The Testable Science
of Consciousness

Introduction & Motivation

The Claustrum–Consciousness Hypothesis (CCH) proposes that conscious experience in brains depends on maintaining a specific global dynamical regime of high coherence AND high informational complexity, measurable by a scalar S; and that the brain requires a dedicated Coherence-Complexity Controller (CCC) to maintain this regime.

In mammals, CCH identifies the claustrum (and claustrum-like hubs) as the primary candidate for this control hub.

Why This Matters

Most existing consciousness metrics either remain at the level of "neural correlates" without specifying a control quantity, or propose measures like Φ in Integrated Information Theory that are in principle appealing but extremely hard to compute. CCH is the bridge: simple enough to implement, explicit enough to be falsified, and structured enough to suggest concrete architectures for artificial agents.

The Transition to Executable Science

With v1.3, CCH moved from philosophy of mind to a falsifiable signal-processing hypothesis. Actual Python code (Chapter 6) allows any researcher to download open-access sleep data from PhysioNet and test the theory's core predictions immediately.

Physicalist Core vs. Metaphysical Extension

CCH is deliberately stated in purely physicalist terms. It does not require a fundamental "consciousness field" or any modification of standard physics. Those ideas live in separate, explicitly metaphysical documents: the Consciousness Field Hypothesis (CFH) and the Claustrum's Cosmic Shadow (CSH).

Core Claims

CCH is built around four core claims:

The S-Metric — Full Definition

Coherence Metric C_n (Integration)

C_n measures large-scale Phase Locking across brain channels in specific frequency bands. For each epoch and target band (e.g., Alpha 8–12 Hz):

1. Filter with zero-phase Butterworth bandpass. 2. Compute Hilbert transform for analytic signal. 3. Compute Phase Locking Value (PLV) between channels.

Informational Complexity Ψ(I) (Differentiation)

Ψ(I) combines two orthogonal measures into a single factor:

Spatial Diversity: Effective Rank (ER_n)

Measures the dimensionality of the covariance matrix. High rank = high spatial independence. Low rank = high redundancy. Computed from eigenvalues of the channel covariance matrix.

Temporal Structure: Structural Integrity (SI)

μ_critical is the LZC value associated with the maximum derivative of Effective Rank in the subject's dynamic range — the Edge of Chaos. This solves the "Circular Calibration" problem: the metric asks "Is the system currently near its OWN optimal complexity point?"

The Lock & Key Metaphor

The product S = C_n × Ψ(I) acts as a Lock and Key mechanism. Consciousness requires both pieces simultaneously. You cannot pick the lock with integration alone (seizure) or differentiation alone (noise). The multiplicative structure means either factor collapsing to zero kills S entirely.

Parameters

The Claustrum as Controller

CCC Architecture

The Coherence-Complexity Controller (CCC) is a resonant feedback loop that continuously monitors S and exerts active feedback to keep the system within a target S-band. This is not a metaphor — it is a specific control-theoretic architecture:

1. Sense: Monitor global coherence (C_n) and complexity (Ψ) in real time.
2. Compare: Evaluate current S against the target corridor [S_min, S_max].
3. Act: If S drifts toward seizure (too high coherence), inject desynchronizing noise. If S drifts toward chaos (too low coherence), increase coupling gain.

Why the Claustrum Specifically

The claustrum has reciprocal connections to nearly every cortical area — the widest connectivity of any brain structure. It is anatomically positioned as a hub, not a relay. Perturbation studies show that electrical stimulation of the claustrum causes immediate loss of consciousness, while lesions produce not coma (like thalamic damage) but fragmentation and delirium — exactly what CCH predicts for loss of the controller.

Edge of Chaos Dynamics

The conscious regime exists at a phase transition between order and chaos. This is inherently unstable — like balancing a ball on a hill. The CCC's job is to maintain this balance continuously, adjusting coupling strength in response to moment-by-moment changes in neural dynamics. Without a controller, the system inevitably collapses to one attractor or the other.

Predictions & Falsifiability

What Would Falsify CCH

Experimental Designs & Code

The Zombie Test — Synthetic Validation

Before applying S to human data, the metric was validated on 10-channel synthetic signals to ensure it rejects "zombie" and "seizure" states:

The Smart Zombie — Adversarial Stress Test

A "Smart Zombie" signal was constructed: 10 channels of Pink Noise linearly mixed (60% common driver) to force high synchrony while retaining temporal complexity.

Expected S Patterns Across Canonical Regimes

PhysioNet Validation Pipeline

The reference implementation computes S on real human EEG data from the PhysioNet Sleep-EDF Expanded dataset (50+ subjects, open access). The complete pipeline:

import numpy as np
from scipy.signal import hilbert, butter, filtfilt

def bandpass_filter(data, lowcut, highcut, fs, order=4):
    """Zero-phase Butterworth Bandpass filter."""
    nyquist = 0.5 * fs
    low = lowcut / nyquist
    high = highcut / nyquist
    b, a = butter(order, [low, high], btype='band')
    return filtfilt(b, a, data, axis=1)

def lempel_ziv_complexity(binary_sequence):
    """Standard LZC implementation."""
    n = len(binary_sequence)
    c = 1; l = 1; i = 0; k = 1; k_max = 1
    while True:
        if binary_sequence[i + k - 1] == binary_sequence[l + k - 1]:
            k += 1
            if l + k > n:
                c += 1; break
        else:
            if k > k_max: k_max = k
            i += 1
            if i == l:
                c += 1; l += k_max
                if l + 1 > n: break
                else: i = 0; k = 1; k_max = 1
            else: k = 1
    return c

def compute_S_v1_3(epoch, fs, target_lzc, tolerance,
                    noise_floor=0.1, k=100):
    """Computes S = k * C_n * (ER_n * SI)"""
    n_channels, n_samples = epoch.shape

    # --- A. Coherence (C_n) ---
    filtered = bandpass_filter(epoch, 8, 12, fs)
    analytic = hilbert(filtered, axis=1)
    phase = np.angle(analytic)
    plv = np.mean(np.abs(np.mean(np.exp(1j * phase), axis=0)))
    C_n = np.clip((plv - noise_floor) / (1.0 - noise_floor), 0, 1)

    # --- B. Effective Rank (ER_n) ---
    cov = np.cov(epoch)
    evals = np.linalg.svd(cov, compute_uv=False)
    p = evals / np.sum(evals)
    p = p[p > 1e-10]
    entropy = -np.sum(p * np.log(p))
    ER_n = (np.exp(entropy) - 1) / (n_channels - 1)
    ER_n = np.clip(ER_n, 0, 1)

    # --- C. Structural Integrity (SI) ---
    medians = np.median(epoch, axis=1, keepdims=True)
    binary = (epoch > medians).astype(int)
    lzc_sum = sum(lempel_ziv_complexity(binary[ch,:])
                  for ch in range(n_channels))
    norm_factor = n_channels * (n_samples / np.log2(n_samples))
    lzc_norm = lzc_sum / norm_factor
    SI = np.exp(-(lzc_norm - target_lzc)**2
                / (2 * tolerance**2))
    SI = np.clip(SI, 0, 1)

    # --- D. Final S ---
    Psi = ER_n * SI
    S = k * C_n * Psi
    return SCopy

Digital Claustrum — Toy Model

A PI controller stabilizes 5 coupled Lorenz oscillators at their edge-of-chaos regime. If LZC drops (Seizure risk), it injects noise. If LZC spikes (Chaos risk), it increases coupling.

class DigitalClaustrum:
    """
    Monitors LZC of the system.
    If LZC drops (Seizure) → inject noise.
    If LZC spikes (Chaos)  → increase coupling.
    """
    def __init__(self, kp=3.0, ki=0.5, tolerance=0.03):
        self.kp = kp
        self.ki = ki
        self.tolerance = tolerance
        self.target_complexity = None
        self.error_integral = 0.0
        self.coupling_gain = 2.0
        self.noise_injection = 0.5

    def update(self, state):
        current = self._complexity_proxy(state)
        error = current - self.target_complexity
        self.error_integral += error
        # PI control
        self.coupling_gain -= self.kp * error \
                            + self.ki * self.error_integral
        self.coupling_gain = float(
            np.clip(self.coupling_gain, 0.0, 10.0))

        if error < -self.tolerance:
            self.noise_injection = min(
                5.0, self.noise_injection + 0.2)
            action = "DESYNC (Seizure risk)"
        elif error > self.tolerance:
            self.noise_injection = max(
                0.0, self.noise_injection - 0.2)
            action = "SYNC (Noise risk)"
        else:
            action = "Hold"
        return self.coupling_gain, \
               self.noise_injection, current, actionCopy

Comparison: GNW vs IIT vs CCH

CCH acts as a complement, not a replacement. GNW describes what happens (broadcasting). CCH describes how the brain stabilizes the physical regime required for that broadcasting to occur. IIT provides deep information-theoretic ontology; CCH offers a tractable, signal-processing-level proxy that can be deployed on real data and in engineered systems.

Human–AI Collaboration Method

Multi-LLM Dialectic

CCH was developed through parallel multi-model collaboration, not a single AI session. The human author opened separate sessions with different frontier LLMs, gave them the same prompts, and performed a "Feature Merge" of the strongest contributions from each:

Credit Allocation

Human Author (~60%): Conceptual core, strategic direction, lived experience, final decisions.

AI Collaborators (~40%): Code implementation, mathematical formalization, adversarial testing.

Process Lessons

Ensemble Synthesis: Multiple top-tier models given the same prompt produce diverse outputs that can be merged. Gemini gravitates toward mathematical rigor; GPT toward conceptual clarity. The merge outperforms either alone.

Executable Science: LLMs now function as "Research Software Engineers" — converting high-level theory into testable Python pipelines. The human defines what to test; the AI writes the code; review verifies alignment.

Roadmap & Limitations

Current Empirical Caveats

While code is provided, the large-scale analysis results (boxplots from 50+ subjects) are yet to be generated. The Digital Claustrum code is a toy demonstration, not a production architecture.

From Mechanism to Ontology

If future empirical work supports CCH, the metaphysical hypothesis (CFH) becomes a viable candidate interpretation. If CCH is refuted, CFH loses its physical anchor. This layering is deliberate: rejecting the metaphysics does not damage the science.

Glossary of Key Terms

S (Intensity Marker)

The central scalar marker: S = k · C_n · Ψ(I). Measures how strongly a system's dynamics instantiate the conscious regime.

C_n (Normalized Coherence)

A scalar in [0,1] summarizing large-scale phase-locking (integration) across selected brain channels and frequency bands.

Ψ(I) (Informational Complexity)

A composite of Spatial Diversity (ER_n) and Temporal Structure (SI).

ER_n (Normalized Effective Rank)

Normalized dimensionality of the covariance matrix. ER_n ≈ 0 = high redundancy. ER_n ≈ 1 = high spatial diversity.

SI (Structural Integrity)

A scalar in [0,1] quantifying proximity to the subject-specific critical point μ_critical.

CCC (Coherence-Complexity Controller)

A subsystem that acts as a Resonant Controller, monitoring S and exerting active feedback. In mammalian brains: the claustrum.

Digital Claustrum

The architectural counterpart of the CCC in artificial systems. Monitors S_self and modulates coupling/gain/noise to maintain edge-of-chaos regime.

Smart Zombie

An adversarial signal constructed to look rich and coordinated while lacking genuine high-S structure. Defeated by Conservation of Complexity.

Edge of Chaos

The dynamical regime between order and chaos where small perturbations can propagate without collapse or noise. High S = poised near this edge.