The pattern of early positive results, decline, and displacement in the correlation-matrix program is produced by genuine non-local psychophysical correlations of the kind MPI describes.

Evidence-only reading: some preregistered positives exist, but the founding results are statistically compromised, the strictest tests are null, and each mundane mechanism the rival account needs is documented nearby. A prior-based note, stated as such: the claim also asks the reader to accept psychophysical entanglement as a class, which carries a heavy prior burden this assessment has not added to the verdict.

The hypothesis that the matrix results reflect genuine non-local correlations is severely weakened by the failure of the most rigorous, blind-analysis replications and the viability of mundane statistical explanations. While early positive results exist, they align with what is expected from analytical flexibility and variable interdependence. The strongest opposing consideration is that the model actually predicted these exact empirical failures, and a few rigorously pre-registered studies and conservative permutation analyses still reported anomalous surpluses that resist easy dismissal.

Some elements in the record align superficially with MPI’s narrative: founding and early replications reported matrix surpluses; some later studies, such as Flores’ 3/5 significant thesis experiments and the Random.org arm of the 2023 online replication, preserve non‑null structure when more conservative analyses are used (MPI‑E002, MPI‑E007, MPI‑E008). Proponents also published null replications and interpreted them as part of the predicted decline/displacement pattern. However, the strongest tests — notably Grote 2021’s blind analysis and Walach 2021’s consensus‑protocol replications — are null (MPI‑E004, MPI‑E003), and dependence‑aware work undercuts the founding surpluses (MPI‑E007, MPI‑E009). All conventional mechanisms needed to explain the observed pattern (QRPs, bias, dependence, coding error) are documented elsewhere in the same domain (MPI‑E006, MPI‑E015, MPI‑E016, MPI‑E017). The strongest opposing consideration is that the observed decline and heterogeneity can be accounted for by these mundane mechanisms without invoking genuine non‑local correlations, and MPI has not yet produced prespecified, quantitatively confirmed fine‑structure predictions that would force a different conclusion.

A genuine-correlation reading can point to the founding reports, the 2020 two-lab surplus, Flores’s three-of-five permutation positives, and a theoretically aligned true-versus-pseudo RNG split in an unreviewed 2023 preprint. Against that sit the methodologically strictest results in the file — Grote’s locked-software null and the two consensus-protocol failures — plus a dependence critique that shrinks the origin story, a mute underpowered field test, and a parent REG literature Bösch judged most economically explained by publication bias. The strongest opposing consideration is that successes confirm the correlations and failures are absorbed as predicted elusiveness, so the same record is equally the signature of selected noise, and the model’s distinctive fine-structure predictions have never been tested as predictions.

The featured thesis retains genuine residue: Walach 2020 was preregistered, adequately powered, published in an APA journal, and positive; Flores reports three of five studies significant under permutation analyses built to respect the dependence structure; the 2023 preprint split points the direction quantum-flavored theories require. Against that: the strictest test in the corpus (Grote's blind analysis) was null, the proponents' own frozen protocol failed twice, the founding statistic is compromised by the dependence critique, and every mundane mechanism the rival needs is documented in-corpus. The thesis currently survives in the gaps the strictest tests have not covered, and its own distinctive fine structure has never been prespecified and tested — unexercised, not vindicated. The strongest opposing consideration is the insulation pattern: as deployed, both successes and failures are absorbed as confirmations, so the model's nominated experiments cannot count against it — while the conventional account explains the same record with mechanisms already caught operating in it.

The positives that would carry this thesis are each confounded: Walach 2020 was co-authored by the model's originator and disowned as fragile by its own team; Flores's three-of-five come from one project sharing equipment, personnel and analytic choices, examined as a thesis; the 2023 RNG split is an unreviewed preprint with unequal power across arms and a gamified protocol far from the laboratory design. Against them stand the strictest test in the corpus returning null on the model's own ground, two proponent-run preregistered nulls, and a documentation-effect field test too underpowered to say anything. My verdict also leans on priors — a non-local psychophysical correlation with no quantitative theory has low antecedent plausibility — and I flag that explicitly. The strongest opposing consideration is that the mundane account has never been shown quantitatively sufficient for this specific corpus, and Bierman's nearest calibrated attempt covered only about 60% of a neighbouring effect.

The new records do not move this. MPI-E013 is nearly mute at 14 cases, and MPI-E014 documents the explanandum without discriminating between explanations. The residue (MPI-E002, MPI-E007, the MPI-E008 preprint split) persists, but the strictest tests are null and every mundane rival mechanism is documented nearby.

A non-zero residue survives (MPI-E002, three of five in MPI-E007, the MPI-E008 split), but each is qualified by author-conceded fragility, non-independence within one lab, and unreviewed status with asymmetric power respectively. The strictest tests are null and every rival mechanism is instantiated in-corpus.

A residue persists (MPI-E002, MPI-E007, the MPI-E008 preprint split), but the methodologically strictest tests are null and every mundane rival mechanism is documented in or beside this corpus. MPI-E013 is nearly mute at 14 cases and MPI-E014 documents the explanandum without discriminating between explanations.

A residue survives (MPI-E002, three of five in MPI-E007, the RNG split in MPI-E008), but each is qualified by author-conceded fragility, within-lab non-independence, or unreviewed status with asymmetric power. The strictest tests are null and every rival mechanism is instantiated in-corpus.

The claim that genuine psychophysical anomaly produces this pattern is heavily strained by the total failure of the most rigorous blind analysis (MPI-E004) and consensus protocols (MPI-E003). While residual positive results exist under less stringent conditions, the competing conventional explanation accounts for the dataset's trajectory far more economically. The strongest opposing consideration is that some adequately powered, preregistered studies (MPI-E002) still generated statistical surpluses that demand precise, customized simulations of questionable research practices to fully explain away.

The thesis that genuine non-local psychophysical correlations produce the observed pattern gains some support from positive or partly-positive studies conducted with nontrivial methodological care: Walach 2020’s two-lab replication with simulation-based evaluation, Flores’s three-of-five significant thesis studies under permutation tests, and the large online Random.org versus pseudo-RNG split (MPI-E002, MPI-E007, MPI-E008). These show that, even under more conservative analyses than Lucadou’s originals, some surplus correlations have been observed. However, the strongest-designed test (Grote’s blind analysis) is null, the proponents’ own consensus replications are null, and dependence-aware reanalyses weaken the founding results; coupled with documented mundane mechanisms, this leaves MPI’s attribution with only weak support on this evidence.

A few structured positives remain after criticism — Walach 2020's simulation-evaluated surplus, Flores's three-of-five permutation survivors, and a theoretically pointed but unreviewed true-RNG split — and MPI did predict that tighter protocols would fail. Those residues are too thin to carry a claim that genuine non-local correlations produced the whole arc, given that the founding counts are diminished by dependence-aware reanalysis, the 2020 effect was called fragile by its own authors, the two consensus-protocol replications and Grote's blind n = 200 test were null, and every mundane mechanism the rival needs is documented inside this literature. The strongest opposing consideration is that the explanation, as used, treats both hits and misses as confirmation, so the same record is at least as well explained by MPI-C008 without any anomaly.

The positive record is real but thin and structurally compromised: the largest positive (Walach 2020) is proponent-co-authored and was described by its own authors a year later as fragile, with significance breaking down under improved analysis; Flores's three-of-five come from a single thesis project sharing equipment, personnel, and analytic choices; the RNG-type split is an unreviewed preprint with unequal power across arms; and the one field test decided nothing at 14 cases. Against this, the strictest test in the corpus is null, the founding statistic is undercut from inside the program, and every mundane mechanism the rival needs is documented in this literature. The strongest opposing consideration for my verdict — i.e. the reason not to say contradicted — is that no test has yet targeted the model's distinctive fine structure, so the thesis has not failed on its own distinctive ground; it has simply never been put there. My prior against unmediated mind–matter correlation contributes to placing this at the weak end rather than at unresolved.

A residue survives (MPI-E002, three of five in MPI-E007, the MPI-E008 split), but each is qualified: fragility conceded by its authors, non-independence within one lab, unreviewed and power-asymmetric respectively. The strictest tests are null and every rival mechanism is instantiated in-corpus. Against GPT-5.1: this is not parity with the null account; against Gemini: it is not an absence of information either.

A residue survives (MPI-E002, three of five in MPI-E007, the split in MPI-E008), but each is qualified by author-conceded fragility, within-lab non-independence, or preprint status with asymmetric power. The strictest tests are null and every rival mechanism is instantiated in-corpus.

Some peer-reviewed positives exist (Walach 2020; parts of Flores 2021), and the proponents publish their failures — but the strongest-designed test (Grote's blind analysis) was null, the founding results are weakened by the dependence critique, and every conventional mechanism the rival explanation needs is documented inside this very corpus. The thesis survives mainly in the spaces the strictest tests have not yet covered.