The Sovereignty Foundation™ — Research paper · SF-RP-2026-006
The Room and the Bullseye
Why a research programme should aim to eliminate rather than to confirm — stated as a proposition about area, not about virtue
The case for falsification is usually made as an argument about intellectual honesty. It can be made instead as an argument about measure. A target that counts as right occupies almost none of the space of possible positions; the complement occupies almost all of it. A method that aims at the first is aiming where it will almost never land, and a method that aims at the second is aiming where it almost always will. This paper states that asymmetry, corrects the naive reading of it — random elimination narrows nothing — and derives from the correction why a pass/fail bar must be written before the run, why scope clauses are the boundary of a claimed region rather than hedging, and why a programme whose decided results are mostly negative is a programme working as designed.
§ 1The asymmetry
Stand in the middle of a round room. Somewhere on the far wall is a bullseye, and the bullseye is being right. You have a dart.
Now look at the room. Almost all of it is not the bullseye. The target is a point with essentially no area; the room is volume. That asymmetry is not a remark about how hard research feels — it is a fact about measure, and it has a consequence most methodology leaves on the table: a throw lands in wrong-space with probability near one.
Fig. 1 — One hundred and fifty throws, uniform over the area. The bullseye is drawn to scale.
So if you set out to be right, you are aiming at the one region you are least likely to hit, and you will spend the programme missing it. If you set out to establish that something is wrong, you are aiming at almost the whole room. That is reachable. It is also evidential in a way confirmation is not: a disproof is a fact about the target, and it does not decay when someone else runs it.
§ 2Why a dart is not enough
Here is where the picture, taken naively, misleads — and the correction is the whole method.
Hitting wrong-space at random tells you almost nothing. Every point in that room is wrong. Learning that one particular point is wrong removes a negligible share of the volume, and a programme of random throws could run for years and narrow nothing. The value of a disproof is not that it landed. It is how much room it eliminates.
Which means the instrument cannot be a dart. It has to be a cut: something that divides the room into a part you keep and a part you discard, whichever way the result falls. And the thing that converts a throw into a division is entirely procedural — the pass/fail bar has to be written down before the run. A bar set in advance is a plane through the room. A bar set afterward is a label on wherever the dart happened to land, and it divides nothing.
Fig. 2 — The same throw, with the bar written at two different times.
This is why pre-registration is load-bearing rather than hygienic, and why a quantity that was optimised for cannot anchor the bar that follows it: a plane positioned after seeing the data passes through wherever you would like it to.
§ 3What a verdict does to the volume
A cell carries one pre-registered bar and returns one of three verdicts. Two of them divide the room. The third does not, and the difference is not a matter of bookkeeping.
An invalid verdict is the throw that missed the room entirely: a precondition failed, and the instrument was not measuring what it was pointed at. It must not be recorded as a failure of the claim. An invalid result removes no volume, and filing it as a disproof shrinks the surviving region on evidence that was never gathered — the one error in this scheme that cannot be detected downstream, because the record will look exactly like a narrowing that happened.
§ 4What accumulates
Cuts compose. Each one that lands removes a half-space, and what remains is the intersection of everything not yet eliminated. The region shrinks, changes shape, and — this is the part worth sitting with — it is not the truth. It is the set of positions that have survived every test so far.
Fig. 4 — Five cuts. What is left has no claim to correctness; only to not having been removed.
A programme run this way produces an unusual record. Ours currently stands as follows.
Fifty-three per cent of decided claims are negative. That ratio is not a sign of a programme going badly. Under this geometry it is what a programme going well looks like, because the negative results are the ones doing the narrowing. A disproven cell is kept with the same standing as a proven one, cited the same way, and never quietly retired.
§ 5What a law costs to state
The geometry makes something precise that is otherwise a matter of taste. When you state a result, you are claiming a region. The scope clauses — at this corpus size, at this number of rows per parameter, under this frame — are the boundary of that region. They are not hedging. They are the difference between claiming what you cleared and claiming the room.
Fig. 5 — Two circles: what the evidence cleared, and what the sentence claims once its conditions are dropped.
It is also why apparent contradictions between campaigns usually dissolve. Read side by side, most turn out to be one law measured at two scopes, where one side carried its clause and the other dropped it. A result that holds, inverts, and dies is not three findings in conflict — it is one curve with three readings taken at different points along it, and the disagreement was an artefact of stating any of them without saying where you were standing.
§ 6The limit
You never hit the bullseye. There is no cut that establishes a position as correct, and no number of survived tests that converts a surviving region into a true one. What you get is a smaller room, and an exact account of how it got smaller.
I think that is the better object anyway. A claim to be right is not checkable by anyone who was not there. A record of what has been eliminated — each bar written before its run, each result sealed after it — can be picked up by a stranger and rerun.
Being right is a position. Having narrowed is a provenance, and provenance is the thing that survives the person who produced it.