Genitives and Determiners*

Carmen Dobrovie-Sorin, CNRS-LLF, Universit Paris 7

Certain languages have two distinct types of genitives:
(1)

synthetic genitives

analytic genitives

English1
Modern Hebrew
Romanian

Saxon genitives
Construct State associates
al-less genitives

of-genitives
el-genitives
al-genitives

The main goal of the paper is to provide an analysis of synthetic genitives of

category Nmax2. In the three languages examined here, such genitives impose
strict constraints on the determiner of the head noun:
(2) In languages with genitive alternations, synthetic genitives of category
Nmax force the determiner of the head noun to be either (i) empty (English,
Hebrew) or (ii) filled with a suffixal definite article (Romanian).
I will propose that the unifying crosslinguistic syntactic characteristic of
synthetic genitives3 (in languages with alternating genitives) is that they occupy a
Spec,Nmax position, i.e., the Specifier of a maximal nominal constituent that
does not project the DP-level of representation (Section 1). In Section 2, I will

* The central idea of my views on genitives was first published in Dobrovie-Sorin

(2000a,b), where a detailed presentation of the Romanian and Hebrew data can be
found. Dobrovie-Sorin (2002) enlarges the empirical basis by taking into account
new Hebrew data brought up by Engelhardt (1999, 2000) and Danon (2001). Here, I
will mainly concentrate on the semantic analysis and its consequences. Thanks go to
Barbara Partee, whose comments on a previous version made me clarify the analysis.
1As used here, the label 'Saxon genitives' refers to prenominal DPs of the form
DP's; the label 'of-genitives' covers sequences of the form of+DP as well as of +
DP's (a friend of John, a house of John's). Because I do not assume that
morphological form is transparent with respect to syntactic structure, the analysis
proposed below for prenominal Saxon genitives does not extend to Saxon genitives
occupying other syntactic positions (e.g., the predicate position or Saxon genitives
governed by of).
2 I leave aside non-maximal genitives such as "modificational" genitives (e.g., an
old man's hat; cf. Quirk et al. (1985), Woisetschlaeger (1983), Munn (1998)).
3In this paper, the label "synthetic genitive" will be used as an abbreviation for
"synthetic genitive of category Nmax".

propose that constituents sitting in (Spec, Nmax) are to be interpreted via the
rule of semantic composition formulated in (3):4
(3) A genitive specifier (i.e., a genitive that sits in (Spec,Nmax)) is interpreted
as the argument of a function from individuals to individuals (type <e,e>),
which yields the individual denoted by the overall possessive Nmax.
According to this rule, nominal projections embedding Specifiers take argumenttype denotation, although no Determiner is present. We can thus explain why
determiners need not be projected in those configurations in which a Specifier is
projected. The fact that determiners are incompatible with Spec,DP (as in
English and Hebrew) follows from their semantic type. Arguably, the possibility
for the definite article to co-occur with a synthetic genitive in Romanian definite
possessives can be explained on the basis of its being a suffix and by assuming
that it does not have the semantic type characteristic of canonical determiners.
1. Synthetic Genitives occupy Spec, Nmax
In Dobrovie-Sorin (2000a,b, 2002) I showed, for each of the three languages
under consideration here, that the current syntactic analyses of synthetic
genitives were problematic on both syntactic and semantic grounds. I argued that
an adequate analysis could be provided by giving up the hypothesis, largely
assumed among GB theorists, according to which maximal nominal projections
are necessarily DPs, i.e., projections of a functional head labelled Det(erminer),
which roughly corresponds to articles and other functional nominal heads.
Below, I will directly introduce my own syntactic analysis.
1.1. English Saxon Genitives
Current analyses of English Saxon genitives are problematic insofar as they
postulate the head noun is governed by idiosyncratic determiners such as 's or
null AGR (see Abney (1986, 1987), in particular). Since no Determiner can be
overtly realized (see (4)), and since the Bare Phrase Structure theory (Chomsky
(1994, 1995)) imposes no X'-theoretical constraint on constituent structure, the
simplified structure given in (4) is the less stipulative possible analysis, based
on the "null" hypothesis that no Determiner is projected. Since the DP-layer is
lacking, the highest projection is a maximal projection of N, hence the label
Nmax. I take 's to be a Case-morpheme, as in Abney (1987):
(4) a students (*the/these/some/two) results
(4)

Nmax

4 For essentially quantificational genitive DPs (e.g., every man's mother), I assume
Quantifier Raising (see Barker (1995)).

Spec, Nmax

N'

DPgen
a student's

results

1.2. Hebrew Construct State Nominals

The label "construct state nominal" (CSN) refers to the phonological reduction
that characterizes the head N5 : bayit > beyt (ha-i) 'house (the man), the man's
house'. Correlatively, the sequence formed by the head N and its "associate"
(corresponding to a genitive DP) has the properties of a phonological word
(Borer (1988)).6 Besides their phonological properties, Hebrew CSNs differ from
Saxon-genitive constructions in the following respects:
(5) a.
The linear order between the head N and the genitive is reversed.
b.
The CS-associate carries no genitive-marking, whereas Saxon
genitives are marked by 's.
c.
The CS-associate is strictly adjacent to the head N.7
Beyond these important differences, Hebrew CSNs resemble Saxon genitives
insofar as they show a clear complementary distribution with overt Det's:
(6) a.
b.

*ha-beyt ha-i
the-house the-man
beyt
ha-i ha-gadol
house the-man the-big
'the man's big house'

Whenever an overt Det is projected, an analytic genitive built with the

preposition Sel must be used:
(7)

ha-bayit el ha-i
the-house el the-man
'the house of the man'

I will leave aside here Hebrew CSNs built with indefinite associates, which raise
particular problems (see Dobrovie-Sorin (2002)).
Following Ritter (1988), it is currently assumed that CSNs involve N-toD raising (see Fassi-Fehri (1989), Siloni (1994, 1997) and Borer (1996) for
Semitic languages, and Rouveret (1994) for Welsh). Relying on this quasi5Adjectives, as well as numerals, may appear as constructs, but we are interested
here only in nominal Construct States.
6Borer (1996) argues in favor of the stronger claim that construct state nominals are
syntactic words.
7In those contexts in which the adjacency constraint is violated, analytic genitives
are used instead of synthetic genitives:
(i)
*beyt ha-gadol ha-i
house the-big the-man
(ii)
ha-bayit ha-gadol
el ha-i
'the house the-big
of the-man'

unanimous hypothesis, Longobardi (1996) takes the rule of N-to-D to constitute

the defining characteristic of CSNs, a type of construction that covers, according
to Longobardi, not only Semitic CSNs, but also Saxon genitives and Romanian
synthetic genitives. Although I am myself convinced that the three constructions
belong to the same abstract type, I will avoid the label "Construct State",
because I do not think that the Semitic construction is closer to the abstract type.
I will instead use the label "genitive specifiers", which is indicative of the
analysis proposed here. Besides the terminology, I will diverge from Longobardi
in giving up N-to-D (but note that although my analysis does not require N-to-D,
it allows for it). For a detailed discussion of the various existing analyses of
CSNs, the reader is referred to Dobrovie-Sorin (2002). My main claim is
compatible with any of the constituent-structures proposed in the literature, but
in order to avoid assuming unnecessary ingredients, I proposed Bare Phrase
structure configurations of the type shown in (6') for (6)b8
(6')

Nmax
Nmax

b.

AdjP

Spec, Nmax

beyt
house

ha-i
the man

ha-gadol
the-big

In allowing for structures of the type in (6'), I go against Kayne (1994) and
follow Giorgi and Longobardi (1991) in assuming that Spec constituents can
appear on the right-hand side of the head.9
The similarity between the configuration in (6') and Saxon genitives is
clear: DP-level is not projected, and the CS-associate occupies the Spec,Nmax
position.10
8 A comparable structure can be assumed for event-nominals, which will however
be left aside because they raise problems that are outside our present concerns.
9The right-merge of Spec might be dependent on the absence of a complement of
N. The idea that in certain languages nouns do not project the N'-level was put
forward by Ghomeshi (1997) in her analysis of the ezafe Persian construction. The
same generalization is strongly suggested for Romanian, by the fact that two
genitives (regardless of whether they are synthetic or analytic) cannot attach to the
same noun.
10This is not to say that genitive-embedding constituents rely on exactly the same
syntactic configuration in English and Hebrew. One striking difference is the
position of adjectives: they are internal to the N-projection in English, but DPadjuncts in Hebrew. Other differences can be observed regarding the rules that select,
from among the arguments of the head N, the argument that is merged in
Spec,Nmax: compare tmunat ha-xamaniyot Sel Van Gogh 'picture the-sunflowers of
Van Gogh' vs Van Gogh's painting of the sunflowers. The generalization seems to be
that in Hebrew non-event nominals, the DP in Spec,Nmax must be selected by the

1.3. Romanian Genitives

Like English and Hebrew, Romanian displays a remarkable alternation between
synthetic genitives (marked with morphological genitive Case, which in
Romanian is formally identical to the Dative) and analytic genitives, made up of
a genitive-marked DP preceded by the inflected genitive particle al/a/ai/ale
(decomposable in an invariable part, a-, followed by the definite determiner: lsgm, -asgf, -iplm, -leplf),11 which agrees with the head N.
The constraint on determiners can also be observed in Romanian, although
under a slightly different form: synthetic genitives require a definite article on
the head N (see (8)); the presence of any other determiner requires an analytic
genitive (see (9)):
(8) N-def GDetgen
a.
b.

casa vecinului/vecinilor
house-the neighbour-theGen/neighbours-theGen
'the neighbour's/the neighbours' house'
casa unei vecine/unor vecine
house-the afem-Gen neighbour/somefem-Gen neighbours
'a neighbour's/some neighbours' house'

(9) Det N al GDetgen

a.
o casa a vecinului/ a unei vecine
afem-Nom house a neighbour-themasc-Gen/a afem-Gen neighbour
'the neighbour's/a neighbours' house'
b.
acest cine al vecinului/ al unei vecine
this dog al neighbour-themasc-Gen/al afem-Gen neighbour
'this dog of the neighbour/of a neighbour'
Note that regarding linear order, Romanian DPs resemble Hebrew DPs: genitives
(both synthetic and analytic), as well as Adjectives (not illustrated here), follow

head N (it can be either the internal argument (Theme) in picture-nouns or the whole
in a part-whole relation, or an inherent possessor). Such differences in linking rules
do not bear on the syntax and semantics of genitive specifiers, which I will argue to
be essentially the same in Hebrew and English.
11 Note that al/a/ai/ale can be analyzed as a definite noun headed by an empty N:
[Ne]-def art (dHulst, Coene & Tasmowski (1997) and Dobrovie-Sorin (2000)), and
correspondingly the analytic genitive can be analyzed as a full DP embedding a
definite head N followed by a synthetic genitive.
headed by a null pro fo category

the head noun. Moreover, although they do not show any phonological
peculiarity, Romanian synthetic genitives show a strict adjacency constraint:12
(10) a.
b.

* casele de piatra vecinului/ unei vecinei

houses-the in stone neighbour-themasc-Gen/ afem-Gen neighbour
casele de piatra ale vecinului/ ale unei vecinei
houses-the in stone ale neighbour-themasc-Gen/ale afem-Gen neighbour
'the houses in stone of the neighbour/of a neighbour'

Dobrovie-Sorin (1987), Grosu (1988, 1994) and Cornilescu (1993), among many
others, assumed a rule of N-to-D, which was suggested by the suffixal status of
the Romanian definite article. Note however that the analysis relying on N-to-D
was related to the GB model, which required N's and D's to be generated in
distinct positions, even if the Determiner was an affix. Within the minimalist
framework, in which words enter the derivation fully inflected, Romanian
definite N's can be directly merged (as inflected words) under the N-node; N-toD can be dispensed with and synthetic genitives target the Spec,Nmax position.
We thus obtain a syntactic representation that is comparable to that of Hebrew
CSNs:
(8')

Nmax
N+Det

Spec, Nmax
DPgen

[N(+D)casa+a]
house-the

vecinului
neighbour-themasc-gen

This does not mean that Romanian projections headed by a definite N never
count as DPs. Consider indeed examples such as (11):
(11)

masa
table-the

Given the standard rules of semantic composition, sequences of the type N+def
art must be analyzed as Det-projections in order to be interpreted as arguments:
the denotation of the Determiner (type <<e,t> <<e,t>t>>) applies to the
denotation of the noun (type <e,t>) and yields a generalized quantifier (type
<<e,t>t>). In order to allow for this type of semantic composition, we may
12 It seems clear that for Romanian at least, the adjacency constraint cannot be due
to the phonological properties of the construction. A unitary crosslinguistic account
for the adjacency constraint can therefore be based on structural properties alone:
linear order, presence or absence of complements of N, as well as properties of
adjectives might be relevant.

assume, following current minimalist analyses, that a null Det is projected, to

which N+def raises for reasons of "feature-checking":
(11)
Det
N+Det

DP
N
tN+Det

In sum Romanian maximal nominal projections headed by nouns carrying the

suffixal definite article, can be either DPs or Nmax projections, depending on the
syntactic environment and on the requirements of semantic composition. The
syntactic structure shown in (8) can be assumed provided that we can make
explicit the rules by which it is semantically composed. Section 2 below is
devoted to this issue.
1.4. Conclusions
To sum up, the morphological distinction between synthetic and analytic
genitives correlates with a difference regarding their syntactic positions:
(12)
In languages with alternating genitives, synthetic genitives sit in
Spec,Nmax (whereas analytic genitives are either complements of N or DPadjuncts (see Section 2.5. below).
2. The Semantic Composition of Genitive Specifiers and the Constraint on
Determiners
Assuming the syntactic analysis proposed above to be by and large correct, let us
now try to understand the constraint imposed on the determiner of the head noun.
What we have to explain can be broken in two parts. The positive generalization
is that a nominal projection embedding Spec,Nmax can have the semantic type
of arguments (type <<e,t>t> or type e) although it is not governed by a
determiner. The answer to this puzzle will be that genitives in Spec,Nmax are
semantically analyzed as arguments of a function from individuals to individuals
(type <e,e>), which returns the individual denoted by the overall possessive
expression. The negative generalization is that Determiners (other than the
definite article) are incompatible with the projection of Spec, DP. This
generalization follows from the semantic type of canonical Determiners.
2.1. The Relational Analysis

Before introducing my own proposal, let me briefly review the current semantic
analysis. Beyond slight variations, Barker (1991, 1995), Partee (1983, 1997) and
Jensen & Vikner (1994) assume that the semantic analysis of Saxon genitives
relies on a relation, which is either provided by the lexical properties of the head
N (see relational nouns such as son, middle, etc.) or else contributed by the
structure itself. Within Barker (1995) and Vikner & Jensen's (2002)
implementation the relation is uniformly contributed by the head noun.13 Objectdenoting nouns (type <e,t>) such as bike are coerced - due to the presence of the
genitive morpheme- into relational nouns, denoting the set of pairs of individuals
(x,y) such that y is a bike that entertains an underspecified relation Rgen with x
(see (13)a). Relational nouns need not be type-shifted, since they provide a
relation as part of their lexical entry (see (13)b, which says that sister denotes
the set of (x,y) pairs that are in the sister-relation):
(13) a.
b.

[[bike]]
[[sister]]

= xy [Rgen (x,y) & bike (y)] (coerced relational noun)

= xy [sister (x,y)]
(lexical relational noun)

Possessive expressions of the type John's bike or John's child are obtained by
applying the denotations in (13)a-b to the individual denoted by John, notated j:
(14) a.
b.

[[John's bike]]
[[John's sister]]

= xy [Rgen (x,y) bike (y)] (j)

= y [ p (j,y) bike (y)]
= xy [sister (x,y)] (j)
= y [sister (j,y)]

(14)a-b denote sets of individuals (type <e,t>): the set of individuals that are
bikes and entertain an underspecified relation Rgen with John, and the set of
individuals that entertain the sister-relation with John, respectively. Because
these expressions are of type <e,t>, they must combine with determiners in order
to yield a generalized-quantifier type of denotation. Since no determiner is
overtly realized, the existence of a null element must be postulated. But no
empirical evidence is adduced in favor of the existence of a null article, which
cannot be found elsewhere in the language. Note furthermore that the relational
analysis incorrectly predicts that synthetic genitives are compatible with any kind
of determiner. In order to avoid this overgeneration, Partee (1983, 1997) and
Barker (1991, 1995) are forced to stipulate a further constraint: the null
determiner governing the head noun can only be a definite article (for Partee), or
a possessive article (for Barker).

13According to Partee (1983, 1997), the "free relation variable" comes in as part of
the meaning of the genitive construction itself, and the head N governing the
genitive preserves its canonical denotation. Partee and Borschev (2003) give an
empirical argument against this analysis and in favor of the type-shifted denotation
of the head N's (see (13)a-b), assumed by Barker (1991,1995) and Jensen and Vikner
(1994).

2.2. Genitive Specifiers as Arguments of a Function from Individuals to

Individuals (type <e,e>)
In Section 1 above possessives embedding synthetic genitives were analyzezd as
in (15'), where no Determiner is projected:
(15) a.
b.
c.
(15')

a.
b.
c.

Mary's
mother
my neighbour's sister
Mary's
house
Nmax (type e)

Spec, Nmax (type e)

N'

Mary's
my neighbour's
Mary's

mother
sister
bike

But how can we obtain argument-type denotation in the absence of a determiner?

Dobrovie-Sorin (2000a,b, 2002) proposed the following rule of semantic
composition :
(16)
If the Specifier of Nmax is filled, the head N denotes a function of
type <e,e>, which applies to the individual denoted by the DP in Spec Nmax and
yields the individual denoted by the overall possessive Nmax.
This analysis is straightforward for those possessives that are headed by nouns
that are lexically specified as functional (mother, capital, middle, etc.):14

(15") a.

[[Mary's mother]] = x y [mother (x,y)] (m)

= y [mother (m,y)]

It should however be stressed that the functional analysis formulated in (16) is

not triggered by the lexical properties of the noun, but rather by the projection of
Spec,Nmax. In case the head noun is not lexically specified as functional, it is
coerced. For relational nouns such as sister, friend, etc., uniqueness is triggered:
14 When governed by lexically functional nouns, genitive DPs are to be analyzed as
arguments of functions of type <e,e> even if they do not occupy the Specifier, but
rather the complement position (see Frege's (1891) analysis of the capital of the
German empire, or Keenan's (1974) analysis of the inside of the bottle, the rest of
the cake, etc.).

10

to one referent of the genitive corresponds a unique referent for the overall
possessive (see (17)a). Object-denoting nouns such as bike are coerced into
denoting functions of the type shown in (17)b:
(17) a.
b.

fsister (x) = x y [sister (x,y)]

fbike (x) = x y [Rgen(x,y) & bike (y)]

The denotations of the overall Nmax constitutents in (15')b-c are calculated by

applying the functions in (17)a-b (denoted by the head nouns) to the individual
denoted by the DP that occupies the Spec,Nmax position. We thus obtain, as the
denotation of Nmax, the unique individual associated to the individual denoted
by my neighbour/ Mary by the functions in (17)a-b:
(15") b.
c.

[[my neighbour's sister]] = x y [sister (x,y)] ([[my neighbour]])

= y [sister ([[my neighbour]],y)]
[[Marys bike]]
= x y [Rgen(x,y) & bike (y)] (m)
= y [Rgen(m,y) & bike (y)]

Genitive specifiers governed by plural head nouns can also be analyzed as

relying on the rule in (16). In this case, the value of the function is a plural
individual (group) notated by a capital letter below:
(18)

[[John's houses]]

= x Y [Rgen(x,Y) & houses (Y)] (j)

= Y [Rgen (j,Y) & houses (Y)]

The overall Nmax denotes the maximal group of houses that is associated to
John by the underspecified relation Rgen.
We still need to make precise the analysis of QPs occurring in the Spec, Nmax
position:
(19)

Each students car will be cheked.

In line with Barker (1995), I assume Quantifier Raising, as shown in (19):

(19')

each x (x is a student) y (car (y) Rgen (x,y)) [y will be checked]

Given that functional application is not subject to any directionality constraint,

the rule of semantic composition stated in (16) applies not only to Saxon
genitives, but also to the Hebrew and Romanian synthetic genitives, which were
shown to sit in Spec, Nmax.
The functional analysis formulated here resembles the relational analysis
proposed in Barker (1991, 1995), Partee (1983, 1997) and Jensen and Vikner
(1994) insofar as a structure-dependent rule of semantic composition is
postulated, but differs from it regarding (i) the syntactic configuration that it
presupposes and (ii) the content of the postulated semantic rule.

11

Like Barker (1995) and Vikner and Jensen (2002), I assume that in the
context of a Saxon genitive, the denotation of object-denoting nouns is shifted,
but under my analysis the head N is not relational (type <e <e,t>>) but rather
functional (type <e,e>). By applying this type of function to the individual
denoted by the genitive specifier we obtain the individual denoted by the overall
possessive Nmax constituent. No determiner is needed for the semantic
composition, which correlates with the syntactic lack of overt determiners in
possessives built with Saxon genitives. Compare the relational analysis, under
which a null Determiner must be postulated.
2.3. (In)definiteness Spread
As observed by Jackendoff (1974), DPs dominating Saxon genitives are
interpreted as definite or indefinite depending on whether the Saxon genitive
itself is definite or indefinite:
(20) a.
b.

There is a dog/*the dog/??John in the garden.

There is a man's dog/*the man's dog/??John's dog in the garden.

(In)definiteness spread can also be observed for Romanian possessives

embedding synthetic genitives and definiteness spread for Hebrew possessives
embedding definite CSNs.15
The explanation that is currently agreed upon in the syntactic literature
(Fassi-Fehri (1989, 1993), Siloni (1994), Longobardi (1996), among many
others) is that the D of the head noun, although phonologically absent, is
syntactically present, in the form of an empty functional head that inherits the
+def or -def features from the genitive DP. This transmission of (in)definite
features was implemented either as copying/percolation or as an agreement
relation between Spec and the empty functional category,16 the identity of which
varies from one author to the other.17 This account is problematic: how can it be
that the def features, although realized only once, are interpreted twice?
Agreement phenomena are exactly of the opposite kind: a -feature (gender,
number, person, Case) is interpreted only once, although it shows up on several
elements.

15 For indefinite CSNs, see Section 4 of Dobrovie-Sorin (2002).

16 Borer (1988, 1996) proposes a somewhat different account, according to which
the percolation of def features is ensured via the incorporation of the genitive DP
onto the head N. This proposal cannot cover Saxon genitives or Romanian genitives.
17Theorists agree neither on the number nor on the label of the different functional
categories: see NumP, AGRP and PossP in Ritter (1991), Siloni (1994), Longobardi
(1996), among many others.

12

Given the analysis proposed here, the percolation of (in)definiteness18 is a

consequence of the functional analysis that characterizes genitive specifiers:
(21) a.
b.

= x y [Rgen(x,y) & farm (y)] (m)

= y [Rgen(m,y) & farm (y)]
[[a neighbour's farm]] = x y [Rgen(x,y) & farm (y)] ([[a neighbour]])
= y [Rgen(z,y) & farm (y) & a neighbour(z)]
[[Marys farm]]

In (21)a, the head N farm denotes a function (x y [Rgen(x,y) & farm (y)]) that
applies to the constant individual denoted by Mary; the value of the function is
itself a constant individual, hence the definite-like interpretation. In (21)b, on the
other hand, the denotation of a neighbour's farm is obtained by applying the
same function to the individual variable x introduced by a neighbour.19 The
denotation of the overall possessive varies with the values of the genitive
specifier itself.20
By definition, the rule of semantic composition that characterizes genitive
specifiers (see (16)) triggers uniqueness: for each value of the genitive there is
only one value for the function. However, the issue of uniqueness is complicated
by pragmatic factors, which lie beyond the main goal of this paper (see Barker
(1995) and Storto (2000)).

2.4. Determiners Block the Projection of Spec,DP

The fact that canonical determiners (i.e., determiners that apply to properties and
yield generalized quantifiers or individuals) block the projection of Spec,DP
follows from their semantic type: by applying Det (type <<e,t> <<e,t>t>>) to
NP (type <e,t>) we obtain a constituent that already has the semantic type of an
argument (type <<e,t>t>), and as such counts as a maximal projection of Det,
which can further expand only via adjunction. Under this view, examples of the
18 Although I continue to use the label "indefiniteness spread", it should be clear
that my analysis does not identify possessives embedding indefinite genitive
specifiers with indefinite DPs. From both a syntactic and a semantic point of view,
the two types of expressions are distinct, which predicts important differences (see
in particular Flaux (1992, 1993)): indefinite expressions are DPs obtained by
composing an indefinite Determiner with a set-denoting constituent, whereas
possessives embedding indefinite genitive specifiers are composed in an entirely
different way, via a function from individuals to individuals. Within this account,
"definiteness vs indefiniteness spread" means "constant vs variable interpretation".
19I assume here the analysis proposed by Discourse Representation Theory (Kamp
(1981), Heim (1982)), according to which indefinites translate as individual
variables.
20 A similar account of (in)definiteness spread can be found in Partee (1983/1997),
Vikner & Jensen (2002).

13

type in (22) are ungrammatical because the syntactic configuration in (22) is

ruled out by the rules of semantic composition:
(22) *John's

the/this/a

(22') *

DP
Spec, DP
DPGen

child

DP<<e,t>t>
Det <<e,t> <<e,t>t>>

NP<e,t>

2.5. Analytic Genitives

When the head noun is headed by a canonical determiner, genitive DPs can be
either N-complements or DP-adjuncts. In the three languages under
examination here, these positions cannot be occupied by synthetic genitives, but
only by analytic genitives:
(23) a.
b.

a picture *John's/of John

this habit *the neighbour's/ of the neighbour's

Analytic genitives are subject to important crosslinguistic variation: whereas

Romanian and Hebrew have only one morphological form for analytic genitives
(al-genitives, and el-genitives, respectively), English distinguishes between
genitives of the form of John and genitives of the form of John's. The
generalization seems to be that these two distinct forms occupy respectively the
N-complement position and the DP-adjunction position.
The complement position of N is subject to a linking constraint (Barker (1991,
1995)): it can be filled only if the head noun is either relational (see son, friend,
etc.) or a nominalization, which inherits the internal argument of the related
verb:
(24) a.
b.
c.

a son/friend/sister/picture of Mary
this translation of the Ilyad
* a house/pen/shirt of John

In examples of the type in (24)a-b, the genitive DP can be analyzed as filling the
argument position of the head N:
(24')

DetP
Det

a.
b.

14

a
this

NP
N

DPGen

picture
translation

of Mary
of the Ilyad

Possessives embedding genitive complements differ from those built with

genitive specifiers insofar as they involve a relation (rather than a function from
individuals to individuals). By applying the head N, which is lexically specified
as relational (type <e <e,t>>), to the genitive DP, we obtain a set of individuals
(type <e,t>) as the denotation of the constituent labelled N. Such a constituent
must combine with a canonical determiner in order to yield a generalized
quantifier:
(24")

DetP <<e,t>t>

Det <<e,t> <<e,t>t>>

a.
b.

a
this

NP<e,t>

N <e <e,t>>

DPGen (e)21

picture
translation

of Mary
of the Ilyad

Note that N'-constituents embedding a complement genitive are also able to

combine with a Spec,Nmax constituent:
(25) a.
b.

John's picture of Mary.

John's translation of the Ilyad.

Examples of this type fall under the functional analysis proposed in section 2.2:
the presence of the Saxon-genitive coerces its sister (here an N'-constituent) into
denoting a function from individuals to individuals:
(25') a.

fpicture of Mary = x y [Rgen(x,y) & picture of Mary (y)]

The denotation of the overall possessive DP is obtained via rule (16), by

applying this function to the individual denoted by Spec, Nmax:
(25") a.

[[John's picture of Mary]] = x y [Rgen(x,y) & picture of Mary (y)] (j)

= y [Rgen(j,y) & picture of Mary (y)]

21 I assume that when it is associated to a genitive complement, of is a semantically

empty preposition (acting as a Case marker), which as such does not change the
semantic type of the constituent to which it attaches. When of attaches to an adjunct
(see (24)d), the change in semantic type could be induced by the s marking on the
genitive (combined with the adjunct position itself).

15

Let us now consider genitives of the form of DPs. Since they are not subject to
the linking constraint illustrated in (24)a-b vs (24)c, genitives of this form are to
be analyzed as DP-adjuncts:
(24) d.

a house/pen/shirt of Johns

(24') d.

DetP
DetP
Det
a

DPGen
NP
house

of Johns

Adjunct genitives are like complement genitives and differ from genitive
specifiers insofar as they involve a relation (rather than a function from
individuals to individuals). However, they crucially differ from complement
genitives insofar as the relation is not supplied by the head N itself, but is instead
contributed by the genitive marker of. Adjunct genitives can be analyzed as
having the semantic type of predicates (type <e,t>), obtained by abstracting over
one of the argument positions of the relation Rgen, introduced by the of: [[of
the neighbour's]] = ?y [ Rgen (the neighbour, y)]. The semantic composition of
adjunct genitives will not be further investigated here. I will simply assume that
it is comparable to the semantic composition of relative clauses.
Romanian and Hebrew differ from English in having only one form of analytic
genitive. Two hypotheses need to be considered: (i) analytic genitives are not
specified for a particular syntactic position, being able to occur in either the
complement or the adjunction positions, depending on the context; (ii) analytic
genitives are always adjuncts. The latter option is strongly suggested for
Romanian by the fact that two genitive DPs cannot attach to the same head N:22
(26) a.

b.

*traducerea lui a Iliadei

translation-the his a Ilyadgen
'his translation of the Ilyad'
*distrugerea orasului a dusmanului
destruction-the city-thegen a enemy-thegen
'the enemy's destruction of the city'

The ungrammaticality of (26)a-b indicates that al-genitives cannot be in the

complement position. This piece of data constitutes evidence in favor of the
22 It should be observed that two genitives may attach to the same N provided that
one of the two is pronominal and prenominal: compare (26)a and (i), where the
pronoun is prenominal:
(i)
prima lui traducere a Iliadei first-the his translation his a Ilyadgen .

16

hypothesis that the N'-level of representation is not projected in Romanian (see

Ghomeshi (1997) on Persian).
2.6. Genitive Specifiers and Determiners: Overview
For the three languages examined above, the morphological distinction between
synthetic and analytic genitives has been argued to correlate with a syntactic
distinction, between the syntactic position of Spec,Nmax on the one hand, and
complement of N and DP-adjunct positions, on the other hand. I have further
argued that a genitive specifier (i.e., a genitive DP sitting in Spec, Nmax) is
interpreted as the argument of a function from individuals to individuals. This
rule of semantic composition accounts for the strict constraint that synthetic
genitives impose on the determiner of the head noun: it can be either null or a
suffixal definite article. The choice of one or the other option cannot be
explained by the semantic analysis. In particular, we do not know why a
(suffixal) definite article is obligatory in Romanian. Answers to these questions
could be found by taking into account the syntactic properties of a given
language, in particular the type of determiner (free morpheme, clitic or affix) and
the position (post- or pre-nominal) of the genitive specifier. For Romanian, the
necessity of a definite article on the head noun co-occurring with a genitive
specifier might be due to the existence of a strict constraint on morphological
case: it cannot be assigned to nouns, but only to determiners. In such a language
then, possessives must be headed by a determiner in order to be assigned
morphological Case. Similarly, the semantic analysis cannot predict which
languages exhibit an alternation between synthetic and analytic genitives.
References
Abney, S. (1986) "Functional Elements and Licensing", paper presented at
GLOW 1986, Girona, Spain.
Abney, S. (1987) The English noun phrase in its sentential aspect, MIT
dissertation.
Ayoub, G. (1991) "La nominalit du nom ou la question du tanwin", Arabica
xxxviii, 151-213.
Barker, C. (1991) Possessive Descriptions, dissertation, University of California
at Santa Cruz.
Barker, C. (1995) Possessive Descriptions, CSLI Publications, Dissertations in
Linguistics series, publication de Barker (1991)
Borer, H. (1988) "On the morphological parallelism between compounds and
constructs", in Booij&Jaap van Marle (eds.) Morphological Yearbook, Foris.
Borer, H. (1996) "The Construct in review", in J. Lecarme, J. Lwenstamm&U.
Schlonsky (eds.) Studies in Afroasiatic Grammar, Academic Graphics, The
Hague.
Boskovic, Zeljko, and Daiko Takahashi (1998) "Scrambling and last resort".
Linguistic Inquiry 29: 347-366.

17

Chomsky, N. (1994) Bare Phrase Structure, Occasional Papers in Linguistics 5,

MIT.
Chomsky, N. (1995) The Minimalist Program, Cambridge, Mass: MIT Press.
Cornilescu, A. (1993) "Notes on the Structure of Romanian DP and the
Assignment of the Genitive Case", University of Venice Working Papers in
Linguistics 3.2.
Cornilescu, A. (1999) "Aspect and Nominalizations: the Case of Romanian", in
Istvan Kenesei (ed), Current Issues in Linguistic Theory, Crossing Boundaries,
John Benjamins, 211-236.
Danon, G. (2001) 'Syntactic Definiteness in the Grammar of Modern Hebrew',
Linguistics.
Dobrovie-Sorin, C. (1987) "A propos de la structure du groupe nominal en
roumain", Rivista di grammatica generativa, 12: 123-152.
Dobrovie-Sorin, C. (2000a) "(In)definiteness Spread: from Romanian Genitives
to Hebrew Construct State Nominals", in V. Motapanyane, ed., Comparative
Studies in Romanian Syntax, Bejamins.
Dobrovie-Sorin, C. (2000b) "De la syntaxe l'interprtation, de Milner (1982)
Milner (1995): le gnitif", Lardreau, G. (ed.) Cahier Jean Claude Milner,
Verdier, Paris.
Dobrovie-Sorin, C. (2002) "From DPs to NPs: A Bare Phrase Structure Account
of Genitives", Coene, M. & D'Hulst, Y. (eds.) From NP to DP. Volume 2: The
expression of possession in noun phrases, Amsterdam/ Philadelphia: John
Benjamins (Linguistik Aktuell/Linguistics Today).
Engelhardt, M. (1999) 'Properties as Arguments', WCCFL 18.
Engelhardt, M. (2000) 'The Projection of Argument-Taking Nominals', NLLT
18: 41-88
Fassi-Fehri, A. (1989) "Generalized IP Structure, Case and VS Word Order", in
I. Laka & A. Mahajan (eds.) MIT Working Papers in Linguistics 10, 75-113.
Fassi-Fehri, A. (1993) Issues in the Structure of Arabic Classes and Words,
Kluwer, Dordrecht.
Flaux, N. (1992) "Les syntagmes nominaux du type le fils d'un paysan: rfrence
dfinie ou indfinie?", le Franais moderne, pp 113-140.
Flaux, N. (1993) "Les syntagmes nominaux du type le fils d'un paysan: rfrence
dfinie ou indfinie?", le Franais moderne, pp 23-45
Frege, G. (1891), Funktion und Begriff, reedited in Gottlob Frege, Funktion,
Begriff, Bedeutung Vandenhoeck and Rupprecht, Gttingen, 1969, 17-39.
Ghomeshi, J. (1997) "Non-projecting nouns and the EZAFE Construction in
Persian", Natural Language and Linguistic Theory 15, 729-788.
Giorgi, A. and G. Longobardi (1991) The Syntax of Noun Phrases.
Configuration, Parameters and Empty Categories, Cambridge University Press.
Grimshaw, J. (1990), Argument Structure, MIT Press, Cambridge, Mass
Grosu, A. (1988) "On the Distribution of Genitive Phrases in Romanian,
Lingustics 26: 931-949.
Grosu, A. (1994), Three Studies in Locality and Case, Routledge, London and
New York.
Heim, I. (1982) The Semantics of Definite and Indefinite Noun Phrases, PH.D.,
Univ. of Mass., Amherst.

18

Hornstein, Norbert (1999) Movement and control. Linguistic Inquiry 30: 69-96.
Jackendoff, R. (1974) An Introduction to the X-bar Convention, ms, Indiana
University Linguistics Club, Bloomington, Indiana.
Jensen, P.A. & C. Vikner (1994). Lexical knowledge and the semantic analysis
of Danish genitive constructions. In S.L. Hansen & H. Wegener Topics in
Knowledge-based NLP Systems, pp 37-55. Samfundslitteratur, Copenhagen.
Kayne, R. S (1994) The Antisymmetry of syntax. Cambridge, MA:MIT Press.
Kamp, H. (1981) "A Theory of Truth and Discourse Representation", in J.
Groenendijk, T. Janssen and M.Stokhof (eds.) Formal Methods in the Study of
Language, Mathematical Centre, Amsterdam.
Keenan, E. (1974) The Functional Principle: Generalizing the notion of
Subject of, in Michael W. La Galy, R.A. Fox and A. Bruck, CLS 10, Chicago,
Illinois, 298-310.
Lbner, S. (1985) "Definites", Journal of Semantics 4, 279-326.
Longobardi, G. (1996) "The Syntax of N-raising: a minimalist theory",
manuscrit, Universit de Venise.
Manzini, R. and A. Roussou (2000). A minimalist theory of A-movement and
control. Lingua 110, 409-447.
Miller, Ph. (1992) Clitics and Constituents in Phrase Structure Grammar, New
York: Garland.
Milner, J.C. (1982) Ordres et raisons de langue, Paris, Le Seuil.
Milner, J.C. (1995) "L'interprtation des gnitifs", in Langues et langage.
Problmes et raisonnement en linguistique. Mlanges offerts Antoine Culioli,
d. par J. Bouscaen, J.J. Franckel, S. Robert, PUF.
Munn (1998), 'The Possessor that Stayed Close to Home", Proceedings of
WECOL, ed. V. Samiian.
Partee, B. (1983/1997) Uniformity vs versatility: The genitive, a case study.
Appendix to Th. Janssen (1997), Compositionality, in: J. van Benthem & A.
ter Meulen (eds) The handbook of logic and language. pp 464-470. Elsevier,
New York.
Quirk, R., S. Greenbaum, G. Leach and J. Svartvik (1985) A Comprehensive
Grammar of the English Language, London : Longman.
Ritter, E. (1988) "A Head Movement Approach to Construct-State Noun
Phrases", Linguistics, 26.6, 909-929.
Ritter, E. (1991) "Two Functional Categories in Noun Phrases: Evidence from
Modern Hebrew", Syntax and Semantics 25, 37-62
Rouveret, A. (1994) Syntaxe du gallois. Principes gnraux et typologie, CNRS
Editions.
Siloni, T. (1994) Noun Phrases and Nominalization, Ph.D. dissertation,
University of Geneva.
Siloni, T. (1997) Noun Phrases and Nominalization, Kluwer.
Storto, G. (2000) On the structure of indefinite possessives. Proceedings of
SALT X.
Vergnaud, J.-R. & M.L. Zubizarreta (1992) The definite determiner and the
inalienable constructions in French and English. Linguistic Inquiry 23. 595-652.
Vikner, C. & P.A. Jensen (2002) A semantic analysis of the English genitive.
Interaction of lexical and fromal semantics Studia Linguistica 56: 191-226.

19

Woisetschlaeger, E. (1983) "On the question of definiteness in "an old man's

book", Linguistic Inquiry 14, 1, pp 137-154.

20