Re: PIE Word Formation (1)

From: Rob
Message: 44091
Date: 2006-04-03

Greetings, everyone.

After lurking for a while, I've decided to come out of my "slumber"
after reading the messages in this new thread. This subject is one
that I find personally interesting. Although I do not agree with all
of Mr. Gasiorowski's (et al.'s) conclusions, they are nevertheless
very well-presented here, and I thank Piotr for taking the time to
make this presentation. What follows below are my criticisms of the
ideas he presents, along with my own tentative hypotheses.

--- In cybalist@yahoogroups.com, Piotr Gasiorowski <gpiotr@...> wrote:
>
> Here's the promised new topic:
>
> ------------------------------
>
> Proto-Indo-European Word Formation (1)
>
> 1. Nomina: Nouns (Substantives) and Adjectives
>
> Preliminaries
>
> PIE nouns and adjectives (including declinable numerals) fall into
> the same classes and are declined in the same way. The general
> structure of a typical PIE nominal STEM (i.e. a noun or adjective
> stripped of inflectional endings), excluding compounds and
> morphologically opaque words, is as follows:
>
> St- = Rt-S1-S2...-SN-; St- (stem), Rt- (root), -S- (derivational
> suffix)
>
> That is, such a stem consists of a further unanalysable lexical
> morpheme (ROOT) followed by a string of SUFFIXES. In the process of
> word-derivation suffixes are attached one at a time, always at the
> end of the string, so that the structure is in fact a hierarchy of
> lexical domains:
>
> [Rt-]
> [[Rt-]-S1-]
> [[[Rt-]-S1-]-S2-]
> ...

No disagreement here. :)

> Nouns and adjectives can be derived not only from other nomina but
> also from verbs and undeclinables. In the minimal case, there are no
> suffixes at all, and the stem seems to consist of a bare root: St- =
> Rt- (e.g. *gWHen- 'slayer' from the verb root *gWHen- 'slay'). Such
> simple structures are known as ROOT NOUNS. Actually, since PIE
> allows "zero derivation" (as in Eng.: to ride --> a ride; a hand -->
> to hand), i.e. a change of word-class without adding a visible
> suffix, root nouns can be viewed as stems derived from roots by
> means of adding a morphological zero (NULL SUFFIX). Null suffixation
> is also possible with more complex structures, e.g. turning
> adjectives into substantives. It may cause an accentual shift (again
> as in Eng.: to convért --> a cónvert)

While I also agree here, I think it's important to note a couple of
additional things. First off, the accentual shift in English is
always one-way, i.e. the verb receives progressive stress (_convért_,
_recórd_) and the noun receives recessive stress (_cónvert_,
_récord_). Second, languages that allow (or that could allow) "zero
derivation" almost always (in my experience) provide for some other
form(s) of "context cues". In the case of English, the context cues
are subject and object pronouns for finite verbs, the preposition "to"
for infinitives, and the articles "a(n)" and "the" for nouns. IE had
a somewhat different set of context cues: pronominal subject endings
and object pronouns for (finite) verbs vs. case endings for nouns.
However, some of the pronominal endings were phonologically identical
to some case endings, namely 2sg *-s vs. nom. sg. *-s and 1sg *-m vs.
nom. sg. *-m. The fact that we see "zero derivation" in the face of
such syncretism suggests that, at an earlier point in time, these
endings were not identical. Also, the syncretism may have contributed
to the extensive attested use of non-zero derivation later on.

> Most suffixes contain a single consonant (or one of a very limited
> set of suffixes), which may be preceded by *e (e.g. *-er-, *-en-,
> *-es-, *-ent-), but in the prehistory of PIE frequently co-occurring
> combinations of primitive suffixes must have been fused into
> phonologically more complex strings like *-went-, *-men-, *-ter-,
> etc., whose internal structure has been obscured, so that they act
> as unitary morphemes. Some of such complex suffixes may originally
> have been second elements of compounds, reduced beyond recognition
> (cf. English -ly < OE -li:c <-- *li:ka- 'shape, form, body').

Ultimately, all inflections have their origins in such circumstances.
The question is how long ago did those circumstances occur, and how
recoverable they are. Generally, the simpler the inflection, the less
recoverable its origin is.

> A stem is called THEMATIC if it ends in *-e/o-. The term is
> confusing, since "theme" is just elegant variation for "stem", so
> all "stems" should be "thematic" by tautology, but the traditional
> nomenclature reflects a mental shortcut: as the _vowel_ in question
> stands at the end of a stem, it is called the THEMATIC VOWEL, and
> the adjective is mechanically applied to the stem itself. In PIE
> nomina the thematic vowel is traditionally written *-o-, since it
> normally acquired this colour across declensional paradigms, but
> originally it must have been a conditioned variant of *-e-; the
> latter quality is visible when exposed in word-final position (as in
> the voc.sg. *wl.kWe vs. nom.sg. *wl.kWos 'wolf', acc.sg. *wl.kWom,
> etc.).

Regrettably, this point is where my disagreement begins. To begin
with, I consider the term "thematic" to be misleading here. From the
evidence I've seen, it does not appear that the "thematic vowel" had
its origin as such. Rather, I trace its origin to the (animate)
genitive ending, *-ós. Furthermore, while *o does indeed appear to be
a conditioned variant of *e (or, more accurately, they are both
conditioned variants of a single earlier vowel, which I mark as *a),
the overwhelming pattern is *e when stressed and *o (or zero) when
unstressed. Thus, any instance of unstressed *e or stressed *o would
seem to need some other explanation. As a result, it does not follow
to me that the vocative ending *-e (and perhaps also the instrumental
ending *-eh) is a conditioned variant of the "thematic vowel"
(otherwise *o).

> Also in collectives (including neuter "plurals") and feminine
> thematic adjectives, when an original thematic vowel is followed by
> *h2, the result is *-ah2 = {eh2}, not {oh2}: *now-a-h2 'new (f.)',
> etc.

In phonemic notation, I consider it better to mark the suffix using *e
instead of *a. Phonetically, however, the vowel quality was certainly
/a/ (that is, a low central vowel). Rather than attributing this
vowel quality to "coloring", I submit that the quality of the original
ablautend vowel was *preserved* next to "h2", which I tentatively
reconstruct as a velar fricative (i.e. /x/).

> In some morphological environments the thematic vowel may be
> "replaced" by *-i/j-, and there are reasons to believe that it
> actually _changed_ into *-i/j- at some point in its prehistory.

To me, it does not follow that the thematic vowel was necessarily
"replaced" by *-i/y-. Instead, I put forth the hypothesis that this
*-i/y- marker was an alternative affix attached to the bare root or
earlier stem.

> The addition of a suffix may result in changing the vocalism of the
> immediate derivational base. In particular, a suffix with a full
> vowel usually "steals" the accent from the stem, causing the latter
> to lose its vowel. Thus from the root *pleh1- 'fill' we get the
> verbal adjective *pl.h1-nó- 'full'. Similar shifts may be caused by
> accented inflectional endings. As a consequence, only one underlying
> vowel per stem is realised in the surface form. This happens
> especially in the most archaic layers of PIE derivatives; in those
> formed at a later date, when ablaut rules were no longer active
> phonetic processes, the base normally retains its vocalism and the
> alternation of strong and weak forms is restricted to the last
> suffix of the base (or doesn't occur at all).

I think it's important to note here that some inferences can be made
as to *why* a full-vowel suffix "steals" the accent. What we see here
suggests that IE, at the stage when said suffixes were added, had an
alternating pattern of stress. Furthermore, the nature of this
pattern must have been such that it caused the behavior that we see.
My own idea thus far is that IE's accent during this stage fell
regularly on (what was then) the penultimate syllable. Subsequent
quantitative vowel reductions obscured this pattern, and later it
ceased to be productive, resulting in the stress being "fixed" where
it already was. Note that, under this hypothesis, the processes of
quantitative and qualitative ablaut are necessarily divorced from each
other.

> In addition to SIMPLEX stems (consisting of one root plus zero, one
> or more suffixes) PIE makes extensive use of COMPOUNDS (two or even
> more stems combined into a single word). In typical compounds, only
> the final element is inflected. Special types of complex stems
> include UNIVERBATIONS (when an original phrase has become
> "fossilised" as a lexical unit, e.g. *[dem-s]-[poti-]
> '[house-GEN.SG.] + [master]'), and REDUPLICATIONS, when a root is
> preceded by a copy of itself. The copy is rarely complete; in most
> cases only the initial consonant is copied, followed by *-e- or
> *-i-, as in *kWe-kWl-o- 'wheel'. By contrast to reduplicated verb
> stems (of which there are several classes), reduplicated nouns and
> adjectives are rare in PIE and mostly belong to "expressive"
> vocabulary. It seems that nominal reduplication ceased to be a
> productive derivational device early in the history of the language.

With regard to reduplications, I think it's at least *possible* that
the patterns we see are reductions from earlier fully reduplicated
forms. Otherwise, I agree with Piotr's observations here.

Thank you again, Piotr, for this well-presented introduction. I hope
that the interesting discussion continues. :)

- Rob