Matched score gap
3.0%median 1.1% · 128 pairsThe result
A one-dimensional score hid a two-dimensional history.
At q44, hold and excursion paths differed by only 3.0% of the score range on average—1.1% at the median. If this scalar described the causal state completely, the branches should have converged. They moved farther apart instead.
State split at return
0.13095% bootstrap [0.125, 0.134]State split at q64
0.15595% bootstrap [0.150, 0.159]Delayed future shift
0.2581.01× an entire repertoire widthThe loop
Return versus hold.
Every comparison started from the same q32 organism. One branch stayed in its original score direction; the other made an excursion and returned. The final action pair was chosen jointly so both branches landed in the same directional half and as close as possible in q44 score.
High return
Hold high versus make a low excursion and return high.
HHH
HLH
Low return
Hold low versus make a high excursion and return low.
LLL
LHL
Memory through release
The hidden split grew monotonically.
The last intervention occurred at q41. From q44 through q64, there was no relaxation toward a common organism. Form-space separation increased at every fixed readout. The scalar score gap, initially tiny, reopened roughly ten-fold by q48 and continued growing.
Organism-state separation
Mean Hellinger distance between loop and hold.
Absolute score separation
The matched scalar ceases to remain matched under free dynamics.
q44 → q64 growth
+0.025state separation increased 19% after returnq48 score gap
0.550versus 0.052 at the matched returnq64 score gap
0.75595% bootstrap [0.567, 0.963]Tightest quarter
0.150q64 state split with mean return mismatch only 0.16%The causal future
The whole future cloud stayed displaced.
At q49 and again at q57, each branch received the same eight challenge seeds plus free continuation. The repertoire did not shrink or expand much. It moved: matched endpoints after loop versus hold were separated by almost exactly one complete repertoire width at both times.
Breadth: nearly unchanged
→
Location: one cloud-width apart
Immediate future shift
0.2581.01× repertoire widthDelayed future shift
0.2581.01× repertoire widthImmediate width change
0.001595% bootstrap [-0.0020, 0.0050]Delayed width change
-0.003295% bootstrap [-0.0075, 0.0007]A representative organism
Same score neighborhood, different body.
vibrant-orbiter-3596 (source 10), ending high, was selected from pairs with under 1% normalized score mismatch and near-median q44 state separation. The first pair is the matched return. The second pair is twenty passages later.
What this means
The scalar is not the causal state.
Two organisms can occupy nearly the same information-score coordinate while carrying different histories, different dynamics, and different maps of what happens next.
A causal phenotype
The result suggests treating an organism’s position in future-repertoire space as a phenotype in its own right. Width measures plasticity; position measures which possibilities are actually nearby.
The score is useful for steering, but it compresses multiple history-dependent causal states onto the same number. Development may be commitment by translation between those hidden sheets—not merely a scalar rising or falling.
The next swings
Now we can hunt this before organism birth.
- Continuous loopSweep through several matched score levels and trace an actual loop in causal-repertoire space. That would turn this return-memory result into a full hysteresis curve.
- Open-loop replayReplay the exact chosen seed histories without feedback. This separates memory in the physical pulse sequence from memory created by state-dependent control.
- Pre-birth scanMeasure repertoire position, not only Φ-like magnitude, before a body becomes coherent. A sudden translation may be the missing precursor to organism emergence.
- Event couplingAsk whether repertoire jumps lead fission, identity turnover, ecological succession, or recovery—our direct bridge back to the Levin-style precursor question.