Heavy snoring isn't always severe sleep apnea. And quiet snoring isn't always safe.

If you've ever Googled "do I have sleep apnea," you've probably read some version of: "loud snoring is one of the main warning signs." That's true. It's also incomplete, in a way that ends up confusing a lot of people.

Loud snoring increases the statistical probability that you have OSA. It does not, by itself, tell you whether you do, or how severe it is. The two pieces of information are loosely correlated, not the same.

I want to walk through why, because the conclusion changes what you should pay attention to in your own data.

What loudness actually measures

Snoring is sound made by air vibrating soft tissue in the upper airway. The loudness of the sound depends on a few things: how much air is moving, how fast it's moving, how much of the airway is partially blocked, and what tissue is doing the vibrating. None of those map cleanly to "how often is your breathing actually stopping?"

Take two scenarios that produce very different loudness readings:

• Loud, continuous snoring with the airway mostly open. Lots of air moving past lots of vibrating tissue. Annoying for your partner. Often not associated with frequent breathing pauses.
• Quiet, occasional snoring with the airway closing fully. Less sound because the airway is collapsing rather than vibrating. The pauses themselves are silent. Sometimes more clinically concerning than the loud version.

The loudness number, on its own, can't tell those two apart. The breathing-pause count can.

The kind of snoring matters more than the volume

This is the part most loudness-only apps miss. Snoring isn't one acoustic phenomenon — different tissues in the airway produce different snore signatures, and the source matters for what you can do about it.

I'll be specific because we built our model around three categories where the source affects what works:

• Soft-palate snoring (the classic rumble) — back of the roof of the mouth vibrating. Often responsive to side-sleep, weight loss, anti-snore pillows. Position-responsive.
• Tongue-base snoring (lower-pitched, irregular) — the tongue partially blocking the airway when muscle tone drops in deep sleep. More associated with breathing pauses than the palatal kind. Also position-responsive, but with a stronger tilt toward "needs evaluation if frequent."
• Epiglottic snoring (quieter, irregular) — flap above the voicebox catching airflow. The most clinically loaded of the three. If your timeline shows frequent epiglottic events, that's a sleep specialist conversation, not a pillow conversation.

Two more categories sit alongside these but our model doesn't fully separate them yet: nasal-source snores (the whistle, often from congestion or anatomy — the path forward is usually ENT, not positional) and heavy breathing without obstruction (loud-sounding breath through an actually-clear airway, which isn't a problem). Both have lower acoustic confidence at the moment, which is why we'd rather not give them a confident label until the model improves.

The volume of any of these can be high or low. The clinical implication of each is very different. That's the whole reason "loudness alone" isn't enough.

Why this trips up loudness-tracking apps

Apps that focus exclusively on snore intensity — peak dB, total minutes — measure something real. They're useful for tracking whether an intervention reduced sound. They tell you whether your partner is going to be happier this week.

What they can't tell you is what kind of snoring you have, or whether the breathing pauses underneath are improving or getting worse. So the very common pattern of "my snoring sounds quieter, I must be sleeping better" is sometimes right, and sometimes the opposite of what the data is showing. More on that pattern here.

If you've used a loudness-only tracker and felt like the picture wasn't complete — you weren't wrong. The picture isn't complete. The breathing-pause count and the snore type are the missing layers.

What to actually watch in your data

Three numbers, in order of usefulness:

1. BRI (the per-hour breathing-event rate) — the closest single number to "is something wrong?" The loudness can be high, low, or anywhere; this is the one to anchor on.
2. Snore type distribution — what fraction of your snore time is palatal vs tongue-base vs epiglottic. Determines what intervention makes sense.
3. Snore loudness trend — useful as a partner-quality-of-life metric and as a coarse signal, but always read alongside the breathing-pause count.

If only one of those moves and the others don't, you have less information than if all three move together.