Exp Brain Res (1999) 129:378–390
© Springer-Verlag 1999
R E S E A R C H A RT I C L E
Jörg Sangals · Herbert Heuer · Dietrich Manzey Bernd Lorenz
Changed visuomotor transformations during and after prolonged microgravity
Received: 8 January 1999 / Accepted: 6 July 1999
Abstract A series of step-tracking experiments was conducted before, during, and after a 3-week space mission to assess the effects of prolonged microgravity on a non-postural motor-control task. In- and post-flight accuracy was affected only marginally. However, kinematic analyses revealed a considerable change in the underlying movement dynamics: too-small force and, thus, toolow velocity in the first part of the movements was mainly compensated by lengthening the deceleration phase of the primary movement, so that accuracy was regained at its end. The observed in-flight decrements in peak velocity and peak acceleration point to an underestimation of mass, in agreement with the re-interpretation hypothesis of Bock et. al. Post-flight no reversals of the in-flight changes (negative aftereffects) were found. Instead, there was a general slowing down, which could be due to postflight physical exhaustion. Key words Manual tracking · Microgravity · Visuomotor transformation · Adaptation · Human · Spaceflight
Introduction Manned space flight is well-known to be accompanied by problems of motor control. Movements have to be performed in an unusual force field and, thus, to obey Jörg Sangals (✉) Institut für Psychologie, Biologische Psychologie, Humboldt-Universität zu Berlin, Hausvogteiplatz 5/7, D-10117 Berlin, Germany, e-mail:
[email protected], Tel.: +49-30-20246787, Fax: +49-30-20246808 Herbert Heuer Institut für Arbeitsphysiologie an der Universität Dortmund, Dortmund, Germany Dietrich Manzey · Bernd Lorenz Institut für Luft- und Raumfahrtmedizin, Deutsches Zentrum für Luft- und Raumfahrt, Hamburg, Germany
other mechanical constraints than those that the motor system has been adapted to during evolution. In addition, the sensory systems that support motor control can change their response characteristics. This is obvious for the vestibular system, but also true for the visual system as well as for proprioception. Thus, prolonged microgravity results in changes of postural control (Kozlovskaya et al. 1981), and even short-term changes of the gravitational field reduce manual skills (Ross 1991). Overall, the change from a normal gravitational field to microgravity is so radical and – on the time scale of evolution – so new, that the basic motor capabilities that survive the change might be more remarkable than the disturbances. However, the disturbances point to the limits of motor control under extreme environmental conditions. Therefore, the study of motor disturbances under prolonged microgravity enhances our knowledge of the general capabilities of the motor system and does not only serve a practical purpose in itself. This perspective, of course, presupposes that the basic characteristics of the motor system do not radically change during prolonged microgravity, but rather that the available adaptive capabilities, which are needed for successful control of movements in the normal gravitational field, are exploited to deal as effectively as possible with the new challenge of microgravity, which has been unprecedented during evolution. Many motor tasks require that visual information is transformed into the appropriate motor commands, e.g., in reaching for an object. Bock et al. (1992, 1996a, 1996b) have summarized the factors that have been invoked in accounts of disturbances of such visuomotor transformations under changed-gravity conditions. These include mechanical, visual, and proprioceptive effects. Assuming that the direction of gravity, if it is present, is aligned with the mid-body axis, mechanical effects are obvious for movements parallel to the median plane (up and down movements), but should not exist for movements parallel to the transverse plane (neglecting the forces orthogonal to the movements). Nevertheless, Fisk et al. (1993) found essentially no differences between
379
movements in both planes performed with eyes closed under hypergravity and hypogravity during parabolic flights. This suggests that direct mechanical effects can be effectively compensated for. However, these findings might hold only for practiced movements of a certain distance and may not generalize to movements that are based on more flexible visuomotor transformations. Errors of visual localization of objects can result from changed oculomotor innervation patterns that are associated with a certain direction of gaze, and they would mainly affect the perception of the “height” of an object relative to the trunk (cf. Bock et al. 1996b). While egocentric location of an object could be modified under changed-gravity conditions, this is unlikely for allocentric location when both objects are imaged simultaneously on the retina, as when the hand approaches an object. In fact, Berger et al. (1997) suggested that visual feedback may gain importance for motor control during prolonged microgravity because vision is relatively little affected as compared with other sensory modalities. This is likely to be correct as long as vision is based on retinal information and does not involve information about eye position as in the perception of egocentric direction. Proprioception clearly is affected by a change of the gravitational field. There seem to be mainly two sources for these changes. The one is the change of torques required to perform certain movements. Even under normal circumstances, torque affects the judgment of limb position (Worringham and Stelmach 1985). Under hypergravity, Lackner and Graybiel (1981) found not only illusory perceptions of actively performed deep knee bends, but these were associated with felt movements of the ground; vision reduced the size of these illusions, but the felt movement of the supporting surface was seen as well. The second source is the change of vestibular input to spinal motoneurons. The gravity sensors of the inner ear have intricate relations to spinal motoneurons in the service of postural control (Wilson and Peterson 1981). According to Lackner and DiZio (1992), sensitivity of muscle spindles is also modulated by vestibular activity. Specifically, they found that vibration of upper-arm muscles induced stronger illusory movements under hypergravity than in a normal gravitational field and smaller illusory movements under hypogravity. The mechanical, visual, and proprioceptive effects that we have considered thus far are fairly immediate: active muscular forces have to be adjusted to the changed environmental force field or the changed environmental force field modifies perceptions that are at least partly based on sensors with response characteristics dependent on active and passive forces (like innervation of extraocular muscles or muscle-spindle activity). In addition, there are more indirect effects. For example, Berger et al. (1997) observed that the accuracy of visually guided arm movements was not reduced during spaceflight, but the movements were slowed down. They attributed the slowing to the development of a “slowmovement strategy”, which serves to reduce reaction forces that act on the trunk and could cause instability.
Watt (1997) reported a reduced accuracy when subjects pointed with the arm to memorized targets. This reduction was much stronger than in a condition in which the target, but not the hand, could be seen before each single pointing response. Thus, the maintenance of an accurate egocentric spatial map seems to suffer from the absence of gravity. In order to maintain a correct representation of the body orientation relative to the visual environment, occasional glimpses on the visual targets might serve to correct the errors that accumulate over time when information from gravity sensors is lacking. The factors described thus far, together with longterm changes of muscle characteristics such as muscular tone, can contribute to impaired motor performance in several tasks, but there are also tasks in which they do not seem to play an obvious role. Manzey et al. (1993, 1995, 1998) found a deterioration of tracking performance that is hard or impossible to account for in terms of the known mechanical, visual, and proprioceptive changes. The control device was a joystick, operated with the fingers. The movements were essentially parallel to the transverse plane, under normal conditions orthogonal to the direction of gravitational forces. Thus, direct mechanical effects as well as a change in proprioception that results from mechanical effects on receptor characteristics are unlikely to play a role for the tracking decrement. In addition, the muscles involved are not invoked by postural control; they are the main targets of the corticospinal tract, so vestibular influences should be weak or absent. Visual effects should play no role because successful task performance does not rely on egocentric visual localization; instead, it requires localization of one object relative to another. Indirect mechanical effects, as suggested by Berger et al. (1997), are unlikely because the mass of the fingers is too small for the reaction forces to represent a threat to balance, so a generalized “slow-movement strategy” is unlikely to be applied to these movements. Finally, the maintenance of a correct representation of the orientation of one’s own body relative to the visual stimuli was not required for successful task performance. However, under microgravity, more intricate proprioceptive changes may occur that could contribute to the impairment of tracking performance. Such changes had been suggested by informal reports of one of the participating cosmonauts, who claimed that the control device was “different” during the mission than it was on the ground. To explore the nature of the factors that might have contributed to the performance decrement in more detail, we modified the tracking task and adapted a task that we had used before for the study of visuomotor transformations (Sangals 1997; Heuer and Sangals 1998). Basically, this was a step-tracking task with continuous and terminal visual feedback during each single movement. A particularly intriguing hypothesis, from which specific predictions for our kind of task can be derived and which otherwise should be fairly insensitive to the effects of microgravity, is the “re-interpretation hypothe-
380
sis” of Bock et al. (1996b). The essence of this hypothesis is that gravity-related changes of weight are interpreted as changes of mass rather than as changes of gravitational force. We shall develop this hypothesis and its consequences for movement characteristics in some detail. Essentially, we conceive of this hypothesis as being based on the presupposition that there is no genuine adaptation to a deviant gravitational field, but rather that the adaptive processes which become invoked under microgravity are those that are needed for efficient motor control under normal gravity. Consider some basic adaptive capabilities of the motor system that are required in a normal gravitational field. There can be little doubt that the control of simple aiming movements involves both open-loop and closedloop processes (Cruse et al. 1990). One possibility to conceptualize the open-loop processes is in terms of a generalized motor program, which in turn can be represented as a prototypical force-time curve (e.g., Schmidt et al. 1979; Meyer et al. 1982). As noted by Kalveram (1991), a generalized motor program in the format of a prototypical force-time curve will be heavily distorted by the peripheral muscle-bone system when the mechanical parameters of the moving limb (friction, stiffness) are not properly neutralized. This is also true for the effects of gravitational force, which vary with the direction of movement and can also vary during a single movement. In fact, the variable effects of gravity are almost perfectly neutralized. For example, Virjii-Babul et al. (1994) found no differences between the kinematic characteristics of upward and downward movements, against and with gravitational force, while the muscular activity was patterned differently to take the different effects of gravity into account. Humans not only almost perfectly adjust to the variable effects of gravity on a moving limb, but also to variations in mass. For example, when objects are lifted, lifting force and grip force are adjusted to the mass, with a major part of the adjustment being predictive, based on visual information provided by the object (Gachoud et al. 1983; Gordon et al. 1991). However, the adjustment is slightly more complicated than a simple rescaling of forces in that the kinematic characteristics of lifts do not remain unchanged. Also, for horizontal movements with different masses, peak acceleration and velocity tend to decline and movement duration to increase when the mass is higher (Lestienne 1979; Gottlieb et al. 1989). Nevertheless, these changes leave the basic form of the velocity profile unchanged, that is, the effects of mass can be eliminated by proper scaling of time and velocity (Ruitenbeek 1984; Bock 1990). When the predictive adjustment is incorrect, rapid corrections occur, both in lifting a mass (e.g., Johansson and Westling 1984) and in moving it along a horizontal path (Smeets et al. 1995). The evidence for the existence of highly efficient predictive and feedback-based mechanisms to compensate for the variable effects of gravitational force and different masses could lead one to expect that these adaptive capabilities of the motor system could suffice for suc-
cessful motor control in changed-gravity fields, provided that there are no errors in visual and proprioceptive information. However, Bock et al. (1996b) have pointed to a problem that is specific to a changed gravity field and, thus, beyond the environmental conditions that have shaped the evolution of motor control. Objects, which can be one’s own limbs, with a certain mass have a certain weight under normal circumstances. While the relation between visual information and estimated mass of an object is flexible to some degree, because of its variable nature, this is unlikely for the relation between weight and estimated mass, because it is invariant due to the invariant gravitational acceleration, which implies a proportional relation between gravitational force (or weight) and mass. Therefore, if there is no genuine adaptation to a changed gravitational field, a change in weight, which results from microgravity (or hypergravity), should be interpreted as a change in mass. This is the essence of the re-interpretation hypothesis of Bock et al. (1996b). An underestimation of mass under microgravity, either because of a wrong weight cue or because of a lacking weight cue, should have consequences for movement characteristics as well as for the perception of mass. First, the parameters of a prototypical force-time curve should be adjusted incorrectly. According to the results on how different masses affect kinematic characteristics, acceleration should be too small and movement duration too short, so that the movement will undershoot the target, given that the true mass is larger than the one to which the system is scaled. Second, the actual acceleration of a mass will not be the one that can be predicted on the basis of force and the estimate based on the weight cue. The relation between active force and acceleration provides another possibility to estimate the mass, and the results of this estimate will deviate from the one based on weight, so that there is a kind of sensory conflict. There are findings both on mass discrimination and on motor control which are consistent with the re-interpretation hypothesis. We shall turn to mass discrimination first. Under changed gravity conditions, there is conflicting or reduced information about mass as derived from weight and from acceleration. Under such conditions, mass estimates should be more variable, as is typical for perceptual judgments that are based on conflicting or reduced cues (e.g., Heuer 1993), and mass discrimination should be impaired. This is indeed the case under hypergravity (Darwood et al. 1991) as well as under microgravity (Ross and Reschke 1982), and, during a 10-day period in space, Ross et al. (1984, 1986) found no clear evidence for adaptation. In addition to impaired discrimination because of conflicting information, perceptions should also be biased toward the mean mass indicated by the different cues. Thus, Ross and Reschke (1982) reported evidence that masses were judged smaller during short periods of hypogravity. Regarding motor control, the re-interpretation hypothesis leads to slightly counterintuitive expectations. For
381
example, under hypergravity, the feed forward control mechanisms should be preset for a mass which is too high. Thus, with a mass that is actually smaller than expected, the initial acceleration and, perhaps, peak velocity should be too high and movement amplitude too large. This is what Bock et al. (1992, 1996a, 1996b) observed. However, they also noted that the effects were smaller than expected from the re-interpretation hypothesis, which indicates the existence of corrections that will mainly affect the later parts of the movements. As we will discuss below, such corrections should correspond to those seen in other cases where masses are unexpectedly small or large. Bock and Cheung (1998) observed increased force levels under hypergravity, even when subjects produced isometric contractions to match certain percentages of their maximal voluntary force. Thus, adjustments of force levels could be a fairly generalized phenomenon in a situation in which all objects, including ones own body parts, seem to increase in mass, as judged from their weight. For microgravity, the re-interpretation hypothesis leads one to expect too small acceleration and velocity. Depending on how efficient corrective processes are, targets should be undershot. Neglecting corrections during each single movement, the effects of microgravity on undershooting as well as on other kinematic parameters should increase with larger movement amplitudes. This becomes evident from the following simplified considerations. Suppose that, in a normal gravitational field, the expected mass, me, equals the actual mass, m, and the prototypical force-time curve, F(t), is adjusted to achieve an expected acceleration-time profile, ae(t), as F(t)=meae(t) . The actual acceleration, then, is a(t ) =
F(t ) = ae (t ) m
because m=me. Now, under microgravity, m is larger than me, according to the re-interpretation hypothesis. Thus, with F(t) set to meae(t), the predicted actual acceleration will be a(t ) =
F(t ) me = a (t ), m m e
and the a priori unknown degree of mass underestimation me/m is given by a(t)/ae(t). [This estimate presupposes that a(t) and ae(t) do not differ in temporal scaling.] Several kinematic parameters, such as peak acceleration and peak deceleration, depend linearly on target amplitude of the movement (e.g., Hoffmann and Strick 1986; Virjii-Babul et al. 1994). Using the above-mentioned logic, one can predict a smaller slope of the linear functions that relate target amplitude to these parameters under microgravity (e.g., a smaller slope of the function relating peak acceleration to target amplitude). More specifically, the slope should be reduced by the factor me/m, since each peak acceleration for each amplitude is scaled by this factor. Similarly, the slope of peak velocity
and, also, of actual movement amplitude as a function of target amplitude should be reduced. The predictions that can be derived from the re-interpretation hypothesis for microgravity seem not to be consistent with experimental data. In particular, under microgravity, overshooting of targets instead of undershooting has been observed (Ross 1991; Bock et al. 1992). However, these observations have been made for movements in the vertical plane, parallel to the midbody axis. Under normal circumstances, the prototypical force-time curves for such movements need to compensate for the effects of gravity (e.g., Kalveram 1991; Virjii-Babul et al. 1994). In a changed gravitational field, these preprogrammed compensations might become malfunctional. Under microgravity, in particular, added forces, which are correct for a normal gravitational field, are likely to be too large, so that overshoots can easily occur. Under hypergravity, in contrast, added forces can be too small, so that the overshoots expected from the re-interpretation hypothesis could be turned into undershoots, as observed by Ross (1991) for example. As a matter of fact, the predictions are less than clear-cut for situations in which corrections for gravitational forces are required under normal circumstances because they depend not only on the deviation of the estimated mass from the actual mass, but also on the deviation of the estimated gravitational acceleration from the actual one. The effects of both deviations tend to counteract each other, so that the net effect depends on their relative sizes. For example, under microgravity, an upward movement would tend to undershoot the target, because of a too-low estimate of mass, but tend to overshoot the target, because the correction for gravity is likely to be based on a too-high estimate of gravitational acceleration. As a consequence, inconsistency of results is not fully unexpected. A closer examination of the changed-gravity situation reveals that final accuracy of movements may not be a very sensitive indicator for the effects of a changed gravitational field because additional factors come into play. In deriving specific predictions from the re-interpretation hypothesis, we have thus far assumed that force is scaled by expected mass to achieve an expected accelerationtime profile ae(t), which is invariant across different masses. This assumption, however, is a simplification. For a smaller expected mass, ae(t) should be compressed in time and enlarged in amplitude. This would correspond to the known effects of different masses: adjustments to different masses are only partly in terms of force scaling, they are also partly in terms of temporal scaling (Lestienne 1979; Ruitenbeek 1984; Gottlieb et al. 1989; Bock 1990). Thus, the reduction of peak acceleration, peak velocity, and peak deceleration should be smaller than the expectation that is only based on an unexpectedly large mass and neglects the temporal changes; in addition, it should be accompanied by earlier times of peak acceleration, peak velocity, and peak deceleration and, of course, smaller distances covered at these points in time. Similar considerations concern the effects
382
of movement amplitude. The scaling of measures like peak acceleration and peak velocity with movement amplitude tends to become steeper at smaller (correctly) expected masses, which could partly compensate for the reduction of slope expected under microgravity. The scaling of duration measures, in contrast, tends to be reduced at smaller (correctly) expected masses (cf. Benecke et al. 1985), so that a reduction of slope under microgravity would be predicted here as well. There are more factors which complicate the predictions that can be derived from the re-interpretation hypothesis. In particular, there are mechanisms that produce rapid corrections when an unexpectedly high mass is encountered (Simmons and Richardson 1992; Smeets et al. 1995). Such mechanisms will not affect the earliest parts of the movements, but later parts. For example, the duration of the deceleration phase of the movements tends to be increased more than the duration of the acceleration phase. Therefore, the predictions based on the reinterpretation hypothesis should hold for kinematic parameters that characterize early parts such as peak acceleration, time to peak acceleration, and distance at peak acceleration, but the effects should fade away as the movements approach their ends. This is basically what has been found for the first movement after an unexpected load change (Bock 1993), even without visual feedback being available (indicating corrections which are based on other than visual cues). What we have sketched is a way to deal with the effects of microgravity by exploitation of known adaptive and corrective mechanisms. Accurate movements, thus, do not presuppose any genuine adaptation to the changed gravitational field. However, although quick adaptive processes should permit accurate movements in our type of task, the dynamic characteristics of these movements should differ from those observed in a normal gravitational field: while the static characteristics of visuomotor transformations could be affected little or not at all by prolonged microgravity, the dynamic characteristics should be different. Our scenario of how the motor system could exploit its known adaptive capabilities to adjust to microgravity takes into account predictive open-loop processes and rapid corrective processes based on proprioceptive feedback. Of course, visual feedback could also contribute to corrections, and this contribution could gain importance under microgravity (Berger et al. 1997). To examine this possibility, we studied performance in the step-tracking task with continuous as well as terminal feedback; in the terminal-feedback condition, visual feedback was presented only when the subject did not move. Under both conditions, movements were required to be accurate in the end, guided by visual feedback. This was a prerequisite for the next target step to occur. However, our interest was not on final accuracy, but on the spatio-temporal characteristics of the primary submovements (until velocity became zero or until the end of the first deceleration phase), which ended before visual information became available again in the terminal-feedback condition.
While the re-interpretation hypothesis allows one to derive fairly specific predictions for the effects of prolonged microgravity, there are no clear predictions for the aftereffects. Typically, aftereffects of visuomotor adaptations are negative (cf. Welch 1978). Thus, one could expect that, post-flight, the expected mass is too high, corresponding to reports that, post-flight, the astronauts experience their own limbs as being too heavy (Ross et al. 1984, 1986). However, predictive force adjustments can be dissociated from variations in subjective heaviness of objects (Gordon et al. 1991).
Materials and methods Subject The subject of this study was a male cosmonaut, 40 years old. Before his space mission, he had never been exposed to microgravity, except during parabolic aircraft flights in his training for the space mission. Apparatus An IBM thinkpad portable computer served to control the experiment. It was equipped with a 12-inch color liquid crystal display. To register the movements of the subject, an aluminum joystick of 67 mm length, grasped with index finger and thumb, was attached to the right side of the keyboard panel with Velcro tape. The joystick lever could be rotated by 60° in the frontoparallel plane only. It could be rotated freely within this single degree of freedom, that is, neither was any specific lever position (e.g., the vertical) marked (visually or haptically), nor was there any elastic mechanism pulling the lever back to any specific position. Joystick motion was registered with an angular resolution of 0.23° (256 bits). The lever was mounted on top of a small aluminum box (59 mm wide, 59 mm long, 35 mm high) to allow the subject to rest his hand on it during the tracking task. During the experimental sessions, the subject maintained a sitting position with the portable computer being attached with tapes to his lap. Task The task was basically a step-tracking task. The subject had to superimpose a rectangular cursor of 3.3 mm width and 6.6 mm height as rapidly as possible on a target position, which was indicated by a vertical line and changed horizontally in discrete steps. The horizontal cursor position was linearly related to the joystick angle: a joystick displacement by 1° in the one direction led to a cursor motion of 2.81 mm in the same direction; when the joystick was in the central (vertical) position, the cursor was also in the central position on the screen. The required final accuracy of each positioning response was defined by the horizontal extent of the cursor rectangle. After each target step, the subject had to complete the positioning response within 4 s with as many submovements as necessary. The criterion for completion was that the cursor remained within the target area for 300 ms. After movement completion, the next target step was presented with a random delay of 0–150 ms. During this delay, cursor and current target remained visible without any indication of task fulfillment to prevent premature movements in the anticipated direction of the following target. There were two different feedback conditions, continuous and terminal feedback. In trials with continuous feedback, the cursor remained visible all the time. In trials with terminal feedback, the cursor disappeared timelocked with movement onset (defined by the smallest detectable change in joystick angle of 0.23°) and re-
383 30
raw data resampled & filtered
25
position [deg]
appeared only after the joystick movement had stopped for at least one sampling interval (16 ms). In order to avoid a flickering cursor display or staircase tracking strategies (to obtain quasi-continuous feedback), two additional criteria for cursor visibility were introduced. After disappearing, the cursor remained invisible for at least 500 ms, and, after reappearing, it stayed visible for at least 100 ms.
20
(a)
15 10 5
Data analysis Joystick position and cursor status (position and visibility) were recorded with a sampling rate of 62.5 Hz and a resolution of 256 bits, one binary corresponding to 0.23°. The positions were then resampled at a rate of 312.5 Hz using cubic-spline interpolation and, afterwards, low-pass filtered with a cut-off frequency of 15 Hz (Butterworth filter of second order, filtering was done in both directions to avoid phase shifts). The filtered joystick signal was differentiated twice to yield velocity and acceleration values; each of these signals was again low-pass filtered. Based on the resampled and filtered data, the following kinematic parameters were determined for each single movement:
0 200
(b)
150 100 50
2
acceleration [deg/s ]
0
(c)
2000 1000 0 -1000 -2000 -3000 0.0
0.1
0.2
0.3
time [s] 2 acc. [deg/s ] vel. [deg/s]
Each trial consisted of a series of 49 different target positions (or 48 target steps). The targets appeared in a pseudorandom order at eight different horizontal positions, which were evenly spaced at distances of 17.3 mm (corresponding to 6.2° of joystick movement) on the horizontal midline of the display, yielding seven target step amplitudes between 17.3 and 121.1 mm in either direction, thereby requiring joystick displacements between 6.2° and 43.2°. Each trial started with the leftmost position, followed by a displacement of the target to the right. The direction of the target displacement alternated from step to step, whereas the amplitudes were determined by a random process in order to balance the coarse amplitudes of target displacements within a trial (Sangals 1997). Therefore, the direction of each target step was predictable, but not its amplitude. The randomization was based on four target zones, each being defined by a pair of two adjacent target positions. Each target step involved a change of the target zone, and the frequency of a step by either one, two, or three zones was identical within a trial. For a given target zone, the position within the zone was then determined randomly. Due to the fact that steps between adjacent targets within the same target zone were not possible, steps of width 173 mm occurred only about half as often as the other amplitudes. The same was true for the maximal step amplitude, because it could be reached from two positions only (in contrast to all other positions, which could be reached from at least four different positions). Data were acquired during 22 experimental sessions, each consisting of four trials with a fixed order of feedback conditions: terminal, continuous, continuous, terminal. The experimental sessions were run on different days before, during, and after the space mission. Seven pre-flight sessions were recorded 57, 56, 55, 20, 11, 10, and 7 days prior to mission launch. The early start of recording and the corresponding temporal spread of pre-flight sessions was caused by several unexpected delays of the mission. The pre-flight data acquisition phase was preceded by a training period of 14 sessions during five consecutive days, and six additional training sessions 47, 31, and 26 days before launch, which were inserted into the pre-flight recording period to maintain training level during the delays of the mission. During the 21-day space mission, recordings were made on mission days 4, 5, 7, 10, 16, and 18. Another six sessions were recorded 1, 2, 4, 6, 10, and 12 days after return to earth. Finally, a follow-up sequence of three sessions 72, 73, and 74 days after return was recorded. Besides the tracking experiment, the cosmonaut was subjected to a series of 27 materials science, and life science experiments. (Sahm et al. 1996).
velocity [deg/s]
Design and procedure
250
(d)
200 150 100 50 0 2000 1000 0 -1000 -2000 -3000
(e)
0
5
10
15
20
25
position [deg] Fig. 1a–e Sample recording of a single movement. Position-time curve (a), velocity-time curve (b), acceleration-time curve (c), velocity-position curve (phase-plane plot) (d), and acceleration-position curve (e)
peak acceleration, time to peak acceleration, distance at peak acceleration, peak velocity, time to peak velocity, distance at peak velocity, peak deceleration, time to peak deceleration, and distance at peak deceleration. All latency measures were defined with respect to movement onset. In addition, the duration and the amplitude of each primary submovement was computed similarly to the procedure reported by Meyer et al. (1988), who defined the end of the submovement as the first point in time at which one of the following criteria is fulfilled: (1) velocity decreases to zero, (2) acceleration changes from negative to positive, or (3) jerk sign changes (indicating a local minimum of acceleration or a local maximum of deceleration). Unlike Meyer et al. (1988), we used only the first two critera to define the end of the primary movement, because jerk was too noisy due to the limited sampling frequency of the joystick positions. Figure 1 shows a sample recording of a single movement in which distance (a), velocity (b), and acceleration (c) are plotted against time. The lower graphs (d) and (e) show velocity and acceleration as a function of position. The kinematic variables essentially represent the values on the abscis-
384 sae and ordinates in these different plots at the minima and maxima of acceleration and velocity. In total, we analyzed 22 sessions of four trials each. Each trial consisted of 48 target steps and 48 corresponding movements. The last movements of all trials were omitted in the statistical analyses because, in several instances, they were incompletely recorded. For the other movements, the kinematic parameters were screened for abnormalities (such as initially wrong directions of a movement, no orderly succession of acceleration and deceleration phase). In total, 4.0% of deviant recordings (of a total of 4136) were identified. As described above, both extreme target distances occurred approximately half as often as the intermediate amplitudes, which were similarly frequent. For the seven distances, in the order from short to long, the actual frequencies of analyzed movements were 276, 635, 704, 641, 656, 671, and 386. There were two options for the unit of statistical analysis. The first option was the individual movement. In this case, the variation across movements within a single trial is considered as random error. We rejected this option for a simple reason. It turned out that there were a number of statistically significant differences between the seven pre-flight sessions. This indicates that there were reliable day-to-day fluctuations in performance even under normal circumstances. As a consequence, if we had compared performance in-flight with some baseline performance, statistically significant differences could have resulted from such fluctuations and not from the effects of microgravity. Therefore, we chose the individual test session as the unit of statistical analysis, which implies that the day-to-day fluctuations in performance are considered as random error against which the effects of microgravity are tested. The advantage of this procedure is that the risk of spuriously significant findings is reduced, but the cost is that short-term influences during single sessions will be missed. For each session and each target distance, we computed the medians of the kinematic variables, separately for the two trials with continuous feedback on the one hand and the two trials with terminal feedback on the other hand. For the statistical analyses, we grouped the 22 sessions into six periods: seven pre-flight sessions, the first three in-flight sessions (in-flight 1), the second three in-flight sessions (in-flight 2), the first three post-flight sessions (post-flight 1), the second three post-flight sessions (postflight 2), and the three follow-up sessions. Thus, the numbers of sessions in the various periods were small, so that only highly consistent effects could be discovered.
Results We consider first the amplitude and duration of the primary submovements and, thereafter, the kinematic parameters as the movements evolved from start to end. Each of the 11 kinematic parameters was subjected to a three-way analysis of variance (ANOVA) with the between-session factor Period and the within-session factors Feedback Condition (terminal vs. continuous) and Target Amplitude. The between-session factor Period was split into four contrasts: each of the two in-flight periods and the two post-flight periods was compared with the combined pre-flight and follow-up periods. It turned out that there were only very few differences between conditions with terminal and continuous feedback. These will be discussed separately after the effects of period and target amplitude have been described for the various kinematic parameters. Each of the kinematic parameters scaled with target amplitude, and the slightly trivial main effects in the ANOVAs will be skipped.
Fig. 2 Amplitude (a) and duration (b) of primary submovements as a function of test period, shown separately for the seven target amplitudes (target amplitudes are marked by horizontal lines)
Amplitude and duration of primary submovements In Fig. 2a, the mean session medians of the amplitudes of the primary submovements are shown separately for each period and target amplitude; the target amplitudes are marked by horizontal lines. In general, the amplitudes of primary submovements fell short of the target amplitudes, more so for more distant targets, and not at all for the smallest target amplitude. There were no systematic differences between periods; only for the second in-flight period did the difference to pre-flight and follow-up amplitudes interact with target amplitude: F(6,97)=3.5, P
