summary of co-possession
Edith A Moravcsik
edith at CSD.UWM.EDU
Thu Feb 18 15:53:15 UTC 1999
/I tried to post this message on LINGTYP a few days ago but nobody
seemed to receive it. When I re-sent it last night, the system
rejected it saying this message had already been posted. It also
told me that if I wanted a second posting, I had to add some new text
to the beginning of the message lest the system recognizes it as
the old message that it (falsely) believes it had already distributed and
consequently rejects it again. Hence this preamble./
---------- Forwarded message ----------
Date: Thu, 18 Feb 1999 09:45:59 -0600 (CST)
From: Edith A Moravcsik <edith at csd.uwm.edu>
To: edith at csd.uwm.edu
Subject: no subject (file transmission)
This is an update on CO-POSSESSION (thanks to Alan King for
the apt term!): i.e. constructions involving one possessum and
more than one non-hierarchically interpreted possessor, as in
English "John's book of prayers", "my part of the house", or "the
girl's share of the estate". Below I will try to summarize all
contributions to the discussion i.e., both those that were sent
before my last summary of January 27 and those that came after,
and also add a few comments. The following colleagues have
participated in the discussion ("p.c." means the messages were sent
to my address rather than to LINGTYP):
Michael Noonan (p.c.)
Kameshwar C. Wali (p.c.)
While a number of interesting general issues have come up in
the discussion - for example, under what conditions can two
constructions be consider to be of the same kind, the power
struggle between syntactic and morphological desiderata, and the
extent to which facts of language structure can be explained at
all - below I will focus on points directly relevant to the
following three questions:
Which are languages that allow for co-possession and
which are ones that do not?
Are there any structural correlates to the presence
and absence of co-possessive constructions in s language?
If there are such correlations, how can they be explained?
WHICH ARE LANGUAGES THAT ALLOW FOR CO-POSSESSION AND
WHICH ARE ONES THAT DO NOT?
The chart below has five columns.
- The first lists the language names.
- The second indicates whether co-possession is grammatical if
the two possessors have the same form.
- The third shows whether co-possession is grammatical
if the possessors have different forms.
- The fourth shows what the form of the non-genitive possessor
- The last column indicates whether the language has
Blanks indicate no information available. K-T abbreviates
Koptjevskaja-Tamm (see references at the end of this message).
LANGUAGE CO-POSSESSION CO-POSSESSION FORM OF THE POSSESSUM
WITH FORMALLY WITH FORMALLY NON-GENITIVE AGREES WITH
IDENTICAL NON-IDENTICAL POSSESSOR POSSESSOR
Basque: awkward; YES make one NO
mostly in possessor
written style appositional
Hebrew: NO NO
English: NO YES NO
Finnish*: NO one possessor
in Elative or
French: YES YES NO
German: YES YES NO
Hungarian: NO NO one possessor YES
Hebrew: NO YES one possessor in YES
(K-T l995, 24)
Italian: YES (K-T l995, 21) NO
Arabic: YES (K-T 1995, 21) NO
Marathi: YES YES
Arabic: YES (K-T 1995, 22) NO
Russian: clumsy one possessor NO
Spanish: YES YES NO
Turkish: NO possessor- YES
Welsh: NO NO
* Co-possession seems to occur in some Finnish constructions,
although, as Hannu Tommola's data showed, not in the translation
of 'John's share of the estate', etc. Kristiina Jokinen (l991, 7)
says that the "specificer genitive" and the "descriptive
genitive" can occur together. Furthermore, Maria Koptjevskaja-
Tamm cites cases where possessors marked for the same genitive
case cooccur (1988, 147; 1990 8, 18).
ARE THERE ANY STRUCTURAL CORRELATES TO THIS THE PRESENCE
AND ABSENCE OF CO-POSSESSIVE CONSTRUCTIONS IN A LANGUAGE?
Five implicational statements have been proposed. In each
statement, the implicatum is the non-existence of co-possessive
constructions or the existence or lack of it of some special
form of such constructions. The implicantia are of course
different and their specification heads each statement.
Names in parentheses indicate the person(s) who either proposed
the generalization or seemed to agree with it. The formulations
are non-verbatim renderings of what has been proposed.
1. IMPLICANS: POSSESSOR-POSSESSUM'S LINEAR ORDER
Co-possession constructions involving multiple
preposed possessors are ruled out. (Alan King)
2. IMPLICANS: POSSESSOR-POSSESSUM ADJACENCY
Constructions that require adjacency of possessor and
possessum, with the possessor strictly on one side of the
possessum, do not, by necessity, allow co-possession
involving possessors which both observe this constraint.
3. IMPLICANS: POSSESSOR-POSSESSUM AGREEMENT
Possessor-possessum agreement rules out co-possession
involving possessors each of which would have to control
agreement on the possessum. (Michael Daniel, Martin
Haspelmath, Alan King, Edith Moravcsik, Hannu Tommola)
Counterexample: Marathi, which, as Kameshwar Wali reported,
has co-possession and also possessor-possessum agreement.
Alan King suggested that Aymara, Yapese, and Chamorro are
some other head-marking (or both head- and dependent-
marking) languages on which this hypothesis could be further
4. IMPLICANS: POSSESSORS OF SAME FORM
If a language has several devices for marking adnominal
relations, co-possession involving possessors of the
same form is ungrammatical or at least avoided. (Michael
Daniel, Alan King)
(Or, a weaker version: If a language allows co-possession
involving possessors of identical form, it also allows co-
possession involving possessors of non-identical forms.)
5. IMPLICANS: AFFIXALLY MARKED POSSESSORS IN CO-POSSESSION
If affixally marked possessors can form co-possessive
constructions, so can adpositionally marked ones.
/For ##6 and 7, see below./
IF THERE ARE SUCH CORRELATIONS, HOW CAN THEY BE EXPLAINED?
Three explanatory principles have been proposed.
I. AMBIGUITY AVOIDANCE
This principle would explain why co-possessive constructions
with identically marked possessors are dispreferred (#4 above).
Co-possessive constructions introduce the possibility of two
kinds of ambiguity:
a/ ambiguity between the interpretation of one of the
two possessors vis-a-vis that of the other
For example: "painting of Lucy of you": who is
the painter or possessor of the picture and who
if the person painted? (Alan King)
b/ ambiguity between the co-possessive and hierarchical
For example: "part of house of John":
- hierarchical (or stacked, or chained) interpretation:
"John" is the possessor of "house"; "house" is the
possessor of "part"
- co-possessive interpretation:
both "John" and "house" are possessors of "part"
Michael Daniel suggested that, at least for dependent-marking
languages, the desideratum to avoid ambiguity of the
second kind is the reason why co-possession is not preferred.
But, as Alan King pointed out, linear order could provide
the necessary differentiation as it does in Basque between
co-possessive and hierarchical meanings:
"house John part": co-possession
"John house part": hierarchical interpretation.
And, as Alan also noted, the ambiguity of the Spanish "la
parte de Juan de la casa" between co-possession and hierarchical
interpretation is simply condoned, which illustrates that we
cannot posit a universal ban on ambiguity in such constructions.
II. GRAMMATICALIZATION: THE ORIGIN OF AGREEMENT MARKERS
This argument would serve to explain why languages with
possessor-possessum agreement do not (often) have co-possessive
constructions (#3 above).
Martin Haspelmath suggested that the reason why languages
with possessor-possessum agreement do not allow for co-
possession is that in such constructions the possessum would
seem to require have two agreement markers; but, since co-
possession would be a rare construction, such agreement would
not get a chance to historically develop.
I still think that in order for us to be able to take this as an
explanatory principle, we would need to answer the question of
why such possessor-possessum-agreement languages do not solve
the morphological issue by either selecting one of the two
possessors for agreement, or dispensing with agreement
altogether. This concern, however, did not seem to be shared by
other participants of the discussion.
III. GRAMMATICALIZATION: THE STRATAL UNIQUENESS LAW
This argument would provide historical explanations for ##1, 3,
and 5 above.
Martin invoked the Stratal Uniqueness Law - i.e., preference for
only one instance of a relational category per clause - and
proposed that it preferentially held for more grammaticalized
categories over less grammaticalized ones. There are two ways in
which this constraint may be instrumental in explaining the
crosslinguistic distribution of co-possessive constructions. On
the one hand, it makes sense of why some subtypes of co-possivve
constructions are preferred over others. First, provided pre-
head positions are in general more grammaticalized in the noun
phrase, it would explain #1: the ban against the cooccurrence of
possessors both preceding the possessum. Second, if possessor-
possessum agreement is a sign of grammaticalization, it would
explain why constructions that show such agreement in general
prohibit co-possession (#3 above). Third, it also predicts #5
according to which adpositional possessors occur in co-
possessive constructions more than affixally marked ones, as
well as that bare possessors are more resistant to occurrence in
co-possessive constructions than affixally marked ones.
The other explanatory use of the Stratal Uniqueness Law is that
the very occurrence of co-possessive constructions of any type
may be predicted by the doubling of other relational terms.
Assuming the scale cited by Martin and adding "Possessor" to the
Subject > DirectObject > IndirectObject > Oblique > Possessor
it would follow that if a language allows for the doubling of
Indirect Objects (or any other term to its left), it would
also allow for doubling of the Possessor. The following may thus
be added to the set of 5 implications above:
6. If in a language sentences may include two Indirect Objects,
noun phrases show co-possession.
The three explanations address 5 of the 6 implications given
above. The only one which has not been addressed is #2 which
says two possessors cannot coccur if the language would require
both of them to be on one side of the possessum and both
immediately adjacent to it. This, however, as Alan King pointed
out, does not need a linguistic explanation since it
simply follows from the notion of adjacency.
In closing, here are a few items for future research on
- For each language where co-possession is not possible, we
would need to check whether each of the two prospective
possessors could by itself express the meaning that it would need
to express in the co-possessive constructions. If not, then
the ban is not on the cooccurrence of the two possessors but
on a particular meaning-form correspondence.
- Pronominal possession has been relatively neglected in
the discussion. Pronominal possessors are often adjectival in
form: is the co-possession involving two adjecitvally expressed
pronominal possessors ever possible (such as "my your picture"
for 'my picture of you')?
- If a language allows for only one genitive in phrases such
as 'John's part of the house', which of the two possessors will
be genitival and which will take on another expression? Although
one might expect a semantic principle to apply here, Alan King
pointed out that in the Basque expression "your photograph of
Lucy", either one or the other of the two possessors can be
adnominal with the other appositional.
- While it is not that easy to find lexical nouns that
one might expect to take two possessors, there is a kind
of common derived noun which one can expect to occur with
two possessors: action nominalizations, such as "the destruction
of the city by the enemy", "the banning of street crime by
officials", etc. As I briefly scanned Maria Koptjevskaja-Tamm's
extensive work on the forms of such constructions across
languages, several relevant points offered themselves.
Masha identifies several subtypes of action nominal
constructions, one of which is the co-possessive type. This
includes two possessors whether of the same genitival form or
utilizing alternative genitives (1988, 145-153; 1990, 7-10)).
She notes that the cooccurrence of two genitives of exactly the
same form is dispreferred (1990, 7) - cf. implication #4 above)
and she also invokes an anti-ambiguity principle (1990, 8).
Furthermore, she found that languages where the possessum
agrees with the posessor do not allow double genitives in
action nominalizations constructions (1988, 197); cf.
implication #3 above. She also points out that no language uses
a construction for action nominalization that is not also used
otherwise (1988, 227) and, in particular, she proposes that
action nominalization constructions are modelled on
constructions involving other two-place nouns (1988, 227): if a
language has double-genitive constructions in action
nominalization phrases, then it also has double genitives
otherwise (1990, 17-18). If this hypothesis is correct, then we
gain an futher implicational statement:
7. If in a language two genitives can occur in action
nominalization constructions, they can also occur in some
other possessive constructions that do not involve
(I.e., if "the destruction of the city of the enemy" then also
"the part of the house of John".)
Jokinen, Kristiina. l991. On the two genitives in Finnish.
EUROTYP Working Papers VII/14. European Science Foundation.
Koptjevskaja-Tamm, Maria. 1988. A typology of action nominal
constructions. Stockholm University.
Koptjevskaja-Tamm, Maria. 1990. Action-nominal constructions
in the languages of Europe. EUROTYP Working Papers VII/7.
European Science Foundation.
Koptjevskaja-Tamm, Maria. 1993. Nominalizations. Routledge.
/Revised version of l988./
Koptjevskaja-Tamm, Maria. 1995. Possessive noun phrases in
Maltese: alienability, iconicity, and grammaticalization.
In EUROTYP Working Papers VII/25 and in Rivista di
linguistica, 8, 1, l996.
More information about the Lingtyp