Part VI · Consequences
# Eight Ways This Is Wrong
A theory of consciousness that cannot die is not a theory. These are ordered from cheapest to most decisive, and the last one would end the programme outright.
> Assumes
> Parts II–V. Each prediction cites the chapter it comes from.
> Delivers
> Eight falsifiable claims, each with a stated measurement and a stated failure condition.
11.1 How to read these
Axiom 2.3 forbids the theory from predicting any new physics, so every prediction below is structural: a claim that some mathematical property of an unmodified physical state corresponds to some phenomenal property. That is a real constraint. It means the theory cannot be rescued by discovering a new force, and it means each claim can be checked with existing instruments.
Each is stated with three parts — the claim, the measurement, and the falsifier. The falsifier is the part that matters; a prediction without one is decoration.
11.2 Spectral predictions
1. Atomicity must beat band power
Claim. From Chapter 6, valence magnitude tracks the atomic mass of the modular spectral measure, not power in any frequency band.
Measurement. From MEG or high-density EEG, remove the aperiodic $1/f$ component (specparam or equivalent), then estimate $\hat{\mathcal{A}}=\sum_k(P_k/\sum P)^2$ on the residual periodic spectrum. Regress momentary valence report on $\hat{\mathcal{A}}$, on band powers, and on global amplitude.
Falsifier. If band power or global amplitude predicts valence better than $\hat{\mathcal{A}}$, symmetry-as-atomicity is dead. This is the cheapest test in the book and could be run on existing datasets.
2. The consonance kernel has a power-law exponent
Claim. From Chapter 7, valence response to paired periodic stimuli follows the kernel $\kappa$, with peak heights decaying as $(pq)^{-\sigma}$ and $\sigma$ expected between 1 and 2.
Measurement. Two-tone and two-flicker pleasantness ratings across a dense sweep of frequency ratios. Fit peak depths against denominator products.
Falsifier. A non-power-law falloff — exponential in $q$, or flat, or with the octave no more consonant than the tritone — refutes equation (7.2).
11.3 Structural predictions
3. Suffering should be non-self-averaging
Claim. From Chapter 8, chronic suffering is replica symmetry breaking, so repeated sampling of the same nominal state should yield a broad overlap distribution with ultrametric structure.
Measurement. Compute pairwise correlation distances between windows of resting-state MEG within a session. Test the ultrametric inequality $d(x,z)\le\max\{d(x,y),d(y,z)\}$ against a null model controlling for ordinary hierarchical clustering.
Falsifier. No ultrametricity excess in depression or chronic pain relative to controls falsifies Axiom 8.1's identification of suffering with RSB.
4. Log-valence variance scales with integration
Claim. From Proposition 9.1, $\mathrm{Var}(\ln|\mathfrak{V}|)\propto M$, the number of bound modes.
Measurement. Within-subject variance of log-transformed intensity reports, against a proxy for integration such as effective dimensionality of the coherent spectrum.
Falsifier. Constant log-variance across widely differing integration levels. Note this is a test of the scaling; log-normality alone is cheap and would not confirm anything.
11.4 Transitions
5. Anaesthesia is a phase transition, not a dimmer
Claim. From Chapter 6, loss of consciousness is a pure-point-to-absolutely-continuous transition in the spectral measure — a localisation–delocalisation transition, with critical behaviour.
Measurement. Track $\hat{\mathcal{A}}$ through a slow, controlled induction. Look for critical scaling and a divergent correlation time near loss of responsiveness.
Falsifier. A smooth crossover with no critical exponent, and $\hat{\mathcal{A}}$ declining proportionally to drug concentration, falsifies the spectral account.
6. Annealing obeys a three-halves law
Claim. From Proposition 8.2, the drive required to shift a rigid state scales as $(1-T/T_c)^{3/2}$.
Measurement. Dose–response across interventions that plausibly implement annealing — psychedelics, intense exercise, meditative absorption — with rigidity calibrated by a pre-intervention measure of $\mathcal{D}$.
Falsifier. A different exponent falsifies the Sherrington–Kirkpatrick reduction, though not necessarily the RSB picture; the mean-field approximation is the first thing to suspect.
11.5 The two hard ones
7. Capacity scales with area, not volume
Claim. From the area law (4.2), the number of simultaneously distinguishable phenomenal distinctions goes as $A/\xi^2$, not $V/\xi^3$.
Measurement. Comparative: capacity should be insensitive to cortical thickness and sensitive to cortical surface area, which vary independently across species, across development, and in specific pathologies (lissencephaly, polymicrogyria).
Falsifier. Capacity tracking volume rather than area. This is the wildest prediction in the book and I flag it as such — but it is not empty, and the dissociation between thickness and surface area is real and measurable.
8. Binding requires a shared field
Claim. From Chapters 4 and 5, two neural populations with no shared coherent field region cannot be phenomenally bound, however tightly they are functionally coupled.
Measurement. A brain-to-brain interface carrying information over a purely digital channel — no shared field region, arbitrary latency and arbitrary bandwidth.
Falsifier. If such an interface ever produces genuinely unified experience across two subjects, this entire model is refuted and the computationalists are right.
> Why prediction 8 is the important one
>
> The other seven test details of the machinery: the right functional, the right exponent, the right order parameter. Prediction 8 tests the architecture. It is the sharpest available discriminator between field theories and functionalist theories of consciousness, because the two make flatly opposite predictions about a case that is becoming technologically approachable.
>
> I would rather have that bet on the table than not, and I would rather it were tested than argued about.
> ### What Chapter 11 established
>
> - Eight predictions, each with a measurement and an explicit failure condition.
> - Predictions 1 and 2 are cheap and could be run on existing data or with simple psychophysics.
> - Predictions 3–6 test the specific functionals and exponents of Chapters 8 and 9.
> - Prediction 7 (area law) is the most exposed; prediction 8 (shared field) is the most decisive.
> ### Exercises
>
> 1. For prediction 1, explain why removing the $1/f$ component is essential rather than cosmetic. What happens to $\hat{\mathcal{A}}$ if it is left in? [2]
> 2. Design a control condition for prediction 5 distinguishing a genuine phase transition from a sigmoid response with a steep slope. [2]
> 3. Prediction 7 assumes $\xi$ is constant across the comparison. Identify a species pair for which that assumption is doubtful, and say how you would check it. [2]
> 4. Prediction 8 requires assessing whether two subjects share a unified experience — a notoriously hard judgement. Propose an operational criterion that does not beg the question. [3]
> 5. Rank all eight predictions by expected information gain per unit cost, and defend the ordering. [2]
For agents
GET /api/library/spectral-panpsychism/ch11.md