Chronobiology is the study of biological rhythms: how living systems generate recurring changes, adjust them to environmental cycles and coordinate them across cells, organs and behaviour. The gold standard is an explanation that separates an internal clock from an externally repeated event, measures timing accurately and shows what the evidence can—and cannot—establish.
A plant preparing for daylight, a cell producing a rhythmic molecular signal and a coastal animal becoming active around the tide all raise a timing question. Their mechanisms are not interchangeable. This guide builds a practical way to investigate them: define the measured event, identify its reference clock, follow the proposed mechanism and test the explanation when the conditions change.
The guide connects the established How Science Works: Chronobiology article with the wider Gold Standard collection. Its learning purpose follows The Gold Standard of Education: make knowledge usable beyond a familiar example. Every numerical dataset below is an invented teaching case, not a clinical reference range or a published experimental result. The examples explain reasoning; they do not prescribe sleep restriction, light exposure, medication or animal experiments.
THE GOLD STANDARD · CHRONOBIOLOGY
Follow biological time from pattern to explanation
Choose a route through the timing question, its mechanism and the evidence needed to test it. All teaching chapters remain open below.
ROUTE 1 · CHAPTERS 1–4
Define and measure biological time
Separate repeating events, period, phase and clock drift.
ROUTE 2 · CHAPTERS 5–8
Follow clock mechanisms
Connect feedback, biochemical oscillation, temperature and entrainment.
ROUTE 3 · CHAPTERS 9–12
Interpret changes across the system
Distinguish phase shifts, masking, coordination and population averages.
ROUTE 4 · CHAPTERS 13–16
Connect organisms and measurement
Explore plants, tidal timing, sleep concepts and sampling design.
ROUTE 5 · CHAPTERS 17–20
Model, diagnose and transfer
Check assumptions, solve a changed-context case and continue learning.
Full chapter index · Try the transfer task · Gold Standard subject hub
Chapter index
Define and measure biological time · Chapters 1–4
Follow clock mechanisms · Chapters 5–8
Interpret changes across the system · Chapters 9–12
Connect organisms and measurement · Chapters 13–16
Model, diagnose and transfer · Chapters 17–20
CHAPTER 1 OF 20 · Define and measure biological time
1. Begin with a measurable recurring event
“The organism has a rhythm” is a starting observation. A useful investigation names the variable: leaf angle, movement counts, a molecular reporter signal, a hormone concentration or the timing of an event. It also states the observation interval and how the measurement was made.
Suppose a camera records a leaf at 08:00 on successive days. The pictures look similar. This is compatible with a daily rhythm, but it is also compatible with a leaf that scarcely moved. Observing once per day cannot reveal what happened between pictures. A repeating measurement schedule must not be confused with evidence of a repeating biological process.
Now imagine measurements across several complete days showing repeated rises and falls. The pattern is stronger evidence of rhythmic variation. Yet the cause remains open: the organism may contain an oscillator, respond directly to changing illumination, or combine both processes. The first graph describes an output; it does not identify the entire mechanism.
The Society for Research on Biological Rhythms places daily, annual and tidal patterns within the field. That breadth matters. Chronobiology is not simply another name for human sleep science, and a 24-hour rhythm is not the only kind of biological timing worth explaining.
Write a question that can fail
Replace “Does this plant know the time?” with “Does the measured leaf-angle rhythm continue after the repeating light cue is removed under suitable controlled conditions?” The second question identifies an observation that could challenge the proposed explanation.
For a learner, the first task is therefore a sentence with four parts: the organism or preparation, the measured variable, the time reference and the comparison. Once those parts are clear, the vocabulary of clocks becomes much easier to use precisely.
CHAPTER 2 OF 20 · Define and measure biological time
2. Keep period, phase, amplitude and mean separate
Period is the elapsed time for a complete cycle. Phase locates a point within that cycle relative to a stated reference. Amplitude describes the size of the rhythmic variation. The mean level describes the signal’s central level. Changing one does not require all the others to change.
Consider the fictional model y(t) = 50 + 10 cos[2π(t − 6)/24], with t measured in hours and y in arbitrary units. Its period is 24 hours. Its maximum occurs at t = 6 hours, then again at 30 hours. Its mean over a complete cycle is 50. Its amplitude is 10, while its peak-to-trough range is 20. The units and definitions are part of the answer.
At t = 6, the cosine is 1, so y = 60. At t = 18, it is −1, so y = 40. At t = 0 and t = 12, it is 0, so y = 50. Four calculations show why amplitude is not the same as the largest recorded value.
If the constant 50 becomes 70, the whole curve moves upward without changing its timing or amplitude. If 10 becomes 5, the oscillation becomes smaller without changing its period. If the peak moves from hour 6 to hour 9, the model has a three-hour delay relative to the original reference.
A useful reporting repair
“Treatment increased the rhythm” is ambiguous. Did it raise the mean, enlarge the amplitude, improve regularity or change the measured peak time? State the actual result. This distinction is useful even before formal statistical modelling: it prevents a visually impressive graph from carrying a claim that its axes do not support.
CHAPTER 3 OF 20 · Define and measure biological time
3. Distinguish a clock from a driven pattern
A lamp switched on every morning can produce a daily pattern in a sensor with no internal timekeeping mechanism. An organism can likewise show repeated behaviour because its environment repeatedly prompts it. A circadian oscillator, by contrast, can maintain an approximately daily rhythm without a continuing daily cue.
The central comparison is between conditions with a repeating environmental signal and suitably controlled conditions without that signal. Persistence supports endogenous timing. It does not make every alternative explanation disappear: temperature cycles, handling schedules, measurement disturbances and other cues must also be considered.
Imagine two fictional devices. Device A lights only while an external switch is on. Device B contains a timer that continues cycling when the switch stops changing. During a repeating switch schedule, their outputs could look identical. Only the changed condition separates the explanations. The analogy clarifies experimental logic; living clocks have more complex inputs and outputs than either device.
What a negative result leaves open
If a measured rhythm disappears, the clock may have stopped, the output may have become disconnected from it, the preparation may be unhealthy, or the measurement may be insensitive. A failed output is not automatically a complete account of the underlying mechanism.
This is why strong work combines time-series evidence with suitable controls and, where available, independent measurements. A second output can help distinguish a general timing change from a fault in one reporting pathway. The confidence of the conclusion should match the number of alternatives actually tested.
Use “daily pattern” when that is what the data establish. Reserve a stronger claim about an endogenous circadian mechanism for evidence that addresses its persistence and regulation. Precision makes the explanation more useful, not less interesting.
CHAPTER 4 OF 20 · Define and measure biological time
4. Calculate free-running drift without mixing clocks
Free-running period is measured when an oscillator is not locked to a repeating external time cue. It is often represented by τ. Environmental cycle length is commonly represented by T. The distinction matters when estimating how timing differences accumulate.
Suppose an invented oscillator has τ = 24.4 hours, and a reference clock repeats every T = 24 hours. Both have a marker at elapsed time zero. The next marker from the oscillator occurs at 24.4 hours, then 48.8, then 73.2. Each successive marker is 0.4 hour later than the corresponding reference marker.
After ten oscillator cycles, its marker occurs at 244 hours. The tenth reference marker occurred at 240 hours. Their corresponding-event difference is four hours. This statement counts cycles explicitly; it does not casually equate ten biological cycles with ten 24-hour days.
At the exact elapsed time of 240 hours, the oscillator has completed 240/24.4 = approximately 9.836 cycles, while the reference has completed ten. The deficit is approximately 0.164 cycle. Expressed as an angle, it is about 59 degrees of one cycle. If expressed in hours on a 24-hour phase scale, it is approximately 3.93 phase-hours. Different reporting conventions give slightly different-looking numbers because they answer different questions.
Use an event table before a shortcut
Write the observed elapsed times and subtract successive events. If peaks are at 5.0, 29.4, 53.8 and 78.2 hours, the three intervals are all 24.4 hours. Subtracting the first peak from the last and dividing by three also gives 24.4. Dividing by four would be wrong: four peaks define three intervals.
That interval-counting error is small on the page and large in its effect. Always distinguish the number of observations from the number of completed cycles between them.
CHAPTER 5 OF 20 · Follow clock mechanisms
5. Follow the evidence from genes to feedback
Biological clocks became more understandable when researchers could alter components and examine what happened to timing. In their 1971 fruit-fly clock-mutant study, Konopka and Benzer reported arrhythmic, short-period and long-period mutants affecting emergence and locomotor rhythms. The result connected inherited changes with measurable timing phenotypes.
Later, Hardin, Hall and Rosbash’s 1990 study connected rhythmic period RNA with feedback involving its protein product. The general lesson is a loop: molecular production changes a component that eventually influences subsequent production. The biological details determine how that loop behaves.
A useful conceptual model has production, accumulation, inhibition and removal. If a regulator accumulates gradually, its effect can arrive after production has already continued for some time. When the regulator is removed, inhibition can weaken and production can resume. Delays, reaction rates and nonlinear responses matter; merely drawing a circular arrow does not demonstrate an oscillator.
Distinguish a component from a complete explanation
Suppose an invented mutation changes both a reporter’s brightness and its apparent peak time. The observation may reflect a timing change, but it could also involve altered reporter production or degradation. A causal claim becomes stronger when other measurements agree and the proposed mechanism predicts the change.
Likewise, a gene associated with a rhythm is not automatically the sole clock gene. It may receive rhythmic input or affect an output. The reasoning task is to locate the component within the causal sequence, then identify the evidence that supports that location.
For the broader study of neural cells and circuits, continue through The Gold Standard of Neuroscience. Here the narrower question is how a mechanism generates and regulates time.
CHAPTER 6 OF 20 · Follow clock mechanisms
6. Recognise different molecular solutions to timing
Not every biological oscillator is built from the same molecular parts. A particularly clear experimental result came from Nakajima and colleagues in 2005: a sustained cycle of KaiC phosphorylation was reconstituted with the cyanobacterial proteins KaiA, KaiB and KaiC plus ATP. Its period showed temperature compensation, and mutant timing in the preparation corresponded with timing measured in cells.
The importance of this example is conceptual. A biochemical preparation can express a timing cycle without requiring a new round of gene transcription for every cycle. It prevents the overgeneralisation that all biological timekeeping must be explained as the same transcription–translation loop.
Reconstitution is also a distinctive form of evidence. Instead of only removing a component from a complicated organism, researchers assemble a smaller system and ask what it can generate. A sufficient set of components under specified conditions is informative, but it does not establish that the intact organism contains no additional regulation.
Separate the minimal system from its environment
Imagine a model oscillator that keeps cycling on a laboratory bench but drifts when connected to a noisy input. The isolated result and the connected result answer different questions. One concerns the ability to oscillate; the other concerns how that oscillator behaves within a larger system.
When reading a paper, identify what has been removed to make the mechanism visible. Has the study simplified tissue organisation, environmental variation, metabolism or communication between cells? That simplification can be the experiment’s strength. Its limits become a problem only when the simplified result is presented as a complete description of every living context.
CHAPTER 7 OF 20 · Follow clock mechanisms
7. Treat temperature compensation as a measured property
Temperature compensation means that clock period remains relatively stable across a relevant temperature range. It does not mean that temperature has no effect on the organism, that every clock variable remains unchanged, or that the mechanism works indefinitely at any temperature.
To see the distinction, use fictional measurements. A preparation has a period of 24.0 hours at 20°C and 23.6 hours at 30°C. Its oscillation frequency is the reciprocal of period. The frequency ratio is therefore (1/23.6)/(1/24.0) = 24.0/23.6, approximately 1.017. Over this ten-degree interval, that corresponds to a frequency Q₁₀ of about 1.017.
Compare a separate hypothetical reaction whose rate doubles across the same interval. Its Q₁₀ is 2. The comparison illustrates relative temperature stability, but it does not explain which molecular reactions produced that stability. A measured property and its mechanism remain distinct scientific jobs.
Check uncertainty before declaring perfect compensation
Suppose each period estimate has appreciable uncertainty. A small apparent difference might be compatible with measurement variation. Conversely, a stable period could coexist with reduced signal amplitude or deteriorating preparation health. Reporting only the preferred summary could hide an important change.
Record the temperatures actually tested, the number of independent preparations, the observation duration and the other variables monitored. Do not extrapolate beyond that range. A graph with two temperatures can support a limited comparison; it cannot map the whole thermal response of a living system.
The Kai-protein experiment in the previous chapter provides a concrete research example of temperature-compensated oscillation. The invented calculation here is an exercise in interpreting a rate ratio, not a reproduction of that experiment’s data.
CHAPTER 8 OF 20 · Follow clock mechanisms
8. Understand entrainment as a stable timing relationship
Entrainment occurs when an oscillator becomes locked to a repeating external cycle with a stable phase relationship. A time-giving environmental cue is called a zeitgeber. The important evidence concerns the relationship over repeated cycles, not simply the fact that an organism responds when a cue appears.
Suppose an invented preparation begins its measured activity two hours after a daily cue for many cycles. A second preparation begins five hours after the same cue. Both may be entrained to the same 24-hour cycle while having different phase angles. Synchronised period does not require simultaneous activity.
A simple event-time model makes the balancing requirement visible. Let eₙ be the timing offset at cycle n, δ the drift that would occur without correction, and c(eₙ) the cue-induced correction. Write eₙ₊₁ = eₙ + δ − c(eₙ). A fixed offset requires c(e*) = δ. This is a teaching model, not a universal equation for organisms.
If δ is 0.4 hour and the cue can provide a suitable 0.4-hour advance, drift can be balanced in this model. If its largest advance is only 0.2 hour, the stated model cannot maintain that fixed relationship. The numbers expose why entrainment has limits rather than being a magical reset command.
Stability is a second requirement
Even where drift and correction balance, a small disturbance must be considered. Does the next correction move the offset back towards its previous value, or move it further away? A balance point can exist without being stable.
For example, near a fixed point let c(e) = δ + 0.5(e − e*). The remaining error is halved each cycle. If the starting displacement is two hours, subsequent displacements are one hour, half an hour and a quarter of an hour. This describes convergence within the assumed model, not a prediction of how quickly a person adjusts after travel.
CHAPTER 9 OF 20 · Interpret changes across the system
9. Read phase shifts relative to biological time
The effect of a cue can depend on the oscillator’s phase when the cue arrives. A phase-response curve summarises the relationship between stimulus timing and the resulting change in phase. It must identify the preparation, stimulus and phase reference; it is not a universal timetable.
Khalsa and colleagues’ human laboratory study measured light-induced shifts using biological phase markers and found phase-dependent advances and delays. The result supports a mechanism in which the state of the receiving system matters. It does not justify copying a laboratory exposure schedule into everyday treatment advice.
Here is an invented calculation that isolates the reasoning. Without an intervention, a marker is predicted at elapsed hour 102. The observed marker occurs at hour 100. Relative to that prediction, it advanced by two hours. If someone instead compares it with a wall-clock reference at hour 96, they may report a four-hour delay. Both subtractions are arithmetically correct, but they use different baselines and answer different questions.
The causal effect of the intervention is the difference from what would otherwise have happened, not merely the difference from a convenient civil-time label. That counterfactual often needs a control condition or a defensible prediction from earlier data.
Record the sign convention
Some graphs assign positive values to advances; other analyses express event-time changes with earlier events as negative. State the convention before interpreting a curve. “A shift of −2 hours” is incomplete unless the reader knows what the sign means.
This is also where Photobiology becomes useful: characterising a light input and characterising its timing are complementary tasks. Colour names or clock time alone do not fully define the biological comparison.
CHAPTER 10 OF 20 · Interpret changes across the system
10. Separate masking from a change in the clock
A stimulus can alter a measured output immediately without changing the oscillator that helps organise it. This direct alteration is called masking. It matters whenever behaviour is used as a convenient window into biological timing.
Consider a fictional activity record. Activity falls sharply during an environmental disturbance and returns afterwards. That immediate fall is not enough to establish a clock shift. To examine timing, compare the later sequence of activity markers with its predicted continuation, while considering other explanations.
Three possible records illustrate the distinction. In the first, activity is suppressed only during the disturbance and subsequent markers resume on the old trajectory. In the second, subsequent markers remain displaced but keep approximately the old interval between them. In the third, the intervals themselves change. These observations suggest, respectively, a transient output effect, a phase change, or a period change—subject to the quality of the controls and measurements.
Do not decide from one point
A single low reading could be suppression, a shifted trough, sensor failure or random variation. Several observations before, during and after the event are needed to distinguish those possibilities. Plotting only the most striking interval can make the wrong explanation seem obvious.
The repair is to preserve the full sequence and ask what each explanation predicts next. If the hypothesis is a lasting phase shift, the displacement should remain visible after the immediate disturbance ends. If the hypothesis is a period change, the offset should continue accumulating. If the hypothesis is equipment failure, an independent measure or reference channel may reveal it.
Chronobiology becomes especially useful here because it treats the trajectory as evidence. The important result may be how a system returns, rather than how dramatically it first reacts.
CHAPTER 11 OF 20 · Interpret changes across the system
11. See an organism as a network of clocks and outputs
In mammals, the suprachiasmatic nucleus, or SCN, plays a coordinating role in circadian timing and receives information related to light from the eyes. Many other tissues also have circadian rhythms. The NIGMS overview provides an accessible introduction to this distributed organisation.
Coordination should not be imagined as every organ doing the same thing simultaneously. Different functions can maintain different phases while sharing a common daily organisation. A timing relationship can be orderly precisely because its components peak at different times.
Evidence that peripheral timing can respond differently comes from Damiola and colleagues’ mouse study. Changing feeding time shifted rhythmic gene expression in peripheral tissues without an equivalent shift in the SCN measurements. This is evidence about the studied preparation and conditions, not a universal instruction about when humans should eat.
A network comparison
Imagine two fictional signals. A peaks at hour 4 and B at hour 10, so B follows A by six hours. After a change, A peaks at hour 6 and B at hour 12. Both shifted by two hours, while their internal separation stayed six hours.
In a different case, A remains at hour 4 while B moves to hour 14. The separation has become ten hours. Calling both cases “the clock shifted” loses the system-level distinction. One is a common displacement of the measured signals; the other changes their relationship.
Choose measurements to match the claim. One signal cannot establish that all tissues shifted together. The Gold Standard of Physiology extends this reasoning into coordinated living function, where a useful explanation connects local changes with the whole system.
CHAPTER 12 OF 20 · Interpret changes across the system
12. Explain how a flat average can hide strong rhythms
A population-level signal combines many components. If those components peak at different times, their variations can partly cancel. The average may become weak even when individual components continue oscillating strongly.
In Welsh and colleagues’ fibroblast study, individual cells maintained rhythms while their phases became dispersed. This helped explain damping in the combined signal. The finding shows why the level of measurement matters: a culture average and a single-cell record can answer different questions.
Use two invented signals to see the arithmetic. Let y₁ = 50 + 10 cos θ and y₂ = 50 + 10 cos(θ + π). Because cos(θ + π) = −cos θ, their mean is exactly 50 at every point. Each signal has amplitude 10, but the mean has amplitude zero. Nothing in this example stopped oscillating.
Now reduce the phase separation to six hours on a 24-hour cycle, equivalent to π/2 radians. The average of two equal-amplitude cosines has amplitude 10 cos(π/4), approximately 7.07 units. It is smaller than either component, but not flat. The mean’s amplitude depends on their relative timing as well as their individual strengths.
Diagnose the correct level
When a tissue signal weakens, ask whether individual amplitudes declined, phases spread out, cell numbers changed or the recording system lost sensitivity. These mechanisms can produce superficially similar averages and call for different explanations.
The mathematics does not prove which mechanism occurred in a real experiment. It proves that one tempting inference is invalid: a flat average alone cannot establish that every component is flat. To distinguish the possibilities, obtain evidence at the level where their predictions differ.
CHAPTER 13 OF 20 · Connect organisms and measurement
13. Connect plant timing with day length and resources
Plants provide a useful route beyond human examples. Their timing must be interpreted alongside light, carbon gain, development and seasonal conditions. A clock can help coordinate a response with an expected environmental cycle, but the benefit depends on the biological and environmental context.
Dodd and colleagues’ 2005 Arabidopsis study compared plants with different clock periods under external light–dark cycles. Matching internal and external periods was associated with advantages in the measured photosynthesis, growth and survival outcomes. The evidence concerns those plants and conditions; it is not a universal percentage gain for crops.
Seasonal timing adds another problem: how can a daily system help discriminate long and short days? Yanovsky and Kay’s Arabidopsis study examined the interaction between circadian regulation of CONSTANS and light-dependent flowering signals. The timing of light relative to internal state mattered, rather than day length being only a total-energy measurement.
An original coincidence model
Suppose a fictional plant model has a sensitive interval from hour 14 to hour 16 after dawn. Under a ten-hour light interval, that window occurs in darkness. Under a sixteen-hour light interval, it overlaps light for two hours. A rule that requires light during the sensitive interval distinguishes the schedules even before total daily light energy is considered.
The model illustrates a logical possibility. It omits molecular regulation and does not predict flowering in an unspecified species. To use it responsibly, identify which part is established for the organism and which part is a teaching simplification.
For the energy-conversion mechanism itself, continue to photosynthesis: light, carbon and evidence. Chronobiology adds the timing relationships around that mechanism rather than replacing its chemistry.
CHAPTER 14 OF 20 · Connect organisms and measurement
14. Extend the question beyond a 24-hour day
Chronobiology includes rhythms shorter and longer than a day. Ultradian and infradian describe those broad timescales. Seasonal, tidal and lunar timing questions require measurements suited to their own environmental cycles; they should not be forced into a daily explanation by default.
In Zhang and colleagues’ study of the marine crustacean Eurydice pulchra, roughly 12.4-hour swimming rhythms persisted under constant conditions, and manipulations disrupting circadian timing did not abolish tidal timing. The evidence supported a separable circatidal timing mechanism in that species.
The important inference is limited and strong: one measured timing system could persist when another was disrupted. It does not imply that every coastal organism uses the same machinery or that all twice-daily patterns are tidal clocks.
Choose the observation window for the question
An invented signal with peaks every 12.4 hours has peaks at 0, 12.4, 24.8 and 37.2 hours. Over the first 24 hours, it can look roughly twice daily. Over a longer record, the difference from an exact 12-hour period becomes visible. Comparing the tenth corresponding peak gives 124 hours versus 120 hours, a four-hour difference.
For an annual question, several days of data plainly cannot establish recurrence across years. For a shorter rhythm, coarse daily sampling can miss nearly everything. Duration and sampling interval must fit the proposed period.
Field interpretation also needs environmental records. A change in seasonal activity might reflect an internal timing process, an immediate resource response, development or several interacting causes. Record what changed outside the organism before naming the clock inside it. Timing science improves ecological explanation by keeping those alternatives visible.
CHAPTER 15 OF 20 · Connect organisms and measurement
15. Keep sleep pressure, chronotype and performance distinct
Sleep timing is related to circadian organisation, but it is not a direct display of one clock variable. Borbély’s two-process model distinguishes a sleep–wake-history-dependent process from a circadian process. The original paper explains how their interaction can account for patterns that either process alone would not capture.
That distinction is enough to prevent a common mistake: two people tested at the same civil time may differ in prior sleep, time awake, internal phase, task experience and circumstances. A difference in their performance cannot be assigned to circadian timing merely because the test occurred in the morning.
Chronotype concerns patterns of timing preference or behaviour assessed using particular methods. It is not a certificate of intelligence, motivation or a permanently optimal study hour. The site’s Chronotype–Time-of-Day Fit guide explores the narrower learning decision and its confounders.
Interpret a study comparison carefully
Imagine a fictional learner scoring 70% on one morning exercise and 85% on one evening exercise. The fifteen-point difference does not identify its cause. Perhaps the evening questions were easier, the topic was practised earlier, the room was quieter or the morning followed insufficient sleep.
An informative comparison needs comparable tasks, repeated observations and records of relevant conditions. Even then, a practical scheduling observation is not a diagnosis of an internal clock. Keep the outcome specific: accuracy on this task under these conditions.
Do not create sleep loss or other harmful conditions to make a personal experiment more dramatic. Ordinary learning records can support modest decisions without pretending to replace specialised research. Persistent sleep difficulties or health concerns require an appropriate healthcare assessment, not an inference from the teaching models here.
CHAPTER 16 OF 20 · Connect organisms and measurement
16. Build a time series that can answer the question
A useful record includes elapsed time, civil time and timezone where relevant, the measured variable, units, preparation identity and environmental conditions. Preserve the original timestamps. Changing clocks, missing observations or inconsistent labels can create apparent phase changes before any biology is analysed.
Sampling design matters as much as the final graph. In the fictional model y = 50 + 10 cos(2πt/24), a measurement every 24 hours always returns 60 if the first observation is at t = 0. The sampled record looks constant even though the underlying model varies between 40 and 60.
Sampling every 12 hours from that same starting point returns 60, 40, 60, 40. But starting at t = 6 hours returns 50, 50, 50, 50. This exact example shows why a minimal-looking sampling schedule can be misleading: its phase relationship to the signal matters. Real signals with noise, sharp features and multiple frequencies need more careful design.
Count independent units correctly
One culture measured one hundred times is one repeatedly observed culture, not one hundred independent biological preparations. Repeated observations describe its trajectory. Independent preparations help assess whether the result generalises beyond that trajectory. Both are useful, but they supply different kinds of information.
Missing values also need interpretation. If a sensor fails preferentially during high activity, the missingness may distort the inferred shape. Filling every gap with a straight line can conceal that problem. Retain an explicit missing-data record and explain any processing choices.
A good analysis begins with a plot of the unprocessed sequence and a log of conditions. Before selecting a rhythm model, look for abrupt instrument changes, long gaps, baseline drift and differences among preparations. A polished fitted curve cannot repair an ambiguous measurement history.
CHAPTER 17 OF 20 · Model, diagnose and transfer
17. Fit a model without letting it invent the biology
A cosine model is a compact way to describe a smooth repeating signal. Its fitted mean, amplitude and peak phase can be useful summaries. They remain model parameters: the real waveform may be asymmetric, have several peaks, change over time or contain unrelated trends.
For the model y(t) = M + A cos[2π(t − φ)/T], define M, A, φ and T before interpreting them. A positive A makes φ the peak time in this convention. If a different formula uses a phase angle with the opposite sign, the reported parameter may look different even for the same curve.
A mean can depend on when you sample
Using the chapter 2 model, observations at hours 6, 7 and 8 are approximately 60, 59.66 and 58.66. Their average is about 59.44, far above the true full-cycle model mean of 50. The discrepancy is produced by sampling near the peak, not by an arithmetic error.
This explains why an ordinary mean from irregularly clustered observations need not equal a rhythm-adjusted model mean. The fitted estimate also depends on whether the model is suitable and whether the data cover enough of the cycle to identify its parameters.
Separate detection from comparison
Finding evidence of a rhythm in group A and failing to find it in group B does not, by itself, establish a statistically supported difference between the groups. Sample sizes and uncertainty may differ. The comparison needs a method that addresses the proposed difference directly.
Likewise, an exact fitted period of 24 hours may have been imposed by the analyst. Ask whether the period was estimated or fixed. Model choice should be reported clearly enough that a reader can distinguish an empirical finding from an assumption used to obtain the curve.
CHAPTER 18 OF 20 · Model, diagnose and transfer
18. Solve a changed-context timing case
The task
A fictional research team measures a molecular reporter in two preparations. Preparation A has successive peaks at elapsed hours 4.0, 28.2, 52.4, 76.6 and 100.8. Preparation B has peaks at 10.0, 34.2, 58.4, 82.6 and 106.8. All values are invented and treated as exact for the arithmetic exercise.
Answer four questions. What period does each record show? What is their phase separation? How late is A’s fifth peak relative to a 24-hour reference sequence beginning at hour 4? If the two reporter signals are ideal sinusoids with equal means of 50 and amplitudes of 10, what happens to the amplitude of their average when their separation is six hours on a 24.2-hour cycle?
Then change the context. A later intervention reduces the population-average amplitude almost to zero. The team concludes that every individual clock stopped. What additional evidence would distinguish that claim from a loss of synchrony? Finally, suppose A’s first post-intervention peak occurs at hour 123.0, while its unaltered continuation predicts hour 125.0. State the phase shift relative to that prediction.
Answer and reasoning
Each adjacent interval is 24.2 hours. Across A’s full record, (100.8 − 4.0)/4 = 24.2 hours; across B’s record, (106.8 − 10.0)/4 gives the same result. There are five peaks but only four intervals. B is consistently six hours later than A at corresponding peaks.
The reference sequence for A is 4, 28, 52, 76 and 100 hours. A’s fifth peak is therefore 0.8 hour, or 48 minutes, later than the fifth reference peak. It is not 1.0 hour late: four intervening cycles accumulated the difference.
For equal sinusoidal amplitudes, the average amplitude is 10 times the absolute value of cos(π × 6/24.2), approximately 7.12 units. The six-hour separation is slightly less than a quarter of this 24.2-hour cycle. Using the 24-hour answer of 7.07 would ignore the changed condition.
| Quantity | Answer | Reason |
|---|---|---|
| A and B period | 24.2 hours | Four equal inter-peak intervals |
| B relative to A | Six hours later | Compare corresponding markers |
| A fifth-peak offset | 0.8 hour later | 100.8 minus 100.0 |
| Average amplitude | Approximately 7.12 units | Equal sinusoids separated by six hours on a 24.2-hour cycle |
| Post-intervention shift | Two-hour advance | 123.0 compared with predicted 125.0 |
The nearly flat population average does not prove that individual clocks stopped. Single-preparation or single-cell records could reveal continuing oscillations with dispersed phases. Independent checks of reporter function and preparation health would help distinguish timing changes from measurement failure. The two-hour advance is relative to the stated unaltered prediction; a later sequence would help establish whether this was a persistent phase change, a changed period or an isolated anomalous peak.
CHAPTER 19 OF 20 · Model, diagnose and transfer
19. Diagnose misleading claims and repair the next step
The most useful correction identifies the missing comparison. “Be more careful” does not tell a learner whether the problem lies in the time reference, the biological level, the measurement or the causal inference.
| Claim | Missing distinction | Useful repair |
|---|---|---|
| It repeats daily, so it is an internal clock | Endogenous timing versus repeated forcing | Examine suitable constant-condition evidence and residual cues |
| Both have a 24-hour period, so they are aligned | Period versus phase | Compare a defined marker against the same reference |
| The signal got bigger, so amplitude increased | Mean level versus oscillation size | Report both mean and amplitude with units |
| The culture average is flat, so every cell stopped | Component behaviour versus population average | Examine individual trajectories and phase distribution |
| A peak moved, so the intervention reset the clock | Immediate output effect versus persistent change | Examine later cycles against a control prediction |
| The model found 24 hours | Estimated parameter versus imposed assumption | State whether period was fitted or fixed |
| One hundred readings mean one hundred replicates | Repeated measures versus independent units | Identify the independent preparation or participant |
| One evening score proves a night-person advantage | Timing versus task and context confounding | Use comparable repeated observations without sacrificing sleep |
Questions readers often ask
Is a circadian rhythm exactly 24 hours? Not necessarily when free-running. “Approximately daily” is the key idea; entrainment can align the expressed timing with an environmental cycle.
Can two rhythms share a period but have different phases? Yes. The chapter 18 preparations have identical intervals and a six-hour separation. Period tells you how often; phase tells you where within a reference cycle.
Does a wearable directly read the biological clock? A movement-based record measures movement and supports inferences about activity patterns. It is not identical to a molecular or physiological phase measurement. Interpret the output at the level the device actually measures.
Does knowing a mechanism establish a treatment or school policy? No. A mechanism can motivate a question. Decisions about outcomes, benefits, harms and feasibility need evidence suited to the particular intervention and population.
Is a smooth graph automatically good evidence? No. Ask about timestamps, sampling, missing data, independent units and alternative explanations. Smoothness is a property of presentation or modelling; it is not a substitute for those checks.
CHAPTER 20 OF 20 · Model, diagnose and transfer
20. Build a learning route that survives a new example
Start by drawing two curves with the same period and different phases. Then draw two with the same phase and different amplitudes. Explain each difference without using the vague phrase “the rhythm changed”. This establishes a vocabulary tied to visible consequences.
Next, work through the event-time arithmetic in chapters 4 and 18. Write elapsed times explicitly and count intervals. Then return to the masking case and predict what later measurements would look like under each explanation. A prediction is a stronger learning check than repeating a definition.
For a data project, use the invented models here or an openly documented teaching dataset. Create a full signal, sample it at different intervals and compare the sampled record with what you know generated it. Change one assumption at a time. There is no need to manipulate anyone’s sleep or expose organisms to unsuitable conditions to learn the central mathematics.
For biological depth, read a primary study with four questions beside it: what preparation was studied, what was measured, what comparison supported the conclusion, and what remained untested? Carry those questions from fruit flies to cyanobacterial proteins, plants and marine animals. The mechanisms change; the discipline of interpretation remains useful.
Continue through the connected subjects
Use How Science Works: Chronobiology for the established subject overview. Follow Photobiology for light–organism interactions, Neuroscience for neural mechanisms and Physiology for coordinated function. The How X Works hub provides routes into other mechanisms and methods.
Sources and how to use them
The linked primary studies establish specific findings: Konopka and Benzer, 1971 on clock mutants; Hardin and colleagues, 1990 on molecular feedback; Nakajima and colleagues, 2005 on reconstituted biochemical oscillation; Welsh and colleagues, 2004 on individual-cell rhythms; and Damiola and colleagues, 2000 on peripheral timing.
Other comparisons come from Khalsa and colleagues, 2003, Dodd and colleagues, 2005, Yanovsky and Kay, 2002, and Zhang and colleagues, 2013. These papers are linked for their particular experiments, not presented as universal rules for all species or people. The original models and calculations in this guide are teaching constructions designed to make assumptions, predictions and limits visible.