See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/12136068

Factors associated with pilot error in aviation crashes

Article  in  Aviation Space and Environmental Medicine · February 2001

Source: PubMed

CITATIONS READS

80 6,403

4 authors, including:

Guohua Li George W Rebok

Columbia University Johns Hopkins University
352 PUBLICATIONS   10,459 CITATIONS    242 PUBLICATIONS   10,226 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advanced Cognitive Training for Independent and Vital Elderly (ACTIVE) View project

Hopkins Memory Research Lab at Fairhaven View project

All content following this page was uploaded by Guohua Li on 03 September 2014.

The user has requested enhancement of the downloaded file.

Factors Associated with Pilot Error in
Aviation Crashes
GUOHUA LI, M.D., DR.P.H.,
SUSAN P. BAKER, M.P.H., Sc.D. (HON.),
JUREK G. GRABOWSKI,M.P.H., AND GEORGE W. REBOK, PH.D.
LI G, BAKERSP, GRABOWSKIJG, REBOKGW. Factors associated
w i t h p i l o t error in aviation crashes. Aviat Space Environ Med 2001;
72:52-8.
Background: The importance of pilot error in aviation crashes has
long been recognized. However, understanding and preventing pilot
error remains the foremost challenge in aviation safety. Objective: This
study aims to identify pilot characteristics and crash circumstances that
are associated with the presence of pilot error in a large sample of
aviation crashes. Methods: Different data files compiled by the National
Transportation Safety Board for 329 major airline crashes, 1,627 com-
muter/air taxi crashes, and 27,935 general aviation crashes for the years
1983-96 were merged; and the presence of pilot error was analyzed in
relation to variables indicating the characteristics of the pilot-in-com-
mand, crash circumstance, and aircraft. Multivariate logistic regression
modeling was performed to assess the associations of individual vari-
ables with the likelihood of pilot error given a crash. Results: Pilot error
was a probable cause in 38% of the major airline crashes, 74% of the
commuter/air taxi crashes, and 85% of the general aviation crashes.
Among the factors examined, instrument meteorological condition and
on-airport location were each associated with a significantly increased
odds of pilot error. The likelihood of pilot error decreased as pilot
certificate rating increased in commuter/air taxi and general aviation
crashes. Neither pilot age nor gender was independently associated with
the odds of pilot error. With adjustment for pilot characteristics and
crash circumstances, flight experience as measured in total flight time
showed a significant protective effect on pilot error in general aviation
crashes. Conclusions: The prevalence and correlates of pilot error in
aviation crashes vary with the type of flight operations. Adverse weather
is consistently associated with a significantly elevated likelihood of pilot
error, possibly due to increased performance demand.
Keywords: aviation safety, human factors, pilot error.
HE IMPORTANCE OF H U M A N factors, particu-
T larly pilot error, in aviation crashes has long been
recognized (9). Overall, about 80% of aviation crashes
and 50% of aviation incidents are attributed to pilot
error (5,11). As aviation hardware becomes more reli-
able due to advanced technology, the relative impor-
tance of human factors in aviation safety is likely to
further increase (15).
Empirical research on pilot error has been limited
primarily to describing the detailed operational and
behavioral acts that contribute to aviation mishaps and
to classifying those deviant acts according to various
taxonomies (15). Although these studies have helped to
delineate the profiles and patterns of pilot error, they
are descriptive in essence and provide little information
about the etiology of pilot error. The lack of epidemio-
logic data on etiological factors underlying pilot error
52
has made it difficult, if not impossible, to develop ef-
fective interventions.
In this study, we examined the prevalence and cor-
relates of pilot error in a large sample of aviation
crashes, using multiple data files compiled by the Na-
tional Transportation Safety Board (NTSB) over an ex-
tended period. The results indicate that the likelihood
of pilot error in a given crash is a function of both
endogenous and exogenous factors. In contrast to the
widely accepted notion that pilot error is entirely an
intrinsic phenomenon of human behavior, this study
reveals that extrinsic attributes, such as weather condi-
tion, are important determinants of pilot error as well.
METHODS
The NTSB reports all civilian aviation crashes using a
set of standard forms. A crash ("accident" in NTSB
terminology) is defined as an event associated with
operation of an aircraft that results in personal death or
serious injury, or substantial aircraft damage. The Fac-
tual Report (NTSB Form 6120.4) is the main data form
for recording information collected from field investi-
gations. There are over 200 data items in the Factual
Report, providing detailed information about the crash
circumstances, aircraft, and pilot-in-command. Also in-
cluded in the Factual Report is a narrative statement of
the facts, conditions and other findings pertinent to the
crash.
Based on information in the Factual Report, the se-
quence-of-events of each crash is delineated by NTSB
investigators according to established protocols. For
From the Department of Emergency Medicine (S. P. Baker, J. G.
Grabowski, G. Li), Johns Hopkins University School of Medicine,
Baltimore, M.D.; and Departments of Health Policy and Management
(S.P. Baker, G. Li), and Mental Hygiene (G.W. Rebok), Johns Hopkins
University School of Hygiene and Public Health, Baltimore, M.D.
This manuscript was received for review in November 1999. It was
revised in September 2000. It was accepted for publication in Septem-
ber 2000.
Address reprint requests to: Guohua Li, M.D., Dr.P.H., who is an
associate professor, Department of Emergency Medicine, Johns Hop-
kins University School of Medicine, 1830 East Monument Street, Suite
6-100, Baltimore, M.D. 21205; ghli@jhmi.edu.
Reprint & Copyright 9 by Aerospace Medical Association, Alexan-
dria, VA.
Aviation, Space, and Environmental Medicine ~ Vol. 72, No. 1 ~ January 2001
CORRELATES OF PILOT ERROR--LI ET AL.

each crash, up to five major events of relative signifi-

cance that led to the crash are recorded in chronological 90

order. For each identified event in the crash sequence, 80

causes and factors related to the aircraft, environment
70
and human are documented using standard coding
schedules (17). A cause is a condition that played an 6O

essential role in the causation of the crash (i.e., without

its presence, there would have been no crash); and a
4O
factor is a condition that aggravated the development of
the crash sequence (i.e., it made the crash more likely or
more severe). Coded data on the sequence of events are 2O
General Aviation
stored in a separate file in the NTSB data system. In this --c-
-~- Comrnuter/AirTaxi
10
study, pilot error is operationally defined as any pilot --o- Major Aidines

performance that was identified by NTSB investigators 0

as the probable cause of one or more events in the crash
sequence.
This study included all crashes of airplanes and he-
licopters that occurred during 1983-96 under regula-
tions 14 CFR 91 (general aviation), 14 CFR 121 (major
airlines) or 14 CFR 135 (commuter/air taxi). Incidents
that resulted in only minor damage to the aircraft or
minor injury to an occupant were excluded from this
study. Data files for the Factual Report and for the
sequence of events were merged by matching the
unique NTSB crash identification numbers. Sequence-
of-events data were available for 94% of all the crashes
recorded by the NTSB.
Analysis was focused on examining the associations
between pilot error and pilot characteristics (age, gen-
der, total flight time, and certificate rating) and crash
circumstances (time and location of crash, basic weather
condition at crash, type of aircraft, and crash severity
index). Since this study included three major categories
of civil aviation that are operated under different reg-
ulations and environments, and that involve different
populations and aircraft, data were analyzed and pre-
sented separately for major airline crashes, commuter/
air taxi crashes, and general aviation crashes. Total
flight time was categorized based on the quartiles for
each of the three aviation groups, with quartile 1 rep-
resenting the lowest total flight time group and quartile
4 the highest total flight time group. Pilot certificate
rating was determined based on the highest category of
license held b y the pilot-in-command. Statistical analy-
ses were conducted using the Statistical Analysis Soft-
ware (SAS Institute Inc., Cary, NC) at both bivariate
and multivariate levels. The logistic regression model
(10) was used in the multivariate analysis, with the
binary outcome variable being whether pilot error was
present in the crash. The goodness of fit was assessed
based on the Hosmer-Lemeshow statistic (10).
The logistic regression model is a standard statistical
method for describing the relationship between a bi-
nary outcome variable and a set of independent vari-
ables. The specific form of the logistic regression model
is as follows (10):
e~0+~,• . +~pxp
It(x) = 1 + e ~~ "' +~p•
1983 1984 1985 1986 1987 7988 1989 19g0 ~991 1992 1993 1994 1995 1996
Year of Crash
Fig. 1. Prevalence rates of pilot error in aviation crashes by aviation
category, United States, 1983-1996.
The logit transformation can make the logistic regres-
sion model a linear regression model (10):
F ,'n'(x) 3
g(x) = lnLl_~Z~J = /3o+/33xl + ... + /3pxp
The logit g(x) is linear in its parameters, and may range
from -m to +% depending on the range of x. The pa-
rameters, /30, /31..... /3p are estimated using maximum
likelihood and interpreted as the log odds ratios.
RESULTS
During the 14-yr study period, the NTSB recorded a
total of 371 major airline crashes, 1,735 commuter/air
taxi crashes, and 29,798 general aviation crashes. Se-
quence-of-events data linkage were available for 329
(89%) of the major airline crashes, 1,627 (94%) of the
commuter/air taxi crashes, and 27,935 (94%) of the
general aviation crashes. Crashes without the sequence-
of-events data were significantly more likely than
crashes with the sequence-of-events data to have oc-
curred away from airports and to have involved less
experienced pilots. However, these two groups of
crashes were similar in the compositions of pilot age
and gender, time of crash, type of aircraft, and basic
weather conditions.
Prevalence o f Pilot Error
Analysis of pilot error was limited to the 29,891
crashes for which the sequence-of-events data were
available. Overall, pilot error was listed as a probable
cause for 38% of the 329 major airline crashes, 74% of
the 1,627 commuter/air taxi crashes, and 85% of the
27,935 general aviation crashes. The prevalence of pilot
error was fairly stable over the 14-yr study period,
particularly in commuter/air taxi crashes and general
aviation crashes (Fig. 1). Pilot error in major airline
crashes that occurred during 1990-1996 appeared to be
less prevalent than in those occurring during 1983-1989
(34% vs. 43%, p -- 0.09) (Table I).

Correlates of Pilot Error

where ~r(x) is the conditional mean of the outcome
variable given x l , . . . , Xpwhen the logistic distribution is At the bivariate level, several pilot characteristics
applied. were significantly associated with pilot error in general

Aviation, Space, and Environmental Medicine ~ Vol. 72, No. 1 ~ January 2001 53
 CORRELATES OF PILOT ERROR--LI ET AL.

TABLE I. PROPORTIONS OF AVIATION CRASHES WITH PILOT ERROR BEING A PROBABLE CAUSE BY PILOT CHARACTERISTIC,
CRASH CIRCUMSTANCE AND AVIATION CATEGORY, U.S., 1983-96.

Major Airlines Commuter/Air Taxi General Aviation

Crashes Crashes Crashes
involving involving pilot involving pilot
Total pilot error Total error Total error
crashes* No. % crashes* No. % crashes* No. %
Pilot Gender
Male 328 126 38.4 1,580 1,161 73.5 2G874 22,900 85.2
Female 1 0 0 47 40 85.1 1061 945 89.1
X2 = 3.2, X2 = 12.1,
p = 0.07 p < 0.001
Total Flight Time Quartile
1 88 40 45.5 427 323 75.6 7,783 7,119 91.5
2 75 21 28.0 402 295 73.4 6,756 5,873 86.9
3 87 34 39.1 402 301 74.9 G719 5,560 82.8
4 79 31 39.2 396 282 71.2 6,677 5,293 79.3
X2 = 5.3, X2 = 2.4, X2 = 480.3,
p = 0.15 p = 0.49 p < 0.001
Type of Certificate
Student/Private 555 425 76.6 16,846 14,900 88.4
Commercial 1,072 776 72.4 5,143 4,195 81.6
Airline Transport 329 126 38.3 329 126 38.3 5,946 4,750 79.9
;r = 3.3, X2 = 330.3,
p = 0.07 p < 0.001
Year of Crash
1983-1989 158 68 43.0 966 717 74.2 15,804 13,738 86.9
1990-1996 171 58 33.9 661 484 73.2 12,131 10,107 83.3
)(2 = 2.9, X2 = 0.2, )(2 = 71.6,
p = 0.09 p = 0.65 p < 0.001
Time of Crash
6:00 AM-5:59 PM 196 73 37.2 1,093 807 73.8 21,849 18,524 84.8
6:00 PM-5:59 AM 133 53 39.8 534 394 73.8 6,086 5,321 87.4
)(2 = 0.2, 3(2 = 0.0, X2 = 26.7,
p = 0.63 p = 0.98 p < 0.001
Location of Crash
Off Airport 128 36 28.1 928 702 75.6 14,703 12,112 82.4
On Airport 198 89 44.9 672 478 71.1 12,955 11,496 88.7
x ~ = 9.3, X2 = 4.1, )r 222.9,
p < 0.01 p = 0.04 p < 0.001
Weather
Visual Meteorological 256 84 32.8 1,257 870 69.2 25,336 21,360 84.3
Instrument
Meteorological 66 39 59.1 340 302 88.8 2,339 2,240 95.8
X2 = 15.3, X2 = 52.7, )2 = 224.0,
p < 0.001 p < 0.001 p < 0.001
Type of Aircraft
Plane 329 126 38.3 1,340 1,012 75.5 26,211 22,491 85.8
Helicopter 287 189 65.9 1,724 1,354 78.5
X2 = 68.4, X2 = 11.4,
p < 0.001 p < 0.001
Crash Severity Index
Non-fatal 284 102 35.9 1,224 860 70.3 22,339 18,625 83.4
Fatal 45 24 53.3 403 341 84.6 5,596 5,220 93.3
= 5.0, X2 = 32.3, X2 = 351.4,
p = 0.03 p < 0.001 p < 0.001

* Totals may vary within groups due to missing data.

aviation crashes and commuter/air taxi crashes (Table in the lowest quartile to 79% in the highest quartile
I). O f t h e g e n e r a l a v i a t i o n c r a s h e s i n v o l v i n g p i l o t s u n - ( p < 0.001) ( T a b l e I). A f u r t h e r e x a m i n a t i o n re-
d e r a g e 20, 9 4 % w e r e a t t r i b u t e d t o p i l o t e r r o r , c o m p a r e d vealed that the protective effect of total flight time
with about 85% of general aviation crashes involving against pilot error existed only in general aviation
o l d e r p i l o t s (p < 0.001) ( F i g . 2). A g e - r e l a t e d v a r i a t i o n i n crashes under visual meteorological conditions (VMC)
t h e p r e v a l e n c e r a t e s o f p i l o t e r r o r w a s s t a t i s t i c a l l y in- (Fig. 3).
significant for major airline crashes and commuter/air General aviation crashes involving student/private
taxi crashes. pilots were more likely to be attributed to pilot error
Total flight time showed an effect on pilot error than other g e n e r a l a v i a t i o n c r a s h e s ( T a b l e I). A l o w e r
only in general aviation crashes, with the prevalence prevalence of pilot error was also observed in commut-
rate decreasing progressively from 91% for pilots er/air taxi crashes involving pilots who held airline

54 Aviation, Space, and Environmental Medicine ~ Vol. 72, No. 1 ~ January 2001
CORRELATES OF PILOT ERROR--LI ET AL.
113o
90
8o
SO
4o
3o
20
-o- General Aviation
10
Co~/AJr Taxi
-o- Major Aidines
o
<20 20-29 30~3e 4O-49 5O-59 >6O
Pilot Age (years)
Fig. 2. Prevalence rates of pilot error in aviation crashes by pilot age
and aviation category,UnitedStates, 1983-1996.
transport certificates, as compared with commuter/air
taxi crashes involving commercial pilots.
Female pilots accounted for 0.3% of major airline
crashes, 3% of commuter/air taxi crashes, and 4% of
general aviation crashes. Partly reflecting the effect of
lesser flight experience, a higher proportion of general
aviation crashes and commuter/air taxi crashes involv-
ing female pilots were caused by pilot error than those
involving rnale pilots (Table I). Female pilots who were
involved in general aviation crashes recorded an aver-
age 888 total flight hours (SD 2,845 h), compared with
2,411 h (SD 4,426 h) for their male counterparts (p <
0.001). A striking gender discrepancy in total flight time
also existed among pilots who were involved in com-
muter/air taxi crashes [mean total flight time: 5,795 h
(SD 4,891 h) for males and 3,257 h (SD 2,269) for fe-
males, p < 0.001].
With regard to crash circumstances, instrument me-
teorological conditions (IMC) were associated with a
significantly higher prevalence of pilot error, irrespec-
tive of the type of flight operations (Table I). An ele-
vated prevalence of pilot error was found in major
airline crashes and general aviation crashes occurring
on airports, as compared with crashes away from air-
ports (Table I). Whereas the proportion (40%) of major
airline crashes that occurred at nighttime (6 p m to 5:59
a.m.) was considerably higher than that for c o m m u t e r /
air taxi crashes (33%) and for general aviation crashes
(22%), the prevalence of pilot error was similar between
daytime and nighttime crashes in each of the three
aviation categories (Table I).
Helicopters constituted 18% of commuter/air taxi
crashes and 6% of general aviation crashes; and pilot
error was less prevalent in helicopter crashes than in
airplane crashes. More than half (53%) of the fatal major
airline crashes were attributed to pilot error, compared
with 36% of the nonfatal major airline crashes (p =
0.03). The prevalence of pilot error was also signifi-
cantly higher in fatal commuter/air taxi crashes and
fatal general aviation crashes (Table I).
Logistic Regression M o d e l i n g
The associations between individual variables and
pilot error were further assessed at the multivariate
Aviation, Space, and Environmental Medicine ~ Vol. 72, No. 1 9 January 2001
level. Since the correlates of pilot error identified in the
bivariate analysis varied with the type of aviation, a
logistic regression model was fitted separately for major
airline crashes, commuter/air taxi crashes, and general
aviation crashes (Table II). With adjustment for pilot
characteristics and other covariates, crashes under IMC
were significantly more likely to involve pilot error than
crashes under VMC, with an estimated odds ratio of 3.9
for major airline crashes, 2.9 for commuter/air taxi
crashes, and 3.8 for general aviation crashes. Other risk
factors of pilot error identified from the multivariate
logistic models included on-airport location for major
airline crashes and general aviation crashes, and being
fatal for commuter/air taxi crashes and general aviation
crashes (Table II).
The effects of pilot characteristics on the likelihood of
pilot error was primarily limited to general aviation
crashes (Table II). The odds of pilot error in general
aviation crashes decreased as pilots' total flight time
and certificate rating increased. The interaction effect of
total flight time and basic weather condition on pilot
error was assessed during the modeling and was not
included in the final model because of statistical insig-
nificance. With regard to pilot age, a significantly ele-
vated odds ratio existed only among pilots younger
than 20 yr in general aviation crashes. There was not an
independent gender effect on pilot error. Among vari-
ables indicating pilot characteristics in commuter/air
taxi crashes, certificate rating was the only one that
showed a significant effect on pilot error, with an esti-
mated odds ratio of 0.7 for airline transport pilots rela-
tive to commercial pilots (Table II). In the multivariate
logistic regression model for major airline crashes, there
were only two pilot characteristics (age and total flight
time) feasible for analysis. Whereas the odds of pilot
error in major airline crashes appeared to be greater for
pilots in the lowest total flight time quartile, there was
not a significant age effect (Table II). Goodness-of-fit
tests based on the Hosmer-Lemshow statistic indicated
that the estimated logistic models presented in Table II
fit the data well.
--o- IMC
VMC
1 2 3 4
Total Flight Time Quartile
Fig. 3. Prevalence rates of pilot error in general aviation crashes by
pilot total flighttime and basic weathercondition, UnitedStates, 1983-
1996.
55
CORRELATES OF P I L O T E R R O R - - L I ET A U
TABLE II. ESTIMATEDODDS RATIOS (OR) OF PILOT ERROR IN AVIATION CRASHES AND 95% CONFIDENCE INTERVALS (CI)
FROM MULTIVARIATELOGISTIC REGRESSIONMODELS BY AVIATION CATEGORY, U.S., 1983-96.

Major Airlines Commuter/Air Taxi General Aviation

OR 95% CI OR 95% CI OR 95% CI
Pilot Age (yr)
<20 1.63 1.06-2.49~
20-29 0.77 0.53-1.12 1.09 0.97-1.23
30-39 1.38 0.68-2.78 1.03 0.75-1.43 1.04 0.95-1.15
40-49* 1.00 1.00 1.00
50-59 1.23 0.68-2.24 1.18 0.77-1.83 1.02 0.92-1.13
-> 6O 1.91 0.83-4.35 1.13 1.01-1.28~
Pilot Gender
Male* 1.00 1.00
Female 2.19 0.95-5.07 1.12 0.92-1.38
Total Flight Time Quartile
1" 1.00 1.00 1.00
2 0.39 0.19-0.80+ 0.98 0.70-1.38 0.66 0.59-0.73t
3 0.68 0.34-1.37 1.02 0.71-1.46 0.50 0.45-0.57t
4 0.82 0.37-1.80 0.79 0.53-1.18 0.43 0.37-0.49*
Type of Certificate
Student/private 1.18 1.07-1.31*
Commercial* 1.00 1.00
Airline Transport 0.71 0.55-0.92~ 0.92 0.83-1.01
Year of Crash
1983-1989" 1.00 1.00 1.00
1990-1996 0.66 0.40-1.08 0.85 0.67-1.07 0.77 0.72--0.83*
Location of Crash
Off Airport* 1.00 1.00 1.00
On Airport 3.09 1.79-5.31+ 0.92 0.71-1.19 2.14 1.99-2.30+
Weather
Visual Meteorological* 1.00 1.00 1.00
Instrument Meteorological " 3.89 2.09-7.26* 2.88 1.99-4.16t 3.80 3.07-4.70*
Type of Aircraft
Plane* 1.00 1.00
Helicopter 0.56 0.41-0.78t 0.97 0.82-1.11
Crash Severity Index
Non-fatal* 1.00 1.00 1.00
Fatal 1.67 0.82-3.39 1.79 1.29-2.48* 3.13 2.79-3.52+

* Reference group.
+Significant at p < 0.01 level.
Significant at p < 0.05 level.

DISCUSSION 1990-96 a d d s m o r e evidence for the g e n u i n e n e s s of the

Aviation safety for U.S. m a j o r airlines has i m p r o v e d
c o n t i n u o u s l y in the p a s t four decades; the fatal crash
rate decreased f r o m a b o u t 1.0 per 100,000 flight h o u r s in
1958 to 0.02 p e r 100,000 flight h o u r s in 1996 (16). Similar
success does not s e e m to h a v e b e e n achieved in other
aviation operations; statistics for the p a s t decade indi-
cate that fatal crash rates for c o m m u t e r s , air taxis, a n d
general aviation r e m a i n e d steady (16). The continuing
d o m i n a n t role of pilot error in general aviation crashes
is remarkable. In a s t u d y p u b l i s h e d in 1942 involving
300 general aviation crashes, H e r b o l s h e i m e r (9) re-
p o r t e d that pilot error w a s a cause in 85% of the crashes,
w h i c h is exactly the s a m e as f o u n d in the p r e s e n t study.
W h e r e a s the l o n g - t e r m historical trend of pilot error
in c o m m u t e r / a i r taxi crashes is less clear d u e to the lack
of data, the p r e v a l e n c e rate (38%) of pilot error in major
airline crashes r e p o r t e d in this s t u d y a p p e a r s to be
l o w e r t h a n that (45 to 67%) f o u n d in the 1960's a n d
1970's b a s e d on international s a m p l e s (15,21). This dis-
c r e p a n c y could be multi-faceted, p r o b a b l y reflecting the
effects of b o t h genuine a n d artificial factors. The finding
that the prevalence rate of pilot error in m a j o r airline
crashes decreased b y 21% b e t w e e n 1983-89 a n d
e n c o u r a g i n g t e m p o r a l trend. Responsible for the de-
cline of pilot error in major airline crashes could be use
of a d v a n c e d technologies (e.g., the g r o u n d p r o x i m i t y
w a r n i n g systems) a n d i m p l e m e n t a t i o n of intervention
p r o g r a m s (e.g., cockpit resource m a n a g e m e n t ) . Artifi-
cial factors, such as possible changes in r e p o r t i n g crite-
ria a n d investigation protocols o v e r t i m e a n d differ-
ences in s t u d y p o p u l a t i o n s across studies, m a y also
h a v e contributed to a certain extent to the o b s e r v e d
decline in pilot error in major airline crashes. G i v e n that
crash risk for m a j o r airlines is c o n t i n u i n g a d o w n w a r d
trend, the decrease in the p r e v a l e n c e of pilot error in
crashes of m a j o r airlines o v e r the p a s t four decades, if
c o n f i r m e d b y further research, s h o u l d be a reason for
o p t i m i s m , a n d the causes u n d e r l y i n g the t e m p o r a l
trend m e r i t rigorous investigation.
Pilot error has b e e n studied p r e d o m i n a n t l y b a s e d o n
the i n f o r m a t i o n processing m o d e l of h u m a n perfor-
m a n c e f r o m a psychological perspective. According to
this model, piloting is a d y n a m i c task involving three
basic elements: i n f o r m a t i o n gathering, decision m a k i n g ,
a n d action. U n d e r this theoretical f r a m e w o r k , research
is often c o n d u c t e d using flight simulators in controlled

56 Aviation, Space, and Environmental Medicine ~ Vol. 72, No. 1 ~ January 2001
CORRELATES OF PILOT ERROR--LI ET AL.
conditions aimed at localizing cognitive performance in
information processing resources, such as memory ca-
pacity, and attentional and reasoning abilities (18).
Studies from this approach are imperative for under-
standing the mechanisms of pilot error and for assess-
ing the effects of physiological and behavioral factors
(e.g., aging and alcohol use) on pilot performance
(13,14). Limitations of this approach are that it over-
emphasizes the importance of intrinsic factors in pilot
error and that its findings are not necessarily valid
when extrapolated to the real-world flight environment
(4).
Results of this study indicate that among factors re-
lated to the pilot, aircraft and crash circumstance, in-
strument meteorological condition is most predictive of
pilot error. This finding holds true for all three catego-
ries of aviation included in the study, and appears to be
consistent with previous reports (2,6,21). Based on an
examination of 400 fatal crashes of "public transport
aircraft," Shuckburgh (21) found that 30% of human
errors identified are incorrect operations in instrument
meteorological conditions. In a survey of German Air
Force pilots, Gerbert and Kemmler (6) found that the
most frequently reported extrinsic factors associated
with pilot error is restricted visibility due to poor
weather conditions. Baker and colleagues (2) compared
a group of pilots who had multiple crash records with
a group of pilots who each had only one crash record
during the same period, and concluded that exposure to
hazardous environmental conditions, such as adverse
weather and unpaved runways, is a major determinant
of crash frequency.
In addition to basic weather condition, crash location
also is associated with the likelihood of pilot error.
Crashes occurring at an airport are significantly more
likely than crashes away from an airport to involve
pilot error. The findings with weather and crash loca-
tion can easily be explained by the stress model, pro-
posed by Selye (19) and refined by Hannkonson (8),
Green (7), and Alkov et al. (1). According to the stress
model, reaction to stressors is a process of adaptation,
which occurs in three stages: the alarm reaction, the
resistance, and the exhaustion. Flight safety is the bal-
ance between performance ability and performance de-
mand. Green (7) outlined three types of stressors that
pilots confront, environmental, acute reactive, and life
events. These stressors affect flight safety b y either in-
creasing performance demand or reducing pilot perfor-
mance ability. A crash occurs at the exhaustion stage of
adaptation when performance demand exceeds perfor-
mance ability. Increased pilot error in instrument
weather conditions and during takeoff and landing is
likely due, to a large extent, to increased performance
demand.
The results of this study suggest that certain pilot
characteristics, particularly variables concerning flight
experience, may play a significant role in pilot error, but
in general, intrinsic factors appear to be less important
than environmental factors. The seemingly perplexing
finding that the protective effect of total flight time
against pilot error exists in general aviation crashes but
not in commuter/air taxi crashes and major airline
Aviation, Space, and Environmental Medicine ~ Vol. 72, No. 1 ~ January 2001
crashes is, in fact, plausible and supported by several
other studies (3,12,20). Characterizing the relation be-
tween flight experience and safety performance is a
nonlinear function: the safety benefit of flight experi-
ence is greater in the early, experience-building stage,
and diminishes as total flight time increases (12). Over-
all, general aviation pilots are far less experienced than
those flying commercial operations. The finding that the
beneficial effect of total flight time on pilot error is
limited only to general aviation can thus be explained
b y the relative inadequacy of experience in general
aviation pilots. It does not imply that flight experience
is unimportant to commercial pilots. Rather, it reflects
the nonlinear nature of the experience-performance re-
lation and the fact that all commercial pilots have al-
ready accumulated a large number of flight hours. That
the protective effect of advanced certificate rating
against pilot error remains significant in commuter/air
taxi crashes illustrates the persistence, albeit diminish-
ing, of the benefit of flight experience. It is worth noting
that neither pilot age nor gender is independently as-
sociated with pilot error in major airline crashes or
commuter/air taxi crashes. The only significant age
effect found in this study is a higher risk of pilot error
among pilots under age 20 who were involved in gen-
eral aviation crashes.
While the findings of this study are strengthened by
the large sample size, the inclusion of three major cat-
egories of aviation, and the multivariate statistical tech-
nique, there is a limitation with the research design that
warrants special attention. The analysis of pilot error
was conditional to the occurrence of a crash. The factors
associated with pilot error identified in this study are
correlates in essence, and thus do not necessarily rep-
resent risk factors. Aviation crashes are rare events.
Presumably, the overwhelming majority of pilot errors
and encounters with adverse weather and other haz-
ardous conditions do not result in an incident or a crash
and thus are unreported. Unless surveillance data on
pilot error and exposures to hazardous conditions, re-
gardless of the flight outcome, become available, risk
factors of pilot error cannot be adequately assessed.
Nevertheless, this study provides valuable data for
understanding the etiology of pilot error and for for-
mulating prevention strategies. It provides evidence
that pilot error is not merely a function of human na-
ture. Rather, it is determined by both endogenous and
exogenous variables in ways that are understandable
and predictable. Environmental stressors that increase
performance demand often play an important role in
the causation of pilot error and aviation mishaps (2,6).
To minimize pilot error, enhancing performance ability
through safety training may be beneficial but has its
limits. Developing and applying technologies that re-
duce pilot performance demand in dangerous situa-
tions appear to be more effective, and may have con-
tributed to the significant decline in the prevalence of
pilot error in major airline crashes over the past four
decades.
ACKNOWLEDGMENTS
Funding for this research was provided primarily by Grants
R01AG13642and R01AA09963from the National Institutes of Health
57
CORRELATES OF PILOT ERROR--LI ET AL.
and in part by Grant R49CCR302484 from the Centers for Disease
Control and Prevention. The authors thank Stanley Smith, Chief of the
Analysis and Data Division, National Transportation Safety Board,
for data assistance and helpful comments; and Ruth Nelson for sec-
retarial support.
REFERENCES
1. Alkov RA, Gaynor JA, Borowsky MS. Pilot error as a symptom of
inadequate stress coping. Aviat Space Environ Med 1985; 56:
244 -7.
2. Baker SP, Li G, Lamb MW, Warner M. Pilots involved in multiple
crashes: "accident proneness" revisited. Aviat Space Environ
Med 1995; 66:6-10.
3. Bennett CT, Schwirzke M. Analysis of accidents during instru-
ment approaches. Aviat Space Environ Med 1992; 63:253-61.
4. Billings CE, Gerke RJ, Wick RL, Jr. Comparisons of pilot perfor-
mance in simulated and actual flight. Aviat Space Environ Med
1975; 46:304-8.
5. Billings CE, Reynard WD. Human factors in aircraft incidents:
results of a 7-year study. Aviat Space Environ Med 1984;
55:960-5.
6. Gerbert K, Kemmler R. The causes of causes: determinants and
background variables of human factor incidents and accidents,
Ergonomics 1986; 29:1439-53.
7. Green RG. Stress and accidents. Aviat Space Environ Med 1985;
56:638-41.
8. Haakonson NH. Investigation of life change as a contributing
factor in aircraft accidents: a prospectus. Aviat Space Environ
Med 1980; 51:981-8.
58
View publication stats
9. Herbolsheimer AJ. A study of three hundred non-selected avia-
tion accidents. J Aviat Med 1942; 13:256-66.
10. Hosmer DW, Lemeshow S. Applied logistic regression. New
York, NY: Wiley & Sons, 1989.
11. Li G. Pilot-related factors in aircraft crashes: a review of epide-
miologic studies. Aviat Space Environ Med 1994; 65:944-52.
12. Li G, Baker SP. Prior crash and violation records of pilots in
commuter and air taxi crashes: a case-control study. Aviat
Space Environ Med 1994; 65:979-85.
13. Morrow D, Leirer V, Yesavage J. The influence of alcohol and
aging on radio communication during flight. Aviat Space En-
vrion Med 1990; 61:12-20.
14. Morrow D, Yesavage J, Leirer V, Tinklenberg J. Influence of aging
and practice on piloting tasks. Exp Aging Res 1993; 19:53-70.
15. Nagel DC. Human error in aviation operations. In: Wiener EL,
Nagel DC, eds. Human factors in aviation. San Diego, CA:
Academic Press, 1988:263-303.
16. National Safety Council. Accident facts, 1998 edition. Itasca, IL:
Author, 1998. p. 124.
17. National Transportation Safety Board. NTSB aviation coding
manual, http://www.ntsb.gov/aviation/codman_intro.htm.
18. Salthouse TA. Theoretical perspectives on cognitive aging. Hills-
dale, NJ: Erlbaum, 1991: 301-49.
19. Selye H. Stress without distress. Philadelphia, PA: Lippincott,
1974: 33-44.
20. Shuch HP. The influence of flight experience on midair collision
risk perception. Accid Anal Prev 1992; 24:655-60.
21. Shuckburgh JS. Accident statistics and the human-factor element.
Aviat Space Environ Med 1975; 46:76-9.
Aviation, Space, and Environmental Medicine ~ Vol. 72, No. 1 9 January 2001