Independent New Hampshire Journalism · Northfield, NH

Schizophrenia Runs in My Family. So I Built a Method to Check My Own Mind.

My mother’s brain assembles connections that are not there. Mine assembles more of them than can all be true. Here is the verification discipline I built to tell the difference, and why it ended up running a newspaper.

Every verification rule this newspaper runs on started as a defense against my own brain. That is not a figure of speech. My mother has schizophrenia. I grew up next to a mind that assembled connections which were vivid, detailed, internally consistent, and not there.

I got half of her genes and, as best I can tell, a working copy of the same engine. My brain makes links faster than it can check them. Most days that is an asset in this line of work, because the job is noticing what does not fit. Some days I catch the engine running hot and I do not trust one thing it hands me.

So I stopped trying to think my way out of the problem and built a procedure instead. Rules I follow whether or not I feel like following them. Rules that do not care how certain I am when I sit down. That procedure is now the editorial standard of Granite State Report, and the reason it works on a police chief is the same reason it works on me. It does not ask who is making the claim.

First, the numbers I went and looked up

For years I carried a figure in my head that I had never checked. That is the exact failure this essay is about, so I checked it.

The strongest genetic evidence comes from twins. A Danish study of 31,524 twin pairs, published in Biological Psychiatry in 2018, put the heritability of schizophrenia at 79 percent. That number sounds like a sentence handed down. Read the next line in the same paper and it stops sounding that way: when one identical twin has schizophrenia, the other has it about a third of the time. Same genome, start to finish. Two-thirds of the time, a different outcome. The authors say plainly that genes alone do not determine who gets sick.

For a child of one affected parent, the published estimates disagree, and the disagreement matters. Irving Gottesman, who built the classic family-risk tables, put lifetime risk at roughly 6 percent. Figures repeated across popular health writing run higher, usually 10 to 13 percent. The most recent work is a 2025 Danish register study in The Lancet Psychiatry, covering more than three million people, and its headline finding cuts against the fear: most mental illness occurs in people with no affected close relative at all. I trust the register data over the older compilations, because it counts whole populations rather than families who came to clinical attention. Every one of those estimates, high or low, says the same thing about the person reading them. The large majority do not develop it.

Then there is timing. A meta-analysis of 192 studies covering 708,561 people, published in Molecular Psychiatry, found that schizophrenia-spectrum disorders peak in onset around age 20 and a half, with a median of 25 and a middle band running to 34. Just under half of all cases have begun by 25. I am 38. That is not a guarantee and I am not treating it as one. It does mean the window where most of this happens is behind me, and it means the vigilance I have carried since my twenties can be spent on something more useful than waiting.

The failure that matters is not seeing the pattern

In 1958 the German psychiatrist Klaus Conrad, writing about the earliest stage of schizophrenia, named the experience of finding meaning everywhere. He called it apophany: an “unmotivated seeing of connections” carrying a sense of abnormal significance. The world starts arriving pre-loaded with hints.

Here is where most people misread that, and where I misread it for years. The problem is not the connection. Everyone makes connections. Human beings are pattern machines, and a reporter who cannot see a shape in a pile of records should find other work.

The problem is when you stop looking.

There is a well-worn experiment called the beads task. Two jars, one mostly red, one mostly blue, hidden from view. Beads come out one at a time and you say which jar they are from. The only real measurement is how many beads you need before you commit. A 2016 review in Schizophrenia Bulletin pooled 55 studies and found that people with psychosis committed after roughly one and a half fewer beads than everyone else, and that their odds of deciding on two beads or fewer ran four to six times higher. Researchers call it jumping to conclusions. It is not a failure of intelligence. It is a failure of patience.

The line in that paper that stopped me was smaller than the headline finding. The same reasoning bias turns up in first-degree relatives of people with psychosis. Not patients. Relatives. Sons and daughters and siblings, walking around undiagnosed, deciding early.

Read that definition again, because it is the whole argument. A delusion is not defined by being false. Being wrong is ordinary; I am wrong every week. A delusion is defined by what it does when the counterevidence shows up, which is nothing. The belief has no handle on the outside. Granite State Report has written before about how the mind builds these explanations and why arguing with the conclusion never works, because the conclusion is downstream of a feeling.

Which gave me a test I could operate. Not “is this connection real,” because I cannot always tell from the inside. Instead: what would prove this wrong, and have I gone and looked for it.

The test score does not save you

I have sat for cognitive testing. I know my numbers, and I am not putting them in this essay, because doing so would be the precise move the essay argues against. An unverifiable figure, supplied by an interested party, offered as evidence about that party. If I would not accept it from a state agency I would not accept it from me.

The research is more interesting than any score anyway, and it points the other direction.

In a study of 1,111 American adults published in Behavioural Public Policy, Dan Kahan and colleagues gave people a hard numbers problem inside a two-by-two table. When it was framed as a skin cream trial, high-numeracy subjects did much better. Same table, same arithmetic, reframed as a study of a handgun ban, and the results split by politics. The people best equipped to read the table were the most polarized of anyone in the sample.

The mechanism is what matters. High-numeracy partisans were about 45 percent more likely to get the right answer when it flattered their side, against 25 percent among the low-numeracy group. They did not lose the ability. They deployed it selectively, working hard when the easy answer threatened them and quitting early when it felt good.

That is the beads task in a suit. Stopping the count because the count so far is pleasing. The clinical population and the well-educated partisan are running the same error, and a strong score does not protect you from it. It makes you a faster lawyer for whatever you walked in believing.

The honest weakness: this finding has been contested. A Western European replication found only partial support, and several researchers have challenged how the effect is supposed to work. The narrower claim survives the fight, and it is enough for me. Raw processing power is no defense against motivated stopping.

The method

None of this requires a clinic. It requires admitting that the part of you doing the checking is the same part that generated the thing being checked, then refusing to rely on it. Here is what I run, on stories and on myself.

1. Write the claim as a sentence that could be false. If I cannot say what evidence would kill it, I do not have a claim. I have a mood with a vocabulary.
2. Name the source before believing it, not after. Not “I read somewhere.” The document, the page, the person, the date. If I cannot produce that, the belief goes in a holding pen and stays there.
3. Take one more draw than feels necessary. The beads research says my instinct to stop arrives too early, so I treat the feeling of being finished as a signal to keep going, not to file.
4. Search for the version that beats me. Deliberately run the query that would prove me wrong. Not the one that would prove me right, which I will find, because I always find it.
5. Count how many people the story needs. Coordination requires many people keeping a secret. Incompetence requires nobody doing anything in particular. When both explanations fit the record, the second is usually the one holding.
6. Date the belief and go back to it. A conclusion I have not re-examined in two years is not a conclusion. It is furniture, and I walk around it without looking.

The rules are unglamorous on purpose. Rules that depend on my judgment in the moment are worthless, because the moment is exactly when my judgment is compromised.

Why this belongs in a newspaper

“Do your own research” has come to mean “trust your own pattern recognition.” It is close to the opposite of research.

Spotting the pattern is the cheap part. It is free, it is fast, and everyone can do it. The expensive part is the disconfirmation, and almost nobody does that, because it feels like losing.

A conspiracy theory and an investigation start in the same place. Both begin with a person who thinks something is off. They part company at the moment the evidence comes back wrong. An investigation dies there. A conspiracy theory eats the contradiction and grows, because the contradiction gets folded in as further proof of the cover-up. That is the unfalsifiability problem wearing a civic costume, and New Hampshire has no shortage of it in town halls and comment sections.

The standard here is that a story dies when the record does not support it, no matter how badly I wanted the record to say something else. Every piece on this site carries a fact-check table where any reader can see which claims were verified and against what. That table exists because I do not expect to be trusted, and I do not think anyone should be.

Being wrong is the cheap outcome

The fear I carried for twenty years turned out to be the least useful thing I owned. Fear does not check anything. It sits there making noise while the same unexamined beliefs go about their business.

What replaced it was a willingness, and it took me an embarrassingly long time to see the difference. Being wrong costs me a correction, some pride, and an hour. That is cheap. I can pay it most weeks and I do.

The expensive outcome is not being able to be wrong. A belief with no handle on the outside, that absorbs every contradiction and comes out heavier. That is the thing I watch for, in the copy and in myself, and it is the only inheritance from my mother’s illness that ever frightened me in a way worth acting on.

So I do not ask whether I am seeing connections. I always am. That is the equipment I was issued. I ask the second question instead, the one that has an answer I can go and get: what would change my mind, and have I looked?

If I ever cannot answer that one, I would consider it a kindness if somebody told me.

— Dexter Dow, Granite State Report

Fact check

Editor’s note: This essay is a first-person account and an argument about method. It is not medical advice, and it is not a diagnosis of the author or of anyone in his family. Risk figures are population statistics and say nothing about any individual. Where published estimates conflict, as they do on the risk to a child of an affected parent, the conflict is reported rather than resolved in favor of the more dramatic number. Two claims rest on how a peer-reviewed review characterizes work it cites rather than on the underlying studies, and are marked ATTRIBUTED. One source, the 2025 Lancet Psychiatry register study, could not be opened directly because the publisher blocks automated access; it is cited in plain text without a link. Every hyperlink in this piece was opened and confirmed live before publication. Mental health is a serious subject and this is one person’s experience, not a clinical guide. Anyone worried about their own thinking, or about someone they love, should talk to a clinician; NAMI New Hampshire’s information line is (800) 242-6264, and the national 988 Suicide and Crisis Lifeline is available around the clock. Granite State Report corrects verified errors promptly and appends a note identifying what changed and when.