I try to keep strictly to my own area, but I'm curious about the relationship between methods of transcription and statistical analysis of Voynichese.
After seeing Claire Bowern's fascinating discussion of H2 in a 2021 video (starts about 29 mins in), I would be very glad if any specialist here in linguistics could give me some idea of how h2 stats might be affected if a person, 'deaf' to finer sounds in a another language, simply transcribed it in forms reflecting how he and his fellow foreigners actually spoke it. (cf 'Arabic' star names in Europe, or English-of-India vocabulary etc.).
What would happen even to stats for the English language if you transcribed a passage as pronounced by a Russian-speaker, or by a German-speaker?
Conversely, what would happen if the only problem for the transcriber was an inability to distinguish some few sounds from others (a number of sounds in India's languages have no parallel in European languages and are mis-heard by Europeans. So too Japanese 'd' and 'l' can sound much the same to many Europeans, who might use therefore use the same letter for both sounds if putting down what they 'heard' but never saw in writing.
[a nice example is seen when the video's subtitles hear 'Voynichese' as a bi-lingual "buena cheese"
]At the extreme, I suppose, would be a tonal language transcribed by someone 'deaf' to tonal differences. This would surely reduce the apparent vocabulary, at least, because with e.g. modern Mandarin the same string would represent words for 'mother', and for 'scold', and 'horse' and “hemp”.
So there's the question: what effect would such failures of communication have on h2 stats?
Has anyone ever done experiments of this kind?
Come to think of it - did anyone ask Claire a question of that kind at the recent zoom conference?
Here's the link to that vid.
(https://www.youtube.com/watch?v=O58h04GHKS8)