Dream experiment methodology

In thinking about how to carry out my Dunne-inspired experiment with precognitive dreams, I’m bothered by two things.

First, there is no control group. Anyone with sufficiently sharp pattern-recognition instincts will be able to find apparent connections between events even when the events are completely unrelated and the similarities are mere coincidence. What we want to know is not whether we can find connections between dreams and future events, but whether we can find significantly more such connections than would be expected by chance — and to know how many chance connections to expect, we need a control.

But what could that control possibly be? Other dreams of mine can’t be used as a control, because the hypothesis is that any of my dreams may have a non-chance connection with any of the events in my life, past or future. One possible control would be to have two dreamers do the experiment together and see whether dreamer A’s dreams match future events in dreamer A’s life more often than dreamer B’s dreams do. This won’t really work, either, though, since naturally my dreams, more so than anyone else’s, will tend to be about the kinds of things that are in my life (sugar gliders, English classes, etc.), making even chance matches more likely.

Even if a suitable control group could be found, experimenter bias would be an inescapable problem. All the pattern recognition is done by the dreamer, who knows which events are from his own dreams and which are from the control group, and he may subconsciously try harder to find links in the one case than in the other.

As far as I can see, there is no solution to these problems. The experiment will simply have to be done without a proper control.


The second thing that bothers me is that deciding whether a particular dream event is meaningfully connected with a particular real-life event requires a subjective judgment call, again leaving the door wide open for subconscious bias.

I think I have a partial solution for this one — not perfect, but probably the best that can be expected given the circumstances. The fact that only one person has any access to the relevant data — dreams and personal experiences — makes it impossible to devise anything like a proper double-blind experiment, but we can get as close as we can.

Here’s what I propose to do: For each dream I log, I will try to come up with and record as many connections as possible with both pre-dream and post-dream waking-life events. Then I will type up an account of the dream and append, in randomized order, accounts of each of the relevant waking-life events. However, each of these events will be described as having occurred prior to the dream. If I write, “Three days before the dream, . . .,” what follows might (for all the reader knows) really be an account of something that happened three days before the dream, but it could just as well be a description of an event which occurred three days after the dream. (The actual temporal distance of the event from the dream — a day, three days, a week, etc. — will be given, but always as if in the past rather than the future.) Also, for each event, I will provide some indication of how commonly events of that sort occur in my life.

I will then send these reports to a few third parties for evaluation. (Since people who are privy to the actual details of my life would be disqualified, I will probably recruit these people online.) Each evaluator will read the account of the dream and the events and will give each event a rating between 0 (no real connection; almost certainly just a coincidence) and 3 (so obviously tied to the dream that I would be astonished to hear that the dream had come first). Averaging the various evaluators’ ratings for each event, I can get some semblance of an objective measurement of which events are closely connected to the dream and which are not. If future events get high ratings about as often as past events do, I would consider that to be evidence of precognition.

This methodology is obviously far from ideal, but it’s the best I can come up with given the slippery and subjective nature of the subject matter. If anyone has any suggestions for improvement, please leave a comment.


As an example of how this would work, here are two more real-life events which I consider to be possibly connected with the snake dream recounted in this post (read it now if you haven’t already, or reread it to refresh your memory). Rate the events in the comments, and after I have a few ratings I will reveal whether each of them happened before or after the dream.

Event A: About a week before the dream, I was playing with the three-year-old daughter of a friend. (Her mother pays me to do this for half an hour every week, and to speak to her in English while we play.) We were playing with salt dough and sculpting various things. I tried to get her to make different things — bowls, snowmen, etc. — but she was only interested in making two things again and again: eggs and worms. Often she would give me a big worm and tell me to “make it a snake” by sculpting a head to put on it. (Cf. the worm/snake in the dream, which looked as if it were made of dough.) As I said, I play with this girl regularly, but this was the first time we used dough or clay together; aside from this, I haven’t sculpted anything in a very long time.

Event B: About a week before the dream, I went to feed my mealworms and found that nearly half of them were dead (probably from overheating, since there was a sudden heat wave at that time). Live mealworms are a golden-brown color, but when they die they quickly turn black (cf. the little black worms in the dream). I usually find a few dead mealworms in the dish every day, but this was the first time I’d ever seen so many of them die at once.

For good measure, here are two more dreams, each with a connected waking-life event:

Dream C: I dreamed that I was about to go outside, but when I opened the door I saw a huge Diatryma stalking about just beyond the front porch. (A Diatryma is a huge flightless prehistoric predatory bird, something like a cross between a secretary bird and a T-rex. They are often depicted eating the dog-sized ancestors of horses. The bird is called Gastornis now, but I still think of it as Diatryma.) I quickly closed the door and went back inside. I went back several times and looked out the glass door, but every time the Diatryma was still there, and I was afraid to go out.

Event C: The day before the dream, I was tutoring a grad student in her living room, and her 9-year-old son was also there, drawing pictures on scrap paper on the coffee table. He showed me one of his pictures and explained it to me. It was “a picture of birds,” but the birds had no wings and were shown walking on the ground. They were not ostriches, though, but round-bodied birds with short necks. One of them was chasing some kind of small quadruped. When I asked him what it was, he said it was a horse.

Dream D: I dreamed that I was teaching a large English class, including one of my real-life students (I’ll call her Joan). I had just started the class, when loud music started to come from my bag. I searched through the bag and found a little travel-size alarm clock (just like one I owned years ago, before I had a cell phone), which was making the noise. I tried to turn it off but couldn’t. Finally I took out the batteries. I started the class again, but someone came in and told me I had to go to a training program right away. I apologized to the students and walked out of the class.

Event D: I was teaching a small group of Taiwanese English teachers, including “Joan.” I had just started the class when I got a text message on my phone, which made an audible beep. I checked it and found that there was an emergency at home and that I had to leave right away. I apologized to the students and walked out of the class.


Please leave a comment rating the four events as described.


Filed under Anecdotes, Dreams, Precognition / Prophecy

10 responses to “Dream experiment methodology

  1. Luther

    I’m a bit confused why past and future would need high marks “about as often” for dreams to be considered precognitive. That would be true of purely random dreams, and of a purely boring life; it would also be false if, say, 2/3 of your dreams are purely dreamified past and the other 1/3 are purely dreamified future.

    It seems to me the goal is to demonstrate that tonight’s dream is more closely correlated to tomorrow’s events than are today’s events: for example, I would consider knowledge of problems that may have led to the emergency in event D to be at least as predictive as your dream.

    An alternate methodology might be a two-part experiment. In the first part you practice writing real-life experiences in a dream-style, and poll people to pick out the real dream. In the second part, for each life event you pick one earlier dream and one earlier experience and write up both as precognitive dreams, then poll people to select which one more closely predicts the event. To reject the null hypothesis (that dreams are not better predictors than the past) it would be necessary for the real dream to be selected as precognitive in the second part more often than dreams could be distinguished from reality in the first part.

    That design has serious problems of it’s own (dreams correlate to events in a different way than does the past) but at least it is comparing predictive power of dreams to the predictive power of something else instead of to the recall ability of dreams.

  2. Bruce Charlton

    You discussion of methodology seems sound. But don’t make thinks so difficult for yourself that the experiment never happens.

    It might be better to do things in several phases, increasing the level of controls at each phase (assuming results are positive), rather than trying to get everything perfect first time.

    This also has the advantage of replicating the results on different independent data sets (different groups of dreams and experiences) – which is more convincing than doing one complex analysis on one single data set.

    For example, I would do the whole things yourself – including the ratings, before I tried to involve third parties/ other people as objective raters. As soon as other people are involved things will slow down – and you will have to consider between rater reliability etc. And if you don’t find anything which satisfies you using your own evaluations, then there would be no need to proceed to this stage.

    In general, the more complex the similarity between dream and experience the less likely it is a coincidence; the shorter the time between dream and the experience the less likely the coincidence, the stranger/ less routine the event which happened in dream and experience the less likely it is a coincidence.

    But a number cannot be allocated to these probabilities, nor is there any reason why it needs a number. Science doesn’t work that way – when people started putting numerical probabilities into human biology, they stopped doing human biology!

    A – 0

    B – 0

    (These events seem too common in your life – although maybe they should get a 1 on the basis that – reversing chronology – you might expect to dream about thinsg taht happen frequently)

    C – 1 – because of the flightless predatory bird and the horse together and the closeness in time.

    D – 0 – seems like common events in life.

    In the end the experiment is a matter of satisfying yourself – if other people are unconvinced (presumably because they think you may be incompetent, dishonest or self-deluding) then they would anyway need to try and replicate the results independently.

  3. Luther, I am taking it as a given that memories of past events influence dreams. I think most everyone accepts that this is true, though of course it would be hard to prove in any rigorous way. If the experiment shows that dreams correlate with past and future events to about the same degree, it would mean that the evidence for dreams being precognitive is about as good as the evidence for their being influenced by memory — and since most people accept the latter, they should also accept the former.

    If, on the other hand, dreams correlate with the future significantly less often than with the past, it would be hard to know what to conclude. We would expect to find some connections just by pure coincidence, but without a control group it’s impossible to say how many chance connections we should expect. For example, say we find that 30% of dreams contain clear references to the past and 10% to the future. How much of that 10% is chance? All of it? There’s no way to tell.

    I considered the idea you mention, of matching life event to life event as a control, but I don’t think it would really work. For one, even if you write them in “dream style,” real-life events are still fundamentally different from dreams, and it would nearly always be easy to tell which was which. For example, every single one of the dreams I’ve related so far includes events which could not possibly happen in real life. Also, we should expect non-random connections between earlier and later events in a person’s life, so it doesn’t really work as a control.


    Bruce, thank you for ratings and suggestions. The fact that you gave almost all of them a 0 suggests that getting third-party ratings is indeed useful. I would have given them all at least 1 myself (obviously, the reason I wrote them down is because I felt they had some plausible connection to the dream), which suggests that subconscious bias is a real problem. But you are probably right that I should do the experiment on my own first and not waste my time on anything more complicated unless I get what seem to me to be positive results.

  4. Bruce Charlton

    My zero ratings were perhaps harsh – I assumed, for example, that because you work as a routine with worms you will dream about worms (more than most people) – I wasn’t confident of a specific relationship between the dream and the event.

    Dunne strongly recommends experimenting before a change in life routine, to try and overcome this.

  5. Luther

    Points taken.

    Events A and B: First, a bit of surprise at the “about a week”: having often a dozen or more dream episodes a night myself, the pool of dreams pulled from seems quite large at that distance. However, assuming your pool is smaller, I’ll give them both a 1: in each case, the correlation appears to be with a side element of the dream, of which there are presumably many others you failed to mention.

    Event C: I’ll give this a 2: close in time, and the main thrust of the dream (seeing a Gastornis) is close; however, lacking is the secondary thrust of the dream: being frightened and blocked by the Gastornis.

    Event D: No time span given, but since I’ve already ignored it in A and B, that’s not a big problem. Joan seems to me no correlation: that real and dream classes share one otherwise-unremarkable student seems all-but inevitable. We are left with match: audible interruption in class and leaving class; and no match: annoying persistent sound buried in bag and in past, repeat interruptions of class, and emergency, and a training program. Add to this preponderence of non-matches the fact that the dream sounds like such a typical inadequacy-fear dream, I’m going with a 0.

  6. Bruce Charlton

    On further thought I would revise my ratings up by a category.

    Zero should probably be used to code for no relation at all.

    1 could refer to a thematic similarity (e.g. work with worms, dream of worms).

    2 could refer to at least two unusual features coincident (a hippogriff with purple spots),

    and 3 to more than two unusual features coincident (a hippogriff with purple spots on a pogo stick).

  7. Luther,

    I see that I forgot to mention the time span for D. The event was the day before the dream. Certainly, the longer the time between the dream and the event, the higher the chance of coincidental “matches.”

    Your comments on the Gastornis dream seem to show a slight misunderstanding of the hypothesis in question. There is no expectation that the “main thrust” of the dream should match the life event; rather, the idea is that future events will appear in dreams (if at all) in much the same way that past events do — that is, distorted and mixed up with a lot of irrelevant garbage, but still clearly recognizable.

  8. Since it seems no one else is going to comment, I’ll go ahead and reveal the correct chronology of the events.

    Event A (making worms and snakes out of dough) happened before the dream of a worm/snake with a doughy texture.

    Event B (finding lots of dead black worms) happened after the dream of finding lots of live black worms.

    Event C (the kid drawing a Diatryma) happened after the dream of seeing a Diatryma.

    Event D (a text message interrupting a class) came before the dream of an alarm clock and an interrupted class.

  9. Extended ( And possibly QuaziRelated Rant ) Prompted by Dream experiment methodology :

  10. Konstantin

    Hi there. I’m making exactly the same experiment myself, and I’m happy I google up your blog. What I decided, is that all I need is to “catch” 2 events, which will _without_ any doubt count as precognition, that is, without my further personal interpretation. Such dreams are rare, I guess, but should exist as theory states. I think that is the most decent way to prove Dunne’s words. Like, some dream he had about “4000” people died, and that actually was in the newspaper next day. Event was not too comon in his life, number matched, the rest matched as well. Something like that. That should give a good start and motiovation to go on, without thinking too much “what if Dunne lied”?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s