Re: [tied] Lemnos stele and Polish semivowels

On Mon, 22 Jan 2001 22:38:05 +0100, "Piotr Gasiorowski"
<gpiotr@...> wrote:

>>If English were as unknown to us as Lemnian (with only a little help
>from obscure Etruscan), I would say that seeing "AT THE AGE OF FORTY
>*AND* LIN..." (which is more like what we have in Lemnian) would
>naturally suggest another numeral. There would of course be a 10%
>chance of being wrong.
>
> I protest again, not as a linguist but as a mathematically literate person. Here we have something slightly more complex than dice-throwing. The stele inscription is a unique specimen of its type; you aren't sure in advance what syntactic structures to expect. Even the way you answer my objection show that you intuitively realise that the likelihood of finding a given category in a given position depends on how much we already know about the language of the text.

Of course.

> Consider the following argument (somewhat simplified for the clarity of exposition). There is a certain likelihood, determined by typological considerations for arbitrary languages -- let's say 50% (though it's possibly higher for related languages) -- that Etruscan and Lemnian have the same decad-unit order. If so, there is also a 50% chance that they use different orders.

There's also the possibility that an arbitrary language uses no
specific (i.e. both) decad-unit orders. Compare the Lemnos stele's
numeral-noun order in <avis s'ialchvis> *and* <s'ialchveis avis>.

>In the former case the dying age (N) of the man MUST be forty (100% likelihood), because none of the words preceding "forty" can be interpreted as a numeral; in the latter case let's accept your reasoning and say there's a 10% chance that it's forty. The total likelihood that N=40 (rather than 40 < N < 50) equals 0.5 x 1 + 0.5 x 0.1 = 0.55 . Your probability of N *not being* 40 is therefore not 90% but a mere 45%! Why? because the fact that we can already identify the numeral 40 and the word <avis> (and know the latter not to be a numeral) gives 40 an unfair a priori advantage over 41, 42 and the rest.

True. Simply said: *if* the decad-unit orders of the Lemnos stele and
of Etruscan are different, there's a 90% probability that what follows
(maras-m) is another numeral. All other things being equal, the total
probability is 90/2=45%. Odds of 90% are too much to expect or to be
comfortable with when it comes to deciphering ancient inscriptions in
poorly known languages. Look for instance at the fact that we cannot
even identify the numeral 40. I think it's 60 (because of Hyttenia).

=======================
Miguel Carrasquer Vidal
mcv@...