The quantitative commitments of the Orch OR literature

Frozen criteriafrozen 2026-08-26 · hash ca78e87d14cc

Fixed before data collection; any later edit fails the site build. A correction would be a new study superseding this one.

search protocol (re-runnable)

Run only after this freeze, by the executing agent's web search plus direct retrieval of primary documents (arXiv, publisher DOI resolution, PubMed Central); deep-research assistants not used. Literal queries: [Orch OR quantitative requirements coherence time]; [Hameroff Penrose 1996 orchestrated reduction microtubules tubulins 25 ms 500 ms]; [orchestrated objective reduction gravitational self-energy E_G tubulin number]; [Tegmark importance quantum decoherence brain processes]; [Hagan Hameroff Tuszynski decoherence biological feasibility recalculation]; [Hameroff Penrose 2014 review Orch OR beat frequencies EEG megahertz]; [Craddock anesthetic terahertz oscillations tubulin clinical potency correlation]; [Hameroff 1998 Orch OR testable predictions]; [orchestrated objective reduction not biologically feasible]; [Orch OR Diosi Penrose spontaneous radiation underground constraint tubulin scenarios]. Per-candidate follow-up: resolve the DOI or arXiv identifier to the primary document; every cell is verified against the primary document before entering the table; searches that return nothing are reported as failed searches (§3.11).

candidates known at freeze time — dispositions pre-committed

Method

One row per (publication, parameter) pair. Each row quotes the paper's stated value with an exact locator (page, section, table, or equation) in the row citation; the stated-as column records whether the number is a requirement, an estimate, an invoked measurement, an assumption, or a prediction — taken from the paper's own wording, never inferred (§3.9). The formulation column ties each number to the 1996 conformational-qubit formulation, the 2014 collective-dipole formulation, or both, so the consistency question the study exists to answer is answerable from the table. Cross-paper comparisons are recorded as rows only where a paper itself makes the comparison; the study's own consistency assessment appears in the findings, in plain language, with no invented precision (§3.13). Parameters ORCH-R008 names that the corpus never fixes — coherence length, entanglement structure, a numerical potency correlation coefficient — are documented as silences in the findings (§3.11). Proponent and critique rows are graded by identical rules. Either outcome publishes: a specification the field can test, or a documented absence of one, which ORCH-R008 notes is itself the most useful finding available at desk cost.

Findings

• The Orch OR corpus does make numerical commitments, and they are arithmetically consistent across two decades: 500 ms with ~10^9 tubulins (1996), 25 ms with 2 x 10^10 (1998, 2014), the same tau x N product throughout because everything derives from tau = hbar/E_G with carbon-nucleus-radius separation — an arithmetic that Derakhshani et al. independently reproduce from the stated inputs. What moved is the physical carrier: the 1996 specification's 0.2 nm conformational displacement was retired in 2014 — with the energetics critique explicitly conceded — leaving only femtometer-scale nuclear separation carried by collective London-force or spin dipoles. The 1996 and 2014 formulations therefore make the same E_G demands but are not the same physical theory, and evidence for conformational switching gathered against the 1996 mechanism no longer bears on the revised one.→ ledger record ORCH-E035
• Tabulating both sides' decoherence numbers in one place: the corpus requires 25 ms for its gamma-synchrony exemplar (or 10^-7 to 10^-4 s per event under the 2014 beat-frequency re-scoping); the critics computed 10^-13 s (Tegmark, on a 24 nm soliton model that both sides agree was not the Orch OR superposition) and 10 fs to 30 ps for anything coupled to electronic motion (Reimers et al.); the proponents' own unconditional recalculation reaches 10^-5 to 10^-4 s — two to three orders of magnitude short of their 25 ms exemplar. The only proponent-side routes to physiological timescales are the explicitly conditional actin-gel screening estimate (10^-2 to 10^-1 s) and a coherence-time inference from unreplicated single-laboratory resonance measurements (10 kHz resonance read as 10^-4 s coherence). No unconditional proponent estimate meets the requirement anywhere in the corpus.→ ledger record ORCH-E036
• Documented silences (the ORCH-R008 question): the corpus nowhere commits to a coherence length, an entanglement structure across tubulins or neurons, a prospective anesthetic potency-correlation coefficient or threshold (Craddock et al.'s R2 = 0.995/0.999 are retrospective fits over eight agents; the non-anesthetic discrimination is stated qualitatively), or a pass/fail threshold for any direct coherence measurement — the acceptance criterion ORCH-R001 would need. The specification the field lacks is not the parameter values (those exist and are internally consistent) but the acceptance criteria that would let any measurement count as pass or fail.
• One report-grain inconsistency surfaced: the 2014 review states that Hagan et al. found microtubule decoherence times of "10^-4 to 10^-3 seconds, or longer", while the 2002 paper's own abstract states 10^-5 to 10^-4 s — a factor-of-ten upgrade in the retelling of the lineage's own result.
• The 2022 underground constraint engages the committed parameter space directly: under the radiation-predicting Diosi-Penrose formulation, the atomic-nuclei separation geometry — exactly the 2.5 fm choice the 2014 revision committed to — is "definitively ruled out", and the partial-separation fallback requires superpositions the authors describe as vastly exceeding anything achieved experimentally. The corpus's quantitative anchor currently survives only in radiation-free OR variants that no experiment has yet tested.
• Coverage and failed searches: all twelve frozen candidates were located and enter the table; no additional specification paper surfaced from the ten frozen queries beyond the pre-declared candidates (the one post-freeze discovery is Tegmark's companion article in Information Sciences, used for kink-model details). Three documents resisted retrieval — the McKemmish 2009 body and Sahu 2013 (publisher paywalls) and the Penrose 1994 book — and their rows say exactly what could and could not be verified.

Limitations

• Verification labels are honest about access: rows rest on publisher abstract pages, open-access and arXiv texts, an archived ScienceDirect page, and author-hosted or institutional mirror PDFs (ai_verified); the Shadows of the Mind row is unverified (book not retrieved). Mirror fidelity is a residual risk for the two 1996/1998 mirrors, partially offset by cross-checking their numbers against independent restatements (Derakhshani et al.; the PMC review).
• The proponent rows are one authorial lineage (Penrose, Hameroff, and collaborators); their mutual consistency is internal consistency, not independent corroboration, and is tagged as such in the rows (AGENTS.md 3.10). The Craddock 2017 rows share coauthors with the specification papers.
• Specification extraction is corpus-bounded by the frozen criteria: numbers appearing only in later solo publications, talks, or popular books outside the frozen candidate set would enter only if surfaced by the frozen queries, and none were. The table is a census of the frozen corpus, not of everything the theory's advocates have ever said.
• The author knew this literature before the freeze; pre-registration cannot prevent pre-hoc criteria tuning (see the freeze PR's panel question). The bound is the known-candidates declaration: every row's publication was pre-declared at freeze, and the single post-freeze discovery is a reprint of a declared candidate's content, not a new source.
• The 1998 mirror is a scanned JSTOR copy with poor OCR in tables; numbers were taken from running text and figure captions only.
• Cells report what papers state, including where the two sides characterize each other's models; this study does not adjudicate whether Tegmark's model was or was not a fair rendering of Orch OR — both characterizations are recorded as stated.