The electrophysiological correlates of wakefulness and sleep we found in the white-crowned sparrow are similar to those described in other avian species [8
]. The amount of REM sleep we report here, however, is lower than previously reported in the white-crowned sparrow [25
]. One possible explanation for this discrepancy is the different criteria used to score REM sleep in the 2 studies. In the previous report vigilance state was sampled across the 24-hour period by scoring the first 4 seconds of each minute, whereas in the current study every 4-sec epoch was scored consecutively across the 24-hour period.
Interhemispheric EEG asymmetries have been described in a number of avian species [40
], but we were unable to detect them in the white-crowned sparrow. This may reflect a genuine species difference, or may be due to technical limitations in our experimental setup. There are relatively small differences in the amplitude of the EEG signal between wakefulness and SWS in the bird, and visual scoring may fail to detect all but the most extreme examples of interhemispheric asymmetries. Another possible explanation is that unihemispheric sleep occurs mainly in some environmental conditions, for instance when birds are under the risk of predation [41
], and may have little immediate adaptive value under standard housing conditions such as those used in our experiments.
The analysis of the relative power spectra in baseline sleep revealed that the EEG power in the SWA range (0.5–4 Hz) was significantly higher in NREM sleep relative to all other behavioral states, as expected. We also observed a marked decline in SWA over the course of the night, with EEG power in the 0.5–4 Hz range being highest in the first 2-h interval after sleep onset and declining steadily afterwards. During drowsiness, the EEG power in the SWA range was significantly higher relative to waking, and significantly lower relative to SWS, thus providing additional evidence that drowsiness can be considered a distinct state. Moreover, SWA in drowsiness also showed a declining trend across the nocturnal period similar to that seen in SWS, suggesting that in both states SWA is homeostatically regulated.
Following 6 hours of sleep deprivation, neither the amount nor the temporal distribution of SWS or drowsiness changed during the 12-hour recovery sleep period relative to baseline sleep. This result is consistent with the effects observed following short-term sleep deprivation in mammals, where often only markers of sleep intensity such as SWA change during recovery sleep, while NREM sleep duration may not increase. REM sleep duration also did not change during recovery sleep, but we observed a significant decrease in the latency to REM sleep in the first 2-hour interval of recovery sleep. In the rat, sleep deprivation of 3, 6 or 12 hours resulted in a small REM rebound [43
], and in the mouse a prominent REM rebound was evident after deprivations as short as 6 hours [7
]. In the pigeon, both 24 hours and long-term sleep deprivation (several days) resulted in a pronounced REM rebound, suggesting that REM is homeostatically regulated in the bird [31
]. Thus, the lack of REM rebound in our study is likely a consequence of the short duration of the deprivation period, rather than a reflection of a lack of homeostatic regulation of REM sleep in this species.
The latency to the first episode of drowsiness decreased significantly during the recovery sleep period, and a trend toward a decrease in the latency to SWS was also observed. The decreased latency to drowsiness is perhaps not surprising given that drowsiness appears to represent a form of vigilant sleep and undoubtedly fulfills some of the requirements of sleep. Although sleep pressure may have been higher following the deprivation, the stress of having a predator (human) in the room for six hours may have induced a conflict between the need for continued vigilance and the need for sleep, and led to an immediate attempt to reduce sleep pressure through means other than frank sleep. Indeed, a previous report of migratory sleeplessness in the white-crowned sparrow indicated that although NREM sleep was reduced considerably during the migratory period, drowsiness was increased significantly, suggesting that drowsiness is capable of dissipating at least some level of sleep need [25
Spectral analysis of the EEG following sleep deprivation also revealed changes in the SWA consistent with results described in mammals. Specifically, relative to baseline, SWA during recovery sleep reached significantly higher values in both SWS and NREM sleep in the first hour after sleep onset. SWA in drowsiness also significantly increased in the first 20 to 40 minutes after sleep onset. Finally, peak SWA in both SWS and NREM sleep during the first 2 hours after sleep onset reached significantly higher levels during recovery relative to baseline, indicating that at the beginning of baseline night SWA values were still below saturation. Interestingly, as in mammals, the first 2 hours of recovery sleep in sparrows were characterized by large increases in EEG power not only in the low frequency range, but also across most frequencies in the 10 to 20 Hz range. These changes in the high frequency range declined in the second 2-hour interval and disappeared entirely by the sixth hour of recovery sleep, and their origin and physiological significance remain unclear. It has recently been suggested, however, that prolonged wakefulness results in stronger cortico-cortical connections, which in turn would drive greater neuronal synchronization during sleep, leading to increased EEG power across all frequencies [17
]. In sparrows, the increase in EEG power in the spindle frequency range (12–16 Hz) is especially intriguing, because spindles were not detected by visual scoring. In mammals, spindles and SWA reflect different but related aspects of cortical bistability, because they are thought to reflect the up states and the down states of the slow oscillation, respectively [46
]. It is currently unknown whether this is also true in sparrows.
In mammals, sleep deprivation is followed not only by an increase in SWA, but also by an increase in the duration of sleep episodes and a decrease in the number of brief awakenings [37
]. Human studies show that SWA during NREM sleep is mainly determined by the prior history of sleep and wakefulness and not by circadian mechanisms [48
]. The evidence for the other markers of sleep consolidation is more limited, but suggests instead that both homeostatic and circadian factors are important [50
]. As shown in Figure , when the first hour of recovery sleep was compared to the first hour of baseline sleep we found trends similar to those described in the mammalian literature (longer sleep episodes, fewer brief awakenings after sleep deprivation). However, when these behavioral measures were compared at the same time of day, changes were either no longer present (duration of NREM sleep episodes), or in the opposite direction (the number of brief awakenings increased after sleep deprivation). Thus, the current data show clear evidence for a post-sleep deprivation increase in sleep intensity as measured by SWA, while the evidence using other markers of sleep consolidation is mixed.
A previous report in the pigeon failed to identify increases in SWA following 24 hours of sleep deprivation [31
], although a significant decline in waking duration following sleep deprivation was reported, suggesting that sleep pressure was indeed higher after the deprivation. As mentioned before, increased sleep pressure following deprivation may be discharged as an increase in sleep duration and/or intensity, and the duration of the deprivation may affect one parameter more than another. Thus, it is possible that the lack of a SWA rebound in the pigeon was due to the relatively long sleep deprivation procedure used in that study (24 hours). In rats, if wakefulness is enforced more than 12–24 hours, a significant amount of NREM may "leak" into the sleep deprivation period [51
]. This raises the possibility that the pigeon was able to obtain significant amounts of SWA during the 24-hour deprivation, either through drowsiness or through local sleep within select neuronal populations. Furthermore, is worth noting that in our study the most significant effects of sleep loss on SWA were seen when we looked specifically at peak SWA or using high temporal resolution (10-min bins), which was not done in the pigeon study [31
]. Further support for the idea of SWA leakage during prolonged sleep loss is provided by an early study of sleep deprivation using constant light in the pigeon [52
]. The authors reported a near elimination of sleep for more than 10 days, with no subsequent increases in either total sleep time or SWA during recovery sleep. Based on these data, they concluded that pigeons do not show a mammalian-like compensatory rebound in SWA following sleep deprivation. However, the overall SWA was actually preserved across the entire sleep deprivation period in constant light, which may provide a reasonable explanation of why no rebound in SWA occurred during recovery sleep. Sleep restriction paradigms that preserve the overall 24-hour SWA amount are similarly not followed by SWA rebound [53
]. Thus, SWA rebound is observed during recovery sleep after short but not after long sleep deprivation most likely because only the latter forces SWA to intrude into waking during the sleep deprivation period. In agreement with this interpretation, a recent report of recovery sleep in pigeons following short-term sleep deprivation (8 hours) shows that, like in sparrows, SWA does increase during recovery sleep relative to baseline sleep [54
During the migratory season, captive white-crowned sparrows reduced nocturnal sleep by ~60%, with no evidence of increased SWA in the small amount of nocturnal sleep they do obtain [25
]. Remarkably, despite this sleep loss, migrating birds remained capable of engaging in adaptive waking behaviors [25
]. The preservation of performance and motivation during the migratory period suggests that migrants may have evolved adaptations to combat the negative consequences of sleep loss, or that natural selection may have promoted alternative mechanisms for sleep-loss compensation. Interestingly, during the migratory period, a significant increase in daytime drowsiness, as well as a trend toward an increase in nocturnal drowsiness was reported [25
]. Although spectral analysis of nocturnal sleep was performed, no quantitative analysis of drowsiness was possible. The present findings suggest that short episodes of drowsiness are associated with a significant amount of EEG activity in the SWA range. Thus, it is possible that daytime drowsiness offers sufficient recuperative opportunities for the migratory white-crowned sparrow, while at the same time only marginally increasing the risk of predation. Interestingly, mammals that exhibit large amounts of drowsiness, such as ungulates, typically show reduced quantities of SWS, lending support to the idea that drowsiness can partially compensate for SWS [55
]. A quantitative analysis of the EEG during migration will help address this issue definitively.