agreement conflict? (from Michael Daniel)
aroginskii at ISF.RU
Sun Jan 31 10:45:53 UTC 1999
Thank you for your comprehensive resume. It seems to me yet that the results
of the discussion may be viewed under somewhat different bias.
All data gathered by you show that in almost all considered languages,
Hungarian, Finnish, Turkish, Hebrew, English, and, to a lesser extent,
Russian and Basque (with the only exception of Spanish, see contribution by
Alan R. King), the same constraint holds true:
If a language has several devices for marking adnominal relations,
constructions with two noun groups with the same marking of adnominal and
related to the same head are ungrammatical or at least avoided.
The set of different adnominal markers, or marking devices is:
s' genitives, _ of _ genitives and simple adposition in English
genitive + posessive suffix on the posessum, adjectivizer, dative in
genitive, elative in Finnish
possessive suffix on the possessum, simple adposition in Turkish
adjective possessor, postposed genitive in Russian
genitive in -en, form in -ko, extraposed gentitive in -en in Basque
In Russian and in Basque doubled expression of adnominal relation by the
same device seems grammatical, although clumsy (see previous discussion).
In Spanish such construction are allowed (according to King's contribution).
I believe this is so because there is only one device commonaly available
(de X) for all types of NPs.
I join the opinion of Matthew S. Dryer that in English the two constructions
(of and 's) are different adnominal relations (because they are formally
different), although they may convey the same semantic content (e.g. part to
whole relation). In Hungarian, according to what you say, the difference
between two constructions is not that one marking device is preposing, and
the other is postposing; but that one is marked by genitive, and the other
Now there are two different questions:
1. Why they are avoided?
2. How they are avoided?
I continue to think that the answer to the first question is different for
head-marking and dependent marking languages.
I believe that in dependent marking languages the reason is that such a
construction would be the source of ambiguity:
X-ARM Y-ARM Z
(where ARM stands for adnominal relation marker) could be interpreted either
as chained genitives (as in John's father's flat - [[X-ARM Y]-ARM Z]) or as
two genitives with
the same head ([X-ARM Y-ARM Z]). Strong arguments for this decision may be
considering data provided by Alan R. King and by the data from Russian - in
Basque and in Russian "double genitive construtions" are allowed only in
those cases where the chained reading is excluded.
I think that in head marking languages your initial solution is
quite OK, and we are not supposed to find out why there are not two
agreement slots (I join Martin Haspelmath' opinion). The example with
and subjective agreement slots
on the verb seems not to be a decisive argument, because transitive verbs
have two different agreement slots, while intransitives have only one; there
are no two slots for the same grammatical relation. The same may be said
about the double case - in most occasions one of the two cases marks the
relation to the immediate head, while the second is "agreement marker".
Presumably, constructions with several genitives related to the same nominal
head by the same head marking device should be looked for not in
agglutinative, but rather in group marking languages.
The answer to the second question (i.e. how are these constructions avoided)
I would suggest that again two different situations are possible:
a) If the relation is supposed to express the same semantic content (e.g.
possession) a coordinated possessor noun phrase would be preferred.
b) If the relation is supposed to express different semantic content (e.g.
possession and part-to-whole relation) than two different marking devices
are preferred (like John's part of the estate); mind that all languages
mentioned above have several different adnominal marking devices.
Although the reasons for genitives avoidance in head marking languages and
in dependent marking languages are different, the way in which dependent and
head marking languages do this seems to be identical.
More information about the Lingtyp