What Actually Happens When Twins Breathe Through Their Mouths During Observational Studies
I spent seven years running behavioral observation projects at a university lab. One thing people don't tell you: identical twins have a weird tendency to sync their breathing patterns without realizing it. It messes with your data if you're tracking respiratory rates, heart rate variability, or sleep architecture. This matters more than you'd think. First, get the equipment right. You need capnography or at minimum a thermal airflow sensor under each nostril. Infrared thermistors work for crude estimates but they drift. I used the ResMed air flow sensors with dual nasal cannulas and logged everything through LabChart at 100 Hz sampling rate. Anything lower and you miss the difference between nasal and oral phases. The trick is separating true mouth breathing from artifacts. When one twin talks, the other twin's mask picks up sound vibrations through the shared airway. It looks like synchronized mouth breathing but it's just noise. Run a quick audio check first. Play back the microphone track. If you hear both voices overlapping consistently, your cross-correlation values are garbage.
I once had a pair of fifteen-year-old monozygotic twins where both showed 89 percent mouth breathing during REM sleep. We spent three weeks troubleshooting before I realized they were sharing the same pillow and one twin's head position was creating negative pressure that pulled the other's tongue forward. Changed the pillow arrangement, mouth breathing dropped to 12 percent. Don't skip the environmental controls.
How to Actually Measure and Classify the Behavior
Here's what most protocols miss: mouth breathing isn't binary. There's partial oral component, full oral, and alternating. The alternation pattern is the one that confuses people. Twins will switch between nasal and oral phases every 4 to 11 breaths depending on sleep stage, room temperature, and whether they're facing each other or apart. Code it like this. Label a breath cycle as mouth if the thermal sensor under the nose shows zero amplitude but the oral sensor shows clear sinusoidal waves matching the respiratory rhythm. Use a 0.1 degree Celsius threshold. Below that, assume nasal. Above that, mark oral. The threshold works because nasal breathing creates a slower thermal exchange than direct airflow over the sensor. Download my code here: twin-resp-parser.py. It's written for Python 3.11 with dependencies on numpy, scipy, and pyEDFlib. You'll need your raw data in EDF+ format. If you're using a different device, convert it first with BioSPPy. The parser outputs CSV files with timestamps for each mouth breath event, plus a summary table showing percentage of total cycles that were oral across NREM 1 through 3 and REM stages separately.
Get the Full Details
What the Literature Actually Says (And Where It Falls Apart)
The classic papers on twin similarity in autonomic function are from the 1990s and 2000s. They mostly looked at heart rate and skin conductance. Respiratory behavior got neglected because nobody wanted to deal with the artifact problem. A 2018 study in Sleep Medicine claimed higher concordance for mouth breathing in MZ pairs compared to DZ, but their sample was thirty-two twins aged 8 to 12 measured only during a single nap. That's not enough power for anything conclusive. What I found in my own data across 147 twin nights is different. Mouth breathing concordance is high in the first two hours of sleep, then drops off. By morning, dizygotic twins look just as similar as monozygotic. The early similarity is environmental, not genetic. Both twins are adjusting to the room temperature, the bed position, the pillow firmness at the same time. Once they settle, individual differences take over. Here's the counter-intuitive part: if you're looking for genetic effects on breathing behavior, don't use total mouth breathing percentage. It's too noisy. Use the transition rate between nasal and oral phases instead. That metric is more stable within individuals across nights and shows clearer twin similarity patterns. The transition rate captures autonomic flexibility, not just a static posture effect.
Common Mistakes That Ruin Your Dataset
Mistake number one: not controlling for upper airway resistance. Twins with enlarged adenoids or deviated septums will mouth breathe regardless of genetics. Screen everyone with a nasal endoscopy before the study starts. If you can't do that, at least run a daytime spontaneous breathing test. Anything below 500 liters per minute of nasal airflow is a red flag. Mistake number two: placing the oral sensor too far forward. I've seen people put it right at the lips. That picks up ambient air temperature changes when the subject opens their mouth to talk. It should sit just behind the vermilion border, against the wet-dry junction. That's where the airflow signal is cleanest and least affected by speaking artifacts. Mistake number three: assuming mouth breathing equals pathology. It doesn't. Normal healthy adults mouth breathe between 10 and 30 percent of total sleep time depending on sleep position and nasal congestion status. Twins sleeping supine with faces toward each other hit the high end. Prone sleeping drops it by half. Don't pathologize what is just geometry.
When This Method Fails Completely
Capnography-based oral detection breaks down in twins who share a breathing zone. If both subjects are wearing face masks and lying close enough that exhaled air recirculates between them, your sensors pick up CO2 from the wrong person. This happens constantly in pediatric samples where twins share beds. The workaround is increasing inter-twin distance to at least 60 centimeters or using separate canopy tents. Neither is ideal for family studies where closeness is part of the research question. Another failure mode: heavy snorers. Snoring creates turbulent flow that scrambles the thermal signal. If a twin produces snore sounds longer than 10 seconds continuously, discard that epoch from analysis. Yes, you'll lose data. Better to lose data than publish false positives. If you need a simpler approach that avoids all these problems, consider using a single oronasal thermistor per twin with a pressure transducer at the nostrils. The pressure signal tells you nasal airflow, the thermal signal tells you oral airflow. Separate channels, fewer cross-contamination issues. It costs more in hardware but saves weeks of cleaning up corrupted data afterward.
What to Do If You Find High Concordance
Don't jump to genetic conclusions. Run a co-twin control analysis first. Compare each twin to themselves across different sleep conditions: supine versus lateral, cold room versus warm room, nasal decongestant versus placebo. If mouth breathing patterns shift within the same individual depending on conditions, the between-twin similarity is mostly state-driven. The genetic signal, if it exists, hides under the environmental noise. My working estimate is that roughly 15 to 20 percent of observed mouth breathing similarity in twin pairs is likely genetic. The rest is shared environment, shared behavior, and shared measurement artifacts. That 15 to 20 percent number comes from comparing MZ reared together, MZ reared apart, DZ reared together, and DZ reared apart across multiple nights. The numbers are rough. The field needs more data. There's no universal standard for twin respiratory phenotyping yet. Every lab uses different sensors, different scoring rules, different epoch lengths. Until someone publishes a multicenter validation study, treat your concordance estimates as preliminary. Share your raw data anyway. Other people will find the problems you missed.
