Sleep With and Without a Glanded Aid: A Two-Night Crossover Sleep-Laboratory Study of 48 Adults Aboard GSV Considered Opinion, 1697–1699 CE

1. Introduction

Sleep is a familiar occasion for glanding. Most adults can secrete a sleep-promoting compound at will, and a large literature records the compounds that are reported to follow from it (Tessaly, 1671 CE). Its entries are descriptions of what users say they feel. They are not measurements of what the sleeping brain does. Whether a glanded night differs in structure from an unaided one, and whether the difference is carried into the following day, has not been tested under controlled conditions.

Three observations make the question worth asking. First, the timing of secretion governs how quickly its effects arrive, so onset is the natural place to look for an effect (Brindavel & Orsaine, 1684 CE). Second, sleep is an activity that Culture residents describe as chosen and valued for itself, and they are not uniformly willing to hand it to a compound (Halsgarde, 1676 CE). Third, an earlier diary study on Chiark Orbital found that Snap, the waking compound, is taken mostly in the first two hours after waking, so any cost of a glanded night would be met by a morning dose of it before it could be noticed. Self-report alone would then understate it.

We therefore ran a controlled crossover study of glanded and unaided sleep in the laboratory of GSV Considered Opinion. The study was carried out between late 1697 CE and mid-1699 CE, and this paper is written from the vantage of 1700 CE; every source we cite predates it. Our primary questions were whether a self-secreted sleep-promoting compound changes the minutes spent in slow-wave sleep, and whether it changes reaction time the next morning. Secondary questions concerned sleep onset, total sleep, REM sleep and wakefulness after onset.

2. Methods

Participants were invited by message from the faculty's list of adults who had agreed to be asked about studies, and nobody was approached outside it. Eligible participants were adults who had used a sleep-promoting secretion on at least three previous nights. Because every participant chose their own compound from their own glands, the study tests the glanded habit as practised and does not test any single compound. We refer throughout to a sleep-promoting secretion and do not name its constituents. Of 70 adults invited, 58 consented (82.9%), 12 declined or did not reply, and 6 of the 58 could not be scheduled, leaving 52 enrolled (89.7% of those who consented). Enrolled participants were assigned by a random list to one of two orders: glanded night first, or unaided night first. Twenty-six were assigned to each.

Each participant spent two nights in the laboratory, fourteen days apart. On the glanded night they secreted a sleep-promoting compound at lights-out and pressed a bedside marker at the moment of secretion. On the unaided night they secreted nothing. Lighting followed the schedule set by the Hub of Chiark Orbital for its residents (Chiark Orbital Hub Records, 1688 CE), a choice made because most participants were residents of, or long-term visitors to, Chiark Orbital and were habituated to its cycle. Nights were recorded by electroencephalography, eye-movement and chin recordings from surface electrodes, in 30-second epochs. Each night was scored by one rater for wake, three stages of non-REM sleep and REM sleep, and the recording was identified only by a code. A second rater scored 14 of the 96 recorded nights, chosen at random, to measure reliability.

Next-morning alertness was measured one hour after waking by a ten-minute reaction-time task (Yestrel & Marrowby, 1681 CE), reported as mean reaction time in milliseconds. In the earlier diary study on Chiark Orbital, 71.0% of Snap episodes fell within two hours of waking. We therefore asked participants to withhold Snap until the task was over, and checked this by self-report at the end of each morning.

Both primary outcomes were analysed as paired differences, glanded night minus unaided night, with a paired t test and a 95% confidence interval from the t distribution with 47 degrees of freedom. The two primary p values were adjusted by the Holm method, and so were the four secondary ones, as a separate family. Carryover and period effects were examined by the standard two-sample comparisons of subject sums and of differences between orders (Dalquist, 1648 CE). Age in centuries was regressed on each participant's difference in sleep onset to look for an age gradient. Standardised effects are given as the mean difference over the standard deviation of the differences (d_z). Inter-rater agreement was measured by Cohen's kappa over all double-scored epochs (Corvane, 1689 CE). Differences were computed from unrounded values, so a difference may vary from the difference of two rounded means by 0.1.

3. Results

Of the 52 participants enrolled, 48 (92.3%) completed both nights and form the analysis sample. Two withdrew after the first night and two lost a recording to equipment faults, two in each order, so 24 participants are analysed in each order. Participants were aged 26 to 290 years, with a median of 97 and a mean of 123.3. Twenty-five were under 100 and nine were 200 or older. Participants reported withholding Snap on all 96 mornings that were analysed.

Sleep onset was shorter, and slow-wave and total sleep longer, on the glanded night (Table 1). Mean sleep onset fell from 20.9 to 9.2 minutes, a difference of 11.7 minutes (95% CI 9.3 to 14.2, p < .001, d_z = 1.38), and 43 of the 48 participants fell asleep sooner, two later and three showed no difference. Slow-wave sleep, the first primary outcome, was 8.0 minutes longer (3.1 to 12.9, p = .002, Holm-adjusted p = .004, d_z = 0.48), and 37 of 48 participants had more. Total sleep time was 20.0 minutes longer (10.6 to 29.3, Holm-adjusted p < .001, d_z = 0.62).

Costs also appeared on the glanded night. REM sleep was 7.1 minutes shorter (95% CI 2.5 to 11.6, p = .003, Holm-adjusted p = .006, d_z = 0.45), and 33 of 48 participants had less. Wakefulness after onset was 1.9 minutes shorter, but the interval included zero (−4.7 to 1.0, p = .19), so we cannot say whether it changed.

Reaction time, the second primary outcome, also showed a cost of the glanded night. Mean reaction time the next morning was 272.1 ms after glanded nights and 265.6 ms after unaided nights, a slowing of 6.5 ms (95% CI 1.1 to 11.9, p = .019, Holm-adjusted p = .019, d_z = 0.35). Thirty of 48 participants were slower after the glanded night. The interval lies above zero, but its lower end (1.1 ms) is close to it.

We found no evidence of carryover or of a period effect on any outcome (all p > .10), though with 24 participants in each order such tests have little power. The change in sleep onset did not vary detectably with age: the slope of onset difference on age was 0.12 minutes per century (95% CI −3.48 to 3.72, p = .95). Across 13,160 double-scored epochs the two raters agreed on 11,580 (88.0%), giving a kappa of 0.83 (approximate 95% CI 0.82 to 0.84).

4. Discussion

In this sample a self-secreted sleep-promoting compound did what its users expect of it. Sleep began about twelve minutes sooner and lasted about twenty minutes longer, and part of the extra time went into slow-wave sleep, while REM sleep fell. The sums are close: twenty extra minutes of sleep, eight of them slow-wave and seven fewer in REM, leave about nineteen minutes for other stages, which we did not analyse. We do not claim to know where in the night the changes occurred.

Reaction time is the result that deserves care. A slowing of 6.5 ms is about 0.3 of the standard deviation between participants on the unaided night, and the lower end of its interval is 1.1 ms. It was found after participants had withheld Snap, so it is not a trivial effect of an unfinished morning dose. It is also a single laboratory task. We cannot say whether such a difference would be noticed in a game, a project or a conversation, and the reaction-time and REM results together are as compatible with a small residual effect of the compound as with a cost of reduced REM sleep. This study cannot separate the two.

The finding also bears on how Culture residents think about glanded sleep. Residents have described unaided sleep as an activity worth having for itself (Halsgarde, 1676 CE), and our participants chose the compound and the night. Our results give that preference some measured content: the aided night is longer and deeper, and the unaided night may leave a slightly quicker morning. Neither is a reason to prefer one over the other. Both are facts a resident may weigh, by choice, as they weigh others.

We make no claim about repeated use. One glanded night after a fourteen-day interval says little about a habit sustained for years, and our results should not be extended to it. The same applies to any particular compound: our participants used their own, and the analysis pools them.

5. Limitations

Only 48 participants were analysed, drawn from a faculty list of adults willing to be asked, and they are unlikely to represent the Culture as a whole. All were recorded in one laboratory, aboard one General Systems Vehicle. The glanded condition could not be blinded, since each participant knew whether they had secreted, and this may have affected reaction time through expectation. Scorers were blind to condition, but the second rater covered only 14 of 96 nights, and the kappa interval treats its epochs as independent, although they come from 14 nights; it is therefore too narrow.

Dose was not measured. Participants chose their own compound and timing, and the marker records only the time of secretion. A fourteen-day interval was judged long enough to avoid carryover, and the data show none, but the tests for it were underpowered. Withholding of Snap before the reaction-time task was reported by participants and was not verified independently. Two participants withdrew and two recordings were lost, and nothing in our data lets us say whether these four differed from the 48.

Holm adjustment was applied within the two primary outcomes and within the four secondary ones, and no protocol is reported for the six outcomes we analysed. We ran no correction for the age regression or the carryover tests. A single night of recording per condition, taken in an unfamiliar room, may also differ from sleep at home. We regard these findings as an estimate for one occasion, to be replicated before they are relied on.