Experience-Based Mitigation of Age-Related Performance Declines: Evidence From Air Traffic Control

Journal of Experimental Psychology: Applied
© 2009 American Psychological Association
2009, Vol. 15, No. 1, 12–24
1076-898X/09/$12.00
DOI: 10.1037/a0014947
EMBARGO: NOT FOR RELEASE UNTIL 12:01 AM EDT MARCH 16, 2009
Experience-Based Mitigation of Age-Related Performance Declines:
Evidence From Air Traffic Control
Ashley Nunes and Arthur F. Kramer
University of Illinois at Urbana–Champaign
Previous research has found age-related deficits in a variety of cognitive processes. However, some
studies have demonstrated age-related sparing on tasks where individuals have substantial experience,
often attained over many decades. Here, the authors examined whether decades of experience in a
fast-paced demanding profession, air traffic control (ATC), would enable older controllers to perform at
high levels of proficiency. The authors also investigated whether older controllers would show dimin-
ished age-related decrements on domain-relevant cognitive abilities. Both young and old controllers and
noncontrollers performed a battery of cognitive and ATC tasks. Results indicate that although high levels
of experience can reduce the magnitude of age-related decline on the component processes that underlie
complex task performance, this sparing is limited in scope. More important, however, the authors
observed experience-based sparing on simulated ATC tasks, with the sparing being most evident on the
more complex air traffic control tasks. These results suggest that given substantial experience, older
adults may be quite capable of performing at high levels of proficiency on fast-paced demanding
real-world tasks. The implications of these findings for global skilled labor shortages are discussed.
Keywords: cognitive aging, human performance, air traffic control, workforce shortage
Supplemental materials: http://dx.doi.org/10.1037/1037/a0014947.supp
A consistent observation in the study of aging and cognition is
offset or compensate for the impact of aging on complex skills or
decline in many perceptual and cognitive abilities across the adult
their component processes (Kramer & Willis, 2003; Morrow, in
life span. Such declines have been observed in both cross-sectional
press). These efforts have met with some success. For example,
and longitudinal studies across a variety of tasks, abilities, and
Charness (1981) found that the effectiveness of search for chess
processes, including measures of perception, working and episodic
moves is unrelated to age for expert players. In the domain of
memory, abstract reasoning, inhibitory processes, multitasking,
typing, older highly experienced typists perform as well as young
and spatial abilities (Craik & Salthouse, 2008; Kramer & Willis,
professional typists, and this high level of performance for the
2003; Salthouse, 2006).
older typists appears to be due to the more effective use of a
Whereas such studies have provided evidence to fuel notions
preview strategy to offset their slower motor processes (Bosman,
that older is not better, there is another body of literature that has
1993; Salthouse, 1984). Masunaga and Horn (2001) used the
suggested that a byproduct of age is experience, which can be
strategy game GO as a “platform” to investigate age and experience
useful for solving complex moral and social problems (Baltes &
effects. In their study, players of varying age and experience levels
Staudinger, 1993). Indeed, over the past several decades, research-
were required to perform a series of tasks deemed either relevant or
ers have examined whether high levels of experience serve to (a)
irrelevant to the game. They found evidence of age-related decline on
reduce age-related decline on basic perceptual, cognitive, or motor
irrelevant tasks and some moderation of age-related differences on the
abilities that presumably underlie complex skills or (b) aid in the
domain-relevant tasks.
development of domain general or specific strategies that can
Other studies have found equal benefit of expertise for older and
younger adults. Age declines in performing standard spatial ability
tests are comparable for samples varying in experience on tasks
Ashley Nunes and Arthur F. Kramer, Beckman Institute and Department
requiring these spatial abilities, either when experience is mea-
of Psychology, University of Illinois at Urbana–Champaign.
sured by self-rating (Salthouse et al., 1991) or when novices are
Ashley Nunes is now at CSSI, Inc., Washington, DC.
compared with professionals whose practice requires these abili-
This research was supported by Grants RO1 AG25667 and RO1
ties (Salthouse et al., 1990, for architects; Lindenberger, Kleigl, &
AG25032 from the National Institute on Aging. Special thanks to Andrea
Baltes, 1992, for graphic designers). Musical expertise does not
Barkauski and Shanquin Yin for their help with data collection, Sharon
reduce age differences on music recall (Meinz & Salthouse, 1998)
Yeakel and Henry Zaccak for technical assistance, and Greg Myles and
or other music tasks (Halpern, Bartlett, & Dowling, 1995). Knowl-
Kathy Fox for logistical assistance. We are grateful to William Horrey,
edge about the topic of a text does not reduce age differences in
Walter Boot, Jason McCarley, Frank Durso, and M. Kathryn Bleckley for
memory for the text (Hambrick & Engle, 2002; Miller, 2003;
their comments and suggestions regarding versions of this article.
Morrow, Leirer, & Altieri, 1992). Although these studies did not
Correspondence concerning this article should be addressed to Arthur F.
Kramer, Beckman Institute, University of Illinois, Urbana, IL 61801.
find evidence for mitigation, they do suggest that older adults
E-mail: akramer@s.psych.uiuc.edu
derive the same degree of expertise benefit as younger adults do.
12

---

EXPERIENCE-BASED MITIGATION
13
Several studies have also examined experience-based sparing in
be addressed by affording older controllers the opportunity to stay
demanding professions. For example, Tsang and Shaner (1998; see
on the job longer. More generally, it may be the case that evidence
also Tsang & Voss, 1996) investigated the ability of pilots and
of experience-based mitigation in a domain such as ATC could
nonpilots of varying ages to perform a series of single and dual
provide an indication of the ameliorative benefits that expertise
tasks. They found reduced age-related differences on some dual
holds as a means of offsetting the detrimental effects of advancing
tasks for older pilots but not for nonpilots. However, experience-
age among operators managing other complex sociotechnical sys-
based mitigation of age-related differences was not found for
tems (e.g., medicine, construction, industrialized operations).
single task performance. Such a pattern of age-related sparing is
not unexpected given the need to frequently time-share and switch
Present Research
between tasks during piloting. Morrow and colleagues have found
that older pilots perform similarly to younger pilots on a variety of
Our study employed older and younger professional air traffic
air traffic communication tasks, particularly when the messages
controllers and age-matched noncontrollers who collectively per-
are contextually relevant, when the pilots are permitted to read the
formed a battery of cognitive tasks and simulated ATC tasks that
messages at their own pace, and when permitted to take notes
varied in difficulty. The tasks in the cognitive task battery were
(Morrow, Leirer, Altiere, & Fitzsimmons, 1994; Morrow, Wick-
selected to provide measures of both ATC domain-relevant abili-
ens, Rantanen, Chang, & Marcus, 2008, see also Taylor, Kennedy,
ties and less relevant measures of different aspects of cognition
Noda, & Yesavage, 2007). Thus, older pilots capitalize on their
(see Wickens, Mavor, & McGee, 1997). Domain-relevant abilities
wealth of domain relevant knowledge to compensate for age-
included inhibitory control, task-switching ability, visuospatial
related deficits in working memory.
ability, working memory, and breadth of attention. Less relevant
Although such results are encouraging, these studies have his-
abilities included processing speed and inductive reasoning. In
torically investigated and subsequently found evidence of
addition to the cognitive battery, ATC tasks were administered
experience-based sparing of abilities that represent only a subset of
to assess problem-solving ability under different levels of time
those needed to succeed in the real world. However, operators
pressure in a variety of different ATC scenarios. We predicted that,
managing complex sociotechnical systems must exercise a variety
although high levels of experience would do little to ameliorate the
of skills and abilities to maintain optimal performance levels. For
detrimental effects of advancing age on cognitive abilities not
example, in aviation, pilots must exhibit sound flight control and
directly related to ATC, experience would offset potential age-
navigation ability, in addition to being able to communicate with
related decrements on those cognitive abilities that are more di-
other pilots and controllers. Hence, observations of experience-
rectly related to ATC as well as on the simulated ATC tasks
based sparing in instances where only a small subset of desired
themselves. More specifically, given the nature of the ATC task,
abilities (e.g., communication) are held up to scrutiny (e.g.,
we expected that the abilities of inhibitory or interference control,
Morrow et al., 1994, 2003, 2004, 2006) limit the generalizability of
task switching, visuospatial processing, working memory, and
such findings.
breadth of attention may display some experience-related sparing
The need to extrapolate such findings to the real world is
of aging decrements. On the other hand, we did not expect pro-
pressing. Consider the profession of air traffic control (ATC),
cessing speed or inductive reasoning to show experience-related
which is experiencing severe staffing shortages on a global scale.
sparing given that these abilities are less germane to the task of
For example, near misses between aircraft, caused as a result of
ATC.
staffing shortages across Russia, Australia, South Africa, and the
In addition to enabling us to examine whether ATC experience
United States, are getting close to becoming midair disasters, and
influences age-related differences in basic cognitive abilities, the
in some cases airlines are being forced to choose between cancel-
inclusion of the cognitive battery enabled us to address, in part, the
ling, delaying, or diverting flights, or having loaded jetliners flying
“confound of nature” (Morrow, in press) that is inherent in most
through uncontrolled airspace (Baguley, 2008).
studies of Experience
Age interactions. That is, because expe-
Such shortages have been exacerbated by the presence of
rience tends to increase with age, there is a natural confound in
decades-old mandatory retirement policies that have been estab-
most cross-sectional studies of age and experience (see Hoyer &
lished over concerns of age-related performance declines. In the
Ingolfsdottir, 2003, for a rare exception). The cognitive battery
United States, for example, controllers must retire by the age of 56,
enabled us to ask whether the older controllers would outperform
with numerous studies demonstrating age-related performance de-
the older noncontrollers on a wide variety of cognitive tasks or, as
clines being cited as justification for enforcing such policies
predicted, on only those tasks most closely related to the skills
(e.g.,Heil, 1999a, 1999b; Mathews & Cobb, 1974; Trites & Cobb,
necessary for efficient ATC.
1962; VanDeventer & Baxter, 1984). However, these studies have
Given that air traffic controllers are generally retired by 56 years
been characterized by (a) reliance on subjective ratings (which can
of age in the United States, we conducted our study in Canada
be biased against older adults; see Cobb, 1968) and (b) employ-
where controllers can work until 65 years of age.
ment of tasks that often do not reflect operational constraints or
afford skilled operators the ability to use strategies acquired over
Method
the years. Thus, it is perhaps unsurprising that previous investiga-
tions have found limited evidence of success among older adults in
Participants
ATC.
Evidence of experience-based sparing in such a complex do-
Thirty-six licensed ATC controllers and 36 noncontrollers
main has far-reaching implications. Within the ATC context alone,
served as participants, with 18 older and 18 younger adults per
concerns over projected severe controller staffing shortages could
group. Older controllers (all men) were between the ages of 53 and

---

14
NUNES AND KRAMER
64 years (M
57.2), had an operational experience range of
Visual spatial processing, which refers to an individual’s ability
between 25 and 38 years (M
34.06), and averaged 1.08 years of
to process objects and events in space, was measured using the
postsecondary education. Younger controllers (15 men and 3
mental rotation task (Cooper & Shepard, 1973). In the present task,
women) were between the ages of 20 and 27 years (M
24.4), had
a trial consisted of the participant determining whether two shapes
an experience range of between 0.25 and 4 years (M
1.64), and
presented on the screen were the same or mirror images of one
averaged 1.65 years of postsecondary education. Controllers were
another. The first shape always appeared upright, whereas the
recruited for participation from five different ATC centers across
second shape appeared at a variety of orientations between 0 and
Canada. Older noncontrollers (all men) ranged in age from 52 to
360 degrees. Following a brief practice, which consisted of 16
64 years (M
57.5) and had an average of 3.46 years of postsec-
trials, the experiment block began, during which 96 trials were
ondary education. Younger noncontrollers (15 men and 3 women)
administered. The manipulation of shape type (namely, same and
were between the ages of 20 and 27 years (M
23.3) and had an
different shapes) and orientation (which varied in 45 degree incre-
average of 3.82 years of postsecondary education.
ments between 0 and 335 degrees) was orthogonal, with six trials
in each of the 16 different conditions. The trial order was random-
Apparatus and Task
ized to avoid possible order effects, and response time served as
the dependent variable.
Participants completed a series of cognitive and ATC tasks.
Working memory, which entails storage of information, opera-
Within each controller and age group, the order in which the tests
tions on the stored information, and retrieval of information, was
from the cognitive and ATC battery were administered was ran-
measured using the operation span test (Turner & Engle, 1989).
domized to avoid order effects. The test battery took approxi-
Participants were asked to remember words while solving a series
mately 6 hr to administer.
of math problems aloud. Following verbalization of a math prob-
lem (e.g., IS (5
1
6?) DOG), the participant had to evaluate
Cognitive Test Battery
whether or not the problem was true and then recite the presented
word (in this case dog), following which they moved onto the next
A number of neuropsychological tests were administered to
math problem. In addition to 3 practice problems presented for
assess basic cognitive abilities. These abilities and the tests used to
familiarization purposes, 15 trial problems were administered,
measure them are discussed below.
each of which varied in length (between two and six) in terms of
Inhibitory control, which refers to the ability of the individual to
the number of words that had to be recalled after a given sequence.
suppress responses to objects in the environment that would nor-
There were three trials of each sequence size. The experimenter set
mally draw attention, was measured using the flanker compatibil-
the pace of the task, and the performance metric used was the
ity task (Lavie & Cox, 1997). Participants complete an easy or
number of words that could be accurately recalled in a given
hard search task with a compatible or incompatible distractor
sequence (with an all-or-none scoring methodology used).
flanking the search display. The compatibility effect from the
Breadth of visual attention, which refers to how broadly partic-
distractor item is an index of “spare” attentional resources not used
ipants can distribute their visual attention and localize targets in
by the search task. In the present task, participants had to deter-
the periphery (Ball, Beard, Roenker, Miller, & Griggs, 1988), was
mine whether there was a square or diamond among a series of
measured using a variant of the useful field of view task. The
circles, which contained a varying number of shapes (set size
participant was asked to detect and localize a briefly presented
manipulation). In addition, a distractor (either a square or a dia-
target among a distractor array. A trial started with the target
mond) sometimes appeared outside the array, with participants
appearing in one of several possible array locations for 40 ms,
being told to ignore it and focus on the instructed task. In addition
following by a mask. The participant then used a mouse to click on
to the practice block, which consisted of 20 trials, the task con-
the region where the target had appeared. The independent variable
sisted of 100 trials, with an even split of small versus large set size
was the eccentricities of the target from the fixation point, with 10,
conditions (50 per condition). Within each half, there was an equal
20, and 30 degree angles being tested respectively. The experi-
distribution of compatible and incompatible trials (25 per condi-
mental session consisted of three blocks of 39 trials, each of which
tion). A trial began with a fixation cross presented for 500 ms,
consisted of 13 trials for each of the three different eccentricities
followed by the presentation of the stimulus array. Trial adminis-
tested, and response accuracy served as the performance metric.
tration was self-paced and response time served as the primary
Processing speed, defined as the speed at which mental opera-
dependent variable.
tions can be performed, was measured using the dot comparison
Task switching, which refers to the ability to effectively switch
task (Salthouse, 1991). Participants compared two nine-dot matrix
processing priorities between two tasks, was measured using the
patterns to determine whether they were identical as quickly as
Trail Making Test (TMT; Army Individual Test battery, 1944).
possible, with responses entered via a keyboard. The task was
The TMT, and particularly Form B, is considered to be a measure
self-paced, with practice consisting of 10 trials; the test comprised
of mental flexibility and task-switching ability. The first portion of
three blocks consisting of 36 trials in each block. The patterns that
the test (Form A) requires the participant to connect circles that
were the same or different were equally distributed across and
contain numbers (1, 2, 3 etc.) in sequence as quickly as possible.
within blocks, and the time taken to provide a correct response
The second portion of the test (Form B) is more complex as it
served as the primary dependent variable.
requires the participant to alternate between number and letter
Inductive reasoning, which refers to the ability of an individual
(1-A, 2-B, etc.) when connecting the circles. The time taken to
to use limited previous experience and observations to make a
correctly complete Form B served as our measure of task switch-
generalized inference, was measured using the letter series test
ing.
(Ekstrom, French, & Harmon, 1976). Participants determined the

---

EXPERIENCE-BASED MITIGATION
15
next two letters in a sequence having already viewed the first six
screen depicted a sector of airspace 60
60 miles; all aircraft had
letters in the sequence. For example, the sequence A B A C A D
a data tag that depicted the aircraft’s altitude and speed, and the
would be followed by the sequence A E. Participants had to choose
heading of the aircraft could be inferred from its vector line (which
the correct answer from one of four possible options. The test
depicts the projected aircraft path). The task, which was designed
consisted of two parts: the practice portion with 5 practice prob-
to mimic traffic flows at an approach control facility, required a
lems and the experimental portion with 25 problems that got
high degree of interaction (manual input) with the computer to
progressively more difficult. Accuracy of the responses served as
ensure that sector rules were followed. Aircraft approaching from
the performance measure.
the west always entered the sector at FL 350 and had to descend to
FL 330 before a specified point. Conversely, aircraft approaching
from the east entered at FL 320 and had to descend to FL 310
ATC Test Battery
before exiting the sector. For a typical handling sequence, the
participant had to first accept the aircraft into the sector, issue the
Participants also performed a series of ATC tasks. The selection
altitude and heading commands when appropriate to ensure min-
and design of the ATC tasks were based, first, on information
imal deviation from the route, and finally hand off the aircraft to
acquired during extensive discussions with subject matter experts,
the next sector. These commands were issued to the aircraft via the
who had an aggregated experience of well over 120 years; second,
flight control interface that was available to the participant when
a thorough literature review of task analysis associated with the
needed by pressing the right mouse button. The main independent
ATC profession was conducted. The resulting set of ATC tasks
variable that was manipulated was airspace load. Under low load,
included conflict detection, conflict resolution, vectoring, and air-
aircraft entered the sector at a rate of one aircraft every 90 s. Under
space management. On each task, practice trials were administered
high load, the time interval was reduced to once every 60 s.
to minimize observable learning effects during the data collection
Aircraft entry at the waypoints was alternated so that if an aircraft
phase of testing.
appeared at the eastern entry fix at the onset of the simulation, the
Conflict detection task.
In this basic ATC task, participants
next aircraft would appear at the western fix. The primary perfor-
made perceptual judgments as to whether two aircraft on a given
mance metric was the aircraft-handling capacity, defined as the
trial would conflict with one another at some future point. A trial
number of times aircraft successfully passed through their assigned
began with two aircraft converging toward one another at the same
waypoint based on issuance of correct control instructions by the
altitude on the radar screen. The position of each aircraft was
participant. Performance was assessed over the course of the
updated once every 6 s (which is the approximate update rate for
“shift,” which lasted approximately 45 min.
current radar systems), and geometries and speeds at which the
During the initial briefing for the vectoring task, participants
aircraft converged varied from one trial to the next (Figure 1a).
became familiar with the interface and simulation platform and
Task difficulty was manipulated by varying the time to the closest
performed the task under varying aircraft entry rates. Practice
point of approach, with increases in time making perceptual judg-
lasted approximately 20 min. During the experimental phase,
ments more difficult given the larger distances generally involved.
participants handled 14 aircraft under low-load and 14 aircraft
Trial type (conflict present trials vs. conflict absent trials) was also
under high-load conditions (with each aircraft serving as a trial).
manipulated, and the time taken to arrive at a correct decision
Given that the aircraft entry rate was related to airspace load, under
served as the dependent variable. The manipulation of conflict
low load, the scenario lasted approximate 30 min, and the high-
presence and time to the closest point of approach was orthogonal,
load phase lasted approximately 12 min. The order in which
with 20 trials in each condition for a total of 120 trials. Trial order
airspace load was experienced by participants was always fixed,
was randomized across participants to avoid order effects, and the
with low load administered first, followed by high load.
task took approximately 40 min to complete.
Airspace management task.
A more complicated variant of the
Conflict resolution task.
A more complicated variant of the
vectoring task and the most complex task in the ATC battery,
conflict detection task was the conflict resolution task in which
the airspace management task required participants to manage the
participants’ problem-solving capability was measured by having
flow of traffic along different airways through a specified sector of
them resolve a series of complex ATC problems. More specifi-
high-altitude airspace (Figure 4a), which was approximately
cally, conflicts between aircraft pairs on a collision course had to
150
150 nautical miles. To do so, aircraft first had to be
be resolved by issuing altitude guidance instructions via a control
accepted into the sector, and their flight paths had to be checked to
interface (Figure 2a). The time to conflict, defined as the time
ascertain which route they were required to traverse. Following
taken by the aircraft to reach the conflict point, was manipulated;
determination of the route, the participant had to ascertain the
the time taken to correctly resolve the conflict served as the
altitude restrictions associated with that route, coupled with deter-
primary dependent variable. There were a total of 30 trials in this
minations of heading commands required to keep the aircraft along
phase, with 10 trials for each of the three time-to-conflict condi-
the route. This information could be obtained by checking visual
tions. Once again, the order in which the trials were administered
reference markers and each aircraft’s data tag (which depicted the
was randomized across participants, and the task took approxi-
call sign and altitude and speed assignments) on the radar screen.
mately 45 min to complete.
The participant was solely responsible for issuing all the altitude,
Vectoring task.
The first of two full dynamic simulation tasks,
heading, and speed assignments (via the flight control interface)
the vectoring task required participants to navigate aircraft ap-
necessary to get the aircraft safely from its entry point to its exit
proaching an entry point, ensuring they stayed within a specified
point. The final task included handing off the aircraft to the next
boundary of airspace and reached their final waypoint without
sector controller when it approached the final waypoint along its
being in conflict with any other aircraft (Figure 3a). The radar
route. Task difficulty was manipulated by varying the entry rate of

---

16
NUNES AND KRAMER
Figure 1.
(a) The conflict detect task: Participants are asked to render simple conflict or nonconflict judgments
regarding aircraft pairs converging toward one another. Of interest is the amount of time taken to safely and
correctly render a decision. (b) On this basic task, there is no evidence to support the experience-based sparing
hypothesis.
aircraft into the sector, resulting in low and high airspace load.
Performance measures included the number of commands issued
Under low load, a total of 10 aircraft entered the sector in 30 min;
to manage the airspace, the number of operational errors that
under high load, 20 aircraft entered the sector. As was the case in
occurred, and aircraft-handling capacity during the “shift,” which
the vectoring experiment, low load always preceded high load.
lasted approximately 60 min.

---

EXPERIENCE-BASED MITIGATION
17
Figure 2.
(a) The conflict resolution task: Participants are asked to resolve potential conflicts between aircraft
pairs by issuing guidance control instructions. Given the limits of human perceptual ability, the task becomes
increasing difficult as participant responses are sought farther from the potential conflict point. Of interest is the
amount of time taken to safely resolve the conflict. (b) We observed that the magnitude of age-related
performance decrement was higher among noncontrollers than that among controllers, providing evidence for the
experience-related sparing hypothesis.

---

18
NUNES AND KRAMER
Figure 3.
(a) The vectoring task in which participants must sequence aircraft along two air corridors around
an airport. The task becomes increasingly challenging as airspace load increases. (b) A significant interaction is
observed between experience and age, with no difference in performance between more and less experienced
controllers but a substantial difference in aircraft-handling performance between young and older noncontrollers.
The task familiarization for the airspace management task took
ATC tasks (such as voice communication, issuance of weather
approximately 15 min, coupled with another 30-min practice ses-
advisories, and physical receipt and handoff of flight strips) whose
sion. Following this, the experimental phase began, with partici-
inclusion may have improved the overall ATC experience.
pants managing the traffic flow for 60 min (30 min under low load
However, there were two reasons why such characteristics were
and 30 min under high load), after which the simulation termi-
omitted. First, we sought to achieve a balance between placing
nated.
participants in an environment that was realistic enough so that the
ATC task battery realism.
The ATC battery was developed in
mitigating effects of experience could be observed among control-
consultation with our subject matter experts to place participants in
lers (Craik & Jennings, 1992; Kirlik, 1995) but did not impede the
as realistic an ATC environment as possible given logistical con-
conduct of the data collection process given logistical constraints.
straints. Each task in the battery had participants rely on their
For example, given that data were collected at a number of
visuospatial, problem-solving, and task management abilities to
facilities, access to the same group of pilots for experimental
attain optimal performance levels, requirements that are not unlike
control purposes would have been difficult. Second and more
those demanded in the current ATC environment. We acknowl-
important, given that the inexperienced controller group had no
edge the absence of additional characteristics in our more complex
prior ATC knowledge, it was important to simplify the ATC task

---

EXPERIENCE-BASED MITIGATION
19
Figure 4.
(a) The airspace management task: Participants are asked to manage the flow of aircraft through a
region of airspace under increasing traffic load. (b) Results show that older noncontrollers issue fewer guidance
control instructions than younger controllers. (c) Closer analysis of the operational error rates also reveals
equivalent performance between older and younger controllers, suggesting that compensatory strategies are used
by the former group to maintain optimal performance levels. (d) This is noteworthy because the number of
aircraft handled by older and younger controllers is nearly equal, this given that an aircraft passes through an
average of four waypoints to reach its destination, which means that younger controllers handled an extra 1.5
aircraft in comparison to older controllers.
sufficiently so that it could be performed by this group while
maintain high levels of performance on a variety of increasingly
ensuring that benefits of prior controller training would also be
more complex and difficult ATC tasks. Means and standard
evident across the range of tasks.
errors for different cognitive battery task conditions are pre-
sented in Tables 1 through 4. Because of space limitations,
Procedure
analysis of variance tables (Tables 5–13) have been included as
an online supplement.
All participants were first prescreened to ensure that they were
in good health, following which they were assigned to their re-
spective age and experience groups. Half of the participants in
Influence of Experience on Cognitive Abilities
each of the four Age
Experience groups received the ATC
Data from the cognitive test battery (Tables 1– 4) were analyzed
battery first, and the remaining half received the cognitive battery
with mixed-mode analyses of variance (ANOVAs).1 In this article,
first. Within each group, the order in which the tests were admin-
we focus on the Age
Experience interactions because these
istered was randomized to avoid order effects. The ATC and
effects are most relevant to our hypotheses. The analyses of the
cognitive test batteries were administered on separate days to
tasks in the cognitive battery revealed age-related declines for both
avoid fatigue. Because of scheduling constraints, time of day was
controllers and noncontrollers on all of the tasks, with evidence of
not accounted for during testing. Finally, given that the entire
experience-based sparing on only two (inhibitory control and
experiment was expected to take approximately 8 hr, regular
visuospatial) of the seven abilities investigated. On these two
breaks were offered to participants.
abilities, a significant interaction was observed between experi-
ence and age, with a smaller age-related difference in cognition
Results and Discussion
In this section, we begin with the analyses of the tasks from the
1 We did not perform a MANOVA with the cognitive battery tasks as we
cognitive battery to determine whether, and to what extent, high
did with the ATC battery tasks. Given the number of factors in the tasks
levels of experience on air traffic control served to reduce age-
employed in the cognitive battery, it was impossible to simplify the
related decline on tests of different perceptual and cognitive abil-
variables such that they would make sense in a MANOVA. However, this
ities. We then examine whether, in the face of age-related de-
is much less of a concern for the cognitive battery given that few of the
clines on basic cognitive abilities, older controllers were able to
tasks should have significant Age
Experience interactions.

---

20
NUNES AND KRAMER
Table 1
Mean Response Times (ms) for the Inhibitory Functioning (Flanker) Task
Flanker
Small set
Large set
Group
Compatible
Incompatible
Compatible
Incompatible
Young controllers
630.67 (17.01)
642.61 (17.42)
965.81 (36.01)
966.19 (31.17)
Old controllers
835.58 (34.55)
837.22 (34.17)
1424.03 (56.99)
1454.58 (62.05)
Young noncontrollers
670.39 (38.89)
688.97 (45.92)
1039.33 (57.25)
1042.17 (55.16)
Old Noncontrollers
821.08 (31.55)
924.53 (66.18)
1501.31 (56.24)
1483.56 (63.29)
Note.
Standard errors appear within parentheses.
between the young and older controllers than between the young
traffic controller, the ATC task-switching task becomes unitized
and old noncontrollers. More specifically, we found that on the
and highly integrated (Kramer, Hahn, & Gopher, 1999; Kramer,
flanker task (see Tables 1 and 5), experience moderated the in-
Wickens, & Donchin, 1985; Logan, 2005), therefore, reducing the
compatibility effect among older participants when the set size was
necessity to switch attention among different subcomponents of
small but not when the set size was large, F(1, 68)
6.26,
2
the ATC task. Similarly, in examining working memory, it may
.33, p
.01. Larger compatibility effects are often reported with
well be the case that experienced controllers rely less on internal
smaller set sizes (Forster & Lavie, 2008). Therefore, the smaller
capacity limited storage and more on external information (e.g.,
set size compatibility effects should be more sensitive to age and
flight progress strips, radar) to maintain the “picture” of the overall
experience than larger set size conditions. Similarly, we also
traffic situation. With regard to the breadth of attention, it is
observed that on the mental rotation task (see Tables 2 and 7),
possible that controllers sequentially focus attention on specific
experience reduced the age difference in response time to mentally
aircraft rather than processing information concerning multiple
rotate geometric shapes for controllers compared with noncon-
aircraft in parallel (as would be consistent with the construct of
trollers, F(1, 68)
3.32, 2
.03, p
.01. The accuracies for the
breadth of attention). In any event, the results for the cognitive
flanker and mental rotation tasks were high and consistent with the
battery suggest, consistent with the extant literature (Morrow, in
response time data.
press; Salthouse, 1984), that experience gained in specific profes-
Consistent with our hypotheses, both inhibitory control and
sions or other pursuits, often over the span of several decades,
visuospatial processing showed experience-based sparing for the
tends to have rather limited effects on only the most domain-
older controllers. Also consistent with our hypotheses, based on an
relevant cognitive abilities. That is, high levels of experience do
analysis of the ATC task, no experience-based sparing was ob-
not broadly provide immunity against the detrimental effects of
tained for measures of processing speed, F(1, 68)
0.01,
2
advancing age on basic cognitive functioning.
.00, p
.96, or inductive reasoning, F(1, 68)
0.03,
2
.00,
p
.87. On the other hand, three other abilities that would appear
Experience and Performance on Simulated ATC Tasks
to be relevant to ATC—namely, task switching, F(1, 68)
0.03,
2
.01, p
.56, working memory, F(1, 68)
3.09,
2
.04,
A multivariate analysis of variance (MANOVA) was conducted
p
.08, and breadth of attention, F(1, 68)
1.06, 2
.02, p
on measures from each of the four ATC tasks to determine whether
.31— did not display evidence of experience-based sparing. It is
experienced-based age-related sparing would be observed. Signif-
certainly conceivable that after many years of serving as an air
icant main effects were obtained for age, F(1, 65)
28.6,
2
Table 2
Mean Response Times (ms) for the Visuospatial Task
Mental rotation (degrees)
Rotation angel/group
360
45
90
135
180
225
270
315
Mirror shape
Young controllers
1057.1 (62.6)
1423.2 (79.4)
1624 (64.3)
1630.5 (91.2)
1783.9 (89.7)
1775.6 (79.8)
1636.4 (81.3)
1383.5 (79.4)
Old controllers
2128.5 (283.1) 2329.7 (279.1) 2432.7 (259.3) 2418.7 (183.7) 2889.4 (281.4) 2549.6 (266.7) 2648.4 (363.3) 2276.1 (261.2)
Young noncontrollers 1076.4 (70.3)
1335.8 (106.9) 1584.6 (116.7) 1612.3 (127.1) 1759.2 (111.2) 1709.2 (128.3) 1688/1 (108.8) 1340.7 (89.2)
Old noncontrollers
2207.4 (194.5) 2467.2 (279.7) 2712.3 (190.9) 2745.6 (226.6) 2821.4 (164.3) 2700.6 (154.7) 2544.8 (145.7) 2174.1 (188.7)
Same shape
Young controllers
853.6 (27.6)
1108.1 (45.9)
1317.3 (37.2)
1415.1 (66.6)
1790.9 (103.0) 1749.6 (82.4)
1507.1 (64.9)
1155.9 (33.9)
Old controllers
1532.2 (179.2) 1682.4 (135.5) 2081.6 (140.1) 2333.8 (286.2) 2903.5 (433.4) 2643.4 (243.1) 2326.8 (226.3) 1874.5 (143.5)
Young noncontrollers
883.5 (43.8)
1079.2 (41.8)
1606.4 (147.8) 1482.6 (81.3)
1742.0 (125.8) 1596.9 (90.5)
1448.5 (83.9)
1086.6 (44.8)
Old noncontrollers
1627.2 (116.7) 1901.3 (188.9) 2256.3 (163.7) 2358.8 (215.6) 2812.4 (220.3) 3028.9 (245.4) 2581.6 (206.8) 2268.4 (167.5)
Note.
Standard errors appear within parentheses.

---

EXPERIENCE-BASED MITIGATION
21
Table 3
and younger controllers were similar, the number of control inputs
Mean Accuracies for the Breadth of Attention Task
on the radar screen by older controllers was lower than that of
younger controllers (Figure 4b). This is noteworthy because anal-
Useful field of view (eccentricity, degrees)
ysis of our qualitative measures, such as operational error rate
(Figure 4c) and aircraft handling capacity (Figure 4d), revealed
Group
10
20
30
few differences in performance between older and younger con-
Young controllers
0.92 (.05)
0.87 (.05)
0.79 (.06)
trollers. This suggests that younger controllers may be issuing
Old controllers
0.60 (.06)
0.53 (.05)
0.45 (.05)
more commands than are necessary compared with the older
Young noncontrollers
0.93 (.03)
0.90 (.05)
0.83 (.04)
controllers, who issue fewer commands while achieving the same
Old noncontrollers
0.51 (.06)
0.45 (.05)
0.39 (.04)
results. Such results are noteworthy because they suggest that older
Note.
Standard errors appear within parentheses.
controllers can compensate for reductions in cognitive ability by
acting in a more measured fashion to achieve performance that
rivals that of their younger counterparts who exhibited better
.64, p
.01, and experience, F(1, 6)
22.4,
2
.58, p
.01,
cognitive ability.
as well as for the Age
Experience interaction, F(1, 68)
7.6,
We note that similar effects were absent for the noncontroller
2
.32, p
.01. Therefore, this analysis was followed by
group, where significant age-related performance impairments
ANOVAs for each of the four ATC tasks.
were observed. These data suggest that in the face of age-related
Older controllers performed quite well on the ATC task, with
decline across many basic cognitive abilities, seasoned older pro-
results from the ATC battery being more consistent with the
fessionals may use alternative strategies (Backman & Dixon,
experience-based sparing hypothesis. That is, experience-based
1992) that employ domain-relevant knowledge to efficiently man-
sparing increased with increasing complexity of the ATC tasks.
age complex sociotechnical systems. We also note a trade-off
The results (Table 10, Figure 1b) from the relatively simple con-
between error rates and aircraft-handling capacity in the noncon-
flict detection task yielded no evidence of experience-based spar-
troller group: More specifically, younger noncontrollers appear to
ing, F(1, 68)
0.01, 2
.00, p
.92. However, the age-related
have handled a higher aircraft volume at the expense of a greater
performance decrement on the conflict resolution response task
error rate.
(Table 11, Figure 2b) was significantly smaller among controllers
In collectively viewing the results from the ATC task battery, it
compared with noncontrollers, F(1, 68)
4.66, 2
.03, p
.03,
is important to note that evidence of experience-based mitigation
suggesting that in comparison to low-level perceptual tasks, the
was observed on three of the four ATC tasks administered (the
benefits of experience-based mitigation may be more evident on
only exception being the conflict detection task). In addition,
more difficult problem-solving tasks.
inspection of the conflict error data on the most complex task
The vectoring task represented a more dynamic environment
(Figure 4c), the airspace management task, revealed little differ-
where decisions had to be made and actions enacted under greater
ence in performance between younger and older controllers. This
time pressure. Here we observed (Table 12, Figure 3b) that, once
effect is noteworthy given longstanding arguments that have pos-
again, high levels of experience ameliorated the effects of age-
tulated that performance lapses are more likely to be exhibited by
related performance declines, F(1, 68)
27.01, 2
.14, p
.01.
older controllers, this given their declining cognitive abilities (al-
More specifically, we found no difference in aircraft-handling
though we note that recent analysis of operational data by Broach
capacity for the young and older controllers. However, this was not
and Schroeder, 2006, has also revealed little evidence of age-
the case for the younger and older noncontrollers. Furthermore,
related differences in operational error rates among controllers).
age-related performance differences increased with increased traf-
Indeed, many of these arguments have been used as the basis for
fic for the noncontrollers but not for the controllers.
establishing and maintaining current mandatory retirement prac-
The airspace management task, which represented the most
tices in the ATC domain. Our results, however, suggest that older
complex and realistic ATC environment, also supported the
controllers are able to reduce their susceptibility to the detrimental
experience-based sparing hypothesis. Our analysis (see Table 13)
effects of age by relying on domain-specific experience. More-
showed that whereas a frequency analysis of control inputs (a
over, in examining performance on the conflict detection, conflict
quantitative metric) revealed that both the type and ratio of com-
resolution, vectoring, and the airspace management task, it appears
mand instructions (altitude, heading, and speed) issued by older
that the experience benefit was manifested in a clearer fashion (as
Table 4
Summary of Means for the Working Memory, Processing Speed, Inductive Reasoning, and Task-Switching Tasks
Working memory:
Task switching
Operation span
Processing speed:
Inductive reasoning:
Group
(no. words)
Dot comparison (ms)
Letter series (%)
Trail making A (s)
Trail making B (s)
Young controllers
16.44 (1.36)
1645.53 (56.83)
0.87 (0.02)
21.92 (1.36)
25.50 (1.95)
Old controllers
13.94 (1.87)
2319.08 (64.17)
0.71 (0.04)
30.34 (1.77)
35.28 (1.87)
Young noncontrollers
23.88 (5.74)
1678.07 (86.23)
0.87 (0.02)
42.08 (1.45)
47.37 (2.68)
Old noncontrollers
14.31 (3.37)
2379.94 (107.32)
0.71 (0.03)
62.84 (4.88)
76.42 (5.17)
Note.
Standard errors appear within parentheses.

---