A Tutorial on Rigorous Scientific Inference: Bayesian Model Selection, the Zero-Patch Standard, and the Deductive Primacy of Axioms Illustrated by the Voynich Manuscript

This tutorial presents a first-principles framework for rigorous scientific inference, grounded in a minimal set of explicit, falsifiable methodological principles. These principles enforce transparency of priors and the strict avoidance of post-hoc modification. We argue that the century-long stagnation in Voynich Manuscript (VMS) research is not a failure of scholar effort or data acquisition, but a systemic artifact of model selection. Specifically, the field has been constrained by the Patching Fallacy: the introduction of unconstrained auxiliary parameters to salvage a hypothesis already contradicted by evidence. By adopting a strict Zero-Patch Standard rooted in information theory and Bayesian probability, we demonstrate how to deduce a prior directly from the topological invariants of the data. When applied to the VMS, this principled discipline rejects linguistic and cryptographic models as probabilistically disfavored. Instead, it identifies the text-generating process as a Structured Reference System (e.g., a Relational Database or Inventory). This assignment is not offered as a settled historical claim but as the information-theoretically minimal explanation under the Zero-Patch constraint derived from the entropy, morphology, and serialization constraints of the evidence.