The impact of gravity on perceived object height

The study investigates whether changing the relationship between the body and gravity—through posture (sitting upright vs. lying supine) or exposure to microgravity—biases visual perception of object height and whether such biases affect precision (uncertainty) of judgments. Prior work has shown gravity can influence perceived size and distance in both parabolic flights and on the ISS, with reports of distortions such as compressed depth and altered width/height matching. Postural manipulations on Earth (supine/prone vs. upright) have similarly affected size perception, suggesting gravity or body orientation plays a role. The authors hypothesized that participants would set the height of a distant square larger when supine or in microgravity than when upright, reflecting an underestimation of perceived height, and that vestibular noise under these conditions would decrease precision. They further examined persistence of any effects after return to Earth and explored possible sex/gender differences.

Prior studies indicate gravity modulates size and distance perception. On the ISS, astronauts have set the height of a cube too short when matching height to width, suggesting elongation in perceived height, and have underestimated depth (front-to-back distance), implying compression of perceptual space. During parabolic flight, the minimum aperture width judged passable was narrower in microgravity than under 1 g, possibly due to a lower subjective eye level; however, long-duration ISS data with the same task found no significant differences, potentially due to adaptation or small samples. On Earth, Harris and Mander showed that when participants were effectively supine (physically or via a pitched visual room), a projected vertical line had to be lengthened (~9%) to match a haptic stick, implying visual objects appeared smaller when supine. Kim, McManus, and Harris in VR with disparity cues found targets needed to be larger when supine (5.4%) and prone (10.1%) versus upright. Viewing targets through the legs (altered posture and eye height) led to underestimation of size with distance (Higashiyama and Adachi), with evidence pointing to posture rather than scene inversion as critical. Somatosensory contributions were tested by comparing land vs. neutral buoyancy underwater; no differences suggested a limited role for somatosensory postural cues in size perception. Mechanistically, microgravity may rescale visual space or remove a gravitational reference frame for interpreting visual input. If vestibular signals directly support size estimation, unusual gravity or supine posture (noisier vestibular gravity signals) should reduce precision. Evidence for posture-dependent changes in precision has been mixed. Gravity may also affect haptic perception similarly to visual perception (Morfoisse et al.), potentially confounding comparisons when a haptic reference is used.

Design: Repeated-measures study with five sessions for astronauts and matched-timing sessions for controls. Sessions: Pre-Flight (Earth), Early ISS (days 2–6 on orbit; astronauts floating), Late ISS (~60 days on orbit; astronauts floating), Early Post-Flight (within 7 days of return), and Late Post-Flight (~90 days after return). On Earth sessions (Pre-, Early Post-, Late Post-Flight), participants completed both sitting upright and lying supine postures (order counterbalanced). Controls mirrored session timing; to simulate microgravity sessions, controls lay supine in sessions 2 and 3.

Participants: Astronauts: 12 (6 women, 6 men; mean age 42.6 years, SD 5.4). Three additional astronauts did not provide full data (1 excluded for timing; 2 delayed). Controls: 20 (10 women, 10 men; mean age 42.6 years, SD 7.2). All reported normal or corrected vision and no vestibular/balance/depth issues. Ethics approvals obtained (York University, NASA, and relevant space agencies); informed consent obtained.

Apparatus: Oculus Rift CV1 HMD (~110° diagonal FOV; 90 Hz). The VR display was head-fixed; both eyes received the same image (no binocular disparity cues). Stimuli were programmed in Unity (2017.1.0f3) and run on an HP laptop with NVIDIA Quadro K610M GPU. Responses via a finger mouse. A steel reference stick (38.1 cm long, 2.5 cm wide, 4 mm thick) was held aligned with the body’s long axis. A cervical neck collar minimized head movements relative to the body.

Stimuli and procedure: Participants were immersed in a head-fixed virtual hallway (3.3 m high and wide), with a dark floor, lighter ceiling, implied horizon, and white Gaussian blobs on walls to enhance perspective cues. The simulated viewpoint corresponded to a constant simulated eye height of 1.65 m at the corridor center. A blue 2D square appeared at one of three simulated distances (8, 12, 16 m), with a floor line aligned to ground the square. On each trial, participants judged whether the square’s height was shorter or taller than the 38.1 cm haptic reference stick they held (but could not see during trials). They responded via mouse (left = shorter, right = taller). Practice trials with very large/small objects verified correct response mapping. Instructions were available in-HMD; on Earth, instructions were also read aloud; in space, SOPs and a “BIG PICTURE” summary were provided.

Staircase procedure: Square size was controlled by paired PEST adaptive staircases per distance (two per distance; total six interleaved per posture and session). One staircase started at 2× reference (76 cm), the other at 0.5× (19 cm). Staircases terminated after 25 trials or 13 reversals. Bounds: 9.5–228 cm. No disparity cues; hallway scene remained fixed relative to body orientation. A video demonstration: https://osf.io/t47nd.

Data processing and psychophysics: For each participant, session, posture, and distance, psychometric functions (cumulative Gaussian) were fit to the full staircase data using R (quickpsy) with differential evolution optimization. The mean provided the Point of Subjective Equality (PSE); the standard deviation (equivalent to the 84.1% JND) indexed precision (higher JND = lower precision). PSEs were converted to PSE ratios by dividing by the reference length (38.1 cm), with ratios >1 indicating that the square had to be set larger to match the reference (i.e., perceived as smaller).

Outlier handling: Conditions were excluded if, within a PEST, ≥5 stimulus presentations hit the upper or lower bounds (9.5 or 228 cm), indicating unreasonable responses. This led to exclusion of one entire second session (supine) for one control and the supine posture of the first session for two controls. No astronaut sessions were excluded.

Statistical analysis: Linear mixed models (lme4 in R) were used separately for accuracy (PSE ratio) and precision (JND). Fixed effects: Session (Pre-Flight, Early ISS, Late ISS, Early Post-Flight, Late Post-Flight), Posture (Sitting, Supine; plus Microgravity for astronauts during ISS sessions), Target Distance (8, 12, 16 m). For JND models, PSE ratio was included as a covariate. Random effects: by-participant random intercepts and random slopes for Session, Posture, and Distance. Significance assessed via bootstrapped 95% confidence intervals for fixed-effect coefficients (confint in R). Where astronaut effects were significant, interactions with cohort (astronauts vs. controls) were tested on the relevant data subset to determine whether effects differed between cohorts. Planned comparisons tested a priori hypotheses about microgravity/posture effects on PSEs and JNDs and post-flight recovery. Data and analysis code are available on OSF: https://osf.io/wvg9z/.

• Across sessions and cohorts, PSE ratios were consistently >1, indicating that the square had to be set larger than the reference stick to appear equal in height (consistent with compressed space in VR).
• Astronauts (accuracy): Model intercept PSE ratio = 1.43 (95% CI [1.15, 1.67]). No significant Sitting vs. Supine differences within any session. Between sessions, PSE ratios were higher at Late Post-Flight than at Pre-Flight for Sitting by 0.37 (95% CI [0.17, 0.57]) and for Supine by 0.30 (95% CI [0.12, 0.49]). Cohort interaction confirmed a significantly larger Late Post-Flight vs. Pre-Flight increase in astronauts than controls for Sitting (interaction 0.53, 95% CI [0.01, 1.02]); not significant for Supine (0.42, 95% CI [−0.03, 0.93]). No significant differences between Pre-Flight and Early ISS or Late ISS were detected.
• Astronauts (precision): No significant differences in JNDs across sessions or postures; microgravity did not reduce precision.
• Controls (accuracy): Model intercept PSE ratio = 1.74 (95% CI [1.42, 2.03]). At Pre-Flight, Supine yielded significantly lower PSE ratios than Sitting by 0.12 (95% CI [0.03, 0.20])—opposite to prior expectations. This difference was not present at Early or Late Post-Flight. No other significant contrasts.
• Controls (precision): At Pre-Flight, JNDs were significantly lower for Supine than Sitting by 0.06 (95% CI [0.03, 0.08]); no differences at later sessions.
• Order effects: Analyses restricted to participants starting with Supine or with Sitting at Pre-Flight were consistent with the main findings; within-session order (first vs. second posture) did not affect PSEs.
• Sex/gender: Exploratory analyses (Supplementary) were largely inconclusive; no firm statements can be made. Overall: No immediate microgravity effect on perceived object height or precision was detected. Astronauts showed a late-emerging increase in underestimation (higher PSE ratios) 60+ days post-flight. Controls showed a transient posture effect at the first session only. A posture-dependent simulated eye-height explanation may reconcile discrepancies with earlier literature.

The findings do not support the hypothesis that microgravity or supine posture acutely impairs precision in height judgments; astronauts’ JNDs were stable across gravity contexts, and controls even showed improved precision when supine at Pre-Flight. Contrary to expectations and prior studies (Harris & Mander; Kim et al.), controls initially set smaller PSE ratios when supine, and astronauts showed no sitting–supine difference across sessions. The authors propose that differences in simulated vs. real eye height may offer a unifying explanation: when seated, participants may have relied on their real-world seated eye height rather than the simulated (fixed) eye height, biasing size scaling differently than when supine, where real-world eye height is ill-defined. VR fidelity and scene cues may further modulate perceived size, as lower environmental fidelity (no disparity, minimal ground plane cues) was associated with larger PSE ratios across studies. Critically, the absence of significant Pre-Flight vs. in-flight differences contrasts with earlier microgravity findings; potential explanations include the larger target distances used here (8–16 m vs. near distances in prior work) and the use of a haptic reference that might be similarly affected by gravity. A novel and concerning result was the late post-flight increase in underestimation in astronauts, not attributable to learning or motivation (which would affect JNDs and were stable), and not seen in controls. Speculatively, prolonged microgravity may induce a strategy that downweights multisensory scaling cues, leading to a persistent ‘flattening’ of visual space after return. These results suggest that while tasks depending on accurate height judgments appear safe soon after arrival in space, latent biases emerging weeks after return warrant attention. The data provide no strong evidence for sex/gender differences in this task.

This study shows no acute effect of long-duration microgravity on perceived object height or judgment precision for distant targets in VR, and no consistent supine–upright differences in astronauts. However, astronauts exhibited robust late-emerging increases in underestimation (higher PSE ratios) 60+ days after return to Earth. Controls showed a transient supine effect and improved precision only at the first session. These findings weaken support for a gravity-as-reference-frame mechanism directly reducing precision and suggest that posture-dependent eye-height cues and VR scene fidelity may explain discrepancies with prior work. Practical implications: no countermeasures appear necessary for acute microgravity effects on object height perception, but space travelers should be cautioned about possible delayed, longer-lasting biases after return. Future work should test immediately upon arrival/return to capture transient effects, increase sample size and power, manipulate simulated and real eye height explicitly, include binocular and richer environmental cues, examine a wider range of distances (including near), and assess whether haptic references are themselves gravity-sensitive.

• Timing constraints: Astronauts were not tested until several days after ISS arrival and not immediately post-landing; transient effects may have been missed.
• Sample size and variability: Modest astronaut sample (n=12) and variability may limit power, especially for detecting in-flight effects and sex/gender differences.
• VR constraints: Head-fixed VR with no disparity cues and impoverished environment likely compressed perceived space and may limit generalizability to real-world viewing; field of view and absence of body/hand visuals differ from prior studies.
• Eye-height confound: Simulated eye height was fixed, but participants’ real-world posture-specific eye height may have influenced scaling differently across postures.
• Haptic reference: If gravity similarly affects haptic length perception, visual–haptic comparisons may mask visual-only gravity effects.
• Staircase/JND estimation: PEST-based procedures can underestimate JNDs, though applied uniformly across conditions.
• Distance differences vs. prior work: Targets were farther (8–16 m) than in studies reporting microgravity effects (often within arm’s reach), potentially reducing sensitivity to eye-height or spatial scaling effects.
• Controls’ ‘microgravity’ sessions were only supine on Earth, not true microgravity.