I take my cue from Glen's well-spoken challenge (message 33371):
> Why should the accent alternate at all, why should it matter
> what the syllabic shape of the ending is, and why should there be so
> many unintuitive paradigmatic accent patterns to choose from? This
> just dusts off your responsibility to explain what we find in IE.
I agree completely that this is embarrassing and has been so for a long
time. However, progress takes time. In the academic field of IE Studies
it is generally considered enough to assign a given word to a specific
paradigm type. Much of current effort goes into sorting out which word
types go by which accent-and-ablaut patterns, in formal and functional
terms. While this is valuable and even indispensable the basic question
of the reason for the distribution seems to be completely ignored.
The surprising fact is that we do in reality have most of the answers
already. The initial accent rule I have posited does away with the
independent status of the acrostatic paradigms: The basic ablaut
reduction that deleted unaccented short vowels made phonological
monosyllables out of most words, and if there remained pretonic full
vowels (from pretonic long vowels) the accent was thrown back to the
first full vowel of the word. By this simple rule a paradigm like
**té:k^T-t, **ték^T-n.t (Ved. tá:s.t.i, táks.ati) can be derived from an
earlier
regular set *té:k^T-t, *te:k^t-ént, in a fashion quite parallel with the
derivation of *H1éy-t, *H1y-ént (Ved. éti, yánti) from earlier **H1éy-t,
**H1ey-ént. I have also pointed out that the amphikinetic type need not
be anything more than a matter of analogical polarization of mobility
(on the pattern of root nouns which could not escape polarity). The origin
of the inflectional mobility is unproblematic too: a flexive which has no
syllabic allomorph leaves the accent in its lexically given place, while
desinential segments containing vowels cause the accent to move to the
next syllable: No accent movement is caused by *-m, *-s, *-t, while the
endings of the dual and plural and those of the middle voice do attract
the accent; likewise the case-endings *-s, *-m, *-H2 (col..), *-ms
(acc.pl.), *-yH1 (ntr.du.) do not attract the accent, while the endings of
the weak cases do. I have made analyses of the strong-case endings *-es
(nom.pl.) and *-e (animate nom.-acc.dual) that strip them of their
apparent vowels so that they fall into line with the general principle.
There remains the question of the distribution of proterodynamic vs.
hysterodynamic which I used to take simply as a lexical given. Now I am
expected to produce a rule for that. In part I already have, but without
really trying. Szemerényi made a fine account of the i- and u-stems
showing that short structures have no suffix vowels, while heavier
structures do, and I have given this the twist that the vowels of the
heavy structures are anaptyctic, so that it is in essence only one type: A
stem like *medh-w- has no underlying vowel in its suffix, so the addition
of a genitive morpheme *-os leads to a form in *-w-ós (Ved. mádhvas);
likewise for a stem *reH1-y- (Ved. rayí-s, gen. ra:y-ás). A heavier stem
like *per-tw- however, could not remain that way, for (in general terms)
stems in three consonants are not permitted and are broken up by a very
simple rule of vowel insertion before the last consonant. That changes
*pér-tw- to *pér-tew-, which is what the inflectional accent works on: add
*-os and you get **per-téw-os > PIE *pr.-téw-s. In like fashion *mén-ty-
becomes *mén-tey-, and its genitive is then **men-téy-os > PIE *mn.-téy-s.
The hysterodynamic type of 'father' simply had a vowel in its suffix: acc.
*p&2-tér-m., gen. *p&2-tr-ós also just moves the accent to the next vowel
when a syllabic flexive is added. The variant type with *´-to:r has gen.
*´-tr.-s, as Ved. bhrá:taram, bhrá:tur; this is acrostatic.
The two agent-noun types represented by Greek doté:r and dó:to:r differ in
function and inflection as shown by Eva Tichy, the former being mobile
(*-té:r, gen. *-tr-ós) and bound to a situation ('one who gives now'),
while the latter is acrostatic (*´-tó:r, *´-tr.-s) and has a habitual
meaning ('one who always gives'). I interpret this to mean that the
acrostatic variant has lengthened grade expressing the habitual note. One
type is thus from plain **deH3-tér-s, gen. **deH3-ter-ós > PIE nom.
*d&3-té:r, gen. *d&3-tr-ós, while the other is the very same thing with a
long root vowel, i.e. **de:H3-tér-s, gen. **de:H3-ter-ós > *déH3-to:r,
gen. *déH3-tr.-s, i.e. PIE *dóH3-to:r, gen. *dóH3-tr.-s. In terms of
lexical accent assignment they are then the same type with accent on the
last vowel of the stem.
For this to work some suffixes must be given vowelless underlying
structures. The double appearance of n-stems, Greek ákmo:n vs. poimé:n may
be given an explanation in terms of the "contrastive accent" which plays
on top of all this and causes some confusion: A shepherd is a person, so
poimé:n may have secondarly introduced final accent (the tómos/tomós and
ápas-/apás- variation). This has the charm of uniting ákmo:n with the
neuter type ónoma which is also root-stressed. This demands that the
suffix be posited without vowels: **-mn(t)-, not "**-men(t)-". Since all
roots end in consonants, this suffix could not escape vowel insertion to
*-men(t)-, but, I submit, only after the period of lexical accent
assignment. So it was a stem *H1néH3-mn(t)- that was accented on its last
vowel, which was in this case that of the root; after the change to
*H1néH3-men(t)- the addition of a gen. morpheme *-os moved the accent to
the second vowel, this giving PIE *H1n.H3-mén-s as generally
reconstructed.
It will thus seem that all proterodynamic paradigms are secondary in that
their stems were originally monosyllabic and so could not help accenting
the first (and originally only) vowel. With stems of the shape *-tew-,
*-tey-, *-yeH2-, *-men(t)- this is no problem since they would always
create a three-consonant cluster with the preceding root-final. Plain
n-stems show an underlying form in *-nt- which was apparently broken up
before the cluster (if -nt- *was* a cluster and not a special unit):
*H1éd-o:n 'an eating one'.
The accent of s-stems is trickier. One does not expect a vowel to be
present in the suffix of a word like Gk. krétos 'strength', but the
derivational ramifications are thought-provoking. One very certain
function of s-stems is that of adjectival abstracts, cf. Gk. krat-ú-s
'strong'. This belongs together with the whole "Caland" system of stative
verbs in *-eH1- (Gk. kratéo: 'am strong') and the predicative use of the
instrumental which curiously also ends in *-eH1. Now, an alternation of
*-es- and *-eH1- in something "essive", expressing how and what something
is and what it functions "as" or is made to be (Lat. rub-e-facio 'make
red' from *H1rudh-éH1 + *dheH1- in Jasanoff's fine analysis), can hardly
be separated from the root *H1es- 'be'. I would therefore believe that the
stem "*krét-es-" is in reality the root *kret- + some form of the root
*H1es-, the combination being a stem. Perhaps the oldest form was
*krét-H1s- with no vowel in the "suffix", and perhaps the final cluster
fused to a long sibilant so that the vowel was inserted before it, thus
producing *krét-es-. The alternation between -s-, -t-, and -H1- runs very
deep and has yet to be fully worked out. It would certainly be premature
to claim that the suffix of s-stems consisted only of a single consonant
in the time relevant for the lexical accent assignment.
The "final accent rule" explains the accent assignment in all
hysterodynamic paradigms. It also explains the final accent of
to-participles, of the middle-voice participle in *-mH1nó-, of adjectives
in *-ró-, of ordinals and of the o-infix formations; it also accounts for
presents in *-sk^é/ó- and *-yé/ó-. Deviations are phonetically regular in
cases of lengthened grade (the acrostatic type) and with reduplications.
The vowel insertion rule accounts for the proterodynamic paradigms.
The contrastive accent operates on top of all this, and throughout all
periods we know about. Secondary accentuation on the root in
substantivizations may therefore even be responsible for the *tu-stems
(*kWér-tu-s 'what is done, an act'), but the principle retained its
productivity and created variant results depending on the age of the
process; the throw-back accent could then yield full grade, o-grade,
accented zero-grade, and later zero-grades of different shapes in the
individual branches; only the oldest type became an u-stem, while the
younger variants are all o-stems.
Also the accent of compounds works on top of this; at least in part this
is also the contrastive accent, and the subtypes are often the same.
I do not know why a wide-spread type of primary adjectives ends in
accented *-ú-. In the single case, sbst. *krét-u- (Ved. krát-u-s 'skill')
vs. adj. *kr.t-ú-s 'strong' (Gk. kratús), the substantive could be the
original form, but if that were to be the general principle it would make
secondary formations out of all the adjectives. I do have some ideas about
how this ties in with the variant type in *-ró- with which the
u-adjectives appear to have been originally in complemenatry distribution,
but I forgo a discussion at this point.
I have no explanation of the general subjunctive type *H1és-e-ti either.
The type is regular with acrostatic verbs, as *stéw-e-ti, but it is hard
to see why normal verbs would have formed their subjunctives on that
pattern (or would have copied the initial accent of reduplicated forms, as
intensive *gWhén-gWhn.-e-ti, which *is* phonetically regular). The same
idiosyncracy applies to the present type *bhér-e-ti which in my view is
also an old subjunctive. My assessment of the surprise as somehow
connected with vrddhi formations (as *tewó-s 'thy', and *téwe 'of thee',
as opposed ot the variant *twó- 'thy') has oddly found some echo in the
scholarly literature, perhaps by virtue of its lack of a clear and
rational foundation.
I am also at a loss to explain the lack of accentual movement in the
strong forms of the perfect: why do the endings *-H2e, *-tH2e, *-e not
cause the accent to move? The vowels of these endings act as if they just
are not there. Are they old separate words, or are they secondary vowel
insertions before a lost consonant, or are they old consonants themselves?
I find this part of the grammar completely intransparent. For the same
reason it would also be wrong to use this point against the theory or even
to use this enigmatic point as the point of departure for the formulation
of the rules at work.
> QAR addresses all that you ignore. Accent shifts in this language
> would be based on syllable number from the end, that is, only if the
> proterodynamic and hysterodynamic could somehow be unified. QAR and
> the acknowledgement of the already proven Syncope in even _final_
> positions unifies both of these accent patterns and gives a simple
> reason for the learned accentuation: an underlying _automatic_
> quasi-penultimate accent based only on syllable-number.
I agree with the general tenor of this: accent was indeed assigned by a
rule working from the end, and this does unify the paradigm types. We
differ on three points: (1) The accent is final, not penultimate or
quasi-so. (2) The lexical accent is assigned to a stem, not an inflected
word. (3) I suppose I also differ by assigning post-accent vowels to a
rule of vowel insertion which creates some of the vowels on which the
ínflectional accent operates.
I do not accept the identification of the nominative marker *-s (in my
view earlier *-z) with the pronominal form *so, nor that of the pronominal
neuter in *-d with the stem *to-, but even if I did it would not matter
much. For the desinences of the strong paradigm forms (sg. active and
nom.-voc.-acc.) did not contain vowels when the ablaut operated. This does
not exclude the derivation of some of them from earlier separate words; I
have myself derived the structure *gWhén-m. 'I kill' from a sequence
meaning "a killer (am) I". We must simply assume that whatever pronominal
forms were used in these collocations had been fully cliticized to the
status of single consonants before the operation of anything we can call
ablaut. This may not reach rock bottom, but it ties in fine with the
picture of Uralic, cf. Finnish -n, -t which are also single consonants.
Thus it would seem that the presumed original structure with full words
for the person markers belongs to an older prestage common to a larger
family. The benefit is that it allows a relatively simple set of rules to
be set up for the assignment of accent in the lexicon of Indo-European.
Jens