0.074, 0.079).

76 out of 81 days had at least 90 stimulations. The 5 days excluded from further analysis were 1 day from monkey L with OFC microstimulations, 1 day from monkey T OFC, 2 days from monkey T dmPFC, and 1 day from monkey T ACCg.

STAR Methods, Inter-individual gaze dynamics analyses, paragraph 2

Counting kept stimulations (reading a), 76 of 81 days have at least 90, and the five others are the days the code names, with the stated composition, 08122019 (L, OFC, 87), 12222021 (T, OFC, 80), 11052021 and 12172021 (T, dmPFC, 64 and 67) and 11042021 (T, ACCg, 50). Counting all stimulations delivered in social runs (reading b) gives 77; the L OFC day has 96.

No difference was observed for pupil size between the two trial types throughout the 700-msec period following trial onset for any of the three stimulated regions.

Results, Microstimulation effects of OFC and dmPFC are not driven by low-level properties of saccades, paragraph 1; Figure 5B legend

What the paper reports

• Wilcoxon signed rank, two-sided, per 10 ms bin, p > 0.05

No 10-ms bin of the 70 gives p below 0.05 in any area; the smallest p are 0.31 (OFC), 0.51 (dmPFC) and 0.11 (ACCg). 2,198 of 19,474 social trials keep M1's gaze in the whole Face for 700 ms; all 27 days of each area have both trial types.

How the test differed from the paper’s method

• Pupil size is the raw file value (a.u.), averaged within 10-ms bins, without baseline correction, as the Methods describe none.

macrosaccade peak velocity (Fig. 5E, OFC

Results, Microstimulation effects of OFC and dmPFC are not driven by low-level properties of saccades, paragraph 1; Figure 5E-H

What the paper reports

• peak velocity, OFC: Wilcoxon signed rank, two-sided, p = 0.848
• peak velocity, dmPFC: Wilcoxon signed rank, two-sided, p = 1
• slope difference against zero, OFC: Wilcoxon signed rank, two-sided, p = 0.665
• slope difference against zero, dmPFC: Wilcoxon signed rank, two-sided, p = 0.904
• slope difference, OFC: permutation test, p = 0.515
• slope difference, dmPFC: permutation test, p = 0.164
• by direction, OFC: Wilcoxon signed rank, two-sided, p > 0.47
• by direction, dmPFC: Wilcoxon signed rank, two-sided, p > 0.31

With reading (a) of the macrosaccade, none of the 14 OFC and dmPFC tests gives p below 0.05. Peak velocity of the first macrosaccade, OFC p = 0.052 and dmPFC 0.88 (paper 0.848, 1); slope difference against zero 0.70 and 0.59 (paper 0.665, 0.904); permutation 0.28 and 0.19 (paper 0.515, 0.164); by direction group the smallest p is 0.071 (OFC CI; paper all above 0.47 and 0.31). With the paper's code's labelling, reading (b), 3 of the 14 give p below 0.05 (dmPFC slope difference 0.030, OFC IC 0.048, dmPFC CC 0.037), so the result depends on how microsaccades are told from macrosaccades. 15,961 of 19,474 trials have a first macrosaccade under (a), 11,354 under (b). The outcome rests on the same assumption as fig5cd-saccade-counts-unchanged, that this detection on the files' unfiltered gaze stands in for the paper's.

How the test differed from the paper’s method

• Assumption, as for fig5cd-saccade-counts-unchanged. The detection on the files' unfiltered gaze, as the paper's code passes it, is taken to stand in for the paper's, whose detected saccades are not described; the detector's output on these data is not physiological (median duration 147 ms), and the labelling of micro- and macrosaccades changes the count of failing tests (0 of 14 under reading a, 3 of 14 under b).
• Saccades as in fig5cd-saccade-counts-unchanged (a Python port of the Otero-Millan et al. 2014 detector, with_inputs). Slopes are least-squares lines over at least 3 saccades per trial type and day; the permutation seed 5 is a standard choice. Hemifields of the saccade's start and end are taken against the centre of roi_m2_face, as for the fixations.

Each data collection day consisted of a total of alternating 10 live social gaze sessions and 5 non-social control sessions on average for monkey L (9-11 social and 4-8 control sessions across all days) and alternating 15 live social gaze sessions and 5 non-social control sessions on average for monkey T (14-15 social and 5 control sessions across all days). Each session lasted 300 sec.

STAR Methods, Experimental setup, paragraph 2

Monkey L has 10.0 social runs per day (9 to 11) and 6.02 control runs (4 to 8); monkey T has 14 social runs on every one of its 36 Eyes days and 5 control runs. Runs hold 295.8 to 297.3 s of valid samples (mean 296.4). All ranges hold, and three of four means; monkey L's control mean of 6.02 is outside 5 plus or minus 0.5, as 14 of its 45 days have 8 control runs. Monkey T's mean of 14 is within the tolerance of 15 but no Eyes day has 15 social runs (the Mouth days have 15).

How the test differed from the paper’s method

• A session of the paper is taken to be one file (one run); the valid run length is the span of non-NaN samples of `t`.

With OFC microstimulations, the monkeys spent more time looking within the partner's whole Face (p = 0.003, stim minus sham difference = 48.9 msec for both monkeys combined; p = 0.008, diff = 66.6 msec for monkey L and p = 0.129, diff = 39.5 msec for monkey T)

STAR Methods, Gaze distance analyses, paragraph 2

What the paper reports

• OFC whole Face, both monkeys: Wilcoxon signed rank, two-sided, p = 0.003
• OFC whole Face, monkey L: Wilcoxon signed rank, two-sided, p = 0.008
• OFC whole Face, monkey T: Wilcoxon signed rank, two-sided, p = 0.129
• dmPFC whole Face, both monkeys: Wilcoxon signed rank, two-sided, p = 0.683
• ACCg whole Face, both monkeys: Wilcoxon signed rank, two-sided, p = 0.848
• OFC Eyes, both monkeys: Wilcoxon signed rank, two-sided, p = 0.044

The tests hold with the code's M1-only events; OFC whole Face p = 0.0025 with stimulation longer (L 0.0067, T 0.15), OFC Eyes p = 0.039, dmPFC 0.80, ACCg 0.93 (paper 0.003, 0.008, 0.129, 0.044, 0.683, 0.848). Under reading (a), the median of day differences, OFC whole Face is 52.8 ms (expected 48.9 plus or minus 4.89) and monkey L 65.1 ms (66.6 plus or minus 6.66), within tolerance, monkey T 46.2 ms (39.5 plus or minus 3.95) is not; reading (b), the mean, gives 62.6, 85.3 and 34.3 ms, none within. dmPFC -14.7 ms matches -14.5; ACCg -2.2 ms against 3.4 and OFC Eyes 12.2 ms against 16.5 do not. With M1-only and mutual events the OFC whole Face test still holds (p = 0.030) but OFC Eyes does not (p = 0.19).

Therefore, we did not find any evidence that the social gaze behaviors of the unstimulated monkey were altered on a short timescale during social gaze interactions.

Results, Microstimulations do not change momentary social attention in the unstimulated animal, paragraph 1

What the paper reports

• M2 social gaze distance, OFC: Wilcoxon signed rank, two-sided, p = 0.118
• M2 social gaze distance, dmPFC: Wilcoxon signed rank, two-sided, p = 0.302
• M2 social gaze distance, ACCg: Wilcoxon signed rank, two-sided, p = 0.064
• M2 inter-looking interval, OFC: Wilcoxon signed rank, two-sided, p = 1
• M2 inter-looking interval, dmPFC: Wilcoxon signed rank, two-sided, p = 0.866
• M2 inter-looking interval, ACCg: Wilcoxon signed rank, two-sided, p = 0.249
• M2 reciprocation latency, OFC: Wilcoxon signed rank, two-sided, p = 0.665
• M2 reciprocation latency, dmPFC: Wilcoxon signed rank, two-sided, p = 0.773
• M2 reciprocation latency, ACCg: Wilcoxon signed rank, two-sided, p = 0.701

M2 social gaze distance, OFC p = 0.12, dmPFC 0.47, ACCg 0.062 (paper 0.118, 0.302, 0.064); the M2-only reading gives 0.11, 0.37 and 0.069. M2 inter-looking interval, readings (a) and (b), smallest p 0.19 (paper 1, 0.866, 0.249). M2 reciprocation latency, reading (b) from the trial time, smallest p 0.46; reading (a), from the start of the M1 look (the mirror of M1's measure), OFC 0.56, dmPFC 0.044 (median 0.11 s longer after stimulation), ACCg 0.99 (paper 0.665, 0.773, 0.701). The criterion asks that no test give p below 0.05 under every reading; 1 of 18 tests (6 distance tests and 3 for each of the four other measures) does, so the claim holds for the distance and the interval and in part for the reciprocation latency.

How the test differed from the paper’s method

• M2's distance is to the centre of `roi_m2_eyes` (M1's Eyes in M2's coordinates); the Methods reading adds the M2 side of mutual whole Face events to m2_all.
• Reading (a) of the M2 inter-looking interval, start of the second minus start of the first M2 whole Face event in the window, follows the Results' "the latency between the second and first whole Face events of M2"; reading (b), from the end of the first, is shown.
• The 188 M2 events whose gaze samples are all NaN have no position and are left out of the M2 distance.

Importantly, the observed effects of OFC and dmPFC microstimulations were not driven by any change in the duration of current looking to partner's Eyes that triggered a microstimulation or sham (Fig. 5A, OFC

Results, Microstimulation effects of OFC and dmPFC are not driven by low-level properties of saccades, paragraph 1; Figure 5A

What the paper reports

• OFC: Wilcoxon signed rank, two-sided, p = 0.302
• dmPFC: Wilcoxon signed rank, two-sided, p = 0.269

Under reading (a), M1-only and mutual Eyes events, the event in progress at the trial is longer on stimulation trials for dmPFC (median day difference 31 ms, p = 0.019; paper 0.269), not for OFC (8 ms, p = 0.61; paper 0.302). Reading (b), M1-only events, gives dmPFC 28 ms, p = 0.028 and OFC p = 0.36. The claim holds for OFC and not for dmPFC. ACCg, outside the claim, gives 0.30 and 0.13. 13,779 of 19,474 trials fall inside an offline Eyes event under (a), 12,647 under (b); the 5,695 others under (a) are spread evenly over areas and trial types (27.3% to 30.1%; stimulation minus sham per day, dmPFC p = 0.069, OFC 0.58), 4,227 of them fall inside another M1 fixation (2,036 outside the face, 1,600 on the non-eye face), and 609 have an Eyes event starting within 20 ms after the trial. Counting those as the trigger gives dmPFC p = 0.020 and OFC 0.58.

How the test differed from the paper’s method

• The triggering fixation is taken as the offline (EyeMMV) Eyes event that contains the trial time; the online fixation that triggered the trial is not in the files.

number of microsaccades (Fig. 5C, OFC

Results, Microstimulation effects of OFC and dmPFC are not driven by low-level properties of saccades, paragraph 1; Figure 5C-D

What the paper reports

• microsaccades, OFC: Wilcoxon signed rank, two-sided, p = 0.456
• microsaccades, dmPFC: Wilcoxon signed rank, two-sided, p = 0.581
• macrosaccades, OFC: Wilcoxon signed rank, two-sided, p = 0.055
• macrosaccades, dmPFC: Wilcoxon signed rank, two-sided, p = 0.23

With the claims file's reading (a), a microsaccade lying strictly within an M1 fixation, microsaccade counts do not change for OFC (p = 0.069; paper 0.456) or dmPFC (0.89; paper 0.581) and macrosaccade counts do not change for dmPFC (0.24; paper 0.230), but OFC stimulation trials have fewer macrosaccades (median day difference -0.069 per post-gaze epoch, p = 0.044; paper 0.055). One of the four tests fails. With the paper's code's labelling, reading (b), none does (smallest 0.15). ACCg, outside the claim, gives p = 0.009 (micro) and 0.004 (macro) under (a). The detector finds 454,642 saccades in the 1,179 social runs (386 per run), median duration 147 ms (95th percentile 1,062 ms), median amplitude 13.7 degrees; in one run the gaze velocity inside fixations has a median of 19.7 deg/s and is under the detector's 5 deg/s saccade limit for 5.4% of samples, so many detected saccades span several fixations. In 202 of the 1,179 runs the MATLAB code's merge of overlapping saccades would not pair starts and ends. The port was not run against the MATLAB code, which is not available here. On synthetic gaze with little noise it finds all 93 saccades with their durations (median 38 ms detected and true); with white noise that gives the data's velocity inside fixations (median 19.8 deg/s, 4.3% under 5 deg/s) it finds 86.7% and doubles their median duration (72.5 against 37.5 ms; 95th percentile 172 ms). The data's durations are longer still, so their noise is not only white. The outcome rests on the assumption stated in the deviations, and the micro/macro labelling changes it (reading b gives 0 of 4).

How the test differed from the paper’s method

• Assumption. The detection here is taken to stand in for the paper's. The paper names the software (Key resources table, Otero-Millan et al. 2014) and says it "captured both canonical microsaccades ... and macrosaccades" (STAR Methods, Control analyses), but not whether the gaze was filtered first, and it reports no property of the detected saccades. The paper's code passes M1's gaze unfiltered (script/detect_microsaccades.m at cc0bf994), so unfiltered gaze is the best-supported choice, and the files' gaze is taken to be the gaze the paper passed. On these data the detector's output is not physiological (median duration 147 ms, 95th percentile 1,062 ms, velocity inside fixations under the 5 deg/s limit for 5.4% of samples in one run), which a filter would change; whether the paper's saccades looked the same is asked of the paper's authors in the findings.
• The dataset holds no saccade events. Saccades are detected with a Python port of the Otero-Millan et al. (2014) detector, from the authors' MATLAB code at a pinned commit (inputs), with that code's defaults, called as the paper's code calls it (script/run_saccade_detection.m and script/detect_microsaccades.m at cc0bf994; M1's gaze in degrees from the first valid time sample, passed as both eyes, NaN samples as blinks). The claim is with_inputs for that software, as the claims file says.
• The port uses scipy's k-means (Lloyd iterations from the same starting centres) in place of MATLAB's kmeans, and merges overlapping saccades as a union in the 202 runs where the MATLAB code would stop with an error.
• The fixations that define a microsaccade are the files' M1 fixations (M1-only and mutual), the EyeMMV events the claims file names; the paper's code detects its own fixations (is_fixation with 20, 10 and 0.03), and its labelling, reading (b), counts a saccade as a microsaccade when its start is at least 50 samples after a fixation's start and before its end (script/label_saccades.m).

the total number of fixations on partner's Eyes per day was comparable to fixations on the RDM stimulus for days involving the three stimulated regions (Fig. S1E, top, all p > 0.30)

Results, opening section, paragraph 2; Figure S1E

What the paper reports

• Wilcoxon signed rank, two-sided, p > 0.3, FDR-corrected

Under reading (a), the day's total, which the criterion judges, Eyes fixations exceed RDM fixations for every area (corrected p 0.00036, 0.000025 and 0.00041 for OFC, dmPFC and ACCg; median differences 177, 202 and 241 per day). A day holds about twice as many social runs as control runs (L 10 and 6, T 14 and 5), so day totals differ. Under the code's reading (b), the mean per run, the counts are comparable (corrected p 0.40, 0.40 and 0.52), as the paper reports; M1-only events give the same pattern under both readings. Whether the claims file's deciding reading (a) misreads the paper, given about twice as many social runs as control runs per day, is asked of a reader in questions/000959.yaml; the outcome is judged on the claims file as it stands.

How the test differed from the paper’s method

• The RDM fixations are the M1 events labelled Eyes in the Dot runs, whose Eyes ROI is the RDM location ("same location and size as Eyes ROI", STAR Methods, Closed-loop microstimulation protocol).

For fixations in the contralateral hemifield, microstimulation effect (stimulation minus sham) of OFC led to significantly smaller social gaze distance in Eyes stimulation compared to Month stimulation (p = 0.009), while no effect was observed for fixations in the ipsilateral hemifield (p = 0.724) or when both hemifields combined (p = 0.799)

STAR Methods, Gaze distance analyses (Mouth stimulation)

What the paper reports

• contralateral, all Eyes days: bootstrap, 1,000 resamples of 5 days with replacement, p = 0.009
• ipsilateral, all Eyes days: bootstrap, p = 0.724
• both hemifields, all Eyes days: bootstrap, p = 0.799
• contralateral, monkey T Eyes days: bootstrap, p = 0.005
• ipsilateral, monkey T Eyes days: bootstrap, p = 0.654
• both hemifields, monkey T Eyes days: bootstrap, p = 0.51

With all 27 OFC Eyes days, contralateral p = 0.078 (a, one-sided) and 0.156 (b, two-sided), 0.059 to 0.070 one-sided over five other seeds (paper 0.009); ipsilateral 0.30 and 0.59 (paper 0.724); both hemifields 0.043 and 0.086 (paper 0.799). With monkey T's 12 days, contralateral 0.037 and 0.074 (paper 0.005), ipsilateral 0.35 and 0.70, both hemifields 0.037 and 0.074 (paper 0.510). The Mouth medians are -6.3 (contralateral), -8.4 (ipsilateral) and -0.24 pixels (both), the Eyes medians -21.2, -12.3 and -17.9. Neither reading gives significance for the contralateral fixations alone; under (a) both hemifields combined is the comparison that is significant, the reverse of the paper's pattern. With the code's M1-only fixations, contralateral p = 0.049 (a) with all days (0.040 to 0.049 over five other seeds) and 0.006 with monkey T's days, ipsilateral 0.30 and 0.35, but both hemifields 0.050 and 0.036, so that fixation set does not give the paper's pattern either.

How the test differed from the paper’s method

• Seed 959 is a standard choice; five other seeds are shown.

We did not observe stimulation effect on the magnitudes of relative causal strength (OFC ... dmPFC ... ACCg

Results, Do microstimulations of OFC also lead to longer timescale modulation of social gaze exchanges?, paragraph 1; Figure S4A legend

What the paper reports

• OFC: Wilcoxon signed rank, two-sided, p = 0.239
• dmPFC: Wilcoxon signed rank, two-sided, p = 0.962
• ACCg: Wilcoxon signed rank, two-sided, p = 0.361

The relative causal strength needs the causal decomposition of Yang, Peng and Huang (2018), whose code runs 2 + 2 x (number of IMFs) ensemble EMDs per trial, each with 10 fixed sifts per IMF. Neither the noise level nor the ensemble size has a default in that code, and its example uses 1,000 ensembles, as does the claims file's reading (a), so both readings cost the same. In the run recorded here, one EEMD of one 150-bin series with 1,000 ensembles takes 15.9 s on one core with PyEMD (6.8 s on four processes; the round-2 run gave 13.1 and 5.5 s, so the timing varies between runs) and yields 7 IMFs, so a trial takes about 255 CPU s, and the 19,474 social Eyes trials about 1,380 CPU hours (about 591 hours of wall time on four processes). A compiled implementation at 10 microseconds per cubic-spline envelope would still need about 104 CPU hours. The notebook limit is one hour. The trials, distances and epochs the analysis would use are all in the dataset.