Time perception in astronauts on board the International Space Station

The study investigates whether long-duration spaceflight alters astronauts’ perception of time, building on evidence that spatial perception (size, distance, depth) and neurovestibular processing are modified in weightlessness. Given proposed overlap in neural substrates for space and time representations, the authors hypothesized that the absence of gravitational reference and the slower pace of movement in orbit would alter temporal perception, specifically leading to underestimation of the relative time between events. Understanding these changes has operational importance for tasks requiring precise timing, such as docking and piloted landings.

Prior work shows weightlessness induces neurovestibular changes (orientation illusions, altered vestibulo-ocular reflexes, motion sickness) and affects cognitive tasks involving spatial orientation and mental rotation. Astronauts have reported ‘time compression’ and ‘space fog’ during missions, and slowed decision-making upon re-entry. Spatial perception studies indicate underestimation of distances in microgravity, likely from adaptive changes in gravitational processing. Theoretical context includes the intertwined nature of space and time and hypotheses of shared cortical metrics. Empirical studies of time perception in space are scarce: an early counting-based assessment during Vostok-2 showed near-accurate 20 s estimates with feedback; a Shuttle study (2–16 s reproduction) found increasing underestimation over flight days and shortly after landing. These suggest brief-interval timing may deteriorate during flight and post-landing.

Participants: 10 ISS crewmembers (9 male, 1 female; mean age 44.1 ± 4.6 years) with no vestibular/oculomotor abnormalities; and 15 ground controls (6 female, 9 male; mean age 43.2 ± 18.8 years). Ethics: Approved by ESA Medical Board and NASA JSC IRB; informed consent obtained. Design: Repeated measures before (three sessions: L−205±51, L−149±55, L−116±45 days), during (approximately monthly: FD17±6, FD46±8, FD71±6, FD99±7, FD134±8, FD164±7), and after flight (R+1, R+5±1, R+9±1). Mission duration 6–8 months (mean 202 ± 28 days). Tasks and apparatus:

• Minute production task (“How long is a minute?”): Using a head-mounted display (Oculus Rift) and a finger trackball on a laptop, subjects pressed ‘go’, waited what they judged to be 60 s, then pressed ‘stop’. Audio instructions via earphones; counting prohibited. On Earth, subjects were seated; in-flight, they were free-floating (reduced proprioceptive/tactile/static vestibular cues). The psychophysics setup matched prior spatial perception experiments.
• Temporal estimation questionnaire (in-flight only for astronauts): Via laptop keyboard, subjects reported durations since: (a) last time performing the test; (b) start of the workday (defined as end of the daily planning conference); (c) lunch; (d) last vehicle docking; and (e) last EVA. Docking/EVA dates from NASA mission documentation. Control procedures: Ground controls performed the minute task seated with identical hardware/software, and estimated durations since last session (days), wake-up (minutes), and breakfast (minutes) recorded in diaries. Session spacing for controls (mean 44.1 ± 10.2 days) matched astronauts’ pre-flight spacing (45.2 ± 28.4 days); wake-up and lunch intervals were comparable. Statistical analysis: Time errors computed as percentage (or days) difference between perceived and actual durations. Linear mixed models (LMMs) assessed: (1) ground-based differences across three sessions and between groups (astronauts vs controls); (2) within-phase session effects (pre-, in-, post-flight) for astronauts; and (3) main effect of flight phase (pre, in, post). Post-hoc pairwise comparisons used Bonferroni adjustments. When preflight measures were unavailable (workday/lunch in astronauts), Mann–Whitney tests compared in-flight astronaut responses to ground controls. For measures without ground controls (docking/EVA), one-sample t-tests assessed deviation from zero error. Analyses conducted with JASP 0.16.1.0 and IBM SPSS 27. Supplementary Table 1 provides fixed/random effects details of LMMs.

Minute production:

• On ground, one minute was produced as 74.1 ± 19.5 s on average; no significant differences across sessions or between astronauts and controls [sessions: F(2,46)=0.58, p=0.56; groups: F(1,23)=0.008, p=0.928].
• Within-phase session effects were not significant (pre: F(2,18)=0.939, p=0.409; in: F(5,45)=0.593, p=0.705; post: F(2,18)=1.212, p=0.321).
• Pooled by phase, there was a significant main effect of flight phase [F(2,108)=15.050, p<0.001]; post-hoc: pre vs in p<0.001; post vs in p=0.002; pre vs post p=0.484. In-flight mean production was 59.6 ± 9.1 s versus 74.5 ± 20.2 s pre-flight, a 20.0% decrease (relative overestimation of 1-min duration in-flight), with near-accurate mean around 60 s during flight. Duration between test sessions (days):
• Intervals similar pre-flight (34–56 d) and FD45 to R+1 (31.0 ± 10 d); longer gap L−116 to FD17 (133.8 ± 43 d).
• Largest errors at FD17 and R+1 (after gravity transitions): mean error −26.0% (SD 24.3).
• No group/session learning effects on ground [F(1,23)=0.023, p=0.881] or between groups [F(1,23)=0.005, p=0.942].
• Within-phase differences significant only post-flight [F(2,27)=4.913, p=0.009], specifically R+1 vs R+4 (p=0.018) and R+1 vs R+8 (p=0.027). No main effect of phase when pooled [F(2,98)=2.390, p=0.097]. Hours-scale durations:
• Start of workday: testing occurred ~4.4 ± 1.0 h after DPC end in-flight and 4.5 ± 1.1 h in controls. No session effects within groups (controls: F(2,42)=0.286, p=0.880; astronauts in-flight: F(5,54)=0.469, p=0.797). Astronauts underestimated duration relative to ground controls (Mann–Whitney p=0.001); overall mean error −14.2% (SD 24.2%).
• Since lunch: testing ~2.7 ± 0.7 h after lunch in-flight and 2.5 ± 0.5 h in controls. No session effects (controls: F(2,42)=0.014, p=0.986; astronauts in-flight: F(5,45)=0.591, p=0.707). Astronauts underestimated relative to controls (Mann–Whitney p=0.037); mean error −19.2% (SD 36.1%). Days-scale event durations:
• Since last docking: actual 26.0 ± 12.4 days prior; mean error +2.2% (<1 day). No session effect [F(5,45)=0.695, p=0.630]; error not different from zero (p=0.797).
• Since last EVA: actual 49.4 ± 15.9 days prior; mean error 2.8 ± 11.9 days (+5.6%). No session effect [F(5,45)=0.312, p=0.903]; error not different from zero (p=0.167).

Astronauts showed a relative overestimation in a 1-minute production task during flight (shorter produced intervals), yet underestimation for hours-scale intervals (since workday start and lunch) and for intervals between sessions when gravity transitions occurred. Accuracy for days-scale, salient events (docking, EVA) remained high. The dissociation aligns with distinct mechanisms for sub-minute versus hours-long intervals, with working memory/clock-like processes for shorter durations and memory-based processes for longer ones. Operationally, minute-level accuracy in flight may be aided by the ISS Onboard Short-Term Plan Viewer, which provides continuous visual time cues, whereas variable daily activities may lead to compression of perceived hours. Relativistic time dilation at ISS speeds is negligible relative to observed effects. Multiple time standards (GMT, Houston, Moscow, mission elapsed time) may confound temporal judgments. The pattern is unlikely to be driven by stress alone, as laboratory stress often yields overestimation, whereas astronauts mainly underestimated hours. Vestibular contributions are plausible: reduced static otolith input and slower movements in microgravity may alter temporal processing, consistent with optokinetic/vestibular stimulation studies that modulate perceived duration and with known overlaps between spatial and temporal processing networks, including right parietal and hippocampal–entorhinal systems. No adaptation across the mission suggests changes establish early (within ~2 weeks) and persist, raising concerns for manual, time-critical operations during and shortly after spaceflight.

This study provides the first systematic assessment of time perception across minutes to days during long-duration ISS missions. Astronauts produced shorter 1-minute intervals in-flight (relative overestimation) and underestimated hours-scale durations (since work start and lunch), while remaining accurate for days-scale, event-anchored intervals (docking, EVA). Effects were most pronounced around gravity transitions and showed no adaptation over months. Findings support overlapping neural representations of space and time and implicate vestibular changes in microgravity. Operationally, altered temporal perception may impact time-critical tasks such as docking and landing. Potential future directions include: delineating neural and vestibular mechanisms underlying temporal distortions in microgravity; evaluating countermeasures (e.g., enhanced temporal cues, training) to mitigate operational risk; expanding sample sizes and mission profiles; and assessing timing performance in tasks directly linked to spacecraft operations.

• Preflight measures for hours-scale durations (since start of workday, since lunch) were unavailable for astronauts due to lack of access to training schedules, limiting baseline comparisons for those measures.
• No ground control comparisons were available for days-scale event estimates (docking, EVA); analyses relied on one-sample tests against zero error.
• Small astronaut sample size (n=10) limits generalizability.
• Potential confounding factors include continuous access to timeline cues (OSTPV), variable daily workload and activity types, differing time standards (GMT/Houston/Moscow/mission elapsed time), and environmental factors (isolation, workload-related stress), which may influence temporal judgments.
• Gravity transitions created unequal intervals between some sessions (e.g., L−116 to FD17), potentially affecting between-session duration estimates.