1. Introduction

We present an approach to one set of physiological issues that will be critical on any long-duration space flight, especially one requiring precision sensory and motor skills on the part of the crew upon arrival at their destination. It is essential that the crew be able to function effectively when first arriving at (for example) Mars, and hence this report pertains to the transient phase upon arrival, rather than to the extended portion of the stay when other long-term adaptive mechanisms will be effective. Due to limited opportunities for sensorimotor experimentation on long-duration flights to date, this is in part a conceptual paper, but nevertheless supported by reports of initial experimental findings from laboratory studies and parabolic flights. We hope to provide a basis for considering the crucial issues in terms of sensorimotor performance on a Mars mission, how best to implement adaptive strategies, and how a supporting ground-based research plan might be designed.

2. The Problem

Any human trip to Mars will be long and complex – minimum six months transit each way plus several months on the surface. Space fights of this duration entail a variety of physiological and psychological challenges: bone loss, cardiovascular and muscular deconditioning, radiation exposure, confinement and isolation. Sensorimotor disturbances also occur, such as impaired hand-eye coordination, balance disturbances, and disorientation. These are most acute during transitions between g environments (e.g., planetary to 0g), but some effects are manifest for weeks or months and lead to maladaptive effects upon later return to a gravitational environment. It is these sensorimotor effects that we address, in particular those related to the balance system.

The vestibular system is the non-auditory part of the inner ear. It consists of specialized structures that sense head rotation (semicircular canals) and linear acceleration (otolith organs), and uses this information on head and body motion to maintain accurate perception of spatial orientation and to generate responses necessary for stable vision and posture. The otolith organs respond to linear acceleration, both tilt relative to gravity and actual translational motion. On Earth, gravity provides a constant and fixed "down" reference, which is sensed constantly by the otolith organs. Thus the various body functions related to spatial orientation and movement – such as walking, maintaining upright posture, recovering from falls – rely on this gravity reference. Prolonged exposure to microgravity during long-duration space flight introduces problems for these systems, since the gravity reference is no longer present. As a result, many sensorimotor responses no longer function normally. As one example, consider the pitch vestibulo-ocular reflex. The vestibulo-ocular reflex (VOR) is a reflexive eye movement that is made when the head moves, to drive the eyes in the head so that they remain stable in space and so can maintain gaze on a fixed point. It is the reflex that, for example, enables someone to read while walking: head motion is sensed by the vestibular system, which moves the eyes so that they remain fixed on the text. When the head movements are made in the pitch plane (as if nodding "yes"), the head continually changes orientation with respect to gravity, and so this response might be adversely altered when gravity level is changed. In particular, we might expect a decreased contribution to the VOR in low g levels and an increased contribution in high g levels. Evidence that this occurs comes from cases in which VOR magnitude is decreased early in orbital flight, upon initial exposure to 0g. As adaptation to 0g takes place, the brain adjusts and learns not to expect a contribution to the VOR from the otolith-sensed gravity component, and the VOR returns to normal. Then, upon first returning to Earth, the now-unexpected gravity contribution produces too large a gain, with recovery to baseline level after several days of re- adaptation Clément et al., 1986)Clément et al., 1986). Concurring results have also been found on a shorter time scale in the low and high g phases of parabolic flight (Karmali and Shelhamer, 2010).

These and other studies demonstrate serious disruptions in sensorimotor function during long- duration flight (Reschke et al., 1997), most notably observed in postflight testing (Bacal et al., 2003; Black et al., 1999; Layne et al., 1997; Reschke et al., 1998; Speers et al., 1998). While there is little or no hard evidence relating sensorimotor deficits to astronaut performance in flight, there are reports from astronauts of hypersensitivity to pitch head movements (Oman et al., 1990; Thornton et al., 1987), and pitch rotation has been reported to disrupt re-adaptation to Earth gravity (Black et al., 1999). A correlation has been noted between decrements in shuttle landing performance and impaired performance in a postflight assessment test that requires the astronaut to rise from a chair and begin walking (Clark, 2002; McCluskey et al., 2001); both tasks entail pitching head movements, and therefore an accurate pitch VOR. Additionally, the head must pitch up and down to look at instruments and controls outside the vertical oculomotor range. With each movement, a synergistic combination of eye and head motions must be made, so that the eyes reach the intended target rapidly and accurately. The effects on the VOR, combined with the tendency to restrict head movements in unusual gravity environments (Oman et al., 1990; Oman et al., 1986) and the likely change in head-movement dynamics in different g-levels (i.e., overshooting in high g), may well combine to produce serious disruptions in eye- head control during pitching motions. This behavior is thus an example of one that could be amenable to the countermeasure approach described below. Although we will not discuss pitch VOR specifically, we do address gravity-mediated effects and vestibulo-ocular function.

3. A Solution

While many sensorimotor processes adapt to changes in g-level, the adaptation process itself can take time (e.g., several days for overt neurovestibular responses (Reschke et al., 1997; Young et al., 1984)). A Mars mission will require transitions between microgravity and planetary- gravity levels, and possibly multiple transitions to and from artificial gravity. These adaptation/ readaptation cycles can present serious operational concerns if insufficient time is available for each adaptation process. Thus context-specific adaptation has been proposed as a possible countermeasure for dealing with these g-level transitions (Shelhamer and Zee, 2003).

Context-specific adaptation (CSA) refers to the ability of an organism to: 1) maintain two different adapted states for a particular response, 2) have each state associated with a specific context state, and 3) switch between the adapted states immediately upon a change in context (without de-adaptation and re-adaptation). One way to think about this is as switching between previously learned "sensorimotor programs" (adaptation states). By themselves, context cues do not normally influence the response states, but with sufficient training, they can enable recollection of the particular state. Veteran astronauts experience fewer and less severe symptoms of adaptation to space flight on subsequent flights (Reschke et al., 1997), possibly because they have associated the appropriate adaptive state with the contextual gravity cue: they may store a sensorimotor program that is appropriate for 0g and one that is appropriate for 1g, and switch between them based on instantaneous g level.

Drawing on this, we discuss CSA as a countermeasure for some of the sensorimotor disturbances of long-duration flight. Considerations of CSA can be used to guide pre-flight training in such a way that gravity-appropriate sensorimotor adaptation states are stored and retained for the duration of an extended flight, with appropriate recall of the adaptation states when needed upon return to a gravity environment (Mars, asteroid, return to Earth). Recommendations can also be made for in-flight activities to maintain adaptation and reduce interference that might diminish adaptation. These considerations are described in more detail in the summary section below.

We use saccadic eye movements as a model system for some of our investigations. Saccades are rapid eye movements that change the line of sight, as when reading, and take the eyes from one object of interest to another. Saccades are ballistic – their duration is so brief (~50 msec) that visual information, which takes longer to process, cannot alter a saccade once it has been initiated. In order to maintain saccade accuracy, a parametric adaptive feedback process exists: the error after each saccade (how far off the intended visual target the eyes land) is used to maintain calibration. We invoke this adaptive process in the lab with a "double-step" paradigm (McLaughlin, 1967), in which the visual target is moved while a saccade is underway. If the target is displaced farther away from the saccade starting position, the saccade will fall short, and if the target is displaced toward the starting position, the saccade will overshoot. In each case, the adaptive process adjusts saccade gain (ratio of saccade amplitude to target displacement) within minutes to place the eyes closer to the displaced target position. (Due to visual masking, these per-saccade target displacements are not perceived, and thus the displacements are interpreted as errors in saccade gain.) Before and after adaptation, gain is assessed by presenting a target which is extinguished during the saccade, providing an open-loop measure (no visual feedback as to saccade accuracy). This saccade-adaptation paradigm is common in studies of oculomotor control and motor learning.

We demonstrated the feasibility of CSA in spaceflight applications by using gravity level as the context cue to switch between stored saccade motor programs. Parabolic flight in NASA’s Reduced Gravity Program provided alternating periods of 0g and 1.8g, of approximately 25 seconds each, 40 times in a flight (Karmali and Shelhamer, 2008), and was the experimental platform for some of the studies presented below.

In the first experiment, we adapted saccade gain to two different values in two different g-levels in parabolic flight (gain-down in 0g and gain-up in 1.8g), and then showed that the g-level itself – the context cue – could recall the previously-learned adapted responses (Shelhamer and Clendaniel, 2002; Shelhamer et al., 2002). While saccade gain is not a response that necessarily needs to change in a gravity-related manner in order to remain functionally useful, this behavior was chosen because saccade adaptation is rapid and does not involve head movements, both of which are appealing for parabolic-flight investigations (head movements in parabolic flight can quickly lead to motion sickness).

Although results are not as clean as those from better-controlled lab studies, we found that saccade gain can be associated with instantaneous g-level. An example is shown in Fig. 1, where one subject performed the adaptation experiment for three consecutive flights. Initial target steps are 20 deg, and the subject begins Flight 1 making saccades of this approximate amplitude, in both g-levels (left-most pair of bars). As adaptation progresses over consecutive flights, saccades in 0g become smaller and those in 1.8g become larger, as requested, until there is a substantial difference between the amplitudes of the saccades in each g-level at the end of the third consecutive flight ("flight 3 end").

This subject performed the same experiment again after a gap of eight months. Of special significance is the fact that g-specific (contextual) adaptation was retained after this intervening period, which is close to that required for a trip to Mars. This is noted in the right side of Fig. 1: initial responses in Year 2, before any adaptation, show an increased gain in 0g and vice versa. This lends support to the idea that adapted sensorimotor responses can be associated with g- level, and can retain that association after many months. A key point here is that this subject did not experience any parabolic flights (exposure to 0g or 1.8g) during the intervening eight-month period, which undoubtedly contributed to retention of contextual adaptation during this period.

Another form of contextual adaptation stems from observations which suggest that ambient environmental setting can also Gain decrease adaptation * p < 0.05 serve as a context. Recent Gain increase adaptation work in humans (Alahyane and Pelisson, 2005) shows that saccade adaptation can be retained for several days after training has ended even with normal (non-reinforcing) intervening experience. While it is not known if the adapted state was in effect while outside the lab between test sessions, it is reasonable to assume that saccades were normal during this time. If so, then this is an example of the laboratory environment serving as a context cue.

We examined this phenomenon in more detail in a prototype experiment with saccade adaptation. Each day, for four consecutive days, gain-decrease adaptation was performed in one room (A), and gain-increase adaptation in a different room (B). On the last adaptation day, gain was tested in each room several hours after the final adaptation trials, and again three days later (Fig. 2). In two subjects, gain decreased by an average of 6% in Room A and increased by 13% in Room B, as assessed on day 4. After the three-day gap, gain was still decreased in A and increased in B, although less so, showing that the room context had become effective enough to recall associated gains, even after several days.

Acquisition of saccade adaptation over three parabolic (to left of vertical flight the following vertical line), after a gap of eight of bars shows the saccade made to a deg, in 0g (hatched (dark bar). (From 2002, used with permission).

We repeated this experiment with adaptation of the vestibulo-ocular reflex (VOR). As with saccades, a simple paradigm is used to invoke VOR adaptation. In this case, the head is moved while a visual target is presented and made to move either more or less than normal relative to the head. If the target moves more than normal, an increase in VOR gain is required in order to maintain gaze on the target, and vice versa. Again, this lends itself naturally to context experiments in which a gain decrease can be associated with one context state and a gain increase with another context state.

The first VOR experiment also used two different rooms. In one experiment, both Rooms A (adaptation) and B (testing only) were small dark experiment rooms. In the second experiment, Room A was again a small experiment room, while Room B (testing) was a large conference room. Adaptation took place in Room A in each case, with a projected visual target driven by head motion as the stimulus for a gain increase (x2); the room was otherwise dark. Testing (no adaptation) took place in Room B, to determine if adaptation would transfer to a neutral room or remain tied to the adaptation environment of Room A. Testing took place in both rooms more than an hour after the last adaptation session. In the first experiment, where the two rooms were similar, there was no context-specificity (Fig. 3A): adaptation transferred completely from Room A to Room B. When the two rooms were different (Fig. 3B), there was significant context- specificity: the adapted VOR became associated with the room in which adaptation took place (A), with less transfer to the neutral room (B).

In the second VOR experiment, two subjects underwent four consecutive days of VOR adaptation in two different experiment rooms (A and B). Subject 1 was exposed to gain-increase adaptation in Room A and gain-decrease adaptation in Room B, and the converse for Subject 2. VOR gains before and after adaptation sessions were assessed each day, with a novel method that does not require the direct measurement of eye movements and provides an approximate measure of VOR gain. (In this "oscillopsia-nulling" task, the subject nulls the apparent motion of a visual target during head movements, and the proportion of head motion that is needed to drive the target so that it appears stationary in space is used to infer a VOR gain: Beaton et al., 2010). Gains in Subject 1 were also tested on the first and last days with eye-movement recording. The last test day is one day after adaptation, providing a test of retention. In each subject there was a modest gain increase and a substantial decrease (Fig. 4), each tied to the respective experiment room.

We also tested the effectiveness of an augmented cue in the course of this experiment, to determine if context specificity would be enhanced. An augmented cue is a secondary context cue – one that is associated with one of the context states in order to make its effect stronger. For this experiment, a radio playing in the background served as the augmented cue, and was present for Subject 1 during gain-decrease adaptation and for Subject 2 during gain-increase adaptation (augmenting Room B in each case). Assessed after adaptation, Subject 1 had a decreased gain in a neutral room (where no adaptation had been imposed) with the augmented cue. Subject 2 had an increased gain with the augmented cue in a neutral room. The augmented cue was effective in pulling adaptation from an experiment room to a neutral room.

4. Summary

We have initial evidence that some sensorimotor responses can be adaptively maintained and tied to g-level over at least eight months. This makes it feasible to train – before flight – a set of skills and behaviors that are g-dependent (such as manual control, piloting, and locomotion) before flight, foreseeing that the later g-onset when arriving at Mars will recall these behaviors appropriately and immediately.

Not only g-level, but also ambient setting, can serve as an effective context. This has both positive and negative implications. Training, as on a piloting or docking task, might be carried out in a cockpit (real or simulated) under the identical g-level in which the task is to be performed, as an aid to tying a specific g-calibration to a specific task and setting. The disadvantage is that care should be taken to avoid "contaminating" this context-specificity by avoiding task practice in the incorrect g-level or cockpit setting. Thus it might actually be best not to practice some procedures that are critical and g-dependent during the 0g phases of the long flight to Mars. This could present a challenging mission-design problem: while it would certainly seem a good use of time to train for the eventual entry and landing as much as possible during the extended trip, and it would seem to be beneficial to do this in the 0g phases, it might in fact be counterproductive. However, refreshing of the sensorimotor task during occasional artificial- gravity (AG) exposure could be very helpful, especially if an appropriate cockpit arrangement and augmented cue are used. A general approach to effective augmented cues is a subject for further research, but these might include specific ambient color schemes, noise and annunciator sound spectra, or instrument configurations. This augmentation should be presented only with the task and g-setting in which it would eventually be needed.

An important issue that we have avoided is the specific gravity level. Earth-based training is obviously performed in 1g, while Mars presents 0.38g. AG may be at Martian level, but it is cheaper in terms of structure and energy to have a lesser magnitude; will this make a difference in maintaining g-specific responses if they are trained at 1g, refreshed at (say) 0.1g, and ultimately needed at 0.38g? Will training on Earth be better if it is performed during Martian- gravity parabolic flight rather than in 1g? It may be that in any case some non-zero g is a strong context relative to zero g, and in fact observations tend to show that 0g is a singularity as far as many neurovestibular responses are concerned. Thus it might be sufficient to treat any g-level as one context state, and 0g as another.

We conclude with a proposed research strategy for context-specific adaptation as appropriate to long-duration flight:

Acknowledgements: Supported by NASA grant NNX10AO19G and NIH grant T32-EB003383.