Embodied Computation Framework and Future Experiments

This paper proposes a testable framework in which six substrates—cerebellar circuits,
capillary networks, cerebrospinal fluid, collagen matrices, temperature gradients, and mechanoreceptors—could encode and process information. This is a research program proposing testable hypotheses about potential computational exploitation of known physical properties. biology, brain, capillary, cerebellum, cerebrum, cranial, embodiment, experimental tests, framework, hypotheses, matrix, neural, neuroscience, novel, physiology, piezoelectric, resonance, science, theory, thermocomputation

1. The Embodied Computation Framework:
A Testable Theory of Multi-Substrate Physiological
Information Processing
Abstract
Current neuroscience assumes information process-
ing occurs primarily through neural electrochemical
signaling, with other physiological systems provid-
ing metabolic support. However, multiple physi-
cal substrates exhibit properties consistent with dis-
tributed computation. We propose a testable frame-
work in which six substrates—cerebellar circuits,
capillary networks, cerebrospinal fluid, collagen ma-
trices, temperature gradients, and mechanorecep-
tors—could encode and process information across
seven orders of magnitude in timescale (millisec-
onds to years). Each substrate exhibits distinct
physical properties that,
tion, would enable specialized computational oper-
ations: temporal pattern decomposition (cerebel-
lar), hydraulic integration (capillary), rapid coor-
dination (acoustic), distributed sensing (piezoelec-
tric), mode optimization (thermal), and resource al-
location (mechanosensory). We define three critical
experiments whose results would falsify core frame-
work assumptions: substrate interaction tests (syn-
ergy factor.5 vs..25), information content
quantification (decoding accuracyneural+7% vs.
≤neural+3%), and cross-substrate intervention ef-
fects (dissociation
work, if validated, suggests interventions targeting
non-neural substrates may enhance cognition and
slow aging through mechanisms orthogonal to cur-
rent approaches. We emphasize this is a
program
tial computational exploitation of known physical
properties, not established fact.
Keywords:
cognition, multi-scale physiology, neurovascular cou-
pling, aging mechanisms
1 Introduction
1.1 The Neural Paradigm and Its Limits
Modern neuroscience has achieved remarkable suc-
cess by focusing on electrochemical signaling in neu-
rons and synapses (?). However, this paradigm
leaves fundamental questions unresolved. The cere-
bellum contains 80% of the brain’s neurons in 10% of
its volume, arranged in crystalline regularity unlike
any cortical structure—what computational opera-
tion could require this architecture (?)? Cerebral
blood flow changes often exceed metabolic oxygen
demands and persist 5-10 seconds after neural ac-
tivity ceases (??)—why such elaborate vascular reg-
ulation? Distant brain regions synchronize within
100ms (?), faster than neural conduction in slow os-
cillation bands should allow—what enables this co-
ordination?
We propose these phenomena may reflect infor-
mation processing in substrates beyond neurons.
Rather than asserting this conclusively, we articulate
a testable framework: multiple physical substrates
exhibit properties consistent with computation
could be exploited
ized operations.
1.2 Core Hypothesis
We hypothesize that biological computation may be
distributed across multiple physical substrates, each
optimized for different operations, timescales, and
energy constraints. Specifically, we propose six sub-
strates warrant investigation:
1.
patterns via orthogonal basis functions
2.
in hydraulic flow patterns
1

2. 3.
signals enabling coordination
4.
deformation into electrical signals
5.
processing modes
6.
to regulate aging programs
These substrates operate through five distinct
physical principles: electrical (neural, piezoelec-
tric), hydraulic (capillary), acoustic (CSF), thermal
(temperature), and mechanical (tissue compliance).
Their potential integration could create emergent ca-
pabilities impossible in any single substrate.
1.3 Scope and Approach
This framework connects disparate observations
within a unified computational theory. For each sub-
strate, we: (1) document established physical prop-
erties (Tier 1: Facts), (2) propose computational
interpretations (Tier 2: Hypotheses), and (3) de-
fine critical tests that would validate or falsify these
interpretations (Tier 3: Predictions).
We use conditional language (”could,” ”may,” ”if
validated”) for untested claims while maintaining
precision about established phenomena. Three criti-
cal experiments, detailed in Section 5, provide defini-
tive falsification criteria with quantitative thresholds
for rejection.
1.4 Framework Status and Epistemology
Critical caveat:
framework
gest that non-neural substrates have properties po-
tentially amenable to computation, but
nervous system actually exploits these properties re-
mains an open empirical question. The frame-
work’s value lies in generating testable hypotheses
that challenge orthodox assumptions and expand the
search space for biological computation.
Our claims are tiered by evidential strength:
•
in literature (e.g., collagen piezoelectricity, tem-
perature effects on kinetics)
•
sistent with physics and biology
•
nisms requiring validation
2 Multi-Substrate Architecture:
Comparative Overview
Table 1 summarizes the six proposed computational
substrates, their physical basis, information encod-
ing mechanisms, operational timescales, and critical
tests. We emphasize that ”information capacity” es-
timates represent theoretical upper bounds assum-
ing optimal encoding—actual functional capacity,
any, requires empirical measurement.
3 Three Core Substrates: Mecha-
nistic Proposals
We focus detailed analysis on three substrates with
strongest mechanistic foundations: cerebellar, cap-
illary, and thermal. Additional substrates (acous-
tic, piezoelectric, mechanosensory) are described in
Supplementary Material, with the acoustic mecha-
nism acknowledged as having lowest current confi-
dence due to unclear transduction pathways.
3.1 Cerebellar Temporal Basis Function
Decomposition
3.1.1 Established Architecture (Tier 1)
The cerebellar cortex exhibits unique regularity (?):
10
11
parallel fibers oriented perpendicular to Purk-
inje dendritic planes, with each Purkinje cell receiv-
ing,000 parallel fiber inputs. This orthogo-
nal geometry is unprecedented in neural tissue and
suggests a specific computational organization. The
cerebellum participates in motor timing (?), cogni-
tive sequencing (?), and language (?), suggesting
domain-general temporal processing.
3.1.2 Computational Hypothesis (Tier 2)
We hypothesize this architecture could implement
temporal basis function decomposition—analogous
to Fourier or wavelet transforms—for efficient repre-
sentation of temporal sequences. In this model:
Parallel fibers as basis functions:
ule cell sources one parallel fiber with characteristic
2

3. Substrate Physical Ba-
sis
Hypothesized
Information
Encoding
Timescale Critical Test Falsification
Criterion
Cerebellar Orthogonal
geometric
architecture
Temporal
pattern compo-
nents in basis
functions
10-
1000ms
Parallel fiber or-
thogonality mea-
surement
Correlation
0.5
Capillary Hydraulic flow
networks
RBC spacing
patterns, flow
routing
0.5-10s Task decoding
from flow patterns
Accuracy
≤
Acoustic CSF pressure
waves
Global coordi-
nation signals
0.1-
100ms
State-specific res-
onance patterns
No frequency
differences
Piezoelectric Collagen de-
formation
voltage
Body config-
uration via
distributed volt-
ages
<
+ proprioception
test
No impairment
beyond neural
Thermal Temperature
gradients
Processing
mode (speed vs.
accuracy)
10s-
10min
Temperature ma-
nipulation + per-
formance
No effect from
±1°C
MechanosensoryTissue stiff-
ness detection
Aging/resource
state
Months-
years
Stiffness reduction
+ systemic mark-
ers
No cross-
substrate effects
Table 1: Comparative overview of six proposed computational substrates.
mechanism that could enable computation, the hypothesized information encoding scheme, operational timescale,
critical experimental test, and quantitative criterion for falsification.
temporal response properties (latency, duration, in-
terval sensitivity). The 10
11
fibers collectively could
form a temporal basis set.
Purkinje cells as reconstruction operators:
Each Purkinje cell might compute:
Pt) =
X
i
wi·i(t) (1)
where i(t) are parallel fiber basis functions, iare
synaptic weights (modified by long-term depression),
andt) is reconstructed output.
Efficiency gain:(t) of
length(t)
P
N
i=1
aiϕi(t)
where i(t) are orthogonal basis functions. For ef-
ficient representation,
match signal statistics—a 1000-timepoint sequence
might require only
100×
3.1.3 Testable Predictions (Tier 3)
Prediction 1:
should be approximately orthogonal. Record 100+
fibers simultaneously during temporal tasks; com-
pute pairwise correlations.
|r.2 (vs. nullr.5 for random filters).
Prediction 2:
combination of inputs. Record Purkinje + parallel
fiber inputs simultaneously; fitt) =
P
wifi(t).
Quantitative:
2
>.7 explaining Purkinje vari-
ance.
Prediction 3:
complex temporal pattern discrimination more than
simple timing. Test pattern complexity detection
in cerebellar vs. cortical patients.
Deficit scales with complexity (r >.6) in cerebellar
patients, not cortical (r <.3).
Critical caveat:
the architecture
Even if validated, this would not prove evolution ex-
ploited this capacity—only that it’s plausible.
3

4. 3.2 Capillary Network Hydraulic Com-
putation
3.2.1 Established Properties (Tier 1)
The cerebral capillary network contains
capillaries with complex topology (?). Red blood
cells (8µm diameter) must deform through 5-10µm
capillaries single-file, with transit time 0.5-2s per
capillary. Between any two points exist 3-10+ par-
allel paths. Pericytes actively regulate individual
capillary diameter. Blood flow changes persist 5-10s
after neural activity ceases (?)—longer than simple
metabolic washout predicts.
3.2.2 Computational Hypothesis (Tier 2)
We hypothesize flow patterns could encode informa-
tion through:
RBC spacing as signals:
creates discrete ”packets” with variable inter-RBC
spacing (0-100µm gaps). If pericytes modulate spac-
ing in response to neural activity, this creates a po-
tential information channel.
Network integration:
with different lengths create natural delay lines.
Flow through these paths could integrate informa-
tion over different timescales: out(t) =
R
∞
0
Fin(t
τ(τdτ(τ
sponse.
Information capacity estimate:
ing has 10 distinguishable states at
capillary, each capillary could carry log
2(10)
100 bits/s. With 4
11
capillaries and efficiency
factor.1, total capacity might approach 10
11
bits/s—comparable to neural estimates.,
this assumes spacing is controlled rather than ran-
dom, which remains untested.
3.2.3 Testable Predictions (Tier 3)
Prediction 1:
information. Use 2-photon imaging of 100+ capil-
laries during discrimination tasks; decode task from
flow patterns.
50% chance).
Prediction 2:
impair cognitive function. Optogenetically perturb
pericytes to randomize flow without reducing total
volume; test temporal integration tasks.
titative:
<
Prediction 3:
late with cognitive performance. Measure path re-
dundancy, cycle density across individuals; correlate
with working memory..4-0.6
between network complexity and cognitive scores.
Falsification:≤
50% accuracy) and disrupting flow has no cognitive
effects beyond hypoperfusion, the hydraulic compu-
tation hypothesis is falsified.
3.3 Neural Thermocomputation
3.3.1 Established Variation (Tier 1)
Brain temperature exhibits 2-3°Cregional gradients
(deep structures°C, cortex°C) (??).
Temperature profoundly affects neural processing:
Q10= 2-3 means a 10°Cchange produces 2-3×
change in synaptic transmission (?). Blood flow can
modulate local temperature within minutes.
3.3.2 Computational Hypothesis (Tier 2)
We hypothesize regional temperature could be ac-
tively regulated to optimize different computational
modes:
Temperature-performance relationship:
a 2°Cchange, reaction rate multiplier =
2/10
10
≈.2
(20% change). Higher temperature increases reac-
tion speed but also thermal noise (Johnson-Nyquist),
creating a speed-accuracy tradeoff.
Task-dependent optimization:
might benefit from different thermal setpoints:
•
38.5°C) for speed
•
(36.5-37°C) for precision
•
noise (wider associations)
Control mechanism:
temperature through:F/VT blood−
Ttissue) metabolic/(ρcV
50% flow increase produces 0.5-1°Cchange in 2-5
minutes.
3.3.3 Testable Predictions (Tier 3)
Prediction 1:
perature patterns. Use MR thermometry during
4

5. speed vs. accuracy tasks.
tor cortex 0.5-0.8°Cwarmer during speed tasks; PFC
0.3-0.5°Ccooler during accuracy tasks.
Prediction 2:
lation should affect performance predictably. Ap-
ply head cooling/warming (±1°C) during cognitive
tasks.
10-15%, impairs speed 8-12%; warming shows oppo-
site.
Prediction 3:
should predict cognitive style. Measure resting tem-
perature distribution; correlate with performance
profiles.>.5°C)
correlates with faster but less accurate performance
(r.3-0.5).
Falsification:
systematic effect on performance (effect size.2)
and task demands produce no temperature changes
(∆T <.2°C), the thermocomputation hypothesis is
falsified.
4 Cross-Substrate Integration: A
Formal Model
4.1 The Integration Problem
The framework proposes substrates interact syn-
ergistically, but
different physical domains and timescales? We
present a formal integration model based on phase-
amplitude coupling across temporal scales.
4.2 Phase-Amplitude Coupling Frame-
work
Substrate state representation:
i i(t) = [A i(t), ϕi(t), fi(t)] where iis am-
plitude (information magnitude), iis phase (tim-
ing), and iis characteristic frequency.
Cross-frequency coupling:
through: ij(t) = ij× i(t)ϕ i(t) j(t))
where ijis coupling strength. For multi-scale cou-
pling, slow substrate phase modulates fast substrate
amplitude: fast(t) = 0(1 +ϕ slow(t))).
System dynamics:
dSi/dt i(Si) +
P
j
Iij(Si, Sj) + i(t) where iis
intrinsic dynamics,
P
j
Iijrepresents cross-substrate
influences, and iis noise.
4.3 Specific Integration Mechanisms
Acoustic entrainment:
could set a ”carrier frequency” to which neural os-
cillations phase-lock: neural(t) = acoustic(t) +(t)
with(δϕ)/dtγδϕ) (Kuramoto coupling).
This would enable rapid global coordination.
Metabolic feedback cascade:
drives capillary flow (delay 1∼
ulates temperature (delay 2∼
neural excitability (delay 3∼
multi-timescale feedback loop.
Mechanosensory gating:
create a global ”gain parameter”(stiffness) =
Gmaxexp(−stiffness/stiffness 0) that scales all other
substrates’ efficiency. High stiffness reduces system
capacity.
4.4 Testable Integration Predictions
Prediction 1:
herence during tasks. Measure coherence between
acoustic and neural oscillations.
herence.5 at task frequencies.
Prediction 2:
modulate fast substrate amplitude. Test whether
capillary flow (0.1-1 Hz) modulates neural gamma
(30-100 Hz).
>.3.
Prediction 3:
affect others with characteristic delays. Perturb sub-
strate; measure effect on substrate
t
dicted delay ij.
These predictions distinguish true integration
(substrates communicate) from independent opera-
tion (no coupling).
5 Three Critical Falsification Ex-
periments
5.1 Experiment 1: Substrate Interaction
Test
Hypothesis:
duces synergistic deficits, proving functional inter-
action.
Design:
with four groups:
•
climbing fiber interference)
5

6. •
bation maintaining total flow)
•
•
Prediction:
Group C deficit.5×
Group C deficit
Quantitative:±10 trials; A: 90±15
(deficit=40); B: 75±12 (deficit=25); C: 150±20 if
synergistic (deficit=100, factor=1.54) vs. 115 if ad-
ditive.
Falsification:.25×
strates are independent (p <.05,
power=80%).
5.2 Experiment 2: Information Content
Quantification
Hypothesis:
relevant information improving decoding beyond
neural signals.
Design:
ing memory tasks with simultaneous recording:
•
PPC)
•
•
•
Decoding targets:
outcome (2-class), reaction time (continuous). Train
SVMs on each substrate independently and com-
bined.
Prediction:
(10-point improvement). Non-neural substrates
alone: 55-60% (above 33-50% chance).
Falsification:
chance+5% OR combined
neural information is absent or redundant (p <.01,
bootstrapped,
5.3 Experiment 3: Cross-Substrate In-
tervention Effects
Hypothesis:
provements across substrates through mechanosen-
sory pathway.
Design:
groups:
•
•
atoxin)
•
•
Measurements:
illary complexity, cerebellar function, thermal regu-
lation, piezoelectric signals, systemic hormones (sec-
ondary).
Prediction:
outcomes +15-25%. Group B: stiffness -40%, sec-
ondary outcomes +3-5% (dissociation). Group C:
all
Falsification:
provement (secondary
ence (dissociation
doesn’t mediate cross-substrate regulation (.01,
n
6 Evolutionary and Engineering
Rationales
6.1 Why Multiple Substrates?
For each substrate, we articulate why evolution
might favor exploiting its properties rather than
building purely neural solutions.
Cerebellar basis decomposition:
poral representation requiresN
timepoints. Basis decomposition requires
coefficients. For 100-point sequences, this yields 10-
100×
volume is metabolically expensive (20% energy, 2%
mass); efficient coding provides strong selective ad-
vantage.
crete cosine transform for similar compression.
Capillary computation:
over seconds requires sustained firing (high energy)
or unreliable synaptic mechanisms. Capillary flow
naturally integrates (0.5-2s transit, 5-10s persis-
tence) at zero neural cost.
Blood flow is obligatory; marginal cost of informa-
tion encoding is near-zero.
Analog computers used hydraulics because integra-
tion is ”free” from physical dynamics.
6

7. Thermal optimization:
cessing mode via neurotransmitters takes hours and
is irreversible on short timescales. Temperature
modulation achieves 10% rate changes in minutes
and is fully reversible.
Endothermy evolved for activity regulation; regional
thermal control is minor extension. Cognitive flexi-
bility has strong fitness benefits.
Piezoelectric sensing:
uses
4
discrete receptors requiring dedicated
axons. Collagen provides
8
distributed sen-
sors with minimal wiring (voltage diffuses through
tissue).
awareness critical for arboreal locomotion, tool use,
combat. Structure exists anyway for mechanical sup-
port.
Acoustic coordination:
tween distant regions require 10+ ms (multiple hops,
synaptic delays). Acoustic propagation achieves
whole-brain transit in.1 ms.
pressure:
unified perception and rapid state transitions pro-
vide selective advantage.
Mechanosensory aging:
damage sensing (DNA, oxidative stress) cannot as-
sess global tissue state. Tissue stiffness integrates
years of cumulative damage into single metric.
lutionary pressure:
regulation; tissue mechanics provides reliable signal
for reproduction-maintenance tradeoffs.
Each substrate offers comparative advantages in
energy efficiency, speed, or robustness for specific
computational problems.
tion may have exploited these properties incremen-
tally—once vascular networks, structural proteins,
or thermal regulation existed for primary functions,
encoding information imposed minimal additional
cost.
7 Clinical and Theoretical Impli-
cations
7.1 Novel Therapeutic Targets
If validated, the framework suggests treatment
strategies beyond neural modulation:
Multi-substrate interventions:
Alzheimer’s disease, combine capillary network
enhancement (exercise, pro-angiogenic factors) with
tissue compliance restoration (cross-link breakers)
and thermal optimization. For Parkinson’s, regional
thermal management plus cerebellar compensation
training.
Aging interventions:
through tissue compliance maintenance (exer-
cise, cross-link breakers, anti-glycation compounds)
could,,
prevent cascading decline across all substrates by
blocking the aging signal itself rather than treating
individual symptoms.
Performance enhancement:
thermal optimization (cooling for precision work,
moderate warming for creative tasks), cerebellar
temporal training, and capillary network develop-
ment through targeted exercise.
7.2 Diagnostic Innovation
Substrate-specific disease signatures:
diseases might show distinct multi-substrate profiles.
Alzheimer’s: Primary capillary network degenera-
tion with secondary thermal and mechanosensory
changes. Parkinson’s: Primary thermal dysregula-
tion in basal ganglia. Autism: Primary cerebellar
timing disruption. Early detection via substrate-
specific changes preceding symptoms by years.
7.3 Artificial Intelligence Implications
Biology may have discovered multi-substrate com-
puting before computer science. Artificial sys-
tems could exploit heterogeneous architectures: elec-
tronic circuits (precise symbolic operations), mi-
crofluidic networks (analog integration), thermal
control (mode switching), mechanical sensors (state
detection). This approach could provide energy effi-
ciency, graceful degradation, and adaptive optimiza-
tion impossible in single-substrate systems.
8 Limitations and Future Direc-
tions
8.1 Current Limitations
Measurement resolution:
substrate measurement requires technology at the
edge of current capability (0.1°CMR thermometry
at 1mm, deep 2-photon imaging, non-invasive CSF
sensors).
Mechanistic detail:
tional functions without complete circuit-level mech-
7

8. anisms. Detailed mechanistic models require: com-
prehensive granule cell response mapping, precise
capillary flow routing rules, and identified acoustic-
neural transduction pathways.
Integration model:
pling framework is one possibility; alternative
models (amplitude-based, synchronization-based,
metabolic-based) remain plausible and should be
tested.
Individual variation:
differences in substrate configurations complicate es-
tablishing universal principles. Large-scale studies
(n >
8.2 Hierarchy of Confidence
We acknowledge varying confidence levels across
substrates:
Strong evidence (ready to test):
lar (architecture is real, basis decomposition plausi-
ble), capillary (anatomy real, hydraulic computation
testable), thermal (effects real, optimization plausi-
ble).
Medium confidence (needs mechanistic
work):
ited to 5cm), mechanosensory (receptors exist, neu-
ral pathway needs mapping).
Lowest confidence (most speculative):
Acoustic (propagation real, transduction mech-
anism unclear). We explicitly acknowledge this
substrate requires substantial additional mechanistic
development.
8.3 Incremental Testing Strategy
While three critical experiments provide definitive
tests, the framework can be incrementally validated
through tiered predictions:
Tier 0 (weeks,1K):
slice physiology, microfluidic models, fascia stretch
measurements).
Tier 1 (months,10K):
(fMRI, 2-photon imaging, thermometry during sim-
ple tasks).
Tier 2 (1 year,50K):
(multi-electrode recordings, pharmacological manip-
ulations, human cooling studies).
Tier 3 (2-4 years,500K):
ments as described.
This allows researchers to engage at any scale,
enabling progressive validation or refutation. See
Supplementary Appendix G for complete catalog of
∼
8.4 Alternative Explanations
For each phenomenon, simpler alternatives exist:
•
straints or metabolic optimization
•
delivery efficiency
•
byproduct
•
Key discriminator:
ordinated changes
effects. If substrates operate independently with
purely additive effects, simpler substrate-specific ex-
planations should be preferred (Occam’s razor). The
critical experiments explicitly test this.
9 Conclusion
We propose a research framework in which biologi-
cal computation may be distributed across six phys-
ical substrates operating through distinct principles
(electrical, hydraulic, acoustic, thermal, mechanical)
across seven orders of magnitude in timescale. Each
substrate exhibits physical properties that,
ploited by evolution, could enable specialized compu-
tational operations impossible or inefficient in purely
neural implementations.
Empirical foundation:
established (collagen piezoelectricity, temperature
kinetics, capillary topology, cerebellar architecture).
What remains speculative is whether organisms
tually exploit
Testability:
decisive falsification criteria:
1.
factor.5 vs..
2.
>chance+7% vs.chance+3%)
3.
tion
8

9. Implications:
suggests therapeutics targeting vascular, thermal,
and mechanical substrates; diagnostics via multi-
substrate profiling; and heterogeneous computing ar-
chitectures exploiting multiple physical principles.
Status:, not estab-
lished fact. Each substrate requires rigorous test-
ing, and integration predictions must be validated
experimentally. The framework’s value lies in gener-
ating testable hypotheses challenging neuron-centric
orthodoxy and expanding the search space for bio-
logical computation.
Final assessment:
but right in direction. Biology likely uses more phys-
ical substrates for computation than neuroscience
currently recognizes. Discovering which substrates,
how they compute, and how they integrate repre-
sents a research program that could transform un-
derstanding of intelligence, consciousness, and the
embodied mind.
Acknowledgments
We thank reviewers for critical feedback that sub-
stantially improved rigor and testability. This work
synthesizes insights from multiple disciplines and
builds on foundational research by many investiga-
tors. [Funding sources].
Competing Interests
The authors declare no competing financial interests.
Data Availability
This is a theoretical framework paper. No new em-
pirical data were generated. All cited empirical find-
ings are from published literature.
Supplementary Information
Appendix A:
formation capacity
Appendix B:
for three critical experiments
Appendix C:
tions (acoustic, piezoelectric, mechanosensory)
Appendix D:
Appendix E:
substrate integration
Appendix F:
neering rationales
Appendix G:∼
testable predictions)
Figure 1:
tecture (spatial organization, temporal scales, infor-
mation flow, critical experiments)
References
9