Box Breathing for Sleep Onset Anxiety: 4 Weeks, a Phone Stopwatch, and What Actually Happened

It was 1:47 AM on a Tuesday in early June 2026 and I had been staring at the ceiling of my Beijing apartment for what felt like an hour. I knew it had been at least an hour because the battery icon on my phone had dropped two percentage points since I had last looked. I rolled over and grabbed the stopwatch app, the one with the plain black background and the red digits. Forty-three minutes since I had turned off the lamp. Forty-three minutes of doing nothing but breathing and being mad that I was breathing and not sleeping.

That is when I downloaded a breathing app at 1:47 AM and started counting.

I am writing this in late September 2026. I have now used box breathing every single night for four weeks. My phone has a log file. The numbers are not flattering, but they are also not bad. Here is what I did, what happened, what didn’t, and whether I think you should try it.

What box breathing actually is (and why a Navy SEAL gets quoted every time)

Box breathing is the simplest structured breathing pattern I have ever tried. You breathe in for four seconds, hold for four seconds, breathe out for four seconds, hold for four seconds. That is one “box.” You repeat it for four minutes, six minutes, or however long you can stand it. The shape is a square on a graph if you draw it — in on the way up, hold across the top, out on the way down, hold across the bottom.

The reason it gets pitched with a Navy SEAL reference is that Mark Divine, a retired SEAL commander, talks about it in his book “The Way of the SEAL.” He claims operators use it to control their heart rate before a mission. The story is good. I am not in a firefight. I am in my bed at 1:47 AM trying to sleep. But the mechanism is the same: structured breathing, deliberately paced, slightly longer exhale than inhale, and a forced hold that interrupts the spiral of “why am I not asleep yet, now I’m more awake, now my heart is going, now I will definitely not sleep.”

The mechanism is parasympathetic. When you extend your exhale beyond your inhale, your vagus nerve gets stimulated and your heart rate drops. The holds in between buy your brain a moment to stop chasing the next anxious thought. I have read about this in two journal papers and one Healthline article. I will link them at the bottom. The short version is: yes, the science is real. The science is also modest. This is not a cure. It is a way to stop escalating.

The setup: what I tracked and how

I downloaded an app called Breathwrk on June 18, 2026. It is not the only option. There is Box Breathe, Breathing Zone, and about forty others. I picked Breathwrk because it was free and did not require a login. I do not want to give a wellness company my email at 1 AM. I also kept the default phone stopwatch open next to the breathing exercise, because I wanted to measure something.

The thing I measured was “sleep onset latency” — the time from “I am lying in bed with the lamp off” to “I am asleep.” I defined “asleep” as the moment I stopped being aware of my phone screen. This is not rigorous. Sleep latency is normally measured in a lab with EEG. I am not in a lab. I have a phone stopwatch and a guess. But it is a guess I can compare night to night, and that is what I cared about.

Every night for 28 nights — June 18 through July 15, 2026 — I logged three numbers in a notes file on my phone: the date, how long it took me to fall asleep, and whether I did the box breathing or not. I missed two nights. I will tell you which two and why.

The four weeks: the actual log

I am going to give you the actual numbers. Not the curated version. Not the version where I edit out the bad nights to make the technique look better. The bad nights are why I trust the good nights.

Week 1, June 18 to June 24. I did box breathing every night. My sleep onset latency was 47 minutes, 38 minutes, 51 minutes, 29 minutes, 62 minutes, 33 minutes, 44 minutes. The average was 43 minutes. My pre-study baseline, going back to a vague memory of March and April, was probably 70 to 90 minutes. I felt a difference by night three. I also felt impatient by night five. Four minutes is a long time to breathe on purpose when you would rather be asleep. Night six, the 62-minute night, I almost quit. I got up, got water, came back, did not do the second round. The 33-minute night I did two rounds, six minutes total.

Week 2, June 25 to July 1. I did box breathing six nights and skipped one (June 28 — I was at a friend’s wedding and drank three glasses of wine, which both helped me sleep and made the measurement meaningless because alcohol changes everything). The latencies were 41, 28, [skipped], 35, 22, 39, 31 minutes. The 22-minute night was the fastest I had fallen asleep since February. I do not know exactly why. It was a Sunday. I had exercised that morning. I had not looked at my phone after 9 PM. And I had done box breathing for six minutes. Take from that what you want.

Week 3, July 2 to July 8. I did box breathing every night. The latencies were 36, 33, 27, 19, 30, 24, 28 minutes. The 19-minute night was on July 5, and I remember it because I was almost startled that I was not still awake. The trend was clearly down. Average 28 minutes.

Week 4, July 9 to July 15. I did box breathing six nights and skipped one (July 12 — I had a deadline at work the next morning and I just lay there rehearsing my to-do list instead, which is a different problem entirely and box breathing does not fix it). Latencies: 25, 21, 18, [skipped], 30, 22, 17 minutes. The 17-minute night was the fastest of the whole study.

So the trend is real for me. Average latency dropped from 43 minutes in week one to about 22 minutes in week four. The 17-minute night at the end was the outlier in a good direction. I am not going to tell you this is going to happen to you. I am a sample size of one, and the weeks I measured also included June and July, which are longer-daylight months in the Northern Hemisphere and my sleep was generally better anyway. But the within-study trend, week over week, was steady.

When box breathing does not work

I want to give you the honest failures, because the wellness internet will not.

It does not work if your problem is not anxiety. If your problem is that you drank two espressos at 6 PM, no breathing pattern will save you. The caffeine half-life is five to six hours. Box breathing will just make you bored while you wait it out.

It does not work if you are ruminating about something specific. One night in week three, I had a fight with my brother over text. I went to bed at 11 PM. I did the box breathing for six minutes. I lay there and the breathing did nothing because my brain was not in a vague “anxious cloud,” it was in a specific loop: I should not have said that, I should have said this, I should call him in the morning, I should not call him in the morning. The structured breathing was the wrong tool. The right tool, in retrospect, would have been writing the unsaid things in a notes app until the loop broke. That is a different post.

It does not work on the first night. I do not care what anyone tells you. If you have been lying awake for an hour and you start box breathing cold, you are still lying awake and you are also breathing in a square pattern. That is miserable. The benefits in my log were cumulative. The first week was barely better than baseline.

It also does not work if you do it with your phone in your face. The app screen has the breathing circle that expands and contracts. That is a visual anchor, which helps some people. But if you do not turn your screen brightness all the way down, the blue-white light tells your brain it is still daytime. I learned this the hard way on night three. I now keep the phone face-down on the nightstand and just listen to the audio cue. The audio cue is a soft inhale sound, a hold sound, an exhale sound, a hold sound. After about a week I did not need the audio anymore. I just counted in my head.

The actual practical setup (the part you actually want)

If you want to try this, here is the version I wish someone had told me on June 17.

Time it for the first time when you are not exhausted. I practiced for two minutes during a coffee break on June 17, the day before I started the study. If you go straight into a four-minute box breath at 1 AM with no practice, the four-second holds feel impossible. They are not impossible. They feel impossible because your lungs are not used to pausing at full inhale.

Pick a count you can hold. Four-four-four-four is the canonical. I could not hold my breath for four seconds at full inhale on day one. I started with three-three-three-three for the first three nights and moved up. This is not a moral failing. It is a lung capacity thing.

Phone face down. Audio cue only. Brightness all the way down. If you are using an app that requires the screen, lower the brightness to nothing and accept that you will be tempted to look at the notification that just came in. You will look. You will lose five minutes. This is why face-down is better.

Do it for six minutes, not four. I found four minutes was just enough for my brain to start ramping back up the anxious thoughts at the end. Six minutes gave me a buffer. If six feels like too much, start with three and add a minute per week.

If your mind wanders, which it will, that is the whole point. You are not trying to clear your mind. You are trying to give your mind something simple to come back to. The “oh, I am not counting, where was I” moment is the technique working. The breath is the anchor. The wandering is the data.

Who this is actually for

Get it if:

• You lie awake for 30 to 90 minutes most nights
• Your main issue is “I cannot turn my brain off,” not “I cannot breathe” (different problem, see a clinician)
• You can commit to 28 nights. Less than that, in my experience, you will not see the trend
• You are not looking for a replacement for therapy or medication — this is a tool, not a cure

Skip it if:

• You have clinical insomnia that has lasted more than three months
• Your anxiety comes with chest pain, shortness of breath, or panic attacks — those need a clinician
• You are not willing to practice when you are not yet in bed. The first-night failure rate is the main reason people quit
• You want an instant fix. The fastest I fell asleep was 17 minutes, and that took four weeks of practice

What I am still doing

I am still doing box breathing most nights. I stopped logging on July 16 because the act of measuring was making me anxious in a different way. I now just do four to six minutes before sleep and that is it. My sleep onset latency is somewhere between 15 and 30 minutes now, on average, though I have had a 50-minute night last week because of a work thing. I am keeping the practice. The phone is face down. The audio cue is off. I just count.

If you try this and you want to send me your numbers, my email is at the bottom of the about page. I am not a clinician. I am a 31-year-old product manager in Beijing who measures things with a phone stopwatch. But I would read your log.

What Remains When the Noise Stops

Box breathing for sleep onset anxiety is the lowest-cost, lowest-side-effect intervention I have tried in 2026. It is not a cure. It is a tool that worked for me, after four weeks, to take my sleep onset from 70-plus minutes to under 30. The honest downside is that it requires consistency, the first week is rough, and if your problem is rumination rather than generalized anxiety, it will not help. If you have 28 nights and a stopwatch, try it.