How to Help Someone Who’s Depressed When Nothing You Try Works

It is Tuesday. Your brother has not come out of his room since Friday, the curtains are still shut, and you are standing in the hallway deciding which version of yourself to be. The patient one, who says nothing and puts a plate outside the door. Or the one who says this cannot go on.

You have tried both. Neither worked, and you are running out of ideas at a rate that frightens you.

There is a lot of writing about mental illness for the person who has one. There is much less for the person in the hallway, and most of what exists arrives as adjectives. Be patient. Be understanding. Don’t judge. All true. None of it tells you what to do with the sentence already forming in your mouth.

There is, though, a large body of research on exactly this. Not on the illness — on the household. It has been running since the 1960s, it is unusually specific about what relatives do, and almost none of it has reached the advice.

The thing researchers measure is how you talk about them

In the 1960s, British researchers noticed something they were not looking for. People discharged from psychiatric hospitals did not do equally well, and the strongest thing separating those who relapsed from those who held was not the severity of the illness or the treatment they had received. It was where they went home to.

So they built a way to measure a home. A researcher interviews a relative, not the patient, and asks them to talk for an hour about the person and the illness. What gets scored is not what the relative believes about mental health. It is the texture of how they speak: how many critical remarks they make, whether the dislike is aimed at the person rather than the behaviour, and whether the illness has swallowed their life so completely that they have nothing else to describe.

Those three things — criticism, hostility, and what the research calls emotional over-involvement — make up a rating with a bloodless name: expressed emotion. A household comes out high or low. It is a description of the relative.

And it predicts what happens to the patient.

The link holds across twelve countries and four continents, at a correlation of r = .31 — small to medium, and stubbornly consistent. That figure comes from a 1998 meta-analysis of 27 follow-up studies in schizophrenia. Two decades later a separate group ran the same literature through a p-curve analysis, which is a technique built to detect whether a set of findings has been quietly massaged into significance, and concluded that it “likely has integrity.”

That is a firmer footing than most of what gets published about families. It is also worth saying exactly what it does and does not claim. It says the way a household talks about someone’s illness is associated with how that person goes on to do. It does not say your household caused the illness. Nobody serious has argued that for fifty years, and the research most often used to imply it says the opposite.

The belief that turns worry into criticism

Here is the part that makes any of this usable.

The criticism in these households is almost never cruelty. It is a theory. Study after study of relatives’ attributions has found the same shape: the relatives who make the most critical remarks are the ones who believe the behaviour is under the person’s control.

And that belief does not fall evenly. It attaches to particular symptoms.

When someone is hearing things that are not there, or is plainly not making sense, families read illness, and answer with worry. When someone will not get out of bed, will not shower, will not reply to a message, will not look for work — families read choice.

Those are also symptoms. In psychosis they are the ones clinicians call negative symptoms; in depression they show up as anhedonia and psychomotor slowing, which are technical words for a person to whom nothing feels worth doing and whose body has slowed to match. From the hallway they look identical to someone who is not trying.

So a family’s criticism lands hardest on exactly the symptoms that most resemble a character flaw and are not one. That is not a moral failing on the family’s part. It is a reading error, and it is the single most correctable thing in this entire literature, because the belief underneath it is wrong in a way that can be checked.

Trying harder is not automatically better

The other half of high expressed emotion is the half nobody expects, and it is the reason this article cannot end at “be kinder”.

Emotional over-involvement is scored when a relative is extraordinarily devoted. Self-sacrificing. Constantly monitoring. Distressed out of proportion to what is in front of them. Organising an entire life around one person’s illness, and describing that as the only thing there is to describe. It is love spent without a limit, and it sits on the same side of the ledger as criticism.

Not because devotion is bad. The proposed mechanism runs closer to this: a household with nothing in it but the illness leaves the person no role except patient, and leaves the relative nothing to draw on when it goes on for years. Both of those make the next bad month worse, and that is the mechanism the coders think they are picking up.

Be careful with this one. Across the literature, criticism is the component that predicts relapse most reliably, and over-involvement is weaker and less consistent — it behaves differently between diagnoses, and between cultures where the same behaviour means something different. What survives all of it is a modest, awkward claim: doing more is not automatically doing better, and a relative with no life of their own is not a stronger support. They are a more brittle one.

What the low-scoring households are doing instead

Warmth is coded separately in this research, and it goes the protective way. So the picture is not a household that has learned to say less. It is a household that has learned to say different things.

Three habits show up repeatedly. None of them are difficult to understand and all of them are hard on a Tuesday.

They treat the symptom as a symptom. Not as a comment on them, not as a decision the person made this morning. The silence at dinner is the illness in the room, and it will still be the illness tomorrow, which is why nobody in the kitchen has to win it.

They keep their own lives running. Work, friends, the Thursday thing they do. Partly because that is what makes ten years survivable, and partly because a person who is unwell needs at least one room in the house where the subject is something else.

They lower the demand in the air. Not the standards. The demand. Fewer questions per hour, fewer plans requiring an answer today, less checking whether the advice was taken. The reason is arithmetic rather than etiquette: every question is a small task, and the person is short of exactly the thing tasks cost.

What to say, and what not to say

Scripts are not magic, and anyone selling you a phrase that works is selling you something. But some sentences reliably do damage and some reliably do less, and the difference is not politeness. It is whether the sentence requires the other person to do the one thing they currently cannot.

The ones that land badly. “You have so much to be grateful for.” “Have you tried going for a walk?” “Everyone gets sad.” Each of these does the same thing underneath: it disagrees with the person’s account of their own condition, which leaves them arguing for how bad they feel instead of talking to you.

Then there is the most common sentence of all, and it is well meant: let me know if you need anything. It sounds generous. It is also an offer that only pays out if the other person initiates, and initiation is precisely what is broken. So the message never gets sent, and then it looks like they didn’t want your help.

Replace it with something that does not need them to start. Name what you have noticed, without a verdict attached, and make an offer that is specific enough to say yes to without a conversation. You haven’t been out since Friday, and I’ve been thinking about you. I’m going to the shop at six. Come or don’t — I’m going at six either way.

The difference matters more than it looks. The first version asks a person with no fuel to generate a plan, which is the expensive part. The second hands them one they can step into.

The one thing worth asking a clinician for by name

Most families never hear that this exists, which is why it goes here rather than in a resource list at the bottom.

Family psychoeducation is a structured programme for relatives — typically around six group sessions run by clinicians, covering how the illness actually works, what the early warning signs of a relapse look like in this particular person, what to do in a crisis, and, not incidentally, what the last two years have been like for you. It is a treatment aimed at the household.

In a 2022 network meta-analysis that pooled 90 randomised trials and 10,340 people, family psychoeducation on its own came out as the most effective family intervention for preventing relapse at twelve months — ahead of more elaborate models that add behavioural and skills training.

Now the part the headline usually drops. The same analysis rated the confidence in that estimate as very low, and a published response in the same journal argued the ranking should not be read without that caveat attached. Individual trials in this area are small: one recent French study that measured caregiver burden and depression falling significantly at six months ran on 25 families.

Your own outcome is part of the picture

The relatives in these studies get measured too, and the numbers are not gentle. Burden, depression and exhaustion in family carers turn up as routine findings rather than edge cases, which is why they are among the outcomes the psychoeducation trials are built to move.

What that looks like is the Thursday five-a-side you dropped in March because you could not be forty minutes away from your phone. Nobody made you drop it. You dropped it because being unreachable felt like a risk you had no right to take, and eleven weeks later there is nothing in your calendar that is yours.

That is the reason the advice about looking after yourself is not a soft addendum. It is usually delivered as permission you are supposed to grant yourself when there is time, and it is not. There is no version of this in which you are the infinite resource, so the relative who has kept one evening is not taking it from the patient. On the only measure this research has, they are the relative who is still standing in year three.

What this evidence does not say

Four things, and they matter as much as the rest.

It is correlational. Households where someone is more unwell are harder to live in, and a more severe illness makes relatives more critical. The arrow runs both ways, and the studies cannot fully separate them.

Most of it is about schizophrenia. That is where the 27 studies and the twelve countries are. The evidence in depression and other mood disorders is real but thinner, and this article has used depression as its running example because that is what most readers are living with, not because the numbers are equally strong there.

It is a group finding. An r of .31 describes a pattern across thousands of households. It does not tell you what will happen in yours, and it is not a scorecard you can rate yourself against tonight.

You did not cause this. Expressed emotion is about what makes a relapse more or less likely from here. It has nothing to say about origins, and the history of blaming families for causing mental illness is a genuinely shameful one that this research does not support.

The bottom line

You are not going to fix this, because it is not the kind of thing a relative fixes, and that is not the job. The job is to be one of the conditions that makes the next few months a little more survivable, which is a smaller task and an achievable one.

Most of it comes down to four moves. Read the flatness as a symptom rather than a decision. Notice what your frustration currently believes, before you speak. Make offers specific enough that nobody has to initiate. And keep enough of your own life that you are still here in two years.

The unglamorous part is that none of that will feel like helping. It will feel like doing less than you should. That feeling is not evidence, and it is the one thing in the hallway worth ignoring.

Keep exploring

Continue reading

Deeper into what you just read.

FamilyThe Part of Parental Burnout That Nobody Says Out LoudHealthStress Does Cause Depression. A Third of the Link Is Something Else.

Explore more

Adjacent ground, if this was useful.

RelationshipsYou Keep Having the Same Fight. Four Habits Decide Whether It Hurts.YourselfResilience Is the Common Outcome, and It Is Not Toughness