Editor for this issue: T. Daniel Seely <dseely
emunix.emich.edu>
Sorry to resurrect a topic from August, but I must have missed this earlier--and just happened to find the relevant issue of LINGUIST while looking for job ads. Richard Sproat had asked for examples of languages whose writing systems allow one to predict the pronunciation (or at least the segmentals) from the spelling and then posted a summary, which suggests to me that the topic is worthy of discussion and is more difficult than might appear. For example, Polish is listed as having such a spelling system (except for words of foreign origin). However, in reality, the situation in Polish is much more complex. Thus, the sequence of letters 'rz' can be either /rz/ or /z^/, the sequence of letters 'ni' before a vowel can be either /nj/ or /n~/, for those speakers (unlike me, but very numerous) who have palatalized labial phonemes, these are never distinguished from sequences of labial + yod except when a we have a prefix ending in a labial before a stem beginning with yod (which happens rarely), the 'zi' can be /zi/ (at morpheme boundaries) or /z'i/ (otherwise), for many speakers the voicing of obstruents before certain suffixes is not what the spelling would seem to suggest, etc. Moreover, since there is certainly no algorithm for distinguishing words of foreign origin from others, to merely say that, in yet other cases than those mentioned, a certain pronunciation is predictable except in words of foreign origin is not going to help with automatic text-to-speech work. There are other examples in the summary which were similarly oversimplified, but Polish is the one I know best, so I will stop here. It seems to me it might be useful to try to see if there are any languages where the pronunciation is really completely predictable from the spelling. I am not sure I would care to wager more than a dollar one way or the other. Alexis Manaster RamerMail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue