We characterized sleep and operant behavior across a full year using species-location-specific exposure to a changing photoperiod. In Figure , a double plotted actogram shows activity for a representative bird over the course of approximately 30 months, including 2005 when the sleep data for this paper were collected. Colored horizontal lines indicate the calendar weeks from which experimental birds were selected. The changing duration of the photoperiod across seasons is indicated by the vertical yellow lines. Daytime activity is shown within the yellow lines, and nighttime activity outside these lines. During winter and summer, activity is completely confined to the light period (area within the yellow lines). In contrast, during spring and fall migratory restlessness, activity extends into the nocturnal period.
Four behavioral states were reliably distinguished based on visual inspection of the EEG and behavioral analysis: wakefulness, drowsiness, SWS and REM sleep. Although the placement of the EEG electrodes was constructed to detect interhemispheric EEG asymmetries, no episodes of unihemispheric sleep were observed. Given the relatively small differences in EEG activity between wakefulness and SWS in the bird, visual scoring of the EEG alone often results in a failure to detect all but the most extreme examples of interhemispheric asymmetries. Furthermore, despite use of several cameras, we could not observe eye closure state 100% of the time, depending on the location of the bird in the cage. Thus, it cannot be ruled out that our failure to detect such asymmetries could be due to technical limitations. However, if any unihemispheric sleep occurred, its duration was likely minimal. Another possible explanation for the absence of unihemispheric sleep in the present study is that this behavior was not expressed because it had no immediate adaptive value under our housing conditions.
Table shows average percentage (mean ± SEM) of 24-hour recording time spent in each behavioral state across season (top), as well as the number of hours birds spent in each vigilance state and the corresponding length of daylight (bottom). In the lower panel REM has been added to SWS and labeled as "sleep", since REM sleep amounts were minimal throughout. Consistent with our previous findings, overall sleep time was reduced in the two migratory periods combined (fall and spring) when compared to the two non-migratory periods combined (summer and winter) (p < 0.05, Tukey HSD). However, sleep duration in winter and summer are markedly influenced by the duration of the nocturnal period (total daily sleep amount in winter is significantly greater than in the summer), (p < 0.05, Tukey HSD); the longer daylight period in the summer results in sleep duration comparable to that observed in both migratory periods. Figure describes the time course of sleep, wakefulness and drowsiness across the day for each season during the dark and light periods. During the two non-migratory periods, sleep and wakefulness are organized by the light/dark cycle; sleep is largely confined to the dark and wakefulness is preferentially expressed in the light. However, sleep in non-migratory summer is consistently initiated shortly before lights-off, whereas in non-migratory winter, sleep is initiated shortly after lights-off. In the two migratory periods, however, the organizing effect of the light/dark cycle on the distribution of sleep and wakefulness is no longer as evident. Specifically, as shown in Figure , during the two migratory seasons, sleep is significantly reduced in the dark period (p < 0.05, Tukey HSD) and significantly increased in the light (p < 0.05, Tukey HSD) relative to sleep in the two non-migratory seasons. Similarly, wakefulness is significantly increased in the dark (p < 0.05, Tukey HSD), and significantly decreased in the light (p < 0.05, Tukey HSD) during the two migratory seasons. Figure highlights the change in sleep-wakefulness expression from light to dark in the non-migratory periods, and the relative lack of change in the migratory periods. In the two non-migratory seasons, as expected, sleep is significantly more likely to occur in the dark than it is to occur in the light (p < 0.05, Tukey HSD) and wakefulness is significantly more likely to occur in the light than the dark (P < 0.05, Tukey HSD). In contrast, during the migratory seasons, the percent of time spent in sleep or the percent of time spent awake does not change as a function of lighting condition. Waking behavior also differed qualitatively in light vs. dark periods during migration, as previously described [1
]. Birds at night engaged in behaviors typical of migratory restlessness (e.g., wing whirring and beak up flight simulation); these behaviors did not occur during the daytime during migration or at any time during the non-migratory days analyzed.
Overall sleep time in the White-crowned sparrow was reduced in migratory periods and in periods of longer daylight
Figure 2 Time course of each vigilance state, wake (red), sleep (purple) and drowsiness (purple), in each of the four seasons. Values plotted as mean number of minutes (± SEM,) of each state in 1-hour intervals. Time 0 represents lights on. Note that the (more ...)
Figure 3 States of vigilance as a function of light and dark. Figure 3A. Box-and-whisker plots comparing percentage of time spent in each vigilance state: Drowsy, Sleep (SWS+REM) and Wake across the seasons, plotted separately for light (yellow background) and (more ...)
Figure shows the relative stability of sleep (i.e., the ratio of sleep bout length to waking bout length) and demonstrates the extent to which the propensity of sleep maintenance was, or was not, preferentially confined to a discrete period within the 24-hours in each season. During the spring and fall migratory conditions, there is no period in the day during which the mean sleep bout duration exceeded the waking bout duration. Furthermore, the relationship between the likelihood of extending a sleep bout and the likelihood of extending a wake bout was biphasic in the non-migratory seasons and generally less organized during the migratory seasons. Note that the smoothed curves in Figure do obscure a brief rise in the relative stability of sleep at about hour 20-21 in the spring and fall. In contrast, in non-migratory winter and summer, there is a clear, consolidated period of sleep, defined as the time during which the smoothed average sleep stability was greater than 1.0. The duration of these periods is markedly different in each non-migratory season. Specifically, in summer, the typical sleep period begins approximately 80 minutes before lights off and lasts for 5 hours. The initiation of sleep before lights off in summer is likely a consequence of the extremely short nocturnal period in this season (2.5 hours). In winter, the sleep period begins approximately 60 minutes after lights off and continues for 11 hours. Daily sleep to wake transitions by season (sleep-drowsy-wake transition treated as a single transition from sleep to wake) did not significantly differ between any seasons or pair of seasons. Numbers of awakenings averaged 271 (± 41) in winter, 205 (± 25) in spring, 193 (± 24) in summer, and 388 (± 55) in fall. Although fall showed the largest number of awakenings, there was not a significant difference between fall and any of the other seasons (nor between any other pair of seasons) on this dimension. The order (fall, winter, spring, summer) by number of awakenings does not speak strongly to hypotheses about the effect of migratory state or season.
Figure 4 Sleep consolidation index for all seasons. Hourly average ratio of sleep bout length to wake bout length was calculated for birds during each season. The blue line represents a smoothed curve through these data. The time during which the smoothed value (more ...)
In DRL procedures, the extent to which an animal makes a premature response, and consequently reduces the number of rewards obtained, putatively reflects impulsivity [11
]. Figure shows box plots for response rate (the ratio of the number of responses to the number of minutes the key light was on; top), and the behavioral inhibition ratio (the ratio of the number of reinforcers to the number of responses; bottom), during each of the migratory and non-migratory seasons. A bird performing the task perfectly should achieve a behavioral inhibition ratio of 1; the lower the ratio, the greater the failure to inhibit behavior. Data for individual subjects were averaged across sessions for each week (or range of weeks) shown in Figure . In general, birds produced lower rates of responding during non-migratory winter and summer and higher rates during migratory spring and fall. In non-migratory periods, average response rates fall within the expected range of 3 responses per minute. During fall and spring migratory seasons, however, average response rates are significantly elevated (t(14) = 3.94, P = 0.001) relative to the two non-migratory seasons. As a result, the behavioral inhibition ratio was significantly lower in the two migratory seasons (t(14) = -6.61, P < 0.001) relative to the two non-migratory seasons. Despite having previously learned to execute the 20 second delay, during the migratory fall and spring this ability was impaired.
Figure 5 Box-and-whisker plots illustrating DRL performance during each season. Top shows response rate (responses/minute) and bottom shows the behavioral inhibition ratio (reinforcers/responses). Horizontal lines within boxes denote medians (50th percentile); (more ...)
Figures , , further characterize changes in operant responding across seasons and more finely illustrate how these changes contribute to the overall effect on the DRL seen in Figure . Figure contains a set of cumulative, single-session records for an individual bird across four seasons and illustrates the changes in response patterns across the year. Although the individual bird represented in Figure exhibited the most extreme seasonal changes in responding, response records produced by other birds followed the same general seasonal pattern. The slope of the graph is an indicator of the rate of responding, with the pen resetting to zero for each 100 responses emitted during the session. The diagonal hash marks indicate when the reinforcers were acquired. During the winter session, rates of responding were low and 17 reinforcers were earned in the 30 min session. The pattern is similar during the summer session, although the response rate and number of reinforcers were slightly higher. Both of these records show steady, paced responding typical for the DRL schedule. During the spring session, however, response rate was extremely high; there were more than 700 responses emitted without a single reinforcer obtained in the 30-minute session. The fall session record is similar to spring although the rate is about half that of spring, with only 2 reinforcers obtained.
Figure 6 Cumulative records for an individual bird from a single session during each season. Diagonal lines extending directly beneath the record (hash marks) indicate food reinforcement. Note there are no reinforcements earned during the spring season. Vertical (more ...)
Figure 7 Performance within sessions on the Differential Reinforcement of Low Rate (DRL) task. Mean ± SEM (n = 15) of number of responses (left axis, dotted lines) and number of reinforcers (right axis, solid line) are plotted for each 3 minute period (more ...)
Figure 8 Relative frequency distribution of interresponse times (IRTs) using a bin size of 2 seconds averaged across birds in each season. Dashed line indicates the time at which IRTs were reinforced. Proportion of reinforced IRTs, those which occur after the (more ...)
Figure summarizes how the entire group of DRL birds performed within the 30-minute sessions during each season. Early session response rates are appropriately low during the non-migratory seasons of winter and summer. In the two migratory periods, response rates are higher. Regardless of season, responding decreased following the first 3-minute epoch in the session, particularly during spring, summer and fall, and this pattern of change did not differ throughout the week. During winter, summer and fall, this decreasing trend in response rate did not produce a concomitant decrease in reinforcement. However, during spring, the within session response decay actually led to a linear increase (p < 0.05) in reinforcement across the session; performance thus "improved" as response rates declined. Despite the "improved" responding during the end of each spring session, performance at the beginning of the subsequent session returned to previous, early-session levels.
Issues in premature responding on the DRL can be further qualified by examining the inter-response time intervals (IRTs) distribution, an analysis of the time intervals a bird waited before emitting a response. Figure shows the relative frequency distribution of IRTs using a bin size of 2 seconds averaged across birds in each season. The proportion of reinforced IRTs, those which occur after the 20 second delay, is shown to the right of the dashed line. In a DRL 20-s schedule, the highest proportion of responses should ideally be in IRT intervals > 20 s, whereas a high proportion of responses in IRT intervals < 20 s indicates premature responding. On average, the smallest proportion of reinforced IRTs occurred during spring, followed by fall, then summer; the greatest proportion of reinforced IRTs occurred in winter. IRT distributions for DRL schedules were bimodal with the modes at the shortest (non-reinforced) and largest (reinforced) bins. Although the temporal placement of the modes differs subtly in each season, there was no definitive trend toward short duration IRTs in the two migratory seasons relative to the non-migratory seasons.
Figure shows the effects of sleep deprivation on performance data by season. The expected migratory increase in response rate is evident for spring but not fall, consistent with the fact that the birds were no longer actively migrating. Sleep-deprived birds tested on the DRL in fall were selected from weeks that did not include those that were used to assess general seasonal performance on the DRL. Regardless of season, the sleep deprivation manipulation affected response rate across conditions (p < 0.05), as well as the number of reinforcers obtained across conditions (p < 0.05). During the sleep deprivation period, response rate was significantly lower relative to pre sleep deprivation baseline during both winter and spring (post-hoc t, familywise p < 0.05). In winter, this resulted in a significant reduction in the number of reinforcers obtained (post-hoc t, familywise p < 0.05). However, in spring, the reduction in response rate produced a significant increase in reinforcers (post-hoc t, familywise p < 0.05). As a result, spring was the only season during which the sleep deprivation experiment produced a significant effect on the behavioral inhibition ratio (post-hoc t, familywise p < 0.05): In this case, we observed a significant increase in inhibition during the actual sleep deprivation relative to baseline (post-hoc t, familywise p < 0.05). For all dependent measures and seasons, recovery levels and post recovery levels did not differ statistically from baseline or from each other.
Figure 9 48 hour sleep deprivation with the DRL birds. Top shows response rate (responses/minute), center shows number of reinforcers acquired, and bottom shows the behavioral inhibition ratio (reinforcers/responses). Data shown are organized by season, Sleep (more ...)