The critical role of the suprachiasmatic nucleus (SCN) in the orchestration of circadian rhythms represents one of the most successful brain-behavior relationships established in mammalian neurobiology [1]. The cellular and molecular bases of circadian rhythmicity are similarly well characterized [2, 3]. Despite these advances, the practical manipulation of human rhythms remains unrealized. Most shift-workers, for instance, fail to reset their circadian pacemakers to effect an alignment of the work shift with the worker's circadian day [4–6]. This misalignment may be a contributing factor in the higher nighttime incidence of industrial accidents and the compromised health of shift-workers [7, 8].

In the laboratory under free-running conditions, humans and other mammals typically undergo the rhythmic alternation between physiological and behavioral states appropriate for day (subjective day) and for night (subjective night) [9]. Although individual rhythmic outputs (e.g., locomotor activity, birdsong etc) might be observed to be bimodally expressed, more proximate measures of clock function (e.g., SCN clock gene expression, light sensitivity of the pacemaker, melatonin secretion, etc) tend to alternate unimodally with a period near 24 h [10–12]. One approach to shift-work involves rapid phase-shifting of the entire circadian program to an optimal compromise between the conflicting demands of work and home schedules [13]. A re-entrainment strategy that manipulates only circadian phase, however, can be easily undone by the entraining effects of bright light exposure in transit to and from work and by exposure to sunlight during the day, both of which serve to reinforce normal diurnal entrainment [6, 14]. Additionally, even effective re-entrainment of circadian phase may have costs in terms of the resulting coincidence of subjective night with social demands during the day.

An alternative solution to the problem of shift-work might involve focusing on change in circadian waveform instead of phase. Under appropriate lighting conditions, hamsters and mice readily adopt stable, multi-modal entrainment patterns that may meet the requirements of shift-workers for separate intervals of alertness during both the night and the day [15–19]. These entrainment patterns, furthermore, do not require the complete avoidance of daytime exposure to sunlight. Specifically, rodents maintained in 24 h light:dark:light:dark cycles (LDLD) may entrain their circadian pacemakers so that nocturnal locomotor activity is programmed during each of the two daily scotophases 12 h apart (i.e., in anti-phase). A series of behavioral and physiological studies has indicated that this reorganization reflects a global splitting of the circadian pacemaker into two oscillatory components that cycle out of phase [20] (unpublished data). The net result is that, in each 24 h period, the subject goes through two short subjective nights interrupted by two short subjective days. This entrainable split activity pattern occurs in each of three rodent species examined, although the stimulus requirements for maintaining stable entrainment apparently differ by species [18].

To date, two factors have been shown to influence the incidence of rhythm splitting in LDLD cycles. First, animals that engage in robust wheel-running in response to scheduled daily exercise are more likely to split their activity patterns than are animals that remain inactive [16, 21, 22], and splitting often coincides with a cage change that provokes wheel-running activity [17]. Second, the presence of dim illumination (e.g., 0.004 – 0.10 lux versus complete darkness) during all dark periods markedly increases the proportion of animals that splits in LDLD [17, 19]. These two effects may be related as animals are sometimes more active in dim light than in complete darkness, although this cannot wholly account for the facilitative effects of dim light on splitting [23]. In the present experiments, we assessed directly whether the feedback from a running wheel was necessary for the induction and maintenance of rhythm splitting. Additionally, we examined how split rhythms changed as a function of the relative length of day and night. Identification of factors that influence rhythm splitting should serve to clarify underlying mechanisms and inform the design of protocols examining the feasibility of similar entrainment in humans. Finally, body temperature (T_b) telemetry allowed assessment of circadian markers that may be mediated by mechanisms separate from those underlying locomotor activity [24–27].

As reported in previous studies, a substantial fraction of each of two hamster species readily adopts bimodal wheel-running rhythms under 24 h LDLD cycles [18, 19, 22, 23]. Three aspects of the present studies validate the use of measures besides wheel-running to identify these split circadian entrainment patterns. In Experiment #1, the simultaneous recording of T_b, general locomotor activity and wheel-running yielded essentially identical rhythms within individual animals. Second, separate groups of Siberian hamsters in Experiment #2 monitored by PIR motion detectors and by wheels generated very similar constellations of activity patterns (i.e., split, unsplit, transiently and delayed splits). Finally, the entrainment patterns identified here with motion detectors also mirror wheel-running patterns reported in other studies [17, 18]. The persistence of the split entrainment state in some hamsters exposed to skeleton photoperiods is particularly informative as it discounts the possibility that split activity patterns derive from a) a single long subjective night rendered bimodal by negative masking by light or b) a single short subjective night plus a second non-circadian component positively masked by the second dark period. The formal basis of this split pattern has been considered in detail elsewhere [15, 17].

The use of measures other than wheel running to identify split rhythms permits testing of the hypothesis that a running wheel is necessary to maintain and/or induce this entrainment pattern. In previous studies, animals that failed to run in a wheel during scheduled daily exposure to novel wheels also failed to later exhibit split home-cage running patterns [21, 22]. Additionally, among animals that did not split immediately in LDLD7:5:7:5, we have repeatedly observed abrupt splitting following a cage change during the daytime dark phase [17, 23], which typically induces robust wheel-running activity. The present results clearly establish that a wheel is not necessary for splitting in either species: in Experiment #1, split rhythms in LDLD persisted in several Syrian hamsters following withdrawal of wheel access, and other hamsters later adopted split rhythms in the absence of a functional wheel when the length of each of the daily photophases was lengthened. Wheel immobilization was not necessarily responsible for the loss of splitting in some hamsters (see Table 1) because rejoining has been observed in the first weeks of chronic wheel access in similar experiments [17]. In Experiment #2, Siberian hamsters exhibited split rhythms with or without access to a wheel at rates that did not differ between groups. Thus, wheels are necessary neither for maintenance nor for induction of split rhythms. These experiments, however, were not designed with sufficient statistical power to rule out a minor influence of a running wheel on split rhythms.

Whereas it was expected that general locomotion and wheel-running – both measures of activity – would correspond closely, we hypothesized that the T_b rhythm might differ from activity rhythms if it was controlled by a distinct oscillatory mechanism as suggested in several, but not all, studies [24–27]. As noted above, T_b rhythms were split in all animals with split activity patterns. Because wheel running likely elevates T_b [28, 29], this bimodal pattern might be expected on the basis of masking alone, and indeed, in this study T_b rhythms were acutely influenced by access to running wheels. The wheel-running masking account can be dismissed, however, by the persistence of split T_b patterns after the wheels were immobilized as has been reported with constant light-induced splitting [30]. Highly significant correlations between T_b and GL values raise the vexing possibility that general locomotor activity, as well as wheel running, can have masking effects [31]. Adjustment of the T_b data to control for the correlation between T_b and GL diminished the amplitude but did not eliminate the rhythmicity of the endogenous T_b rhythm, suggesting that an activity-independent T_b rhythm is split in the same fashion as that of activity. Of course, the estimate of the activity-independent T_b rhythm is only as good as the statistical model relating T_b and GL. Use of a number of different statistical models (e.g., log linear regression; polynomial regression; use of various lag times between T_b and GL and GL integration times [32]), did not alter our basic conclusion.

As illustrated in Fig. 1 and in previous studies [17], splitting of Syrian hamster rhythms under LDLD7:5:7:5 is not always stable nor was activity always equally distributed between scotophases. In Experiment #1, an increase in splitting incidence was noted after transfer to LDLD9:3:9:3, and no split animal ever rejoined under this photoperiod suggesting that this lighting condition may promote stable splitting. This finding accords with other of our unpublished data from wheel-running hamsters. The same trend was apparent in Siberian hamsters of Experiment #2 where splitting was numerically most common under nights shorter than 5 h. Under the shortest nights, however, split rhythms largely disappeared and were later recovered in LDLD9:3:9:3. These observations are consistent with a model of splitting in which compression of subjective night reduces the stability of the unsplit pacemaker [15, 23], and thus a split anti-phase entrainment pattern is preferred. The phase-jumping of split rhythms of several Siberian hamsters in skeleton photoperiods further suggests that the individual split components are not fully stable under 3 h scotophases. Further work will be necessary to determine whether a circadian parameter related to "stability" can be operationalized and convincingly distinguished from known parameters such as period, phase and phase response to light.

The multiple oscillations that underlie splitting under LDLD do not share the same physiological basis as those observed after prolonged exposure to constant light [33]. Whereas the latter case is associated with anti-phase oscillations of the left and right SCN, the former appears to exhibit no such laterality [34](Meyer-Bernstein, unpublished observations). Whether the two LDLD oscillations correspond to the as yet unidentified dual oscillators proposed to underlie photoperiodism, however, remains an open question. At a formal level of analysis, distinct "evening" and "morning" oscillators nicely account for the photoperiodic regulation of activity duration (α) of unsplit animals [12, 35]. Because similar photoperiodic regulation of α is seen for each component of the split rhythms in Experiment #2, each of the two split components may be similarly generated by a complex pacemaker comprising multiple oscillators. Supporting this inference, the relationship between phase angle of entrainment and photoperiod for PIR data in Experiment #2 parallels that of unsplit pacemakers [36], albeit non-significantly in the case of the daytime component. Thus, the circadian system likely comprises not just two functional oscillators, but many, which may couple to generate a variety of entrainment states.

Because all mammalian circadian pacemakers likely derive from multiple oscillators in the SCN, we suspect that this split entrainment pattern could be induced in other species under permissive conditions. The practicality and benefit of this entrainment regime to human subjects await evaluation. At least superficially, it may better meet the needs of shift-workers than present phase-shift protocols that attempt to schedule sleep during the real day without achieving full-scale phase shifts of subjective night. With a split circadian rhythm, shift-workers might schedule sleep in two equal intervals sandwiched between a late shift and daytime activity. If the human and hamster pacemakers respond similarly, bright light exposure before and after each sleep phase would reinforce the split pattern rather than compromise entrainment as is the case in some phase-shift protocols. It also remains to be determined whether this circadian arrangement entails any physiological costs, and if so, whether these outweigh potential benefits. From a translational perspective, we are encouraged that split rhythms can be observed under a broader range of conditions than initially identified. Human splitting protocols should consider the influence of factors identified in animal studies: nocturnal illumination, photoperiod and activity levels. Future animal work will assess the influence of age, endocrine status and photoperiodic history.

Entrainment of novel circadian waveforms is possible with the use of exotic lighting conditions and may represent a productive new strategy for chronotherapeutics. While prior studies point to a role of locomotor activity in inducing split activity patterns, access to running wheels is not necessary to induce or to maintain these entrainment patterns. Instead, split entrainment is influenced by the relative length of light and dark phases of the 24 h LDLD cycle.