Prepared for the Office of the Chief Scientific and Technical Advisor for Human Factors Federal Aviation Administration
by Battelle Memorial Institute - JIL Information Systems, January 1998
This document provides a brief review of the scientific research relating to issues of pilot fatigue arising from crew scheduling practices. A massive amount of research has been conducted on such issues as the environmental conditions that contribute to the occurrence of fatigue, acute and chronic sleep debt and their effects on performance, and the influence of the circadian cycle on alertness. This paper attempts to identify major trends in this literature that might be of value in addressing scheduling regulatory issues.
The paper is organized into seven sections. The first section, "What is Fatigue," attempts to provide a functional definition of fatigue that serves to define the scope of issues that need to be considered, including variables that contribute to the occurrence of fatigue and methodologies for assessing the impact of fatigue on human functioning.
Section two, "Indications and Effects of Fatigue," briefly reviews the human performance and physiological indicators of fatigue. The intent is to identify possible decrements in performance that could have a safety impact. This section also briefly addresses the complexities involved in measuring fatigue levels. As this section explains, fatigue is a complex concept that does not always produce expected measurable decrements in performance.
Section three, "Fatigue and the Aviation Environment," addresses the issue of fatigue within the aviation environment. Before changes are made to existing regulations, the question of whether there is a problem that needs to be resolved should be addressed. Available research on the extent of fatigue within the aviation environment is reviewed. In addition, factors that complicate the assessment of the extent of the fatigue problem in an operational environment are also described.
A pilot’s level of alertness at any time depends upon a complex interaction between a number of variables. Four variables, in particular, need to be considered: time on task, time since awake, any existing sleep debt, and the pilot’s own circadian cycle. Section four, "Standard Duty Period," describes the research trends pertaining to time on task and time since awake while section five, "Standard Sleep Requirements," addresses acute and chronic sleep debt, including recommendations for sleep debt recovery. Section six, "The Circadian Cycle and Fatigue," which looks at the research on circadian cycles and their implications for back-of-the-clock and transmeridian flying. Finally, section seven, "Augmented Crews," looks at the limited data on the use of augmented crews to extend duty periods.
What Is Fatigue
The objective of the regulations proposed in the NPRM is to identify scheduling constraints that will minimize the impact of pilot fatigue that arises from duty time and sleep debt due to crew schedules. The term, "fatigue," has yet to be defined in a concrete fashion (Maher & McPhee, 1994); Mendelson, Richardson & Roth, 1996). Fatigue, as addressed in the human performance literature, refers to "deterioration in human performance, arising as a consequence of several potential factors, including sleepiness" (p. 2). Sleepiness, in contrast, has a more precise definition: "Sleepiness, according to an emerging consensus among sleep researchers and clinicians, is a basic physiological state (like) hunger or thirst. Deprivation or restriction of sleep increases sleepiness and as hunger or thirst is reversible by eating or drinking, respectively, sleep reverses sleepiness" (Roth et al., 1989, cited by Mendelson, Richardson & Roth, 1996, p. 2).
In keeping with current thinking on the concept of fatigue, Maher and McPhee’s approach is used here:
"Fatigue" must continue to have the status of a hypothetical construct, an entity whose existence and dimensions are inferred from antecedent and consequent events or variables" (p. 3-4).
This means that fatigue is treated as a concept that occurs in response to predefined conditions and has physiological and performance consequences. The antecedent conditions of interest here include:
* Time on task, including flight time and duty period duration
* Time since awake when beginning the duty period
* Acute and chronic sleep debt
* Circadian disruption, multiple time zones, and shift work.
The objectives of this document are to review the scientific research in order to:
* Identify the impact of these antecedent variables on human performance
* Relate these variables to appropriate physiological measures that have been demonstrated to be accompanied by decrements in human performance
Identify, to the extent possible, limitations and requirements concerning duty period durations, minimum sleep requirements, etc. that should be reflected in the regulations.
Indications and Effects of Fatigue
The massive literature on fatigue has identified a number of symptoms that indicate the presence of fatigue, including: increased anxiety, decreased short term memory, slowed reaction time, decreased work efficiency, reduced motivational drive, decreased vigilance, increased variability in work performance, increased errors of omission which increase to commission when time pressure is added to the task, and increased lapse with increasing fatigue in both number and duration (Mohler, 1966; Dinges, 1995). Many of these symptoms appear only after substantial levels of sleep deprivation have been imposed. A review of the literature that involved fatigue levels likely to be experienced by pilots suggests that a common fatigue symptom is a change in the level of acceptable risk an individual will tolerate.
Brown et al. (1970) had subjects drove for four 3-hour sessions. The performance measure used was a count of the number of occasions in which the subject executed what the experimenter considered a risky passing maneuver. When driving performance between the 1st and 4th sessions were compared, a 50% increase in the occurrence of risky passing maneuvers in later sessions, when subjects were presumably more fatigued, was obtained.
This change in the level of acceptable risk was confirmed by Barth et al. (1976) and Shingledecker and Holding (1974) who found that fatigue caused subjects to engage in greater risk taking activity in an effort to avoid additional effort. In the Shingledecker and Holding study, subjects performed 36 choice-of-probability (COPE) tasks, which involved locating a fault in one of three removable banks of one-watt resistors, each with varying degrees of probability that the bank had failed. Twenty-eight days separated the first and last three sets of six trial blocks. In this interim, the experimental group received 24 to 32 hours of continuous work on different monitoring-type fatiguing tasks immediately preceding the second trial block set, while the control group did not. The experimental group was found to shift their selections toward riskier, but less effortful strategies, and made more errors when compared with their own non-fatigued results or control group results. Also, subjects who reported they were tired, although not exposed to intentionally fatiguing activities, behaved similarly. Barth et al. performed a similar experiment, except that fatigue was induced by either a variable pitch/speed bicycle ergometer or a treadmill.
In the aviation domain, this strategy of avoiding effort when fatigued has recently been reported. Neri et al. (1992) found a change in strategy toward risk taking in naval pilots during carrier landings. Risk taking behavior also appears in the form of over reliance on automated systems (Graeber, 1988). This increased passivity, which takes the form of a mental aversion to or avoidance of further effort, is common in both the sleep deprived state and when the individual is experiencing the diurnal low point for body temperature during the circadian trough (Hamilton et al., 1972).
A report of some of the occurrences moments before the crash of the aircraft carrying Commerce Secretary Ron Brown further illustrates the type of inaction typical of fatigue (Newman, 1996). Although the pilots detected an error on approach a full minute before the crash, they made no attempt to correct the error—a common characteristic of fatigue. This is due to a reduced level of adherence to one’s normal standard and a reduced ability to cognitively make a connection between cause and effect. One may recognize a problem but not translate its effect due to lack of full comprehension of the situation or simple failure to initiate an action.
Related evidence exists that fatigued workers are satisfied with lower performance and that perceived errors go uncorrected. There is a "loss in the ability of the worker to perceive and adjust to new aspects of the task. The worker seems unable to shift quickly and effectively from one subpart to another" (Broadbent, 1953; cf. Horne, 1988). The latter quality has been found to be a factor when aircraft crews are concentrating on one problem and allow other problems to develop due to neglect.
In the case of the 1985 China Airlines Flight 006 mishap, the pilot became focused on the loss of power in one engine, neglecting other flight duty tasks. Major structural damage and 2 serious injuries occurred when the aircraft experienced more than 5 g’s during its uncontrolled descent from 31,000 feet to 9,500 feet, before control was regained (Lauber & Kayten, 1988). Contributing fatigue factors to the accident were the Captain’s failure to properly monitor the airplane’s flight instruments, over-reliance on the autopilot after the loss of thrust due to engine failure, and performance of duties during the Captain’s circadian trough. The accident occurred 4 to 5 hours after the time he had been beginning sleep during the 6 nights preceding the accident.
In the Guantanamo Naval Base accident, the pilot was so focused on finding a strobe light that he failed to respond to other crew members’ warnings that they were approaching a stall speed (NTSB Aircraft Accident Report, 1993). In an investigation of Air Force C-5 mishaps or near mishaps, it was reported that 55 percent were related to attentional focus problems and 24 percent to decision making problems (Majors, 1984).
Some symptoms of fatigue are similar to other physiological conditions. For example, with fatigue one’s ability to attend to auxiliary tasks becomes more narrow, very much analogous to the effects of alcohol (Huntley et al., 1973; Moskowitz, 1973), hypoxia (McFarland 1953), and heat stress (Bursill, 1958). Dawson and Reid (1997) evaluated performance after 17 hours awake and found performance degraded to a level equal to that caused by a blood alcohol concentration (BAC) of 0.05 percent. At 24 hours, performance decrements were equivalent to that of a 0.10 BAC. After ten hours of sleeplessness, the decline in performance averaged .74 percent per hour.
Finally, Harrison and Horne (1979) found that sleep loss resulted in a difficulty of generating the ideal word or phrase for the idea or thought the person wanted to convey. In addition, there was a loss in intonation and an overall dullness which suggested loss of interest. The authors suggest that this may very well result in personal communication problems in real life situations.
Effects of Fatigue and Sleep Loss on the Brain
Sleep is mainly a restorative process for brain function. Horne (1991) states that this restoration is primarily a function centered on the cerebral cortex of the brain. This is consistent with the findings of Perelli (1980), who found that a high time since awake significantly increased the threshold for information processing. Pternitis (1981) found that dominant EEG frequencies in power plant operator shift workers showed a progressive decline, with each shift beginning in the morning and continuing to night shift. Morning shift employees showed EEG readings of 12-30 Hz, evening shift workers 6-12 Hz, and those on duty during the night shift, 2-6 Hz. Gevens et al. (1997) has shown that observable performance decrements are preceded by observable EEG brain wave changes that clearly indicate decreasing attentional focus. These EEG changes are observable some time before noticeable performance decrements occur. Howitt et al. (1978) measured EEG activity in operational pilots and found that under high workload situations the fatigued pilots’ EEG rose to only half the level of those displayed by fresh pilots.
Another physiological measure of fatigue and sleep is brain glucose levels. All tissue of the body, whether it be heart muscle, kidneys, lungs, or the brain, works electrochemically, and conforms to one principle: the more work done, the more fuel used. Thus, by measuring glucose utilization, oxygen consumption, and blood flow in the brain, areas which are very active during various tasks can be determined.
Thomas et al. (1993), using positron emission topography (PET) scan has provided strong physiological evidence that sleep loss is accompanied by a decrease in brain glucose metabolism. The areas most involved were the prefrontal cortex, the inferior parietal cortex, and thalamus. During 48 hours sleep deprivation, the overall brain glucose utilization declined 7 percent, while in the areas of higher order thinking declines ranged from 10 to 17 percent (Thomas, 1997). Although these reductions seem relatively minor over a 48 hour period, Gold (1995) recently found that comparatively small blood glucose changes could significantly enhance cognitive performance in a variety of subjects including healthy young adults, elderly, and severe states of pathology such as Alzheimer’s and Downs Syndrome patients.
PET scans of recovery sleep, taken sequentially through the night and synchronized with EEG changes, show that slow wave sleep appears to have its greatest effects on the same brain areas that Thomas et al (1993, 1997) showed were most affected by sleep loss (Braun et al., 1997). This indicates that areas of the brain involved in alertness, attentional focus, concentration, short term memory, drive and initiative, problem solving, complex reasoning, and decision making are the greatest beneficiaries of deep sleep (Lamberg, 1996).
Since the front brain is responsible for analysis of information, judgment, planning, decision making, and the initiation of actions, it is not surprising that NTSB found decision making abilities suffered with high time since awake.
The orderly planning and sequencing of complex behaviors, the ability to attend to several components simultaneously, and then flexibly alter the focus of concentration, the capacity for grasping the context and gist of a complex situation, resistance to distraction and interference, the ability to follow multi-step instructions, the inhibition of immediate but inappropriate response tendencies, and the ability to sustain behavioral output… may each become markedly disrupted (Restak, 1988).
Many of the functions described by Restak are the same functions necessary to a pilot’s ability to competently fly an aircraft.
Although the studies just listed do show performance decrements due to fatigue, other studies have shown no effect (e.g., Rosenthal, 1993), particularly when sleep loss levels up to 24 hours, or small chronic partial sleep loss levels of only one or two hours per day are used. The lack of definitive results in partial sleep deprivation studies may be due to differences in testing procedures. Rosenthal tested on four separate occasions, whereas others tested only once per day. In a more severe sleep deprivation study, Thorne (1983) made the testing instrument the primary task, which lasted 30 minutes of each hour. As sleep loss became increasingly greater, subjects became slower. Therefore, the time to complete the self-paced task increased about 70 percent, and at times doubled.
Evans et al. (1991), in a review of fatigue in combat, clearly stated that studies using embedded testing, such as Thorne (1983), Angus and Heslegrave (1985), and Mullaney et al. (1981), consistently show greater effects of fatigue and sleep loss performance decrements than short duration isolated intrusive tests. Belenky et al. (1986) notes that continuous embedded testing reveals larger performance decrements sooner than does intermittent testing. In Angus and Heslegrave (1985), analysis of results found a 28% decrement in encoding/decoding performance and a 43% decrement in logical reasoning after 24 hours awake. Haslam (1982), using non-embedded testing, found no decrements and 29%, respectively.
The greater sensitivity of embedded testing is not surprising given that they measure performance for a more prolonged period. Brief, intrusive psychometric tests, in contrast, are novel and act as a rest break, distraction, and temporary stimulus, thereby increasing short term mobilization of effort thus boosting performance. The use of such an instrument would function similar to the effect Chambers (1961) found in an industrial output study where output remains higher when a worker was switched to different jobs periodically than to stay at one job.
Another explanation for the varying effects of performance due to fatigue is that performance is, in part, dependent upon the circadian physiology of the subject. Subjects experiencing circadian dysrhythmia or operating during their circadian trough are more likely to yield substandard performance.
Also, motivation can play a major role in the relationship between fatigue and performance. "Both experimenter and subject motivation can have a large impact on results, particularly in the behavioral and subjective domains. Motivation effects are frequently most apparent near the end of studies (where performance improvement is sometimes found) but also may account for the difficulty in showing decrements early in periods of sleep loss" (Bonnet, 1994, p. 50).
In addition to embedded testing, other parameters considered to increase sensitivity in testing for fatigue and sleep loss performance decrements include continuous performance, prolonged vigilance, and multiple task jobs, similar to what is shown to work in decrement due to noise (Belenky 1986; Dejoy, 1984). Self-paced tasks have been reported to be less affected by sleep loss than tasks that are faster work-paced (Johnson & Naitoh, 1974). Fatigue effects tend to be minimal when tasks are self-paced, brief, highly motivating, and feedback is given. On the other hand, tasks which involve sustained vigilance and attention, the use of newly acquired skills, and new information retention tend to challenge short term memory. This is because work-paced tasks accelerate the rate of information processing, thereby decreasing the reserve capacity of brain function.
Roth et al. (1994) support long monotonous objective testing and the MSLT as good measures of sleep loss decrement and sleepiness, respectively. McFarland (1953) considered the deterioration of skills over time a promising framework for the study of fatigue. This has recently been attempted in aviation research by Neville et al. (1992) through the use of flight data recorders for measuring parameters of flight over time. This procedure may be the best avenue yet for truly measuring performance decrements in an operational setting.
Performance measures have obvious value for assessing the effects of fatigue and sleep-related variables. Microsleeps are another useful approach. Microsleeps were first recognized by Bills (1931) and were first called "blocks." Over the intervening years they have also been called "gaps," "lapses" and, more recently, "microsleeps." The physiological drive to sleep can result in a microsleep lasting a few seconds to a few minutes. The latter terminology is the result of EEG recording showing that during these lapses in information processing, subjects momentarily slip into a light sleep. This occurs with the eyes open and usually without the knowledge of the individual, an observation first reported by Miles (1929). Bonnet and Moore (1982) found that before 50 percent of normal subjects became consciously aware of falling asleep, they had been asleep two to four minutes. These intermittent lapses in consciousness impair performance by leading to errors of omission due to missed information. In serial tasks that are work paced, microsleeps can also lead to error of commission and, if frequent enough or long enough, can lead to loss of situational awareness.
Microsleeps have been shown to be a useful approach to assessing the effects of time of day on sleepiness levels. EEG brain wave changes confirm that pilots experience greater sleepiness and decreased alertness between 2:00 to 4:00 a.m. (Gundel, 1995). Alpha waves in EEGs indicate micro events or micro sleeps and have been found to be three times greater during night than during day flights (Samel, 1995). Samel et al. (1997) found that during outbound flights, pilots experienced 273 microsleeps or an average of 1.38 microsleeps per pilot per hour. On return flights the following night, pilots experienced 544 microsleeps or 2.47 microsleeps per hour per pilot. Both feelings of fatigue and the occurrence of microsleeps increased as duty time progressed. Rosekind et al. (1994) also observed micro sleep in pilots and a progressive increase as flights progressed, particularly in the latter portion of the flight. These findings confirm both the physiological occurrence of microsleeps in commercial aviation pilots, and the accumulative nature of fatigue in successive night operations.
The beneficial effects of taking breaks have also been demonstrated by measuring microsleeps. Workers performing continuous tasks without breaks (Bills, 1931; Broadbent, 1958) or suffering from sleep loss began to demonstrate signs of micro sleeps much sooner than those with rest breaks or getting adequate rest, respectively (Kjellberg, 1977b).
The research cited in this section suggests that fatigue may be a factor in the aviation environment due to direct performance decrements and, indirectly, through microsleeps that disrupt pilot functioning. The next section looks at data relating to the occurrence of fatigue in the aviation environment.
Fatigue and The Aviation Environment
The unique characteristics of the aviation environment may make pilots particularly susceptible to fatigue. Environmental factors such as movement restriction, poor air flow, low light levels, background noise, and vibration are known causes of fatigue (Mohler, 1966). In addition, the introduction of advanced automation into the cockpit has changed the nature of the job for many pilots. Hands-on flying has been replaced by greater demands on the crew to perform vigilant monitoring of these systems, a task which people tend to find tiring if performed for long periods of time. For example, Colquhoun (1976) found that monotonous vigilance tasks could decrease alertness by 80 percent in one hour, which is correlated with increased EEG theta activity or sleep-like state. Since physical activity and interest in the task can help to minimize the decline in performance due to continuous work and sleep loss (Wilkinson, 1965; Lille, 1979), automation may contribute to increased drowsiness in pilots suffering from fatigue or sleep loss. Also, as will be shown below, these cognitive-based activities may be susceptible to the effects of fatigue.
Although these environmental characteristics are suggestive, the actual extent to which fatigue is a safety issue needs to be assessed. A study of ASRS incident reports suggested that 21% of incidents were fatigue-related. This figure was challenged by Baker (1996), who pointed out that the database is a biased system due to self reporting, and the data were further biased by the researchers’ interpretation of the reports. Kirsch (1996) argues that the actual ASRS estimate is four to seven percent. Graeber (1985) clarifies the situation as follows:
An initial analysis of NASA’s Aviation Safety Reporting System (ASRS) in 1980 revealed that 3.8 percent (77) of the 2006 air transport crew member error reports received since 1976 were directly associated with fatigue (Lyman & Orlady, 1980). This may seem like a rather small proportion, but as the authors emphasize, fatigue is frequently a personal experience. Thus, while one crew member may attribute an error to fatigue, another may attribute it to a more directly perceived cause such as inattention or a miscommunication. When all reports which mentioned factors directly or indirectly related to fatigue are included, the percentage increases to 21.1 percent (426). These incidents tended to occur more often between 00:00 and 06:00 [local time] and during the descent, approach or landing phases of flight. Furthermore, a large majority of the reports could be classified as substantive, potentially unsafe errors and not just minor errors.
In a study of flightcrew-involved major accidents of domestic air carriers during the 1970 through 1990 period (NTSB, 1994), one conclusion pertained directly to the issue of fatigue: "Half the captains for whom data were available had been awake for more than 12 hours prior to their accidents. Half the first officers had been awake more than 11 hours. Crews comprising captains and first officers whose time since awakening was above the median for their crew position made more errors overall, and significantly more procedural and tactical decision errors" (p. 75). This finding suggests that fatigue may be an important factor in the carrier accidents. Because the study involved only domestic carrier accidents, it remains unclear as to whether other fatigue-related factors, such as long flight times and circadian disruption due to multiple time zones would also appear as causative factors. On the basis of this study, the NTSB recommended that the FAA address the issues of flight duty times and rest periods.
Although the results of this study are suggestive, the actual impact of fatigue has yet to be determined. Since no real effort has been made to identify the effects of fatigue in accident and incidence investigation, it is difficult to assess the magnitude of the problem. In addition, it is possible that self-reporting systems, such as ASRS, may be affected by the inability of people to accurately assess their own fatigue levels (Sasaki et al., 1986; Richardson et al., 1982; Dinges, 1989). Subjective evaluations of sleepiness have not been found to be reliable except in extreme sleepiness. Rosekind and Schwartz (1988) noted that the scientific literature generally demonstrates a discrepancy between subjective reports and psychophysiological measures, the result being underestimations of one’s level of sleepiness (cf. Dement & Carskadon, 1981). Dement et al. (1978) and Roth et al. (1994) reported that some subjects judged themselves alert, when in fact they were in the process of falling sleep.
Graeber et al. (1986), summarizing the collaborative efforts between European, Japanese, and American investigators to evaluate sleep in long haul aircrews, reported that subjective evaluations are sometimes erroneous as to the true nature of the psychophysiological state of sleepiness. These results were obtained in two separate studies by Dement et al. (1986) and Sasaki et al. (1986). Mullaney et al. (1985) also found that subjects subjectively felt that they performed better under sleep loss conditions when paired with another subject, when in reality it had no effect on actual performance decrements. Rosekind et al. (1994) found pilots unable to subjectively evaluate changes in performance due to a short inflight nap. Although pilots did show physiological improvements in alertness, they could not subjectively notice a difference. Belenky et al. (1994) points out that due to the psychophysiology changes in higher order cognitive judgment areas with fatigue and sleep loss, these changes automatically preempt ones ability to evaluate his or her own performance accurately.
One possible reason for these findings is that the presence of certain factors masks sleepiness and the absence of other factors unmasks sleepiness. Environmental factors that have a masking affect include noise, physical activity, caffeine, nicotine, thirst, hunger, excitement, talking about something interesting, etc. For example, Howitt et al. (1978) found that sleep deprived pilots in operational settings felt no noticeable fatigue once flight preparations were under way and flight commenced. This explanation is supported by research that used the multiple sleep latency test (Dement et al., 1986, Sasaki et al., 1986; Rosekind et al., 1994; Roth et al., 1994). In contrast to the subjective evaluation, the multiple sleep latency test asks subjects to quietly lie down, close their eyes and try to sleep. This in essence removes many of the masking factors, whereas subjective alertness in relation to EEG recording appears to have better correlation because both can be recorded in the same environmental setting. Ogilvie et al. (1989) reported that subjective sleepiness responses to the Sanford Sleepiness Scale only reached significance when subjects were entering stage I sleep. Thus it may be that when EEG alpha and theta activity appears there is truly a feeling of sleepiness.
Although masking reduces perceived feelings of sleepiness, it does not counteract the effects of fatigue on performance. Kecklund and Akerstedt (1993) conclude that although sleep-deprived subjects may not feel their sleepiness or fatigue due to environmental variables, the sleep pressure is still latently present.
Standard Duty Period
The first regulatory issue that needs to be addressed concerns the duration of the standard duty period. "Standard" is used here to refer to duty periods that do not involve window of circadian low (WOCL) effects or time zone changes. The primary focus of the standard duty period issue addresses the buildup of fatigue as a function of performing the various tasks involved in a duty period. Six factors that may need to be considered are:
* Time on task
* Time since awake
* Task type
* Duty period extension
* Cumulative duty times
* Environmental factors.
Each of these factors is discussed below.
There appears to be some consensus that the effects of time-on-task on performance are difficult to assess (e.g., Maher & McPhee, 1994) and are affected by a number of variables, including time of day, the nature of the task, the subject’s motivational level, and if fatigue or sleep loss are already present (Dinges & Kribbs, 1991; Maher & McPhee, 1994; Mendelson, Richardson & Roth, 1996). In spite of this, performance on many laboratory tasks follows a similar curve (Vries-Griever & Meijman, 1987): relatively low starting performance, followed by optimal performance, which then declines due, presumably, to fatigue. The points at which optimal performance begins and then starts to degrade varies with the task. For some cognitive tasks, optimal performance is achieved after about five hours, then declines to its lowest levels after 12 to 16 hours on task (Spencer, 1987; Nicholson, 1987). Some tasks, such as monitoring tasks that require high levels of vigilance, show performance decrements after shorter durations. Colquhoun (1976) found that monotonous vigilance tasks could decrease alertness by 80 percent in one hour based on increased EEG theta activity which correlates with a sleep-like state. Reductions in task performance over time are also accompanied by an increased need to sleep, as shown by Lisper et al. (1986), who found that car drivers showed an increased likelihood of falling asleep after 9 hours of driving.
Time-on-task measures for a single task may have limited applicability to the aviation domain as the pilot’s job involves performing a number of tasks during a given duty period. Switching between individual tasks may override some of the effects of fatigue due to time-on-task. Studies which have investigated the effects of extended shift durations on worker performance may be relevant as they assess fatigue and performance as a function of the set of tasks that are performed during a shift rather than performance decrements that accrue on a single task. In a manufacturing environment (Rosa & Bonnet, 1993), the number of errors made was relatively high at the beginning of the shift, then decreased because of re-familiarization with the task. Optimal levels were reached within a few hours, then declined over the eight-hour shift. In general, workers on 12-hour shifts became considerably more fatigued than in more traditional eight- to 10-hour shifts (Rosa & Colligan, 1987). This finding has been confirmed in nurses (Mills et al., 1983), industrial shift workers (Colligan & Tepas, 1986), night shift workers (Rosa & Colligan, 1987), sea watch workers (Colquhoun, 1985), and truck drivers (Hamelin, 1987). The latter study also found an increase in the number of accidents that occur when 12-hour shifts are used.
This increased likelihood of accident risk due to long duty periods has been found in other studies. The relative risk of an accident at 14 hours of duty rises to 2.5 times that of the lowest point in the first eight hours of duty. Askertedt (1995) reports accident risks to be threefold at 16 hours of duty, while Harris and Mackie (1972) found a threefold risk in just over 10 hours of driving. These levels of risk are similar to that associated with having narcolepsy or sleep apnea (Lavie et al., 1982), or a blood alcohol level of 0.10 percent. Wegmann et al. (1985), in a study of air carrier pilots, argued for a duty period of 10 hours with 8.5 hours or less of flight duty period.
Time Since Awake
The results of an NTSB analysis of domestic air carrier accidents occurring from 1978 to 1990 suggest that time since awake (TSA) was the dominant fatigue-related factor in these accidents (NTSB, 1994). Performance decrements of high time-since-awake crews tended to result from ineffective decision-making rather than deterioration of aircraft handling skills. These decrements were not felt to be related to time zone crossings since all accidents involved short haul flights with a maximum of two time zones crossed. There did appear to be two peaks in accidents: in the morning when time since awake is low and the crew has been on duty for about three to four hours, and when time-since-awake was high, above 13 hours. Similar accident peaks in other modes of transportation and industry have also been reported (Folkard, 1997). Akerstedt & Kecklund (1989) studied prior time awake (four to 12 hours) and found a strong correlation of accidents with time since awake for all times of the day. Belenky et al. (1994) found that flight time hours (workload) greatly increase and add to the linear decline in performance associated with time since awake.
The effects of task type, as they contribute to the buildup of fatigue, need to be considered from two perspectives:
* Whether certain activities can be excluded from duty period time
* Whether certain activities are inherently more fatiguing and may need to be restricted.
The current regulations regulate only flight time. No limits are provided for duty time. The regulations proposed in the Notice of Proposed Rulemaking 95-18 (NPRM) allow for the concept of "assigned time," which also is unregulated as to maximum limits. The extent to which activities categorized as non-flight time or assigned time contribute to fatigue has yet to be empirically ascertained. However, it is clear that these activities would contribute to fatigue in the form of time since awake. Consequently, it may be appropriate to limit these activities in either of two ways:
* With respect to when they occur relative to flight time so as to avoid pilots achieving high time-since-awake levels during flight time periods.
* Provide maximum levels for these activities comparable to duty period time levels.
The second issue pertaining to task type concerns activities which are known to be inherently more fatiguing. One such activity is the approach and landing. Gander et al. (1994) found that increases in heart rate occurred during the approach and landing phases when compared with other duty period activities. Because heart rate increase is a common measure of workload, this suggests that proposals to limit landings for flights that have other known fatigue factors (e.g., time since awake, window of circadian low, extended flight duty periods) may be appropriate.
The relationship between task type and fatigue buildup in the aviation domain remains to be determined. The demands placed on long-haul pilots are clearly different from those of the regional carrier pilot flying many legs in a propeller-driven airplane with limited automation. Flights across the ocean typically involve a single leg of six or more hours. The main task-related fatigue sources in this case are boredom and cognitive fatigue due to vigilance. The regional pilot, in contrast, may be more susceptible to fatigue due to the high workload involved in performing six or more takeoffs and landings. For this reason, it may prove necessary to develop separate regulations that are appropriate for each major type of operation.
Duty Period Extensions
The research cited on duty period duration suggests that duty periods at or above 12 hours are associated with a higher risk of error. This factor, together with the time-since-awake factor, suggests that extended duty periods also involve a higher potential for crew error. In determining maximum limits for extended duty periods, consideration also needs to be given to other fatigue-related factors that could contribute to excessive fatigue levels during extended duty periods, including number of legs, whether the flight impinges on the window of circadian low (WOCL), and time since awake.
Cumulative Duty Time
No data were found that provide guidance for maximum duty times over longer time periods, such as one month or one year.
The physical environment of the cockpit is a source of other factors that can contribute to fatigue (Mohler, 1966). Factors such as vibration, poor ventilation, noise, and the availability of limited automation can contribute to the buildup of fatigue or accelerate its onset when coupled with time since awake, number of legs, and whether the flight involves the WOCL. This may have implications for regional carrier pilots who fly propeller-driven aircraft.
The research cited suggests an increase in the likelihood of error as duty periods are extended beyond 12 hours. This finding is especially critical for extended duty periods which are likely to occur under conditions (e.g., weather) that, in and of themselves, may increase the probability of crew error.
The interactions between multiple fatigue-related factors must also be considered. Separately, duty period duration, time since awake, number of legs, and environmental factors contribute to fatigue buildup. When any one of these factors reaches a high level, consideration should be given to reducing the maximum allowable levels on these other factors. Time since awake also has obvious implications for reserve assignments and for pilots who commute.