Do birds sleep in flight?
The following review examines the evidence for sleep in flying birds. The daily need to sleep in most animals has led to the common belief that birds, such as the common swift (Apus apus), which spend the night on the wing, sleep in flight. The electroencephalogram (EEG) recordings required to detect sleep in flight have not been performed, however, rendering the evidence for sleep in flight circumstantial. The neurophysiology of sleep and flight suggests that some types of sleep might be compatible with flight. As in mammals, birds exhibit two types of sleep, slow-wave sleep (SWS) and rapid eye-movement (REM) sleep. Whereas, SWS can occur in one or both brain hemispheres at a time, REM sleep only occurs bihemispherically. During unihemispheric SWS, the eye connected to the awake hemisphere remains open, a state that may allow birds to visually navigate during sleep in flight. Bihemispheric SWS may also be possible during flight when constant visual monitoring of the environment is unnecessary. Nevertheless, the reduction in muscle tone that usually accompanies REM sleep makes it unlikely that birds enter this state in flight. Upon landing, birds may need to recover the components of sleep that are incompatible with flight. Periods of undisturbed postflight recovery sleep may be essential for maintaining adaptive brain function during wakefulness. The recent miniaturization of EEG recording devices now makes it possible to measure brain activity in flight. Determining if and how birds sleep in flight will contribute to our understanding of a largely unexplored aspect of avian behavior and may also provide insight into the function of sleep.
Virtually, every animal studied spends a large proportion of each day asleep, a motionless state of reduced environmental awareness (Rattenborg and Amlaner 2002). The daily occurrence of sleep, even under risky circumstances, indicates that sleep serves an important function (Lima et al. 2005). Although controversy persists over the specific function(s) of sleep, most evidence indicates that during sleep, the brain performs functions that are essential to sustaining adaptive brain function during wakefulness (Rechtschaffen 1998; Krueger and Obál 2002; Van Dongen et al. 2003; Stickgold and Walker 2005; Siegel 2005; Tononi and Cirelli 2006). The daily occurrence of sleep and the relationship between sleep and waking brain function has led to speculation over whether birds that engage in nonstop flights lasting longer than a day sleep in flight (e.g., Ashmole 1963). The behavioral and electrophysiological recordings required to detect sleep in flight have not been performed, however, rendering the evidence for sleep in flight circumstantial. Moreover, the belief that such birds must sleep in flight was recently challenged by the finding that captive songbirds in a migratory state are able to function adaptively despite sleeping nearly two-thirds less than when in a nonmigratory state (Rattenborg et al. 2004). Apparently, songbirds have evolved a mechanism to temporarily dispense with a large amount of sleep during migration, thereby, raising the possibility that other birds employ a similar adaptation to remain awake during long flights (see also Lyamin et al. 2005).
The following review examines the evidence for sleep in flying birds. After describing the sleep states exhibited by birds, I discuss whether the neurophysiological changes that accompany each state are compatible with flight. I then examine the evidence for prolonged flights in the various birds in which sleep in flight has been proposed. Finally, I discuss the obstacles and potential solutions to recording sleep in flight and the implications of determining whether birds sleep while flying. I intend to emphasize the idea that the question, “Do birds sleep in flight?” is not just a phenomenological one but one which has fundamental implications for understanding a largely unexplored area of avian behavior and the function of sleep.
Definition of sleep
A discussion of sleep in flight necessarily requires an understanding of what constitutes sleep. In general terms, sleep is a behavioral state instantiated by brain states fundamentally different from wakefulness. As a result, sleep can be defined behaviorally and electrophysiologically. From a strictly behavioral standpoint, a sleeping animal displays the following: 1) behavioral quiescence, 2) a species-specific posture which usually includes eye-closure, 3) increased arousal threshold to stimulation, 4) rapid reversibility to wakefulness with sufficient stimulation (Piéron 1913; Flanigan 1972), and 5) an increase in sleep duration and/or intensity after sleep deprivation (i.e., homeostatic regulation) (Tobler 1985). In addition, in many animals, sleep occurs during a specific portion of the circadian rhythm (Ball 1992). Based strictly on the behavioral definition of sleep, birds in flight do not sleep because they are not quiescent. Regardless, the finding that cetaceans can swim during an electrophysiologically defined state in which one cerebral hemisphere sleeps, while the other remains awake, has already challenged this aspect of the behavioral definition of sleep (see below). Thus, although sleeping animals are usually quiescent, swimming cetaceans and flying birds may be exceptions.
Electrophysiological sleep states
In birds and mammals, sleep behavior is associated with two electrophysiologically distinct sleep states, slow-wave sleep (SWS) and rapid eye-movement (REM) sleep (reviewed in Rattenborg 2006). Although changes in several physiological parameters distinguish wakefulness from sleep, and SWS from REM sleep, the electroencephalogram (EEG) has received the most attention (Steriade 2003). The EEG measures the summation of excitatory and inhibitory postsynaptic potentials generated by neural tissue underlying the EEG electrodes. During wakefulness, the EEG exhibits low-amplitude, high-frequency activity arising from tonic high levels of asynchronous neuronal activity. In contrast to this activated pattern observed during wakefulness, the EEG during SWS displays high-amplitude, low-frequency activity, reflecting the synchronous alternation of neurons between a hyperpolarized state with no action potentials and a depolarized state with action potentials occurring at a high rate comparable to wakefulness (Steriade 2003). In REM sleep, the EEG reverts back to an activated pattern similar to that observed during wakefulness. During mammalian REM sleep, brainstem neurons also discharge at a high rate comparable to wakefulness, but unlike wakefulness, brainstem activation does not result in overt behaviors (Siegel 2004). Neuronal activity in the pons and adjacent midbrain simultaneously inhibit and remove excitation to motoneurons, thereby, causing a complete loss of postural muscle tone. The REM sleep-related reduction in muscle tone also occurs to some extent in birds, presumably through similar mechanisms.
Compatibility of sleep and flight
Several lines of evidence suggest that at least some types of sleep may be possible during some types of flight. Regarding the type of sleep, it is most commonly suggested that birds engage in USWS during flight, but bihemispheric sleep may also be possible. Similarly, while sleep during soaring flight seems most likely, sleep may also occur during flapping flight. In the following sections, I discuss neurophysiological factors that may determine whether each type of sleep occurs during each flight strategy.
Unihemispheric slow-wave sleep
Based on these observations in dolphins, the bilateral locomotor activity needed to sustain soaring and flapping flight seems feasible during avian USWS. However, the ability of the open eye and awake hemisphere to process visual information during USWS in flight is unclear. Unlike dolphins, EEG activity in the hemisphere connected to the open eye is usually intermediate between that of unequivocal wakefulness and unequivocal SWS during avian USWS (Fig. 1) (Rattenborg et al. 1999a,b, 2001). Given the intermediate nature of the EEG, processing of visual information might also fall somewhere between wakefulness and sleep. In the only study to assess visual processing during avian USWS, a video image of a looming object presented to the open eye induced rapid escape behavior in mallards (Rattenborg et al. 1999a,b). While this test showed that birds are able to respond adaptively to threatening stimuli during USWS, the extent to which birds process more subtle aspects of their environment in this state remains unknown (Lima et al. 2005). The degree of interhemispheric EEG asymmetry observed in sedentary birds, however, may not be the same as that occurring in flying birds. Given the capacity to modulate sleep intensity between the two hemispheres while sedentary (Rattenborg et al. 1999a,b), birds may also be able to increase EEG activation and visual processing in the awake hemisphere during USWS in flight.
Regarding visual processing during USWS, it should also be mentioned that the two eyes and associated brain hemispheres do not process visual input in the same manner (reviewed in Vallortigara and Rogers 2005). Of particular relevance for understanding USWS in flight, a bird’s ability to navigate depends on the eye used. In homing pigeons (Columba livia) forced to use only one eye by occluding the other eye, homing performance was higher in birds using the right eye and left hemisphere, than in birds using the left eye and right hemisphere (Prior et al. 2004). Similarly, in the European robin (Erithacus rubecula), migratory orientation depends on visual detection of the magnetic field by the right eye and left hemisphere (Wiltschko et al. 2002). Given this laterality in the avian visual system, birds in flight may prefer to engage in USWS with the right eye open and left hemisphere awake, rather than USWS with the left eye open and right hemisphere awake. Alternatively, hemispheric differences in processing information may have a limited effect on which hemisphere sleeps, if brief awakenings of the left hemisphere are sufficient for a bird to assess and respond to its current position.
Bihemispheric slow-wave sleep
Most discussions on sleep in flight have focused on USWS, but BSWS might also be possible, particularly when frequent visual monitoring of the environment is unnecessary. BSWS during soaring flight seems most likely because birds would only need an ability to maintain their wings in a position that sustains flight. Indeed, sedentary birds and mammals seem readily capable of maintaining postures that oppose the pull of gravity during BSWS (Amlaner and Ball 1994; Zepelin et al. 2005). Flapping flight may also be possible during BSWS. In dolphins, the presence of coordinated bilateral locomotion during USWS shows that a limb connected to a sleeping hemisphere can function adaptively, presumably through brainstem mechanisms unaffected by the sleeping state of the neocortex. Thus, it is conceivable that bilateral locomotion can also occur during BSWS. Indeed, albeit under pathological conditions, sleepwalking humans ambulate while the neocortex displays EEG activity characteristic of SWS (Zadra et al. 2004). Moreover, because the brainstem is sufficient to induce flapping flight in decerebrated birds (Salzen and Parker 1975), bihemispheric sleep may not interfere with flight. Finally, birds may be able to visually navigate between periods of BSWS in flight by intermittently opening one or both eyes, much in the same manner that birds briefly interrupt sleep while sedentary to look for predators (Lendrem 1983, 1984; Gauthier-Clerc et al. 1998, 2000, 2002; Dominguez 2003).
In contrast to SWS, the general reduction in skeletal muscle tone that occurs during REM sleep in mammals and, to a lesser extent, birds, is incompatible with several behaviors that rely on postural muscle tone, probably including flapping and soaring flight. Rats forced to sleep on a small platform surrounded by water, fall off as soon as they transition from SWS to REM sleep, an outcome that is often used to deprive rats of REM sleep (Van Luijtelaar and Coenen 1986). In horses (Equus caballus), elephants (Elephas maximus) and giraffes (Giraffa camelopardalis), SWS occurs while standing or lying, but REM sleep only occurs while lying, presumably due to the reduction in muscle tone (Ruckebusch 1972; Tobler 1992; Tobler and Schwierin 1996). In cows (Bos taurus) and sheep (Ovis aries), rumination persists during SWS, but stops during REM sleep (Ruckebusch 1972). Although northern fur seals (Callorhinus ursinus) are able to keep their nostrils above the water surface by paddling the flipper connected to the awake hemisphere during USWS, paddling stops and the seal’s head sinks below the surface during REM sleep (Lyamin et al. 1996). This may account for the marked reduction in REM sleep when sleep in the water is compared with that on land. Sea otters are also able to keep their head above water during SWS, but not REM sleep (Lyamin et al. 2000). A similar conflict between REM sleep and locomotor activity may have led to the large reduction or complete elimination of REM sleep in cetaceans (Mukhametov 1995). In birds, REM sleep is also associated with behavioral signs of reduced muscle tone, including gradual dropping of the head towards the ground, swaying in standing birds, and sliding of the wings off the side of the body (Dewasmes et al. 1985; Rattenborg and Amlaner 2002). Even thermoregulatory behaviors such as panting and shivering stop during REM sleep in birds and mammals, in part due to the reduction in muscle tone (Heller et al. 1983; Parmeggiani 2003). Based on these observations, the REM sleep-related reduction in muscle tone may also interfere with the flight mechanics of birds.
Before entirely ruling out the possibility for REM sleep in flight, however, there are some differences between REM sleep in mammals and birds that may permit small amounts of REM sleep to occur in flight. In contrast to mammals, where REM sleep-related reductions in muscle tone are readily detected in electromyogram (EMG) recordings, the EMG in birds rarely shows reductions in activity during REM sleep, a finding replicated in several laboratories examining a variety of species (reviewed in Dewasmes et al. 1985; Rattenborg and Amlaner 2002). This seems paradoxical because behavioral signs of reduced muscle tone are readily apparent during REM sleep. A study of the domestic goose provides a potential explanation for this apparent paradox (Dewasmes et al. 1985). As in other birds, when geese slept with their head facing forward the head dropped, despite relatively high levels of neck EMG activity. However, EMG atonia similar to that observed in mammals was readily apparent when the head rested on the birds’ back. Thus, in the absence of adequate support, geese, and presumably other birds, increase muscle tone during REM sleep, thereby, partially counteracting the gravitational pull on the head. Although the head still drops, the fall appears more gradual and less extreme than if complete muscle atonia were present. Interestingly, the duration of REM sleep episodes was not influenced by the head position of the geese, suggesting that neither the elevation in muscle tone nor the drop of the head disrupt REM sleep (Dewasmes et al. 1985). In conjunction with the observation that REM sleep episodes in birds typically only last several seconds (reviewed in Rattenborg and Amlaner 2002), birds may obtain some REM sleep in flight if they are able to maintain muscle tone at a level sufficient to keep the wings in a position that sustains flight.
Finally, based on a hypothesis recently proposed to explain the small amount or absence of REM sleep in cetaceans (Siegel 2005; see also Horne 2000), the need for REM sleep in birds may be greatly reduced during long flights. According to this hypothesis, activation of the brainstem was an evolutionarily early function of REM sleep, and functions associated with REM sleep EEG activation were added more recently. Indeed, in the echidna (Tachyglossus aculeatus) and duck-billed platypus (Ornithorhynchus anatinus), members of an evolutionarily ‘ancient’ group of egg-laying mammals, the monotremes, REM sleep occurs in the brainstem while the neocortex shows high-amplitude, low-frequency EEG activity indicative of SWS, rather than the low-amplitude, high-frequency activity that typically occurs during REM sleep in placental and marsupial mammals and birds (Siegel et al. 1996, 1999). In dolphins, REM sleep-related brainstem activation may not be necessary because the brainstem is sufficiently activated during bilaterally symmetrical swimming during USWS (Siegel 2005). Similarly, in birds engaged in long flights, particularly flapping flights, the need for REM sleep may be greatly reduced because the brainstem is constantly activated. As suggested for dolphins, this would hold true even if birds engage in USWS or BSWS in flight, as long as the brainstem is sufficiently activated. Functions associated with REM sleep-related EEG activation may be achieved by the same mechanisms that cetaceans employ to dispense with neocortical activation. If birds engaged in long flights do, in fact, require less REM sleep, we would not expect to see a postflight rebound in REM sleep, as observed after sleep deprivation in sedentary birds (Tobler and Borbély 1988). In summary, it seems unlikely that birds engage in much, if any, REM sleep in flight, either because REM sleep is incompatible with flight or the need for REM sleep is greatly reduced during flight.
Candidates for sleep in flight
Of all the birds thought to sleep in flight, perhaps the common swift (Apus apus) has received the most attention. The idea that swifts spend the night on the wing arose from observations of swifts ascending into the sky above their colony after sunset and descending the following morning (Edwards 1887; Masson 1930; Weitnauer 1952, 1954, 1960; Lack 1956). Although collisions with and observations from airplanes showed that swifts did, in fact, fly at night (Weitnauer 1952, 1954–1956; Lack 1956), it was not until Weitnauer used radar to track swifts throughout the night that aerial roosting was confirmed (Weitnauer 1960, 1980; Bruderer and Weitnauer 1972). Nevertheless, these radar studies were unable to determine whether individual birds flew for an entire night or the number of consecutive nights spent in flight.
Recently, Tarburton and Kaiser (2001) used radio tracking to follow individual swifts in flight for periods up to two consecutive nights before loosing the signal. The birds participating in aerial roosting were primarily pre-breeders; breeding adults roosted, and presumably slept, in the nest box and only engaged in aerial roosting toward the end of the fledging period. Although the tracking of pre-breeders was discontinuous, due to temporary signal loss, all but one of the pre-breeders tracked were always encountered while in flight, regardless of whether it was day or night. The exceptional individual roosted at night on a building, a behavior rarely observed at this study site. Even fledglings typically departed from the nest for the first time after sunset and spent their first night outside the nest in flight. In other reports, recently fledged juveniles have been found roosting on foliage or other substrates more often than adults, however (Holmgren 2004). Non-aerial roosting occurs primarily during low ambient temperatures, suggesting that juveniles are more energetically vulnerable than adults, and therefore, more likely to resort to non-aerial roosting, presumably a less energetically expensive behavior than aerial roosting. Although there is substantial evidence showing that pre-breeding adult common swifts often spend the night in flight, the average and maximum length of time spent in continuous flight remains unknown. In conjunction with the results from tracking studies, the fact that observations of nonbreeding birds roosting terrestrially are noteworthy suggests that aerial roosting is the typical nocturnal behavior for common swifts, at least during the breeding season. Aerial roosting may also occur in other species of swifts, but the evidence is not as substantial as that for the common swift (reviewed in Holmgren 2004). The evidence for sleep in flying common swifts, however, is circumstantial, depending solely on the notion that if swifts must sleep, such sleep must occur on the wing.
Aside from spending the night in flight, perhaps, the only other piece of evidence for sleep in flight comes from a radar study of the flight patterns of common swifts at night. Bäckman and Alerstam (2002) tracked individual swifts flying at night for periods up to 1 h. In general, the swifts flew into the wind, a flight pattern that reduced displacement from the colony. Interestingly, the swifts’ orientation oscillated with a period of 1–16 min, around the direct heading into the wind, a flight pattern expected under wind speeds lower than the swifts’ lowest flight speed, assuming that the swifts’ goal was to remain over the colony. However, this oscillation occurred at all wind speeds, including those higher than the swifts’ flight speed. Bäckman and Alerstam (2002) suggest that this flight pattern may reflect oscillations in the level of alertness, possibly related to sleep in flight. Weaving from left to right of the direct heading may result from a sensory bias related to flying alternately with only the left or right eye open. Indeed, pigeons flying with one eye occluded tend to veer toward the side with the open eye (Prior et al. 2004).
Diurnal passerines migrating at night
Twice a year, normally diurnal passerines engage in nocturnal migratory flights between their breeding and nonbreeding range (reviewed in Berthold 2001). Although the duration of nocturnal flights shows much inter- and intra-specific variation, radio tracking studies of Catharus thrushes show that individual birds may fly for much of the night for several consecutive nights (Cochran 1987; Bowlin et al. 2005). Although there is insufficient data on daytime activity to determine whether songbirds in the wild sleep more in the day after nocturnal flights, birds may spend a significant amount of time feeding to replenish energy stores and watching for predators in a novel environment, thereby, limiting time available for daytime sleep (Moore et al. 2005). Pressure to sleep in flight is probably greatest in passerines that cross large ecological barriers, such as oceans, which preclude landing. In perhaps the most extreme case, each fall blackpoll warblers (Dendroica striata) and, to a lesser extent, other warbler species, appear to fly nonstop from the northeast coast of North America to the northern coast of South America, a 3,500 km flight across the Atlantic Ocean estimated to last 82–88 h (Williams et al. 1978). Although individual birds have not been tracked throughout the flight, several lines of evidence support the contention that blackpolls complete this flight nonstop. Blackpolls depart from the coast in a southeast direction over the Atlantic Ocean with energy reserves estimated to be sufficient to complete the flight without refueling (Nisbet et al. 1963; Cherry et al. 1985). Moreover, radar studies show that flocks of songbirds typically fly over Bermuda in a southeast direction without stopping before turning south somewhere between Bermuda and Antigua and continue south until making landfall near the northern coast of South America (reviewed in Williams and Williams 1978). Blackpolls either delay sleep until landing in South America, or sleep in flight.
Palaearctic passerines, such as the garden warbler (Sylvia borin), cross the Mediterranean Sea and Sahara Desert using various approaches that may be influenced by the need to sleep. Although individual birds have not been followed, radar studies suggest that some individuals fly nonstop across the Mediterranean Sea and Sahara Desert, a flight estimated to last 42 h, whereas, others stop and rest quietly with closed eyes (H. Biebach, personal communication), presumably sleeping, during the day in small shaded spots in the desert, far from suitable oases (Biebach et al. 1991). Birds found resting in the desert have sufficient energy reserves and hydration to complete the desert crossing (Bairlein 1985; Biebach et al. 1986), and do in fact, depart the next evening (Biebach et al. 1991). Then why do birds stop in the desert? In conjunction with other factors (Biebach et al. 2000; Klaassen and Biebach 2000), a bird’s prior sleep history may influence their decision to either stop in the desert or complete the crossing nonstop. Compared to individuals that acquired sufficient sleep before initiating a crossing, those with a sleep deficit may be more likely to stop in the desert to sleep (see Schwilch et al. 2002).
The sleep patterns of passerines have not been measured directly during migration in the wild. However, a few studies have examined the behavioral state of songbirds during the migratory season in captivity. At the time of year when songbirds normally migrate in the wild, captive birds flap, or whirr their wings while holding onto a perch, a behavior referred to as migratory restlessness or Zugunruhe (reviewed in Berthold 2001). In birds held under constant environmental conditions for several years, the timing of migratory restlessness drifts, or free runs, relative to the timing of migration in wild birds, indicating that it is controlled by an endogenous circannual rhythm (Gwinner 1996; Gwinner and Helm 2003). Migratory restlessness seems to reflect the natural migratory tendency of a given species because the duration of migratory restlessness within and across nights is correlated with the flight behavior of conspecifics in the wild (Gwinner 1986). Thus, the sleep patterns of captive songbirds may provide insight into how birds sleep during migration in the wild.
The electrophysiological recordings of white-crowned sparrows suggest that songbirds have a capacity to dispense with a large proportion of their sleep during the migratory season. An obvious caveat is that the sleep patterns observed during migratory restlessness in captivity, where true flight was not possible, may not be the same as that in free-flying birds. Unlike captive birds, those flying in the unobstructed night sky might obtain some sleep in flight. Nevertheless, if songbirds rely heavily on a strategy of sleeping during migratory flights, captive songbirds exhibiting migratory restlessness would not be expected to sustain long periods of wakefulness at night without adversely affecting cognitive performance in the daytime.
Virtually nothing is known about the behavioral state of passerines during migratory flights in the wild. During nocturnal flights, songbirds emit flight calls and presumably listen for those from other birds (Farnsworth et al. 2004). Although these behaviors likely occur when a bird is awake, periods of sleep could still occur between calls, and the calls of other birds might be detected during sleep, particularly USWS. Collisions with man-made structures do not seem to reflect lapses in attention related to sleep, but rather, attraction to and entrapment within the illuminated airspace surrounding lighted structures on cloud covered nights (Cochran and Graber 1958; Larkin and Frase 1988). Finally, although brief pauses in flapping flight occurring more frequently toward the end of nocturnal flights in Catharus thrushes may reflect brief periods of sleep, they may also indicate changes in flight patterns associated with descent (Bowlin et al. 2005).
Many shorebirds seem to engage in transcontinental and transoceanic migratory flights spanning thousands of kilometers and encompassing multiple days and nights of nonstop flight (reviewed in Piersma et al. 1990, 2005; Williams and Williams 1999; Van de Kam et al. 2004). In perhaps the most extreme case, a recent study suggests that bar-tailed godwits (Limosa lapponica baueri) complete a nonstop, 5- to 6-day, 11,000-km flight from western Alaska to New Zealand and eastern Australia during their fall migration (Gill et al. 2005). Although direct measurements of the movements of individual birds engaged in such flights have not been reported, several lines of evidence, including flight distance simulation models and the low occurrence of godwits on intervening islands, support the contention that this flight is completed nonstop. Anecdotal observations of bar-tailed godwits and red knots (Calidris canutus) sleeping upon arrival in New Zealand, rather than feeding, as observed in conspecifics that had arrived earlier, suggest that shorebirds are at least partially sleep deprived during the flight (see Schwilch et al. 2002).
Albatrosses spend much of their lives soaring at sea, a lifestyle that has led to speculation over whether they sleep on the wing. Tracking studies of several species conducted in the last 15 years have confirmed the remarkable flight capacity of albatrosses (Jouventin and Weimerskirch 1990; Croxall et al. 2005) but also indicate that albatrosses may sleep on the ocean surface. Although albatrosses may travel hundreds of kilometers in a 24-h period, commuting and foraging flights occur primarily during the daytime, at least, in the breeding season (Prince and Morgan 1987; Jouventin and Weimerskirch 1990; Weimerskirch et al. 1997; Weimerskirch and Guionnet 2002). At night, yellow-nosed (Diomedea chlororhynchos), sooty (Phoebetria fusca), black-browed (Diomedea melanophris impavida), grey-headed (Diomedea chrysostoma), and wandering albatrosses (Diomedea exulans) float on the water and probably sleep because they do not seem to feed on the surface at night (Weimerskirch and Wilson 1992; Weimerskirch et al. 1994; Weimerskirch and Guionnet 2002). Interestingly, the amount of time spent flying during the day is correlated with the amount of time spent on the water during the subsequent night (Weimerskirch and Guionnet 2002). This may reflect a need to digest food acquired during the previous day and a compensatory increase in sleep resulting from increased time spent awake while in flight. Even during commuting flights between the nest and foraging areas, black-browed albatrosses stop and spend much of the night floating on the water (Weimerskirch and Guionnet 2002). If albatrosses were able to obtain sufficient sleep in flight, the parents would be expected to minimize the interval between chick feedings by flying nonstop during the commuting legs of foraging trips. Nocturnal roosting on the water during commuting flights and the increase in time spent on the water after days with more flight suggest that albatrosses accrue a sleep deficit during flight that is replenished while sleeping on the water or land.
The relationship between day flight and night floating may arise from the complete absence of sleep in flight or the acquisition of insufficient amounts of sleep in flight. For example, while albatrosses may obtain some SWS in flight, either unihemispherically or bihemispherically, REM sleep may not occur in flight. Although the wings of albatrosses have a shoulder-locking mechanism that allows the birds to hold their wings in a horizontal position optimal for soaring with minimal muscular effort (Pennycuick 1982; Meyers and Stakebake 2005), it is unclear whether this mechanism is sufficient to hold the wings in a position that sustains flight during REM sleep. As a result, the relationship between day activity and nocturnal roosting may reflect the need to recover the components of sleep that are either not possible or occur in a less effective manner in flight. In summary, although albatrosses might obtain some sleep in flight, particularly during periods of rapid, long-distance flights in the nonbreeding season (Croxall et al. 2005), they also seem to accrue a sleep deficit during flight that is replenished during sleep while floating on the water at night.
Sooty (“wideawake”) tern
The sooty tern’s (Sterna fuscata) vernacular name, wideawake tern, is attributed to their tendency to fly over the breeding colony at night while emitting a call that sounds similar to “Wide-a-wake” (Schreiber et al. 2002). Whether the vernacular name accurately describes their behavior while at sea remains unclear, however. As early as 1891, Taylor speculated over whether and where sooty terns rest outside the breeding season (Scott 1891). Ashmole (1963) later noted that although sooty terns remain at sea around their breeding colony on Ascension Island during the 2-month nonbreeding season, they never landed at the colony, even though the second closest landmass was 1,100 km away. Based on the absence of sooty terns on their breeding colony outside of the breeding season and the birds’ aversion to landing on the water, Ashmole (1963) suggested that they sleep on the wing. Although the amount of time spent away from land has not been measured directly, banding studies show that juvenile sooty terns engage in long transoceanic flights. Birds hatched on the Dry Tortugas, off the southern coast of Florida, USA, migrate southeast along the northeast coast of South America and then turn east and fly to the west coast of Africa, a transatlantic flight of 4,000 km (Robertson 1969).
In addition to the evidence for transoceanic flight, there is also evidence to support the contention that sooty terns rarely land on the water. In most water birds, preen oil from the uropygial gland provides water proofing for the feathers. In sooty terns, however, the uropygial gland produces preen oil with less lipid content than observed in terns that frequently land on the water (Johnston 1979). The low lipid content probably accounts for the fact that sooty terns become waterlogged rapidly when placed in water (Watson and Lashley 1915). Moreover, wet sooty terns experience difficulty in taking off from the surface (Mahoney 1984), a finding that probably accounts for their reluctance to land on the water. The aversion to landing on water during apparent extended periods of time at sea suggests that the sooty tern either sleeps on the wing or dispenses with it altogether.
Postflight recovery sleep
Birds that engage in prolonged flights may require a period of postflight recovery sleep, even if sleep occurs in flight. Although REM sleep seems less likely to occur during flight than SWS, the quantity and quality of SWS may also be reduced by the physical act of flying and frequent brief arousals needed to navigate. As a result, birds may need to recover lost REM sleep and SWS after flights. Obviously, if birds do not sleep at all in flight, periods of recovery sleep may be required after extended flights. It is even possible that the need for sleep accrues more rapidly during flight, if brain activation, or use, during wakefulness in flight is greater than that during wakefulness on land (Shaffery et al. 1985; Vyazovskiy et al. 2004; Huber et al. 2004). Elevations in body and brain temperature occurring during flapping flight (Bernstein et al. 1979; Adams et al. 1999) could also increase the rate at which a sleep deficit accumulates, as observed for SWS in mammals (Horne and Reid 1985; Bunnell et al. 1988; Horne and Moore 1985; Morairty et al. 1993; Gao et al. 1995; reviewed in McGinty and Symusiak 1990). Baring adaptations to permit SWS and REM sleep to occur during flight in a manner comparable to sedentary sleep or to circumvent the need for sleep during flight and postflight recovery sleep altogether, birds may be partially or totally sleep deprived during long flights, and therefore, need to recover lost sleep upon returning to land. As discussed for albatrosses, the correlation between day flight time and night roosting on the water is consistent with the notion that at least some birds accrue a sleep deficit during flight (Weimerskirch and Guionnet 2002).
Periods of undisturbed recovery sleep may be particularly important for a bird’s future performance during wakefulness and overall fitness. Although controversy persists over the specific function(s) of sleep (Vertes and Siegel 2005), most research indicates that sleep is critical for maintaining adaptive brain function during wakefulness (Rechtschaffen 1998; Krueger and Obál 2002; Van Dongen et al. 2003; Stickgold and Walker 2005; Siegel 2005; Tononi and Cirelli 2006). For example, converging lines of evidence indicate that one function of sleep is to consolidate and enhance memories acquired during the preceding period of wakefulness (reviewed in Stickgold and Walker 2005). Periods of postflight recovery sleep may, therefore, be needed to process information acquired during the previous flight. Disturbances during this period may either lengthen the amount of time needed to recover lost sleep and process new memories or irreversibly interfere with the processing of information acquired during the previous flight, thereby, disrupting future performance during wakefulness. Sleep disruption may also affect performance by interfering with other sleep-related functions that sustain adaptive brain function during wakefulness and by reducing a bird’s level of alertness and ability to sustain attention during wakefulness (Van Dongen et al. 2003).
Disruption of other proposed, or as yet unknown, functions occurring in sleep might also adversely affect fitness. For example, sleep seems to play an important role in the prevention of and recovery from infection in mammals (Majde and Krueger 2005). Thus, disruption of postflight recovery sleep may lead to an increased chance of infection and death and the transmission of pathogens. Birds may be most vulnerable to this effect of sleep disruption during migration when both their sleep deficit and exposure to pathogens may be highest, the later due to their tendency to congregate during migration. Understanding the sleep patterns of birds during and after flight and the impact of sleep disruption during recovery sleep on subsequent waking performance and health might have implications for the conservation of birds and other animals in which periods of undisturbed sleep are essential to fitness (see Wikelski and Cooke 2006).
From a behavioral ecology perspective, if flight does, in fact, increase the need for sleep, then birds may weigh the benefits of flight against the costs paid in increased sleep when determining whether to engage in long flights. Periods of intense recovery sleep resulting from long flights might be particularly costly because they reduce time available for adaptive waking behaviors, including the detection of threats (Lima et al. 2005). A study of herring gulls (Larus argentatus) demonstrates the potential tradeoff between the benefits of flight and the cost paid in increased sleep (Shaffery et al. 1985). Herring gulls provided with food at their nest remained sedentary and slept less than non-fed conspecifics that had to fly between foraging sites and the nest. Increased brain use during flight and foraging or flight-induced elevations in brain temperature, may account for the greater need for sleep in the gulls that were not fed. Interestingly, non-fed gulls allowed intruders to remain on their territory longer than fed gulls, a potential cost of the flight-induced increase in time spent sleeping. Based on these findings, factors that increase flight time, such as increased foraging time due to decreased prey abundance, may cause a bird to abandon a particular area not only because the energetic costs of foraging are high, but also because the cost of sleeping longer or deeper become too high. The number of birds in such an area may also decline due to increased predation on birds that are sleeping longer or more deeply.
Additional scenarios can be readily envisioned wherein the costs of flight paid in increased sleep might affect the flight behavior of birds. For example, given the potential relationship between flight-induced elevations in brain temperature and sleep, birds might attempt to minimize the flight-induced increase in sleep by choosing to fly at times and altitudes that reduce brain heating and subsequent sleep. In the case of migrating passerines, in conjunction with other factors, a relationship between air temperature, brain temperature, and sleep may have selected for flying at night when air temperatures are lowest.
Do birds sleep in flight? The unequivocal answer is we simply do not know. While several lines of evidence indicate that certain species spend long periods of time flying, and therefore, might need to sleep in flight, the behavioral and electrophysiological parameters that define sleep have not been measured in these birds. Simply demonstrating periods of nonstop flight that exceed what is considered to be the normal capacity to remain awake is not sufficient to demonstrate sleep in flight. As recently shown, songbirds, such as the white-crowned sparrow, have the capacity to function adaptively on limited sleep during migration (Rattenborg et al. 2004). Consequently, birds that engage in even longer flights, lasting several days, may employ a similar adaptation to postpone or dispense with sleep. My intent here is not to propose that birds do not sleep in flight, but simply to dispel the notion that they must sleep in flight.
Obviously, the greatest obstacle to determining whether birds sleep in flight is the difficulty inherent in measuring sleep in a flying animal. Direct measures of brain activity are needed to determine whether sleep occurs in flight and the type of sleep employed. Until recently, EEG recording devices were too large for a bird to carry in flight. However, Vyssotski et al. (2006) recently developed and successfully deployed an EEG and a global positioning system (GPS) logger on pigeons during short, daytime homing flights to their loft. While in flight, the EEG showed bihemispheric activation indicative of wakefulness. Because pigeons usually fly during the day and sleep at night, they were not expected to sleep during these short daytime flights. Nevertheless, this study demonstrates the feasibility of recording sleep-related EEG activity from free-flying birds.
Recordings of sleep-related EEG activity in combination with behavioral signs of sleep (e.g., eye closure and an elevated arousal threshold) would provide the most conclusive evidence for sleep in flight. Direct observations of flying birds can be made in wind tunnels, such as those specifically designed for birds (Pennycuick et al. 1997), at the University of Lund (Sweden), or the Max Planck Institute for Ornithology—Seewiesen (Germany). In contrast, observations of birds flying in the wild may be particularly difficult to obtain. Although a small video camera and logger mounted on the back and aimed at the head might capture sleep-related behavior in free-flying birds (Takahashi et al. 2004), the utility of this approach is limited by the low temporal resolution of current devices.
REM sleep may be the least likely and hardest to detect brain state in flying birds. Even in sedentary birds, determining whether brief periods of EEG activation reflect REM sleep or wakefulness depends on behavioral observations of reduced muscle tone (e.g., head dropping) because avian EMG recordings rarely show clear reductions in activity during REM sleep similar to that observed in mammals. Even with direct observations of birds in a wind tunnel or video recordings of free-flying birds, however, it may be virtually impossible to detect REM sleep if birds are able to maintain muscle tone, and therefore, head position, during REM sleep in flight (see above). A similar situation may contribute to the inability to detect unequivocal REM sleep in cetaceans (Mukhametov 1985; Lyamin et al. 2002, 2004).
In addition to the EEG, other measures of brain activity may reveal a bird’s brain state in flight. For example, in mammals (Cirelli et al. 2004), fruit flies (Cirelli et al. 2005), and presumably other animals, a large proportion of the genes expressed in the brain are expressed as a function of behavioral state. Consequently, the gene expression profile measured after a flight may be used to determine brain state during flight (see Shimizu et al. 2004), much in the same way that doubly labelled water is used to measure energy expenditure in passerines engaged in nocturnal migratory flights in the wild (Wikelski et al. 2003). Profiles similar to those observed during sedentary sleep would suggest that sleep occurred during the previous flight. Alternatively, high levels of genes expressed in wakefulness would indicate an absence of sleep in flight. In the later case, the elevated expression of specific wakefulness genes or the expression of novel genes not normally observed during wakefulness or sleep may provide clues to the biomolecular pathways employed to dispense with or delay sleep. Ultimately, an understanding of how birds achieve such theoretical periods of sleeplessness promises to provide insight into the function of sleep.
The notion that some birds fly nonstop for several days, and therefore, must sleep in flight, is a long-held belief in the popular culture. Until recently, it seemed reasonable to assume that such birds had to sleep in flight. Indeed, some types of sleep (USWS and even BSWS), but not all (REM sleep), seem compatible with flight. However, the recent discovery that white-crowned sparrows dispense with a large proportion of their sleep during nightly migratory flights raises the possibility that birds which engage in even longer flights employ a similar, albeit more extreme, adaptation to postpone or dispense with sleep while in flight. The use of new miniaturized data loggers to record the EEG from birds flying in a wind tunnel or in the wild will reveal whether sleep occurs in flight. Determining if and how birds sleep in flight will contribute to our understanding of a largely unexplored aspect of avian behavior and may provide insight into the function of sleep.
I am grateful to the late Ebo Gwinner who inspired the work on sleep and migration. I also thank Dolores Martinez-Gonzalez and Martin Wikelski for their thoughtful comments on the manuscript and Theo Weber for assistance with the figures. The Max Planck Institute for Ornithology—Seewiesen supported this work.