A morphology-only target became a later difference in causal responsiveness.
In the open starting state, expansive-versus-compact control made no clear gate-score difference when feedback ended. After 128 untouched steps, it did: +0.00530 [+0.00222, +0.00813].
The late information difference is not a trivial readout of “how visually different were the bodies?” In the open state, organism-level morphology and gate effects were essentially uncorrelated at +128.
Visible form moved first. Information geometry followed on its own clock.
Visible expansive minus compact body
The controller created an immediate morphology split. By +128, the population interval includes zero.
Expansive minus compact gate score
The closed starting state shows an earlier transient tendency. The open starting state develops a clear difference only at +128.
The original information gate also decays slowly.
Untouched open minus closed gate score
The open–closed gate difference falls from +0.01681 to +0.00644, but remains visible after 128 release steps.
So the system carries at least two memories at once: the older developmental mode, and a newer target-dependent difference written by morphology feedback.
The images converge before the causal response does.
Source 34 is mechanically closest to the panel mean in the late open-state body and gate effects. Top row: compact target. Bottom row: expansive target.
Information geometry and form are taking turns carrying the organism’s history.
This is much closer to the phenomenon we came looking for: a hidden causal organization that precedes form, shapes form, and can itself be rewritten by the history of form-generating interactions.
Now test whether the returned information state changes the next body.
Give the late compact and expansive branches the same fresh action after their visible forms have converged. If their bodies diverge again according to the hidden gate difference, we have completed the causal loop: information → body → information → body.