dyads and co-compounds
bernhard.waelchli at ISW.UNIBE.CH
Mon May 24 13:32:06 UTC 2004
I fully agree with Nick that dyad constructions are different from dvandva
compounds (or co-compounds, as I call them in my thesis to appear at OUP
"Co-Compounds and Natural Coordination". The Sanskrit term is problematic
because "dvandva" is actually an example for reduplication 'two-two > pair'
rather than for a coordinating compound. Moreover, Sanskrit dvandva
compounds are rather exotic typologically, because in most languages where
they occur co-compounds are not just simply words, they usually have some
properties of phrases as well and are word-like rather than just words,
which is why many Russian sources on Uralic and Turkic languages use "parnye
slova" "pair words" rather than compound).
Dyads and co-compounds have different typological distributions.
Co-compounds are characteristic especially of the languages of continental
Eurasia (becoming more and more frequent as one moves closer toward
continental East and South East Asia), and partly New Guinea and
Meso-America. (If anybody knows of languages with dvandva compounds outside
Eurasia, New Guinea, Meso-America plus Quechua I would be very grateful for
data!) In Australia, where dyads are very common (see Nick's Cologne
Arbeitspapier), co-compounds seem to be quite rare (Australian languages
with co-compounds are Wik-Mungkan and Kuku Yalanji). However, there is
certainly some functional affinity, since co-compounds very often express
pairs of relatives, such as MOTHER-FATHER for PARENTS,
ELDER-SIBLING-YOUNGER.SIBLING for SIBLINGS. Generally, many co-compounds
contain converses as parts (Cruse 1986: 231ff, X is X to Y and Y is Y to X),
this is not restricted to kinship terms (e.g. GUEST-HOST > 'guest and host',
FRONT(PART)-BACK(PART) of a house > 'house with two rooms'). But
co-compounds are absolutely not restricted to converses.
It seems that in some very few cases dyads can develop from co-compounds
(e.g., Kuku Yalanji, some Yi/Loloish languages [Tibeto-Burman] such as Akha
and Lalo; Bradley 2001). As far as I can see there is no positive or
negative typological correlation between the frequency of co-compounds and
To Nick, the characterization of co-compounds as 'X and Y,
characteristically paired' characterizes actually only one semantic type of
co-compounds, there are quite some others as well, for instance 'X and Y >
always/everybody/everywhere', 'X and Y are a set of which X and Y are
prototypical members', 'X or Y or some entity closely associated to X or Y'
etc. Co-compounds are often also verbs (e.g. EAT-DRINK, READ-WRITE), but all
this only strengthens Nick's argument that co-compounds and dyads are
More information about the Lingtyp