Visual measures of perceived roll tilt in pilots during coordinated flight and gondola centrifugation

2.3.1Centrifuge

The centrifuge experiments were performed in the dynamic flight simulator (Wyle Laboratories Inc., El Segundo, CA) at the Aviation Physiology Laboratories in Linköping. This centrifuge has a radius of 9.1 m. Its rotation is anticlockwise (as seen from above).

Facing forwards in the gondola, the subject was fixed in a cockpit seat by means of safety belts. The head was not restrained (as this would not be allowable in the aircraft) but the subject was instructed to avoid head movements by keeping the back of the head against a head rest. The rest had a u-shaped vertical groove for the back of the head and was padded with 1 cm of foam rubber. It was adjustable in the vertical and antero-posterior directions, making it natural and comfortable for each subject to keep the head stationary in the upright position during the entire recording periods.

The roll position of the gondola was computer-controlled so that the resultant gravitoinertial force vector was always parallel with the subject’s median plane. In addition, during acceleration and deceleration of the centrifuge the pitch position of the gondola was adjusted with respect to the tangentially acting inertial component; thus, also in pitch, the orientation of the subject did not change with respect to the resultant gravitoinertial force vector. Planetary acceleration and deceleration was 7.8°/s². From stationary the 2-G level was attained in 10 s. During the experiments the gondola was completely darkened but the subject was observed in infrared light by means of a video camera and could always communicate with the experimenter via a two-way intercom system.

2.3.2Aircraft

The in-flight experiments were carried out in a high-performance jet trainer (Saab Sk60). The two standard catapult chairs had been replaced by simple seats, giving place for two extra seats behind. Pilot flying (an experienced fighter pilot) was sitting in the left front seat with the experimenter to the right. The test subject was seated in the left rear seat. To the right of him was a technician managing data collection. All wore headsets and could communicate freely with each other. The subject stabilised his head against a head rest, similar to that used in the centrifuge. In addition, securing the head to the face-mask device (see below) ensured that the sagittal plane of the head was persistently aligned with the long axis if the aircraft and that rotational movements of the head were minimized.

The air speed was 290–310 kn. The pilot aimed at entering, in a coordinated manner, left turns with a bank angle of 60° (2 G) within 10 sec and to maintain each turn for 6 min. During the turns, the pilot’s manoeuvring was based primarily on the attitude indicator and instruments presenting the G level and any deviation between the G vector and the aircraft’s vertical plane. By slowly increasing the altitude, we avoided the vibrations which can occur when passing through the vortex of the aircraft’s own trajectory. Exiting the turns was also performed in 10 sec and with the resultant G vector persistently in the head-to-seat direction. The resulting trajectory would ideally be circular with respect to the air, not as projected on the ground. As a rule, each turn was preceded and followed by at least 2 min of straight-ahead level flight.

Complete darkness was created in the following way. The test subject wore a modified diver’s mask, the glass of which had been removed. The mask was connected to the device with the luminous line via a light-proof flexible tube (diameter 20 cm), consisting of an external layer of reflecting plastic and aluminium foil and an internal layer of black velvet. Light-proof ventilation channels permitted breathing through the nose. The subject could put on and remove the mask and the head set without assistance.

Resultant Gz level and roll attitude were recorded with a frequency of 5 Hz by means of a 3DM-GX3-45 Miniature GPS-Aided Inertial Navigation System (LORD Microstrain®, Williston, VT). This device also provided a recording of the aircraft’s trajectory relative to the surface of the Earth. In addition, the instruments were continually observed by the experimenter who also made notes on relevant indications.

2.3.3Recording of the SVH

In front of the subject (at eye level and straight ahead, 50 cm from the subject’s eyes) there was a line of red light-emitting diodes, 75 mm long and 1.7 mm wide, mounted on the axle of a digital servo (DSR 1015, Thunder Tiger Corp, Taichung City, Taiwan). The axis of rotation coincided with the subject’s naso-occipital (visual) axis. The servo was controlled by a microprocessor (Arduino, UNO with a program in C). Every time the line was switched on, the subject adjusted it, using two push-buttons on a remote control, so that he perceived it as horizontal (i.e. so that it corresponded with his spontaneous imagination of the horizon of the external world). If the subject kept one of these buttons pressed, the rotation of the line was 11°/s; by briefly tapping the buttons the subject could adjust the orientation of the line in steps of 0.2°. When pleased with a setting the subject pressed a third button, which extinguished the line. With an accuracy of 0.1° the deviation of the line from the gravitoinertial horizontal (i.e. the transversal plane of the gondola or aircraft) was automatically recorded. It was then instantaneously offset 8–26° (randomly), alternately clockwise and counter-clockwise with respect to the subject’s latest setting, and switched on again after a latency of 1 s.

Programming for the recording of data from the microprocessor was performed in LabView (National Instruments Corporation, Austin, TX) on a HP ProBook 6570 (Intel Core i5, 2.60 GHz) connected to the microprocessor via a network cable.

As rule, recording of the SVH commenced at least 1 min before acceleration of the centrifuge or the entering of a turn with the aircraft. After deceleration of the centrifuge, and after the exiting of a turn with the aircraft, recording proceeded for at least 2 min at 1 G.

4.1.1The magnitude of indicated roll tilt

During a co-ordinated 2-G turn (bank angle 60 degrees), pilots with similar education and flight experience showed a substantial inter-individual variability in the visual measure of perceived roll tilt; the range was approximately 40 degrees. In addition, there was a striking correspondence between data obtained in the gondola centrifuge and during co-ordinated flight. For both the initial and final SVH tilt there was a high and significant correlation between data from the centrifuge and aircraft, i.e. each individual tended to show similar values in the two stimulus situations. Considering group means and inter-individual variability the present values are compatible with those of earlier studies on experienced fighter pilots and military helicopter pilots [26]. This suggests that the basic flight education might have a relatively great impact on learning processes associated with curved flight.

4.1.2Response patterns

An additional observation was that the pilots did not simply differ with respect to the magnitude of indicated tilt. In spite of the small number of subjects, the study design, with repeated 2-G exposures and a great number of data points, made it possible to discern two principal response patterns. One of these is characterised by a tilt of the SVH which is large in the beginning of the 2-G exposure but then declines, approaching zero by the end of the 2-G exposure. In these cases there was, as a rule, an after-effect, with negative values in conjunction with deceleration of the centrifuge or exiting the turn with the aircraft. Presumably, these subjects responded to the canal stimuli associated with acceleration and deceleration of the centrifuge as well as with the entering and exiting of turns with the aircraft. The decline during the 2-G plateau would reflect the fading memory trace of the initial canal stimuli. The after-effect seems to be a counterpart to the so-called leans, i.e. an illusion of banking that can occur during straight-ahead and level flight after a rapid correction for a turn, entered involuntarily and slowly [3]. In contrast to common characterisations of the leans, the present after-effects were short-lasting, although the magnitude could be as large as the SVH tilt in the beginning of the preceding 2-G period.

The second major pattern consisted in a tilt of the SVH that remained at a similar magnitude during the entire 2-G exposure. No after-effects were seen in these cases, but there was another conspicuous phenomenon, namely errors - and even sudden reversals - in the direction of the SVH tilt. As discussed below, although pilots may associate the sensation of increased G load with being in a coordinated turn, the somatosensory input does not reveal the direction of the turn. Therefore, pilots who tend to rely on this information may experience an ambiguity in the perceived direction of banking.

4.2.1Characteristics of non-pilots

In non-pilots, acceleration of the gondola centrifuge typically induces a tilt of the SVH that initially approximates 30–40 per cent of the real gondola inclination and then declines exponentially with a time constant of 1-2 minutes, approaching an asymptote that rarely deviates more than 5 degrees from zero. Further, there is a considerable inter-individual variability; at 2 G the initial SVH tilt may range from near-zero to more than 40 degrees [22, 23]. A suggested explanation for the general underestimation of the roll tilt angle is that the graviceptive systems, which during gondola centrifugation persistently signal that the subject is upright in roll, contradict the semicircular canal response to the swing out of the gondola [7, 22, 29]. Accordingly, one factor behind the inter-individual variability could be that subjects differ with respect to the relative dependence on input from the semicircular canals and otolith organs.

4.2.2Characteristics of pilots

At group level, experienced pilots differ from non-pilots in three major ways [26]. Firstly, a considerable fraction of pilots show responses that are greater than in the most extreme non-pilots; SVH tilts approximating the real tilt of the gondola are frequent among pilots. Secondly, it is common that the SVH tilt does not decline with time. Thirdly, also among experienced pilots there is a great inter-individual variability; the final SVH tilt may range from near-zero to values approximating the gondola inclination.

4.2.3Reproducibility.

Studies with repeated testing have shown that differences between subjects are not due to uncertainties in the measurement procedure. In non-pilots, tested with an interval of 1-2 weeks, the reliability coefficient for the initial and final deviation was, respectively 0.86 and 0.87; the corresponding values for student pilots were 0.97 and 0.98 [20]. Also in the present material, there was a high correspondence between data from the two centrifuge occasions, in most cases separated by several weeks. Thus, it must be assumed that the inter-individual variability in the visual measure of perceived roll tilt represents true and comparatively stable individual characteristics.

4.3.1The bodily sensation of increased weight

In earlier studies, several pilots have described that they associate the sensation of increased bodily weight and strain in the head-to-seat direction with being in a co-ordinated turn - and that an increase in G level is often accompanied by the visual impression of an external horizon, tilted with respect to the own transversal plane [20, 26]. Similar statements - and “pilot-like” SVH responses - have been made also by non-pilots with proficiency in motorcycle riding or certain sports activities [27].

Given the fundamental role of graviception, it might seem paradoxical that an increase in the magnitude of the G vector, acting in the head-to-seat direction, would lead to a sensation that this G vector is not pointing “downwards”. Nevertheless, analogues of co-ordinated flight are common in animal life. Running along a curved path would not be compatible with postural control if the perceived direction of the resultant G vector were taken as a representation of the Earth gravity field. Rather, the constancy of the Earth gravity field combined with the equivalence between gravity and inertia entails that a neural representation of the magnitude of the G vector may facilitate the task of attaining an adequate combination of speed, trajectory and lateral body tilt.

Further, during gondola centrifugation the decline in the SVH tilt among non-pilots occurs with a greater time constant, i.e. more slowly, at higher G levels [22]. Thus, although the gravitoinertial force vector, acting in the head-to-seat direction, is in conflict with the initial canal message (which induces the sensation of roll tilt) the increase in the G vector seems to promote the memory of the canal message. Also this observation suggests that the tendency in pilots to associate an increased G load with a bank angle has a natural basis and would not require any radical or “unnatural” re-interpretation of vestibular and somatosensory input.

4.3.2Spatial orientation learning during flight

Since the persistence of the SVH tilt is common in pilots, it must be assumed that the development of this phenomenon is driven by consequences of practical significance (positive feedback); an improvement in spatial orientation would reduce the cognitive workload of flying. But how could this be reconciled with the large inter-individual variability in the magnitude of the SVH tilt? One possibility is that pilots simply differ in constitutional properties underlying the development of a sense for the relationship between G load and bank angle. Alternatively, under ordinary flight conditions the value of this sense would not depend so much in quantitative accuracy. Even if not providing an approximate measure of the bank angle, it might support situational awareness in a qualitative way by maintaining the experience of being in a turn.

However, since the increase in G level per se cannot serve as a clue regarding the direction of banking, the value of an ability to translate G level into bank angle would be dependent on other sources of information, e.g. the semicircular canals. Thus, if the entering of a turn occurs slowly, or in case vibrations interfere with the pilot’s ability to detect the canal stimuli, then the G-dependent mechanism for roll-tilt perception may yield a direction error.

The occurrence of two distinct response patterns suggests that inter-individual variability is not merely a matter of quantity but that individuals may differ also with respect to the sensory modality underlying perceived roll tilt during co-ordinated flight. Pilots in whom the SVH tilt is large in the beginning of the 2-G exposure but then declines substantially with time apparently respond to canal stimuli but do not use the sensation of G load as an indication that the roll tilt of the gondola or aircraft persists. In contrast, a G-sensitive pilot who misperceives the direction of banking would have neglected the canal input. Also the abrupt transitions in the direction of indicated roll tilt by subject H during a single 2-G period apparently occurred in contradiction with a canal input representing motionlessness. Thus, the identification of these two response patterns supports the notion that humans may have an inherent tendency to prioritize canal or otolith input. Even if not always advantageous for spatial orientation, such a tendency might be more economic, in terms of neural processing, than interpretations based on both sources of information.

4.4.1Semicircular canal stimuli in roll, yaw and pitch

Nevertheless, the correspondence in perceived roll tilt between the two systems might appear problematic to reconcile with certain earlier findings in the centrifuge. Namely, the magnitude of perceived gondola inclination is dependent not only on the roll-plane angular-displacement component per se (the swing out of the gondola) but also on the pattern of the angular-velocity components in yaw and pitch, determined by the angular velocity of the centrifuge about its main axle and the subject’s increasing tilt with respect to this axle [21, 25, 26].

An ideal observer, seated upright facing forwards in the gondola, would perceive the initial yaw-left rotation as the entering of a curve, and interpret the increasing sensation of pitch- backward rotation as a consequence of his own change in roll orientation with respect to an Earth-fixed axis of rotation. The roll angular-displacement component would facilitate that interpretation. It has, in fact, been established that the brain’s processing of sensory information on self-motion involves tacit knowledge of physical constraints [13] or estimates of probabilities [11]. Thus, if a subject undergoing acceleration in the centrifuge was capable of regarding the main axis of centrifuge rotation - revealed initially via the yaw stimulus - as an Earth-fixed reference, then the proportions of the yaw and pitch components would convey information about the Earth horizontal plane.

The significance of this source of information is reflected in the fact that the ability to perceive the change in gondola position during acceleration is dependent on the subject’s heading position. In experiments with 8 non-pilots the SVH response was 21±8° in the forward position but only –7±10° degrees when the subjects were facing backwards [21]. In a follow-up study, we compared 8 non-pilots with 9 experienced fighter pilots [26]. In the forward position, the initial SVH tilt was greater in pilots (31±16°) than in non-pilots (17±6°). Also in the backward position the pilots showed a greater initial SVH tilt (–12±17°) than the non-pilots (–5±7°). In both groups the responses were significantly greater in the forward position. Conspicuously, whereas the SVH declined with time (5 min at 2 G) to near-zero in non-pilots, the pilot group showed a negligible decline in the forward position (from 31±16° to 26±23°) but an increase when facing backwards (from –12±17 to –24±23°).

4.4.2Familiarity of the motion pattern as a whole

One explanation for the difference in perceived roll tilt between the forward and backward position is based on concepts adopted from Gestalt psychology [9, 10]. Briefly, when the subject is facing forwards, the pattern formed by the canal stimuli in yaw and pitch would be more familiar; in conjunction with other stimulus components it will contribute to an experience of the motion pattern as a meaningful whole. And if the subject has an adequate imagination of the nature of the motion pattern as a whole, he or she will more likely discern and estimate a single component, e.g. the change in roll position. In contrast, when the subject is facing backwards, the pattern of yaw and pitch angular velocity (i.e. transition from yaw to pitch-forward) would be unfamiliar, which would interfere with the perception of the roll stimulus. Accordingly, the subject’s ability to comprehend the nature of a complex stimulus may be reflected in the perceived magnitude of a single stimulus component [25].

4.4.3Differences in canal stimuli between aircraft and centrifuge

The significance of the canal stimuli in yaw and pitch for the perception of roll position evokes concerns regarding the possibility of re-creating the entering of a co-ordinated turn by means of a centrifuge-based simulator. Since the radius of a 2-G turn, performed with an aircraft, is much greater than that of a centrifuge, the yaw and pitch components will typically be only about 10 per cent in the aircraft, although the change in roll position can be performed with the same angular velocity as in the centrifuge. Therefore, the contribution of angular velocity in yaw and pitch on the perception of roll tilt makes it seem likely that the perceived bank angle would be smaller during real flight than during a centrifuge run.

Nevertheless, during the entering of a turn with an aircraft, the proportions between the yaw and pitch components, i.e. the transition from yaw to pitch-backward, will correspond to the roll tilt in the same way as in the centrifuge. Thus, if the yaw and pitch components are above the stimulus threshold for the canals they will be redundant with the roll plane stimulus and thereby facilitate the perception of a change in roll attitude. This would explain why the SVH tilts were similar in the two systems despite the quantitative difference in the two angular- velocity components. However, since vibrations may increase the perception threshold for canal stimuli, there would be a greater likelihood that the yaw stimulus, indicating the beginning and direction of a turn, will be neglected during certain flight conditions.

4.4.4The magnitude of the resultant G vector

As regards the increased gravitoinertial force vector, there is no reason to doubt the similarity between a centrifuge run and a real turn with an aircraft. Accordingly, the tendency of pilots to associate an increased sensation of weight with a bank angle was similar in the aircraft as in the centrifuge. However, since the G vector per se does not reveal the direction of roll tilt, one might consider factors that would make errors in the perceived direction of banking more likely in aircraft. As mentioned, the initial yaw-plane canal stimulus is much weaker in the aircraft, and the pilot’s ability to discern canal stimuli may be impaired by vibrations. Another aspect is that the centrifuge run commences with a jerk and a tangential acceleration, which inevitably alert the pilot, who may become keener to the coming stimulus pattern. Because of such differences it cannot be concluded without further notice that testing in a centrifuge would give a true reflection of the prevalence or significance of this kind of spatial disorientation in aviation.

4.5.1Quantitative evaluation of simulated flight manoeuvres

No simulation system can reproduce real flight in perfect detail. The evaluation or adjustment of simulation algorithms often involve verbal judgments by experienced pilots. Quantitative recordings of specific components of spatial orientation may be a valuable complement to verbal accounts. The precision of conscious vision, as well as the close connections between vision and the sense of balance, makes a visual indicator particularly useful for the study of spatial orientation. With the contrivance for achieving complete darkness, it can easily be mounted also in an aircraft. The findings of the present experiments strongly indicate that in spite of certain differences between a real co-ordinated turn and a centrifuge run, the perception of roll tilt in pilots does not differ between the two conditions. Thus, the centrifuge may be considered an adequate means for simulating the vestibular stimulus of co-ordinated flight. Issues that remain to be addressed are the significance of neck proprioception and the vestibular effects of head movements. In the present study head movements were avoided (although the head was not completely restrained) whereas during real flight the pilot’s visual scanning of cockpit instruments or of the external environment require frequent head movements. Thus, Coriolis phenomena and G-excess effects [3] may contribute to the disorientation caused by coordinated flight per se.

4.5.2Inter-individual variability

The large inter-individual variability in a group of pilots who had undergone the same strictly scheduled education and were at a similar stage in the career raises questions regarding, on the one hand, underlying learning mechanisms and, on the other hand, practical consequences. One issue is the propensity of pilots to “translate” the sensation of G load into a bank angle. If this ability had been driven by the sensory impressions and demands of ordinary flight, then, the major advantage, in terms of cognitive economy, would not be dependent on quantitative accuracy. Rather, the mere association of increased G load and banking, as a qualitative experience, would serve to maintain the pilot’s awareness of being in a co-ordinated turn.

Nevertheless, the magnitude of perceived bank angle may become a determining factor in rare situations where high cognitive workload or acute psychological stress impairs the pilot’s ability to manoeuvre based on rational scanning of the instruments [1, 24]. Thus, demonstrations in a centrifuge, with recordings of the subjective visual horizontal, may not only provide insight into sensory mechanisms underlying spatial orientation but could also contribute to the pilot’s awareness of personal limitations. As to the question whether the ability to perceive the bank angle during co-ordinated turns can be improved by training in a centrifuge, a basic requirement would be quantitative evaluation also in aircraft.