In mammals, the master circadian pacemaker is located within the paired SCN of the anterior hypothalamus. When the SCN are lesioned, animals become arrhythmic (17–19). Furthermore, if fetal SCN are transplanted into an SCN lesioned animal, circadian rhythms are restored (20), with a period determined by the donor animal (21). The SCN show circadian variations in electrical activity and firing rate over 24 h, with high activity during the subjective day and low activity during the subjective night (22). Individual SCN neurons oscillate with a period of roughly 24 h when dissociated from the rest of the SCN tissue, indicating that these rhythms are generated at an intracellular level rather than occurring as an emergent network property (23). While the role of the SCN in driving circadian rhythms in physiology and behavior in mammals was established in the 1970s, it was not until the late 1990s that the molecular basis of these rhythms was established.

The underlying mechanism generating intracellular circadian rhythms is a transcriptional-translational feedback loop (TTFL) comprising positive, negative, and accessory limbs (24). The positive limb consists of the core clock proteins, CLOCK and BMAL1, which both contain a basic helix-loop-helix domain and bind together to form heterodimers. These in turn bind to E-box enhancer regions of Per1-2 and Cry1-2 genes to promote their transcription (Figure 1). The negative limb comprises the translated PER and CRY proteins which translocate back into the nucleus and directly interact with the CLOCK/BMAL1 complex to inhibit transcription. In turn, the transcription of PER and CRY proteins are reduced, and the proteins are also actively broken down, leading to re-activation of transcription by CLOCK/BMAL1. In addition to the core loop, an accessory loop is also driven by the CLOCK/BMAL1 activation. The Rev-erbα gene is transcribed and produces the orphan nuclear receptor REV-ERBα, which activates a ROR response element in the promoter of Bmal1 to inhibit its transcription. As PER and CRY interact with CLOCK/BMAL1 to inhibit transcription, Rev-erbα falls as well, disinhibiting Bmal1 transcription, allowing levels to rise again (25). CLOCK/BMAL1 heterodimers also drive the transcription of a large number of other genes which contain E-box enhancers in their promoter region, termed clock controlled genes (CCGs), thus allowing the clock to influence a wide range of cellular functions (26). The characterization of the molecular basis of circadian rhythms reflects one of the best examples of how genetic mechanisms can give rise to complex behavior. Indeed, it is remarkable that changes in a single gene can give rise to changes in the period of the circadian clock, or even arrhythmicity.

A clock is of no use unless it can be set to the right time. In mammals, the master SCN clock is entrained to the external environment by time cues (zeitgebers), the most important of which is light detected by the eyes. Indeed, loss of the eye abolishes entrainment (27, 28). The SCN receives input from the retina via the retinohypothalamic tract (RHT), which allows the light to adjust the phase of its endogenous rhythms to match that of the environment (29). Studies on the photoreceptors mediating circadian entrainment demonstrated that mice can still phase shift their locomotor activity rhythms and suppress pineal melatonin in response to light even in the absence of the classical rod and cone photoreceptors that mediate vision (30, 31). These findings suggested the existence of a novel retinal photoreceptor, in addition to the well-characterized rods and cones. It was subsequently shown that a subset of photosensitive retinal ganglion cells (pRGCs) expressing the photopigment melanopsin (OPN4) are intrinsically light sensitive and form the primary projection to the SCN (Figure 1) (32). However, as pRGCs receive input from the outer retina, their output via the RHT depends upon both their intrinsic responses as well as extrinsic signals from rods and cones (33, 34). As such, mice lacking melanopsin (Opn4^−/−) display normal circadian entrainment, but attenuated phase shifting responses to light (35, 36). However, mice in which pRGCs are genetically lesioned are unable to entrain to light (37). Since the identification of pRGCs, it has become clear that these cells mediate more than just circadian entrainment, and are involved in a range of non-image forming (NIF) responses to light, including the pupillary light response, regulation of sleep–wake timing, photophobia, light aversion, and cognitive function, as well as influencing image forming responses, such as visual adaption (38). Melanopsin-expressing pRGCs project to multiple brain targets, including the intergeniculate leaflet, olivary pretectal nucleus, medial amygdala, lateral habenula, and superior colliculus, suggesting that different NIF responses may involve different neural projections (38, 39).

The primary neurotransmitters of the RHT are glutamate and pituitary adenylate cyclase activating polypeptide, which are released at synapses in the SCN in response to photic stimuli (40, 41). This results in increases in calcium concentration and firing rates in SCN neurons. Increased calcium concentration activates intracellular signaling pathways [e.g., cyclic AMP (cAMP) and PKA], converging on the phosphorylation of cAMP response element binding protein, which translocates to the nucleus, binding to cAMP response elements in the promoters of Per1 and Per2, increasing their transcription. This results in the molecular clock in SCN being either advanced or delayed. Although this link between light input, membrane events, and the TTFL has been characterized, the mechanisms by which the TTFL regulates membrane potential are poorly understood (3, 42). While great progress has been made in understanding the molecular basis of photoentrainment, this model is almost certain to be incomplete.

The SCN has widespread projections throughout the brain, including to the septum, the anterior paraventricular thalamus, and to multiple hypothalamic nuclei including the subparaventricular zone, ventromedial hypothalamus, dorsomedial hypothalamus, and pre-optic area (43). The SCN also projects to the paraventricular nucleus, whereby it modulates circadian rhythms in neuroendocrine and autonomic function (44, 45). Furthermore, molecular circadian rhythms are not confined to the SCN. Studies in the late 1990s showed that rat fibroblasts, which had been cultured for 30 years, were capable of rhythmic clock gene expression following a serum shock (46). Furthermore, clock gene reporter studies demonstrated that tissues throughout the body displayed rhythmic clock gene expression, including the liver and adrenal glands (47). These peripheral clocks are thought to play a key role in regulating local tissue physiology (48), with the SCN coordinating circadian timing throughout the body via a combination of neural, paracrine, hormonal, and behavioral signals. The identification of peripheral clocks throughout the body led to a fundamental change in our understanding of circadian rhythms, demonstrating that temporal organization is embedded in the physiology of virtually all cells, tissues, and organs.

The quantity, quality, and timing of sleep and wakefulness are regulated by both a homeostatic and a circadian process (termed Process S and C, respectively) (50). These processes interact to produce periods of wake and sleep during the day (Figure 2). This conceptual model has been useful in interpreting disturbances of sleep/wake regulation, and has been validated by quantitative predictions (61–63). The homeostatic process gradually accumulates during prior wakefulness, and dissipates during sleep. This process is highly correlated with the power of slow wave activity (SWA) on the EEG during NREM sleep, characterized by frequencies in the 0.5–4 Hz range (64). The mechanisms underlying this homeostatic process are unclear. However, a number of putative sleep factors build up in the brain during prolonged wakefulness and dissipate during sleep and these may mediate the homeostatic process (65). Perhaps the best known example is adenosine, which increases in the basal forebrain during wake and dissipates during sleep, and has been proposed to account for the action of adenosinergic drugs, such as caffeine, on sleep (66, 67).

The circadian process varies throughout the 24 h day, allowing wake and sleep at alternate phases of the cycle (68). The primary driver of the circadian process is thought to be the SCN, as following SCN lesions endogenous rhythms in rest/activity are abolished, but the homeostatic regulation of sleep remains intact (69–71). The SCN appears to primarily provide a wake-promoting signal during the active period as lesions increase the total amount of sleep time (72). Studies in which internal desynchrony is induced in rats using 22 h days have also shown that rhythms in NREM sleep can be dissociated from rhythms in body temperature and REM sleep (73). More specifically, the circadian timing of REM sleep has been shown to be associated with clock gene expression in the dorsomedial SCN (74). The SCN likely mediates process C via direct and indirect projections to hypothalamic and brainstem nuclei (including locus coeruleus, VLPO, and orexin neurons), which control levels of arousal and sleep (75, 76). The ventral subparaventricular zone appears to be a key node in this pathway, as lesions of this nucleus result in profound reduction in rhythms of locomotor activity and sleep (77). Together, the homeostatic and circadian processes determine the timing of sleep and wake, and disturbance of either process has the potential to disrupt sleep and influence subsequent waking performance.

In addition to the role of homeostatic and circadian processes in the regulation of sleep, light exposure also directly modulates sleep. While light increases arousal and alertness in diurnal species such as humans, it promotes sleep in nocturnal rodents. Studies in rats have shown that light exposure suppresses activity and results in increased sleep, whereas darkness results in increased wakefulness (49, 78). One of the many projections of melanopsin pRGCs is to the sleep-promoting VLPO (39). This observation, coupled with the role of melanopsin pRGCs in the regulation of numerous NIF responses to light, led to studies of acute sleep induction in mice lacking melanopsin (Opn4^−/−). Initial studies suggested that melanopsin-deficient mice show impaired sleep induction in response to nocturnal light exposure, suggesting that melanopsin plays a key role in mediating sleep induction and maintenance in response to light (79–82). However, these findings were not consistent with data showing that nocturnal light exposure in rodents produces a rise in plasma corticosterone (83), and that these effects on adrenal corticosterone are via the SCN, but independent of effects on the clock (84). Recent studies using different wavelengths of light have shown that short-wavelength 470 nm (blue) light results in delayed sleep onset, coupled with behavioral light aversion and elevated plasma corticosterone, and that this arousal response is attenuated in melanopsin-deficient mice. By contrast, longer wavelength 530 nm (green) light of the same intensity resulted in reduced arousal responses and more rapid sleep induction. Consistent with previous studies, sleep induction in response to green light was attenuated in melanopsin-deficient mice (85). These data are consistent with recent data using chemogenetic activation of pRGCs that produce behavioral arousal, rather than sleep (86). Furthermore, these findings are also consistent with the alerting effects of light described in humans (see Direct Effects of Light on Cognitive Processes).

The finding that melanopsin may be involved in the regulation of the sympathetic nervous system and subsequent arousal is perhaps not surprising, as pRGCs provide a major input to the SCN which is known to regulate sympathetic function. Indeed, this is the primary pathway via which pineal melatonin synthesis is regulated (87). By contrast, explaining why melanopsin-deficient animals show impaired sleep induction in response to light is more challenging. One potential explanation is that melanopsin has recently been shown to be involved in light adaptation (88). If rod/cone signaling normally mediates sleep induction, in the absence of melanopsin responses to bright light stimuli may quickly saturate leading to impaired responses compared to wild-type mice (85). Rather than impaired photic input, an alternative explanation is that the impaired sleep induction in melanopsin-deficient mice may simply reflect a reduced requirement for sleep. Support for this hypothesis comes from data from melanopsin knockout mice showing reduced delta power during the dark phase as well as reduced accumulation of delta power following sleep deprivation. These findings suggest that the need for sleep increases at a slower rate in melanopsin-deficient mice (82). If this is indeed the case, it would suggest that differences in homeostatic sleep could account for impaired sleep induction in melanopsin-deficient mice, rather than deficits in light input to the VLPO as has previously been suggested.

Additional experimental variables may influence sleep induction in response to light. The environmental context is almost certain to influence acute sleep induction, as light exposure in the home cage may produce quite different effects on sleep in comparison to a novel environment. For example, in novel environments, such as an open field or novel object testing arena, sleep induction is not observed in response to light (89, 90). In addition, it should also be considered that while c-Fos has been used as a marker of VLPO activation during sleep (91), induction of Fos in response to light may simply reflect the subsequent sleep/wake status of the animal rather than providing a marker of light input as has been widely used in the SCN (92). While there is a limited retinohypothalamic projection from melanopsin pRGCs to the VLPO (39, 93, 94), it is quite possible that the direct effects of light on sleep may be mediated via other neural pathways.

In summary, as well as circadian and homeostatic processes, light can also directly modulate sleep. However, future studies are required to understand how circadian and homeostatic processes interact to influence acute sleep induction in response to light, as well as the detailed neural pathways involved. These direct effects of light on sleep, and conversely alertness, are clearly important for the effects of light on cognitive processes.

In humans, cognitive function shows variation over the 24 h day, starting off low in the morning, maintaining high levels until habitual bedtime, apart from a dip in the afternoon. This pattern is related to both sleep and circadian processes. Forced desynchrony, in which subjects are exposed to a light schedule to which they are unable to entrain, have been used to investigate the role of circadian and homeostatic regulation of cognitive processes, such as alertness, vigilance, working memory, sleepiness, and mood. These studies show circadian rhythms in all of these different cognitive processes, which also decline with time awake (96, 97).

Again, data from animal studies are less conclusive. Some studies report increased performance during the subjective night in aversive (98) and appetitive tasks (99, 100), whereas others found better performance during the subjective day (89, 101, 102). These contradictory findings may reflect the nature of the behavioral tasks employed. Originally, it was shown that rodents performed best on behavioral tasks when training and testing times were matched, which may reflect state/context dependent learning (103, 104). To date, a major limitation of circadian studies of different cognitive processes is that while circadian time is carefully controlled, the preceding sleep/wake status of the animal is rarely considered.

In summary, processes such as learning and memory certainly appear to be under circadian control to some extent, and this circadian regulation has the potential to influence the effects of light on cognition.

Human studies have demonstrated an important role of light in the regulation of alertness. Imaging studies have also shown that light exposure can influence cortical and subcortical networks involved in cognitive processes, such as attention, arousal, and memory (105–110). In addition, a number of studies have shown that short-wavelength light (470 nm or lower) is associated with the increased suppression of melatonin, reduction in subjective sleepiness, reduced reaction times and changes in EEG power in the delta–theta frequency range (111, 112). These findings are consistent with recent studies on the effects of light-emitting devices on subjective alertness, EEG and sleep latency (113). The primary cognitive effects of light on appear to be via increased alertness—which is typically measured using subjective rating scales as well as tests of sustained attention such as the psychomotor vigilance task, a simple reaction time task. Overall, these studies suggest an increase in subjective ratings of alertness in response to light, though whether these findings always translate into increased cognitive performance is less clear (114–116). However, in cognitive tasks where sustained attention is necessary, light may be expected to exert greater effects.

Despite the wealth of human studies on the acute effects of light on alertness, remarkably few animal studies have investigated the effects of light on cognitive performance. Studies on the acoustic startle response in rats have shown that this response is enhanced by increasing light exposure (117). Subsequent studies investigated the effects of light on tone-cued fear conditioning, finding that light enhances freezing responses in wild-type mice (118). Short pulses of white light at night in mice have been shown to improve consolidation of contextual fear learning and long-term potentiation (119). Under other conditions, bright light exposure impaired spatial navigation performance on a water maze task in BALB/c mice, which was associated with increased anxiety and elevated corticosterone levels (120). Most recently, studies on spontaneous object recognition show that bright light during the test phase impairs recognition performance, regardless of the light level during the sample phase [although see Ref. (121)]. These effects on recognition memory are abolished in both mice lacking rods/cones as well as in mice lacking melanopsin, suggesting that integrated responses from both systems mediate the effects of light on performance (90, 118). Light may also influence emotional processes such as mood, and recent studies using aberrant LD cycles have shown that these may give rise to depression-like behaviors (122) (see T-Cycles). These data add a further complexity to the effects of light, which may involve circadian and non-circadian effects (9, 123).

In summary, data from humans show clear effects of light on alertness, and in some cases, also on performance in tests of sustained attention. However, despite a number of rodent studies exploring the effects of light on different cognitive processes, no consistent effects have emerged. Light may certainly affect the outcome of laboratory tests of learning and memory in rodents. However, the direction and amplitude of any effect may depend on the nature of the test and the different cognitive processes involved. The difficulty of characterizing mechanisms in human studies, combined with the lack of consistent effects in animal models, has led to a lack of any detailed understanding of the photoreceptor contributions and underlying neural pathways involved in such responses.

Given the key role of the circadian system in the regulation of sleep, any disruption of the circadian system is likely to influence subsequent sleep/wake timing. Sleep has been suggested to play a role in cognitive performance (124), and sleep disruption is known to impair multiple aspects of cognition, including arousal, attention, and working memory (125). As such, the effects of sleep disruption must also be taken into account when considering the effects of light on cognitive processes, particularly where circadian function is affected.

In rodents, sleep deprivation influences several aspects of memory, which have been assessed using a variety of behavioral tests. Total sleep deprivation in the first 5 h after training has been reported to reduce contextual fear memory, despite otherwise adequate sleep (126–129), but produce no effect on tone-cued fear memories (126). Using the platform-over-water REM sleep deprivation method, contextual fear memory has been suggested to be impaired, again with no effect on tone-cued fear memories (130–133).

Spontaneous object recognition is a highly tractable test of learning and memory in rodents (134, 135) that has also been widely used to study the role of sleep in learning and memory. Studies have shown that both object-recognition and object-location memories are impaired by 5–6 h of sleep deprivation after training (136–139), with an apparently crucial window at 3–4 h (140). Another test that has been used to study the effects of sleep deprivation is the Morris watermaze, which relies on aversive immersion in water to motivate animals to find a hidden platform. While this requires spatial learning and is sensitive to hippocampal damage, other brain regions and strategies are also important (141). 4 h of selective REM sleep deprivation immediately following training has been suggested to impair performance (142). However, this finding is equivocal with some studies agreeing (143–145) and others disagreeing with the results (146). This may be due to differences between protocols favoring different search strategies and/or brain areas. Spontaneous spatial recognition in the Y-maze has also been studied following sleep deprivation. Similar to other behavioral paradigms, performance is impaired by 12 h total sleep deprivation prior to training (147).

In addition to sleep duration, sleep architecture is also important for memory. Humans sleep in a consolidated bout once a day and progress through many cycles of REM and NREM sleep during the night. If this is fragmented, daytime function is impaired and increased sleepiness occurs (148). By contrast, rodents sleep in multiple short bouts, consisting of both REM and NREM sleep, throughout the day and night, with a greater amount of sleep during the light phase, primarily due to increased length of sleep bouts during the day (149). Disturbing sleep architecture with regular waking prevents normal completion of sleep bouts, resulting in increased sleep pressure despite no change to the total sleep duration (150). Sleep fragmentation also impacts cognitive processes. Mice subjected to sleep fragmentation for 15 days, induced by being disturbed by a bar across the cage every 2 min, have poor learning and retention in the Morris watermaze (151). Similarly, optogenetic activation of hypocretin neurons fragments sleep without altering total sleep time, and when fragmented in the first 4 h following training, causes deficits in object-recognition memory (152). Importantly these studies suggest that even when sleep timing and total sleep duration may remain comparable, fragmented sleep can give rise to impaired performance in specific cognitive processes.

Together the data described above suggest that sleep disruption impairs specific aspects of learning and memory. However, these effects are not straightforward, with sleep deprivation, selective REM deprivation and sleep fragmentation having subtly different effects on different behavioral tests. Some of the cognitive deficits that occur as a result of sleep loss may arise as a result in specific changes in synaptic function, particularly relating to glutamatergic signaling and synaptic plasticity. One example of this is the role of the GluA1 AMPA receptor subunit, encoded by the Gria1 gene. The GluA1 subunit is important in both AMPA receptor trafficking and synaptic plasticity (153–155). Critically, GluA1 levels in synaptoneurosomes in both the cortex and hippocampus have been shown to be elevated following prolonged wakefulness (156). This supports the synaptic homeostasis hypothesis, whereby wakefulness is associated with a net increase in synaptic strength, which is subsequently renormalized during sleep (157). Data from mice lacking GluA1 may provide some insight into the consequences of these changes in synaptic plasticity that occurs during sleep. GluA1-deficient mice show unimpaired performance on associative, long-term memory tasks, such as the Morris watermaze. By contrast, these animals show selective deficits in short-term habituation to recently experienced stimuli (158, 159). These findings suggest that during waking synaptic GluA1 levels increase, reflecting an ongoing habituation and reduction in attention. This hypothesis suggests that sleep may be important for the restoration of attentional performance (160).

In conclusion, while the disruption of sleep undoubtedly influences cognitive function, the specific cognitive processes affected and the underlying mechanisms involved are not straightforward. However, when considering the effects of light—either directly or via its effects on the circadian system—researchers should always be aware that effects on cognition could arise due to a concomitant disruption of sleep.

Arousal is another key factor that may influence the outcome of studies on cognition. In this context, rather than simply being awake, arousal refers to a state of physiological alertness resulting in increased attention and cortical activity, along with changes in motivation and emotional state. Again, light may modulate arousal either directly, or indirectly via the circadian modulation of arousal. Arousal responses are thought to be mediated via projections from the SCN to the dorsomedial hypothalamus which are then relayed to the ascending arousal system, including the noradrenergic locus coeruleus, cholinergic laterodorsal tegmental nuclei, dopaminergic ventral tegmental area and serotoninergic raphe nuclei (75). In addition, the widespread projections of the melanopsin pRGCs may also be important in the effects of light on arousal, including the lateral habenula, medial amygdala, and subparaventricular zone (9, 39). In addition to the regulation of the ascending arousal system, physiological arousal may also be accompanied by increased activity of the sympathetic branch of the autonomic nervous system, resulting in widespread changes in physiology, particularly relating to cardiovascular and adrenal function. Whether these effects of light on arousal are independent or interrelated remains unclear.

The relationship between arousal and cognitive performance is complicated. Yerkes and Dodson (161) reported that with simple learning tasks there was a positive linear relationship between arousal and performance (i.e., the higher the level of arousal, the better the task performance). However, as the difficult of the task increased, an inverted-U shaped relationship between arousal and cognitive performance was observed, with optimal performance requiring an optimal level of arousal (161). As such, rhythms in arousal will influence where an individual sits on the arousal-performance curve, and are likely to contribute to the rhythms seen in behavioral performance in animal studies. As a result of the differences in baseline arousal, the response to stressors such as handling, restraint, and environmental noise will differ (162, 163). When baseline arousal is low, increased arousal may be expected to result in improved performance, but when baseline arousal is high, increasing arousal further may impair performance (Figure 3).

The circadian control of adrenal glucocorticoids via the classical hypothalamic–pituitary–adrenal (HPA) axis is well known (164, 165). As well as the circadian regulation of adrenal glucocorticoids, light has also been shown to directly modulate glucocorticoid release (83), and may also exert different effects on arousal depending upon wavelength (85). This acute response to light does not involve the classic HPA axis, instead relying upon modulation of the sympathetic nervous system (83). While transient increases in glucocorticoids in response to stressors—such as light exposure—are a normal physiological response (termed “allostasis”), long-term exposure to such stimuli can result in fundamentally different responses (“negative allostasis”) (166). Such chronic stress increases baseline glucocorticoid levels and attenuates the amplitude of glucocorticoid rhythms, both at an ultradian and circadian level (164, 167). Elevated glucocorticoid levels are known to affect cognitive processes, such as learning and memory (168). As well as changes in plasma glucocorticoid levels, cardiovascular markers such as heart rate may provide useful markers of arousal. Furthermore, it has recently been suggested that spontaneous fluctuations in locomotor activity may provide a useful marker of generalized arousal (169, 170), which may be beneficial for future studies in this field.

Given the role of the circadian clock in regulating the ascending arousal system as well as the autonomic nervous system, as well as the direct effects of light on these systems, changes in arousal state should also be considered when investigating the effects of light on cognitive processes.

Constant conditions are frequently used in circadian research to study free-running circadian rhythms. While constant darkness allows animals to organize their behavior exclusively according to their internal circadian clock, constant light has been used to study light input as well as a means of producing circadian disruption.

Shifting the LD cycle under which animals are housed has been used to mimic the sudden shift in time-zones produced by jet-lag. Acute jet-lag protocols typically involve a single advance or delay in the LD cycle. In addition, chronic jet-lag—involving repeated shifts of the LD cycle—has also been used as a model of circadian disruption.

As the circadian clock is not exactly 24 h, it is adjusted on a daily basis by the prevailing LD cycle. This entrainment process can be challenged using LD cycles whose length differs from 24 h – termed “T cycles”. A range of different T cycles have been used, where T refers to the day length. For example, an 11 h light, 11 h dark LD cycle is referred to as T22.

Due to the widespread adoption of artificial light, nocturnal light exposure is increasingly common. To study the physiological consequences of light exposure during the dark phase, protocols have been used in which animals are exposed to light during their normal dark (active) phase.

A final protocol that has been used to study circadian disruption in Siberian hamsters (Phodopus sungorus) is the use of DPSs.