Enhancing Spatial Orientation in Novice Pilots: Comparing Different Attitude Indicators Using Synthetic Vision Systems [original]

Enhancing Spatial Orientation in Novice Pilots: Comparing

Different Attitude Indicators Using Synthetic Vision Systems

Alice Gross & Dietrich Manzey

Technische Universitaet Berlin, Berlin, Germany

Spatial disorientation (SD) is a common factor in aviation accidents, especially in novice pilots. An

experiment was carried out to determine which of four different attitude indicator concepts in combination

with two different display backgrounds (abstract vs. synthetic landscape) proves to be the most beneficial

for novice pilot performance. Inexperienced pilots had to recover from unusual attitudes by using the

standard moving-horizon display, a moving-aircraft display, a frequency-separated display, and a “mixed”

display, with the latter two representing hybrid concepts with movements of both aircraft symbol and

horizon bar. Participants performed the task of recovering from unusual attitudes most efficiently with

hybrid display concepts, suggesting that these display concepts prevent figure-ground reversals and

associated pilot errors. Outcomes of the study suggest that the implementation of hybrid display concepts

as a backup option when unwillingly entering Instrument Flight Conditions could be a solution for

preventing SD in novice pilots.

INTRODUCTION

Loss of spatial orientation (SO) during aircraft

navigation is a common factor in fatal aviation accidents

(Comstock et al., 2003, Gillingham & Previc, 1996). Collins

and Dollar (1996) found that 80.2% of aviation accidents

associated with spatial disorientation occur in instrument

meteorological conditions (IMC), when pilots have to fly

under instrument flight rules (IFR), navigating by reference to

the attitude indicator (AI) and other instruments only. From

these aviation accidents 85% were the “result of collision with

the ground, water or structure” (p.7). Especially untrained and

novice pilots who do not have a certification for flying in IMC

tend to experience difficulties in flying in adverse weather

conditions: Nearly half of all weather-related accidents result

from pilots attempting to continue visual flight rules (VFR;

navigating solely by reference to outside visual cues) flight

into IMC. When continuing to fly under VFR in IMC the

probability of having a fatal accident increases to 83%

(Roscoe, 2004).

While flying under IMC, pilots need to rely on the AI, an

instrument which displays visual information about the

aircraft’s pitch angle (nose-up or nose-down) and bank angle

(tilting of the aircraft to one side). By doing so it provides

crucial information about aircraft attitude, so that the pilot

does not have to rely on what he sees (or does not see) when

looking out of the window. Conventionally, the AI consists of

a small symbol depicting an aircraft and a horizon bar, which

divides the instrument into two halves. The top half

representing the sky is usually blue. The bottom half

representing the earth’s surface is usually brown. Additional

degree marks on the display representing pitch and bank angle

are also common.

There are several possible design options to convey pitch

and bank information via the AI and it remains an open

question how to depict this information in a most compatible

way (e.g. Previc & Ercoline, 1999; Yamaguchi & Proctor,

2010). The standard way of designing an AI is the ‘inside-out’

or ‘moving-horizon’ display. It represents roll and pitch

movements by its consequences in terms of what one would

see if looking at the outside world through a porthole in front

of the aircraft. That is, the aircraft symbol remains fixed and

the artificial horizon-line in the AI rotates or moves upwards

or downwards corresponding to the apparent movements of

the real horizon line if looking outside from the cockpit. This

display fulfills what has been referred to as the “principle of

pictorial realism”, which states that a “display should look like

or be a pictorial representation of the information that it

represents” (Wickens, 2003, p. 152). However, it does not

confirm the competing “principle of moving part” which

requests that the movements of the display corresponds to the

movement of the aircraft, as well as to the steering movement

of the pilot. This latter principle is better reflected by the

“outside-in” or “moving-aircraft” display which has been used

in Russian aircraft for a long time. The moving-aircraft

display only depicts pitch movements by means of upward and

downward movements of the artificial horizon-bar and roll

movements by means of movements of the aircraft symbol. A

schematic depiction of both display formats is provided in

figure 1.

Figure 1: Schematic representation of moving aircraft display

(a) and moving horizon display (b) in an ascending right turn.

A number of empirical studies have investigated the

compatibility and human performance consequences of these

two different AI designs. Whereas no differences were found

with respect to attitude tracking, i.e. situations where pilots are

constantly checking the display and making control inputs to

correct small deviations in order to maintain a given attitude

(Yamaguchi & Proctor, 2010), the moving-aircraft display was

usually found to be significantly better ( more suitable) when

pilots had to recover from suddenly occurring unusual

attitudes (Johnson & Roscoe, 1972; Lee & Myung, 2013;

Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 1033

Roscoe, 1968). Specifically, novice pilots, not yet trained for

instrument flying, committed much more reversal errors in

their initial correction movements and/or needed longer time

to recover when flying with a moving-horizon display

compared to a moving-aircraft display. Seemingly these pilots

intuitively misinterpreted the movements on the display as

representing aircraft movements. This suggests a dominance

of the principle of moving part over the principle of pictorial

realism and has led researchers to call the usability of horizon-

moving displays into question (e.g. Previc & Ercoline, 1999).

One possible explanation for the disadvantages of moving-

horizon displays, that has already been put forward during the

early years of flying, involves a figure-ground reversal effect:

Within the context of extreme flight attitudes and a moving-

horizon AI, the pilot no longer sees his aircraft as the mobile

part in the world but his display as the mobile part which

moves against the stable cockpit panel background (Grether,

1947). This could lead to a figure-ground reversal in the

pilot’s mental model, ultimately letting him to believe the part

he has influence over via control inputs is the horizon bar

instead of the aircraft symbol. It instantly seems plausible that

such an effect might be responsible for the research results

reviewed above taking into account that most, if not all, of the

research was based on early generation of moving-horizon

displays which typically represented single round instruments

of comparatively small size. However, in current generations

of glass-cockpits the size of displays has significantly

increased which now provides new possibilities of integrating

more realistic images of the real world in the AI by means of

synthetic vision system (SVS) technology. It seems obvious

that the use of SVS might strengthen the “pictorial realism” of

a moving-horizon AI, thus making the figure-ground

relationship in the display less ambiguous. Accordingly it

might be assumed that SVS technology might diminish or

even reverse the disadvantages of moving-horizon displays

compared to moving-aircraft displays. Moreover, the design of

hybrid displays might be possible. These displays combine

motion relationships from both the moving-horizon and the

moving-airplane display, having both parts in the AI move in

certain relationships to each other, thereby making control

reversals less probable. One such concept has already been

proposed by Roscoe and colleagues (frequency-separated

display; Roscoe & Williges, 1975; Roscoe, Corl & Jensen,

1981).

The scope of this paper is to revisit the compatibility issue

of AI display design in the context of SVS technology. For

this purpose, student participants without prior flying

experiences were required to perform attitude recoveries with

four different AI designs, i.e. moving-horizon, moving-

aircraft, and two hybrid designs, and either of two

backgrounds, i.e. abstract vs. synthetic landscape.

METHOD

Participants

All participants were TU Berlin staff or students recruited

through opportunistic sampling. A total of 30 participants, of

which 14 were male and 16 were female, took part in the

study. The average age of participants was 25.96 years (SD =

3.2). None of them had any prior knowledge of flying,

whether in a simulator, nor in real life. Participation was

compensated with a payment of 5 Euro (about 6.80 US$) per

person. An experimental session took around one hour.

Apparatus and Tasks

The research simulator was situated in a laboratory of the

Technical University of Berlin. Four computers were

connected over a local area network to generate the primary

flight display (PFD) including the AI instrument, a

navigational map and the view out of the window. The fourth

computer was needed to start and stop flight scenarios via the

software UltraVNC version 1. The research simulator

consisted of a fixed base mock-up replication of a Cessna 172

Skyhawk SP G1000 Cockpit which was placed on top of a

desk. The open source flight simulator FlightGear was used as

the simulation model. A Saitek Pro Flight Yoke System USB

steering yoke was screwed onto the desk in front of the left

monitor on which the PFD was simulated. The right monitor

showed a navigation map of the terrain, pilots were flying

above. The view out of the window was projected onto the

wall above the cockpit mock-up. 16 unusual attitude

recoveries were flown with every AI, which had to be

performed as quickly as possible. These included recoveries of

four different bank angles (30, 60, 90 or 120 degrees)

simulating a surprising change of aircraft attitude, which were

presented four times each. Each bank angle was presented as a

tilt to the left or right and with pitch either 15° up or 15°

down. Each attitude change was presented in an unpredictable

way after some time of stable horizontal flight.

Design

The study was conducted as a 4(type of display) x 4(bank

angle) x 2(display background) mixed design. The first factor

was defined as within-subjects factor and involved four levels:

Participants flew with a moving-horizon display, a moving-

aircraft display, a frequency-separated display, and a “mixed”

display. The frequency-separated display differs from the

abovementioned displays to that extent that it does not only

depict pitch and bank information but aileron information as

well. Every aileron input executed is reflected in a

corresponding movement of the aircraft symbol in the same

direction as the steering wheel. Pitch and bank attitudes are

indicated conventionally in the same ways as in a moving

horizon-display. In the mixed display, both the airplane and

the horizon symbol move in a certain ratio to each other,

thereby depicting roll and pitch angles. The angle between

aircraft and horizon line does depict the actual bank angle of

the aircraft. The different types of displays are presented in

figure 2. The second factor was also defined as a within-

subjects factor and included the different degrees to which the

aircraft banked. The aircraft could either bank to a degree of

30°, 60°, 90° or 120°. The third factor involved the two

different display backgrounds and was operationalized as a

between-subjects variable. One half of the participants flew

with a SVS background whilst the other half of the

Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 1034

participants flew with a classical blue-sky brown earth

background (see figure 3). In summary, every participant flew

16 recovery tasks with every AI and only one display

background, i.e. performed a total of 64 single tasks.

Figure 2: AIs (with SVS backgrounds) depicting a climbing

turn to the right, from upper left in clockwise direction:

moving-aircraft display, moving-horizon display, mixed

display, and frequency-separated display.

Figure 3: Classical (left) and SVS display background (right).

Performance measures

Performance measures were derived from log-files, which

contained the complete steering input of each participant.

There were three dependent variables. The first dependent

variable was time to initial control input. This was defined as

the time from the unusual attitude presentation to the first

control input that was recorded. This dependent measure is

important because it represents the time it takes to recognize

and process the aircraft attitude portrayed via the different

attitude indicators and backgrounds.

The second dependent variable was the total recovery

time, which was defined as the time it takes to bring the

aircraft in a stable position minus the time to initial control

input. Due to the fact that in this study all pilots were novices

and had no experience in flying an airplane or holding it in a

stable position whatsoever, a stable position was defined as

holding the aircraft between a pitch of ±5° and a bank of ±5°.

The aircraft had to be held between these ranges for at least

2.5 seconds in order to be rated as a success. This dependent

variable is important because it shows how an instrument

supports pilots in the process of bringing an aircraft back to a

stable straight and level flight after being confronted with an

unusual attitude.

The third dependent variable was the rate of reversal

errors. Errors were defined as a control input that caused the

aircraft to turn even further to the side it was already banked

to. Shortly speaking, if an aircraft was tilted to the right, the

correct control input would have been to steer the yoke to the

left and vice versa. If pilots committed an error while

recovering from unusual attitudes, this would have meant that

information conveyed by an AI was not easy to interpret or

even ambiguous.

Procedure

There was an accommodation phase as well as a practice

session before the data collection sessions started.

Accommodation phase started with one minute of free flight

during which participants could become acquainted with the

simulator without further instructions from the researcher. To

guarantee a similar level of understanding of movement

relationships between the yoke and the aircraft, as well as of

proficiency in steering the aircraft, a standardized text was

read to the participants to describe several flight tasks that

were to be flown. During the experiment, participants had to

recover from unusual attitudes, which were introduced

randomly. They had to master this task using four different

AIs. Attitude changed every 20 seconds, leaving exactly this

amount of time to the participant to recover to straight and

level flight. Flight level was automatically reset to 6000ft for

each attitude change to prevent participants from losing too

much altitude during the course of the 16 attitude recoveries

per AI.

Hypotheses

In accordance with earlier findings it was assumed that

the moving-aircraft AI in combination with the classical blue-

sky-brown-earth background would be beneficial in terms of

fewer initial steering errors and shorter reaction times (RT)

compared to the moving-horizon display. Accordingly, it was

theorized that participants would commit more errors and have

longer RTs with the moving-horizon AI. Furthermore, it was

assumed that these differences in performance measures

would disappear or even reverse when each of the two attitude

indicators are combined with the SVS background. A

computer generated terrain on the PFD should enable the

novice pilots to maintain a stable mental model of the flight

situation, thereby having shorter reaction times and

committing less control reversal errors, bringing performance

measures for the two displays more in line with each other.

It was further theorized that overall performance with

both the frequency-separated display as well as with the mixed

display in terms of errors and RTs would be better than with

the two classical display types, independent of the display

background. This was assumed because both types of hybrid

displays fulfill both relevant compatibility principles, i.e. the

‘principle of the moving part’ and the ‘principle of pictorial

realism’.

Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 1035

RESULTS

Time to initial control input. Overall, participants needed

0.9 seconds to give an initial control input with the SVS

display background (SD=0.29) and 0.78 seconds to give an

input with the classical background (SD=0.26). With each

display being separately, it took participants 0.93 seconds to

initiate a control input with the moving-aircraft display. With

the moving-horizon display, the frequency-separated display

and the mixed display it took 0.82, 0.84 and 0.76 seconds

respectively. The 4(display type) x 4(bank angle) x 2(display

background) analysis of variance (ANOVA) did not confirm

that the participants’ reaction to unknown attitude presentation

was faster with the classical background than with the SVS

background (F(1,28)=2.53, p=.122). However, it revealed a

significant main effect of display type (F(1.7, 47.8)=4.85,

p=.016). Post-hoc Bonferroni paired comparisons of display

type revealed that participants were faster in giving initial

control input with the mixed display than with the moving

aircraft display (p<.05). Other comparisons did not prove to be

significant. In addition, the main effect of bank angle, became

also significant, (F(2.2, 62.5)=8.16, p=.000). Means and

standard errors for this effect are shown in figure 4 (left side).

Responses to the more extreme shifts of bank angles (120° and

90°) were faster than responses to sudden shifts of bank angles

by 30°. No significant interactions were found

Total recovery time. The 4x4x3 ANOVA of total recovery

times yielded significant main effects of display type,

(F(9,84)=14.64, p=.000) and bank angle, (F(3, 84)=37.21,

p=.000), as well as a significant display type x bank angle

interaction (F(5.67, 158.7)=23.75, p=.000). Overall, mean

total recovery times were shorter for the hybrid displays than

the moving-horizon or the moving-aircraft display. A priori

planned post-hoc comparisons (Bonferroni) revealed that

participants were faster recovering from unusual flight

attitudes with the mixed display (4.82 sec) than with the

moving aircraft display (5.65sec; p<.05) and the moving

horizon display (5.91 sec; p<.05), as well as with frequency

separated display (5.19 sec) than with the moving horizon

display (p<.05). Means and standard errors for recovering

different bank angles with different display types are shown in

figure 4 (right). Bonferroni post-hoc paired comparisons of

bank angles revealed that participants were overall faster in

bringing the aircraft back to straight and level flight from a

bank angle of 30° than they were for all other bank angles (all

p<.05). The interaction effect was due to the fact that the

recovery times with the moving-aircraft display turned out to

be the slowest compared to all other display types for bank

angles of 30°-90°, yet the quickest for the most extreme bank

angle of 120°. No significant main effect of display

background (F(1,28)=.69, p=.415) nor any interaction

involving this factor was found.

Errors. Errors were defined as an initial control input

that caused the aircraft to turn even further to the side it was

already banked to. Each participant had to react to 64 sudden

attitude changes, thus 64 errors could be committed by each

participant. In total, 230 errors were committed by all 30

participants. Error rate per display type showed that most

errors (78) were committed with the moving aircraft display.

With the moving horizon display, the frequency separated

display and the mixed display, error rates were 69, 46 and 37,

respectively. When further dividing errors, not only per

display type but also per display background, it was found that

overall fewer errors were committed with the SVS background

than with the classical display background. However, the

ANOVA did not reveal this main effect of display background

significant, (F(1,28)=1.93, p=.175).

Figure 4: Time to initial control input per display type and bank angle

(left). Total recovery times per display type and bank angle (right).

There was a significant main effect for display type

(F(1,28)=4.86, p=.004). Bonferroni post-hoc paired

comparisons of display type revealed that participants

committed more errors when recovering from unusual flight

attitude with the moving aircraft display than with the

frequency separated (p<.05) and the mixed display (p<.05).

Furthermore, significantly more errors were committed when

using the moving horizon display than when using the mixed

display. No significant main effect was found for bank angle

(F(1,28)=1.01, p=.365). Moreover, a significant two-way

interaction between factors type of display and bank angle was

found, (F(5.652, 158.265)=2.857, p=.013). This interaction

seemed to have resulted from the fact that participants

committed significantly more errors when recovering from a

120° bank angle using the moving aircraft display compared to

when using one of the other three displays.

DISCUSSION

The scope of this study was to compare several AI

concepts along with two display backgrounds in terms of

novice pilot performance. More specifically it was

investigated to what extent SVS display backgrounds would

reverse earlier findings, which point to a general advantage of

a moving-aircraft display, in favour of the standard moving-

horizon display. It was also explored to what extent different

types of hybrid displays would provide general advantages

independent of the display background.

The first hypothesis was not supported by the data.

Contrary to our expectation, the moving-horizon display

generally led to equal or even better performance than the

moving-aircraft display. The only exception was the recovery

from extreme bank angles of 120° which were performed

quicker with the moving-aircraft display. These results

emerged independent of whether the background of the AI

was abstract or a synthesized picture of the environment. It

contradicts earlier findings, which indicate an advantage of

moving-aircraft display designs when using an abstract AI

display background (e.g. Gardner & Lacey, 1954; Previc &

Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 1036

Ercoline, 1999). One possible explanation for this discrepancy

could lie in the differences between the design of classical

AIs, investigated in the earlier studies, and the general design

of PFD as used in our research. When looking at the body of

research that has been conducted on this topic, it becomes

obvious that most of it was carried out between the 1940’s and

the 1980’s. AIs used in earlier studies were small, round

instruments that were not integrated with other elements of the

cockpit. Due to their smaller size and their clear cut

delimitation to the cockpit panel, it seems plausible that these

types of instruments were particularly prone to the figure-

ground reversal effect described in the introduction, which

have been proposed to explain the advantage of moving-

aircraft configurations (Grether, 1947). Accordingly, it is

probably much more intuitive for pilots to link their steering

movement directly to the movements of the display (as in a

moving-aircraft design) than a reverse coupling (as in the

moving-horizon display). In the current study a much larger

PFD design was used which not only differed from the earlier

displays with respect to its size but also to the obvious

presentation of the artificial horizon as background in relation

to the instrument information. All instruments of the PFD

were superimposed onto the display backgrounds without

having the airspeed indicator, the altimeter and the heading

indicator highlighted through a black background as it is done

in conventional PFDs. Although most PFDs make use of a

black background to highlight parts of the PFD, designs

similar to the one used in this experiment are produced and

employed for example by Garmin and Cessna. The

nonexistent boundaries between display background and

instruments could have prevented pilots from having figure-

ground reversals with the moving-horizon display: By looking

at a large and coherent display, the illusion of looking out of

the window is stronger than when looking at a small round AI.

Independent of whether land and sky were presented in an

abstract or more natural way (SVS) this feature could have

been supportive for creating the effect of looking out of the

cockpit window, thus decreasing the differences between

moving-aircraft and moving-horizon displays. Even more

important and interesting than the effects for the moving

horizon vs. moving aircraft displays are the effects found in

the present study for both types of hybrid displays. Given that

the hybrid displays used in the present study confirmed both,

the principle of the moving part and the principle of pictorial

realism, it was assumed that they should provide advantages

for novice pilots independent of the display background. Our

results provide at least partial support for this assumption.

Analysis of times to initial control input showed that

performance with the mixed display was indeed significantly

faster than with the moving-aircraft display. Furthermore, total

recovery times were significantly faster with the mixed as well

as the frequency-separated display than with the moving-

horizon display. Similarly, performance with the mixed

display was also significantly faster than with the moving-

aircraft display. Finally, significantly more errors were made

when using the moving-aircraft and the moving-horizon

display than when using a hybrid display. Overall, this pattern

of results suggests that the naïve participants used in the

present study performed best with the mixed display. It is to

be noted that this type of display was the most artificial one

because it combined display movements of the aircraft symbol

as well as the horizon but none of these movements

corresponded to the real world. By moving half the angle of

the “real” rolling movement of the aircraft or the perceived

horizon, only the final angle between the banked aircraft and

the horizon corresponded to the true relationships. Obviously,

this type of design provided two advantages which made it

intuitive and easy to understand for the participants: (1)by

combining the two movements relationships into one display,

the two design principles were integrated (2) extreme

deviations from a horizontal attitude of the aircraft where

depicted in only moderate angles on the display, thereby

making it comparatively easy to identify the direction of

necessary steering actions quickly. Even though such a mixed

display might be most confusing for pilots trained in

instrument flying, it seems that it might support other pilots

best in cases of unforeseen and rare occasions where it is

necessary requirements to correctly identify and correct the

attitude of their plane, based on instrument information only.

Overall, outcomes of this experiment make a re-evaluation of

earlier experimental outcomes advisable in the light of

progressed technical development in cockpit instrumentation,

associated altered AI attributes, as well as new possible design

options for AIs.

REFERENCES

Collins, W.E., Dollar, C.S. (1996). Fatal General Aviation Accidents

Involving Spatial Disorientation: 1976-1992. Federal Aviation

Administration Office of Aviation Medicine, Washington, DC.

Comstock, J.R., Jones, L.C. & Pope, A.T. (2003). The Effectiveness of

Various Attitude Indicator Display Sizes and Extended Horizon

Lines on Attitude Maintenance in a Part-Task Simulation. Human

Factors and Ergonomics Society Annual Meeting Proceedings,

Aerospace Systems, 5, 144-148.

Gardner, J.F. & Lacey, R.J. (1954). An experimental comparison of five

different attitude indicators. WADC Technical Report No. 54-32,

Wright Air Development Center.

Gillingham, K.K. & Previc, F.H. (1996). Spatial orientation in flight. In R.L.

DeHart (Eds.), Fundamentals of Aerospace Medicine (2nd ed.).

Baltimore: Williams & Williams.

Grether, W.F. (1947). A discussion of pictorial versus symbolic aircraft

instrument displays (Memo. Rep. No. TSEAA-694-8B). Wright-

Patterson Air Force Base, OH: Air Material Command.

Johnson, S.L. & Roscoe, S. N. (1972). What moves, the Airplane or the

World? Human Factors 14, 107-129.

Lee, B.G. & Myung, R. (2013). Attitude Indicator Design and Reference

Frame Effects on Unusual Attitude Recoveries, The International

Journal of Aviation Psychology, 23(1), 63-90.

Previc, F.H. & W.R. Ercoline (1999). The "Outside-In" attitude display

concept revisited. International Journal of Aviation Psychology,

9(4), 377-401.

Roscoe, S.N. (1968). Airborne displays for flight and navigation. Human

Factors 10, 321- 332.

Roscoe, S.N. (2004). Moving Horizons, Control Reversals and Graveyard

Spirals. Ergonomics in Design 12(4), 15-19.

Roscoe, S.N., Corl, L. & Jensen, R.S. (1981). Flight Display Dynamics

Revisited. Human Factors, 23, 341-353.

Roscoe, S.N. & Williges, R.C. (1975). Motion Relationships in aircraft

attitude and guidance displays: A flight Experiment. Human

Factors, 17, 374-387.

Wickens, C.D. (2003). Aviation Displays. In P. Tsang & M. Vidulich, (Eds.)

Principles and Practices of Aviation Psychology. Lawrence

Erlbaum Associates, Mahwah, NJ, pp. 147-199.

Yamaguchi, M. & Proctor, R.W. (2010). Compatibility of motion information

in two aircraft attitude displays for a tracking task. American

Journal of Psychology, 123, 81-92.

Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 1037