A
Deep Dive: Precognitive Lucid Dreaming
Can we access information about events before they happen?
This is a fascinating subject because it brings together three areas that are rarely examined together:
Lucid dreaming – becoming consciously aware that you are dreaming while the dream is still happening.
Precognition – apparently obtaining information about an event before that event occurs.
The nature of time and consciousness – whether awareness is entirely confined to the present moment or whether information might somehow be accessible across time.
The research described by Robert Waggoner is particularly interesting because it attempts to investigate precognition within a lucid dream, using a future image-selection process.
But we need to distinguish three things:
Having a dream that resembles a later event.
Demonstrating that information about a future event was obtained before it happened.
Establishing a mechanism through which such information could travel backwards in time.
These are three increasingly demanding scientific claims.
The study offers an opportunity to investigate the second question, but it does not yet establish the third.
Let's explore the research, its methodology, the implications and the scientific limitations.
1. First, what is genuinely new about this research?
The paper is titled:
Future Dreams of Electric Sheep: Case Study of a Possibly Precognitive Lucid Dreamer with AI Scoring.
It was written by Julia Mossbridge, Dave Green, Christopher C. French, Alan Pickering and Damon Abraham, and published in the International Journal of Dream Research in 2025. The October 2026 IONS article is discussing that earlier publication.
The central innovation is not simply asking people whether they have dreamed about the future.
It is combining three things:
A participant who can deliberately enter lucid dreams.
A target-selection procedure in which the target is selected after the dream report has been submitted.
AI-based assessment of similarities between dream descriptions and target material.
This is important because ordinary precognitive dream reports have a major scientific problem: they are usually identified after the event.
For example:
You dream about a house burning down. Three weeks later, you hear about a house fire.
You might feel that your dream predicted the event. But how do we know you would have remembered that dream if the fire had never occurred?
This is called retrospective selection.
We remember the apparent matches and tend to forget the thousands of dreams that did not match anything.
The Mossbridge study attempts to address this problem by recording the dream before the target is selected.
That is a meaningful methodological improvement.
An important clarification about the word "future"
There is a subtle but crucial point in this particular experiment.
The target was randomly selected after the dream, but the target images and stories already existed in an online archive dating from 2006–2016.
Therefore, the experiment was testing whether the dreamer could obtain information about a randomly selected future target, not whether he could predict an event that had not yet happened in the physical world.
The future element was the selection of the target, rather than the creation of the target event itself.
This distinction matters when interpreting the results.
2. Why lucid dreaming could be particularly interesting
A lucid dream occurs when you become aware that you are dreaming while remaining inside the dream.
Ordinary dreaming often involves accepting whatever happens without questioning it.
In lucid dreaming, something changes.
You might suddenly recognise:
The environment is a dream.
The people around you are dream characters.
You are not physically located where you appear to be.
You can sometimes influence what happens next.
This creates a fascinating possibility for research.
Instead of waiting for a spontaneous dream that might contain an apparent prediction, a lucid dreamer can deliberately set an intention.
For example:
Tonight, when I become lucid, I will ask to see the image that will be randomly selected tomorrow.
This is effectively an experimental instruction given to the dreaming mind.
Lucidity versus dream control
An important distinction is that becoming lucid does not automatically mean you have complete control over a dream.
There are three separate abilities:
Ability | Meaning |
|---|---|
Dream awareness | Knowing you are dreaming |
Dream agency | Intentionally deciding what to do |
Dream control | Successfully changing dream events or surroundings |
Someone can become lucid but still be unable to control the dream.
Research published in September 2026 also highlights this distinction, noting that dream control can be deliberate or expectation-driven and that individual differences are substantial.
This matters because the Mossbridge experiment depends on the dreamer being able to achieve a particular mental task while lucid.
3. The experiment: How did Dave Green attempt to access the future target?
The participant, Dave Green, was an experienced lucid dreamer and artist.
His task was to enter a lucid dream and deliberately seek information about an image that would be randomly selected after he had recorded his dream.
The process was relatively straightforward.
The experimental sequence
Enter a lucid dream
Green recognises that he is dreaming.
Create a computer or smartphone
Within the dream, he looks for a screen.
Request the future target
He intends to see the image that will be randomly selected the following day.
Record the experience
He wakes up, writes a detailed transcript and creates drawings.
Submit the dream before target selection
The transcript is sent to the researcher acting as the target selector.
Randomly select the target
A random-number procedure selects an item from the existing pool of 478 targets.
Compare the results
The researchers compare the dream with the selected target using human and AI judging methods.
The study consisted of two sets of five trials, giving ten formal dream trials in total.
Why this sequence matters
The most important protection is that the dream transcript was submitted before the target was selected.
This reduces the possibility that the dreamer consciously adjusted the description after seeing the target.
However, it does not eliminate every possible source of bias or information leakage.
That is why independent judging, secure time-stamping and tightly controlled target selection are so important.
4. The target images: What counted as a successful prediction?
The target pool contained 478 previously collected images and associated stories.
The researchers used several different ways of assessing whether a dream corresponded to its target.
This is particularly important because dreams rarely reproduce reality like photographs.
They often contain:
Symbolic representations.
Emotional impressions.
Unfamiliar places.
Fragmented images.
Distorted people and objects.
A mixture of literal and metaphorical content.
The researchers therefore explored both visual and semantic matching.
Example B: The house and the researcher's home
Coincidental resemblance
The researcher's own English townhouse.
This example is especially interesting psychologically.
The dream drawing seemed to resemble the home of one of the researchers, who had been considering putting it up for sale.
However, the paper reveals a critical detail: the dreamer did not remember the exact image clearly and searched online for a townhouse picture to use in his drawing.
The researcher later recognised a resemblance to his own home.
This was an interesting coincidence, but it was not an independently established precognitive hit.
It illustrates how easily personal associations can become part of the interpretation of dream material.
5. The AI judges: A potentially important development
This is arguably the most technically interesting part of the study.
The researchers used five different AI text-embedding models:
OpenAI Ada.
OpenAI L3.
Nomic.
Snowflake.
BAAI.
These systems were not simply asked, "Does this dream predict this image?"
Instead, they converted dream descriptions and target descriptions into mathematical representations called embeddings.
What is an embedding?
An embedding converts language into a numerical representation that captures aspects of its meaning.
For example:
A dream mentioning a stadium, women and a competition.
A target describing a women's rugby match.
The embedding system attempts to represent the semantic similarities between those descriptions.
The researchers then used cosine similarity to compare the resulting numerical representations.
Simplified example of semantic matching
Dream description
Women competing in a large public arena.
AI semantic analysis
Target description
A women's sporting competition in a public stadium.
Calculate similarity
What were the results?
Judging method | Correct matches |
|---|---|
Skilled human judges | 3 out of 10 |
Unskilled human judges | 1 out of 10 |
Best AI performance | 5 out of 10 |
With five possible target choices for each dream, random selection would have an expected accuracy of 20%.
The best AI result was therefore above the nominal chance expectation.
However, there is an important statistical qualification.
The AI analysis involved five embedding models and four different combinations of dream text, dream images, target text and target images.
The best result of five correct matches out of ten had a reported binomial p-value of 0.033, without correction for the multiple comparisons.
The authors explicitly describe these AI analyses as exploratory.
Why this matters
AI may help researchers assess dream material more consistently than human judges.
But AI similarity is not the same as evidence of precognition.
A language model can find connections between two descriptions because both contain broad themes such as:
Movement.
Conflict.
Water.
Family.
Buildings.
Groups of people.
The more flexible the interpretation, the easier it becomes to find apparent similarities.
The challenge is to demonstrate that the correct target consistently scores higher than decoys, using a scoring system fixed before the results are known.
6. The most important statistical result: One particularly strong match
The researchers also compared dream material with a much larger database containing 2,707 possible targets.
This was intended to provide a more demanding test than simply choosing between the five targets in each experimental set.
One result stood out: the backgammon/rugby dream.
The researchers reported a very high semantic similarity between that dream's text and its corresponding target.
The paper reports a Z-score of 4.99 and discusses a corrected probability below 0.00012 in its expanded analysis. The abstract reports a different threshold, below 1.2 × 10⁻⁵, for the AI findings. These figures relate to different statistical presentations and should not be treated as interchangeable.
Why is this interesting?
It suggests that the particular dream-target correspondence was unusual relative to the comparison database used by the researchers.
But there are three reasons to be cautious:
The result emerged from an exploratory AI-scoring approach.
The strongest correspondence was identified among multiple dream-target comparisons and analytical choices.
The study involved only one participant and ten formal trials.
A statistically unusual match is worth investigating, but it is not equivalent to demonstrating that the dreamer received information from the future.
The next experiment needs to reproduce the result using a scoring method that is fixed in advance.
7. Why judging dreams is such a difficult scientific problem
This is a major issue that Robert Waggoner correctly identifies.
Dreams are not usually literal descriptions of reality.
Suppose somebody dreams of:
A circular building.
A group of people.
A large body of water.
A red object.
A feeling of danger.
Later, the target is a photograph of a rescue operation beside a flooded stadium.
Could that be a match?
Possibly.
But those five dream elements could also be associated with hundreds of other images.
This creates a problem called interpretive flexibility.
The more freedom judges have to interpret symbols, the more opportunities there are to identify a resemblance after seeing the target.
Literal versus symbolic scoring
Literal approach | Symbolic approach |
|---|---|
A tree matches a tree | A tree represents growth |
A house matches a house | A house represents security |
Water matches water | Water represents emotion |
A sporting event matches a sporting event | Competition represents conflict |
Neither approach is automatically correct.
The difficulty is that symbolic interpretation can become almost unlimited unless the rules are established in advance.
My recommendation
A future study should use two separate scoring systems:
System 1: Objective visual matching
Assess identifiable features:
Number of people.
Colours.
Shapes.
Objects.
Buildings.
Landscapes.
Activities.
System 2: Semantic matching
Assess broader themes:
Competition.
Travel.
Conflict.
Family.
Celebration.
Danger.
The results should be reported separately.
That would help researchers discover whether any apparent precognitive information is primarily visual, thematic or symbolic.
8. Dream telepathy: Is connecting with another person's mind easier than accessing a future target?
The article introduces another fascinating area of research: dream telepathy.
The International Association for the Study of Dreams (IASD) has a history of dream telepathy contests inspired by the work of Montague Ullman and Stanley Krippner.
The general experimental idea is different from the Mossbridge study.
In a dream telepathy experiment:
A sender receives or views a target image.
The sender attempts to mentally communicate the image.
A sleeping participant dreams.
The participant reports the dream.
Judges compare the dream against the target and alternative images.
The sender is therefore part of the experiment.
In precognitive lucid dreaming, by contrast, the dreamer attempts to obtain information about a target that has not yet been selected.
The Clare Johnson and Beverly D'Urso example
The story in Waggoner's article describes a lucid dreamer who encountered the telepathic sender in a dream.
The dreamer heard references to shouting and trees.
The eventual target was a photograph of giant trees in Sherwood Forest.
This is an intriguing anecdote.
But it is important to remember that a retrospective example from a contest is not equivalent to a controlled experiment establishing telepathy.
Why might the distinction matter?
There are two very different hypotheses:
The dreamer may somehow receive information associated with another person's present thoughts.
The dreamer may somehow receive information about a future random selection.
The first is a telepathy hypothesis.
The second is a precognition hypothesis.
They should be tested separately rather than treated as one phenomenon.
The fact that a dreamer experiences a person delivering information in a dream does not establish that the information originated from that person's mind.
9. Could the dreaming mind process information differently from the waking mind?
This is a particularly interesting question even without assuming precognition exists.
Dreaming is associated with changes in attention, memory, emotional processing and sensory experience.
During ordinary dreams, the brain can generate extraordinarily complex environments, people and situations.
During lucid dreams, some of the usual limitations of ordinary dreaming change because the dreamer becomes aware of the dream state.
This raises an interesting research question:
Could lucid dreaming provide a different way of organising, retrieving or associating information?
There are several conventional explanations worth considering.
A. Unconscious information processing
The brain continually processes information that does not reach conscious awareness.
Dreams may combine fragments of memory and previously encountered information in unexpected ways.
A dream that later resembles an event might therefore reflect information already encountered, even if the dreamer does not consciously remember encountering it.
B. Pattern recognition
The brain is exceptionally good at identifying patterns.
This is useful for survival, creativity and problem-solving, but it can also create connections between unrelated events.
C. Memory reconstruction
Dream memories are unstable.
When we recall a dream, we may unintentionally reorganise its content in light of subsequent events.
This is one reason that recording a dream immediately upon waking is so important.
D. Expectation and intention
If a lucid dreamer repeatedly asks to see a future image, that intention may influence the imagery generated by the dream.
This is established as a plausible psychological process, even though whether the resulting imagery contains genuinely future-derived information is a separate question.
None of these explanations has been demonstrated to account for the particular study results. They are competing possibilities that further experiments should examine.
10. The major scientific limitations
I think this is where we need to look most carefully.
The researchers themselves acknowledge that the results are preliminary. They describe the evidence as weak but encouraging and explicitly state that a confirmatory experiment is needed.
There are six particularly important limitations.
1. Only one dreamer
Ten trials involving one participant cannot establish that lucid dreaming generally provides access to future information.
Even if Dave Green genuinely demonstrated an unusual ability, it might be unique to him, or it might reflect chance.
2. The AI analysis was exploratory
The researchers tested multiple AI models and different combinations of dream and target material.
This creates a multiple-comparisons problem.
If enough models, scoring approaches and analytical methods are tried, some will produce apparently impressive results by chance.
3. The original human judging method was flawed
The two skilled judges had to distribute their scores across five possibilities so that their scores added up to 100.
The researchers subsequently recognised that this made the results difficult to analyse statistically.
Consequently, the original preregistered judging results could not provide a straightforward test against chance.
4. The target pool already existed
The 478 targets were historical images and stories.
Although the selection occurred after the dream, the information itself already existed.
Future studies should use genuinely novel, securely controlled target material to reduce the possibility of previous exposure.
5. The pilot studies matter
The researchers conducted two encouraging pilot studies before the formal ten-trial experiment.
Pilot work is normal and useful, but it means the formal results should be considered in the context of an earlier exploratory process.
6. The strongest individual match is not the same as consistent performance
The backgammon/rugby correspondence was particularly striking.
But the other dream-target pairings were much less convincing.
A genuine precognitive ability should ideally produce repeatable performance across many trials, not depend on one especially impressive example.
11. What would a genuinely convincing precognitive dreaming experiment look like?
I would design the next study around the strengths of this research while addressing its weaknesses.
Proposed next-generation experiment
Recruit multiple experienced lucid dreamers
Include at least 30 participants, with a larger sample if practical.
Prepare a new target database
Use thousands of carefully vetted images that participants have not previously encountered.
Record dreams before target selection
Time-stamp and securely lock each transcript and drawing.
Select targets using independent randomisation
A separate researcher or automated system selects the target only after the dream has been securely recorded.
Prevent feedback during data collection
Participants should not see previously selected targets until the study is complete. This prevents later dreams being influenced by knowledge of earlier targets.
Use independent judging
Multiple blinded judges and preselected AI models score dream-target matches against decoys.
Publish every result
Include misses, ambiguous matches, exclusions, statistical methods and all preregistered outcomes.
Independent replication
Another laboratory repeats the entire protocol without relying on the original research team.
An additional experiment I would particularly like to see
Compare three different conditions:
Condition | Task |
|---|---|
Ordinary dreaming | Participants simply record their dreams |
Lucid dreaming | Participants seek the future target while lucid |
Waking intention | Participants attempt to visualise the future target while awake |
This could help answer whether lucidity itself is important.
If the lucid dream group consistently performed better than the other groups, that would provide a more specific reason to investigate the dreaming state.
If all three groups performed similarly, the explanation might lie elsewhere.
12. Can we practise precognitive lucid dreaming?
There is a practical question here that I think is worth exploring.
Even without assuming precognition is real, lucid dreaming is a recognised phenomenon that people can practise.
The following approach would allow someone to investigate their own dreams while maintaining a useful scientific record.
A structured dream-investigation routine
Before sleep
Set a clear intention to recognise when you are dreaming.
Decide on one simple question to explore.
Avoid deciding what the answer should look like.
During a lucid dream
Confirm that you are dreaming.
Ask to see a randomly selected target.
Observe what appears without forcing an interpretation.
Remember the colours, objects, people, shapes and emotional impressions.
Immediately upon waking
Write down everything you remember.
Draw the imagery, even if it seems nonsensical.
Record the time and date.
Do not edit the original report after seeing the target.
After target revelation
Compare the dream with the actual target.
Record both matches and mismatches.
Have somebody unaware of the correct target judge the results.
Keep a record of every trial, not just the interesting ones.
The most important rule is to record everything before looking at the target.
That protects the integrity of the experiment and helps distinguish genuine advance information from retrospective interpretation.
I would also avoid sacrificing sleep or using sleep deprivation to induce lucid dreams.
13. The deeper philosophical implications: Is time as straightforward as we think?
Now we arrive at the most fascinating question.
If precognitive dreaming were eventually demonstrated under rigorous conditions, it would raise fundamental questions about how information relates to time.
There are several theoretical ideas worth distinguishing.
A. The block universe
In the block-universe interpretation associated with relativity, past, present and future can be described as parts of a four-dimensional spacetime structure.
The present is not necessarily a uniquely privileged universal moment in the mathematical description.
However, this does not mean that humans can access information from the future.
The block universe is a way of understanding spacetime, not a demonstrated explanation for precognitive dreams.
B. Retrocausality
Some interpretations and approaches to quantum physics investigate whether future measurement conditions can enter the mathematical description of earlier events.
This is called retrocausality.
It is a serious theoretical topic, but it does not establish that dreams can access future information.
C. Timeless or non-local consciousness
Some philosophical and speculative approaches propose that consciousness may not be entirely restricted to the individual brain or the ordinary experience of time.
These ideas are interesting, but they do not currently have an established experimental mechanism explaining precognitive dreaming.
D. The neuroscience explanation
An alternative is that dreams are products of the brain's extraordinary capacity to combine memory, expectation, emotional information and pattern recognition.
This explanation does not require information to travel backwards through time.
At present, none of these approaches has been demonstrated to explain the findings of the Mossbridge study.
14. My overall assessment
Where the evidence stands
Lucid dreaming is a real phenomenon.
People can become aware that they are dreaming, and some can deliberately influence their dream experiences.
The Mossbridge study provides an unusual preliminary result.
The AI scoring methods identified several dream-target correspondences, with one particularly strong example.
Lucid dreamers can reliably access future information.
The study is too small and its exploratory analysis too extensive to establish this conclusion.
Consciousness can transcend ordinary time.
This is a possible philosophical interpretation, not a finding established by the experiment.
My five principal conclusions
1. The methodology is promising. Recording dreams before random target selection is a meaningful improvement over retrospective dream anecdotes.
2. Lucid dreaming offers a valuable research environment. It gives participants an opportunity to establish an intention and deliberately seek information.
3. AI could improve dream research. Semantic analysis might make it easier to assess large quantities of dream material, although its scoring methods need to be independently validated.
4. The statistical evidence remains preliminary. Ten trials involving one dreamer are insufficient to establish precognition.
5. The relationship between dreaming, consciousness and time deserves further investigation. But we should not mistake an intriguing result for a demonstrated mechanism.
15. The question I find most interesting
There is something particularly intriguing about the difference between the two approaches in this research.
In ordinary dream telepathy, the participant is supposedly attempting to receive information associated with another person's mind.
In precognitive lucid dreaming, the participant is attempting to receive information associated with a future random selection.
Both approaches involve the possibility that information might become available to the dreaming mind through a process that conventional explanations do not readily account for.
But they are fundamentally different hypotheses.
If future experiments demonstrated that lucid dreamers could reliably identify targets selected only after their dream reports were securely recorded, the implications would be significant.
It would raise a question that goes beyond dreaming:
Is the human mind limited to processing information from its immediate physical environment and its existing memories, or could there be some additional relationship between consciousness and the structure of time?
We do not currently have evidence sufficient to answer that question affirmatively.
But it is a question that can be approached experimentally rather than simply philosophically.
Three directions for an even deeper investigation
1. The science of precognition
Explore the Maimonides dream experiments, presentiment research, remote viewing, dream telepathy and the strongest criticisms of each.
Explore the wider precognition research2. Lucid dreaming and consciousness
Explore REM sleep, brain activity, metacognition, meditation, dream control and the possibility of using lucid dreams for deliberate investigation.
Explore lucid dreaming and consciousness3. Time, consciousness and quantum physics
Examine retrocausality, the block universe, quantum entanglement and theories connecting consciousness with time.
Explore consciousness beyond ordinary timeMy suggestion: The most revealing next step would be to examine the historical Maimonides dream experiments alongside this new lucid-dreaming research. That would help establish whether the current findings represent a genuinely new development or another chapter in a much longer scientific investigation into the possibility of precognitive dreaming.
No comments:
Post a Comment