In late February 2023, a prominent neurogenetics lab at a midwestern university announced a new circadian mutant. The headline in a popular science outlet read, "Clock gene discovered in flies," and the paper showed mutant flies free-running with a period near 24 hours, while wild-type controls behaved normally. The finding seemed clean, a textbook main effect of a single gene. Within a year, two independent labs had tried to replicate it and failed. The phenotype did not appear. One of those labs, after months of troubleshooting, discovered the culprit: a thermostat. Their rearing room ran at 25°C, not 22°C, and that two-degree difference was enough to erase the effect entirely.

A Setpoint, Not a Strain, Made the Headlines

The original study, published in the journal Genetics, described a gene that, when mutated, lengthened the free-running period of Drosophila's circadian rhythm. The authors used standard activity monitors, vials with food, and a 12:12 light-dark cycle. They reported the rearing temperature as 22°C, a common setpoint in many fly labs, but they did not specify where in the room that temperature was measured. The equipment, the protocol, the statistical tests, all looked standard.

When the replication attempt began, the team at a west coast institute ordered the same mutant stock from the Bloomington Drosophila Stock Center. They followed the methods section closely, using the same type of activity monitor and the same analysis software. They ran the assay for ten days, as described. The mutant flies showed a period indistinguishable from wild type. They repeated the experiment, varied the light intensity, checked the food recipe. Nothing worked. After several months, a postdoc on the replication team noticed a discrepancy: the original lab's rearing room was set to 22°C, but the replication lab's room was set to 25°C, a temperature many labs use to speed up development. The difference had not been flagged in the methods, because both labs simply wrote "reared at constant temperature" in their protocols. The replication team contacted the original lab, who checked their thermostat logs and confirmed: the setpoint was 22°C, but the calibration had drifted by about 1.5°C over a year. Once the replication lab lowered their room to 22°C and allowed the flies to acclimate for two generations, the mutant phenotype reappeared. The effect was real, but it was conditional on temperature. The original headline had implied a simple genetic cause, but the actual story was an interaction between a mutation and an environmental variable that most labs treat as a nuisance.

This incident is not isolated. Circadian biology has a long history of temperature sensitivity, and Drosophila are particularly responsive to even small thermal shifts. The replication crisis in psychology and economics has focused on statistical practices and p-hacking, but in model-organism research, the quiet variables are often physical: a thermostat, a shelf position, a seasonal change in building ventilation. As one reviewer put it, "We chase effects across labs, but we rarely chase degrees Celsius."

How a Rearing Room’s Climate Rewrites a Fly’s Clock

Drosophila development is exquisitely sensitive to temperature. At 25°C, the fly life cycle takes about 10 days; at 18°C, it stretches to nearly 19 days. The circadian clock, which is cell-autonomous and molecular, also responds to temperature through a mechanism called temperature compensation. This process keeps the period roughly constant across a range of temperatures, but it is not perfect. The period of the clock can vary by a few minutes per degree, and in some mutant backgrounds, the compensation is impaired.

Most fly labs maintain rearing rooms at 22°C or 25°C, depending on the experiment. The choice is often practical: warmer rooms speed up experiments, cooler rooms slow them down. But the setpoint is rarely treated as a biological variable. Labs record "22°C" in their methods, but they rarely specify whether that is the room air, the shelf surface, or the inside of a vial. Air circulation, incubator heat, and even the number of people in the room can create gradients of several degrees.

In the replication failure described above, the difference was not just the setpoint. The original lab placed their activity monitors on a shelf near a window, where the afternoon sun raised the local temperature by 1–2°C. The replication lab used an interior room with stable HVAC. The flies experienced different thermal histories, even when the nominal temperature was the same. This is not an exotic scenario; it is a routine reality in aging buildings.

Circadian output is measured via locomotor activity, typically by placing individual flies in glass tubes with a small infrared beam. The flies move, and the beam breaks are counted over days. The resulting time series is analyzed to estimate the free-running period in constant darkness. This assay is sensitive enough to detect a 0.5-hour difference in period, but it is also sensitive to subtle environmental perturbations. A fly that is slightly too warm may have a slightly shorter period, regardless of genotype.

Temperature compensation is a well-known feature of circadian clocks, but it is not absolute. In natural populations, flies from warmer climates have slightly different temperature sensitivity than flies from cooler climates. Laboratory strains, which have been in captivity for decades, may have adapted to their rearing temperature. A strain maintained at 22°C for years may behave differently when moved to 25°C, even without any genetic change.

The Original Experiment’s Hidden Variable

In the original study, the mutant flies seemed to free-run at 24 hours, while wild-type controls showed a period of about 23.5 hours. The difference was small but consistent, and the statistics were solid. The authors genotyped the flies after the behavioral assay, which is standard practice, but they did not record the exact temperature in the behavioral room during the assay. They assumed the room was at 22°C, because that was the setpoint on the thermostat.

But the thermostat was on the wall, and the activity monitors were on a bench two meters away. A heat-generating incubator nearby raised the local temperature by 1–2°C. The flies were effectively assayed at 23–24°C, not 22°C. The mutant's period lengthened at that temperature, while the wild type remained within its normal range. The effect was real, but it was a genotype-by-environment interaction, not a main effect.

In the replication lab, the flies were assayed at a true 25°C, and the mutant's period was indistinguishable from wild type. The mutation's effect was masked at higher temperatures. This is a classic example of a conditional phenotype, but the original authors had not tested that condition. They had simply used the standard temperature in their lab, which happened to be 22°C, and they had not measured the actual temperature at the assay site.

This is not a story of fraud or sloppiness. It is a story of how a routine variable can become invisible. The original lab had used 22°C for years, and their equipment was calibrated. But the calibration was done for the room, not for the vials. The difference between 22°C and 24°C is small enough to be ignored in most protocols, but large enough to flip a behavioral phenotype.

Some researchers argue that this kind of environmental sensitivity is a feature, not a bug. If a mutation only shows a phenotype at a specific temperature, that tells us something about the mechanism. But it also means that a single-lab result, without rigorous environmental reporting, is not a reliable foundation for a general claim. The original paper's title implied a universal clock gene, but the data only supported a temperature-dependent effect.

A Replication Attempt That Turned Cold

The replication lab that eventually solved the puzzle had a different culture. They had been burned before by a failed replication in a different field, and they had adopted a habit of logging everything. Every rack in their rearing room had a small data logger, recording temperature and humidity every 15 minutes. They had this system because a previous postdoc had noticed that the room's temperature fluctuated with the building's HVAC cycles, and they wanted to control for that.

When they ordered the mutant stock, they raised the flies at 25°C, their standard. They ran the behavioral assay, saw no effect, and initially assumed the mutation was a false positive. But their data logs showed a clear difference: the original lab's flies had developed at 22°C, and the replication flies at 25°C. They decided to test the temperature hypothesis directly, rearing a new batch at 22°C. The phenotype reappeared.

The cross-lab communication was not immediate. The original lab was defensive, initially suggesting the replication was flawed. But when they checked their own logs, they found that their thermostat had drifted, and the room had been running at 22.5°C instead of 22°C for several months. The drift was small, but it was enough to alter the assay temperature, because the monitors were on a bench near a heat source. The original lab acknowledged the discrepancy and shared their raw activity data.

The replication lab then did something unusual: they published a detailed account of the temperature-dependent effect, with both labs' data. They showed that the mutant's period was longer at 22°C, but not at 25°C, and that the effect was reproducible within each temperature. The original authors later added a note to their paper, clarifying that the phenotype was temperature-sensitive. The community moved on, but the lesson lingered.

This incident echoes a broader pattern in science. A grant reviewer's scoring rubric can shape which behavioral datasets get collected, and a thermostat can shape which phenotypes get observed. In both cases, the procedural details, not the hypothesis, determine the outcome.

Questioning the Temperature-Centric Explanation

While the temperature difference is a compelling explanation, it is not the only possible one. The replication lab's success at 22°C could have been influenced by other factors that covaried with the temperature change. For instance, the flies were allowed to acclimate for two generations at the lower temperature, which might have altered their microbiome or epigenetic state. The original lab's stock might have been inadvertently contaminated or genetically drifted, although both labs used the same stock from the repository. The replication lab also used a different brand of fly food, which could interact with temperature to affect development.

Moreover, the original lab's thermostat drift was not the only difference. The original lab's activity monitors were placed near a window, exposing the flies to natural light cycles that could have affected their circadian rhythm even in constant darkness. The replication lab used an interior room with no windows, so the light regime was more controlled. The original lab's room also had a higher humidity level, which could influence fly physiology and behavior.

Some critics have pointed out that the replication lab's decision to switch to 22°C was based on a single hypothesis, and they did not systematically vary other environmental factors. A more rigorous approach would have been to perform a factorial experiment, varying temperature, humidity, light, and food independently. Such an experiment would have provided stronger evidence for the temperature-specific effect, but it was not done due to time and resource constraints.

Nevertheless, the temperature hypothesis is supported by the fact that the phenotype reappeared when the temperature was lowered, and the original lab's own logs confirmed the drift. The mechanism is plausible: circadian temperature compensation is known to be imperfect, and many mutations affect the clock's temperature sensitivity. The interaction between the mutation and temperature is likely real, but the exact contribution of other variables remains unknown.

The Craft of Controlling What You Don’t Measure

Standard protocols for Drosophila circadian assays list temperature as a constant, often "22°C ± 1°C". But they rarely specify where that measurement is taken. A room's air temperature can differ by 2–3°C from the surface of a shelf, and the inside of a vial can be another degree warmer, due to metabolic heat from the flies. The difference between 22°C and 24°C is within the tolerance of many protocols, but it is outside the tolerance of a sensitive circadian assay.

Some labs have started to treat temperature as a biological variable, not a nuisance. They place data loggers on each rack, sometimes inside empty vials, and they record temperature continuously throughout development and behavior. They calibrate their thermostats quarterly with a NIST-traceable probe, and they log the calibration dates. This is not required by most journals, but it is becoming a standard in labs that do replication work.

Open-source tools have made this easier. Small, cheap temperature loggers can be programmed to record every 10 minutes and export data as CSV files. Some labs share these logs on public repositories, so that a replication attempt can verify the thermal history of the flies. This is a form of environmental metadata, analogous to the metadata that climate scientists use to archive their raw data. A Fortran 77 routine from 1987 might still gate which climate models get archived, but in fly labs, the gating variable is often a temperature sensor.

The craft of controlling what you don't measure is about humility. You cannot control every variable, but you can measure the ones that are likely to matter. Temperature is one of the most likely, because it affects every aspect of fly physiology. Humidity, light intensity, and air flow also matter, but they are less often reported. A reviewer who asks for rearing temperature data is not being pedantic; they are asking for the information needed to interpret the result.

There is a cost to this rigor. Data loggers, calibration probes, and additional metadata collection take time and money. Some labs resist, arguing that their experiments are robust to small variations. But the replication failure described here shows that what seems robust in one lab can be fragile in another. The trade-off is between efficiency and reproducibility, and the balance is shifting.

What This Means for Future Circadian Studies

The immediate implication is that rearing temperature must be reported precisely, with details about where and how it was measured. A single number "22°C" is not enough. The measurement location, the calibration method, and the range of variation over the experiment should be included in the methods section. Some journals have begun to require this for model-organism studies, but it is not yet universal.

Replication audits should request raw environmental logs, not just the final data. If a replication fails, the first question should be: what were the environmental conditions in both labs? This is a simple check that can save months of troubleshooting. The Shapley supercluster's 2017 velocity map fractured when one analyst replotted its redshifts, and a circadian phenotype can fracture when one lab replots its temperature logs.

Meta-analyses of circadian studies could adjust for thermal history, if the data were available. Currently, most papers do not report the temperature during the behavioral assay, only the rearing temperature. This is a gap that could be filled with a standard reporting checklist. Some researchers have proposed a set of minimum environmental reporting standards for fly work, including temperature, humidity, and light intensity, with the exact sensors used.

Funding bodies may start to require environmental metadata as part of data management plans. The National Institutes of Health and other agencies already ask for data sharing plans, and they could extend that to include environmental logs. This would increase the cost of compliance, but it would also increase the value of the data for replication.

The mutant phenotype from the original study may still exist, but it is now understood as conditional. The gene is not a "clock gene" in an absolute sense; it is a gene that, when mutated, disrupts the clock at 22°C but not at 25°C. This is a more nuanced and interesting finding. The focus shifts from a main effect to an interaction, and that is a more realistic picture of biology.

A Practical Checklist for Fly Labs and Reviewers

For labs that study circadian rhythms in Drosophila, the first step is to calibrate thermostats quarterly with a NIST-traceable probe. This is a simple procedure that can be done in an afternoon, and it prevents the slow drift that caused the original discrepancy. The probe should be placed where the flies actually live, not just on the wall.

Second, record temperature at the vial level, not just the room level. A data logger placed on a shelf, inside an empty vial, can capture the microclimate that the flies experience. This is especially important in rooms with incubators, which can create heat gradients.

Third, note seasonal variations in building HVAC. Many labs have rooms that are warmer in summer and cooler in winter, even with climate control. A log of temperature over the course of a year can reveal patterns that affect experiments.

Fourth, share environmental logs with replication requests. If another lab asks for your protocol, include the temperature data. This is a small act of transparency that can prevent a whole class of replication failures.

Reviewers should ask for rearing temperature data as a matter of course. A paper that reports a new circadian phenotype without specifying the temperature at the assay site is incomplete. Reviewers can also ask for the calibration method and the range of temperature variation.

Finally, consider temperature as a biological variable, not a nuisance. The interaction between genotype and temperature is often more informative than the main effect. In the case of the "clock gene" that was not, the temperature sensitivity revealed something about the mechanism. The field would benefit from more studies that deliberately vary temperature, rather than treating it as a constant to be controlled.