Confinement has no effect on visual space perception: The results of the Mars-500 experiment

Abstract

People confined to a closed space live in a visual environment that differs from a natural open-space environment in several respects. The view is restricted to no more than a few meters, and nearby objects cannot be perceived relative to the position of a horizon. Thus, one might expect to find changes in visual space perception as a consequence of the prolonged experience of confinement. The subjects in our experimental study were participants of the Mars-500 project and spent nearly a year and a half isolated from the outside world during a simulated mission to Mars. The participants were presented with a battery of computer-based psychophysical tests examining their performance on various 3-D perception tasks, and we monitored changes in their perceptual performance throughout their confinement. Contrary to our expectations, no serious effect of the confinement on the crewmembers’ 3-D perception was observed in any experiment. Several interpretations of these findings are discussed, including the possibilities that (1) the crewmembers’ 3-D perception really did not change significantly, (2) changes in 3-D perception were manifested in the precision rather than the accuracy of perceptual judgments, and/or (3) the experimental conditions and the group sample were problematic.

Astronauts during spaceflight are exposed to a visual environment that differs in many important aspects from the familiar environment on Earth. For instance, during exposure to weightlessness, the gravitational frame of reference is absent. The directions up and down are no longer unequivocally defined, and the perception of upright becomes relatively more body-defined (Clément, Arnesen, Olsen, & Sylvestre, 2007; Glasauer & Mittelstaedt, 1998). In such perceptual ambiguity, the task of spatial navigation, or even a simple search for objects on board, becomes challenging (Clément & Reschke, 2008; Oman, 2007). Moreover, in the confined environment of a spacecraft or space station, all visible objects stay within a few meters from the observer’s viewpoint, and the visibility of farther objects is precluded. This visual field restriction does not allow a crewmember to perceive objects relative to the position of the horizon, thus making size and distance estimation rather difficult (Rand, Tarampi, Creem-Regehr, & Thompson, 2012). In addition to microgravity and confinement, astronauts are also exposed to a variety of other stressors during spaceflight, such as altered light:dark cycles, sleep deprivation, and high mental and physical workloads (Clément & Reschke, 2008), which altogether may adversely affect their 3-D perception.

A few psychophysical experiments have attempted to determine how crewmembers’ three-dimensional (3-D) perception is changed by the short-term microgravity generated during parabolic flights. In an experiment by Clément, Lathan, and Lockerd (2008), subjects were asked to adjust the length of a particular dimension (height, width, or depth) of a 3-D cube presented in a virtual-reality environment until all three dimensions had the same apparent length. The data demonstrated that subjects judged the cube to have a smaller height, slightly larger width, and longer depth. The authors attributed those errors in perceived dimensions to a general tendency to perceive the straight-ahead direction as being lower in microgravity than in normal gravity, which is in accordance with frequent astronaut reports indicating that spacecraft interiors look longer and higher than they really are (Lathan, Wang, & Clément, 2000). Furthermore, Villard, Garcia-Moreno, Peter, and Clément (2005) demonstrated a decrease in the strength of geometric illusions based on perspective, such as the Ponzo, Müller-Lyer, and horizontal–vertical illusions, in a microgravity environment. The results of all these studies indicate that microgravity has a substantial effect on the observers’ 3-D perception. Still, it is important to bear in mind that no study has yet tested the microgravity-induced changes in 3-D perception during spaceflight.

Even less is known about the effect of exposure to a confined environment; research has not explored such effects in astronauts or in other subjects experiencing confinement (e.g., antarctic winter-over participants, submariners on long-duration missions, prisoners in solitary confinement, participants of closed-system experiments [e.g., Biosphere 2, BIOS-3], or bed-confined patients or elderly confined to their domestic environments). There is some indirect evidence of such an influence. For instance, submariners are forbidden to drive cars immediately after returning from a mission (Clément & Reschke, 2008). Subjects in sensory deprivation studies often comment on a lack of depth perception after completing their confinement (Zubek, 1969). And antarctic winter-over expedition members may tend to overestimate the size of observed objects and underestimate their distance (Smith, 1966). Is it thus reasonable to expect that, as a consequence of the experience of being confined, the 3-D perception of participants would become temporarily impaired?

We had the opportunity to longitudinally investigate the perceptual performance of six volunteers participating in the Mars-500 project, an isolation study that simulated a flight to Mars. The crewmembers spent 520 days in a sealed facility that was designed to resemble a spacecraft. The interior volume of the habitat was only 550 m³, and the tubular space of the facility was further split into a number of rooms, including six individual quarters, a kitchen, and a storage space (see Fig. 1). With the exception of the narrow transfer tunnel with hatches, their view was restricted to no more than 5 m.

Fig. 1

Diagram of the facility.

A series of psychophysical experiments examining the crewmembers’ visual space perception, particularly their responsiveness to perspective cues, was conducted.¹ In the first experiment, we investigated whether and to what extent the crewmembers were able to ignore 3-D contextual information when estimating the relative 2-D lengths of the lines that stood in front of the background scene. In contrast, the second experiment focused on whether crewmembers could extract and apply perspective information from the 3-D environment when matching the size of objects that are a natural part of the visual scene. In the third experiment, we investigated the perceptual salience of farther versus closer objects in the visual environment using a change-detection task. Finally, in the fourth and fifth experiments, which were only conducted shortly after the isolation ended, we tested whether the crewmembers’ visual space was isotropic in every direction. All experiments were designed to test the hypothesis that, as a consequence of obscured open-space views, the crewmembers would be less responsive to perspective visual cues. (Alternatively, however, one might expect a negative adaptation aftereffect—i.e., strong responsiveness; see, e.g., Ghuman, McDaniel, & Martin, 2010.)

General method

The subjects of the Mars-500 project, which was conducted at the Institute of Biomedical Problems in Moscow from June 2010 to November 2011, were six male crewmembers—three Russians, one French, one Italian, and one Chinese. Their ages ranged from 27 to 38 years (M = 32.0). All of the subjects reported normal to corrected-to-normal vision and were naive to the purpose of the experiments. The control group consisted of seven undergraduate students, who were examined under laboratory conditions.

Each subject ran the experiments on his own personal laptop in his living quarter. The stimuli were presented on a 15.6-in. monitor set at a resolution of 1,600 × 900 pixels. The subjects viewed the photographs used as stimuli binocularly from a distance of approximately 50 cm, and on average, the stimuli subtended a visual angle of 15º. Since the photographs were obtained from the Internet, the focal lengths varied considerably, ranging from 6.3 to 300 mm. The experimental session was arranged so that the crewmembers opened the experimental software, chose the particular experiment, read the instructions, and started the experiment. When the crewmembers finished all of the experiments, they uploaded the data files to the server.

The subjects took part in a total of 10 experimental sessions. The baseline session was conducted two weeks before the mission began, eight sessions took place during confinement, and the last session took place seven days after confinement ended. The schedule of the “in-flight” sessions was as follows: Week 2, Month 2, Month 3, Month 5, Month 8, Month 10, Month 13, and Month 15. The control group had an identical schedule.

Each session required approximately 45 min to complete the presented set of tasks. The experimental battery consisted of nine experiments examining the crewmembers’ visual space perception, but also their face and object recognition. The depiction of the latter experiments, however, is outside the scope of the present study. In any particular session, the crewmembers conducted only some of the experiments in the battery, typically five or six. Consequently, unequal numbers of measurements were taken during each experiment, and also individual experiments were performed during various phases of isolation. Our experimental battery was never the only item presented to the crewmembers; instead, the crewmembers received the battery as part of a series of various tests and experiments on various topics that had been scheduled for a given day.²

During the course of confinement, the three experiments were conducted to test the crewmembers’ visual space perception skills.

Experiment 1: 2-D relative length

Method

In this experiment, the stimuli were two lines drawn in red on a photograph of a natural scene (see Fig. 2). The lines were not an actual part of the scene. Still, the presence of the background scene was expected to affect the perceived length of the lines, in the sense that it could induce the illusion effect (see below). The task was to adjust the 2-D lengths of both lines until they matched, regardless of the visual context of the scene. The crewmembers responded by pressing the up and down arrow keys, which simultaneously increased the length of one line and decreased the length of the other line. The starting lengths of the lines were randomized and ranged between 3 and 8 cm.

Fig. 2

Sample stimuli from Experiment 1. The two lines presented formed either a perpendicular or a parallel configuration. The task was to adjust both lines to the same length, regardless of the visual context of the scene in the background.

Four two-line configurations were used: The lines either were parallel and set to the (1) horizontal or (2) vertical direction, or were perpendicular, with the upper line (3) horizontal or (4) vertical. Each crewmember performed the task four to six times for each pair of lines. A total of 20 trials were obtained for all conditions: 4 configurations × 5 repetitions. The task was scheduled at baseline, Month 2, Month 10, and Month 15.

Notice that the configuration composed of two parallel lines corresponded to the configuration of the Ponzo illusion, especially when 3-D scenes eliciting a strong impression of depth were present in the background (see, e.g., Kilbride & Leibowitz, 1975). Furthermore, the configuration of two perpendicular lines corresponded to the configuration of the horizontal–vertical illusion.

In this experiment, we hypothesized that, as a consequence of the crewmembers’ lack of exposure to open-space views throughout the confinement, they might perceive the portrayed depth as being less pronounced. Thus, the visual context of the scene would have a reduced effect on their perceptual judgments of the relative 2-D lengths. In other words, we expected that the crewmembers would gradually become less susceptible to geometric illusions.

Results

For each condition, Table 1 shows the accuracy and precision of the judgments. For a visualization of the same data, see Fig. 3. The accuracy is given by the average mean (M) computed from the means of five repeated judgments by the same observer in each condition and each session. The observers’ precision is given by the average coefficient of variation (SD/M) computed from the individual coefficients of variation in each condition and each session (for an explanation of the procedure, see Lappin, Shelton, & Rieser, 2006).

Table 1 Summary of results for the four conditions of Experiment 1: Horizontal parallel lines, vertical parallel lines, perpendicular lines with the upper line vertical, and perpendicular lines with the upper line horizontal

Fig. 3

Results from Experiment 1. For an explanation of values and symbols, see the note to Table 1.

The relative 2-D lengths of the lines were consistently misperceived by all crewmembers. The lengths of the parallel lines were generally judged with greater accuracy than the lengths of the perpendicular lines (t = 3.03, p < .01), which is consistent with previous findings (see Norman, Todd, Perotti, & Tittle, 1996, for a review). In the former case, the crewmembers adjusted the top line to be too short by an average of 5.2%. In the latter case, on the other hand, the vertical line was consistently overestimated. The mean deviation was 9.0%. These magnitudes of under- and overestimation in the crewmembers were, on average, lower than those of the control group (difference between means, t test: for parallel lines, t = –3.09, p < .01; for perpendicular lines, t = 0.61, p = .54), and also lower than the results obtained in previous studies on the Ponzo illusion (Fineman, & Carlson, 1973; Li & Guo, 1995; Prinzmetal, Shimamura, & Mikolinski, 2001) and the horizontal–vertical illusion (Craven, 1993; Mamassian & de Montalembert, 2010). Importantly, however, the crewmembers’ judgments were already generally more accurate in the first session, and remained rather stable across all sessions. A repeated measures analysis of variance (ANOVA) with Number of Sessions as a factor showed no significant difference across either the average means [F(3, 17) = 0.83, p = .62] or the average coefficients of variation [F(3, 17) = 0.32, p = .98]. For each condition, a regression analysis was computed to evaluate the changes in each subject’s data as a function of time. All regression slope coefficients were found not to differ significantly from zero (t tests for the four experimental conditions in the crewmembers: t = 0.278, 0.016, 0.683, 0.609; t tests for the four experimental conditions in the control group: t = 0.654, 0.560, 0.740, 0.880).

Experiment 2: 3-D relative size

Method

In this experiment, photographs that had been selected to suggest significant depth were used. Each photograph included two or more human figures standing at apparently different distances from the camera position (see Fig. 4). With the exception of one figure in the scene, all figures were of a size appropriate to the scale. These figures served as references for the test figure, whose image size was intentionally set to be too small or too large with respect to the scale of the scene. The crewmembers were instructed to change the size of the test figure by pressing the up and down arrow keys so that the size of the test figure would fit the scene. Each crewmember was tested for 31 trials at any given session. The task was scheduled at baseline, Month 2, Month 5, and Month 15.

Fig. 4

Sample stimuli from Experiment 2. One figure in the picture, indicated here by the arrow, was presented at a disproportionate size. The task was to adjust the size in the picture to fit the scene.

We were well aware that this task can in principle be solved without 3-D perception, simply by aligning the eyes of the test figure with the eyes of other figures in the photograph and/or with the horizon. To prevent this artifactual strategy, the set only included photographs in which the figures were shown from above or below—that is, the figures’ eyes were above or below the horizon.

We hypothesized that unlike the control subjects, the crewmembers would perform this task with low accuracy and high variability, especially after spending a considerable amount of time in confinement.

Results

Table 2 shows the averaged means (M) and coefficients of variation (SD/M) for the crewmembers and the control group in each condition and each session. For a visualization of the same data, see Fig. 5. As in the previous experiment, the perceptual judgments were not dramatically different in the two groups. A two-way ANOVA with Number of Sessions and Group as factors revealed no significant effect of sessions (F = 1.24, p = .22) or of groups (F = 3.05, p = .10).

Table 2 Summary of results for the four conditions of Experiment 2: Near and far testing figures, and testing figures seen from above and below

Fig. 5

Results from Experiment 2. For an explanation of values and symbols, see the note to Table 2.

The size judgments were found to be largely dependent on the relative order of the human figures within the 3-D scene. The size of the test figure was systematically set too small when it was positioned closer in the scene than the reference figures (mean = –3.1%). Conversely, the size of the test figure was systematically set too large when it was positioned farther away in the scene than the reference figures (mean = 4.5%). The data for the near and far conditions were significantly different for both groups (difference between the means in the crewmembers, paired t test: t = –10.72, p < .01; difference between the means in the control groups, paired t test: t = –7.65, p < .01).

Moreover, the perceived size of the human figure changed with the camera position (see the relevant parts of Table 2). The figure was consistently set too small when it was observed from below (mean = –3.4%). Conversely, the figure was consistently set too large when it was observed from above (mean = 3.2%; for an explanation of this perceptual bias, see, e.g., O’Shea & Ross, 2007). Again, the differences between the means in the below and above conditions were significantly different for both groups (difference between the means in the crewmembers, paired t test: t = 15.90, p < .01; difference between the means in the control group, paired t test: t = 10.32, p < .01).

The ANOVA revealed no interaction between the relative order of the stimuli and the camera position (F = 0.22, p = .64).

According to our expectations, the crewmembers should have perceived less depth in the 3-D scenes. Consequently, the crewmembers were expected to gradually start to adjust the size of the near test figures to be too small, whereas the adjustments to the size of the far test figures were expected to be too large. Adjustments to the size of a figure observed from below were expected to be too small, whereas adjustments to the size of a figure observed from above were expected to be too large. This pattern of data was observed in the baseline session and remained unchanged across all sessions. A repeated measures ANOVA with Number of Sessions as a factor showed no significant difference across the average means [F(3, 18) = 1.90, p = .06] or the average coefficients of variation [F(3, 18) = 0.96, p = .50]. For each condition, a regression analysis was computed to evaluate the changes in each subject’s data as a function of time. All of the regression slope coefficients were found not to differ significantly from zero (t tests for the four experimental conditions in the crewmembers: t = 0.505, 0.177, 0.926, 0.554; t tests for the four experimental conditions in the control group: t = 0.973, 0.521, 0.546, 0.680).

Experiment 3: Change detection

Method

In this experiment, the crewmembers were presented with 18 pairs³ of continually alternating near-twin images of real-world scenes that were separated by a brief blank field. Each image was displayed for approximately 250 ms, and each blank was displayed for approximately 80 ms. In the typical change blindness paradigm, a single change is made within the visual scene in terms of the color, location, or size of a given object (Rensink, O’Regan, & Clark, 1997; Scholl, 2000; Simons & Rensink, 2005). In the present study, in fact two changes were made instead of one. One change was localized in a near-distance region, whereas the other was in a far-distance region (see Fig. 6). Notice that the latter change was always positioned in a location of the image that exceeded the commonly perceived distances in the interior of the facility (between several and several hundred meters from the camera position). Importantly, both changes were identical concerning the image size, and thus the more distant change depicted a larger 3-D object. In addition to size, other image parameters, such as the color and brightness of the changing elements, were also harmonized, and the position of changes was balanced from left to right to guarantee that both changes would be equally perceptually salient.

Fig. 6

Sample stimuli from Experiment 3. Two black circles in each image indicate the changing elements in a pair of alternating images. The task was to detect a change as quickly as possible.

The crewmembers’ task was to find the change as quickly as possible and to click the mouse button at the place on the screen where the change occurred. The response triggered the presentation of the next trial. Each trial had a time limit of 45 s. If the crewmembers failed to detect the change within this time limit, the next image was automatically displayed. We evaluated which of the two changes was preferred. Notice that we assumed that the crewmembers would detect only one of the two changes in the image. To minimize the likelihood that both changes would pop out simultaneously, the following restrictions were applied: The scenes in the experimental set were complex and full of details, the changed regions were reasonably small (i.e., they never occupied more than 3% of the overall image area), two changes were never close to each other in the image, and the changes were not localized symmetrically around the vertical or horizontal axes. The task was scheduled at baseline, Month 3, Month 8, Month 13, and postisolation. An entirely different set of images was used in each session.

We hypothesized that, with the increasing time of the confinement, the ratio of near detected changes to all detected changes would gradually increase with increasing confinement time. This relationship would indicate an impairment of size constancy in the sense of a decrement in the crewmembers’ sensitivity to information localized at a far-distance region.

Results

In the change detection task, the crewmembers successfully detected the changes in 97.6% of all images, requiring an average of 11.6 alternations (7.7 s). The change-detection performance of the control subjects was comparable (96.7%); nonetheless, they noticed changes more slowly, with an average of 13.2 alternations (8.7 s). However, a repeated measures ANOVA with Group as a factor showed a nonsignificant effect on the mean reaction time [F(4, 16) = 1.65, p = .089].

In this experiment, two changes were in each visual scene; one change was localized at a near-distance region, whereas the other change was placed at a far-distance region. The crewmembers were expected to detect near and far changes in a way similar to the control subjects at the beginning of confinement, and to display progressively less sensitivity to detecting the far changes. However, the data revealed that both experimental groups had greater success detecting near changes than detecting far changes as early as in the baseline session, and this trend persisted in all subsequent sessions. The difference in the ratio of near detected to all detected changes between the two groups was not statistically significant in any of the sessions (see the values of the χ ² tests in Table 3), and never exceeded 14%. For a visualization of the data, see Fig. 7.

Table 3 Results from Experiment 3

Fig. 7

Results from Experiment 3. For an explanation of values and symbols, see the note to Table 3.

In addition to the described experiments that were conducted during isolation, two other experiments were performed seven days after isolation ended.

Experiment 4: Bisection

Method

In this experiment, black-and-white photographs of urban scenes were presented to the crewmembers. All of the photographs were taken using a Panasonic Lumix DMC-FX550 camera with a focal length of 25 mm. For each photograph, two landmarks such as corners or doors were selected and marked by small yellow crosses. The crosses represented two (fixed) endpoints of an imaginary 3-D line connecting two points in the scene located at different distances from the observer’s viewpoint.

The crewmembers were instructed to move a third, adjustable cross that was initially positioned at a random location along the imaginary 3-D line by pressing the appropriate buttons on the keyboard and to set the cross at the perceived midpoint (see Fig. 8). To successfully perform this task, perspective information must be considered, and the matching cross must be placed closer to the more distant of the two endpoints in the image; that is, it must be biased upward. As the orientation of the 3-D line approached the radial direction, the matching cross ought to be shifted closer toward the more distant endpoint. The in-depth orientation of the line defined by the crosses varied between 20º and 80º in the individual photographs.

Fig. 8

Sample stimuli from Experiment 4. Two crosses represented the endpoints of an imaginary 3-D line. The task was to place the third cross at the perceived midpoint of the line.

Each crewmember was tested for 33 trials. We hypothesized that the accuracy and precision of the responses would be low as a consequence of the prolonged experience of confinement.

Results

In the bisection task, the deviation of the crewmembers’ bisection settings from the Euclidean bisection point was computed. Table 4 shows the mean deviations of the settings and the setting variability, as expressed by the interquartile range for the 3-D adjustment. All crewmembers had bisection settings that deviated significantly from the Euclidean bisection point (binomial test, p < .001). The judgments were typically biased downward in the image from the Euclidean bisection point, indicating that the perceived midpoints were too close to the observer (see Lappin et al., 2006; Trommershäuser, Maloney, & Landy, 2003, for a comparison). The same perceptual bias was present, although slightly less pronounced, in the control group (binomial test, p < .001). Still, a significant difference between the two groups was observed (paired t test: t = –3.84, p < .01), which, however, cannot be attributed to the specific environment of the Mars-500 project. Notably, the underestimation of the distance of the midpoint was greatly exaggerated in the case of one crewmember (5006). On almost half of his responses, a perceived midpoint was even below the midpoint in the image, indicating a negative perspective effect. When the data from this crewmember were excluded from the analysis, the difference between groups was greatly reduced.

Table 4 Results from Experiment 4

Experiment 5: L-shape

Method

The only open-field experiment in the experimental battery was conducted in the hall of the Institute of Biomedical Problems. The hall was approximately 25 m long and 3 m wide. A clear linear perspective was provided by the edges of the floor, walls, and ceiling. The stimulus configuration consisted of three yellow plastic bottle caps (2.5 cm in diameter) that were placed on the ground plane and formed an L-shape from the observer’s perspective (i.e., one length interval was at a frontal orientation and the other was aligned with the observer’s line of sight). Two bottle caps were frontally oriented and defined a reference. The distance between the bottle caps was fixed to 1 m. The crewmembers’ task was to position a third bottle cap so as to adjust the length of the depth interval (comparison) to match the length of the frontal interval (standard) (see Fig. 9). This task was accomplished by directing a laser pointer to the perceived location of the target cap.

Fig. 9

Scheme of Experiment 5. Three bottle caps were lying on the ground plane and formed an L-shaped configuration from the observer’s perspective. The task was to match the length of the depth interval to correspond with the length of the frontal interval.

Four different egocentric distances of the standard interval—that is, 3, 7, 11, and 15 m—were used in each environmental setting, indicating that the majority of the judgments were made on distances that exceeded those in the interior of the facility. Additionally, we used two different orientations of the in-depth with respect to the standard frontal interval; one orientation had the depth portion of an L-shape directed toward the observer, and the other orientation had it directed away from the observer. Each crewmember performed the L-shape task twice at each distance and at each relative orientation of the two intervals. Sixteen trials were obtained for each crewmember: 4 distances × 2 orientations × 2 repetitions.

In this experiment, we expected a decrease in the accuracy and precision of the responses with increasing stimulus distance from the observer in both experimental groups. For the crewmembers, however, the slope of the decrement was hypothesized to appear much steeper, especially beyond distances of approximately 5 m.

Results

Figure 10 illustrates the results of the L-shape task; see Table 5 for the numerical results. The abscissa is the egocentric distance of the frontal interval, and the ordinate is the ratio of the in-depth and frontal intervals that were judged to be equal. The two curves represent the data for the crewmembers and the control subjects. The ratios are all well above 1.0, indicating that the in-depth intervals had to be made larger in order to be judged equal to the frontal intervals. The perceptual inequality of the frontal and in-depth intervals increased with egocentric distance. The average depth-to-width ratios were 1.18, 1.41, 1.54, and 1.57 for the individual viewing distances.

Fig. 10

Results from Experiment 5. For an explanation of values and symbols, see the note to Table 5.

Table 5 Results from Experiment 5

Concerning the difference between the performance of the crewmembers and the control subjects, a repeated measures ANOVA with Group as a factor showed a nonsignificant effect, on both the average means [F(1, 11) = 0.68, p = .61] and the average coefficients of variation [F(1, 11) = 0.93, p = .50].

General discussion

Living in an isolated and confined environment for prolonged periods represented a stressful situation for the participants. Stressors arising from the specific environmental attributes, including confinement to a small area, impossibility of escape, limited mobility, social isolation, altered photoperiod, and perceptual monotony (Arias & Otto, 2011), may have negatively affected not only the participants’ behavior and subjective well-being, but also their cognitive functioning.

Systematic data on the possible changes in visual perception as a consequence of a prolonged experience of isolation are missing. To our knowledge, this topic has only been studied with participants of sensory deprivation experiments and spaceflight crewmembers. However, those environments present other factors besides confinement, which may influence the performance in perceptual tasks, and therefore these findings cannot be directly attributed to the experience of confinement.

Sensory deprivation experiments compare a subject’s performance in visual tasks prior to and after such an experiment. No significant changes in depth perception, size constancy, form discrimination, or visual acuity in participants were observed (for a review, see Zubek, 1969). It is important, however, to be aware of issues that are specific to these experiments: (1) The visual understimulation in sensory deprivation experiments is enormous; (2) the period of the confinement does not exceed one or two weeks; and (3) the validity of some methods used in those experiments has been questionable: For instance, conclusions about participants’ depth perception have been formulated only on the basis of the results of a Howard–Dolman stereoacuity test, in which the observers aligned two movable vertical rods until they appeared to be equidistant from the observer.

In astronauts, numerous physiological changes in visual function occur during spaceflight—for example, in visual acuity, contrast sensitivity, and eye movements—particularly during the first days of flight (the results of numerous studies are summarized in Clément & Reschke, 2008). Surprisingly, however, previous studies have not described fundamental changes in the crewmembers’ perceptual performance. Léone (1998) examined mental rotation skills; de Schonen, Léone, and Lipshits (1998) observed face inversion effect; and Benke, Koserenko, Watson, and Gerstenbrand (1993) presented several visuospatial tasks; and all of these studies failed to detect significant changes. On the other hand, numerous anecdotal reports from astronauts can be found describing serious difficulties in distance estimation and 3-D navigation (see, e.g., Oman, 2007).

When designing our experiments for the Mars-500 project, we considered the above-mentioned list of stressors and hypothesized that the most important factors for the crewmembers’ visual performance would be (1) perceptual monotony and (2) confinement to a restricted area, primarily due to the following reasons:

• 1. The visual environment in the facility was strictly constant, with the same fluorescent lighting, furniture, types of activities, routines, and people. This consistency sharply contrasts with a rich and visually stimulating natural environment. The lack of novel stimuli and prolonged exposure to a monotonous environment was expected to negatively affect the flexibility of the visual system—that is, the crewmembers’ sensitivity to stimulus variability—and this may also have led to deficits in perceptual learning.

• 2. During the stay in the facility, the crewmembers were confined to a small living space with the notable lack of a visible horizon. The objects were not perceived relative to the position of a horizon, and the eye level thus remained undetermined. This situation may have critically influenced their 3-D perception, because information about eye level and the angle between the line of sight to the observed object and a horizontal line of sight is crucial to determining an object’s absolute distance (e.g., Ooi, Wu, & He, 2001; Sedgwick, 1986). In fact, not only angular declination could not be fully used in this visual environment, but also other perspective depth cues, such as linear perspective and texture gradient. Instead, the crewmembers’ 3-D perception may have become increasingly based on binocular disparity and other physiological cues.

A series of experiments were designed to explore both of these potentially afflicted areas of the crewmembers’ visual perception. However, in the present study, we focused solely on the topic of responsiveness to perspective visual cues, as examined by numerous, to some extent independent, tasks. Fully preserved 3-D perception would facilitate successful performance in some experiments and hinder performance in other experiments. Hopefully, the results from the individual experiments would provide a complex assessment of the possible changes in the crewmembers’ 3-D perception.

However, one needs to be aware of limitations when interpreting results that arise from the nature of the Mars-500 project. With the exception of Experiment 5, all of the experiments were computer administered. This means that the crewmembers estimated photographic rather than real-world stimuli. How might the restriction of Experiments 1–4 to photographic images have affected their depth judgments? Surprisingly, only little is known about the specifics of depth perception in photographs. Several studies have examined the accuracy of egocentric and exocentric distance perception. Kraft, Patterson, and Mitchell (1986) reported distortions in the depth judgments made when viewing photographs of natural scenes (see also Watanabe, 2006). Typically, sagittal depth is foreshortened, whereas lateral width is not. Koenderink, van Doorn, and Wagemans (2011) observed significant differences between the depth ranges obtained with various methods. Ohnaka and Matsuda (2008) found that the perceived distance between the camera and the object was mostly underestimated. Bengston, Stergios, Ward, and Jester (1980) showed that viewing photographs from too large a distance increases judged pictorial depth. On the other hand, no study, at least to our knowledge, has investigated the accuracy of perceived size (see also Stefanucci, Lessard, Geuss, Creem-Regehr, & Thompson, 2012).

The experience of long-term confinement was found to have very little or no effect on the crewmembers’ perceptual judgments. The crewmembers had no difficulties with fully perceiving depth in a scene, as was manifested by the impact of scene presence on 2-D length judgments (Exp. 1). Their 3-D size adjustments appropriately reflected the scale of the visual scene (Exp. 2). Also, information that was contained in the far array remained as salient as information in the near array (Exp. 3). Visual space was compressed in the in-depth dimension relative to the frontal dimension (Exps. 4 and 5; see also Baird & Biersdorf, 1967; Gilinsky, 1951; Loomis, Da Silva, Fujita, & Fukusima, 1992; Loomis & Philbeck, 1999; Loomis, Philbeck, & Zahorik, Loomis et al. 2002). How should these results be interpreted?

• 1. The first possibility certainly is that the data may depict the stability of perceptual performance throughout the mission—that is, the crewmembers’ 3-D perception may not have changed significantly, even after a year and a half in confinement.

• 2. Rather than unaffected 3-D perception, the data may reflect the crewmembers’ approaches to the experiments. Given that the crewmembers did not perform the tasks under the supervision of experimenters, and that each experimental task was conducted several (four to five) times during the isolation, the crewmembers may have found a way to accomplish the tasks even without 3-D perception (see also Point 3).

• 3. The conditions for conducting the experiments were far from optimal, for several reasons. Our experiments were never conducted individually. Rather, they were administered as part of a larger battery of tests and experiments scheduled each day. Thus, the simple fact of doing many miscellaneous tasks potentially affected the crewmembers’ performance. Factors such as poor motivation, fatigue, or problems with sustained attention could have resulted in less reliable responses.

• 4. The group sample was potentially problematic. We noticed that the crewmembers largely differed from the controls and from participants in other experimental studies. Even in the baseline session, the crewmembers had much shorter reaction times than the controls and presented typically differing patterns of data. Moreover, there were very large differences in performance between individual crewmembers, which may have been caused by the ethnic heterogeneity of a sample consisting of two Western Europeans, two Eastern Europeans, and two East Asians (one of the Russians was originally from Tajikistan). Notably, for example, the Asian participants gave markedly different responses in the bisection and change detection tasks than did the rest of the sample.