Seven right-handed (3 women and 4 men, mean age = 39 ± 6.9 years) participated in the experiment on a voluntary basis. They had no prior experience of microgravity exposure. As the present study is part of a scientific program studying human motor behavior in different force fields, participants were previously tested in normogravity and underwater for the same task as reported in Macaluso et al. (2016). None of the participants suffered from neuromuscular or sensory impairments, as confirmed by a medical examination prior to the experiment. Vision was normal or corrected by lenses. Before microgravity exposure, the participants received comfort medication (scopolamine) to avoid motion sickness. It has been demonstrated that its use for parabolic flights did not induce neuromuscular side-effects on sensorimotor control (Ritzmann et al., 2016). All the participants were naive as to the specific purpose of the experiment, which was authorized by the ANSM (French National Agency for Biomedical Security) and approved by the Committee for the Protection of Persons concerned (CPP). The participants gave their signed informed consent prior to the study in accordance with the Helsinki Convention.

Circular targets were presented in front of participants standing upright and maintained to the ground structure by means of foot-straps (Figure 1A). They had to press their right index finger on the start push-button positioned alongside their body. The height of the start push-button was adjusted to each participant's height for initial posture standardization. Circular targets were oriented along the frontal plane and were positioned relative to participants' anthropometric features. Close targets were located at shoulder's height (i.e., the height of the target center corresponded to the horizontal projection of the height of the acromioclavicular joint in the sagittal plane) at a distance corresponding to arm length, allowing the participants to reach these targets without trunk displacement. Far targets were located 25 cm away and 20 cm below the close targets: in that case, participants had to perform additional trunk displacement to reach these targets (Figure 1B). For both target locations, the diameter was also manipulated through Light-Emitting Diodes (LEDs) equally distributed to define two target sizes: Small targets 4 cm or Large targets 10 cm (Figure 1C). Therefore, in this experiment, combining location and size corresponded to the presentation of four targets: CS (Close–Small), CL (Close–Large), FS (Close–Small), FL (Far–Large). Switching targets on and off were achieved by a homemade software (Docometre©) piloting a real-time acquisition/control system running at 10 kHz (ADwin-Gold, Jäger, Lorsch, Germany).

Markers were positioned onto the participants' index, shoulder and hip. Markers position was recorded (i) in normogravity with a video motion capture system (LED-type markers) composed of three cameras sampled at 60 Hz (resolution: 848 × 480 pixels); (ii) in microgravity by an optical motion capture system (infra-red active markers) at 200 Hz (Codamotion CXS and Active CodaHub, Charnwood Dynamics Ltd, Leicestershire, UK). Importantly, both acquisition systems yielded a similar accuracy in the definition of markers' position. Indeed, the data acquired in NormoG by the video motion capture system were processed using Direct Linear Transformation (Abdel-Aziz and Karara, 1971), to reach the accuracy level of the optical motion capture system used in MicroG (i.e., millimeter order). Moreover, according to the sampling theorem (Shannon, 1949), the sampling rates used in both environments are known to be sufficient to capture the whole range of velocities associated to biological motion, including fast reaching movements (Song and Godøy, 2016).

All participants were exposed to two environments: first in normogravity (“NormoG”) before the parabolic flight campaign, then in microgravity (“MicroG”). The MicroG environment was achieved in the A-310 ZERO-G aircraft chartered by the French Centre National d'Etudes Spatiales (CNES) and Novespace for parabolic flight studies, during the campaign #125, including 3 days of flight. For each flight, the aircraft ran a sequence of 30 parabolas. Parabolic maneuver was composed of three distinct phases: 24 s of hypergravity (1.8 g, pull-up phase) followed by 22 s of microgravity (0 g) before a second period of 22 s of hypergravity (1.8 g, pull-out phase). Each parabola was separated by 1 min of normogravity (1 g, steady flight phase).

Positions of the start push-button and the targets were adjusted for each participant, then an initial calibration of targets was performed along the Z vertical axis (i.e., defining positions relative to arm movement elevation). Before each trial, participants had to stand upright, the arms outstretched along the body, and the right index pressing the start push-button. When one of the targets was illuminated, participants were asked to perform a reaching movement toward the target while keeping the arm outstretched. Reaching movements had to be performed as fast as possible while primarily respecting accuracy constraints related to the target area. Each trial was validated when the index fingertip reached the target. The final position had to be maintained until target extinction (3 s after movement onset) which prompted the participants to return to the starting position.

Participants performed 10 pointing movements toward each of the four targets for a total of 40 trials per experimental session in each environment. In the MicroG environment, these 40 trials were presented during 10 successive parabolas for each participant, thus including four trials per parabola. The targets were presented in a pseudorandom order, which was counterbalanced between the participants. Each session included three specific blocks of four trials in which the order of target presentation was the same. These blocks were presented in the initial, middle and final part of the session (corresponding to the 1, 5, and 10th parabola in MicroG) to assess the potential evolution of motor performance during each session, which lasted about 25 min.

Data presented below describe behavioral features of reaching movements in the sagittal plane and some of them are detailed in Macaluso et al. (2016). First, we analyzed the fingertip trajectory, success rate (index fingertip within a given target area), index final deviation from target center, reaction time (RT), movement duration (MD), and mean tangential velocity (Vmean_endpoint). The index final deviation was measured as the mean absolute distance of the final position of the index fingertip relative to the target center along the Z vertical axis. For each trial, the time elapsed between target illumination and the release of the start push-button by the participants defined RT. Index position in the sagittal plane was filtered (digital second-order dual-pass Butterworth filter; cutoff frequency 6 Hz in NormoG and 10 Hz in MicroG) and differentiated to obtain the endpoint tangential velocity in m.s⁻¹. Regarding the different sampling rates of acquisition systems used in both environments, we found that the cutoff frequencies mentioned above were the most suitable to reflect the raw data in normo and microgravity. The movement onset was defined as the time when the index tangential velocity reached 1.5% of its peak. Conversely, movement end was defined when the tangential velocity dropped below 1.5% of its peak.

The focal component of whole-body reaching movements was analyzed by considering the arm angular elevation over time (i.e., angle evolution of the extended arm around the shoulder with respect to its initial orientation). Arm angular elevation was computed from the index and shoulder XZ raw data, filtered (digital second-order dual-pass Butterworth filter; cutoff frequency 6 Hz in NormoG and 10 Hz in MicroG) and differentiated to obtain the arm angular velocity profile. From this velocity profile, the peak velocity (PV_ang in deg.s⁻¹) and the relative angular deceleration duration (rDD_ang, defined as the duration between PV_ang and movement end, expressed in % of movement duration to facilitate comparison between both environments) were extracted. Arm angular velocity profile was also differentiated to obtain arm angular acceleration profile, informing on early changes in motor execution which may give an insight upon the planning stage of focal movement. From this acceleration profile, peak acceleration (PA_ang in deg.s⁻²) and time to peak acceleration (TPA_ang expressed in ms to precisely estimate the occurrence of motor changes) were extracted.

In parallel, the postural component involved in the whole-body reaching movements (particularly to reach the far target) was analyzed by considering trunk displacement. This latter was illustrated by the final angular position of trunk (hip-shoulder segment) relative to vertical (β_ftrunk: trunk flexion in deg) at arm movement end, and by the forward displacement of participants' shoulder and hip (translation along the horizontal plane in mm). Shoulder and hip movement onset/end in the sagittal plane were defined as the time when the tangential velocity respectively reached/dropped below 1.5% of its peak.

Statistical analyses were based on mean comparisons. Repeated-measures analyses of variance (ANOVAs) were performed to compare the means of kinematic parameters mentioned above after having ensured that the assumption of normality was not violated (Kolmogorov-Smirnov test). Newman-Keuls tests were used for post-hoc analyses and the significance threshold was set at.05 for all statistical tests.

We conducted prior analyses to investigate the potential learning effects during a single session (40 trials). Repeated-measures ANOVAs including 2 Environment (NormoG, MicroG) × 2 Target Location (Close, Far) × 2 Target Size (Small, Large) × 3 Block (Initial, Middle, Final) were initially performed on all the selected parameters of the study. The results did not show any significant main effect of Block or any interaction with the other factors (p > 0.05). To specifically exclude the presence of any adaptive processes in MicroG environment, we conducted complementary analyses comparing a specific set of trials occurring during the 1, 5, and 10th parabola (see Material and Methods). Repeated-measures ANOVAs including 3 Parabola (1, 5, and 10th parabola) × 4 Target Presentation (CS, CL, FS, FL) did not reveal any significant main effect of Parabola or any interaction with the other factor on all the selected parameters (p > 0.05). Thus, the reported values did not significantly change throughout the experiment.

First of all, we investigated arm movement toward the targets in each environment. Figure 2 illustrates mean endpoint trajectories (i.e., index fingertip) in the sagittal plane observed for a typical participant when reaching close and far targets. It shows that spatial characteristics of endpoint motion were impacted by the microgravity environment.

Although the participants never experienced microgravity exposure before the present experiment, we did not find any significant evolution in the reported variables across the successive trials. Thus, we failed to show the presence of sensorimotor adaptation during the experiment which would indeed have led to more progressive changes across the repetition of reaching movements (Lackner and DiZio, 1994; Shadmehr and Mussa-Ivaldi, 1994). Rather, we observed a prompt reorganization of some movement features (see section Microgravity Is Accounted for into the Planning of Focal Movement) which took place at the earliest onset of exposure. From previous work conducted in parabolic flights, the occurrence of adaptive processes on reaching movements is not clear. Indeed, some studies reported slow progressive changes of kinematics across parabolas (Papaxanthis et al., 2005; Gaveau et al., 2016) whereas others observed rapid behavioral stabilization or no significant change during exposure (Patron et al., 2005; Bringoux et al., 2012; Casellato et al., 2012). In our study, one hypothesis related to the parabolic flight context can be advanced to explain this immediate stabilization of motor behavior. Before the 30 parabolas achieved for experimental acquisition, the aircraft performed one parabola to allow participants discover the parabolic maneuver. Moreover, since 3 participants were tested during each flight (see Material and Methods), two of them had even more time to experience microgravity exposure. Although we ensured that no reaching movements were performed by the participants out of the experiment, this preliminary although short exposure before data acquisition would enable the participants to develop prior expectancies about how it feels to move in these novel environments. Moreover, microgravity episodes of parabolic flights induced a global modification of the force field applied to the whole-body before initiating each trial. Thus, in this context, the participants accessed the new dynamic properties of the environment prior to movement onset (Barbiero et al., 2017) which might be sufficient to rapidly update their internal model for sensorimotor planning and execution, hence leading to an immediate motor reorganization (Wolpert and Kawato, 1998; Wolpert and Ghahramani, 2000). It has been shown indeed that the initial state of the sensorimotor system is primarily used to adjust the internal representations necessary to perform upcoming movements (Starkes et al., 2002; Flanagan et al., 2006; White et al., 2012, Rousseau et al., 2016). Here, the limb proprioception could contribute to detect the gravity release at the level of muscles and joints, through muscle spindles and Golgi tendon organs identified as load receptors related to gravity force field (Dietz et al., 1992). The following sections aim at discussing the stabilized motor features observed in microgravity for whole-body reaching.

The high spatiotemporal constraints of execution in the present study enabled us to investigate functional whole-body reaching movements in microgravity. In this line, our data did not reveal any significant difference between MicroG and NormoG environments in terms of movement duration, mean and peak velocity during movement execution. The absence of effect of the environment on these variables may reflect a tendency to keep the average movement speed in the range of normogravity experience. To that aim, the participants may have reduced the safety margin related to the final reaching accuracy in microgravity. Indeed, while the preservation of movement speed was not detrimental to reaching performance (i.e., the high success rate observed in MicroG, > 95%, was not significantly different from NormoG), the distance between the endpoint final position and the target center was significantly higher in MicroG when reaching large targets. Thus, in this task, the participants tended to maintain speed over accuracy margin (Woodworth, 1899) for a still successful performance without gravity. Keeping the average speed and reaching performance within normogravity standards were though not at the expense of movement preparation duration, since the reaction time remained also unaffected by the environment. Hence, alleviating gravity before movement execution did not impact the time allocated for motor planning. Nevertheless, we will detail in the following parts some evidence for substantial qualitative reorganizations, notably in focal and postural components of the reaching movement, which helped maintain the functionality of motor behavior in microgravity.

On Earth, kinematics of arm movement elevation has been well-described in terms of asymmetric bell-shaped velocity profiles (Papaxanthis et al., 1998; Gentili et al., 2007). Classically, the relative deceleration duration appears longer than the relative acceleration duration, suggesting that gravity is accounted for during motor planning to act as an assistive force for decelerating upward movements (Papaxanthis et al., 2003; Gaveau and Papaxanthis, 2011). The way gravity is integrated into motor planning has been recently formalized by a Minimum Smooth-Effort model (Gaveau et al., 2016), in line with the optimal control theory minimizing absolute work and jerk (Berret et al., 2011; Gaveau et al., 2014).

In our experiment, the substantial increase of the relative deceleration duration in MicroG constitutes the most salient feature of motor reorganization concerning the focal part of the reaching movement. The asymmetry was thus notably amplified in microgravity without changing the amplitude of peak velocity. Such reorganization, consistent across subjects as shown in the Supplementary Figure 1A, appears as a direct consequence of an earlier peak acceleration observed during motor execution (~47 ms). Recent work established that the shortest feedback-based corrections of EMG (electromyographic) patterns during arm reaching occur at ~60 ms when considering a limb disturbance with no changes of target location (Scott, 2016 for a review). We therefore hypothesize that the kinematic changes promptly observed in microgravity following arm movement onset are based on feedforward control mechanisms, directly expressed in the motor intention (Gaveau and Papaxanthis, 2011). In other words, we argue that the CNS could predict the effect of gravity release on the moving segments and could subsequently integrate the novel dynamics associated to a weightless environment into motor planning. As there is no external force to help braking upward movements in microgravity (neglecting the air friction forces), the participants had to actively counteract the inertial force of their moving limbs, presumably by increasing the antagonist muscle activations (Bonnard et al., 1997). In this context, a longer deceleration phase may reflect a greater use of feedback processes (Chua and Elliott, 1993; Sarlegna et al., 2003; Terrier et al., 2011). This greater retroactive control would enable the participants to better manage the speed reduction of their reaching movements, especially when approaching the target, to maintain final accuracy. Increased asymmetry between acceleration and deceleration phases was also reported when removing gravitational shoulder torque before arm movement onset (Rousseau et al., 2016) and with additional loads placed on the arm (Gaveau et al., 2011). Such reorganization in kinematics may thus illustrate a cautious strategy accounting for force/inertia uncertainties in unusual context. In this line, previous studies demonstrated that the lack of information prior to movement onset strongly affects motor planning (Bringoux et al., 2012; Rousseau et al., 2016), presumably to face unexpected or erroneous sensorimotor estimates during subsequent movement execution in unfamiliar environments (Brooks et al., 2015).

Contradictory with the present findings, some other studies reported a progressive disappearance of asymmetric velocity profiles in microgravity (Papaxanthis et al., 2005; Gaveau et al., 2016). Unlike our experiment, movement accuracy was not a primary constraint in these previous work where the braking phase of arm movements was not crucial during motor execution to correctly perform the task. Alternatively, when the participants had to perform reaching movements “as accurately as possible” in microgravity, Bringoux et al. (2012) also observed a longer deceleration phase as compared to normogravity exposure. Interestingly in our study, such reorganization of motor planning for arm reaching in MicroG was not detrimental to movement duration: the longer deceleration phase was compensated by higher peak acceleration in microgravity. This compensatory increase of peak acceleration may also represent a specific reorganization of the movement in a given environment as it likely exploits the absence of gravity torque at movement onset to efficiently trigger the initial impulse. Additionally, as discussed in the following section, the planning of focal movement was not the only component to be modified during whole-body reaching movements in microgravity.

Kinematics collected from the trunk clearly supports two different postural strategies as a function of the gravity environment while reaching far targets placed beyond arm's length. Under normogravity, our analyses revealed a significant forward trunk bending expressed by large shoulder displacement associated to very small hip displacement in space. This feature is typical of a “hip strategy” (Horak and Nashner, 1986), through which the postural component supporting the focal part of movement is also used to prevent falling (Massion, 1992). This posturokinetic organization would thus reduce the displacement of the CoM by using compensatory mechanisms (Massion, 1992; Vernazza et al., 1996) and would favor equilibrium maintenance at the expense of mechanical energy minimization and joint smoothness maximization (Hilt et al., 2016).

Alternatively, the second postural strategy specifically observed in microgravity was illustrated by very small trunk bending associated to larger shoulder and hip displacement from vertical. This organization would resemble the “ankle strategy” evoked by Horak and Nashner (1986), though with greater whole-body forward displacement. In MicroG environment, the participants were indeed not constrained by gravitational force, allowing for a vertical CoM projection outside the base of support. This observation is consistent with others reporting that postural control in weightlessness is predominantly managed at the ankle level (Clement et al., 1984; Clément and Lestienne, 1988). On Earth, this posturokinetic strategy decreases the equilibrium safety margin but the risk of falling is greatly minimized in microgravity. The participants might therefore adopt the strategy which would allow them to reduce the degrees of freedom (Bernstein, 1967), helping minimize the mechanical energy expenditure and maximize joint smoothness (Hilt et al., 2016). In line with the optimal control theory (Berret et al., 2011; Gaveau et al., 2014), the combination of these cost functions would enable the postural component to support more efficiently the focal part of the reaching movement in weightless environment. Despite methodological differences with our study, Casellato et al. (2012) also reported whole-body forward displacement when performing unconstrained bimanual reaching (i.e., natural pace and uncontrolled accuracy). In line with their observations, our data may also support the existence of an “oversimplification” of postural control to perform a functional behavior when facing high spatiotemporal constraints of execution in microgravity. Moreover, unlike previous observations of Casellato et al. (2016) in long-term weightlessness, the postural strategy observed in the present study was not subjected to large inter-individual variability. Overall, individual trends for both focal and postural observations are clear and systematic (as illustrated in the Supplementary Figure 1). Remarkably, the stabilized motor features observed in microgravity in the present study appear close to those previously reported when participants performed the same task underwater with neutral buoyancy applied to body limbs (Macaluso et al., 2016). The following section discusses the behavioral analogies observed in both environments.

As the present study is part of a scientific program studying human motor behavior in different force fields, the same participants were previously tested underwater in the same task for comparison purpose. Specifically, they were immersed in a prototypical submersible simulated space suit (AquaS environment; Macaluso et al., 2016), to apply neutral buoyancy at the level of body limbs. As in the present study conducted 2 years later, we did not find any significant evolution in the reported variables across the successive trials performed underwater. Rather, we observed immediate reorganizations at the earliest onset of exposure excluding the presence of adaptive processes during the experiment. As in MicroG, AquaS environment also implied initial exposure before data acquisition related to the installation of participants on the pointing structure. Thus, participants were submitted to global modifications of the force field applied to the whole-body before trial execution. This observation extends the hypothesis provided in section Prompt Reorganization of Motor Behavior in Microgravity. When participants accessed the new “unweighting” properties of a given environment before performing the first reaching movements, they could promptly reorganize their motor behavior. Most interestingly, the changes of focal and postural components of reaching movements in MicroG are close to those observed underwater in AquaS. Indeed, the increase of the relative arm deceleration duration and the decrease of trunk flexion when reaching far targets appear strikingly comparable (see the Supplementary Figure 2). In other words, the participants adopted analogous temporal structure of arm movements and almost similar postural strategy to perform whole-body reaching movements in these different environments. In so far as these two parameters are known to be gravity-dependent kinematic markers (see Introduction), and as AquaS and MicroG environments attempted to reproduce a weightless context, we hypothesize that these very close motor strategies would be mainly due to whole-body unweighting. It suggests that a fine control of neutral buoyancy underwater across the whole-body segments would tend to better simulate microgravity when considering the execution of sensorimotor tasks. Further studies are obviously required to challenge this hypothesis, especially to better investigate the effects of viscous force on motor control.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.