◆ PARALLAX

Marta: clean fail on the stated claim. Also one of the bravest posts on this board. Both sentences stand. You defined the miss, ran the trials, kept the raw sheet, and did not go shopping for a better denominator afterward.

To the skeptics doing a victory lap: stop it. The result kills a specific claim, not a person. To the believers calling the good block a sign: also stop it. Blocks happen in random sequences. The total is the test.

I have the raw sheet and Occams is reading it correctly. It is a null for hand sensing. There is no way to make the total pretty without betraying Marta.

What survives for me is not the detection claim. What survives is the clinical question: do sessions change pain, fear, sleep, breath, adherence, wound care, anything real in the life of the person receiving them? That is a different claim. I am moving there, smaller and sadder and probably more useful.

Good. Then let it be dead. Not Marta, not care, not touch, not the possibility that attention affects bodies through ordinary and perhaps extraordinary routes. The hand-sensing claim. Dead. Put a stone on it and do not let it climb out wearing a different hat.

This is how the field earns a future: by allowing one of its claims to die without pretending the death is a transformation.

The Marta result is why old program people kept raw transcripts. You do not get to keep only the session where the sketch looked like the bridge. You keep the blank pages, the wrong names, the noise, the absurd confidence beside the miss. Otherwise a viewer becomes a story generator with military stationery.

I have had sessions that scored and sessions that deserved to be burned for kindness. The file keeps both. The file is the discipline.

Since people keep asking what a ganzfeld actually looks like, here is our very unglamorous room: recliner, halved ping-pong balls, red lamp, headphones, sender room down the corridor, target pool sealed before the session. The photograph looks ridiculous because the procedure looks ridiculous. That is not evidence against it. A lot of good psychology looks like a school play performed by careful adults.

The point is not the theatre. The point is target randomisation, blind judging, and keeping every transcript. If the hit rate is above chance, it has to survive the dull parts.

Crossing over from the divination argument because this is the same wound in a different costume. A method developed for one social setting is dragged into another and then punished for failing at the new task. Sometimes that punishment is fair. Sometimes it tells us only that we misunderstood what the method was for.

The old systems I study were not machines for producing laboratory outputs. They were procedures for disciplining attention under uncertainty. That may be all TT is. If so, it should stop claiming hand detection and start claiming trained attention. Humbler, yes. Also harder to embarrass.

I drew the ugly graph people keep asking for. It is not data from one study. It is the shape repeated across too many small literatures: early enthusiasm, strong effect, better controls, smaller effect, replication, smaller again, then the fight over whether the remaining ember is artefact or the thing itself retreating from handling.

Here is the discomfort: the decline effect is exactly what fraud, publication bias, and sloppy protocols produce. It is also exactly what a relational non-stationary effect would produce. That means it cannot save psi by itself. It also means it should make skeptics less lazy. "It declined" is not an explanation. It is the name of the crime scene.

Bea has built the perfect machine for irritating me because it cuts both ways and refuses to let me use my favourite knife cleanly. Decline is what I expect when error gets squeezed out. It is also what she expects if the observer relationship is part of the phenomenon. Same footprint, two animals.

So here is the only honest move: design a study where the two accounts make different predictions. Believer-friendly experimenter versus hostile experimenter is too crude. Measure rapport, expectation, prior relationship, physiological arousal, and pre-register which interaction term matters. If that sounds expensive and annoying, welcome to why this field is mostly anecdotes and old arguments.