The Optimum Appears as Signal Zero, Not Error Zero: The Information Gate and Why Bodies Flee Their Own Optimum
Abstract
Part 4 of a series. Part 1: 10.5281/zenodo.23587879 · Part 2: 10.5281/zenodo.23587670 · Part 3: 10.5281/zenodo.23526074 This article states the information-gate principle of a body-control framework developed over 24 years of field instruction, and shows that four of the eight predictions registered in Part 1 are facets of it. The gate. Motor control laws update as π_new = π_old + η·g·δ, where δ is the prediction error and g is a gate indicating whether that error reaches the learner. The update has no term for structure and no term for repetition count. The optimum is a sensory null. The most efficient posture requires the least supporting moment and therefore delivers the least signal to mechanoreceptors. For an inverted pendulum, the smallest detectable tilt is θ_min = S_th / (m·g·l): a dead band around the optimum whose width scales with the sensory threshold and shrinks with added mass. In the field, adding load works better than instruction for low-sensitivity individuals — raising plant gain rather than supplying the answer. A rationale corrected in this article. The author's field claim is that low-sensitivity individuals do not rest at the optimum but settle in inefficient postures. The original rationale, aversion to self-state uncertainty, fails when formalized as a trade: it predicts that low-sensitivity individuals flee no more, and on most measures less, than others. The replacement, from the author's field account, is not a trade: the habitual posture is already there and the optimum is invisible. A habit is corrected only if it lies outside the dead band; inside, the cost difference never arrives and the nearest existing motor program with the least discomfort persists. This matches use-dependent learning (Diedrichsen et al., 2010), and links the dead band to fatigue: when information or resources run short, the body falls back to its most practiced circuit. Two alternatives are kept and made testable: risk avoidance in an unstable plant (consistent with known lean-away under postural threat), and a passive dead-band equilibrium θ_passive = θ_min·K/(K − m·g·l), confirmed numerically, whose ratio to θ_min does not depend on sensory threshold. Three ways the gate closes — threshold, channel competition (mirror, external feedback), interrupt (pain, reflex) — and three matching ways to open it: amplify the signal (load, error exaggeration), remove the competing channel (no mirror, fade feedback), stop the interrupt. A section for RL and control readers poses the structure as questions: the optimum here is a null of observation rather than of reward. Four falsifiable predictions are registered (numbered from 13, continuing the series): - Prediction 13. After several days of practice standing 2–3 cm in front of or behind the EMG-minimizing optimum, unassisted stance stays shifted toward the practised side, more so in individuals with higher plantar sensory threshold (practised side × threshold interaction). Passive and risk accounts predict no lasting effect; use-dependent learning predicts no dependence on threshold. - Prediction 14. Lean-away under postural threat (known) is larger in individuals with higher plantar sensory threshold (threshold × height interaction). Predicted by the habit and risk accounts, not by the passive or uncertainty accounts. - Prediction 15. The retention benefit of error exaggeration is larger in individuals who cannot detect their own error at baseline. - Prediction 16. Gate type and effective intervention cross over: threshold-type individuals improve with signal amplification but not channel removal; competition-type individuals the reverse. Refutations are welcome. The weakest part is the corrected rationale, which has the form of use-dependent learning and matches the author's field account, but has not yet been linked to sensory threshold by any observation; the risk account remains a live alternative. The RL section has not been checked against literature. The author has no formal research training. Corrections will be issued as new versions, with earlier versions kept retrievable. The article body is in Japanese. An English summary, the registered predictions and the reference list are included in the document. Contact: [email protected]
Related
- Exploiting Low-Rank Structure in Genomic Relationship Matrices for Computationally Efficient Optimum Contribution Selection
- Optimum foraging area in a three-trophic food chain
- Compression through internal and external information bottlenecks resolves information intractability across brains, bodies and worlds
- Rhythms in Decoding Do Not Signal Pulsing Information
- Partitioning convergent modules separates phylogenetic signal from ecological information in mosaic fossils
BibTeX
@misc{kuriyama2026optimum,
title = {{The Optimum Appears as Signal Zero, Not Error Zero: The Information Gate and Why Bodies Flee Their Own Optimum}},
author = {CHIAKI KURIYAMA},
howpublished = {Zenodo (CERN European Organization for Nuclear Research)},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23609856},
url = {https://doi.org/10.5281/zenodo.23609856},
note = {CC BY},
}
RIS
TY - UNPB
TI - The Optimum Appears as Signal Zero, Not Error Zero: The Information Gate and Why Bodies Flee Their Own Optimum
AU - CHIAKI KURIYAMA
PY - 2026
DA - 2026/10/24
PB - Zenodo (CERN European Organization for Nuclear Research)
DO - 10.5281/zenodo.23609856
UR - https://doi.org/10.5281/zenodo.23609856
AB - Part 4 of a series. Part 1: 10.5281/zenodo.23587879 · Part 2: 10.5281/zenodo.23587670 · Part 3: 10.5281/zenodo.23526074 This article states the information-gate principle of a body-control framework developed over 24 years of field instruction, and shows that four of the eight predictions registered in Part 1 are facets of it. The gate. Motor control laws update as π_new = π_old + η·g·δ, where δ is the prediction error and g is a gate indicating whether that error reaches the learner. The update has no term for structure and no term for repetition count. The optimum is a sensory null. The most efficient posture requires the least supporting moment and therefore delivers the least signal to mechanoreceptors. For an inverted pendulum, the smallest detectable tilt is θ_min = S_th / (m·g·l): a dead band around the optimum whose width scales with the sensory threshold and shrinks with added mass. In the field, adding load works better than instruction for low-sensitivity individuals — raising plant gain rather than supplying the answer. A rationale corrected in this article. The author's field claim is that low-sensitivity individuals do not rest at the optimum but settle in inefficient postures. The original rationale, aversion to self-state uncertainty, fails when formalized as a trade: it predicts that low-sensitivity individuals flee no more, and on most measures less, than others. The replacement, from the author's field account, is not a trade: the habitual posture is already there and the optimum is invisible. A habit is corrected only if it lies outside the dead band; inside, the cost difference never arrives and the nearest existing motor program with the least discomfort persists. This matches use-dependent learning (Diedrichsen et al., 2010), and links the dead band to fatigue: when information or resources run short, the body falls back to its most practiced circuit. Two alternatives are kept and made testable: risk avoidance in an unstable plant (consistent with known lean-away under postural threat), and a passive dead-band equilibrium θ_passive = θ_min·K/(K − m·g·l), confirmed numerically, whose ratio to θ_min does not depend on sensory threshold. Three ways the gate closes — threshold, channel competition (mirror, external feedback), interrupt (pain, reflex) — and three matching ways to open it: amplify the signal (load, error exaggeration), remove the competing channel (no mirror, fade feedback), stop the interrupt. A section for RL and control readers poses the structure as questions: the optimum here is a null of observation rather than of reward. Four falsifiable predictions are registered (numbered from 13, continuing the series): - Prediction 13. After several days of practice standing 2–3 cm in front of or behind the EMG-minimizing optimum, unassisted stance stays shifted toward the practised side, more so in individuals with higher plantar sensory threshold (practised side × threshold interaction). Passive and risk accounts predict no lasting effect; use-dependent learning predicts no dependence on threshold. - Prediction 14. Lean-away under postural threat (known) is larger in individuals with higher plantar sensory threshold (threshold × height interaction). Predicted by the habit and risk accounts, not by the passive or uncertainty accounts. - Prediction 15. The retention benefit of error exaggeration is larger in individuals who cannot detect their own error at baseline. - Prediction 16. Gate type and effective intervention cross over: threshold-type individuals improve with signal amplification but not channel removal; competition-type individuals the reverse. Refutations are welcome. The weakest part is the corrected rationale, which has the form of use-dependent learning and matches the author's field account, but has not yet been linked to sensory threshold by any observation; the risk account remains a live alternative. The RL section has not been checked against literature. The author has no formal research training. Corrections will be issued as new versions, with earlier versions kept retrievable. The article body is in Japanese. An English summary, the registered predictions and the reference list are included in the document. Contact: [email protected]
ER -