minimal pairs (was: PIE e/o Ablaut)

[Stanley Friesen]

At 10:42 PM 3/29/00 +0000, proto-language wrote:

>Dear Larry and IEists:
>[PR]

>Now I am really confused. I would have thought that /h/ could be established
>by many minimal pairs like [her] / [per] and /ng/ by many minimal pairs like
>[bang] / [ban], along the lines of your dictionary's: "The existence of such
>a pair demonstrates conclusively that the two segments which are different
>must belong to two different phonemes."

But none of those minimal pairs requires that /h/ and /ng/ belong to
different phonemes.  After all, there are allophones for which the rule
really is based on initial versus final position.  For instance, in High
German, the allophone of /d/ at the end of a word is voiceless.

Minimal pairs only establish the *particular* sounds involved as
distinct.  They do not resolve the issue with regards to other sound pairs.

[PR]

>So what other condition is "*sufficient*" to establish a phonemicity?

I think you mean 'necessary', since a minimal pair *is* sufficient (by
itself it establishes the two sounds which distinguish the two words as
distinct phonemes).  It is, however not *necessary*, since /h/ and /ng/ are
treated as distinct phonemes even in the absence of a minimal pair
demonstrating their distinction.

I am not sure there *is* any single necessary condition.

[PR]

>Ah, a diplomatic answer.

>Well, a rule that has been proposed to account for IE /o/ is that it results
>when the (tonal-/stress) accent of an /e/ is shifted to another syllable.

Now, the problems with this are various.

First, the rules for allophones must refer to *synchronic* features.  A
shifted accent is no longer present as a conditioning feature.  (However, a
rule referring to the position of a syllable *relative* to an accent is
allowable, as in the rule on Proto-Germanic reflexes of PIE
obstruents).  So, by its very form, postulating the *removal* of a
conditioning factor, this model requires the *reconstructed* PIE *o to be a
phoneme.

Second, in many cases the "accent shift" is postulated *solely* on the
grounds of the existence of /o/.  This is circular reasoning.  Internal
reconstruction can only take you so far!

Finally, there are a number of cases of reconstructed PIE /o/ for which an
accent shift is dubious, at the very least.  (Many of these are
"grammatical" /o/s, which are only possible if it is already a phoneme).

>Additionally of interest is that no IE verbal root seems to contain /o/.

That goes to show it was a relatively recent acquisition as a phoneme, not
that it was not one.

I do agree, PIE *o must be recently generated, probably from more than one
source.

>So, applying your insight, is IE /o/ a phoneme or not?

In PIE proper, yes, it is.

--------------
May the peace of God be with you.         sarima at ix.netcom.com