(urth) Grand Unified Theory

[António Pedro Marques]

Dan'l Danehy-Oakes wrote (25-08-2010 21:48):

> As regards pronunciation, though, for all I know the ancient Romans may
> (though VERY improbably) have pronounced "crupta" as we pronounce
> "blanko."
>
> Given how much variation takes place in just the English of England, and
> how much it shifts in a single lifetime, I find it hard to conceive that
> we know how anything was pronounced two millenia before the invention of
> recording technology. (...) Can we know how, say, koine was pronounced,
> or do we merely build edifices of more-or-less probable speculations?

You're quite right to be skeptic in these matters, and it is true that what 
people who deal with history today do is to build such edifices. But there 
is a reason why linguists are confident, and that is the coherence of 
linguistic processes. Language change, as observed scientifically during the 
last couple of centuries, isn't random. It follows patterns ammenable to 
rational analysis and actually exhibits a huge degree of regularity. Now if 
one looks to what is attested in the last millennia, exactly the same 
processes appear to be taking place. It stands to reason that a knowledge of 
the modern forms and corroborating evidence from historical documents 
(either, as you asked for, the rare but existing descriptions of sounds by 
contemporary writers, or the way other languages borrow words, or even 
sometimes the way languages are transcribed in different alphabets than 
their own) is a solid basis to derive guesses that are so convincing as to 
become the default hypothesis. It's very, very far from looking at the 
letters they used and ascribing to them a pronunciation similar to what we 
use today for those letters or ones similar to them. In fact modern 
pronunciations, when resulting from a historic development, are seldom equal 
to the hypothesized original, though they can shed light on it.

Sound-based writing systems such as greek's present an advantage: they are a 
reflection (seldom perfect, though usually one can see where the defects 
are) of the language's phonemical structure, unconsciously derived by native 
speakers. Having their phonemical structure available is a bliss in terms of 
understanding how it may have evolved and even what kind of regional 
variation one can reasonably expect - because we can observe how modern 
languages behave in that regard, what types of sounds are likely to be 
present in a given system (e.g. if there are only N vowels, the most 
probable hypothesis is that those don't include some given ones), and what 
type of different pronunciations may be expected to appear without a break 
in the spelling tradition. Of course, all (or most) conjecture, but there 
comes a point when conjecture is so much in agreement with everything else 
that it takes very good reasons to doubt it.

English is perhaps a misleading example, for a variety of reasons. It looks 
alphabetical, but in fact it's full of homographs that are not homophones 
and the reverse. This is a historical artifact, and a rare one at that. 
Where old languages may have been subject to the same phenomenon, it was 
certainly to a much smaller extent, if any. Spelling usually *is* a guide to 
pronunciation, the more so in languages which didn't carry a burden of 
millennia of established orthographic subsystems from a handful of other 
languages (as english does).
Also, english is almost all diphthongs, which though legitimate isn't all 
that common, language-wise. You hardly have a 'pure' vowel in english, and 
when you multiply that across dialects you end up with a very unusual 
situation indeed.