# Brash–Levin article evidence audit: independent replication and predictive credit

**Nathan Sweet, Sweet Rationalism · November 8, 2026**  
**Affected articles:** WordPress post 892, “The Chladni Plate Alternative to Platonic Necessity” (published), and post 218725528, “The Chladni Plate Challenge to Michael Levin’s Biological Platonism” (draft).  
**Related record:** `BRASH_REVISION_AUDIT_2026-11-08.md` and its 23-revision source archive. This document is a subsequent, separately dated evidence qualification; it does not alter the historical snapshots in that archive.

## Result

The previously used phrase “validated prospective biological predictions” exceeded the source check if readers could reasonably take *validated* to mean independently replicated. The 2016 **hnf4/hh** rescue is a model-generated prediction reported as experimentally confirmed by the same coauthored project. It is better evidence than mere post hoc curve fitting, provided the prospective order is accurately reported, but the materials located in this search do **not** establish testing by a separate, unaffiliated research group. The same authorship/independence distinction applies to the much-cited Durant et al. (2017) planarian results.

An unrelated researcher, Patrick Seebold, did publish a **2025 OSF preprint** claiming reproduction of the **two-headed phenotype following gap-junction inhibition in Dugesia japonica**. That is partial outside corroboration. It is not shown by the accessible abstract and personal laboratory account to reproduce the specific 2017 cryptic-state probabilities, membrane-voltage findings, or physiological reset conditions. No peer-reviewed independent replication of those specific features was located in this targeted search. The OSF full-text fetch returned an access error in this pass, so its methods, sample sizes, and analytic independence were not independently audited from the full manuscript.

## Claim-specific evidence matrix

| Claim / target | Originating or relevant source | Outside evidence located | Credit at present | What would strengthen it |
|---|---|---|---|---|
| Electrical influences on planarian regenerative polarity | [Marsh & Beams (1952)](https://doi.org/10.1002/jcp.1030390203), predating Levin | Original independent historical research on electric-field interventions | Independent evidence for **the broad phenomenon**, not for Levin's later exact protocols | Modern independent replication with matched electrical fields and quantitative outcomes |
| Head-versus-tail identity can be switched experimentally | [Petersen & Reddien (2008)](https://doi.org/10.1126/science.1149943), RNAi in a separate research group | Independent genetic-mechanism evidence of polarity plasticity | Supports the phenomenon under a **different causal intervention**, not replication of octanol effects | Matched causal models combining genetic and voltage interventions |
| Multiple heads after transient gap-junction blockade, with later recuts | [Oviedo et al. (2010)](https://doi.org/10.1016/j.ydbio.2009.12.012), with Levin as collaborator | A separate paper within Levin's research programme | Repeated/extended within-programme observations, not independent-lab confirmation | Unaffiliated replication of the protocol, washout, and recutting series |
| A two-headed **D. japonica** phenotype after gap-junction inhibition | [Durant et al. (2017)](https://doi.org/10.1016/j.bpj.2017.04.011) and earlier associated work | [Seebold (2025) OSF preprint](https://doi.org/10.31219/osf.io/e27mq_v1), unaffiliated authorship and separate experimental setting; [first-person methods account](https://patrickseebold.substack.com/p/the-two-headed-worms-in-my-basement) | **Partial independent, non-peer-reviewed corroboration** of a broad intervention–phenotype link | Independent peer-reviewed repeats, with protocols, controls, raw outcomes, and drug specificity |
| Cryptic normal-looking planarians later yielding two-headed offspring with approximately 75:25 stochastic outcomes, voltage differences, and physiological reset | [Durant et al. (2017)](https://pmc.ncbi.nlm.nih.gov/articles/PMC5443973/) | No matched unaffiliated reproduction found in this pass; related work from the collaborating programme is not an external replication | A substantive original-team report, not yet independently confirmed at each detailed target | Preregistered third-party replications of the distributions, lineage tracking, voltage imaging, and reset interventions |
| Model-predicted <em>hnf4</em> plus <em>hh</em> RNAi rescue | [Lobo, Morokuma & Levin (2016)](https://doi.org/10.1093/bioinformatics/btw299) | No unaffiliated reproduction of the specific rescue prediction found in this pass | **Reported prospective within-project model test**, with uncertain external reproducibility | Independently specified/locked model, blinded preregistered experiments by an unaffiliated lab, held-out interventions and transfer |
| Independent Platonic source entering through the physical interface | [Levin's own public account](https://thoughtforms.life/symposium-on-the-platonic-space/) and [Ingressing Minds](https://doi.org/10.3390/philosophies11050161) | None of the above experiments is a discriminating test of the extra source | No ontological upgrade from biological result alone | An explicit source–interface map making validated predictions that differ from equally constrained physical/organizational rivals |

## Independence rule

For the purpose of replication, **independence is a property of the evidence-generating process**, not researchers' metaphysical sympathies. The strongest check uses researchers outside the originating collaboration, fresh specimens or samples, independently executed protocols, transparent exclusions, blinded scoring where possible, separately analyzed data, and prospective commitments. Coauthors at different universities remain one collaborating test. A preprint by an unaffiliated investigator can provide outside evidence without yet meeting the standards of a mature, peer-reviewed multi-lab replication.

Published internal repetitions and technical replicates matter. They can identify laboratory consistency and some sources of error. They do not alone estimate portability across personnel, equipment, strains, settings, or unmeasured protocol differences. A failed outside attempt also needs competent methods and adequate power before it should seriously count against the original claim.

## Publication corrections

- In post **892**, replaced broad claims that the work had earned *validated* prospective predictive success with carefully delimited language about the 2016 within-project prediction and the missing unaffiliated confirmation. The 2017 findings are attributed to the originating team, and the 2025 partial preprint and older independent polarity research are distinguished from replication of the full cryptic-state protocol. The November 8 revision note contains a short replication addendum.
- In draft post **218725528**, qualified introductory and concluding experimental assertions; inserted the same claim-specific replication boundaries and sources without turning the essay into a narration of its editing history. The post remains a draft.
- Both articles preserve attribution to original experimental achievements and the central distinction between a result and an additional Platonic causal interpretation. No claim of a demonstrated nonphysical source is inferred from replication or from its absence.

## Update / loss conditions

This status is a **search-limited assessment as of November 8, 2026**, not proof that an independent study does not exist. Raise confidence in a specific empirical result when an identifiable unaffiliated team reproduces the same protocol and measurements with credible controls. Lower it if competently powered, methodologically matched repetitions fail. Raise confidence in a prediction model only with prospective, independently scored held-out outcomes; an accurate mechanistic prediction does not automatically support an extra ontology. Correct any row if a newly located source demonstrates that an allegedly independent dataset actually shares the original investigators, animals, measurements, or unreported analysis choices.