Tonal and stress accents

Jens Elmegaard Rasmussen jer at cphling.dk
Sat Apr 1 14:59:48 UTC 2000


On Sun, 26 Mar 2000, petegray wrote:

> Jens said:

>> There was stress before PIE, and stress
>> after PIE, why not _in_ PIE?

To this Peter replied:

> I find this argument more persuasive than your earlier ones, Jens, but we
> are still left with the evidence of Greek, which does not appear to have had
> a stress accent until well after the classical period.   Sanskrit accent is
> also described as pitch, but I know the details less well, and cannot argue
> that stress is precluded by the evidence, the way I would wish to for Greek.
> I know that the Sanskrit grammarians use only pitch terms, not stress, and
> that there is no sign of the stress effects on adjacent syllables that we
> would expect to find with a stress accent.

> If these two languages did indeed have no stress accent, then your logic
> seems shakier - it is not necessarily true that the PIE stress accent
> continued unabated from the earliest time.

Thank you, I see your point very well, Peter, but I still find it hard to
believe that stress (of the time when zero-grades were born) changed into
high pitch (as assumed for PIE), and that high pitch later changed back
into stress everywhere it got a chance. Could the reason why we don't see
syllable weakenings caused by the accent in Sanskrit and Greek (etc.) not
simply be that these effects had already happened, so that we are looking
at the results? Latin has a lot of weakenings caused by the ealier initial
stress, but none yet caused by the historical accent on the penult or
antepenult; still, all Romance languages testify to this accent and enough
have strong reductions caused by it to prove that it existed already in
their common prestage. I don't know of any weakenings caused by the fixed
stress of Polish or Czech, nor Finnish for that matter. I wouldn't be
surprised, however, if there _were_ degrees of duration or aperture
("sonority") that would show in fine measurings. Do stress accents have to
be accompanied by _instant_ alternations of phonemic proportions? Can't
changes begin subphonemically and go ignored for long periods? - I agree
that we cannot dismiss high pitch for the PIE accent syllable, but only
then accompanied with some prominence of the kind we call stress, for that
can be reconstructed too.

   Actually, there are a few accentrelated phonological changes in
Greek. Most notably, the sequence CRHC comes out different dependent on
the position of the accent. Born without proper vowels, the sequence is
normally unaccented, and so gives Gk. CRe:C/CRa:C/CRo:C (with H1, H2, H3
respectively), but if the accent is shifted onto the syllabified sonant to
mark a change of part of speech, the result is CaRaC in the case of H2
(less well known for the other laryngeals), cf. pairs like
vb.adj. kma:to's (Att. kme:to's) 'worn out' : subst. ka'matos
'fatigue'; thna:to's (Att. -e:-) 'dead' : sbst. tha'natos 'death'; quite
possibly also vb.adj. gne:to's 'born' : sbst. ge'nesis from
*g^nH1-to-/-ti- with different placing of the accent. Does this difference
point to pitch or stress? I'd vote for stress any day, for the following
reason:

   Phonetically, a syllabified sonant differs only from an asyllabic
sonant in one respect, viz. with regard to duration. In English, the final
syllable of rotten is articulated just like the n of not, only a bit
longer, this giving the impression on experienced listeners of an
additional syllable. Thus, the phonetic protoform of kma:to's must have
been something like *k^m:xto's, which in Greek assimilated the final part
of the segment [m:] to the articulation of the "laryngeal" (i.e., velar or
uvular) low spirant that followed, this giving something in the order of
[kmaxto's] whence, by the usual loss of laryngeals with compensatory
lengthening, the output form kma:to's as we find it to be. Now, if the
accent is shifted to the first syllable we get a substantive *k^m':xtos;
the further development of this form into ka'matos can be described in
various ways. I am on record for postulating two prop- vowels, before and
after the cluster RH, i.e. something *ka'mxatos with coloration of the
extra vowels by the laryngeal, so that these new vowels stay when the
laryngeal is lost before vowel to leave precisely ka'matos. However, I now
see a somewhat different possibility, namely normal development of the
"syllabic m" into /a/ and a change to /ma/ of the final part of the old
sequence mH2. That would come about by itself if, as would be natural, an
accented syllabic sonant is a bit longer than an unaccented one, for then
we get something like *k^m':mxtos, from where we expect a pre-Greek
development like *ka'mxtos, whence, with the general change of H2 to a
between consonants in Greek, the end-product ka'matos. I would still
prefer the first of these alternative proposals, since syllabification of
laryngeals is very strikingly similar all over and so just cannot be
relegated to the individual histories of the separate branches (although
this is precisely what has spread like a bushfire in the consensus among
scholars of Indo-European in the past decade).

   Still, whichever way one chooses to view the change from *k^m':xtos to
ka'matos, /a'ma/ obviously differs from /ma:/ by being more resolutely
energetic; the segment is given a maximum of sonority right at the outset,
as opposed to what it would get "by default" from the unmarked development
of the underlying sequence anyway. That looks to me more like energy, i.e.
stress, rather than tonal height.

   Also, the Greek accent limitation rules smack more of stress than of
tonal height. One can easily understand that if you give the stress accent
all you've got, you don't have breath for very much more word-length after
the outburst, while it would be hard to understand that you could not go
on speaking lower-key syllables after you had spoken a syllable on a high
note - unless, of course, the high note is _very_ high and thus also
demands all your energy. Barring such typological extremes, this looks
very much like stress on the advance.

   I know that the Greek situation is not PIE, nor is that of Iranian.
There are some suggestive indications of PIE stress in two facts, however:
(1) The shortening of -VH to -V in vocatives: Ved. de'vi, Slav. z^eno, Gk.
nu'mpha; (2) the gemination in hypochoristics, *attos, *attikos 'daddy'.
These phenomena are both easily explained on the basis of a massive
concentration of articulatory energy on the initial in words or wordforms
used in calling or even shouting (the fashionable explanation of the
vocatives by "laryngeal loss in absolute final" is just another case of
collective, but irrational parrotlike repetition of some guru's reply
given on the spur of the moment). Since the vocative has initial accent
irrespective of the lexical place of the accent in the word-stem concerned
(thus even Ved. pi'tar, Gk. pa'ter), we here find accent placing and
stress-reflex correlated in a most illustrative way (nore also the
gemination in the vocative-based Lat. Juppiter).

   Perhaps the Gordian knot is to be cut in the following way: There was a
scale of phonetic registers in the language, the extreme cases being
shouting and enclisis which display quite plain reductions, while in
between the extremes the normal way of giving prominence to a syllable was
by a somewhat higher tone - and, I'd say, a limited amount of extra
articulatory energy (stress). In the further development of the languages,
the emphatic pronunciations became the normal ones in many languages, and
so the stress-based systems of Modern Greek and Pashto grew out of earlier
systems in which the stress element was slighter. The stress element
cannot have been nil, for then it would be incomprehensible that the
syllable to receive stress would always be the one that earlier had the
high tone. Note that the element of pitch cannot have been overly dominant
in Vedic or Old Greek either, for - like stress - it is ignored in poetry
which is based on syllable length only. This leaves us with a PIE system
in which the prominent syllable of a given word was normally pronounced on
an only slightly higher note and with only a slight degree of additional
articulatory force, while in emphatic variants the stress part came to
dominate more radically.

How's that for a attempt to integrate the observations we can make?

Jens



More information about the Indo-european mailing list