Convergent Corroboration and the CCT Taxonomy: A Methodological Account of Evidence Rating and AI Use in an Independent Neuroscience Research Programme

No existing evidential standard was designed to assess a theoretical framework that generates novel predictions across multiple independent clinical conditions from a single mechanistic criterion. Randomised controlled trials assess specific interventions. Systematic reviews assess accumulated evidence on single questions. Peer review of individual papers assesses whether a paper's claims are supported by the evidence it presents. None was built to assess a framework whose evidential standing depends not on any single confirmatory study but on convergence across independent lines from multiple disciplines — none of which were constructed to support it. This paper describes the Convergent Corroboration Taxonomy (CCT), a four-level evidential rating instrument (CCT-1 Speculative through CCT-4 Confirmed) developed within the Return Constraint Model (RCM) research programme to address this gap. The CCT's logical foundations are those of convergent corroboration methodology — the same structure that established Harvey's circulation account, Darwin's theory of natural selection, and plate tectonics — formalised and applied consistently across a multi-domain neuroscience programme. Three structural safeguards against self-serving bias are described: the independence criterion (evidence must come from researchers with no knowledge of the programme), the falsifiability criterion (every claim is rated for what would refute it, not only what supports it), and the construction domain exclusion (the framework is assessed on cases it was not built to explain). The CCT is illustrated through its application to the RCM programme, which has applied the same thalamocortical gate mechanism criterion across twelve neurological conditions generating a taxonomy of gate failure modes. Two independent external confirmations — research groups publishing results consistent with RCM predictions without knowledge of the programme's existence (Wehmeyer et al., 2025, Molecular Psychiatry; Dai et al., 2024, Frontiers in Neuroscience) — are described as CDL Class 5 evidence under the CCT framework. Associated published programme outputs include: the MND gate drive failure account (Conway, 2026b, Zenodo DOI: 10.5281/zenodo.21071366) and the Tourette syndrome account (Conway, 2026c, Zenodo DOI: 10.5281/zenodo.19113847). The paper also provides a full account of the role of Claude (Anthropic) as a research instrument in the programme — distinguishing between tool use (literature searching, evidence register construction, document quality control, iterative review and challenge) and intellectual authorship (theoretical framework, hypotheses, predictions, CCT ratings), all of which are the author's own. The honest limitation of the CCT scale — that it has not been externally validated as a rating instrument — is stated directly. This paper is offered as a methodological resource for independent researchers facing the same evidential challenge, and as a companion methodological account to the RCM programme papers.