A FLIGHT SIMULATOR STUDY TO EVALUATE MANUAL FLYING
SKILLS OF AIRLINE PILOTS
Andreas Haslbeck
1
, Paul Kirchner
1
, Ekkehart Schubert
2
, and Klaus Bengler
1
1
Institute of Ergonomics, Technische Universität München,
2
Institute of Aeronautics and Astronautics, Sec-
tion Flight Guidance and Air Transportation, Technische Universität Berlin
This paper reports an experimental study with the objective to assess pilots’ raw-data-based flight perfor-
mance which is affected by long-term practice and structured training. Fifty-seven airline pilots with differ-
ent levels of aviation experience scheduled on an Airbus fleet, representing contrary levels of practice and
training, had to fly a simulated 45 minutes approach and landing scenario while flight performance data
were objectively recorded. The level of practice and training was found to have a significant influence on
manual flying skills. Pilots with low levels of practice and training showed a large variance in manual flight
performance; pilots with high levels of practice and training demonstrated high and homogenous perfor-
mance.
INTRODUCTION
For today’s aviation experts, manual flying is a critical issue.
The aviation industry, primarily aircraft manufacturers and air
carriers, is trying to manage the trade-off between safe and
economic flight operation. On the one hand, they put emphasis
on the training of their pilots’ manual skills, but on the other
hand, obligatory manual training hours are to be reduced with
every new type of aircraft introduced, in order to reduce the
effort for necessary type rating training sessions. In preventing
accidents, the pilot’s manual flying skills are considered as the
last line of defense: if all the automation breaks down and
manual operation becomes necessary, a pilot is still in charge
of conducting a safe landing on his own. Two very prominent
disasters, indicating lacking manual flying skills are Air
France flight 447 and Asiana Airlines flight 214 (BEA, 2012;
NTSB, 2014)
Although much research has been done in the field of aviation,
empirical studies on manual flight performance in specific
comparable scenarios are still rare. Some of the more promi-
nent works in this area are discussed below. In the present
paper, the question is addressed whether airline pilots can
maintain sufficient manual flying skills by recurrent training
and daily flight practice over the course of a pilot’s career.
The experimental results are derived from a flight simulator
study, which was performed in cooperation with a major Eu-
ropean airline.
Acquisition of manual flight
Manual control of an aircraft is an active task relative to when
pilots monitor the aircraft under automation (Flach, 1990;
Sarter & Woods, 1994); also known as a closed-loop control
problem (Field & Harris, 1998; Wickens, 2003). Manual fly-
ing is a psychomotor process requiring more than operating
the control stick of an aircraft. Three main stages of infor-
mation processing have to be considered in manual flying:
perception, cognitive processing, and response execution
(Childs & Spears, 1986). One model frequently referred to for
these sequent stages was founded by Wickens (Wickens &
Hollands, 1999). In flight school, pilots learn and intensively
train these active processes before they are introduced to au-
tomation, which then switches their task as a pilot from han-
dling an aircraft to managing it (Childs & Spears, 1986; JAA,
2006). From this point on, pilots are faced with automation
induced skill degradation (Balfe, Wilson, Sharples, & Clarke,
2012), caused by the automation taking over the responsibility
for tasks previously performed by the human operators (Par-
asuraman & Riley, 1997).
Automation-induced changes on the flight deck
As flying becomes more automated, pilot´s manual flying
skills degrade. This inverse relationship is primarily caused by
the automation altering the active flying task to a passive mon-
itoring task (Sarter & Woods, 1995). The introduction of early
glass cockpits (late 1970s), flight management systems, and
fly-by-wire control (late 1980s) in commercial aviation were
significant automation milestones. Billings (1991) also de-
scribed these changes in terms of information, management,
and control automation. Automation helps the human operator
in difficult situations when incapacity, workload, fatigue or
inaccuracy occur – just to name a few. Well known ironies of
automation describe negative automation effects like skill deg-
radation (Billings, 1991; Endsley & Kiris, 1996) or reduced
operator vigilance (Endsley, 1999).
Evidence-based experimental studies to assess degrading
manual flying skills
In the mid-1980s, empirical studies shed some light on this
issue. While a large number of experiments explored exposure
to automation and related situation awareness, only a few at-
tempts were made at investigating and measuring the devel-
opment and degradation of manual flying skills under automa-
tion.
An early effort to warn of diminishing flying skills was made
by Childs and Spears (1986). They postulated a concern that
ineffective perceptual processes lead to deteriorated motor
responses. Sarter and Woods (1994) reported a study focusing
on pilots’ mental models of the flight management system.
Their findings revealed that these mental models do not ac-
Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 11
Copyright 2014 Human Factors and Ergonomics Society. DOI 10.1177/1541931214581003
count for monitoring skills and, to some degree, manual skills,
as they require active interventions by the pilot, like the loss of
glideslope information. The transition from conventional flight
decks towards automated ones was evaluated in a simulator
study by Veillette (1995). In an experiment with non-
voluntary airline pilots, he measured manual flight perfor-
mance across varying degrees of automation. The results of
this study showed significant differences in manual flight per-
formance between the two groups: the pilots accustomed to
automation had significantly larger deviations from the ideal
flight paths – evidence of degrading manual skills due to au-
tomation. A more recent study analyzing manual flight per-
formance was introduced by Gillen (2008). His comparison
showed that pilots’ self-assessment delivered higher ratings
than pilots were able to perform in the simulator experiment –
pilots’ confidence in their own skills was subject to a bias. In
addition, the pilots tested performed below certification stand-
ards, which means these subjects would have failed in a certi-
fication situation.
Ebbatson (2009) showed in a large-scale analysis of manual
flight performance data a correlation between manual flying
skills and practice rather than overall flight experience. Recent
flight practice including manual flying occurring a few weeks
prior to the experiment had more influence on the measured
performance than flight hours accumulated over a pilot’s en-
tire career. Finally, Ebbatson suggests replicating his experi-
ment with a group of long-haul pilots, where he assumes even
stronger effects between manual flight performance and prac-
tice would be found.
These studies all had specific foci, but cannot deliver a com-
prehensive view on the performance of pilots with a low level
of practice and only few opportunities for training, like pilots
in long-haul operation. Attempting to fill this gap, the follow-
ing study was conducted to focus on long-haul pilots. Consid-
ering different approaches already mentioned, the current
study addresses the following aspects specifically (see also
Haslbeck et al. (2012)):
• A randomized sample for manual flying experiments
is necessary to avoid self-selection and volunteer bi-
ases, thus, participants should not be chosen on a
voluntary basis. Otherwise pilots with fairly high
skill levels tend to participate in such experiments.
• A highly realistic and valid standardized setting for
experimental simulator studies is needed, using a cer-
tified full flight simulator with motion effects e.g.
from light turbulence weather effects and having real
air traffic control instructions, which would force pi-
lots to handle a higher workload by distinguishing
between remote messages and their own messages in
radio communications.
• The difficulty of a scenario should deliver tasks that
can be fulfilled but should also give participants a
chance to fail due to their manual flying skills.
• The highly standardized progress sequence of the ex-
perimental simulation scenario should ensure that, in
general, all participants face the same technical, envi-
ronmental, and organizational conditions.
METHOD
Research Question and Aim of the Study
The main research question is: How do practice and training
influence manual flying skills? The concept ‘level of practice
and training’ (according to the German expression
Trainiertheit) stands for the manual flying skill level of pilots
and is affected by the following three aspects:
• passed time since initial flight school to account for
long-time skill degradation;
• daily flight practice to consider aspects of on-the-job
training of skill;
• the effect of flight simulator training lessons, when
selected flying tasks and maneuvers are repeatedly
practiced and tested under supervision.
Flight experience, for example in terms of flight hours, and the
level of practice and training are inversely proportional: while
their flight hours are continuously rising, for most pilots, sec-
tions including active handling are rare especially on the long-
haul. In this context, experience is rather meant as declarative
knowledge how to solve problems and tasks than implicit
knowledge or skill how to fly an aircraft. Therefore, it is ex-
pected that long-haul captains (CPTs) would have a lower skill
level than short-haul first officers (FOs), because it has been
longer since they attended flight school including systematic
initial flight training, and they have a significantly lower fre-
quency of recent flights than their short-haul colleagues.
Participants
To investigate the influence of practice and training, younger
FOs on short-haul schedules and elder CPTs on long-haul ser-
vice were chosen at random (stratified random sample), to
establish an extreme groups design, representing two typical
populations on both evaluated fleets. All participants occupied
the same seat as they do in line operation. Two Airbus-type
qualified full-flight simulators (JAR-FSTD A) were used for
this study because of the very comparable cockpit designs and
the resulting ease of transferring between different types
(communality). 27 male CPTs participated, representing a low
level of daily practice and training but a high level of opera-
tional experience. Their simulator was operated in an Airbus
A340-600 configuration. For the other group, representing a
high level of daily practice and training but a low level of op-
erational experience, 30 FOs (27 male, 3 female) took part in
this experiment in an Airbus A320-200 simulator. They should
have been in line operation for about five years. Two random-
ly selected CPTs reported sick and were replaced by two
equally qualified but voluntary CPTs. All participating pilots
experience four simulator events per annum. The CPTs had
more operational tasks (executive decisions) in their last two
simulator sessions prior to the experiment, while the FOs were
said to have experienced manual flying tasks. This means that
the kind of training for both groups of pilots ideally met the
experiment’s demands. Table 1 shows the demographical data,
showing flight experience as overall flight hours and years
since flight school, as well as the number of individually per-
formed landings within the past 30 days. Participating pilots
Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 12
were scheduled for the experiment by their company’s flight
operations department, so participation was part of their ser-
vice schedule and not on a voluntary basis. All pilots were
airline pilots (ATPL) and in service of the cooperating airline.
All CPTs held the type rating for A330/340 family aircraft,
and the FOs held the type rating for A319/320/321 family
types.
Table 1. Demographical data of participants (mean values).
age overall flight
hours
indiv. landings
in past 30 days
years since
flight school
CPTs (n=27) 50.4 15,019.7 3.4 24.6
FOs (n=30) 30.4 3,373.9 16.6 4.5
Procedure
The participating pilots were prepared the same way as for a
regular flight, wearing pilot uniforms and bringing their daily
used computers for the electronic flight bag system. Subjects
were always instructed to be the pilot flying (PF). A confeder-
ate pilot monitoring (PM) was instructed to have a passive but
cooperative role, and not to cause errors. These confederate
pilots (two alternating for each group of participants) were
also scheduled by the partner airline on the correspondent
fleet. The first subject began the experiment (three subjects
per night) approximately two and a half hours after the starting
time. The whole procedure resembles longer flights with land-
ings during the early morning hours, representing long-haul
flights from the east or mid-range flights operated with short-
haul aircraft in the partner airline (Haslbeck et al., 2012).
Scenario
After an uneventful flight from the east toward Munich Air-
port, the PF returned from his last break to perform the ap-
proach and landing 25 minutes prior to scheduled touchdown.
All flight crews had to perform a missed approach before in-
tercepting the ILS (guide beam provided by the instrument
landing system) for a second time. At this time the approach
mode could not be armed and the autopilot was disabled by a
scripted event. After this point, the pilots had to perform all
flying activities manually without the flight director and auto-
pilot assistance. When the localizer was manually intercepted
– providing runway centerline guidance – the measurement of
manual flight performance started. A hand-flown landing (raw
data ILS) with touchdown ended the 45-min. scenario.
Dependent Measures
Pilots were instructed to act and fly according to standard op-
erating procedures of their airline (including licensing stand-
ards) – the same as in a real flight. Flight performance data
were objectively measured by the flight simulator’s data re-
corder. Here, deviations from the ideal glide slope (vertical
guidance), and localizer (lateral guidance) were measured.
These metrics represent the resulting system performance ac-
cording to a control loop including the pilot and the aircraft
(Morris & Miller, 1996). Flight path deviations can be consid-
ered in two different ways: measurement of absolute values
with averaging afterwards or comparing maximum deviations
to licensing standards. Both approaches were pursued and are
subsequently shown.
All data for the localizer and glide slope are standardized to
aberrations in dots, a unit which can be monitored on the pri-
mary flight display in the cockpit and which gives pilots in-
formation about their actual attitude with respect to the ideal
approach path. The individual glide slope variations for all
participants are observed from 3,000 ft. AGL (above ground
level) down to 200 ft. AGL, whereas localizer aberration is
considered significant from 3,000 ft. AGL to the model height
of the aircraft above the threshold of 50 ft. AGL.
According to partner airline manuals, guidelines and laws
(JAA, 2006), a maximum variance of one dot deflection on
each side of the primary flight display localizer and glide
slope scale must be maintained on precision approaches at all
times. All measures can be directly compared to pilots’ licens-
ing standards (JAA, 2006; Ebbatson, 2009). To complement
the comparison of maximum deviation values to legal stand-
ards, the root mean square error (RMSE) was calculated for all
pilots to give a combined measure of their accuracy, equally
weighting mean error and standard deviation (Hubbard, 1987;
Flach, 1990). Here this measure is taken to express the differ-
ences between both groups, rather than to distinguish between
the directions of both the localizer and glide slope.
RESULTS
Pilots´ manual flying performance in terms of maximum local-
izer and glide slope deviations from the manually flown ILS
approach is shown in figure 1 and 2. These two diagrams
evaluate pilots’ skill against licensing standards (± 1dot max.).
Figure 1: participants’ maximum localizer deviations
Figure 2: participant’s maximum glide slope deviations
Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 13
Each bar represents the maximum deviation for each pilot
from the target value in figure 1 and 2. Positive localizer devi-
ations imply a horizontal drift to the right of the runway cen-
terline, with negative deviations correspondingly to the left.
Vertical drift information is provided via the glide slope indi-
cator. Positive deviances equal an aircraft position higher than
the ideal glide path, while negative deviances in contrast rep-
resent a lower than ideal position of the airplane. Given the
results depicted in Figure 1, six out of 57 subjects violated
restrictions on the allowed localizer variance. For glide slope
deviation, eight out of 57 participants could not perform with-
in the acceptable limits. A total of nine different pilots (15.8
%) did not meet the mandatory skill test requirements in this
scenario. Relative to test-person groups, seven (25.9 %) out of
27 CPTs did not fulfil at least one of the binding ILS deviation
parameters, while two (6.7 %) out of 30 FOs did not. For lo-
calizer and glide slope deviations, the RMSE is shown in Fig-
ure 3.
Figure 3: RSME for localizer and glide slope deviations
Table 2: statistical analysis of ILS deviations data
mean Independent
t-Test
Mann-Whitney
test
t (df); p; r U; z; p; r
max.
localizer
devia-
tions
left
CPTs .254
t(36.12)=2.55
2; p< .008;
r= .39
U=249.5;
z=2.489;
p= .006; r= .33
FOs .143
right
CPTs .646 t(55)=2.873;
p= .003;
r= .36
U=70; z=5.356;
p< .001; r= .71
FOs .298
RMSE localizer
CPTs .025
t(23.86)=5.19
1; p< .001;
r= .73
U=28; z=5.629;
p< .001; r= .78
FOs .010
max.
glide
slope
devia-
tion
high
CPTs .796 t(55)=2.077;
p= .021;
r= .27
U=154;
z=4.013;
p< .001; r= .53
FOs .481
low
CPTs .471 t(55)=3.08;
p< .002;
r= .38
U=211;
z=3.102;
p= .001; r= .41
FOs .271
RMSE glide
slope
CPTs .035
t(24.48)=5.49
3; p< .001;
r= .74
U=25; z=5.684;
p< .001; r= .79
FOs .015
The two groups were compared using parametric as well as
non-parametric tests, as assumptions of normality could not be
met for all deviation data. Both tests using an alpha-level of
.05 indicate highly significant differences between the two
groups’ manual flying performances. In addition, the effect
sizes expressed by a point-biserial r show moderate to very
large effects.
DISCUSSION AND CONCLUSIONS
Evidence was found that participating CPTs with a lower level
of practice and training (table 1) cause larger deviations from
the ideal approach parameters than their more practiced FO
colleagues; even if one limitation to this study is the fact that
A320 and A340 differ in size, weight, and handling character-
istics. From a technical or aeronautical perspective, these are
completely different types of aircraft, while seen from a hu-
man factors view, both are successive milestones on a pilot’s
career and the human-machine-interfaces thinly differ because
of the communality design principle of Airbus planes. When
thinking about the differences in both types, the A340 is over
four times heavier than its smaller counterpart and so it has
larger flight path inertia. However both types are controlled by
comparable roll and pitch rates, realized by larger control sur-
faces. In addition, the A340’s higher flight path inertia might
be an advantage in case of external perturbations like gusty
wind. Based upon licensing standards on the one hand and
flight operation realities on the other hand, all pilots have to
perform within the same limits. Licensing standards neither
differ between CPTs and FOs nor between long-haul or short-
haul. Moreover both groups of pilots as well as both types of
aircraft can use the same airports and runways. Thus require-
ments for manual aircraft handling are the same for all pilots.
Another argument for choosing this comparison between
A320 FOs and A340 CPTs is the assumed maximum range
between the pilots’ different skill levels (table 1) to utilize an
extreme groups design.
That degrading manual flying skills have been observed in this
rather small sample of pilots suggests that this is likely more
prevalent than one could have suspected. As participants were
active professional pilots, the results should be valid for other
airline’s personnel. In several cases, the deviations from ideal
performance are large enough that pilots would have even
failed a check situation – a dramatic finding that could reflect
inadequate maintained skills. As training lessons normally
cover equal contents for short-haul and long-haul pilots in
longer sequences, one can assume that differences in manual
flight performance are instead a consequence of everyday
flight practice. This hypothesis is also supported by Ebbat-
son’s (2009) study: accordingly, recent flight practice result-
ing from frequent flight operations, seem to be the most im-
portant factor in maintaining manual flying skills. Duncan,
Williams and Brown (1991) have found some comparable
insights in a real driving car experiment, “that adequate driv-
ing skills cannot be assumed, even for the `average´ experi-
enced motorist, simply because they once were mastered
[…]”. In comparison, simulator training can instead teach the
right techniques for handling the aircraft (Buckley & Caple,
2009).
A further limitation to this study is that the level of practice
and training is confounded with pilot’s age and experience.
Tsang (2003) describes and cites findings in her comprehen-
sive review that “older, experienced individuals do not neces-
Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 14
sarily perform more poorly than their younger counterparts in
tasks specific to their domain of expertise.” Taylor, Kennedy,
Noda, and Yesavage (2007) have reported a study investigat-
ing performance changes of pilots with different age and also
under regard of different levels of expertise. Their results indi-
cate no strong decline of landing skills by experienced (ATPL)
pilots over time. In spite of these findings, in a field experi-
ment with airline pilots, their level of practice and training will
always be partially confounded with age and experience. For
the measurement of manual flying skills, operational flight
experience plays only a minor role and in an airline’s daily
operation, experience and the level of practice and training
normally develop contrarily: CPTs have accumulated a vast
amount of flight experience but experience only very few op-
portunities to practice flight skills – neither in simulator ses-
sions nor in real operation. In spite of these limitations, the
results of this study deliver a highly valuable picture of pro-
fessional pilots’ ability to manually control an aircraft.
Long-haul operation with its high degree of automation and
pilots’ long exposition to automated systems, was shown to
have an eroding effect on manual flying skills; pilots with re-
duced flight duties and part-time schedules, like management
pilots or ones who are on parental leave, should be kept in
mind. Some examples to be supposed to airlines to implement
strategies against deteriorating skills: additional simulator
training sessions as well as type rating trainings concentrating
on manual aircraft handling; combining short-haul and long-
haul operation for long-haul pilots (mixed-fleet flying), espe-
cially for CPTs suffering from a lack of practice opportunities.
For human-machine-interface designers, the approach of adap-
tive automation (Parasuraman, 2000) could also lead to a more
flexible and dynamic task sharing between human and auto-
mation in the near future.
Future studies should further operationalize and analyze the
influence of simulator training sessions. In addition, further
groups of pilots with medium levels of practice and training,
like short-haul CPTs and long-hauls FOs could complement
insights in pilot’s manual flying skills.
Balfe, N., Wilson, J. R., Sharples, S., & Clarke, T. (2012). Development of
design principles for automated systems in transport control. Ergonomics,
55(1), 37–54.
BEA. (2012). Final Report on the accident on 1st June 2009 to the Airbus
A330-203 registered F-GZCP operated by Air France flight AF 447 Rio de
Janeiro - Paris. Le Bourget.
Billings, C. E. (1991). Human-Centered Aircraft Automation: A Concept and
Guidelines: NASA-TM-103885. Moffet Field, California.
Buckley, R., & Caple, J. (2009). The theory & practice of training, 6th edition
(6th ed.). London: Kogan Page.
Childs, J. M., & Spears, W. D. (1986). Flight-skill decay and recurrent train-
ing. Perceptual and Motor Skills, 62(1), 235–242.
Duncan, J., Williams, P., & Brown, I. (1991). Components of driving skill:
experience does not mean expertise. Ergonomics, 34(7), 919–937.
Ebbatson, M. (2009). The Loss of Manual Flying Skills in Pilots of Highly
Automated Airliners (PhD Thesis). Cranfield University, Cranfield.
Endsley, M. R. (1999). Level of automation effects on performance, situation
awareness and workload in a dynamic control task. Ergonomics, 42(3),
462–492.
Endsley, M. R., & Kiris, E. O. (1995). The Out-of-the-Loop Performance
Problem and Level of Control in Automation. Human Factors, 37(2), 381–
394.
Field, E., & Harris, D. (1998). A comparative survey of the utility of cross-
cockpit linkages and autoflight systems' backfeed to the control inceptors of
commercial aircraft. Ergonomics, 41(10), 1462–1477.
Flach, J. M. (1990). Control With an Eye for Perception: Precursors to an
Active Psychophysics. Ecological Psychology, 2(2), 83–111.
Gillen, M. (2008). Degradation of Piloting Skills (Master's Thesis). University
of North Dakota, Grand Forks.
Haslbeck, A., Schubert, E., Onnasch, L., Hüttig, G., Bubb, H., & Bengler, K.
(2012). Manual flying skills under the influence of performance shaping
factors. Work: A Journal of Prevention, Assessment and Rehabilitation,
41(Supplement 1/2012), 178–183.
Hubbard, D. C. (1987). Inadequacy of root mean square error as a perfor-
mance measure. In International Symposium on Aviation Psychology, 4th
(pp. 698–704).
Joint Aviation Requirements - Flight Crew Licensing (Aeroplane), Joint Avia-
tion Authorities 2006.
Morris, T. L., & Miller, J. C. (1996). Electrooculographic and performance
indices of fatigue during simulated flight. Biological Psychology, 42(3),
343–360.
NTSB. (2014). Crash of Asiana Flight 214 Accident Report Summary: De-
scent Below Visual Glidepath and Impact With Seawall. Public Meeting of
June 24, 2014. Washington, DC.
Parasuraman, R. (2000). Designing automation for human use: empirical
studies and quantitative models. Ergonomics, 43(7), 931–951.
Parasuraman, R., & Riley, V. (1997). Humans and Automation: Use, Misuse,
Disuse, Abuse. Human Factors, 39(2), 230–253.
Sarter, N. B., & Woods, D. D. (1994). Pilot Interaction With Cockpit Automa-
tion II: An Experimental Study of Pilots' Model and Awareness of the
Flight Management System. The International Journal of Aviation Psychol-
ogy, 4(1), 1–28.
Sarter, N. B., & Woods, D. D. (1995). How in the World Did We Get into
That Mode? Mode Error and Awareness in Supervisory Control. Human
Factors, 37(1), 5–19.
Taylor, J. L., Kennedy, Q., Noda, A., & Yesavage, J. A. (2007). Pilot age and
expertise predict flight simulator performance: A 3-year longitudinal study.
Neurology, 68(9), 648–654.
Tsang, P. S. (2003). Assessing Cognitive Aging in Piloting. In P. S. Tsang &
M. A. Vidulich (Eds.), Principles and practice of aviation psychology
(pp. 507–546). Mahwah, N.J: Lawrence Erlbaum.
Veillette, P. R. (1995). Differences in aircrew manual skills in automated and
conventional flight decks. Transportation Research Record, (1480), 43–50.
Wickens, C. D. (2003). Pilot Actions and Tasks: Selections, Execution, and
Control. In P. S. Tsang & M. A. Vidulich (Eds.), Principles and practice of
aviation psychology (pp. 239–263). Mahwah, N.J: Lawrence Erlbaum.
Wickens, C. D., & Hollands, J. G. (1999). Engineering psychology and human
performance (3rd). Upper Saddle River, NJ: Prentice Hall.
Proceedings of the Human Factors and Ergonomics Society 58th Annual Meeting - 2014 15
