From: Piotr Gasiorowski
Message: 15001
Date: 2002-09-03
----- Original Message -----From: richardwordinghamSent: Tuesday, September 03, 2002 2:08 PMSubject: [tied] Re: Question about rules> What has this got to do with reality? The matches may be good, but will they be perfect? Can you get a regular derivation of spoken Modern English 'one' from Old English 'a:n'? No. (There's a dialect in which 'oats' undergoes the same non-standard sound change.)No. But (1) an explanation of this irregularity does exist (the borrowing of a dialectal form), which means that 'one' is not a case of straight-line inheritance; (2) the expected outcome survives in <only> (and in a number of dialects); (3) the word _is_ an exception and regular reflexes of OE a: (as in bone, home, boat, loaf, rope, soap, road, etc.) are overwhelmingly more numerous than any exceptions. We can never hope for a sound change to be perfectly regular, but this particular one is as reasonably regular as anyone could wish. My point is that whatever the wrinkles, Modern English is generally derivable via Proto-Germanic and Old English, not independently from Proto-Indo-European, and that the linguistic evidence for that is overwhelming.> Can you _regularly_ derive the Romanian word for 'six' from Latin? No (
http://groups.yahoo.com/group/cybalist/message/14398 ), nor from PIE. (The mismatch here is due to the sort of pattern making that seems to plague the numerals.) Moreover, there will always be words that cannot be traced back to the ancestral language.Again, it isn't individual words that matter but the overall regularity of the correspondence. It doesn't mean that exceptions can be ignored. Ideally, we should have a plausible explanation of each. This _is_ circular (we establish a pattern and invent special explanations for any deviations from it, thus saving regularity ex post), but it isn't _viciously_ circular, since the explanations cannot be arbitrary and the regularity must be at least satisfactory. If exceptions are too numerous or too troublesome, we know that something is wrong with the model (Grassmann's and Verner's laws were formulated to remedy such a problem).
> The question then becomes the more difficult one of how one recognises that something is seriously wrong. When is the fault in reality and when is it in the model?The answer will usually depend on the specific conditions of a given language. For some extinct languages we know enough about their grammar and vocabulary, their regional diversification, external contacts, etc. to be able to recognise the effects of paradigmatic levelling or borrowing without the slightest difficulty. The less we know, the harder your question becomes, of course.
> I still think the relevant key points are that:
> (i) honest false models have been seriously attempted.
> (ii) they have been identified as such despite the path from A to B
being short.
> The interesting part of an answer would be the details of (ii).
Certainly.Piotr
P.S. I was writing in haste and wrote nonsense:Note that the parent-offspring relation (supported by a battery of correspondences which do not involve any irreversible sound changes) is obvious in the case of Latin and any of the the Romance languages...I wanted to write "... which don't involve any irreversible sound changes operating in the prohibited direction".P.