The chocolate may not cause the migraine, and you can test your own patterns

Key takeaways

• In a 2003 electronic-diary study of 97 people with migraine, most who noticed early warning signs such as tiredness went on to have a migraine within three days at least half the time, so a craving can be the start of an attack rather than its cause.
• When researchers in Copenhagen exposed 27 people to the bright light or hard exercise they blamed for their migraines with aura, only 3 had that kind of attack.
• A study of more than 11 million Medicare doctor visits found no link between rain and visits for joint or back pain, while a smartphone study found humid, windy, low-pressure days raised the chance of a painful day by about 20%, enough to turn a 5-in-100 chance into about 6 in 100.
• In a 2024 blinded trial, people who believed they were sensitive to gluten had no more symptoms after eating gluten unknowingly than after eating none.
• Before you cut a food or blame a trigger, count the times the pattern failed as well as the times it held, then test it on planned days with everything else held steady.

Ask people with migraine what sets off their attacks, and chocolate often comes up. In one study summarized in a 2020 review, 61.7% of patients had heard that chocolate could trigger a headache, though only 14.3% said it had happened to them¹.

The belief is easy to understand. Picture a hypothetical evening: you crave chocolate, eat a few squares, and wake the next morning with a pounding headache. The chocolate looks guilty.

But a migraine often starts before the pain does. The American Migraine Foundation (AMF) describes an early phase called the prodrome, which can last hours or even days, and it names food cravings as one of the symptoms most typical of that phase². The craving may have been the migraine’s opening move.

When researchers have tested chocolate directly, it has mostly come up empty. In a 1997 study described in the review, 63 people with headaches ate either 60 grams of chocolate or carob, a substitute that tasters couldn’t tell from chocolate, without knowing which. Headaches were no more common after the chocolate¹.

The review’s authors found too little evidence to call chocolate a migraine trigger. They urged doctors not to tell migraine patients to avoid it by default¹.

Our companion piece on correlation and causation covers health headlines. This one is about your own data: a diary, an app, or a hunch built over years. It can teach you a lot, and it can fool you in predictable ways.

Warning signs can arrive days before the pain

In 2003, a team led by neurologist Peter Goadsby, then at University College London, gave 97 people with two to eight migraines a month a hand-held electronic diary³. For three months, they logged non-headache symptoms every day³.

The American Academy of Neurology, which publishes the journal Neurology, summarized the results. Of the people who noticed warning signs, 72% saw a migraine follow within three days on at least half of those occasions³. The most common signs were tiredness (72%), trouble concentrating (51%) and a stiff neck (50%)³. Goadsby said the brain may begin to change as much as three days before the pain³.

That reshapes how to read your own records. Anything you do during those early hours can look like a trigger: reaching for sweets, sleeping badly, snapping at someone, skipping a workout because you feel drained. Researchers call this reverse causation, when the supposed effect is causing the supposed cause.

Your brain counts the hits and forgets the misses

In 1996, Donald Redelmeier and Amos Tversky followed 18 people with arthritis for more than a year⁸. Each named the weather conditions that worsened their pain. The researchers found no statistically significant link between the patients’ pain and those conditions⁸.

Then they showed 97 college students pairs of random sequences that had no connection at all. The students tended to see correlations anyway⁸. The authors proposed that the belief about arthritis and weather grows partly from our habit of finding patterns where none exist⁸.

Here’s one way to see why. Every possible trigger sorts your days into four boxes:

• you were exposed and had symptoms,
• you were exposed and felt fine,
• you weren’t exposed and had symptoms,
• you weren’t exposed and felt fine.

Memory loves the first box. A rainy day with an aching knee feels like evidence. A rainy day with a fine knee leaves no impression, and neither does an aching knee on a sunny day. A real pattern only shows up when you count all four.

Rain, pain and two very different studies

The weather question has been tested at enormous scale. In 2017, a team led by Anupam Jena matched Medicare claims with local rainfall records⁵. They looked at more than 11.6 million visits to general internists by about 1.55 million adults aged 65 or older between 2008 and 2012⁵.

After adjustment, 6.35% of visits on rainy days involved joint or back pain, compared with 6.39% on dry days⁵. The gap was tiny and ran the opposite way from the belief⁵. The authors noted a limit: claims data can’t capture pain that people manage at home without seeing a doctor⁵.

A 2019 study took the opposite approach, asking people directly every day. In the “Cloudy with a Chance of Pain” project, more than 13,000 people across the United Kingdom signed up through a smartphone app, and the team analyzed 2,658 who reported on most days⁶⁷. Most had arthritis; others had fibromyalgia, migraine or nerve pain⁷.

The researchers found small but statistically significant links between pain and humidity, air pressure and wind, with humidity mattering most⁶⁷. Rainfall and, on average, temperature weren’t linked, and mood didn’t explain the results⁷.

On a damp, windy, low-pressure day, the chance of a painful day was about 20% higher than on an average day⁷. The University of Manchester’s illustration: a 5-in-100 chance on an ordinary day would become about 6 in 100⁷.

Lead author Will Dixon noted that about three in four people with arthritis believe the weather affects their pain⁷. Both studies can be right. A real effect can be small, and a strong belief can grow from a small effect plus a memory that keeps only the hits.

Five ways your own records can fool you

The symptom came first

This is the prodrome problem. If tiredness, cravings or irritability are early signs, then whatever you ate or did during them gets the blame²³.

Something else changed at the same time

Picture a hypothetical reader who gets headaches on Saturdays and blames Friday-night pizza. But Saturdays also bring sleeping in, a later first coffee and the drop in stress after a long week. Any of those could be the real factor, or none of them. Researchers call this confounding.

Cycles can line up with habits, too. NINDS, the federal neurology institute, suggests that people who menstruate note period dates in a headache journal¹⁴. That record can help you tell a monthly pattern from a food that happens to show up at the same time.

You went looking when things were at their worst

A widely used 2005 explainer in the International Journal of Epidemiology describes regression to the mean as a statistical effect that can make natural ups and downs look like real change⁹. Unusually high or low readings tend to be followed by ones closer to your average⁹.

People usually hunt for a trigger, or start a remedy, after a bad stretch. The following weeks are likely to look better whatever you change.

You checked so many things that something matched

Every comparison you make gives luck another chance to produce a match. Picture tracking 12 foods against 5 symptoms. That’s 60 comparisons.

If each one has even a 1-in-20 chance of lining up by accident, you’d expect about three false alarms, even if no food matters at all. That’s our own arithmetic, not a study finding: 60 divided by 20. An app that scans dozens of factors will nearly always flag something.

You expected it

Expecting harm can produce real symptoms. This is the nocebo effect, the flip side of placebo, and the gluten studies below show its strength.

When suspected triggers face a fair test, many fade

In 2013, Danish researchers including Anders Hougaard and Jes Olesen of the University of Copenhagen recruited 27 people whose migraines come with an aura, a phase that can bring flashing lights, shimmering patterns or blind spots²⁴. Each said bright or flickering light, hard exercise, or both, set off their attacks⁴. The researchers exposed them to those triggers in the lab, including intense exercise and 30 to 40 minutes of bright, flashing or flickering light⁴.

Only about 11% (3 people) developed a migraine with aura, and another 11% had a migraine without aura⁴. Light by itself brought on no aura attacks at all⁴. A lab session may not copy real life, where several stresses can pile up. Still, the results suggest that a trigger people feel sure about often fails a direct test.

In an editorial alongside the study, Goadsby suggested that some behaviors blamed as triggers may be the brain responding to an attack already under way⁴.

Gluten shows the same pattern. In a 2013 study by Jessica Biesiekierski and colleagues, 37 people with IBS who believed they were sensitive to gluten, and who had tested negative for celiac disease, first spent two weeks on a gluten-free, low-FODMAP diet¹⁰. FODMAPs are certain carbohydrates that ferment in the gut. Symptoms improved¹⁰.

Then, without knowing which was which, they ate a high-gluten, low-gluten or gluten-free diet for a week each. Symptoms got worse on all three, by similar amounts¹⁰.

A 2024 trial in the Lancet Gastroenterology & Hepatology separated expectation from gluten directly¹¹. Researchers in the U.K. and the Netherlands randomly assigned 84 adults with self-reported gluten sensitivity to four groups. Each group was told it would eat gluten-containing or gluten-free bread, and each was given one or the other¹¹.

Those who expected gluten and got it scored highest for gut symptoms, 16.6 on a 100-point scale¹¹. Those who were told their bread was gluten-free but got gluten scored 6.9, about the same as people who expected and got none, 7.4¹¹. The authors called this a nocebo effect, though they couldn’t rule out some effect from gluten itself¹¹.

None of this means the symptoms are imaginary. The pain, bloating and fatigue are real. What fails the test is the explanation.

The four-box trigger check

Use this with a diary you’ve already kept for at least a month. Fill it in for one suspected trigger at a time.

First, write down your time window before you look at the data. For example: “a headache within 24 hours of eating chocolate.”

	Symptom within the window	No symptom
Exposed	Box A: ___ days	Box B: ___ days
Not exposed	Box C: ___ days	Box D: ___ days

Then compare two numbers:

• Risk when exposed = A divided by (A + B).
• Risk when not exposed = C divided by (C + D).

This is a hypothetical example. You ate chocolate on 10 days and had a headache after 3 of them, a 30% rate. On the 20 days without chocolate, you had headaches on 6, also 30%. That’s no pattern, even though those three chocolate headaches felt convincing.

If the exposed rate is clearly higher, ask four more questions before believing it:

1. Did early warning signs, such as a craving, come before the exposure?
2. Did something else change on those days, such as sleep, stress, travel or your cycle?
3. Did you start tracking during an unusually bad stretch?
4. How many possible triggers did you check before this one stood out?

A fair pattern test you can run at home

If a suspected trigger survives the four-box check, test it on purpose. This short version borrows from formal single-person trials; our article on self-trials gives a full worksheet.

1. Name one suspect and one symptom. Write the time window and what result would convince you, before you start.
2. Plan the exposure days in advance. Flip a coin each morning, or set a calendar, so your mood that day doesn’t decide.
3. Hide it if you can. For a food, ask a helper to serve it mixed into something, on some days and not others, without telling you which.
4. Hold everything else steady. Keep sleep, caffeine and exercise as regular as you can, and note anything that changes anyway.
5. Rate the symptom the same way every day, at the same time, from 0 to 10.
6. Run enough rounds. One reaction proves little. Aim for several exposure days and several comparison days.
7. Decide, then recheck. If the rule you set is met, adjust, and test again in a few months.

Some situations call for a clinician, not a home test:

• Possible food allergy. MedlinePlus, from the National Library of Medicine, says not to try an elimination diet if a food has ever caused severe symptoms, and never to provoke a reaction or bring back such a food on your own. Allergy challenges belong under medical supervision¹². Wheezing or trouble breathing after eating calls for 911¹².
• Suspected gluten problems. NIDDK says doctors don’t recommend going gluten-free before celiac testing, because the diet can affect results¹³.
• Headache warning signs. NINDS advises seeing a doctor right away for a sudden, severe headache, a headache with weakness or confusion, or a new pattern of headaches that worsen over days or weeks¹⁴.
• Prescription medicines. Never stop or skip one to test a theory without talking to your prescriber.

Why a fair test honors what your body tells you

Paying close attention to your body is a strength, and natural-health traditions have long prized it. A fair test takes that attention seriously enough to check it.

Wrong triggers have costs. People give up foods they love, bend their schedules around weather forecasts, or feel guilt over a square of chocolate. Meanwhile, the real cause can go unnoticed.

A tested pattern is worth keeping, and a disproved one frees you. Either way, you end up trusting your own body more, not less.

What the evidence shows

Claim	Evidence rating	What it rests on
Early migraine signs such as fatigue can appear hours to days before the pain	Promising but early	One diary study of 97 people; AMF patient guidance23
Chocolate is a common migraine trigger	Evidence says no, or unsafe	Small blinded trials mostly found no effect; a 2020 review found insufficient evidence1
Self-reported light or exercise triggers reliably bring on migraine with aura	Evidence says no, or unsafe	3 of 27 people had that kind of attack in a lab test4
Rain increases joint and back pain visits	Evidence says no, or unsafe	More than 11.6 million Medicare visits5
Humid, windy, low-pressure days slightly raise the chance of pain	Promising but early	One large smartphone study; modest effect67
People see patterns in random data, and swings back toward average can look like change	Strong evidence	A classic experiment with random sequences8; a well-established statistical effect explained in an epidemiology journal9
Gluten itself causes symptoms in most people who report gluten sensitivity	Not enough evidence	Two blinded trials point to FODMAPs and expectation; a gluten effect isn’t ruled out1011

Questions readers ask

The bottom line

Your own records are valuable, but they don’t come with labels saying what caused what. Early warning signs, other changes, bad stretches that pass, too many comparisons and expectations can all make an innocent food, habit or weather pattern look guilty.

Count all four boxes, ask the five questions, and test a suspect on planned days with everything else held steady. You’ll cut fewer things out of your life for no reason, and you’ll be more likely to find what really matters.

Related reading

• Chocolate tracks with Nobel prizes, and other lessons in telling a link from a cause
• Feeling better after trying something isn’t a test, but you can run one
• Leaky gut can be measured, but most “heal your gut” protocols haven’t been tested
• Memory rewrites your symptoms, and a diary keeps the record straight
• Your wearable’s best number is your own normal, not the average

Sources

1. Nowaczewska M, Wiciński M, Kaźmierczak W, Kaźmierczak H. “To Eat or Not to Eat: A Review of the Relationship Between Chocolate and Migraines.” Nutrients, 2020;12(3):608. pmc.ncbi.nlm.nih.gov
2. American Migraine Foundation. “Timeline of a Migraine Attack.” AMF, 2018, updated August 2024. americanmigrainefoundation.org
3. American Academy of Neurology. “Migraine Warning Signs Predict Attack” (press release on Giffin NJ, Ruggiero L, Lipton RB, et al., “Premonitory Symptoms in Migraine: An Electronic Diary Study,” Neurology, 2003;60(6):935-940). March 24, 2003. aan.com
4. American Academy of Neurology via ScienceDaily. “Migraine Triggers Less Potent Than Thought” (summary of Hougaard A, Amin FM, Hauge AW, Ashina M, Olesen J, “Provocation of Migraine With Aura Using Natural Trigger Factors,” Neurology, 2013). January 23, 2013. sciencedaily.com
5. Jena AB, Newhouse RL, Olenski AR, Molitor D, Miller N. “Association Between Rainfall and Diagnoses of Joint or Back Pain: Retrospective Claims Analysis.” BMJ, 2017;359:j5326. pubmed.ncbi.nlm.nih.gov
6. Dixon WG, Beukenhorst AL, Yimer BB, et al. “How the Weather Affects the Pain of Citizen Scientists Using a Smartphone App.” npj Digital Medicine, 2019;2:105 (UCL Discovery record). discovery.ucl.ac.uk
7. University of Manchester. “Cloudy With a Chance of Pain: Smartphone Study Shows Pain More Likely on Humid, Windy Days.” October 24, 2019. manchester.ac.uk
8. Redelmeier DA, Tversky A. “On the Belief That Arthritis Pain Is Related to the Weather.” Proceedings of the National Academy of Sciences, 1996;93(7):2895-2896. pmc.ncbi.nlm.nih.gov
9. Barnett AG, van der Pols JC, Dobson AJ. “Regression to the Mean: What It Is and How to Deal With It.” International Journal of Epidemiology, 2005;34(1):215-220. academic.oup.com
10. Brett AS. “Gluten Sensitivity Without Celiac Disease: Real or Not?” NEJM Journal Watch, September 19, 2013 (summary of Biesiekierski JR, et al., Gastroenterology, 2013;145:320). clinician.nejm.org
11. de Graaf MCG, Lawton CL, Croden F, et al. “The Effect of Expectancy Versus Actual Gluten Intake on Gastrointestinal and Extra-Intestinal Symptoms in Non-Coeliac Gluten Sensitivity: A Randomised, Double-Blind, Placebo-Controlled, International, Multicentre Study.” Lancet Gastroenterology & Hepatology, 2024;9(2):110-123 (White Rose repository record). eprints.whiterose.ac.uk
12. MedlinePlus, National Library of Medicine. “Food Allergy.” Medical Encyclopedia, reviewed October 29, 2024. medlineplus.gov
13. National Institute of Diabetes and Digestive and Kidney Diseases. “Diagnosis of Celiac Disease.” NIH, last reviewed October 2020. niddk.nih.gov
14. National Institute of Neurological Disorders and Stroke. “Headache.” NIH, last reviewed March 13, 2026. ninds.nih.gov