A pocket guide for the experimenter studying daily rhythms in memory-related behavior

Abstract

The study of circadian rhythms requires a deep understanding of how biological systems shift on a daily timescale. Similarly, the study of memory requires an appreciation of how memory mechanisms change across time. These complexities create serious challenges for examining circadian rhythms in memory-related behavior, an undertaking in which both sets of considerations must be integrated and assessed simultaneously. In this pocket guide, we present an experimental design optimized to distinguish circadian influences on different stages of the memory process. Our approach distills ideas from a complex body of literature, offering a simple path forward in circadian memory research.

INTRODUCTION:

More than fifty years of behavioral studies point to circadian rhythms in the encoding, molecular consolidation, maintenance, and retrieval of long-term memories (Hartsock & Spencer, 2020). A parallel and rapidly expanding body of work supports a role for neural circadian clocks. These data suggest that circadian rhythms in learning and memory processes are not merely an intriguing phenomenon arising in isolated circumstances, but rather a pervasive and fundamental feature of memory function (Krishnan and Lyons 2015; Rawashdeh et al. 2018; Hartsock and Spencer 2020). Yet, when we began to think about memory rhythms in our own research, we encountered the challenge of deducing best practices from a disparate literature. Here we provide a pocket guide for the experimenter seeking to incorporate a rhythms perspective in studies of memory-related behavior.

A PRIMER ON CIRCADIAN RHYTHMS:

There are three broad temporal categories of biological rhythms: circadian (daily, such as the sleep-wake cycle), ultradian (shorter-than-daily, such as hourly secretory pulses of glucocorticoid hormones), and infradian (longer-than-daily, such as the human ovulatory cycle) (Hartsock et al., 2022). We will focus on circadian rhythms (see Box 1 for a circadian glossary). By definition, circadian rhythms are endogenously-generated daily cycles in behavior and physiology. They influence virtually all biological processes that have been studied to date, including various aspects of metabolism, cardiovascular function, and behavior (Rijo-Ferreira & Takahashi, 2019). In turn, disruptions in circadian rhythms are implicated in medical conditions ranging from obesity to Alzheimer’s disease (Geng et al., 2025; Rijo-Ferreira & Takahashi, 2019). Basic science research, including behavioral studies, has shown a tight link between circadian rhythms and memory function (Krishnan and Lyons 2015; Rawashdeh et al. 2018; Hartsock and Spencer 2020). Such work has prompted an effort to explicate the mechanisms by which the circadian system modulates memory in health and disease.

Box 1: Circadian Glossary.

Circadian rhythm:

a near-24-h biological cycle that is endogenously generated and sustained

Clock genes:

the genetic elements whose products constitute the molecular clock

Diurnal rhythm:

a near-24-h biological cycle that is regulated through non-endogenous mechanisms or whose endogenous regulation has yet to be experimentally demonstrated

Entrain:

synchronize (an oscillatory process)

Local clock:

a tissue-level circadian oscillator outside the SCN

Master clock:

a tissue-level circadian oscillator in the suprachiasmatic nucleus of the hypothalamus (SCN) that coordinates local clocks throughout the rest of the body

Molecular clock:

the transcriptional-translational feedback loop that comprises the cyclic expression patterns of the the core clock genes and has a period of approximately 24 h

Nychthemeral rhythm:

a near-24-h biological cycle that is regulated through non-endogenous mechanisms or whose endogenous regulation has yet to be experimentally demonstrated

Phase:

the timing of peaks and troughs within an oscillatory process compared to a reference, or a specific point within an oscillatory process

Zeitgeber:

time-giver (tr. German); typically an external entrainment cue that sets the timing of a circadian rhythm

Zeitgeber time (ZT):

the time of day in relation to ths environmental light-dark cycle, with ZT0 signifying the time of light onset and ZTX signifying the time X hours after light onset

Circadian rhythms depend on molecular clocks distributed throughout the body. The timekeeping abilities of molecular clocks arise from a cellular transcriptional-translational feedback loop, which consists of daily oscillations in the expression and function of clock genes. In mammals, the core clock genes include Bmal1, Clock (and its forebrain substitute Npas2), Period (Per1/2/3), and Cryptochrome (Cry1/2) (Reick et al., 2001; Takahashi, 2017). A heterodimer formed by the BMAL1 and CLOCK/NPAS2 proteins induces transcription of the Period and Cryptochrome genes, whose protein products then inhibit further transcription by BMAL1 and CLOCK/NPAS2. This feedback mechanism constitutes a cell-level oscillator with a cycle length of approximately 24 h. By controlling cellular transcription and engaging in other direct molecular interactions, the molecular clock anchors cellular functions to the time of day.

A small hypothalamic brain region called the suprachiasmatic nucleus (SCN) acts as a circadian master clock to coordinate molecular clocks throughout the rest of the body (Colwell, 2011; Reppert & Weaver, 2002). By way of the retinohypothalamic tract, periodic lighting cues provide input to molecular clocks within the SCN and set the phase of SCN clock gene expression cycles (Güler et al., 2008; Reppert & Weaver, 2001). Thus, the SCN can align its functions with the solar day (Colwell, 2011). So powerful is the influence of light on the SCN that biological rhythms researchers refer to light as the primary zeitgeber (German: time-giver) for the circadian system, employing the terminology of zeitgeber time to convey the number of hours since light onset in circadian experiments (Ashton et al., 2022). Compared to molecular clocks in other tissues throughout the body, molecular clocks in the SCN are unique in their ability to entrain directly to periodic environmental light. In turn, SCN neurons exert modulatory control over a variety of peripheral signals, which can be either diffusible, such as glucocorticoid hormones (Spencer et al., 2018), or non-diffusible, such as the firing of pre-autonomic neurons in hypothalamic regions outside the SCN (Buijs et al., 2006; Colwell, 2011). Daily oscillations in these SCN-orchestrated signals serve to relay solar timing information to cells throughout the rest of the body.

Molecular clocks have been identified in virtually all tissues outside the SCN, including the heart, liver, and non-SCN brain regions (Colwell, 2011; Reppert & Weaver, 2002). Individual cells in tissues outside the SCN can temporarily maintain oscillations in clock gene expression independent of the SCN, but they rely on SCN-orchestrated signals to coordinate and sustain these oscillations (Hastings et al., 2003; Reddy et al., 2007; Welsh et al., 2004). Coordinated rhythmic clock gene expression among cells within a tissue gives rise to a tissue-level clock, or local clock, which appears to be necessary for proper tissue function, such as for memory (Hasegawa et al., 2019; Kwapis et al., 2018; Ozburn et al., 2015; Parekh et al., 2019; Woodruff et al., 2018). Local clocks may control a unique set of physiological processes from one tissue to another, likely facilitating tissue-specific circadian rhythms in function (Kumar et al., 2024; Wolff et al., 2013; Yamazaki et al., 2000; Yoo et al., 2004). In the brain, this arrangement may underlie a region- or circuit-specific coordination of memory processes (Hartsock & Spencer, 2020).

We have sought in our research to understand how local clocks in the brain contribute to circadian rhythms in memory-related behavior. Still, such rhythms can arise through a variety of mechanisms. For example, daily fluctuations in SCN-orchestrated signals can directly affect memory processes (Liston et al., 2013). Similarly, biological functions such as feeding, which are themselves modulated in part by the circadian system, can influence various aspects of memory neurobiology (Mandal et al., 2018). Whatever the mechanism for generating circadian rhythms in a particular memory-related behavior, thoughtful experimental design can help to uncover the underlying neurobiology.

A PRIMER ON MEMORY NEUROBIOLOGY:

The neurobiology of memory can be divided roughly into four sequential processes: encoding, molecular consolidation, maintenance, and retrieval (Frankland et al., 2019; Guskjolen & Cembrowski, 2023; Kandel et al., 2014). Encoding takes place during a learning event and comprises the immediate changes in neural function that support information processing. Strong or prolonged neural activation during encoding can trigger the transcription and translation of new cellular substrates that preserve changes initiated during encoding; these processes can last up to about 48 h and together constitute molecular consolidation. Maintenance can be viewed as the collection of neural functions that sustain changes produced by consolidation. Retrieval then involves the reactivation of neural cells involved in the memory trace. Together, these neurobiological processes support the formation (encoding ➔ consolidation), retention (consolidation ➔ maintenance), and recall (maintenance ➔ retrieval) of long-term memories.

Accumulating evidence suggests that each of these processes can fall under circadian influence. For example, one study assessed whether zebrafish could learn to associate an aquarium compartment with mild electric shock at a different rate depending on the time of day, illustrating faster encoding when training occurred during the usual active phase of the rest-activity cycle (Rawashdeh et al., 2007). Another found that sea slugs, whether diurnal or nocturnal, could better remember an operant task if learning occurred during the active phase, providing evidence for a circadian rhythm in molecular consolidation (Lyons et al., 2005). A third revealed that contextual threat retention/retrieval is stronger during the inactive phase in mice, regardless of the time of learning (Chaudhury & Colwell, 2002). This particular set of studies was among the first to deploy the experimental designs and interpretation guidelines upon which our pocket guide has been built. We have provided a comprehensive review of such behavioral studies elsewhere (Hartsock & Spencer, 2020).

Circadian rhythms appear also in many neurobiological mechanisms that support memory processes. For example, the protein kinase A (PKA) and mitogen-activated protein kinase (MAPK) cascades exhibit robust basal and activity-dependent daily oscillations in the rodent hippocampus (Eckel-Mahan et al., 2008; Phan et al., 2011; Rawashdeh et al., 2016; Shimizu et al., 2016; Wardlaw et al., 2014). Learning-induced activation of the transcription factor CREB – a downstream effector of the PKA and MAPK cascades – also fluctuates on a circadian timescale (Eckel-Mahan et al., 2008; Jilg et al., 2019; Rawashdeh et al., 2016), and this oscillation is abolished in Period1 clock gene knockout mice (Rawashdeh et al., 2016). Accordingly, clock gene disruptions local to the hippocampus have been shown to impair spatial, social, and contextual memory functions (Hasegawa et al., 2019; Kwapis et al., 2018). Such data strongly suggest that circadian rhythms in hippocampus-dependent memory functions are generated by molecular clock influences on hippocampal memory mechanisms. Related work has demonstrated that the circadian regulation of memory neurobiology is commonly observed across other brain areas and memory paradigms (Hartsock & Spencer, 2020; Krishnan & Lyons, 2015; Rawashdeh et al., 2018).

EXPERIMENTAL DESIGN IN MEMORY RHYTHMS RESEARCH:

As memory researchers interested in rhythms, we are tasked with two responsibilities. First, we must identify time-of-day differences in memory function. Second, we must determine which of the neurobiological events that support memory exhibit rhythmicity. In this section, we describe an approach for tackling these challenges. We have arrived at this approach by integrating ideas from the extant literature on memory-related behavior (Hartsock & Spencer, 2020; Krishnan & Lyons, 2015; Rawashdeh et al., 2018), and we have used this approach in our own research to characterize circadian rhythms in conditioned threat extinction (Hartsock et al., 2023). Here we explain the considerations behind our approach and offer suggestions for its broader implementation. For simplicity, we focus on experimental designs involving a single training session, with any follow-up being a memory recall test. The principles we discuss, however, can be applied in designing more complex experiments.

Assume we want to ascertain whether the learning and/or memory of a conditioned tone-shock association exhibits time-of-day differences in hamsters. Our ultimate goal in such an investigation would be to identify whether underlying processes of encoding, molecular consolidation, maintenance, or retrieval might follow a daily rhythm. An attractive initial strategy might be to use two experimental groups. Each group would undergo acquisition training trials (conditioned associations of the tone with shock) during either the daytime or nighttime. Twenty-four hours later, both groups would be given a subsequent memory recall test to determine whether the conditioned stimulus (tone) could produce the conditioned response (defensive freezing behavior) in the absence of the unconditioned stimulus (shock). We would subsequently analyze freezing during training and testing sessions as a behavioral correlate of the tone-shock association.

If we were to observe greater freezing in earlier acquisition trials during the daytime, this would indicate a time-of-day difference in the rate of acquisition of the tone-shock association. We would interpret the finding to support a time-of-day difference in the encoding of the tone-shock association, given that encoding takes place during the initial training session. However, if we were to observe greater daytime freezing upon re-presentation of the tone during testing 24 h later, our interpretation would be much less clear. The reason is that this experimental design does not allow us to distinguish whether a rhythm observed during testing owes to the time of day of training or the time of day of testing. In this design, any group trained during the daytime might exhibit strong recall, regardless of the time of testing. Likewise, any group tested during the daytime might exhibit strong recall, regardless of the time of training. This ambiguity would prevent us from identifying whether memory consolidation, maintenance, or retrieval exhibited a rhythm, and a follow-up experiment would be needed. For example, in our own work on conditioned threat extinction, our initial publication illustrated a clear rhythm in extinction recall but did not allow us to pinpoint the neurobiological process under circadian control (Woodruff et al., 2015). Thus, although this design can provide a straightforward method for a first experiment, we suggest a slightly more sophisticated design that can provide substantially more information.

There are four essential considerations for upgrading the current experiment to this stronger one. First, crossover conditions are required for resolving whether a time-of-day difference observed during testing depends on the time of day of training versus testing (Chaudhury & Colwell, 2002; Decker et al., 2007; Lyons et al., 2005; Rawashdeh et al., 2007). The most basic design includes training at two times of day and testing at two times of day, with crossover conditions arranged such that testing is scheduled at two times of day corresponding to each training time. For example, tone-shock pairings might take place either during the day or at night. Hamsters trained during the day could be tested after 24 hours (again during the day) or after 36 hours (at night). Conversely, hamsters trained at night could be tested after 24 hours (again at night) or after 36 hours (during the day). These crossover conditions would allow the determination of whether a rhythm during testing owes to the time of day of training versus testing.

But in this scenario, we would still risk missing a rhythm. Imagine that there is a rhythm in tone-induced freezing, with a peak at sunrise and a trough at sunset. If we were to test memory recall only at mid-day and mid-night, our sampling would occur during the fall and rise of the rhythm but would omit the peak and trough, creating a false impression that no rhythm is present (Figure 1A). For this reason, the second essential consideration for such experiments is the sampling frequency. The best practice is to schedule several time points throughout the 24-h day for both training and testing (Figure 1B), allowing us to observe all phases of a putative rhythm in memory-related behavior. Circadian research has employed a range of sampling frequencies, with inter-sample intervals of 4 h being common and some studies calling for even shorter intervals (Hughes et al., 2009, 2017; Xin et al., 2021). For experiments on memory-related behavior, we suggest an inter-sample interval of 6 h. In our experience, this longer minimum interval is necessary for feasibility, reducing the number of subjects needed for crossover conditions and allowing flexibility for time-intensive memory tasks.

Figure 1. Sampling frequency during assessments of memory-related behavior.

Left panel If memory performance happens to be sampled at the peak and trough of an oscillation, a time-of-day difference will be observed (red data points). However, if memory performance happens to be sampled during the rise and fall of an oscillation, a time-of-day difference will be missed (black data points). For this reason, the best practice is to include more than two sampling points when characterizing a rhythm in memory-related behavior. Middle panel Sampling across the day with four or more data points ensures that a rhythm can be visualized. Right panel In tests of memory recall, it is important to test not only at 24 h after acquisition, but also at 48 h, which facilitates analysis of changes in memory-related behavior from one oscillation to the next. This right-most panel is our minimum recommended sampling frequency for tests of memory recall.

The third essential consideration is that 24-h fluctuations in memory recall can create a false impression of memory decay (or strengthening) in repeated recall tests. Scheduling recall tests across a full 24-h window, however, enables comparison of two oscillations that occur in succession. This can be achieved by including an additional memory recall test at 48 h after acquisition training (i.e., a full 24 h after the first of multiple memory recall tests). For example, imagine that our tone-shock association indeed exhibits a 24-h oscillation during cued recall testing but also, through repeated presentations of the tone alone, undergoes memory extinction, a secondary learning and memory process that produces a new tone-no shock association (Bouton et al., 2021). We could distinguish the 24-h oscillation in threat recall from memory extinction by comparing freezing at 24 h to freezing at 48 h, identifying how the memory-related response changes not only within a 24-h oscillation (assessing the oscillation across time points from 24 to 48 h), but also from one oscillation to the next (directly comparing the 24-h and 48-h time points). This approach has been used previously for precisely this purpose, illustrating extinction across repeated threat recall tests in mice (Chaudhury & Colwell, 2002). The strategy makes obvious whether a rhythm exists for memory recall and controls for changes in memory-related behavior across repeated recall tests (Figure 1C).

The fourth essential consideration is the possibility of time-stamping, a phenomenon in which memory recall is best at a particular interval after learning has occurred (Holloway & Wansley, 1973). For example, perhaps our hamsters recall the tone-shock association most strongly at 24-h intervals after training, regardless of the time of day. While time-stamped memories are believed to rely on a time-of-day marker generated by the SCN (Hut & Van der Zee, 2011), they differ from other circadian-regulated memory processes in that memory recall depends on the interval after training rather than the time of day of testing. To identify time-stamping, it is necessary to examine the time interval between training and testing in a systematic fashion, making sure to assess the same set of train-test intervals across experimental groups.

Figure 2 illustrates an experimental design optimized to address these four considerations. Such an experimental design can accommodate within- or between-group measures. A between-group design can facilitate cleaner data interpretation but tends to create logistical and feasibility challenges due to the unwieldy number of subjects required. On the other hand, a within-group design necessitates repeated-measures testing within each group, making it important that the recurring memory test does not itself significantly alter the memory, as through extinction, reinforcement, practice effects, or reconsolidation. One potential workaround for within-group recall tests is to use short memory “probe” trials, which limit new learning by minimizing the number of memory stimulus presentations, as in our work on conditioned threat extinction (Hartsock et al., 2023).

Figure 2. An experimental design optimized for memory rhythms research.

In this experimental design, training occurs at one of four times across the 24-h day. Corresponding to each training session, memory testing occurs five times, with the first test of memory recall taking place 24 h after training. Testing sessions are scheduled across the full 24-h day, with train-test intervals ranging from 24 to 48 h. This design enables the discrimination of rhythms in training versus testing, includes enough sampling time points to reveal a rhythm, accounts for changes in memory performance from one oscillation to the next, and identifies time-stamping.

DATA INTERPRETATION IN MEMORY RHYTHMS RESEARCH:

Deciphering the results from such an experiment can present its own challenges. This is the one drawback of producing a dataset so rich with information! Fortunately, the patterns that emerge can be quite distinct.

When training sessions are scheduled across the day, a fluctuation in the rate of acquisition indicates a putative circadian rhythm in encoding (Figure 3A). This can be observed only in learning paradigms that require multiple acquisition trials, such as maze learning, Pavlovian conditioning, or a bar-pressing discrimination task. Keep in mind, however, that many common paradigms do not require multiple trials for acquisition and therefore do not provide a within-session measure of acquisition rate, as is the case for object or place recognition tasks. When multiple acquisition trials are present, a fluctuation in the number of trials to reach criterion is a good indicator of putative circadian regulation.

Figure 3. Example results from an experiment on memory rhythms.

A When multiple groups are scheduled to undergo memory acquisition at different times across the day, a time-of-day difference in learning may become evident in the form of time-of-day differences in the rate of acquisition. In this example, acquisition occurs most rapidly at ZT6. B Memory performance during testing may depend on the time of day of training. In this example, memory performance is strongest when training occurs at ZT12, regardless of the time of day of testing. C Memory performance during testing may depend on the time of day of testing. In this example, memory performance is strongest when testing occurs at ZT12, regardless of the time of day of training. D Memory performance during testing may depend on the time interval between training and testing. In this example, memory performance is strongest at 24-h intervals after training, regardless of the time of day of training or testing. E Memory performance during testing may depend on a combination of factors. In this example, memory performance is strongest when training occurs at ZT18 and testing occurs at ZT12.

A time-of-day difference that emerges across testing sessions must be interpreted with attention not only to the time of training, but also to the time of testing. If a time-of-day difference across testing sessions depends upon the time at which training occurred, the time-of-day difference must reflect a rhythm in encoding or consolidation processes (Figure 3B). In this case, if there is no time-of-day difference in the rate of acquisition across training sessions, we can safely conclude that the time-of-day difference observed during testing reflects a rhythm in consolidation processes. If a time-of-day difference is observed across initial training sessions, however, and if its phase matches that observed across testing sessions, it is ambiguous whether the time-of-day difference observed during testing reflects a rhythm in encoding versus consolidation processes. If the phase of the time-of-day difference observed across training sessions does not match the phase of the time-of-day difference observed across testing sessions, then the time-of-day difference observed across testing sessions is likely to reflect a rhythm in consolidation processes. If a time-of-day difference across testing sessions depends instead on the time of testing, the time-of-day difference reflects a rhythm in maintenance and/or retrieval processes (Figure 3C). Rhythms in maintenance and retrieval cannot be discriminated in experiments that are strictly behavioral, given that the strength of memory maintenance and the strength of memory retrieval both contribute to the strength of memory recall during the testing session. Finally, if we observe a time-of-day difference across testing sessions that depends on the time interval between training and testing, we can interpret this as time-stamping (Figure 3D).

Of course, the presence of a rhythm in one process does not preclude the presence of a rhythm in another. Where multiple processes are circadian regulated, their influences are superimposed (Figure 3E). For example, one study illustrates that conditioned contextual threat recall is stronger in mice trained during the inactive phase, but only when they are tested also during the inactive phase (Eckel-Mahan et al., 2008). This study shows that rhythms in threat recall are not present in rats trained during the active phase, suggesting that rhythms observed in inactive-phase-trained mice represent a combined influence of rhythms in both learning and recall. Given that each memory process invokes its own neurobiological mechanisms, the ability to discriminate circadian rhythms in one versus another, or to illustrate rhythms in multiple processes, can provide a clear direction for follow-up studies on the underlying neurobiology. The logic of this section is summarized in Figure 4.

Figure 4. Flow chart for interpreting a memory rhythms experiment.

At left is shown a flow chart for characterizing time-of-day differences that occur across training sessions. At right is shown a flow chart for characterizing time-of-day differences that occur during testing sessions.

ASSESSING THAT A RHYTHM IS CIRCADIAN:

Once a time-of-day difference has been demonstrated, multiple criteria must be met to identify the phenomenon as a true circadian rhythm (Johnson, 2004). First, we must show that the putative circadian rhythm can sustain a cycle length of approximately 24 h in constant conditions. This shows that the rhythm emerges through endogenous mechanisms rather than as a response to daily environmental signals. Environmental signals typically held constant for circadian studies include light, temperature, and humidity. Regarding light, many considerations are vital (Box 2). For instance, constant darkness or constant dim light must be used instead of constant bright light, given that constant bright light disrupts SCN circadian function (Daan & Pittendrigh, 1976; Ma et al., 2007; Ohta et al., 2005). If a rhythm can persist without these external cues, this provides evidence that the rhythm can be generated through internal mechanisms. However, if a rhythm turns out to occur only in response to daily environmental signals, or if the rhythm has not yet been shown to meet circadian criteria, the rhythm is more appropriately labeled as a nychthemeral rhythm (i.e., a rhythm exhibiting a cycle length of approximately 24 h) or, though less precise but more common in parlance, a diurnal rhythm (i.e., a rhythm occurring daily).

Box 2: Considerations About Light in Circadian Experiments.

When other environmental variables are held constant, a 24-h rhythm that can persist in the absence of daily periodic lighting cues is considered to emerge endogenously. Hence, rhythms researchers often use constant darkness to test for circadian regulation. Constant bright light is not a viable alternative, since it desynchronizes molecular clocks in the SCN (Daan & Pittendrigh, 1976; Ma et al.. 2007; Ohta et al., 2005). Even constant dim light may be problematic, with dim light at night impairing memory physiology (Bedrosian et al., 2011; Fonken et al., 2012).

Laboratory rodents lack the retinal L-cone receptor, which is sensitive to long light wavelengths in humans. Thus, a longstanding assumption has been that dim red light can be used to observe rodents during the dark phase without activating the SCN and producing circadian disruption (Fall, 1974; Finley, 1959; Hanifin et al., 2006). Yet several studies have questioned the extent to which laboratory rodents are insensitive to dim red light. The rat retina responds to red light (Niklaus et al., 2020), rats can perform operant discrimination under red light (Nikbakht & Diamond, 2021), and dim red light can acutely activate the SCN (Stritzel et al.. 2025) and suppress melatonin secretion (Dauchy et al., 2015). Consequently, red light should be used sparingly, if at all, and held constant across experimental groups. Near-infrared lights and infrared-sensitive cameras are a good option for behavior monitoring in total darkness. Additional light sources on computers, fire alarms, behavior boxes, and other electronics should be turned off or covered with electrical tape or aluminum foil.

Even brief light pulses during the dark phase can disrupt the circadian system (González, 2018). Therefore, after optimizing room lights for housing and behavior, accidental light exposure remains a major concern. It is often necessary to educate animal care staff on lighting needs to ensure that light exposures do not occur during cage changes or other routine care. Similarly, animals must be protected from light during transportation to and from behavioral testing, necessitating light-related training for lab members. Under even the best of circumstances, mistakes can still happen. Light monitors can be installed in the housing room to identify accidental light exposures. To verify circadian stability, daily activity patterns can be monitored with running wheels or movement detectors. Any use of dark-phase light exposure should be reported, including the light wavelength, light intensity, and duration and portion of the dark phase during which light exposure occurred.

Constant darkness is generally not feasible for human studies. Instead, forced desynchrony is a go-to method (Wang et al., 2023). In this paradigm, circadian rhythms are systematically separated from the sleep-wake cycle. This is achieved with a non-24-h sleep-wake cycle that is different enough from 24 h in length that the circadian system cannot entrain to it. By teasing apart sleep and circadian rhythms, forced desynchrony facilitates the identification and assessment of circadian rhythms distinct from rhythms organized by the sleep-wake system. Detailed discussion of forced desynchrony has been provided elsewhere (Wang et al., 2023).

Second, we must show that the putative circadian rhythm can entrain to relevant periodic environmental stimuli. The most commonly utilized entrainment signal, on grounds both pragmatic and physiological, is light. The entraining influence of light on a putative circadian rhythm is often assessed by shifting the timing of periodic lighting to a new schedule or maintaining a subset of animals on a nonstandard light cycle upon entry into the vivarium. One typical phase-shifted or nonstandard lighting schedule is the reversed light-dark cycle, in which the daily lighting schedule is adjusted by 12 h. Bear in mind that the circadian system is slow to adjust its internal timing, and best practice is to wait at least the same number of days as hours shifted before determining whether a biological process has become entrained to a new lighting schedule. (In fact, a rapid shift to a new light cycle suggests that the process may depend on light rather than the circadian system.) In any case, if a rhythm can entrain to the timing of an external signal, this provides additional evidence that the rhythm is circadian. When planning experiments, note the possibility also of light masking, the tendency for nocturnal rodent behavior to differ under light compared to darkness (Mrosovsky, 1999; Von Gall, 2022). Most rodent behaviors are sensitive to light masking, making it difficult to disentangle effects of light from circadian influences. One way to address this concern is to hold lighting conditions constant across groups, with all behavior being assessed in darkness (Box 2).

Third, one would ideally show that the putative circadian rhythm exhibits temperature compensation, which means that it can maintain the same cycle length throughout a range of physiological temperatures. Although this feature of the circadian system is highly conserved across a variety of biological systems and functions (Narasimamurthy & Virshup, 2017), it is rarely considered with regard for memory-related behavior. Mammals are homeothermic, so only under severe environmental conditions could mammalian molecular clocks experience a temperature variation extreme enough to alter uncompensated metabolic processes. On the other hand, poikilotherms (insects, reptiles, etc.) might offer a means to examine temperature compensation in memory-related behavior, but we have no knowledge of any such reports. Temperature compensation becomes more relevant for mammalian studies conducted in vitro, given that the enzyme-based cellular mechanisms supporting circadian function are resistant to changes in temperature (Ishiura et al., 1998; Izumo et al., 2003; Nakajima et al., 2005). These findings suggest that, like the biological processes underlying circadian rhythms in other functions, those underlying circadian rhythms in memory may be remarkably stable against fluctuations in temperature. Thus, despite the limited applicability of temperature compensation for studies of mammalian behavior, the topic merits mention for its potential role in non-mammals and its importance at the molecular level.

Finally, although not strictly required as a circadian criterion, we can show that the putative circadian rhythm depends on a molecular clock, be that clock in the SCN, local to a particular brain region, or present elsewhere in the organism. We find this strategy to be the most informative for elucidating a circadian mechanism, especially in cases where the clock of interest is manipulated locally. Multiple studies have now demonstrated that memory functions can be altered through the manipulation of local clocks in the brain (Hasegawa et al., 2019; Kwapis et al., 2018; Ozburn et al., 2015; Parekh et al., 2019; Woodruff et al., 2018), illustrating the dependence of many memory functions on local circadian mechanisms. We have outlined these criteria for circadian regulation in Box 3.

Box 3: Criteria for Circadian Regulation.

• The putative rhythm persists in constant environmental conditions, such as constant darkness

• The putative rhythm can entrain to periodic environmental stimuli, such as shifts in the light-dark cycle

• The putative rhythm is resistant to changes in body temperature (generally not considered for mammallan behavior)

• Optional: The putative rhythm depends upon a molecular clock in the SCN, local to a particular brain region, or elsewhere in the organism

CONCLUDING REMARKS:

A vast literature details many fascinating phenomena contributing to circadian rhythms in memory. To our knowledge, no resource yet exists to organize this information for the purpose of clarifying research protocols. We have offered a framework to address this need, providing examples and instructions to guide experimental design and interpretation as one moves from behavioral to mechanistic investigations.

Several factors may operate on both circadian rhythms and memory function, including stress, sex differences, and feeding schedules. Additionally, measures of memory-related behavior during the active versus inactive phases are often confounded by the timing of sleep. These types of variables are particularly difficult to study in a circadian context, given that they necessitate the inclusion of additional treatment groups in experiments already pushing subject capacity. Nonetheless, they deepen the clinical relevance of circadian work by offering insights into how circadian rhythms set a backdrop for other environmental elements to influence memory. Extricating these interactions represents a major goal for circadian memory research, and we suggest that circadian characterizations represent a first step toward incorporating other variables.

Based on our own experiences, we can appreciate the technical, logistical, and financial challenges inherent to such work. Simultaneously, we see immense potential for discovery. A variety of animal models and memory paradigms have yet to be scrutinized from a rhythms perspective. Further, within models and paradigms, reports often disagree as to which neurobiological aspects of memory fall under circadian control or what time of day corresponds to peak circadian phase (Hartsock & Spencer, 2020). A central objective for future work will be to standardize methods across laboratories – a goal that might be facilitated through the adoption of a common experimental framework. Whether you are looking ahead as a new or seasoned researcher, we hope this pocket guide provides a starting point.