A Round-Trip Encoding Model for the Voynich Manuscript

We present a round-trip encoding model for the Voynich manuscript—a bidirectional system achieving 99.6% coverage across 38,204 words. The model describes the encoding as Cₙ = f(Pₙ, Pₙ₋₁), where ciphertext depends on both the current and previous abstract plaintext symbols. We identify an alphabet of 21 symbols with 18 word-initial classes, demonstrate deterministic decryption, and validate through cross-validation (97.4%). Unlike prior generative approaches, our model produces exact original ciphertext through round-trip transformation: decrypt(encrypt(P)) = P. The abstract plaintext symbols are structural placeholders—we identify the encoding mechanism but not the underlying language. External evidence is needed to complete any decipherment.