The entry says that six wristwatches recorded a flight that nobody described while it ran. This piece sets out what the watches were, how far they can be trusted, and what they recorded for each man, with the numbers as the Penn group printed them in the Proceedings of the National Academy of Sciences in January 2013 and in the appendix to that paper. The crew are the letters the paper gave them, a to f, assigned at random; the paper says nothing about which letter is which man, and neither does this.

The instrument

The Actiwatch Spectrum, made by Philips Respironics, is a watch-sized box worn on the wrist of the non-dominant arm. It shows the time. Inside is a piezo-electric sensor that produces a voltage whenever the wrist’s acceleration changes. The voltage is sampled thirty-two times a second; the largest change in each second becomes that second’s activity value, and the sixty values of each minute are added into one count. A running baseline filters out constant accelerations such as gravity. Beside the sensor is a light meter, recording the average illuminance at the wrist in lux, also once a minute. Each watch was calibrated by the manufacturer so that counts could be compared between watches.

The six put the watches on at 13:46 on average on 3 June 2010, the earliest at 10:28 and the latest at 14:46, and took them off at 18:02 on average on 4 November 2011, between 17:57 and 18:08. Each man’s watch was exchanged twice during the mission before its battery ran down. Once a week, after the evening vigilance test, each man copied the week’s data to an SD card together with his test results, his questionnaire answers and the videos of his face; the cards were “jettisoned from the Mars 500 facility weekly”, retrieved, uploaded to a secure site in Philadelphia and “immediately checked by the programmed algorithms (Pulsar Informatics, Inc.) and by investigators at the University of Pennsylvania for data completeness and integrity.”

The first and last days were less than full days and were dropped. One hour per man was deleted to allow for the clocks going back in October 2010, and one added for their going forward in March 2011; in October 2011 Russia abolished the change, and the laptops, which had no internet connection and did not know, changed the time anyway and had to be corrected by hand. Allowing for all that, the group expected 4,485,336 minutes of record and got 4,396,333, which is 98.02 percent; the missing 1.98 percent, from downloads and equipment failures, ranged from 0.5 to 6.0 percent across the six. Of what was recorded, 63.8 percent was scored as active wakefulness, 31.0 percent as sleep and 5.3 percent as waking rest.

How far to trust it

The scoring is done by a proprietary algorithm in the manufacturer’s software, Actiware 5.59, which sorts each minute into active wake, sleep or waking rest. The group checked it by eye and corrected obvious misclassifications, less than 2.8 percent of the minutes (0.8 to 5.2 percent across the six). They also validated the algorithm separately, on twenty-two healthy adults of about the crew’s age living for fifteen days each in a sealed laboratory on a sixteen-hours-up, eight-hours-down schedule, with the sleep verified by polysomnography and the waking by continuous observation. Over 115 days of such comparison the algorithm identified wakefulness correctly 96.2 percent of the time and sleep 97.0 percent; its overall accuracy was 96.4 percent. Half its errors were confusions between waking rest and sleep, both of which are lying still. Within the eight hours in bed it overestimated sleep by 26.4 minutes a night, because a man lying awake looks, to a wrist, like a man asleep. The group therefore reports the crew’s mean sleep as 7.39 hours a day uncorrected and estimates it at 6.95 hours corrected, and uses the uncorrected figures throughout, since the correction has its own uncertainty of plus or minus eight minutes. Everything below is uncorrected.

The crew as a whole

The paper divides the mission into four quarters of 130 days. Across them, time in active wakefulness fell by 7.0 percent, from the first quarter to the fourth, which is 1.12 hours less active waking per man per day; the fall was steep in the first three months and gradual for the next thirteen. Sleep rose by 8.4 percent, or 0.59 hours a day: 7.12 hours in the first quarter, 7.23 in the second, 7.50 in the third, 7.71 in the fourth. Waking rest rose by 50.4 percent, or 0.54 hours a day. The intensity of movement during waking fell steeply in the first quarter and gradually for a year. The trends held on work days and rest days alike, and in the night (22:00 to 08:59) and the day (09:00 to 21:59) taken separately. Summing the sleep: 5,550.6 hours in the first quarter for the six together, 5,639.5 in the second, 5,850.3 in the third, 6,012.8 in the fourth; the second half of the mission held 673 hours more sleep than the first.

Two periods ran against the trend. In the first forty days the crew averaged less than 6.78 hours of sleep, lower than any later forty-day period, under the highest workload ratings of the mission. And in the eighty days up to and including the landing on Mars, mean sleep fell from 7.33 to 7.13 hours and the crew rated themselves more tired. Otherwise the line ran one way until the final twenty days, when active wakefulness and the intensity of movement rose sharply and sleep and rest fell, relative to the two sixty-day periods before; the controllers said there was no extra work, the workload ratings did not rise, and the explanation given was “psychological anticipation of mission end.”

The vigilance test ran alongside. Each man did the three-minute PVT-B twice on one day of each of the seventy-four weeks, 888 tests in all, with the laptop camera filming his face. Across the crew, response speed improved and lapses fell, especially in the second half, as the sleep rose. There was no cumulative deficit. One man’s errors are the exception, below.

The six, one by one

Crewmember a slept the most: 7.94 hours a day on average, rising from 7.57 in the first quarter to 8.61 in the fourth. From early in the mission his sleep was split, a night-time anchor and a daytime nap, and the nap grew: his daytime sleep rose 50.8 percent in the second half. He kept a 24-hour period, because the main sleep always came at night. He rated his sleep as good, and better as the mission went on. He slept more than ten hours on 12.2 percent of mission days; in the first month his sleep had begun 1 hour 59 minutes later than before the mission.

Crewmember b is the twenty-five-hour man. On two nights measured a fortnight before closure he fell asleep at 00:51 and 02:01 and napped both days. In the first week his sleep began at 02:30 on average, the latest of the six, then 02:59, 04:23 and 05:12 in the following weeks; after sixteen days his sleep was beginning after four in the morning. The spectral analysis of his sleep and waking, in ninety-day windows moved ten days at a time, shows a dominant period of 24.72 hours in the second quarter lengthening to 25.06 in the fourth and averaging 24.98 across the mission, with a smaller 24-hour peak from his attendance at breakfast between eight and ten. His sleep averaged 7.55 hours a day and he slept more than ten hours on 12.6 percent of mission days, including one sleep of about ten hours on day 501. For 1,959 hours, 15.7 percent of the mission, he was asleep when every other man was awake, or awake when every other man was asleep. The watch recorded the cause as well as the effect: he was awake under room light in the hours after midnight when light delays the clock, from the first thirty days on and periodically afterwards. In the post-mission interview he said he was “occasionally” prone to late sleep, napped to compensate, and “goes to bed very late”, especially on holidays; he did not consider it a disorder. The group’s reading is an evening chronotype, present before the mission, that the dim lighting could not hold to twenty-four hours.

Together, a and b were out of step with the rest of the crew for 2,498 hours, 20.1 percent of the mission. They were the only two whose sleep onset in the first month was two hours or more later than before the mission.

Crewmember c slept 7.25 hours a day, a monophasic night-time sleeper on a 24-hour cycle, with the fewest sleep-quality complaints after d and the lowest error counts after a and b. He showed, in the words of the 2014 paper, “neither behavioral disturbances nor reports of psychological distress.”

Crewmember d was the most active of the six while awake, by a factor of 1.85 over the rest, and also one of the best sleepers: 7.79 hours a day, with the most frequent ratings of good sleep and the lowest rest time. His vigilance errors were low. The paper offers him as the demonstration that “high wake activity levels and adequate sleep to maintain alertness are not incompatible in long-duration mission confinement.”

Crewmember e slept 7.26 hours a day, rising from 6.61 in the first quarter to 7.89 in the fourth, on a 24-hour cycle. His sleep-quality ratings were the second-worst in the crew, and he accounted for ten of the fifty-one test videos that showed sleepiness. His questionnaires, in the 2014 paper, are another story, told in The other instruments.

Crewmember f slept the least: 6.54 hours a day, 6.80 in the first quarter, 6.55, 6.33 and 6.48 in the next three, “due to a worsening sleep onset insomnia.” His sleep-quality ratings were the worst of the six and so were his ratings of stress, tiredness and workload. He made 828 vigilance errors across the mission, lapses and false starts together, against 69 for a, 54 for b, 96 for c, 119 for d and 198 for e; he accounted for 64 percent of the crew’s lapses and 48 percent of their false starts. Thirty-nine of the fifty-one videos that showed slow eyelid closures were his. He reported difficulty performing the test after 38 of his 148 bouts, 25.7 percent, where the others reported it after 1.6 percent. The group’s reading is “chronic partial sleep deprivation throughout the mission.”

Four of six, then, with “one or more of the following problems: disrupted sleep–wake periodicity (n = 1), increased displacement of sleep into the diurnal period (n = 2), performance deficits associated with chronic partial sleep deprivation (n = 1), and frequent reductions in perceived sleep quality (n = 2).” The problems “occurred early in the mission and persisted unabated.”

The light

The watches also recorded the light, and the group measured the lamps. The modules were lit by fluorescent tubes whose spectral power distribution, identical throughout the facility, was weak between 446 and 477 nanometres, the blue to which the circadian system is most sensitive. At the wrist, during waking, the crew were under less than 177 lux for 90 percent of the mission, under 412 lux for 99 percent, and reached 981 lux only 0.05 percent of the time. The average while actively awake fell by 25.6 percent, from 104.8 lux in the first quarter to 78.0 in the fourth; the crew controlled their own lighting. Light during sleep did not change, and light during rest rose slightly in the final quarter.

ESA’s blue-light experiment, run by Luzian Wolf’s team, exposed the crew to blue-enriched light from about eight in the morning to seven at night between days 439 and 499 and gave them red-tinted glasses; the crew’s video diary of 1 September 2011 shows the modules lit blue. The Penn group did not run it and, under the partners’ agreements, did not know how far the crew complied. They compared the sixty days of blue light with the sixty before and found no statistically significant difference in sleep or waking.

Two corrections and a file

On 25 January 2013 the journal printed a correction: the vertical axis of the paper’s Figure 2A had been labelled “cumulative wake activity (×10⁴)” instead of “(counts/min × 10⁷)”, and that of Figure 2D “(h × 10²)” instead of “(×10²)”. “These errors do not affect the conclusions of the article.” The data were deposited as an Excel file on a University of Pennsylvania server, without names.

Sources

• Mathias Basner, David F. Dinges and colleagues, “Mars 520-d mission simulation reveals protracted crew hypokinesis and alterations of sleep duration and timing”, Proceedings of the National Academy of Sciences 110 (7), 2635–2640, 2013, with its Supporting Information Appendix: the Material and Methods (the instrument, the clocks, the 4,396,333 minutes, the validation study), Tables S4 (sleep onsets), S5 (sleep by quarter), S6 (sleep quality) and S7 (errors), Fig. S6 (the lamps) and Fig. S7 (crewmember b), and the Results on light and on the final twenty days. doi:10.1073/pnas.1212646110, PDF with appendix, correction
• Mathias Basner and colleagues, “Psychological and behavioral changes during confinement in a 520-day simulated interplanetary mission to Mars”, PLOS ONE 9 (3), e93298, 2014, CC BY, for the per-crewmember summary of the sleep data, d’s factor of 1.85, f’s shares of lapses and false starts, and c and d. doi:10.1371/journal.pone.0093298
• European Space Agency, Mars500 Information Kit, February 2011, “Effect of blue-enhanced light on alertness and sleep-wake behaviour” (Luzian Wolf and colleagues); and “Mars500 video diary 14: Welcome to the blue light zone!”, 1 September 2011. information kit (PDF), video diary 14