can birds fly while sleeping

A common myth once held that albatrosses could fly for years at a time, eating and drinking and mating on the wing, landing only to lay their eggs. Modern science does not support this old wives’ tale, but the verifiable truth about avian flight behavior is almost as impressive. The Gray-headed Albatross can circle the globe in only 46 days, making numerous pit stops along the way. And rather than the albatross, it’s the Alpine Swift that holds the record for the longest recorded uninterrupted flight by a bird: One logged more than 200 days in the air as it hunted flying insects on its wintering range in the skies over West Africa.

These legendary flights raise a flurry of questions about how the birds pull off such feats, and chief among them is the question of sleep. For many years, scientists conjectured that long-ranging birds could sleep while aloft, despite having no real evidence to support this claim. Until now, that is. A new study about the Great Frigatebird, published earlier this month in Nature Communications, supports the conventional wisdom—but in a surprising way.

The Great Frigatebird might not have the incredible range of the Alpine Swift, but its aerial feats are astonishing in their own right. On their wandering flights, frigatebirds can stay aloft for up to two months without touching down on land or water. More importantly, while out at sea, they couldnt even take a break even if they wanted to; unlike most other seabirds, frigatebirds cant swim, becoming waterlogged and eventually drowning if they do encounter water. Its this inability to stop and get some rest while floating that has caused scientists to suspect the bird might sleep while flying, and its why Niels Rattenborg of Germanys Max Planck Institute for Ornithology (and other colleagues) chose to study their sleep patterns.

Rattenborg was also drawn to frigatebirds for logistical reasons. One nesting population of the species in the Galapagos Islands is “quite tame” after years of constant observation, he says. Rattenborg and his team found it relatively easy to capture 15 of the birds to implant electroencephalographs (EEGs) into their skulls. Because EEGs measure electrical activity in the brain, the researchers were able to tell when the birds were awake or asleep. An implanted accelerometer clued them into how fast and in what direction the animals flew.

When they downloaded the data from the tiny devices a week later, the researchers found that while frigatebirds do sleep while flying, they sleep very little—about 45 minutes each day in short ten-second bursts, usually after dark. By contrast, on land, the birds sleep one minute at a time throughout the day and night for a total of roughly 12 hours each day.

While sleeping mid-flight, frigatebirds don’t go completely on autopilot; the birds often sleep with only one side of their brain, leaving the other side awake. Most animals that sleep half-brained do so to stay alert for predators, but frigatebirds have no natural predators in the sky. Rattenborg suspects that they remain half-awake to prevent mid-air collisions, though none were observed during the study.

Much like hawks or eagles, frigatebirds soar by circling thermal updrafts to gain altitude before gliding straight for long distances, slowly losing altitude until it’s time to climb again. All of the sleep recorded in the study occurred during the upwards-circling portion of the flight; the birds didn’t sleep at all while gliding down. Paper co-author Alexei Vyssotski of the University of Zürich, who designed the implantable EEG/accelerometers and performed some of the bird surgeries, says that while it may be more complex, catching a thermal updraft is also the safest part of a flight. “An animal cant collide with the water surface when the altitude rises,” he says.

The discovery that birds do in fact sleep on the wing, even if only in short, infrequent bursts, confirms a long-standing scientific theory about avian biology. It also adds to the growing literature about the necessity and nature of sleep in general, even in humans. A few years ago, Rattenborg discovered that Pectoral Sandpipers can survive and even thrive for weeks with very little sleep, and his decades-old finding about Mallard Ducks half-brained sleeping patterns inspired research about why people sleep poorly in hotels and other unfamiliar places—because one side of our vigilant brain stays a little awake to keep watch.

Could humans also benefit from many short naps over long periods of time? Leonardo Da Vinci is alleged to have slept only 90 minutes a day, in short fifteen-minute bursts every four hours. Maybe he was onto something that frigatebirds already knew.

Pledge to stand with Audubon to call on elected officials to listen to science and work towards climate solutions.

Asymmetric sleep linked to circling flight

The interhemispheric asymmetry in EEG slow wave activity (SWA; 0. 75–4. 5 Hz power) varied during SWS in flight. We employed an asymmetry index [AI=(left hemisphere SWA–right hemisphere SWA)/(left hemisphere SWA right hemisphere SWA)] to classify SWS as bihemispheric (BSWS; ?0) in order to measure the frigatebirds’ use of USWS. 3<0. 3) or asymmetric (ASWS; ?0. 3?AI?0. 3), with an absolute AI?0. 6 indicating USWS28. All types of SWS occurred in flight (Fig. 2a,b). In flight, there was a greater proportion of SWS made up of ASWS (71 57±3. 96%, s. e. m. ) than on land (47. 64±2. 38%; P<0. 002, two-tailed paired Student’s t-test), and the proportion of ASWS that was made up of USWS (flight, 47 27±5. 30%; land, 24. 96±2. 26%; P<0. 003, paired two-tailed Student’s t-test). The presence of BSWS on the wing suggests that ASWS is not necessary to maintain the aerodynamic control of soaring or gliding flight and thus likely serves other functions, even though SWS was more asymmetric in flight.

During ASWS18, the eye opposite the more awake hemisphere is typically open due to the nearly complete crossing of input from the eyes, allowing birds to sleep and monitor for threats at the same time17. Accelerometer recordings in frigatebirds indicate that ASWS performs a comparable role during flight. The majority of the time during wakefulness and SWS on land, frigatebirds maintained a straight head (85 91±3. 19%, s. e. m. , and 89. 67±2. 18%, respectively, P=0. 056, paired two-tailed Student’s t-test; Fig. Head position was categorized as straight when acceleration along the sway axis fell between the dashed vertical lines in Figure 3a), as shown by the distribution of sway axis values with a cluster around zero. 3a (?0. 175g0 and 0. 175g0; standard acceleration of free fall g0=9. 80665, m s?2). Moreover, during wakefulness, acceleration along the sway axis displayed a unimodal peak at zero in flight (75 42±2. 05%, s. e. m. ). But during SWS in flight, as opposed to SWS on land, the distribution of sway axis values was tri-modal, with two additional peaks reflecting acceleration to the left and right and one peak centered around zero (Fig. 3a). The birds’ radial acceleration as they turned in either direction was the main cause of their acceleration to the left and right (wing angle, 18). 75±0. 48°, s. e. m. ; P=6. 4 × 10?15, paired two-tailed Student’s t-test; Fig. 3b), likely reflecting soaring on rising air currents13,21,26. It’s interesting to note that the birds during ASWS were more likely to circle in the direction of the side with higher SWA (ASWS-L, to left, 65). 31±5. 07%, s. e. m. , to right, 8. 76±1. 46%, P=1. 1 × 10?7; ASWS-R, to left, 9. 31±1. 81%, to right, 66. 59±6. 47%, P=1. paired two-tailed Student’s t-test, with a size of 2 × 10?6), but there was no bias for circling to one side (to the left, 35 22±4. 31%, to right, 35. 35±4. 83%, P=0. 73, paired two-tailed Student’s t-test; Fig. 3c,d). We found smaller asymmetries in gamma activity (30–80 Hz power), a frequency linked to visual attention, during SWS in addition to asymmetries in SWA29. Opposite to SWA, during SWS with asymmetric gamma (?0. 1>AI>0. 1) The birds moved quickly in the direction of the lower gamma (left gamma) 71±2. 02%, s. e. m. , to right, 65. 76±6. 21%, P=6. 5 × 10?6, paired two-tailed Student’s t-test; left gammaP=1. 2 × 10?6, paired two-tailed Student’s t-test), but there was no bias for acceleration toward one side (to the left, 37) during SWS with symmetric gamma. 50±3. 74%, to right, 31. 77±4. 20%, P=0. 96, paired two-tailed Student’s t-test; Fig. 3e). The frigatebirds had their eyes open toward the direction of the turn, based on the more awake EEG activity (lower SWA and higher gamma) in the hemisphere opposite the turn (Fig. 3b), presumably to watch where they were going.

The distribution of awake and SWS 4 s epochs, encompassing all birds together, at varying sway accelerations (0 02g0 bins) on land and in flight at night. In both the awake and SWS states, the values on land (top) were centered around zero, suggesting that the birds maintained their head posture. While the birds were flying while awake and holding their heads and wings straight, SWS, for the most part (70 57%), was characterized by circling flight to the left and right, as evidenced by the sway acceleration. 175g0 and >0. 175g0 (dashed vertical lines). (b) Diagram illustrating the head and wing angles in relation to the horizon during leftward-circling flight, as determined by the accelerometry The corresponding brain state (see below) is also shown. (c) Recording demonstrating the correlation between acceleration along the sway axis and asymmetric SWS (ASWS); during ASWS, there is increased EEG slow wave activity (SWA; 0 75–4. 5 Hz power) in the left hemisphere (ASWS-L), the sway axis displayed high values, signifying a leftward spiral, and when SWA was higher in the right hemisphere (ASWS-R), the sway axis displayed a rightward spiral. Same bird as in Fig. 2. (d) The correlation between the type of SWS in flight and sway acceleration for all birds combined The data from (a) are divided into three categories: bihemispheric SWS (BSWS), ASWS-L, and ASWS-R, as defined in the main text. (e) Link between flight mode (sway acceleration) and SWS during flight for data from (a); gamma activity (30–80 Hz power) is partitioned based on the interhemispheric asymmetry; asymmetric gamma with greater gamma in the left (AGamma-L; AI?0). 1) or right (AGamma-R; AI??0. 1) hemisphere and bihemispheric (symmetric) gamma (BGamma; ?0. 1<0. 1). (b) illustrates the general relationship between circling flight, brain state, and probable eye state18. It shows the relative difference in EEG SWA between awake (green) and sleeping (blue) hyperpallium. The green arrows are circling to the left and indicate the general direction of visual flow.

Similar content being viewed by others

Camille Testard, Sébastien Tremblay, … Michael L. Platt

Benjamin Gorko, Igor Siwanowicz, … Stephen J. Huston

Kana Okano, Jakub R. Kaczmarzyk, … Jeffrey C. Grossman

Numerous birds, such as swifts1,2,3,4, sandpipers5,6, songbirds7,8,9,10, and seabirds11,12,13, fly nonstop for several days or weeks at a time. Since most animals suffer from sleep deprivation, it is widely believed that birds satiate their daily need for sleep while flying15. But the possibility that birds skip sleep entirely during lengthy flights was raised by the recent finding that some birds can function adaptively for several weeks despite sleeping for much less time16. Therefore, evidence of extended flights does not automatically imply evidence of sleep during flight; neurophysiological recordings of the alterations in brain activity that are indicative of sleep are necessary to address this issue. Additionally, these recordings are necessary to determine the types, amounts, and intensities of sleep as well as the possible implications of flight-related sleep adaptations for our understanding of sleep’s functions. It is unknown if birds sleep on the wing because recordings of brain activity during extended flights are not available15.

In response to shifting ecological demands, birds on land can alternate between sleeping with one hemisphere at a time and sleeping with both hemispheres simultaneously17,18. Birds that experience unihemispheric slow wave sleep (USWS) maintain an open eye directed toward potential threats while maintaining a connection to the awake hemisphere. Dolphins can swim in this state and utilize USWS to observe their surroundings19. Thus, birds in flight may depend on USWS to preserve environmental awareness and aerodynamic control of the wings, all the while getting the sleep required to maintain focus during waking hours. We tested this hypothesis in great frigatebirds (Fregata minor).

Despite spending weeks or months flying over the ocean, frigatebirds are not known to rest on the water, as Darwin noted during his visit to the Galápagos Islands20.12,13, 21 After more than a brief encounter with the water, taking off is challenging due to their long wings, poorly webbed feet, and reduced feather waterproofing. Great frigatebirds depend on large predatory fish and cetaceans to push prey above the surface, including flying fish and squid, in order to capture food12. While earlier research identified probable feeding episodes—that is, slow flying close to the surface—mainly during the day12,21, under favorable circumstances, feeding may also take place at night22, as frigatebirds follow ocean eddies that indicate opportunities for foraging both during the day and night23. Because of this, frigatebirds must be awake at all times when over the ocean due to ecological requirements.

We show that frigatebirds can sleep in flight using one hemisphere at a time or both at once by observing their brain activity while they are over the ocean. While on the wing, frigatebirds sleep in both ways, but their asymmetry in sleep is greater than it is on the ground. It is possible that frigatebirds use unihemispheric sleep to monitor their direction of travel because they sleep primarily while circling in rising air currents and keep their eye directed toward the direction of flight. Even though they can sleep on the wing, flying frigatebirds sleep much less, less deeply, and for shorter periods of time than they would on land. This suggests that the ecological demands for attention that flying frigatebirds must meet are typically not met by sleeping unihemispherically. The prevailing belief that significant daily sleep is necessary to maintain adaptive performance is challenged by the capacity to maintain cognitive function on little sleep.

Sleep scoring and EEG analysis

REMLogic software (Natus Medical, Pleasanton, California)16 was used to score the amount of time spent awake, in SWS, and during REM sleep on all days during the flight that had stable EEGs. Every recording made after landing again was also scored, including the brief landings between two flights that birds 1 and 5 were shown to make (Supplementary Figs 1 and 3). One or more epochs of a particular state, separated by a single epoch of a different state, were considered to constitute a bout of that state. All scored days served as the basis for the wake, SWS, REM sleep, and total amount of time spent in each state. The night with relatively high signal quality and a lot of sleep was the focus of the EEG’s spectral analysis (see Supplementary Figs 1–8). Using the fast Fourier transform, all four s artifact-free epochs for each state were examined (0 25 Hz bins) applied to Hamming-windowed data. Fourier coefficients obtained for ranges 0 were used to estimate SWA and gamma power. 75–4. 5 and 30–80 Hz, respectively. For statistical comparisons, gamma power and SWA medians were employed. Quartiles for group medians shown in Fig. 4b, as well as Supplementary Figs 14b and 15b, are bootstrap estimates. The relationship between interhemispheric asymmetries in SWA and gamma and the mode of flight (Fig. 3d and 3e) were based on the sleep-deprived night. Furthermore, for the final night of the flight, SWS-related SWA was computed in order to identify any possible variations in sleep intensity during the flight (Supplementary Fig 15b).

The accelerometer recordings revealed two predominant patterns during flight (Fig. 2a). Flapping flight was characterized by large sinusoidal oscillations (?2. 5 Hz) in the heave and surge axes, which represent separate beats of each wing. On the other hand, during gliding and soaring flight, the three axes were mostly flat or displayed slow oscillations that were probably caused by a combination of respiratory movements and fine maneuvers (see Fig. for an expanded view of SWS). 2a). During daytime gliding and soaring, the head moved in small, frequent, and fast horizontal movements on top of these slow oscillations. Occasionally, the bird would drop and there would be a sharp decrease in acceleration along the heave axis, most likely as a result of the wings folding momentarily (Supplementary Movie 4). Last but not least, there were sporadic high-frequency activity bursts in all axes at the same time, which most likely represented preening as seen in birds flying over the colony and on the nest.

Previous studies12,13,26 and our own observations (Fig. 1d), show that frigatebirds exhibit two major flight trajectories; rising in circles (soaring) and straight gliding down. In addition to identifying flapping flight, the accelerometer was useful for discriminating circular from straight flight (Fig. 1e). During both types of flight the absolute air-referenced flight speed averaged over significant time intervals (>4 s) is constant (Fig. 1e). Thus, the tangential (co-directed with the speed vector) acceleration is zero in both flight modes. When the animal flies straight the total acceleration felt by the accelerometer is produced only by the gravity vector g (standard gravity, 1g0=9.80665, m s?2). However, during circular flight additional centripetal (radial) acceleration, ar=V2/R (V, speed; R, radius of the trajectory) is added to the acceleration of gravity: . As rotation lies approximately in the horizontal plane, the two acceleration vectors are orthogonal to each other and total acceleration, . Thus, to determine whether the trajectory is straight or not, it is sufficient to measure total acceleration, low-pass filter it to remove the influence of wing flapping and compute radial acceleration from this equation. Radial acceleration above 0.175g0 corresponded to circling flight (Fig. 3, Supplementary Fig. 16). Total acceleration in circling flight was 1.057±0.003g0 and radial acceleration was ?0.340±0.009g0 (s.e.m.; see Supplementary Table 1 for values for individual birds). The bank (wing) angle during soaring was measured as arccos(g/atot) and was 18.75±0.48° (s.e.m.). However, for our EEG analysis it was also important to know whether the bird was rotating to the right or to the left. This information was obtained by measuring radial acceleration with the accelerometer mounted on the bird’s head with one axis (that is, sway) directed laterally. Because frigatebirds keep their heads straight during both flight modes, we were able to determine radial acceleration directly from the accelerometer without additional transformations. However, to confirm this claim and to increase the accuracy of the radial acceleration measurements we also performed computations without this assumption. The accelerometer was attached to the birds head in a way such that one axis was orthogonal to the tangential plain of the bird skull and another was directed laterally. Projection of total acceleration on to the tangential plain of the bird skull clearly shows three clusters corresponding to straight and circling flight, with turning to the left and right (see data from one example bird in Supplementary Fig. 16a). To simplify this analysis, we rotated the axes of the head-fixed coordinate system to have one axis directed to the ground during straight flight; however, in the recording examples shown in Figs 2 and 3, Supplementary Fig. 13, acceleration is shown in the original axes of the accelerometer. The following analysis shows that the skull surface tangential plane deviated by 29.86±0.68° (s.e.m.) from the horizon (see Supplementary Table 1; see also Fig. 1a). As a first step we down-sampled the acceleration data to 25 Hz (from original 200 Hz) to decrease computation time. We then filtered out high frequencies by applying a low-pass finite impulse response filter (0.1 Hz; span 40 s). The input data were processed both in the forward and reverse directions and the resulting sequence had precisely zero-phase distortion and doubled filter order. Then, we computed principle components (PCs) in 3D space without mean subtraction. The first PC pointed in the direction of the gravity vector, the second—in the lateral (radial) direction, and the third—in the direction of the speed vector. Because we found that accuracy of the PCs determination can be affected by outliers, mainly due to episodes when the bird drops down with acceleration in the direction of the first PC <0.95g0, we excluded such points and recomputed the PCs again. In the horizontal plain of the second and the third PCs (Supplementary Fig. 16b), clusters corresponding to rotation to the left and right were aligned relative to the coordinate axes. The best separation was observed along the second PC corresponding to sway acceleration. The vertical lines drawn at sway accelerations ±0.175g0 reliably separate the three clusters in all birds. Because we wanted to compute rotations of the head relative to straight flight, we repeated the PC analysis, but for points representing straight flight only. Coordinates of the first PC gave the direction to the ground during straight flight. The angle between this direction and skull surface normal is the skull angle shown in Supplementary Table 1. We rotated the coordinate system a second time to have one axis in the direction of the first PC (Supplementary Fig. 17). In this head-fixed coordinate system, during circling flight, the absolute value of lateral (sway) acceleration was 0.321±0.008g0 (s.e.m.), acceleration in the direction of the flight (surge) was 0.028±0.005g0 and vertical (heave) acceleration was 1.006±0.001g0 (see Supplementary Table 1). Assuming zero tangential acceleration as before, we computed the angle of the head turn in circling flight (2.137±0.184°, s.e.m.) relative to straight flight and the direction of the axes over which the turn was performed (right–left: 0.626±0.096°, beak–tail: 0.521±0.111°, down–up: ?0.209±0.033°, signs are valid for the case when the animal turns left, but absolute values represent averaged quantities for left and right turns taken together, see Supplementary Table 1). To simplify interpretation of the head turn we computed angular deviations of the head-fixed vector pointing upwards in the lateral (right–left) and anterior–posterior (beak–tail) directions. These deviations were 1.033±0.252° and 1.444±0.233° (signs correspond to the left turn as before). As shown in the table, bank angle (wings-to-horizon) was computed with the assumption that total acceleration was orthogonal to the plane of the wings. This assumption was verified by placing accelerometers on the backs of two magnificent frigatebirds together with the GPS logger in a pilot study (Supplementary Fig. 18). In these two birds, total acceleration during circling flight was 1.053 and 1.067g0. Standard deviations of sway acceleration were 0.013 and 0.016g0, and standard deviations of surge acceleration were 0.033 and 0.036g0, respectively. Thus, the standard deviation of the total acceleration vector in the lateral direction was 0.71° and 0.85° and in the anterior–posterior direction it was 1.80° and 1.38°. Taking the 95% confident border as a more conservative estimate, we obtained 1.45° and 1.67° for sway and 3.60° and 2.81° for surge. These angles are much smaller than the angle of the wing plane to the horizon (18.32° and 20.41°). Thus, our assumption about orthogonality of the plane of the wings to total acceleration is correct.

FAQ

Can birds fall while sleeping?

How long do birds sleep for?

What happens if a bird gets tired while flying?

Do birds stand while they sleep?