Languages can be deciphered by statistical techniques not because they are "logical" or have a "system" (well uhm you're sort of right, actually). In natural la

Languages can be deciphered by statistical techniques not because they are "logical" or have a "system" (well uhm you're sort of right, actually). In natural languages there is a certain distribution of each phoneme and syllable (let's say due to an unknown reason, but this is observed in all languages[1]), which makes this distribution to somewhat correlate with certain letters and symbols if we're lucky. E.g. suppose we know /i/ phoneme is 90% of vowels in a language and we see that the letter <e> is used about 70% of the time in a large corpus of this language. Our best guess is /i/ is transcribed as <e>. If this phoneme to symbol correlation is too noisy (e.g. in the case of Voynich Script or in this case) then we'll have harder time using statistical techniques. [1] There are certain universal phonological "Universals" that applies to 99.99% of languages. https://en.wikipedia.org/wiki/Linguistic_universal https://en.wikipedia.org/wiki/Linguistic_universal But still these don't let us find the entire structure of a language, and we have to use other techniques. Such as comparative study to other languages, finding loan words we know (e.g. finding the word "computer" or "Jesus" in a text of unknown language with unknown alphabet)