Michaela Haul and Eberhard Gwinner

Max-Planck-Institut fur Verhaltensphysiologie, Von-der-Tann-Str. 7, D-82346 Andechs, Germany Abstract House sparrows (Passer domesticus) can synchronize their circadian rhythms of locomotion and feeding with times of periodic food availability In contrast to most mammals, which synchronize only a specific part of their circadian system with feeding cycles and thus express two distinct activity components, house sparrows synchronize their circadian activity rhythms as a whole with the food zeitgeber. Previous results had indicated that feeding cycles act as comparatively weak zeitgebers for house sparrows. In the present study, therefore, we investigate whether feeding schedules are weak zeitgebers in general or whether their impact on the circadian system of the birds depends on the degree of food restriction. A detailed analysis of the synchronization pattern under the different experimental conditions should help to clarify whether house sparrows use a different mechanism for food-synchronization than mammals. House sparrows were kept in continuous dim light and exposed to different feeding schedules with daily food access durations ranging from 8 to 20 h. Many birds lost synchronization and exhibited free-running rhythms in locomotor and feeding activity when the daily food access duration was lengthened but became synchronized when the feeding duration was shortened. The interpretation that short food access durations represent stronger zeitgebers than long food access durations was supported by the occurrence of large negative phase-angle differences during long daily feeding schedules, contrasting with small and sometimes positive phase-angle differences under short food access durations. There were no indications that house sparrows possess a specific foodentrainable circadian oscillator as mammals do. Rather, periodic food availability seems to be a zeitgeber for the whole circadian system, which, hence, can be synchronized both by light and food. An explanation for such different mechanisms of food-synchronization is offered in the feeding ecology of these animals. Birds may evaluate the importance of a specific feeding schedule as a zeitgeber either from temporal information on the duration of the daily food access time or from energetic considerations. The phase-angle differences associated with the different feeding schedules and the maintenance of daily activity times may ensure an appropriate temporal integration of behavior with specific conditions. Nonsynchronized birds exhibited masking-induced feeding activity, which might represent an alternative means of adjusting to feeding cycles when synchronization cannot occur.

Endogenous circadian rhythms enable organisms to synchronize behavioral and physiological processes with periodic events in their environment. Such internal timing devices provide means for using resources that are available only at specific times of the day, such as certain kinds of food or prey (reviewed in, e.g., Daan, 1981). Responses of the circadian system to daily rhythms in food presentation have been investigated in various species, ranging from invertebrates to primates. These studies revealed that most animals can synchronize-at least fractions of their circadian system-with times of food availability, the pattern of synchronization being highly species or even strain specific (for reviews, see Boulos and Terman, 1980; Aschoff, 1986; Mistlberger, 1994).

When exposed to restricted feeding schedules most of the mammals studied develop a robust and pronounced bout of running activity several hours prior to the time at which food becomes available. This synchronized activity pattern has been interpreted as anticipatory activity (reviewed in Mistlberger, 1994). In most of these studies, however, another component of activity was also present that was not synchronized by the feeding schedule but free ran with its own period. The occurrence of such dissociated activity components has been alluded to the existence of two distinct circadian oscillators that regulate behavior. Subsequent lesion experiments supported the concept that in these animals, anticipatory activity is in fact governed by a separate circadian oscillator outside the major circadian pacemaker, the suprachiasmatic nucleus (SCN; Stephan et al., 1979 a, b; Boulos et al., 1980; Stephan, 1981). Hence, in mammals the distinction could be made between locomotor activity, which is controlled by a "light-entrainable oscillator" (probably the SCN), and not synchronizable by food, and anticipatory activity, which is regulated by a "foodentrainable oscillator" (FEO).

Recent studies have shown that invertebrate species like honeybees (Apis mellifera; Frisch and Aschoff, 1987), and nonmammalian vertebrates such as pigeons (Columba livia; Abe and Sugimoto, 1987) and house sparrows (Passer domesticus; Hau and Gwinner, 1992) can also synchronize their circadian activity rhythms with periodic food availability. However, although these species may show activity prior to the time of food access, there is as yet no convincing evidence for a separate food-entrainable oscillator.

Despite the apparent differences in the circadian response to periodic food availability within the vertebrates studied so far, detailed investigations of food synchronization in nonmammalian species were lacking. When analyzing the effects of feeding cycles on vertebrate circadian systems, Rusak (1981) and Aschoff (1987) suggested that the interspecific variation in reaction to a food zeitgeber might reflect ecological adaptations of circadian systems. To follow up these ideas, it is necessary to examine which kind of feeding cycles actually act as circadian zeitgebers and to characterize the stimuli animals may use to evaluate the strength of a food zeitgeber. Such results could then provide information about the functional context of food synchronization and on the adaptation of circadian systems to different ecological needs.

In house sparrows, our previous study had suggested that food-at least in the specific experimental schedule used-acted as a rather weak zeitgeber for the circadian rhythms of locomotor and feeding activity of house sparrows (Hau and Gwinner, 1992). This conclusion was derived from (i) the finding that sparrows synchronized with feeding cycles only within a narrow range of periods (at best between 23.5 and 25.0 h) and (ii) the observation that the phase-angle differences between the onset of the birds' behavioral rhythms and the onset of the daily feeding time changed rather drastically when the period of the feeding cycle was varied. Both results are characteristic for synchronization with a weak zeitgeber (Aschoff and Pohl, 1978; Aschoff, 1981). Yet the weak zeitgeber action of the feeding cycles in our previous study might have originated from the fact that food access was hardly restricted, feeding durations ranging from 11.15 to 12.5 h per day That this schedule did not impose a severe restriction in food intake was also suggested from the behavior of the birds, which did not use the entire time of food availability

Therefore, in the present experiment we decided to investigate whether, in house sparrows, food acts as a weak zeitgeber in general or whether the zeitgeber strength of feeding cycles depends on the duration of food availability and, hence, on the degree of food restriction. For this purpose, house sparrows were held in constant dim light and exposed to feeding schedules with daily food access durations ranging from 8 to 20 h. The synchronization pattern of the birds was observed together with the phase-angle differences assumed under the different feeding schedules.

Zeitgeber stimuli, inherent in feeding cycles, could be temporal or metabolic; that is, either the duration of the daily food access time or the impact of the feeding schedule on the energy balance could be evaluated by the birds. To estimate the energy balance of the house sparrows under the feeding schedules, food intake and body mass were measured at regular intervals.

By examining the occurrence of anticipatory activity, that is, activity prior to the time of food access, we analyzed whether the birds make use of a separate FEO for synchronization with feeding cycles. Rats always exhibit anticipatory wheel-running activity, regardless of the feeding cycle and its phase with respect to their activity rhythm (e.g., Honma et al., 1983; Aschoff et al., 1983). Hence the appearance of a regular anticipatory pattern seems to be a specific feature of an FEO, and its occurrence in the present experiment could provide evidence for its existence in house sparrows.

MATERIALS AND METHODS

Animals and Housing

Nine, wild-caught, female house sparrows were individually kept in cages of 81 x 50 x 31 cm. Cages were situated in light- and sound-proof chambers equipped with fans for ventilation. Light was provided from the ceiling of the chamber from an incandescent lamp. Light intensity measured at the top of each cage was initially about 100 lux during the light phase (only during the introductory period) and 0.3 lux during the dark phase as well as during constant dim light conditions (LLdim). Food pellets (chick starter mash, BayWa) were provided from a feeder at which the birds had to push open a swing door to reach the food. At set times controlled by an automatic timer, the feeder was locked electromagnetically to prevent food access. Water was provided ad libitum. Water was renewed, food replenished, and cages cleaned twice a week at irregular times of the day, preferentially during the birds' activity time. Once a week in each experimental phase, birds were weighed (usually shortly prior to the onset of food access). Food intake was determined by weighing the feeder regularly before and after food replenishment, taking into account pellets that had been spilled.

Experimental Procedure

During the introductory phase, birds were held in a light cycle of 12-h light and 12-h dark (LD 12:12, lights on at 06.00 h) with food ad libitum. After 1 week, food was restricted (restricted feeding: RF) to 12 h coinciding with the light phase (see also Fig. 1 d). Ten days later, the light schedule was changed to LLdim, which was maintained until the end of the experiment. Birds were kept on the RF 12:12 schedule (RF 12:12/I) for another 12.5 weeks. Daily food access time was then increased to 16 h (RF 16:8) for a period of 9.5 weeks, followed by a food access time of 20 h (RF 20:4) for 5 weeks. Thereafter, feeding time was restricted again to the initial 12 h (RF 12:12/II) for 8.5 weeks, after which it was shortened to 8 h per day (RF 8:16) for 7 weeks. Finally, the birds were again exposed to RF 12:12 (III) for another 7 weeks. During the first week in RF 8:16, some birds rapidly lost weight and therefore received extra food outside the usual feeding time for a maximum of 2 days.

Activity Recordings

Locomotor activity was recorded by means of ultrasound motion detectors (Biebach et al., 1985). Feeding attempts were monitored with the help of infrared light beams in front of the feeder to enable measurement of feeding activity even when the feeder was locked. Actual feeding events, that is, movements of the swing door, were recorded by a light-beam system inside the feeder. Data were collected and stored in 2-minute bins on-line by a computer

Data Analysis

Activity recordings of individual birds were plotted monthly as actograms (Fig. 1). Data were obtained from 8 birds, of which 2 (#1, #3) provided data only for the first three or four experimental phases. To determine the period of the activity rhythms and hence the state of synchronization under a particular condition, Chi^sup 2^ periodogram analyses (Sokolove and Bushell, 1978) were performed for a 20-day period during the second half of each experimental stage. Birds were considered to be synchronized when the dominant period of their activity rhythm as revealed by periodogram analysis differed by no more than 0.1 h from the zeitgeber period of 24 h. In some cases, high levels of activity during the food access time obscured periodogram analysis; these actograms were then counterchecked by visual inspection. Birds that were not completely synchronized but modulated the period of their activity rhythms about the zeitgeber phase ("relative coordination"; Enright, 1965) were considered not to be synchronized. For synchronized birds, onset and end of the daily activity phase (a) were estimated from the actograms by three independent observers unfamiliar with the specific experimental treatment. For eachbird and experimental phase, means were calculated from these data and used in further analyses. Body masses and food consumptions are given as means of the last 2-3 measurements during each experimental stage. For the comparison of activity levels between RF conditions, the sum of individual daily activities over the last 3 weeks of each experiment was calculated. We did not obtain data on body mass during the RF 16:8 condition. Statistical analyses were performed using SPSS for Windows (SPSS, Inc., Chicago). Sample sizes for statistical tests were usually 8, except for repeatedmeasures tests (Cochran and repeated-measures ANOVA and subsequent post hoc tests), where two birds (#1 and #3) could not be included because of missing data (n = 6). Repeated-measures ANOVA (in SPSS notation MANOVA with CONTRAST = DIFFERENCE) was performed after examining the data with a Kolmogorov-Smirnov one-sample test for normality Some repeated-measures ANOVA results had to be corrected with the Huynh-Feldt epsilon for unequal variance in the covariance matrix. Post hoc planned pairwise multiple comparison tests were performed using the least significant difference method described by Sokal and Rohlf (1981). Two-tailed tests were used.

RESULTS

Synchronization Pattern

The birds changed their synchronization state with the different RF schedules (Cochran's Q-test, p = 0.01; Table 1; Fig. 1; Fig. 2). However, preceding RF conditions apparently also influenced the synchronization state of the birds during the next feeding schedule. This is suggested by the fact that different numbers of birds synchronized their activity rhythms to the three (identical) RF12 schedules.

To estimate differences in zeitgeber strength between feeding schedules more adequately, changes in synchronization state resulting from the changes in RF schedule were determined (Fig. 3). We predicted that after lengthening the food access time, the birds should lose synchronization, whereas when food access time was shortened, birds should gain synchronization. Birds that did not change their synchronization state were considered neutral in terms of our prediction. None of the birds reacted contrary to our predictions: All house sparrows lost synchrony or did not change their previous synchronization state when the daily feeding time was prolonged and became synchronized or did not change when food access time was shortened (Fig. 3).

Phase-Angle Difference During the Different RF Schedules

Onset of Activity

The phase-angle difference in onset of the locomotor activity rhythm of synchronized birds (yt) relative to the time of food availability changed with the feeding schedule (Fig. 4 a, b). As the response of the birds to the three RF 12:12 conditions was similar (Fig. 4a), only the initial RF 12:12/I condition was used for statistical analysis. The number of synchronized birds in the condition RF 20:4 was too small (n = 2) for statistical analysis. Hence differences in rH, during RF 12:12/I, RF 16:8, and RF 8:16 were analyzed with a repeated-measures ANOVA. The RF schedules significantly altered PsiH,onset, F(2, 6) = 13.38, p < 0.05, n = 4, with locomotor activity starting earliest during RF 8:16, and earlier during RF 12:12 than during RF 16:8 (see also Fig. 4b). Phase-angle differences in the onset of feeding activity with respect to the time of food availability (PsiF,onset) showed a trend similar to that for locomotor activity but were not significantly different during the different RF schedules, F(2, 6) = 3.15, p > 0.1, n = 4. Midpoint of Activity

A second measure of phase-angle differences is obtained by calculating the difference between the midpoint of zeitgeber time and midpoint of activity time. The resulting phase-angle differences are less susceptible to changes in overall activity time (Aschoff, 1965). Phase-angle differences in midpoint of activity showed a similar trend to the changes in onset of the activity rhythms, but differences between experimental stages were less pronounced and not statistically detectable, F(2, 6) = 3.99, p > 0.07, for locomotor activity, and F(2, 6) = 0.74, p > 0.5, for feeding activity.

Feeding Time

The negative phase-angle differences during some RF schedules indicate that the house sparrows did not use the whole time available for feeding. Hence we determined the time synchronized birds actually spent feeding during the different RF schedules. Actual feeding times were longest during RF 20:4, inte mediate during RF 12:12, and shortest in RF 8:16, F(2, 6) 7.38, p < 0.05 (Fig. 5a). The actual feeding time deviate from the duration of the potential feeding time offered by the RF schedule mainly during long food access times (Fig. 5b).

Body Mass and Food Intake

Mean body mass did not differ significantly b, tween RF schedules, although lowered body masses were suggested during RF 8:16, F(1.35, 6.77) = 3.10, p > 0.1, n = 6. In contrast to mean body mass, mean daily food consumption differed between RF schedules, F(3.96, 19.8) = 17.99, p < 0.001, n = 6 (Fig. 6a). Post hoc comparisons revealed that birds had increased food intake during the first two conditions (RF 12:12/I, RF 16:8), as compared to the succeeding conditions, and increased it again during RF 12:12/III. To evaluate the rate of food intake during RF schedules, food intake was divided by the actual feeding time determined above. The rate of food intake changed with RF schedules, being lowest in RF 20:4 and increasing dramatically in RF 8:16, F(2.04, 10.19) = 45.86, p < 0.0001 (Fig. 6b). Activity Time

Daily activity times (a) of synchronized birds were maintained at similar durations, and no significant differences in a between experimental stages were detectable (repeated-measures ANOVA again only with RF 12:12/I, RF 16:8, RF 8:16; locomotion: F[2, 6] = 1.27, p > 0.3; feeding: F[2, 6] = 3.06, p > 0.3, n = 4; Fig. 7).

Amount of Activity

Locomotor activity per hour of actual feeding time tended to increase toward short food access times, but there was no significant difference between RF schedules, F(1.48, 7.38) = 4.51, p > 0.05 (Fig. 8). Feeding attempts, that is, activity in front of the feeder, per hour of actual feeding time were highest under short food access times and lowest under long food access times, F(4.36, 21.78) = 7.9, p < 0.0001 (Fig. 8). Rate of swing door movements showed a similar pattern as feeding attempts and differed between RF schedules, F(3.4,17) = 7.72, p = 0.001 (Fig. 8).

DISCUSSION

Does the Zeitgeber Strength of Feeding Cycles Vary With Daily Food Access Time?

The synchronization pattern of the birds suggests that the zeitgeber action of RF cycles is dependent on the daily time of food availability Zeitgeber strength of feeding cycles was lowest during long food access times, but increased toward short food access times (Fig. 1). Support for this interpretation is not only derived from the number of birds that were found synchronized in each condition (Fig. 2) but also from the manner in which the birds changed their synchronization states between RF schedules. House sparrows either started to free run when the food access time was lengthened or resynchronized when the food access time was shortened (Fig. 3). This behavior is fully consistent with the prediction that the birds should synchronize better under strong zeitgebers (short food access times) and tend to lose synchronization with weak zeitgebers (long food access times). The findings that in some conditions the birds did not change their synchronization state ("reacted neutrally"; Fig. 3), and that two birds failed to synchronize even to the potentially strongest RF schedule with only 8 h food per day, might be explained by the circadian phenomenon that rhythms are less easily "caught" from free run by a zeitgeber than "held" in a synchronized state (Wever, 1965; Hau and Gwinner, 1992). The present results confirm our previous interpretation that a feeding schedule with 12 h food per day is not a very strong zeitgeber for the circadian system of house sparrows (Hau and Gwinner, 1992).

Additional indications for different zeitgeber strengths of the various RF schedules could be obtained from the phase angles in the midpoints of the activity rhythms assumed by synchronized birds. Under RF schedules with long food access times, birds expressed relatively large negative phase-angle differences and became active several hours after food had become available (Fig. 4). In contrast, under RF schedules with short food access times, the phases of the activity rhythms and the zeitgeber more or less coincided (Fig. 4). The occurrence of relatively large changes in phase-angle differences is compatible with the notion of feeding cycles being weaker zeitgebers, whereas the decrease in phase-angle differences can be interpreted as reflecting an increase in zeitgeber strength (Aschoff and Pohl, 1978; Aschoff, 1981).

It is interesting to note that an almost linear correlation exists between food access time and the strength of a food zeitgeber (Fig. 4b), whereas for a light zeitgeber any deviation from a 1:1 ratio in the LD phases is considered to reduce its strength (Aschoff and Pohl, 1978; Aschoff, 1981). This observation suggests that zeitgeber strength depends on specific parameters inherent in the quality of the respective zeitgeber and that circadian responses are different under the different zeitgeber modalities.

Possible Mechanisms Underlying "Anticipation* in House Sparrows

Detailed examination of the onsets of the house sparrows' activity rhythms should reveal their degree of "anticipation" under feeding cycles of supposedly different strengths, so that it can be compared to the behavior of mammals. When synchronized with feeding cycles, rats always exhibit vigorous running activity prior to the onset of the daily food access time (e.g., Stephan, 1981; Aschoff et al., 1983). In contrast, in the present experiment, house sparrows showed activity prior to the food access time only when exposed to the first 12-h feeding schedule and to the 8-h feeding schedule (Fig. 4). Additionally, such "anticipatory" behavior only occurred when the birds' phase-angle differences, as evaluated from the midpoints of activity rhythms relative to the midpoint of the feeding schedule, were minimal, that is, close to the zeitgeber phase. This indicates that activity prior to the feeding time was only shown when the synchronized circadian system had changed its phase relation to the zeitgeber, whereas in rats anticipation is independent of the light-entrainable activity and its phase-angle differences. Hence activity prior to the food access time in house sparrows appears not to reflect the action of a separate food-entrainable oscillator as is the case in mammals. Instead, it is expressed as a consequence of positive phase-angle differences of the synchronized circadian system. Under certain conditions, it may also result from the birds' tendency to maintain activity time (a) at a constant duration (Fig. 7), so that activity precedes and follows a brief food access time, for example, in RF 8:16 (see Fig. 1).

The results of the present study support our previous conclusion that synchronization with periodic food availability is achieved via a different mechanism in house sparrows than in rats (Hau and Gwinner, 1992). House sparrows synchronize their circadian system as a unit with feeding cycles, whereas rats only respond with a specialized part of their circadian system for which anticipatory activity is characteristic. Results comparable to those of the house sparrows have been obtained for pigeons (Abe and Sugimoto, 1987; but see discussion in Phillips et al., 1993) and canaries (Aschoff, 1987). Although the data do not exclude the existence of a FEO in birds, it could be hypothesized that the differences between mammals and birds in their responses to periodic food availability reflect qualitative differences in circadian organization. That food-synchronized activity rhythms in birds are generated and regulated by a different mechanism than anticipatory response to feeding cycles in mammals is also suggested by the finding that in rats the duration of anticipation does not systematically change with the length of the daily food access time (Honma et al., 1983; Stephan and Becker, 1989). This could either mean that in these studies the zeitgeber strength of the specific feeding cycles used did not change as perceived by the rats, although daily food access times were varied between 2 and 4 h (Honma et al., 1983) or between 4 and 12 h (Stephan and Becker, 1989), or that the duration of anticipation in rats is not related to zeitgeber strength.

Which Cues Convey the Information on Zeitgeber Strength of Feeding Schedules?

Two nonexclusive mechanisms by which the house sparrows might perceive the strength of a feeding schedule are conceivable: birds may (i) estimate the duration of the food access time, exploiting this temporai information, or (ii) measure the energetic impact of exposure to periodic food availability, thus exploiting energetic information. Although the birds did not completely use the long food access times, their actual daily feeding time varied between 8 and 15 h (Fig. 5). These differences in feeding times appear sufficiently large to be of potential use as temporal information to the birds. Alternatively, for assessing their energy reserves, the birds could have used body mass as a rough indicator of their physical state. However, body mass did not change significantly with the RF schedules. This could either indicate that the RF schedules did not impair the birds' energy balance or that by adjusting their feeding behavior and digestion to the experimental conditions, the birds were able to compensate for food restriction. A behavioral compensation is likely in view of the increased feeding rates under short food access times (Fig. 6b). This might have enabled the maintenance of daily food consumption at a level similar to that under the preceding conditions (Fig. 6a).

Additional studies should aim at providing further data to decide which factors convey the relevant information about a food zeitgeber to the circadian system of the house sparrows. It is conceivable that a combination of temporal and energetic cues are used. In rats, energetic aspects of feeding schedules seem to play a role in synchronization. In these animals, "entrainment of activity to periodic food access does not require conditions that impose severe energy deficits" because anticipatory activity emerges and synchronizes with the feeding schedules under a wide range of food access times (Stephan and Becker, 1989). However, additional investigations suggested that complex and interacting variables such as nutrient intake, meal size, and energy balance affect the motivational state of the animals and induce and synchronize anticipatory activity (Mistlberger and Rusak, 1987; Mistlberger et al., 1990; Persons et al., 1993; Mistlberger, 1994).

Furthermore, although such a possibility has been discarded in rats (Mistlberger, 1990,1994), the activity or arousal induced by the restricted feeding schedule may be the underlying mechanism of synchronization with feeding cycles.

Functional Aspects of Synchronization With Feeding Times

The synchronization pattern and also the phase angles shown by the birds under the different feeding schedules can be interpreted in an adaptational context. When conditions are good, there is no immediate need to synchronize with times of food availability and to use the whole food access time. Harsh conditions, in contrast, require adequate adjustments, in that the feeding time must not be missed and must be used completely for energy intake. A functional aspect of food synchronization may also lie in the maximization of food efficiency through the preparation of the digestive system to the time of food availability.

The different mechanisms in food synchronization as exemplified by house sparrows and rats may reflect a circadian adaptation to the foraging ecology of each animal. The house sparrow, as an almost exclusively day-active species, may not need to separate food synchronization from light synchronization. Rats, in contrast, are considered to be opportunistic feeders for which an independent adjustment to feeding cycles and light cycles might be more advantageous. The existence of a separate FEO would allow rats to synchronize with times of food availability without compromising a circadian timing with light-dark cycles.

The observation that the birds did not compress their activity time (a) down to below 8 h under RF 8:16 (Fig. 7) illustrates an interesting characteristic of the circadian system. Aschoff (1962) introduced the term "Aktivitatssoll" ("activity quota") to describe the phenomenon that activity levels of animals do not decrease to minimal values even under optimal conditions (i.e., in the laboratory) where such activity is not necessary, for instance, to search for food. A similar behavior has been reported for birds exposed to LD cycles with different photoperiods, which decreased activity time only to a certain lower limit in accordance with photoperiod (in the laboratory: West and Pohl, 1973; in the Arctic: Aschoff, 1969; Aschoff et al., 1970). Such inability to compress activity time down to a certain level is most obviously expressed under weak zeitgeber conditions (West and Pohl, 1973). It might ensure the realization of behavioral needs under zeitgeber conditions that otherwise crucially limit the available activity time. In the present study, activity outside the feeding period could also serve as a kind of "sampling" behavior, which allows the birds to explore food availability before and after the usual feeding time. It might represent a mechanism by which birds can rapidly notice changes in environmental feeding schedules. Alternatively, it might reflect a program evolutionary adjusted to average conditions.

The behavior of nonsynchronized birds illustrates a circadian mechanism other than synchronization that provides adjustment of behavioral rhythms to feeding cycles: although the activity rhythms of some house sparrows free ran under certain experimental conditions, so that their activity periods drifted away from the food access time, these birds still managed to feed during the food access time. The means by which the house sparrows accomplished such food intake outside their circadian activity time is presumably "masking." Masking effects result from a direct action of external stimuli on the overt rhythm without (necessarily) affecting the underlying circadian oscillator (Aschoff, 1960). In the present study, we suggest that feeding cycles enhanced feeding activity ("positive masking"; Aschoff, 1960) during the time of food availability in birds that would otherwise-as dictated from their circadian phases-have missed feeding time, or at least parts of it (e.g., Fig. 1 c). Hence, if birds cannot synchronize with cycles in food availability for whatever reasons, masking might be an alternative way to cope with the environmental rhythmicity

ACKNOWLEDGMENTS

This study was supported by a stipend of the Max-Planck-Gesellschaft to Michaela Hau and by a grant of the Deutsche Forschungsgemeinschaft to Eberhard Gwinner. We wish to thank Marcel Klaassen in particular for statistical advice; Martin Wikelski, Sabine Heigl, Heribert Hofer, and our collegues in the department for technical support, helpful discussions, and valuable criticism on earlier stages of the manuscript Ann Biederman-Thorson kindly improved the English.