Part of The exploration series
EXPLORATION 07 · TESTING
Observe first, interpret later. Use blind targets, record impressions before feedback, use predefined scoring and let misses count just as seriously as hits.
Method before explanation. The first question is not where unusual information came from, but whether verifiable information is actually present under controlled conditions.
Exploration 07 · testing
If an impression seems to contain information you could not have known through the ordinary senses, the next step is not belief — and not dismissal either. The next step is better testing.
Exploration 06 asked the larger question: can consciousness sometimes receive information beyond the ordinary sensory route? This exploration does not start with the explanation again. It starts with method.
Goal: first investigate whether verifiable information is actually present. Only then does the question of its source become interesting.
Experience → information → effect → explanation
With unusual experiences, four different questions are often collapsed into one step.
What actually appeared in your experience: image, feeling, word, dream, impression?
Did a verifiable detail correspond to something you did not know at the time?
Across multiple well-controlled attempts, does that happen more often or more accurately than chance would suggest?
What source or mechanism would best explain that effect?
An unusual experience answers only question 1. A good hit may say something about question 2. Only a robust pattern across multiple tests makes question 3 interesting. Question 4 comes last.
Before choosing a test, one more distinction is needed: are you testing information access or probability influence? They are different claims.
You test whether a response corresponds to information that was not available through the ordinary route. Remote viewing and precognition belong here.
You test whether predefined Intent shifts the distribution of a genuinely uncertain outcome across many trials compared with what you would expect without that influence.
A good remote-viewing test therefore cannot establish probability influence. And an unusual outcome distribution does not prove that someone received information about the future.
Discovery and evidence are not the same thing
In your first sessions it is fine to discover what kind of signals you seem to get. Perhaps visual images are weak but spatial impressions are strong. Perhaps words arrive only after an initial bodily feeling.
That is exploration. You use the results to design a better protocol.
But once you want to treat a result as a test, define in advance:
Changing the protocol afterwards is not forbidden. But the changed version becomes a new test, not a repair of the old result.
Protect the raw impression
The first impression is often vague: cool, vertical, movement, open space, heavy, red, tension.
The brain quickly wants to turn that into a story:
“High. Hard surface. Wind. A lot of space below.”
“A bridge.”
If the target later turns out to be an observation tower, some raw characteristics may have been useful while the label “bridge” was wrong.
So first record what you actually experienced. Only afterwards add what you think it means.
A practical format is:
R: raw impression · I: interpretation · C: confidence.
Prevent retrospective rewriting
One of the easiest ways to fool yourself is to let your memory of the earlier impression shift after you see the target.
So create a fixed version before feedback:
After the reveal you may annotate, but do not change the original response.
Feedback may change your next model. It may not change your previous answer.
No information leakage
If you are testing whether information becomes available beyond a known sensory route, ordinary information routes need to be excluded as well as possible.
You do not know the target during the session.
The person interacting with you also does not know the target and therefore cannot unintentionally provide cues.
For a solo experiment, a computer can randomly select the target only after your response has been locked. Then the current target could not even have leaked through subtle cues.
The stronger the claim, the more important it becomes to make ordinary information routes practically impossible.
Prevent predictable patterns
If you or a helper chooses the targets, preferences and patterns can emerge without anyone noticing.
So whenever possible, use a predefined pool and random selection.
For forced-choice tests that might mean:
4 possible targets → one randomly selected target → chance level per attempt = 25%
For free-response descriptions, you can use one target plus several randomly selected decoys. An independent judge who does not know which one is the real target can then rank the matches.
Randomization makes the target impersonal. Blinding prevents information leakage. Strong tests usually need both.
No elastic scoring
Vague information is easy to make fit afterwards. “Water”, “movement” and “something large,” for example, occur in many possible targets.
So define in advance how scoring works.
You choose from predefined options. The chance expectation is clear.
An independent judge compares your response with the target and decoys without knowing which one is real.
For your own qualitative scoring, you can predefine categories such as shape, movement, temperature, material, color, indoor/outdoor, scale and sound.
A scoring method invented after the result mostly tests how creatively you can match afterwards.
No highlight reel
Human memory is poor at neutrally tracking striking events. A spectacular hit remains memorable. Twelve boring misses easily disappear from the story.
So:
A good method does not protect your results from misses. It makes the misses informative.
Confidence is data too
Before feedback, give each impression a simple confidence score, for example 0–100%.
Across multiple attempts you can then investigate something more important than accuracy alone:
Are your strongest certainties actually correct more often?
If your 90% impressions are no better than your 40% impressions, your internal certainty feels more convincing than it is informative.
Research on metacognition shows that confidence and objective performance can diverge. Feedback can also improve confidence calibration even when underlying perceptual sensitivity does not change.
That makes calibration a useful training target: not “becoming more certain,” but knowing better when your information is reliable and when it is not.
Replication over anecdote
A single striking correspondence can arise by chance, especially when many possible details are available for retrospective matching.
So run a predefined number of attempts without changing the rules in between.
With forced-choice designs, results can be compared with the known chance level. With free-response descriptions, a predefined blind-ranking method or another standardized score is needed.
The more degrees of freedom you have in target selection, scoring and analysis, the more trials you need before a pattern becomes convincing.
For serious statistical evidence, sample size and analysis should also be specified in advance. For a personal experiment, the simpler starting point is: run a fixed batch and only then judge the whole set.
“Something involving water will happen” is cheap
Some statements have a high probability of fitting in one way or another.
If you often dream about work, relationships, driving, money or illness, striking overlaps with daily life will sometimes happen naturally.
So ask:
A hit is stronger when it is more specific, pre-recorded, independently verifiable and less likely by chance.
Test a domain with real feedback
You are not testing whether you “are intuitive.” You are testing a specific predictive task under repeatable conditions.
Minimize information leakage
An even stronger solo-test variant: have the computer randomly choose the real target from the pool only after your response has been locked.
Then you are testing information correspondence when the target was not available during the session for ordinary leakage.
A cleaner precognition test
Open predictions about “what will happen in my life tomorrow” are difficult to score. A future randomly selected target is much cleaner.
If this performed above chance, it would create a more interesting information problem than a target that already existed before the session.
Even then, MBT's model of possible futures would not automatically be the correct explanation. First you have an effect; only then do you have a theory question.
That is not a precognition test. You are no longer asking which outcome will appear later, but whether Intent can measurably shift the distribution of a genuinely uncertain outcome.
Here you are testing probability influence: Intent → outcome distribution. In precognition you are testing information access: possible future → response. Keep the two claims separate.
Same rule, different context
Record the dream immediately after waking, before seeing news, messages or the selected target. Use predefined targets if you want to test information.
A convincing relocation experience is not itself an information test. Add a hidden, randomly selected target that could only be identified from the claimed location.
Show the sender a random target while the receiver remains blind. No contact or feedback during the session; afterwards use forced choice or blind ranking.
How impressive the state feels is a separate variable. The information claim becomes stronger only through pre-recorded, verifiable correspondences.
Test what the state changes
If you suspect that meditation or a quieter mental state helps, make that itself testable.
For example:
10 minutes of quiet attention before the target task.
The same target task without prior meditation.
Randomize or pre-alternate the conditions and compare after enough sessions:
Then you are not testing “did meditation give me access to the larger system?”, but a more modest question: does this training condition change the quality of my task performance?
MBT should tolerate disconfirming evidence too
MBT includes the Psi Uncertainty Principle: Campbell's idea that paranormal effects in physical reality generally remain surrounded by enough uncertainty for ordinary causality to remain dominant and for “plausible deniability” to remain possible.
That is relevant as part of the MBT model. But in a personal experiment it creates an epistemic risk:
Hit → “evidence for psi”
Miss → “the Psi Uncertainty Principle blocked it”
If every possible outcome confirms the hypothesis, the test teaches you very little.
So define in advance what would actually lower your confidence in the claim. For example:
You do not need to reject MBT in order to require that your own experiment is allowed to fail.
Detection ≠ meaning
Even if no anomalous-information effect is ever found, this way of investigating trains something directly relevant to HSP.
You practice:
That is exactly the stance that makes a human system more updateable.
Make it boring enough to become useful
Use the same fields for every attempt:
| Field | Example |
|---|---|
| Date / time | 04 Sep 2026 · 10:15 |
| Test type | Hidden target |
| Target ID | T-1847 |
| Condition | 10 min meditation |
| Raw impression | cool · vertical · movement · outside |
| Interpretation | water / tower? |
| Confidence | 45% |
| Response locked at | 10:24 |
| Feedback | waterfall beside high rock face |
| Predefined score | 3/5 features |
| Hit / miss / rank | rank 1 of 5 |
| What did I learn? | raw spatial features better than object labels |
Do not change the original fields after feedback. Only add a new feedback or learning layer.
Match the stakes to the evidence
A personal experiment can be interesting without being suitable for important decisions.
Do not use uncertain impressions as the sole basis for medical, financial, legal or other high-impact decisions. Do not use them to make confident claims about other people that you cannot independently verify.
The more uncertain the information source and the greater the potential harm, the less weight any single intuitive or unusual impression should carry.
You can use a signal as a reason to look more carefully, seek additional information or ask a question. That is different from treating it as fact.
Method before explanation
If you want to know whether unusual experiences contain genuinely verifiable information, the test should be designed to give you as little room as possible to make the answer fit afterwards.
Use randomization and blinding. Separate raw impression from interpretation. Record confidence. Count misses as seriously as hits. Work in batches. Do not change scoring after you know the target. Define in advance what would count against your hypothesis.
The strongest conclusion from a good first test is not “MBT is true.” It is much simpler: “under these conditions there was / was not measurable information beyond what we reasonably expect from ordinary routes or chance.”
Only then does the source question come next.
Open science, metacognition and the MBT boundary
For transparency and pre-recording:
For confidence, feedback and calibration:
For the current research boundary around ESP:
That review highlights issues including small samples, inadequate multiple-comparison control, analytical flexibility, heterogeneous findings and limited replication, and concludes that definitive conclusions remain premature.
For MBT itself:
Methodological discipline does not prove MBT. It mainly prevents an interesting hypothesis from appearing stronger than the data allow.