PIE Word Formation (5): Compounds (Part 1)

COMPOUNDS (Part 1)

PIE made extensive use of composition as a process of word formation.
Some types of compounds were fully transparent, while others had become
partly or completely obscured. There is a transitional zone between
composition and affixation: a frequently occurring compositional element
may lose its independent lexical identity and evolve into a prefix or a
suffix.

A typical PIE compound is a concatenation of (at least) two stems (I
shall refer to them as M1 and M2, i.e. the first and the second member).
As with simplex stems, all inflectional morphemes are attached at the
end, which means that M1 remains uninflected.

***ENDOCENTRIC COMPOUNDS***

One important type of compound is called ENDOCENTRIC. In this type, one
of the members (in PIE, M2) is the HEAD, expressing the basic meaning of
the compound. The other element (in PIE, M1) is the MODIFIER, narrowing
down the meaning of M2.

An endocentric compound is called DETERMINATIVE ("tatpurus.a", in
Sanskrit terminology), if the head is a noun and the modifier functions
like an attributive adjective, a prepositional phrase or an adverbial
expression, e.g. "toothbrush" = "brush for the teeth, dental brush" and
"rainstorm" = "storm accompanied by rain"; the compound is then
equivalent to a noun phrase with M2 as its head noun. Cf. Skt.
gr.há-pati- 'house-master, Gk. pró-domos 'vestibule', Lat. mus-cerda
'mouse-droppings'. The "is a kind of" test, while not infallible, may be
of some help in distinguishing these compounds from other types:

A TOOTHBRUSH is a kind of BRUSH
A BLACKBIRD is a kind of BIRD

Apart from substantival tatpurus.as there are are also similar compounds
in which the head is an adjective or participle, e.g. "thunderstruck" =
"(as if) struck by thunder", "home-made" = "made at home", or
"snow-white" = "as white as snow". All kinds determinative compounds
generally tend to have an accented initial member in PIE, even if M1 is
the negative particle *n.-, as in the common subtype comprising negated
verbal adjectives like Ved. á-kr.ta- 'undone' < *n.'-kWr.-to-; similarly
in sú-s'ruta- 'very famous' < *h1sú-k^lu-to-, devá-hita- 'arranged by
gods' < *deiwó-dH&1to- and other compounds with a *-no-/*-to- adjective
or a *-ti- noun in M2 position. Little as is known about phrasal
prominence in PIE, I suspect the M1 accent of such compound is of
contrastive origin, distinguishing between true compounds and phrases,
cf Engl. móving-van, bláckboard, vs. a moving ván, a black bóard.

A particularly important type of endocentric compound in reconstructible
PIE is the VERBAL COMPOUND WITH A GOVERNED FIRST ELEMENT, comparable to
English "God-fearing" = "(one) that fears God", "woodcutter" = "someone
who cuts wood" or "bus driver" = "someone who drives a bus". Here, M1 is
the object of the verb which underlies M2.

There are three common types of such compounds in PIE. One of them has a
root agent noun in M2 position, as in Sanskrit nr.-hán- 'man-slayer'
(PIE *h2nr.-gWHé:n, gen. *h2nr.-gWHn-ós) or Lat. au-ceps 'bird-catcher'.
M2 is accented in the strong cases (including the nom.sg.) and has the
e-grade of the root. The second type, highly productive though not
necessarily very old, involves an accented thematic verbal adjective (or
agent noun) of the *bHoró- type. Cf. Gk. drutómos 'woodcutter' <
*dru-tomh1ó-s or Ved. madHu-dogHá- 'sweetness-milking' <
*medHu-dHougHó-. Both types are composed of elements which also occur on
their own. The vocalism of M1 may be reduced if M1 is an ablauting stem;
otherwise its compositional form is the same as the free one, though
it's possible that in thematic M1s the stem-final vowel was originally
reduced to *-i- in the most archaic layer of such compounds.

The third type, apparently more primitive than the other two, involves a
"compositional root noun" in the zero grade. If the nil-grade verb root
ends in a syllabic liquid, nasal or laryngeal, the root is extended with
*-t- (e.g. *-bHr.t- 'carrying', *-stut- 'praising', *-kWr.t- 'making',
etc.), but the extension is not found after other consonants, e.g.
PCelt. dru-wid- 'tree-knowing' or Skt. veda-vid- 'knowing the Vedas'. In
Vedic, M2 is accented, but given its zero grade and the fact that the
accent in Greek is on M1, this accentuation is probably analogical
(modelled on the types described above); if so, compounds like these
originally consisted of an accented modifier plus an unaccented head.
The same type of "extended root noun" can be found in compounds where M1
plays an adverbial function or is a negative particle, e.g. Ved.
su-dyut- 'shining beautifully' < *h1su-dju-t- or di:rgHa-s'rut- 'hearing
from afar' < *dl.h1gHo-k^lu-t-. It seems likely that those apparent
"root nouns" are in fact compositionally reduced variants of *-nt-
participles (as proposed by Olsen).

Piotr