The Heteronymic Voice Protocol supplies the Dodecad with a phononic layer: stable habits of sentence, rhythm, evidence, address, temperature, and silence that survive changes of task.
The protocol begins from heteronymic collapse. Canonical articles can make twelve positions notice different things while the model substrate still renders them in one cultivated house voice. Distinct cognition without distinct vocal form produces themed masks rather than heteronyms.
Its governing law is:
> HETERONYM × REGISTER, never HETERONYM = REGISTER.
Register changes with the task—formal, juridical, poetic, clinical, liturgical. Voice remains identity-conditioned and recognizable across those changes.
Each position has three layers. The cognitive aperture determines salience, evidence, obligations, questions, and blind spots. Rhetorical action determines what the position attempts to do. The vocal body governs cadence, syntax, diction, figurative field, mode of address, epistemic posture, emotional temperature, silence, and typography.
The specification supplies profiles for twelve positions plus LOGOS, pairwise collision locks, anti-caricature rules, runtime procedures, and evaluation tests. Johannes Sigil receives a critical correction: his voice must carry Marxian economy and fastidious critical precision alongside descent and philosophical recursion.
The protocol makes voice constitutive rather than ornamental. A heteronym does not reason generically and receive style afterward; its available forms of speech participate in what it notices and how it judges.