My grandmother was at the sink, washing up, with her back to me.

That was the whole thing. That was the dream.

The kitchen was her kitchen — the brown tiles, the window over the sink that never shut properly, the radio on. She was doing the dishes the way she always did, which was badly and at speed. She half-turned and handed me a tea towel, the way you hand someone a tea towel when you have handed it to them a thousand times and do not need to look.

And in the dream I knew, with complete calm, that she had been dead for two years. I knew it the way you know your own address. It just did not seem like the sort of thing you would bring up. It would have been rude, somehow — like correcting someone's grammar at a funeral.

So I took the tea towel and dried a bowl, and she said something about the neighbours, and then I woke up at ten past six with my chest aching and the specific, unmistakable feeling of having just been somewhere.

It stayed with me all day. It is, I think, still with me, or I would not be writing this.

Now here is the thing I could not stop turning over. Nobody in that dream said anything that mattered. There was no message. No instruction, no warning, no farewell, nothing you could write down. A woman washed up and handed me a tea towel.

And yet it was obviously about something. It landed like something. Every human being who has had a dream like this knows exactly what I mean, and knows that "your brain was consolidating memories" does not quite cover it.

So I went looking — properly, in both directions. What are dreams, according to the research that measures them? And what did the traditions that have been taking dreams seriously for three thousand years actually claim?

I expected the two to be at war. They mostly aren't. On the biggest question they turn out to agree, and to have agreed for a very long time.

Part OneWhat Are Dreams Made Of?

Start with the single most surprising finding in the whole field, because it reorganises everything else.

Dreams are not weird.

We think they are, because the weird ones are the ones you tell people at breakfast. But when researchers collect thousands of dream reports and code them systematically, only about one in ten contains anything that scores as genuinely bizarre — an impossible object, a person who is somehow two people, a body that changes shape.¹ The rest are people you know, in places you know, doing things that are broadly possible.

What dreams are is discontinuous. In one analysis, 81% of dreams contained an abrupt, complete change of scene — you are in the kitchen and then, with no transition and no surprise, you are on a train.² That is the actual texture of dreaming. Not impossible: unedited. There are no establishing shots.

The rest of the numbers are just as counterintuitive.

Sources for the four figures, in order: Domhoff's review of bizarreness coding; Fosse and colleagues on episodic fragments; the same study on complete replays; Hall and Van de Castle's normative sample. Full details in the references.

Look at the middle two together, because they are the most important pair of numbers in this article.

Roughly two thirds of dreams contain recognisable pieces of your recent waking life. Fewer than two in a hundred replay an actual episode as it happened.³

That pair is the real answer to what are dreams made of: yesterday, taken apart. Your sleeping brain is not filing footage. It is taking the day apart and building something else out of the parts. Which is exactly why my grandmother could be at a sink she has not stood at for years, handing me an object I have not touched since, in a kitchen that no longer exists — and why none of that struck me as odd at the time.

Why it comes out looking like that

The brain in dreaming sleep is not a quieter version of the waking brain. It is a differently configured one, and the pattern is oddly specific.

Running hot: the emotional centres, including the amygdala. The visual association areas — the parts that assemble images rather than receive them. The medial prefrontal cortex, which handles self-referential thought.

Switched down: the dorsolateral prefrontal cortex, which does logic, working memory and self-monitoring, and the orbitofrontal cortex, which does impulse control and reality-checking.⁴

Read that as a job description and you can more or less predict the output. A system that is emotionally driven, generating images, thinking about itself, and running with its fact-checker and its editor both off duty is going to produce something imagistic, emotionally weighted, self-centred and strangely uncritical.

It is going to produce, in other words, a dream.

And it does something odd to your associations

Here is my favourite experiment in dream science, and it is a small, clean, strange one.

Wake somebody out of REM sleep and immediately test how their word associations are working. In waking life, strong associations dominate — show someone hot and cold comes fastest. Straight out of REM, that flips. Weak, distant associations get processed faster than strong ones. The effect was replicated across three separate protocols.⁵

It shows up in performance too. Woken from REM, people solved 32% more anagrams than when woken from non-dreaming sleep.⁶ And in a nap study, only the group that reached REM improved at finding the hidden link between three unrelated words — and only for items they had been primed on beforehand. Not general alertness. Specifically associative.⁷

So the dreaming brain is not just running with the fact-checker off. It is running with the connections deliberately loosened — reaching for the far-apart thing rather than the obvious one.

Hold onto that, because it is about to do a lot of work.

Part TwoWhat the Traditions Said

Here is what I assumed I would find: a lot of omens. Symbol dictionaries. Snake means betrayal, water means change, and so on down the list.

That material exists, and I will come to it. But it is not what the serious texts spend their time on. When you read the traditions at their most careful — the Talmud rather than the almanac, Gregory rather than the folk charm — the striking thing is how often they land on the same answer to the first question.

Where does a dream come from?

Overwhelmingly: from you.

I have left out traditions that simply do not say much about dreams, rather than padding the list. And a caution on the Wendat material: it survives because missionaries wrote it down, often while trying to discredit it, so it reaches us through an unfriendly filter.

The hedges are the real story

Now, honesty. These same traditions also claim dreams can come from outside and can foretell things, and that claim is not some fringe accretion — it is central and canonical. Joseph before Pharaoh. Daniel before Nebuchadnezzar. The Talmud's "a dream is one sixtieth of prophecy."⁸ The hadith that a true dream is one part of prophethood, most commonly transmitted as a forty-sixth.⁹ Whole Puranic chapters of omens. An entire Greek healing industry built on sleeping in temples.

Any article claiming the traditions "really" meant the psychological reading is quietly deleting most of the evidence. So I won't.

But here is what genuinely struck me, and it is the thing I did not expect. Every one of these traditions built hedges against its own predictive claim — from the inside.

• The Talmud: dreams have no legal force whatsoever, and even a true dream is padded with nonsense — "just as it is impossible for the grain without straw, so too it is impossible to dream without idle matters."
• Islamic law: a dream cannot establish a ruling. Not even a true one. Revelation is closed.
• Sirach, quoted approvingly by Gregory: "Dreams have made many to err, and hoping in them they have been deceived."
• Aquinas permits reading dreams for natural causes but condemns "extending beyond its possible limits" — which is a precise description of overreach.
• Nagasena: only one of six kinds is true.
• Aristotle: mostly coincidence.

That is not a set of credulous cultures. That is a set of cultures arguing with themselves and installing safety rails — because they could all see that dream-reading gets abused.

Part ThreeThree Things They Got Right

So: where do the old texts and the modern data actually meet? Both spent centuries on the same question — what are dreams, and where do they come from — and they meet in three places. I have graded them honestly rather than generously.

Notice what has quietly happened here. The traditions' metaphysics — dreams as post from elsewhere — hasn't survived. But their phenomenology — their description of what a dream is like and what it tends to be about — has held up rather well, because they were unusually careful observers of something they took seriously and we mostly don't.

Part FourThe One They Got Wrong

Which brings us to prophecy, and I want to handle it carefully, because a great many thoughtful people have had a dream that later came true and have never quite been able to explain it away.

The short version: controlled tests do not find it. A sleep-laboratory study that specifically recruited people with a history of precognitive dreams found no evidence for dream precognition at all.¹⁵ That study was run by a researcher who had herself previously reported a small positive result and then failed to reproduce it — which is roughly the most honest thing a scientist can do.

But the interesting question is not whether it happens. It is why it feels so much like it happens — and the answer turns out to be flattering to nobody and insulting to nobody.

The arithmetic nobody runs

You have four to six REM periods a night. Call it something on the order of 1,500 to 2,000 dreams a year, the vast majority of which you will never recall. Multiply by everyone you know. Then by everyone in your country.

Now notice that there is no rule about which dream has to match which event, no time limit on the match, and no standard for how loose the resemblance can be. A dream about a hospital counts as a hit if anyone is ill within the month.

Against numbers like that, a striking coincidence isn't unlikely. It's scheduled. The mathematician J. E. Littlewood made this point in general terms: given how many small events a person registers in a day, a one-in-a-million coincidence should happen to each of us about once a month.¹⁶ Dreams are a coincidence-generating machine of exactly the right size.

And the part that is genuinely not your fault

This is the finding that changed how I think about the whole question.

Researchers gave people diary excerpts pairing dreams with later events — some where the event confirmed the dream, some where it contradicted it — and then tested what they remembered. Participants recalled more than twice as many confirming pairs as disconfirming ones.¹⁷

Here is the part that matters. Believing in the paranormal made no difference to the effect. Sceptics did it just as much as believers. This is not credulity. It is a standard property of human memory: the hit is memorable because it is a story, and the ninety-nine misses were never stories, so they were never filed.

Add one more layer. Most dreams evaporate within minutes of waking unless written down — so the surviving sample of dreams you can compare against later events is already selected almost entirely for the ones that turned out to be worth remembering.

And it is worth saying that this is not a modern insult to an ancient belief. Aristotle got there in the fourth century BCE — most apparent prophecies, he wrote, are "mere coincidences." He was pointing at the same hole two and a half thousand years before anyone could measure it.

Part FiveWhat To Do With a Dream

So you wake up with something still on you. What now?

The traditions and the research converge on a surprisingly practical answer, and it starts by throwing away the thing everyone reaches for first.

Put down the dream dictionary

Not because it's disreputable, but because the traditions themselves thought so. The book of dream meanings sold under Ibn Sirin's name was not written by Ibn Sirin — it descends from a compilation made some seven centuries after his death. The Duke of Zhou's Interpretation of Dreams is a folk almanac, not a Daoist classic, and isn't in the Daoist canon. Artemidorus, the most sophisticated dream interpreter of the ancient world, built his whole method on the principle that a symbol means nothing apart from the person who dreamed it — and demonstrated it by listing six women who all dreamed of giving birth to a serpent and whose sons turned out completely differently, each according to his mother's circumstances.

The fixed-meaning table is the shallowest thing every one of these traditions produced, and it is the only part that made it into the app store.

Ask the three questions that actually work

What replaces it is simpler and, in my experience, considerably more useful. Given what we know — that dreams are built from your own material, weighted toward what is emotionally live, and drawn to what is unfinished — the useful questions are these:

• What was the feeling? Not the plot — the feeling. The plot is scaffolding. If the dream is a picture of an emotion, the emotion is the part to name first, and it is usually easier to name than you expect.
• Where in my actual life is that same feeling? This is the only interpretive move with real support behind it. You are not decoding a symbol; you are looking for the waking situation that carries the same emotional signature.
• What has been unfinished for a while? Because that is where dreams demonstrably go — toward the thing you haven't resolved rather than the thing you've mastered.

Three questions, no codebook, and every one of them keeps the authority with the dreamer rather than the interpreter. Bar Hedya would have hated it.

The one dream technique with real clinical evidence

If dreams were meaningless noise, changing them on purpose shouldn't do anything. It does.

Imagery Rehearsal Therapy is the treatment for recurring nightmares, and it is the only one the American Academy of Sleep Medicine rates at its highest level of recommendation.¹⁸ The method is almost absurdly simple: you write down the nightmare, then rewrite it with a different ending, and rehearse the new version for a few minutes a day while awake. That's it. Nine randomised trials support it.

A 2022 study pushed it further. Everyone did the rewriting; half also had a sound played quietly during REM sleep at home that had been paired with their new, better ending. That group had fewer nightmares and more positive dream emotion, still holding at three months.¹⁹

I find this quietly remarkable. The most evidence-based thing we can do about dreams involves deliberately changing the image — which is as close as modern medicine gets to agreeing that dream imagery is a workable material rather than exhaust.

If nightmares are frequent, distressing, or following a trauma, this is worth raising with a doctor rather than doing alone — nightmare disorder affects around 4% of adults and is treatable, and up to 80% of people with PTSD experience nightmares.

And if the dream was about somebody who died

These are extremely common, they are almost always gentle rather than frightening, and people who have them tend to describe them the way I did — as a visit rather than a memory.

I don't think you have to decide what it was. That is genuinely the position I have landed on, and it is not a fudge.

What the science says is that your mind reached for the person you are missing, built them out of everything you have of them, and put you in a room together, because that is what minds do with what is unfinished. What the traditions say is that a dream comes from the thoughts of your own heart. As far as I can tell, those are the same sentence in two vocabularies — and neither of them requires the experience to have been less than it felt.

So — what are dreams? The closest I can get to one sentence, after all of this: pictures your own mind builds at night, out of the rooms and people it keeps, weighted toward whatever you have not finished feeling.

A woman washed up and handed me a tea towel. I have decided that was enough.

One practical footnote, since it is the single thing that changed this for me: write them down. I did not notice what my dream was made of until I had it in front of me in my own handwriting a week later — and I would never have spotted that the same kitchen kept turning up if I hadn't been logging them in Dreamery. Almost everything interesting about your dreams lives in the pattern across them, and the pattern is invisible until you keep a record.