What happened six hours ago

The strongest single factor in one of the largest datasets wasn't the pressure right now. It was the drop six hours earlier.

The strongest single finding in one of the largest datasets on weather and headache wasn't what the pressure was at the moment the headache came on. It was what the pressure had done six hours earlier.

The finding comes from the Japanese study with 4,375 users and 336,951 hourly logs over a whole year. Among the factors that were statistically significantly linked to headache in both statistical models, a marked pressure drop six hours earlier was one of the clearest. Not the pressure right now. Not the morning pressure alone. The drop that was already over by the time the headache set in.

The study also found that specific patterns in the pressure over a full six days were linked to headache. So the strongest single signal lay hours in the past. Some of the others lay days in the past. None of them lay in the present, where you'd look for them yourself.

Meteorology measures over a shorter span than that

There's a fixed standard for how you even talk about a "pressure drop". Weather services use the concept of pressure tendency: the net difference between the pressure at the start and end of a period — usually the three hours preceding an observation — plus how the pressure moved along the way: falling, rising, or steady.

Three hours is the period meteorologists normally look back over. The finding that best explained headache in the Japanese dataset lay twice as far back. That's not just outside your own memory. It's outside the timespan even the field's own standard measure of a pressure drop covers.

You're looking in the wrong place

Try checking in while the headache is there. Is the air pressure low? Is it falling? That's the natural place to look. But if the actual trigger was a drop that was already over six hours ago, you're looking in the wrong timeframe. The pressure may well have recovered, be holding steady, or be on its way up, while your head hurts from something that happened while you were sitting in a meeting or asleep.

It's not a flaw in your attention. It's that cause and effect here are separated by hours, and people aren't good at holding that span of time open in memory.

Even a model struggles

That number is worth holding onto when talking about pressure drops and headache. If a statistical model — with access to all the data a person can never keep track of themselves — only reaches 0.59, it's not for lack of effort. Migraine attacks come on unpredictably enough to limit the chances of reacting in time, even with machine help. The model behind the number wasn't new — it had already been built. It was testing it, on a brand-new dataset from two American university hospitals, that produced 0.59.

A delayed, weak signal that only explains part of the story to begin with isn't something a human brain can piece together by thinking back in the evening.

This is why Aneroid doesn't ask straight away

That's the background for why Aneroid, by default, waits 45 minutes after the pressure has moved before asking how you're doing. Not because 45 minutes is a magic number, but because a question asked right at the moment of the pressure change would have the same blind spot as your own memory: it would be looking for the link at the wrong moment.