A study says exercise helps depression. Another says meditation reduces anxiety. Your friend swears by cold showers.
But does it work for you?
The only way to know is to test it yourself. Not casually, but systematically. That's what an N-of-1 experiment is.
What's an N-of-1 experiment?
In medical research, N refers to the number of participants. Most clinical trials have hundreds or thousands of people. An N-of-1 trial has exactly one: you.
The concept comes from clinical medicine, where doctors sometimes run formal experiments on individual patients to find the best treatment. You serve as your own control group. Instead of comparing your results to other people, you compare your results to yourself under different conditions.
This isn't just guessing or paying attention to how you feel. It's structured experimentation with data collection, defined interventions, and before/after comparison.
Why individual results matter
Clinical trials tell you about averages. "On average, people who took this medication improved by 5 points on the PHQ-9." But averages hide variation.
Some people in that trial improved by 15 points. Some didn't improve at all. Some got worse. The average is 5, but that doesn't mean you will improve by 5.
Mental health interventions have particularly high individual variation. What works brilliantly for one person does nothing for another. Genetics, lifestyle, history, circumstances: all of these affect response.
The only way to find out where you fall is to test it.
The basic structure
A simple N-of-1 experiment has three phases:
A (Baseline): Measure your mental health before changing anything. This is your control period. Take assessments for 2-3 weeks to establish a stable starting point.
B (Intervention): Add the thing you're testing. Keep measuring. Run this phase for at least 4-6 weeks, long enough for effects to show.
A (Return to baseline, optional): Stop the intervention. Keep measuring. If your scores return to baseline levels, that's stronger evidence the intervention was responsible for the improvement.
This is called an ABA design. The second A phase helps rule out coincidence: maybe you improved for reasons unrelated to the intervention.
What you need
A specific intervention. Vague doesn't work. "Exercise more" isn't testable. "Run for 20 minutes, 3 times per week" is testable.
A consistent measure. Use the same mental health assessment throughout. The PHQ-9, GAD-7, or DASS-21 work well. Take it at the same time each week.
Enough time. Most mental health interventions need 4-8 weeks to show effects. Supplements might need longer. Don't conclude too early.
One variable at a time. If you change three things simultaneously, you won't know which one helped. Test one intervention per experiment.
Example: testing morning exercise
Week 1-3 (Baseline A):
Normal routine. No new exercise.
Take GAD-7 every Sunday evening.
Scores: 12, 11, 13 (average: 12, moderate anxiety)
Week 4-9 (Intervention B):
Add 20-minute morning run, Monday/Wednesday/Friday.
Continue weekly GAD-7.
Scores: 11, 10, 9, 8, 7, 8 (average: 8.8, dropping to mild)
Week 10-12 (Return to Baseline A):
Stop running. Return to normal routine.
Continue weekly GAD-7.
Scores: 9, 11, 12 (average: 10.7, climbing back up)
Conclusion: Morning exercise appears to reduce anxiety by about 3-4 points. When stopped, scores return to baseline. This suggests a real effect.
Strengthening your experiments
The basic ABA design is good. These additions make it better:
Multiple cycles. An ABAB design (baseline, intervention, baseline, intervention) provides more evidence. If scores drop during both B phases and rise during both A phases, coincidence becomes less likely.
Longer phases. More data points per phase means more confidence in the pattern. Six weeks per phase is better than three.
Blind yourself (when possible). Hard with exercise, but possible with supplements. Have someone else put your supplement in unmarked capsules alongside identical placebos. Take one each day without knowing which. This controls for placebo effects.
Track confounders. Note sleep, stress, major life events. These help explain outliers and unusual data points.
What N-of-1 experiments are good for
Testing interventions with individual variation:
- Exercise (type, timing, frequency, intensity)
- Diet changes (cutting alcohol, adding fish, eliminating foods)
- Supplements (omega-3, vitamin D, magnesium)
- Sleep changes (bedtime, wake time, caffeine cutoff)
- Routines (meditation, journaling, social time)
Optimizing existing treatments:
- Does taking medication in the morning vs. evening matter?
- Does adding exercise to therapy help?
- What's the minimum effective dose of something that works?
Resolving uncertainty:
- "I think caffeine makes my anxiety worse, but I'm not sure"
- "I feel better when I exercise, but maybe it's just placebo"
- "This supplement might be helping, or I might be wasting money"
What N-of-1 experiments are not good for
Acute conditions. If you're in crisis, you need professional help, not experiments.
Serious symptoms. If your PHQ-9 is 20+, work with a doctor. Self-experimentation is for optimization, not treatment of severe conditions.
Irreversible interventions. You can't undo surgery. You can stop taking a supplement.
Replacing medical care. An N-of-1 experiment might help you and your doctor decide between options, but it's not a substitute for professional guidance.
Common mistakes
Changing too many things. "I started exercising, cut sugar, and began meditating." Great for your health, terrible for knowing what worked. Test one thing at a time.
Concluding too quickly. You feel better after one week of the intervention. That might be real, might be placebo, might be coincidence. Give it 4-6 weeks minimum.
No baseline. You start the intervention immediately and feel better. Better than what? Without baseline data, you don't know.
Inconsistent measurement. You take the PHQ-9 one week, the DASS-21 the next, then skip two weeks. This data is hard to interpret. Same assessment, same schedule, every time.
Expecting too much. A 5-point improvement on the PHQ-9 is clinically meaningful. You might not feel dramatically different, but the data says something real changed.
The deeper point
Most health advice is based on population averages. "Studies show that X helps with Y." But you're not a population. You're one person with your own biology, history, and circumstances.
N-of-1 experiments let you generate evidence about your responses. Not what works for the average person in a study. What works for you.
This takes more effort than just trying things and seeing how you feel. But it gives you actual data: numbers you can look at, trends you can see, conclusions you can trust.
You can run real experiments on yourself. All you need is a clear question, a consistent measure, and enough patience to let the data accumulate.
Getting started
1. Pick one thing to test. Something specific and reversible. "30 minutes of walking, 5 days a week" or "no alcohol for 6 weeks."
2. Choose your measure. GAD-7 for anxiety, PHQ-9 for depression, DASS-21 for both plus stress.
3. Establish baseline. 2-3 weeks of measurement before you change anything.
4. Start the intervention. Keep measuring weekly.
5. Run it long enough. 4-6 weeks minimum.
6. Look at the data. Did scores change by 5+ points? Did they change in the direction you expected?
7. Optional: stop and confirm. Do scores return to baseline? If so, the intervention was likely responsible.
The experiment doesn't have to be perfect. Imperfect data is better than no data. And over time, you'll learn more about what actually moves the needle for your mental health.
Run your experiments with Survey Doctor
Survey Doctor makes it easy to take weekly assessments and track your scores over time. Set up a consistent measurement schedule, run your baseline, start your intervention, and see the data. Start your baseline assessment now.