The Silence Paradigm

The von Neumann-Wigner interpretation concerns collapse at the moment of measurement. T₂ tracks the decay of off-diagonal coherence in the density matrix, which happens whether or not anyone measures the system. These are not the same process, so a null on T₂ does not rule out an effect in measurement-outcome statistics.

T₂ is therefore used as an operational proxy for any putative consciousness-coupling channel, chosen because it is continuous, high-throughput, and carries a quantitative noise floor from the existing metrology literature. A secondary arm records four quantum random number generators (QRNGs) in the measurement-outcome channel, but it is exploratory, underpowered for the small effects reported in prior consciousness-RNG work, and not treated as a second confirmation. A null on T₂ constrains the hypothesis. It does not eliminate it.

In quantum foundations, not parapsychology. Decoherence experiments routinely characterise T₂ against electromagnetic, thermal, vibrational, and magnetic perturbations. None has included the observer's cognitive state as a controlled independent variable verified by EEG. Whether the result is a signal or a tight upper bound, it is new within quantum foundations.

What the prior literature contributes is method, not evidence. The Princeton PEAR laboratory, the Global Consciousness Project, and Radin's RNG and double-slit studies reported small deviations associated with focused attention. That work has not achieved broad acceptance; the Bösch, Steinkamp, and Boller (2006) meta-analysis found effect sizes correlating inversely with study quality, the classic signature of systematic bias. This protocol builds on none of its positive claims. It is cited only for the lessons that shaped this design: pre-registration before collection, independent blinded analysis, EEG-verified states rather than diffuse intention, a direct coherence measurement rather than RNG statistics, randomised ordering, and explicit criteria for what counts as a null.

Not a claim to have solved the hard problem of consciousness. Not a claim about dark matter, dark energy, or cosmology. Not dependent on any single interpretation of quantum mechanics. Not powered to reach the effect sizes reported in prior consciousness-RNG work, which test a different measurement channel entirely.

It changes the question being asked. A single-condition test asks whether T₂ differs with a meditator in the room, a question dominated by slow instrumental drift. The eight-cell design asks instead whether T₂ takes a specific shape at C6 and C7 and nowhere else. No environmental confound produces a signal that tracks EEG-verified minimal phenomenal experience and cessation but not mental arithmetic, ordinary rest, or focused attention.

It internalises its own controls. Baseline, C1, C2, and the thermal dummy are reference cells in the same randomised block, not separate validation studies. They calibrate the instrument floor, the high-excitation endpoint, the human-presence baseline, and the warm-mass confound within the same schedule.

Each condition is a falsifier for the others. The protocol does not commit to a single site. It tests the predicted gradient and lets the shape of the result discriminate. Steps at both C6 and C7, with C7 higher, is the two-step pattern. An effect concentrated at C7 with C6 sitting near C1 to C5 is a single-point cessation effect. A smooth rise from C3 to C7 that does not survive adjustment for the somatic covariates is a depth-of-practice signature. A flat line everywhere supports environmental decoherence theory and yields a pre-registered upper bound.

It answers the artifact objection with structure rather than assurance. To mimic the predicted result, a shielding failure or a drifting sensor would have to vary in lock-step with EEG-verified cognitive states across randomised trials, leave every synchronously monitored environmental channel flat, and reproduce the specific two-step shape. Real shielding failures do not behave that way. They track time of day, temperature, mains frequency, and building activity, all of which are recorded.

A session runs all seven conditions in randomised order: 10 minutes of empty-chamber calibration before ingress, observer entry and 15 minutes of ingress settling, then each condition preceded by a 5-minute settling buffer, 15 minutes per period for C1 to C6 and 30 minutes for C7, then a brief post-session baseline. Total session length is approximately three hours. One session per observer per 24 hours. C7 is placed last wherever randomisation permits so that its four-step exit protocol does not fall mid-chain; this is a partial restriction on the randomisation and is declared as such. Empty-chamber and thermal-dummy sessions contain no observer, cannot be chained, and run as separate observer-free sessions on the same randomised schedule. The empty chamber runs Sb = 45 times and the thermal dummy runs once per session-per-participant count, so at the registered target of 10 participants and 15 sessions each the total is 150 observer sessions and 60 observer-free sessions, or 210 chamber-days at one session per chamber per day.

Why chained. Chaining does three things. Every condition in a chain shares one chamber state, so slow instrumental drift is common to the chain and differences out of the within-participant contrasts. It holds the participant burden at 15 to 30 laboratory days rather than the hundreds an unchained design would require, which is the difference between a schedule a host can accommodate and one it cannot. And it holds the practitioner's condition of the day constant across the whole axis rather than spreading it over weeks. The cost is that order and carryover effects become live, so position within the chain is recorded and pre-registered as a covariate.

Write the total decoherence rate as Γ_total = Γ_env + Γ_cog, where Γ_env collects every characterised environmental channel and Γ_cog is the putative consciousness-linked contribution. Placing the empty chamber below C6 and C7 in the predicted ordering requires Γ_cog < 0 at the two deepest conditions: adding a conscious observer in a deep state to a shielded chamber makes the sensor more coherent than the same chamber with nobody in it.

That is the strongest claim the framework makes, and it is pre-registered as a directional assumption layered on von Neumann-Wigner rather than derived from it. It is stated plainly because the weaker alternative is easy to mistake for it. The suppression pattern, in which C6 and C7 exceed C1 to C5 but the confidence interval on the sign discriminator excludes the pre-registered protective effect size, is reproduced exactly by a purely somatic account: a warmer, more mobile body at C1 degrades the sensor, and a stiller, cooler body at C6 and C7 returns it toward baseline, with no consciousness-coupling term at all. The framework commits to the harder prediction and pre-registers the easier one as a distinct, weaker outcome. A reversed sign is pre-registered disconfirmation and is not reinterpreted after the fact.

Frozen classifier. EEG verification is the inclusion criterion for the two deepest conditions, so the protocol pre-registers a positive, quantitative classifier with named channels, named bands, explicit amplitude and coherence thresholds, and a defined scoring window for C6 and C7. It is defined by what the state is, never by exclusion, so a deviant trace that fails C5 is not thereby reclassified as C6 or C7. Where the C7 signature varies across practitioners, both branches are specified in advance as a disjunctive criterion.

Frozen estimation pipeline. The T₂ estimation pipeline is specified and locked before collection alongside the classifier. Exploratory pilot recordings may anchor thresholds and fit parameters; once set, both are lodged in the pre-registration. Pilot data are never pooled into the confirmatory dataset, never set the empty-chamber baseline, and pilot participants are excluded from the confirmatory sample where the pool permits.

Positive control. The chamber establishes that the measurement is quiet. It does not establish that it is sensitive, and only the second licenses a null to be read as an upper bound: an experiment that detects nothing because nothing is there and one that could not have detected anything produce identical data. Calibrated dephasing noise is injected on the existing microwave control line at an amplitude pre-registered to shorten T₂ by at least five times the minimum detectable effect, with a physically independent check via a calibrated offset to the Layer 1 thermal liner setpoint. It runs at commissioning and weekly, on the empty chamber, with no condition labels. It is a gate, not a covariate. If T₂ does not respond, Phase 0 does not proceed.

Observer physiological covariates. Respiration, electrocardiogram, surface electromyography, and a body-directed thermal channel are recorded on the same GPS-disciplined time base and pre-registered as nuisance covariates. These change systematically across the axis, which turns the stiller-body objection into a testable control. A consciousness-linked effect is credited only if it survives adjustment for them.

Three primary tests, evaluated together at family-wise alpha = 0.005 one-sided with Bonferroni correction (per-test alpha ≈ 0.00167): the Jonckheere-Terpstra ordered test on C1 to C7, a rank test that is more powerful than an ANOVA when the predicted alternative is an ordering rather than a bare difference (the empty chamber and the thermal dummy are not members of this ordering); Contrast 1, [C6, C7] against [baseline, C1 to C5]; and Contrast 2, C7 against C6. A Bayesian track runs in parallel, where a Bayes factor above 10 counts as strong evidence and above 100 as decisive, with the mirror thresholds of 1/10 and 1/100 counting as evidence for the null. A finding counts as fully pre-registered evidence only where both tracks agree in the same direction; divergence is reported as divergence. All other cross-condition comparisons are exploratory.

Contrast 1 is deliberately hard to pass. Pooling the empty chamber into the comparison group raises that group's mean, because the baseline is itself predicted to sit above C1 to C5. A comparison against C1 to C5 alone would be easier to clear. The pooled form is kept anyway, so that the primary test cannot be accused of choosing its own comparison group once the predicted ordering is known.

Contrast 2 is underpowered in Phase 0 and stays in the family anyway. A null on it is uninformative rather than disconfirming. It is retained rather than demoted because moving a pre-registered directional test out of the family after the design is fixed is a researcher degree of freedom the protocol does not grant itself. Its primary form is computed on a duration-matched 15-minute C7 window anchored to EEG-verified state entry rather than to clock start, because the C7 signature includes an entry transient. The full 30-minute window and the final 15 minutes are pre-registered as secondary windows and reported alongside without selection among them. Falsification of the primary prediction rests on the ordered test and Contrast 1.

Secondary test D, the sign discriminator. [C6, C7] against the empty-chamber baseline alone, one-sided, at alpha 0.005. This is the only pre-registered test that separates the protective prediction from the suppression pattern: Contrast 1 pools the baseline into its comparison group, the ordered test excludes the baseline, and Contrast 2 compares only C7 with C6. It sits outside the primary family, does not enter the Bonferroni correction, and is not powered for: its power is 0.32 at the registered target, since it carries the empty-chamber cell at full weight rather than at one sixth and is therefore the test most exposed to the empty-chamber session count. Its estimate and confidence interval are reported in every case. Its weights are −1 on the empty-chamber baseline and +½ on each of C6 and C7, summing to zero.

Its interpretation is pre-registered as a three-way classification made on the confidence interval, not on the p-value alone, against delta, the pre-registered protective effect size of d = 0.3. Protective: Contrast 1 positive and test D reaches its threshold, with the Bayes factor crossing in the same direction. Suppression: Contrast 1 positive and the interval on test D excludes delta. Unresolved: the interval spans both zero and delta, in which case no inference is drawn about the sign of Γ_cog. Because test D is not powered for in Phase 0, reaching 0.32 at the registered target, unresolved is the expected outcome and is pre-registered as such. A near-zero point estimate whose interval does not exclude delta is recorded as unresolved and never as suppression; inferring the absence of a protective term from an underpowered test is an inference from absence, which this protocol declines to make here for the same reason it declines to make it for Contrast 2. Phase 1 promotes this test into the primary family and powers for it.

Model. Linear mixed-effects, which estimates the condition means while accounting for the fact that repeated sessions from one participant are not independent observations. T₂ is the dependent variable, condition is a fixed effect, and participant is a random effect (random intercept, with a random slope where estimable). The two observer-free cells contain nobody, so they carry no within-participant structure; they enter under a session-date random effect capturing chamber drift, interleaved with observer sessions on the same randomised schedule. Comparisons that cross that boundary are between-cluster and carry larger uncertainty. One consequence is carried explicitly in the power model. The empty chamber holds no observer, so a single shared baseline estimate enters every participant's Contrast 1, and its error does not shrink as participants are added. The per-participant term and the shared baseline term are therefore separated, and the number of empty-chamber sessions is registered as its own parameter, Sb = 45, decoupled from the sessions-per-participant count. Empty-chamber sessions need no participant and no recruitment, so Sb is instrument time rather than enrolment.

No theory supplies a predicted effect size, so the experiment is not powered against a theoretical target. It is powered to place a bound. The target minimum detectable effect is d = 0.3 on Contrast 1, meaning a shift of three tenths of a standard deviation. That figure is a feasibility-driven choice, not a theoretical one.

Participants, not sessions, are the binding constraint. The power analysis treats person-to-person variability in the deep-versus-shallow effect as a parameter τ and computes power across τ in {0.05, 0.10, 0.15, 0.20}, spanning near-uniform practitioners to substantial heterogeneity. At τ = 0.10 with 30 sessions, 8 participants reach 0.82 power on Contrast 1 and 10 reach 0.95; at τ = 0.15, 10 reach 0.80 and 12 reach 0.92. No realistic session count rescues a four- or five-person study.

Stopping rule. The registered target is P = 10 participants, with P = 8 as the minimum acceptable. P = 10 is the smallest count that satisfies the rule at the session floor, reaching 0.83 power at S = 15 and τ = 0.10 where P = 8 reaches 0.62. P = 8 cannot reach the rule at the floor at any number of empty-chamber sessions, its ceiling there being 0.66, and satisfies it only at S = 30, so enrolling at the minimum obliges the higher session count as a requirement rather than a preference. A firewalled pilot (2 to 3 practitioners, 5 to 10 sessions each) estimates τ and the per-session noise; the session count is then chosen so simulated power for Contrast 1 is at or above 0.80 at d = 0.3 across the pre-registered τ range, with a floor of 15 sessions. Sessions are counted on classifier-eligible sessions rather than on attempts, so that the rate at which C7 is attained cannot silently inflate the power figures. No interim analyses. If the required participant count exceeds the recruitable expert pool, the Bayesian track becomes the primary readout, the detectable effect is recalculated at the actual figures, and Phase 0 reports the shallower bound it achieved, with the empirical heterogeneity and noise model it establishes carried into the Phase 1 registration.

What a null constrains. At the design point, a Phase 0 null bounds consciousness-T₂ coupling at roughly d ≥ 0.24 to 0.29 depending on enrolled participants and pilot-estimated τ, reported alongside the test result whether positive or null, and only if the positive control has passed. It does not reach the far smaller effects reported in prior consciousness-RNG work, which test a different measurement channel; a null here does not refute them.

The three Phase 1 hypotheses. H1-1, deepened test. The Phase 0 family retested on the purified sample, with the sign discriminator promoted into the primary family and Contrast 2 adequately powered, at a per-test threshold recomputed for four tests. H1-2, coupling-model scaling. Purification changes the intrinsic coherence time by a measured factor R on the same platform with the same people. The ratio of the Phase 1 to the Phase 0 fractional effect on Contrast 1 is predicted to equal R under a fixed-added-rate coupling and 1 under susceptibility scaling, and the two are separated by whether the confidence interval excludes the other model's prediction. If both phases are null the comparison is made on the achieved bounds instead, as an instrument consistency check. H1-3, noise spectral density. Sweeping the CPMG or XY8 pulse spacing tunes a bandpass filter across the noise spectrum. Any condition-dependent component is predicted to concentrate below roughly 100 Hz, where human physiology operates, and to fall away at higher filter frequencies. A flat sweep, or a component concentrated where no physiological correlate exists, is pre-registered evidence for an instrumental interpretation rather than a consciousness-linked one. This turns the standard objection to dynamical decoupling, that it may suppress the very signal being hunted, into the measurement itself.

The project is in the pre-experimental phase. The theory, the protocol, the engineering blueprint, and the statistical companion are complete, publicly archived with DOIs, and pre-registered on OSF before any data collection. No data have been taken. What is missing is a host institution.

The design carries a graded set of consciousness conditions rather than a single meditative state, because the relationship between meditation style and any observer effect is not empirically established and a single-state test could not distinguish one from instrumental drift. C6 is treated as a condition distinct from both non-directed openness and full cessation, following the Minimal Phenomenal Experience literature (Metzinger, 2020). The sign of the predicted effect is stated explicitly, the protective and suppression patterns are separated by a dedicated pre-registered test, an observer-free thermal-dummy cell isolates the warm-mass confound, and a pre-registered positive control gates any null.

• The Silence Paradigm (Paper 1), 10.5281/zenodo.21850626
• The Silence Experiment (Paper 2), 10.5281/zenodo.21850613
• Shielded chamber design specifications (TSP-ENG-001), 10.5281/zenodo.21817766
• Phase 0 power analysis and study design table, 10.5281/zenodo.21850635
• Pre-registration, OSF 10.17605/OSF.IO/F658T

A host institution. A supervising principal investigator in quantum sensing, quantum foundations, or contemplative neuroscience, with an affiliated institution behind them. A first hosted measurement can be small: a student-run T₂ study on existing NV hardware, requiring no new build and none of the budget below. The full Phase 0 figure applies only to a dedicated chamber. Phase 1 needs no separate facility, since it runs on the same chamber and the same sensor platform with a different diamond sample, so host adoption is the only thing standing between the protocol and both phases.

Phase 0 funding. ~$469,700 (range $420,000 to $490,000) for the foundational test: NV-center, full dual mu-metal architecture, 64-channel EEG, SQUID monitoring, and the pre-registered eight-cell test at family-wise alpha = 0.005. The figure is the computed bill-of-materials total and is the marginal cost to a university host, excluding facility overhead. Potential sources include FQXi, the Templeton Foundation, and Templeton World Charity Foundation.

Peer review. Paper 2 will be submitted as a Registered Report, a format in which the design is reviewed and accepted before any data exist, to a journal that accepts it (Royal Society Open Science, PLOS ONE, Collabra: Psychology). Both papers will also be posted to arXiv (quant-ph or physics.gen-ph). Zenodo and OSF provide archiving and pre-registration; they are not a substitute for peer review, and the papers are labelled accordingly.

I come to this work from an unusual angle. My professional background is in construction and engineering, which is where the chamber design comes from: the same discipline of materials, tolerances, and layered systems that goes into building anything that has to hold its shape under real-world loads. The five-layer 3 m cubic chamber was designed independently and is published in full so it can be reviewed, critiqued, and built.

Alongside that, I have a sustained interest in neuroscience, contemplative practice, and the measurement problem. I am a long-time meditator with an active practice across the techniques the protocol describes: focused attention, open monitoring, non-directed openness, the threshold state, and cessation. That practice is what gave me the vocabulary for the two-step prediction at C6 and C7, and what convinced me the distinction between MPE and full cessation was operationally real enough to pre-register as separate conditions.

I am not an academic physicist. The theoretical framework, the protocol, and the engineering specifications were developed independently. The project exists to be reviewed, critiqued, and, if it survives that, adopted by people with the institutional infrastructure to run it. The goal is straightforward: build the instrument to the highest standard available resources allow, run the experiment with full transparency, and let the data decide.

ORCID 0009-0002-3904-8557 · X @CliftonBaconNY