Eleven days after his sister last thought he was doing well, he was using again. She says he chose it. He says the shift moved to nights in April, the pharmacy switched him to a generic that did not sit right, his ride on Thursdays fell through twice, and by the second week he was sleeping four hours a night and had not spoken to anybody in nine days.
They have had this argument four times in six years and neither has ever repeated the other's version back.
Both of them are invented. Putting the argument on made-up people means I am not describing anybody's family, and not anybody's recovery but the general shape of one.
They are arguing about where the cause sits, which is what every argument about addiction turns out to be. Two claims about that argument have been tested, and the results go opposite ways.
Changing the causal story people are told has not been shown to reduce blame. That sentence is doing careful work. The studies did not find that blame stays put. They failed to find that it moves, and those are different claims.
Changing the word used for the person does move blame. That has been tested twice and came out the same direction both times, and neither paper reports a number I could reach for how large the effect is.
Why the disease argument exists
The reason anyone fights about whether addiction is a disease is that the frame is supposed to move the attribution. Off character, onto biology, and therefore off blame. That is the argument, made in good faith by people who have watched other people be treated badly.
The scientific dispute underneath it is real and unresolved. One camp holds that addiction is an acquired disease of the brain and that the model has produced better treatment. Another holds the evidence does not support the model to the extent its advocates claim. Marc Lewis makes the sharpest version: addiction changes the brain, and brains are supposed to change, so the finding does less work than it appears to.
That fight is not getting settled here. What can be settled is narrower: whether the disease frame does to blame what its defenders say it does.
What has actually been tested
Before the evidence, one distinction that decides how much of it applies.
The studies below tested neurobiological explanations, meaning a brain-based causal story. That is not the same thing as the disease frame. A person can call addiction a disease without saying anything about brains, and Marc Lewis says a great deal about brains while denying it is a disease. Everything that follows is about the brain story specifically.
A 2018 meta-analysis gathered 26 studies of what happens when somebody is handed a brain-based explanation for a mental health problem. Nineteen were correlational and produced no significant overall effect. The seven experimental ones did produce effects.
Set against people given no such explanation, people given one wanted more distance from the person, judged him more dangerous, and were more pessimistic about his recovery. Blame did not move.
Four limits belong with that result.
The blame estimate is the weakest of the four by a distance. It came out at -0.24, with an interval running from -0.73 to +0.26. That is consistent with a substantial reduction in blame, with no effect at all, and with a small increase. Four studies could not tell, and nobody should read it as proof that blame stays put.
None of the seven experiments was about addiction. They studied depression, panic disorder, psychopathy, and mental illness in general. Carrying the pattern across to addiction is my move and not the paper's.
The authors say outright that they had too few studies of any single condition to test whether the pattern differs by diagnosis.
And the meta-analysis was testing a model that made two predictions. A biological explanation was supposed to lower one kind of stigma by making people think the person could not help it, and raise another by making them think he is a categorically different sort of creature. Only the second half showed up. The authors report that neurobiological explanations were not linked to reduced blame in either the experimental or the correlational studies.
The one study that is about addiction
A 2024 study put 755 people in front of a randomized vignette describing alcohol use disorder, opioid use disorder, problem gambling, or diabetes, then measured what each participant believed causes addiction and asked what they thought of the person described.
Believing addiction is a disease predicted nothing about stigma toward alcohol use disorder or opioid use disorder. For problem gambling it predicted more stigma, and that was the only condition where disease belief predicted anything at all.
Believing addiction is a moral failing predicted higher stigma for all three, strongly and consistently.
The design is worth stating precisely, because I described it loosely in an earlier version. The vignette was randomized. The beliefs were measured rather than assigned. So the belief and the stigma were observed together in the same people, and nothing here establishes that changing somebody's belief would change their judgment.
One thing that does move
Kelly and Westerhoff randomly assigned 516 clinicians to read a vignette about a man in legal trouble over alcohol and drugs. Half the vignettes called him a substance abuser. Half said he had a substance use disorder. Nobody saw both versions, and two thirds of the room held doctorates.
Two of the three things measured did not move. The clinicians did not rate him as more of a social threat, and they did not differ on whether he should be treated rather than punished.
What moved was culpability. The ones who read abuser agreed more that the man was personally to blame and that punitive measures were called for.
A companion study put the same two terms to 314 members of the public and found the same direction on culpability, plus the social threat difference the clinicians did not show. That study showed each person both labels, which is a weaker design for this purpose, because seeing both invites you to notice what is being tested.
What I cannot tell you is how large any of this is. Neither paper reports an effect size in anything I could reach, and the full text of the clinician study is closed. The direction is what I am claiming. If somebody tells you a single word transforms clinical judgment, that is more than these two papers support.
Which is a problem for the first half of this post
If one word moves culpability in five hundred clinicians, then framing can move blame.
What has failed to move blame is one particular kind of framing: the causal story about brains, tested in seven experiments about other conditions, with an interval too wide to call. The label did what the explanation could not, and I would rather say that plainly than let the two halves sit next to each other and hope nobody notices.
The comparison I cannot make
I applied different standards to the two literatures, and the comparison at the center of this post may not survive it. On framing, I have a meta-analytic estimate with an interval running from a substantial reduction in blame to a small increase. I decline to conclude anything from it. On labels, I have two studies, one available to me only through its abstract. Neither reports a magnitude I could reach. Then I set the second finding against the first and say the label did what the explanation could not.
That is a comparison between a null with wide bounds and a direction with no size attached. My own effect-size rule does not permit it. The narrower claim is defensible: two experiments found a difference in stated culpability after a single word changed, and four experiments about other conditions could not tell whether a brain-based story moves blame at all. Both are ratings on a vignette rather than evidence about how anyone was treated.
What the sister and the brother are doing
Three questions are in that kitchen and they are being argued as one.
Do the brain changes count as a disease? No measurement settles that. What does a given frame do to how he is treated? The evidence above addresses that question. What is owed to him regardless of where the cause sits? No measurement touches that question, and it is what most of these arguments are actually about.
They do not disagree about the eleven days. In four conversations neither of them has said which question they are having.
So say which one. Before the next round, name whether you are arguing about whether addiction is a disease, about what the words do to how he gets treated, or about what you owe him regardless. The first two have evidence behind them. What you owe him has none, and it is usually the argument you are actually having.