Lob Ner 2011

Journal of Semantics, 28, 2011: 279–333
doi:10.1093/jos/ffq022
Advance Access publication January 25, 2011
Concept Types and Determination

SEBASTIAN LÖBNER
Heinrich-Heine-Universita¨t Düsseldorf
Abstract
Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

The article offers a systematic account of four basic types of nouns, of four basic types
of nominal determination and of the interaction of noun type and determination
type. The four basic noun types are sortal, individual, relational and functional; they
correspond to the four logical types Æe,tæ, e, Æe,Æe,tææ and Æe,eæ on the one hand, and to
four types of concepts on the other. The four basic types of nominal determination
are singular definite, indefinite (a variety of determinations including simple
indefinite), absolute and relative, where possessive determination is a combination
of relative reference and possessor specification. Only inherently unique concepts are
available for singular definite reference, only nonunique ones for indefinite use, only
nonrelational concepts for absolute reference and only relational ones for relative use.
However, determiners of each type can be combined with nouns of each lexical type:
if the type restrictions are not met, the noun is shifted to an appropriate type. On this
basis, congruent and incongruent determinations are distinguished. Statistical,
typological and historical evidence is presented that corroborates the hypothesis that
incongruent determination is a marked option and involves type shifts of the noun.
1 INTRODUCTION
This article is an attempt to present the outlines of a semantic theory of
determination, based on the distinction of four basic conceptual lexical
types of nouns: sortal, individual, relational and functional. It will be
argued that the function of determination essentially is to form a sortal,
individual, relational or functional concept on the basis of the lexical
meaning of the CNP1—with the possible addition of further
conceptual content. The lexical type of the noun and its type under
determination need not match. On the contrary, determination can
yield any type of NP concept from any lexical type of noun. Therefore,
the focus of the theory will be on the interaction of lexical noun
meaning with determination.
The proposal is confined to NPs in referential use, be it
nonquantificational or quantificational, but it excludes predicative
1
Following the convention in Abbott (2010), the term CNP—common noun phrase—is used for
operands of determination.
The Author 2011. Published by Oxford University Press. All rights reserved.
For Permissions, please email: journals.permissions@oup.com.
280 Concept Types and Determination
and generic NPs. This is in line with the basic idea that the function of
determination is essentially to fix the type of reference for the NP.
Another range of phenomena not to be focused on is nominal
quantification proper. As I argued elsewhere,2 neither definites nor
indefinites constitute instances of quantification proper. There remains
a very small range of quantifying determiners proper: those that
correspond to the standard quantifications of predicate logic, that is,
expressions for ‘every’, ‘all’, ‘both’, ‘some’ or ‘no’.3
Thus, the scope of the article is restricted to indefinite, definite,
relative and absolute determination with sortal, individual, relational

and functional nouns in referential use. It is claimed that the theory
holds for both explicit and implicit determination. Based on general
notions of the conceptual type of noun and NP meanings and the role
they play in reference, it applies to all languages. Whether ‘‘concepts’’
are considered cognitive representations or mathematical constructs is
a foundational issue that by and large can be neglected here. The
distinctions and observations discussed here apply to both approaches.
2 LOGICAL TYPES OF NOUNS: BASIC DISTINCTIONS

2.1 Sortal, individual, relational and functional nouns and concepts
Nouns, including pronouns and proper names, have lexical meanings
of different logical types. The four basic types are sortal, individual,
relational and functional:
Sortal nouns are unary predicate terms, of type Æe,tæ. They

characterize their potential referents in terms of properties.
Their meanings are sortal concepts. In a given context of
utterance, there may be zero, one or more entities the noun
denotes.
Sortal nouns constitute the prototypical subtype of nouns. They
occur with all modes4 of determination; they naturally take adjectives
and attributes; they are syntactically simpler in that they do not need
2
See Löbner (1985: §2, 1987: §2, 2000: §2) for the nonquantificational character of ‘‘particular’’
definite NPs in general, including plural definites; Löbner (1987: §3, 1990: §3) for the
nonquantificational analysis of ‘‘particular’’ indefinites; Löbner (2000: §4) for a nonquantificational
account of simple ‘‘characterizing’’ (generic) indefinites such as singular indefinites with indefinite
article or bare plurals and mass nouns. The terms particular v. characterizing are used in the sense of
Krifka et al. (1995).
3
I use italics for quoting linguistic expressions, and single quotes for meanings and concepts;
double quotes are used in the sense of ‘what is called . . .’.
4
By ‘‘modes of determination’’ I mean varieties, or types, of determination such as definite,
indefinite, quantificational and so on.
Sebastian Löbner 281
complements such as possessor specifications. Sortal nouns include

terms for natural kinds as well as for kinds of artefacts, but these are by
no means the only lexical subtypes.
Individual nouns are individual terms, of type e. They include
proper names and personal pronouns, along with nouns for
unique institutions in a context of utterance: pope, US president,
Georgia, weather and temperature. Their meanings are individual
concepts that assign a unique referent to every appropriate
context of utterance.
Due to their inherent uniqueness,5 the default use of individual

nouns in argument position is singular definite; plural, indefinite and
quantificational use requires special contexts.
Relational nouns are binary predicate terms, of type Æe,Æe,tææ.6

Their meanings are binary relational concepts, involving
a further argument in addition to the referential argument.
Relational nouns characterize their referents in terms of
a particular relation to some other object. Since this object is
usually specified by means of a possessive construction, we will
call it the ‘‘possessor’’.7 In the case of relational nouns proper,
the characterizing relation is not necessarily to-one.8 In a given
context of utterance, and for a given possessor, there may be any
number of potential referents, including none.
Examples of relational nouns proper include most kinship terms,
terms for nonunique parts and a broad variety of deverbal nouns. Due
to their lack of uniqueness, relational nouns proper occur with
indefinite as well as definite determination, in singular or in plural (if
5
I am not happy with the term unique(ness), but there does not seem to be a better choice. In
German, I would use the term eindeutig, which means something similar to unambiguous. A
description is ‘‘eindeutig’’ if it can apply to only one case. A description is ‘‘unique’’ if there is just
a single case it applies to, but there may be more such cases possible. In a given context, girl with a red
hood may amount to a unique description because it may happen that there is exactly one such case in
that context; but the description would not be eindeutig. Pope is eindeutig because the concept
‘pope’ is such that there can be only one such person. (Even at the times when there were two popes,
each one claimed to be ‘‘the’’ pope.) Thus, ‘‘Eindeutigkeit’’ entails uniqueness but not conversely. I
hope that the discussion will leave no doubt that the term unique is used here in the stronger sense of
eindeutig. In Löbner (1985), I made a point of using nonambiguous or unambiguous instead of unique.
In this article, I will often need a term for the opposite case; obviously, ambiguous would be
misleading, and nonunambiguous a terminological monster.
6
Relational and functional nouns of higher arity will not be considered here.
7
Talking of the second argument as a ‘‘possessor’’ actually constitutes overgeneralization.
Obviously, there are alternative means, in particular for deverbal nouns, but also for others (cf.
English examples such as ticket to Rome or desire for apples).
8
That is, injective, or right-unique.
they are count nouns). For referential use, the possessor argument needs
to be saturated.
Functional nouns are unary function terms, of type Æe,eæ.
Their meanings are functional concepts, involving one
argument: the possessor. Functional concepts are restricted to
a domain of appropriate possessors and of appropriate contexts
of utterance. The value for a given argument, in a given context
of utterance, constitutes the uniquely determined referent of the
noun.

Functional nouns include relational role terms such as mother, author
and president; terms for unique parts such as head, top and cover; and
terms for abstract aspects, or dimensions: age, size, price, meaning, name
and odour. They combine the nonprototypical features of uniqueness
and relationality; they require saturation of the possessor argument and
are predisposed for singular definite use.
The type distinction is relativized by polysemy: a noun represents
a certain type only with respect to a given lexical reading. For example,
child has a relational reading ‘direct descendant of ’ and a sortal reading
‘nonadult’; moon has an individual reading ‘moon of the earth’ and
a relational reading ‘natural satellite of ’; subject can be a grammatical
term (functional), it can mean ‘member of a state’ (relational) or
‘individual, person’ (sortal).
It is convenient to relate the type distinction to two binary features,
[U] for inherent uniqueness and [R] for inherent relationality: sortal
nouns are [–U][–R], individual nouns [+U][–R], relational nouns
[–U][+R] and functional nouns [+U][+R]. These features are not
meaning components in the sense of feature semantics but just
descriptive abbreviations.
2.2 Types are specified in the lexicon

It is often argued that nouns of whatever subtype can be used for
relative and absolute reference, and for unique and nonunique
descriptions, whence their lexical type should be considered unspecific.
I argue that these lexical types of nouns are distinct since they are
defined in terms of logical types of concepts that constitute their
meanings. They are as distinct as unary v. binary predicates, or
predicates v. individual concepts in predicate logic. It is not possible to
assign a meaning to a noun while leaving open the logical type. Uses
that seem to contradict the type assignment, it will be argued, are
enabled by type shifts.
There are only two ways to technically evade fixing a lexical

concept type. One could assign every noun each of the four types, an
option that is clearly psychologically unrealistic because it violates laws
of economy. Alternatively, one might try to arrive at a uniform type by
assuming general underdetermination. If inherent uniqueness is left
open, the resulting superclasses would be sortal and relational. If in
addition the relationality distinction is left undetermined, the result is
relational. Thus, an underdetermination account would end up with
‘‘relational noun’’ as the only type, which is certainly not the universal
type of noun.9

2.3 Uniqueness is a semantic property of individual and functional
nouns
To see the difference between sortal and individual nouns, let us
consider a minimal pair in Japanese: kare (‘he’) v. otoko (‘male human’).
Kare is a third-person singular pronoun exclusively used for male
persons. The use of pronominal kare is definite, either deictic or
anaphoric. Kare carries a sortal specification ‘male person’, but it is
primarily an individual concept: the expression is used as though this
sortal description were uniquely specifying one referent in the given
context of utterance. The noun otoko has the same sortal content, but it
can be used indefinitely or definitely. Thus, the meanings of kare and
otoko differ only in that the former constitutes an individual noun and
the latter a sortal noun. This minimal pair illustrates that the difference
between individual and sortal concepts is not a matter of descriptive
content but a matter of concept type as such. The logical meaning
representations would be:
(1) a. kare ix (person(x) ^ male(x))
b. otoko kx (person(x) ^ male(x))
Proper names, another group of individual nouns, carry the sortal
specification that the name of the referent is that word; further sortal
information can be part of the meaning, such as ‘female person’. The
crucial point about proper names, again, is that they are definite in
referential use. Thus, a logical representation of the female proper name
Kim would be:
(2) Kim ix (person(x) ^ female(x) ^ name(x) ¼ [kim])
9
See also the remark on the asymmetry among the four types at the end of section 4.4.
Individual nouns other than proper names and pronouns can be

classified into subtypes: role terms (US president, pope), terms for
institutions (catholic church), terms for unique objects (sun, moon, earth)
or singular events (World War I); also, abstract terms for certain aspects
of the ‘‘world’’ such as ‘‘date’’, ‘‘temperature’’ (in the sense of
‘temperature of the air’), ‘‘weather’’ and so on. The referent of an
individual noun depends on the context of utterance, including a time
index, a location index and a constellation of facts (i.e. a ‘‘world’’ in the
sense of possible world semantics). The range of indices for which an
individual noun refers to a particular entity may vary widely. Some

have almost globally uniform reference (US president, Ghana), while
others like mosque, station, exit and so on have a particular referent only
in smaller domains. For such individual nouns, the index for fixing the
unique referent has to be determined more precisely. Proper names, in
particular given names, have a small range of uniform reference. This
does not contradict the assumption of semantic uniqueness of proper
names since uniqueness is relative to a given context of utterance. The
crucial difference between [+U] and [–U] nouns is this—unique nouns
‘‘say’’: this is the description of the referent, in the given context of
utterance there is exactly one that fits it. [–U] nouns say: this is the
description of the referent (it need not be unique).
For certain subtypes of [+U] nouns, uniqueness is due to their
descriptive content. For example, terms for concepts such as ‘age’,
‘temperature’ (in the general, functional sense) and ‘shape’ are
semantically unique because for logical reasons an entity can have
only one age, temperature or shape. Equally, role concepts such as
president, pope, dean and father define their potential referents in terms
of characteristics that entail that they can apply to only one individual.
A frequent objection to the approach taken here is that for any noun
that is supposed to be semantically unique there are contexts of
utterance where the uniqueness can be cancelled, whence uniqueness
should be considered a mere implicature or explained otherwise when
it comes about. Matushansky (2008) argues that the lexical type of
proper names is essentially Æe,tæ because this type fits the predicative and
nondefinite uses of proper names;10 the default uses of proper names in
argument position carry an implicit definite article. My argumentation
will be converse (section 6.2.2): proper names express individual
concepts; they are lexically of type e. If they are used predicatively or
10
For the sake of simplicity, I am disregarding the naming relation that Matushansky postulates as
a further meaning component. The question of uniqueness is independent of this assumption. See
Matushansky (2008) also for reference to further approaches that analyse proper names as not
inherently unique.
with indefinite or quantificational determination, they undergo a type

shift from type e to Æe,tæ; if used with a definite article, the article is
semantically redundant.
Certain subtypes of sortal nouns are frequently shifted to unique
concepts:
(3) a. Beware of the dog! (written on a board attached to a garden
gate)
b. He brought his toothbrush.
The sortal concept ‘dog’ can be shifted to an individual concept for

a dog that plays a certain role (pet, watchdog, etc.) in a particular range
of contexts of utterance.11 Similarly, sortal furniture terms like table and
bed can be used for established unique reference in certain
surroundings. ‘Toothbrush’ is a concept for a certain type of artefact
with a purpose specified. Toothbrushes are mostly used by one person
exclusively; this results in a mapping from toothbrushes to persons,
whence sortal ‘toothbrush’ can be shifted to the functional concept ‘the
object with the sortal characteristics of a toothbrush which possessor
uses to clean his/her teeth’. These particular shifts occur frequently
with large classes of nouns and without drawing on particular
contextual information. They interfere in a principled way with the
lexical meanings of nouns that undergo these shifts.
2.4 Relationality is a semantic property of relational and functional

nouns
The fact that [+R] is a lexical property of relational and functional
nouns is well established (cf. Barker forthcoming). The crucial criterion
is the presence of an additional ‘‘possessor’’ argument. Depending on
the grammar of the language, specification of the argument may or may
not be syntactically obligatory. In English, it is not, but there is
a number of Native American languages, where it is (see sections 6.2.3
and 6.2.4). Even if possessor specification is grammatically optional, it is
mandatory for semantic and pragmatic reasons. Possessor arguments
need to be taken care of in one way or other, if not by explicit
specification, then by mechanisms such as existential saturation,
abstraction or possessor retrieval from the context.12
11
See Gerland & Horn (2010) on ‘‘permanently established individual concepts’’.
12
Partee (1997) argues that there are ‘‘intransitive’’ relational nouns, such as stranger, which do not
allow possessor specification. If this claim is correct, for such nouns the possessor determination
would always be pragmatic.
If a noun meaning provides for an additional argument, this is not

just an additional lambda operator. The meaning also specifies a relation
to the referential argument: there is a ‘‘lexical’’ possessor relation for
[+R] nouns. For example, in tooth of a dog, the relation between the
referent and the possessor is that the former is a tooth of the latter. For
[–R] nouns, no predefined possessor relation comes with the meaning
of the noun: my stone can mean ‘the stone I am holding in my hand’,
‘the stone I threw’, ‘the stone I am carving’, ‘the stone I am sitting on’,
‘the stone I have been talking about’ and so on. In view of the
optionality of the possessor argument, the presence or absence of

a lexically specific relation to possessor is a useful diagnostics for [+R]
nouns.
The lexical relation given for [+R] nouns can be overridden by
a pragmatically given relation (Partee & Borschev 2002). For example,
‘‘my tooth’’ could be a dog’s tooth that I wear on a string around my
neck. In such cases, a shift applies that derelativizes the noun by taking
care of the lexical possessor argument; for example, the possessor
argument of tooth in my tooth may be saturated existentially. The
resulting [–R] concept is then added some contingent relation to the
possessor [see (59) in section 6.2.4].
It has been observed that there are privileged relations for
possessive uses of certain classes of [–R] nouns, in particular terms for
artefacts. Using qualia theory (Pustejovsky 1995), Vikner & Jensen
(2002) derive the general availability of certain readings from the
lexical qualia structure. For example, my book lends itself naturally to
the readings the book I am reading and the book I wrote because the
qualia component specifies reading and writing as typical actions
performed with books. Shifting such nouns to [+R] concepts (cf. the
case of toothbrush mentioned above) is quite common. Although
these shifts are very frequent, one would not assume that a concept
such as ‘toothbrush’ involves an additional argument: reference to
a ‘‘toothbrush’’ is possible without taking care of any possessor
argument.
2.5 Meaning and context

I consider the meaning of an expression, for example, a proposition, as
that which results from the application of regular semantic composition
to lexical meanings and grammatical information. No extralinguistic
information enters composition. The proposition of the sentence I am
tired is the same, independent of the context of utterance. When an
expression is uttered in a given context of utterance, its reference gets
fixed and it takes on utterance meaning. The utterance meaning of I am

tired depends on the context of utterance.
As to the notion of context: ‘[n]othing short of an extremely
comprehensive theory of linguistic communication could provide
a general solution to the problem of context dependence’, state Stanley
& Szabó (2000: 220). So far, such a theory does not exist. It was stated
above that the reference of individual and functional nouns depends on
a time–location–world index. This is a necessary assumption, but an
index still does not make up a full context. What is needed is a dynamic
theory of context, where the context changes with every utterance as

well as with the course of time and the development of facts. It is
obvious that, for example, anaphoric definites can only be fully
analysed in a dynamic context framework. DRT (Kamp & Reyle 1993)
and dynamic semantics (Groenendijk & Stokhof 1991) are theories that
cover some aspects. In addition, a theory of context must cover general
communicational aspects such as the indication and negotiation of the
relevant context among interlocutors. Gumperz’s (1982) theory of
contextualization takes initial steps in this direction. An interactional
theory of context would also cover aspects such as topicality and
salience.
3 TYPES OF NOUNS AND TYPES OF DETERMINATION (1):

DEFINITE AND INDEFINITE
3.1 Basic correspondences (1): uniqueness and definiteness
The properties that distinguish the types of nouns, that is, uniqueness
and relationality, correspond to types of determination and reference.
Clearly, uniqueness is linked to definiteness, and relationality to
possessive determination. However, the correspondence between noun
type and modes of determination is not straightforward.
For the theory to be developed here, we first have to settle the
question: Which modes of determination are ‘‘natural’’ with which
types of CNPs? The second question will be: Which type of reference
results from the application of certain determinations? By a ‘‘natural’’
case of determination, I mean a determination that operates on the
CNP in its original lexical meaning without coercion. Later, we will
refer to the ‘‘natural’’ cases of determination as ‘‘congruent de-
termination’’ (section 5). In this section, I will restrict the discussion to
uniqueness.
For [+U] CNPs, singular definite determination is the only type that
is natural, as it is in harmony with the feature of inherent uniqueness.
For [–U] CNPs, there is a wider range of determinations, that is, all
those that require, or at least admit, a number of cases other than one.
These include the following:
(4) modes of determination that are natural with inherently [–U]
CNPs
determination examples in English
simple indefinite a(n) Nsing, bare plural, bare mass
nouns
unspecific indefinite some Nsing
free choice any N

negative indefinite no N
contrastive demonstrative this, that Nsing
interrogative which N
plural (plural)
numerical cardinal numerals
quantitative many, several, some (plural,
mass), etc.
quantificational every, each, all, both
Simple indefinite NPs have one referent when they are interpreted
at the utterance level, but there may be a choice among several single
cases to which the bare description would apply. Free choice,
interrogative and contrastive demonstrative determinations involve
the possibility of more than one case; negative indefinite determination
of less than one. Plural NPs have one complex referent that is
composed of possibly more than one single case. For indefinite plural
NPs, there may be a choice among several complex referents. For
definite plural NPs, the complex referent is unique. Particular
quantificational determination involves reference to a complex domain,
out of which nonunique cases are referred to.
Common to all types of determination in (4) is the fact that in the
given context the number of single cases may not equal 1; this is why
nonunique input is required. Independently, the output may be an NP
with unique or nonunique reference. I will therefore distinguish
between I(nput)-uniqueness and O(utput)-uniqueness: I-uniqueness
relates to the uniqueness of the CNP, O-uniqueness to the reference of
the NP.
(5) Input-uniqueness and Output-uniqueness
determination I-unique O-unique
singular indefinite no no
plural indefinite no no
plural definite no yes

contrastive demonstrative no yes
singular definite yes yes
It follows from these observations that if the CNP is just a noun, the
determinations in (4) are natural with [–U] nouns, while singular
definite determination is natural with [+U] nouns. This difference in
terms of natural determinations adds essentially to the characterization
of noun types in section 2.1.

3.2 The function of determination (1): definite and indefinite
The next step reverses the perspective of nouns and determination. We
will analyse types of determination as operations on nominal concepts.
The approach taken is a natural extension of the theory of definiteness
developed in Löbner (1985). There it was proposed that definite
determination is not only natural with [+U] CNPs but in fact the
definite article has the function to indicate that the CNP is to be
construed as a unique concept: reference with definite NPs is reference
in terms of an individual or functional concept. The approach explains
why [+U] CNPs go naturally with singular definite determination:
their regular meanings are congruent with the concept type indicated
by the definite article. Therefore, definite NPs with [+U] CNPs were
called ‘‘semantically definite’’ in Löbner (1985). If inherently [–U]
CNPs are used for definite reference, for example, in anaphoric use, the
CNP concepts are enriched with contextual information in order to
render the individual or functional concept required. These cases were
called ‘‘pragmatically definite’’. The approach is based on three
assumptions:
(i) The type of reference of an NP token is defined in terms of
a conceptual type.
(ii) Definite articles indicate the conceptual type of reference (as
[+U]).
(iii) The token interpretation of the NP builds upon the lexical
meaning of the CNP; if the concept type indicated by the
article differs, the CNP concept is shifted to fit.
The theory of definiteness will now be extended to determination
in general. The essential function of determination is to map its input
CNP onto a specific conceptual type for reference. O-unique
determinations map the input onto a unique concept, O-nonunique
determinations map the input onto a nonunique concept and so on.
Using the O-characteristics of modes of determinations, we can set up

the following initial table, arrows representing the mappings.13
(6) determination output
simple definite / [+U]
simple indefinite,
unspecific indefinite,
free choice / [–U]
The only function of simple definite or indefinite determination is
to fix the output type. The other cases mentioned in (4) involve

additional steps. Demonstrative determination requires [–U] input to
be mapped on [+U] output. Plural and quantitative determination
requires [–U] input and maps it onto [–U] output with additional
quantity specification.
(7) determination input output
demonstrative [–U] / [+U]
plural, quantitative
numeral [–U] / [–U]
Plural definites result from two steps: first plural on a [–U] input,
then definite determination:
(8) plural definites input output
1. plural [–U] / [–U]
2. definite / [+U]
Quantificational determination is achieved by means of genuine
(logical) quantifiers such as ‘all’, ‘every’, ‘each’, ‘both’ or with numerals
or quantity expressions as ‘‘weak’’ quantifiers. Particular quantification
involves reference to the domain of quantification, a totality of single
cases. This is explicit in partitive constructions, for example, some of
NP, and implicit if the CNP is directly combined with the determiner.
Reference to the domain of quantification with count CNPs is plural
definite (Löbner 2000: 258 ff.). The partitive construction partitions
the domain into a set of cases for selection. Adding a quantificational,
quantitative or numeral element selects cases out of the set, for
predication. Every selects arbitrary individual cases out of the set; all
selects arbitrary members of a cover of the set;14 three selects three
13
If there is no entry to the left of the arrow, this means that the mapping carries no input
restriction. Equivalently, ‘/ [+U]’ can be replaced by ‘[6U] / [+U]’.
14
See Löbner (2000: §§3.3, 3.4) for the difference between ‘‘every’’ and ‘‘all’’ quantification.
arbitrary individual cases and so on. The total series of mappings

involves four steps:
(9) quantification input output
1. plural [–U]/[–U]
2. definite /[+U]
3. partition [+U]/[–U]
4. quantitative/numeral [–U]/[–U]
Steps 2 and 3 together constitute the partitive component of
particular quantification; their characteristics are [–U] / [+U] /

[–U]. The [+U] output of the first step corresponds to the ‘‘partitive
constraint’’. The output of the second step, that is, of the partitive
as a whole, is [–U], reflecting the ‘‘anti-uniqueness’’ of partitives
(cf. Barker 1998 for both observations, Chierchia 1995: §2 for the
latter).
In addition to the type mappings, determiners or equivalent
functional elements may carry further semantic or functional in-
formation such as classification, quantity specification or deictic
distinctions.
3.3 Explicit and implicit (in)definiteness

Many languages do not mark simple definiteness and/or indefiniteness.
In Dryer’s (2008a) sample of 566 languages, 281 have a definite article
distinct from demonstrative words or a definite affix on the noun, 55
use a demonstrative word as marker of definiteness and 230 do not
mark definiteness. For indefiniteness, Dryer (2008b) states that out of
473 languages, 269 do not mark indefiniteness, 91 have an indefinite
article different from the numeral for 1, 90 use this numeral as a marker
of indefiniteness and 23 languages mark indefiniteness with a noun
affix. Apparently, there are sufficient other means for deciding on the
uniqueness of reference. Such means include information structure, as
referential topics tend to be definite.
If languages mark definiteness or indefiniteness, marking is usually
incomplete. English does not mark definiteness of proper names, nor
indefiniteness for mass nouns and plurals. We consider cases where
(in)definiteness is not marked, but reference is (non)unique for other
reasons, as carrying (in)definite determination implicitly.
Plural is not universally marked either. For a sample of 291
languages, Haspelmath (2008) states that in 133 languages marking
plural is obligatory for all (count) nouns, 40 languages have obligatory
plural marking for animate nouns, in other languages plural marking is
optional for all nouns or for animate nouns; 28 languages totally lack
nominal plural.
Among the other modes of determination mentioned in (4),
numerical, quantitative, interrogative and negative determinations
cannot be implicit because they all involve a further function in need
of explicit expression. The same holds for contrastive demonstratives,
which, in addition to unique reference, indicate a deictic distinction
such as proximity v. distality, or relation to speaker v. addressee (Diessel
2008). Free-choice indefiniteness emphasizes nonuniqueness; such
emphasis, too, needs explicit expression. Whether there are implicit

quantifiers is a complex problem, connected to fundamental assump-
tions about semantics. Many analyses of, for example, plurals, generic
indefinites or donkey sentences, are based on the assumption of implicit
quantification. I consider such analyses as a deviation from the principle
of compositionality (Löbner 2000: 269 ff., 280 ff.). In my view,
quantification is necessarily explicit; for the constructions with ‘‘hidden
quantifiers’’, in fact a nonquantificational account is adequate. Simple
definiteness and indefiniteness present a choice between elementary
types of concepts (basically IC and SC, respectively; see section 5.5).
Compared to these standard determinations, quantification proper,
terms of universal or existential quantifiers, is a heavy instrument that
not only affects the type of reference of the NP but also the
predicational structure of the whole sentence by imposing a second-
order predication on it. It is plausible that such functions are in need of
explicit expression.
To sum up, among the modes of determination mentioned in (4),
simple indefiniteness (including nonspecific), simple definiteness and
plural may be implicit. The assumption of implicit determination does
not constitute a commitment to null determiners. The only
assumption made here is that determination, that is, the reference
type assigned to a given NP token, is not left open in communication:
in other words, a referential noun token will be assigned either unique
or nonunique, either relational or absolute reference, and in this sense
‘‘carries’’ a specific type of determination on whatever basis it is
determined.
3.4 A criterion for definite determination

The fact that definite determination may be implicit calls for an
independent test. Such a test can also be used for deciding on the
nature of determiner elements whose function is still to be assessed.
Lyons (1999: 15 ff.) cites four tests for (in)definiteness in English:
(10) a. subject of existential there constructions

There was a bottle / *the bottle rolling on the floor.
£
b. £adj as NP (be) constructions
Clever as *a girl / the girl is, she’ll notice.
c. certain predicative possessive constructions
*A house / the house is mine.
d. partitive constructions
Some of *houses / the houses are damaged.
These tests are basically heuristic in nature. It is not clear how the
traditional characterizations of definiteness in terms of familiarity,

uniqueness or inclusiveness would predict the behaviour of (in)definite
NPs in such constructions. In addition, as Lyons (1999: §1.2.1, §6.2)
points out, these criteria do not distinguish accurately between definite
and indefinite NPs. Existential there constructions also exclude
quantificational NPs with ‘‘strong’’ quantification such as ‘all’, ‘every’,
‘both’ or ‘most’. The adjectival construction in (10b) allows for generic
indefinites; the predicative possessive construction in (10c) is good with
quantificational subjects (cf. some of the houses are mine); partitive
constructions admit certain indefinite NPs. Also, the tests are more or
less language specific.
The test proposed here directly derives from the analysis put
forward: since definite NPs are of type e, their logical behaviour must
be that of individual terms. The criterion was first proposed in Löbner
(1985); Löbner (2000: 215 ff.) provides a comprehensive discussion.
Basically, a natural language NP behaves like an individual term in
predicate logic iff it does not exhibit a difference between Boolean
sentence negation and predicate negation, or between Boolean
sentence conjunction and predicate conjunction.15 For a formal
£
definition, let £n + p represent the syntactic combination of an NP

n and a predicate expression p, for example, a VP; let ^ represent
predicate conjunction and : predicate negation.
(11) a. negation condition
For the NP n and arbitrary predicate term p:
£
£n + p is false iff £n + (: p) is true

b. consistency
£
If £n + p is true, necessarily £n + (:p) is false.

£
Or: £n + p and £n + (:p) cannot both be true.

15
Boolean sentence negation yields a sentence with opposite truth-values; it is presupposition
preserving. Boolean predicate negation yields a predicate with the same domain of application,
assigning opposite truth-values for every argument. Boolean conjunction of sentences and predicates
is defined analogously.
c. completeness
£
If £n + p is false, necessarily £n + (:p) is true.
£
Or: £n + p and £n + (:p) cannot both be false.
Negation of a predication over a particular object is equivalent to
stating the complementary predication. Negation of a predication with
a nonunique argument term does not amount to a denial of a particular
case. Let us assume that leave and stay are complementary predications:
(12) negation test
a. Sue didn’t stay. 5 Sue left.

b. neg (One man stayed.) ¼ No man stayed. P One man left.
If Boolean sentence negation is available in the language considered,
a sentence is false iff its negation is true; (11a) then means that sentence
negation and predicate negation are equivalent, and conditions (11b)
and (11c) break the equivalence down into the respective pair of
entailments. Quantificational NPs of the type ‘all’, ‘every’, ‘each’,
‘both’, ‘most’, ‘some’ and ‘no’ all exhibit different truth-conditions for
predicate (‘‘internal’’) negation and sentence (‘‘external’’) negation.
‘‘Strong’’ quantifiers, intuitively those that claim truth for more
than half of the cases, fulfil the consistency condition but not
completeness. Indefinites are inconsistent: in the same context of
£
utterance both £n + p and £n + (:p) may be true. They are also

£
incomplete: £n + p and £n + (:p) may both be false.

(13) a. conjunction condition
For the NP n and arbitrary predicate terms p1 and p2:
£
£n + (p1 ^ p2) is true iff £n + p1 and £n + p2 are both

true.
b. upward monotonicity
£
If £n + (p1 ^ p2) is true, then necessarily £n + p1 and

£
£n + p2 are both true.

c. conjunctivity
£
If £n + p1 and £n + p2 are both true, then necessarily

£
£n + (p1 ^ p2) is true.

Intuitively, the conjunction condition holds for uniquely referring
NPs because all tokens of such an NP in the same context refer to the
same referent. If two predications are made in the same context using
the same uniquely referring NP argument term twice, they amount to
the conjoined predication over the argument. By contrast, tokens of
a nonuniquely referring NP, even in the same context of utterance,
may refer to different referents [cf. (14a) v. (14b)]:
(14) conjunction test

a. Sue is blind and Sue is a spy. 5 Sue is a
blind spy.
b. One man is blind and one man is a spy. P One man is
a blind spy.
Again, (13a) is the conjunction of (13b) and (13c). According to
(13a), sentential conjunction is equivalent to predicate conjunction;
(13b) holds if predicate conjunction entails sentence conjunction,16 and
vice versa for (13c). The condition of upward monotonicity is fulfilled
by all NPs that are intuitively positive: simple indefinite, definite,

demonstrative, lower bound quantitative and quantificational NPs, but
not negative determiners or quantifiers of the ‘no’ or ‘(only) few’ type.
Consequently, this property does not distinguish indefinites from
definites, unlike conjunctivity, which is fulfilled by definites, but not by
indefinites.
Of all types of NPs, those expressing universal quantification (‘all’,
‘every’, ‘each’, ‘both’) come closest to fulfilling the logical criteria for
uniqueness: they fulfil the two conjunction criteria and the criterion
of consistency. The only condition they violate is completeness.
Barwise and Cooper (1981: 183 f.) define definiteness by a criterion
that is equivalent to the conjunction of the three criteria fulfilled by
universal quantifiers.17 Lyons (1999: 11 f.), too, considers the definite
article essentially equivalent to all. This view is logically inadequate in
that it disregards the fact that definite NPs, unlike universal
quantifiers, fulfil the completeness criterion. Sentential and predicate
negation coincides for definite NPs. It must be emphasized that this
also applies to plural and mass definites. In these cases, Boolean
negation yields an all-or-nothing contrast due to a presupposition of
indivisibility (Löbner 2000: 239–51). The sentence in (15a) is true iff
every single one of ‘‘the kids’’ is eating spinach, and it is false iff those
kids are all not eating spinach, which is what the negation in (15b)
expresses.
(15) a. The kids are eating spinach.
b. The kids are not eating spinach.
£
16
Upward monotonicity is usually defined as follows: if p entails p’ then £n + p entails
£
£n + p’ . This is equivalent to (13b): let n be upward monotone according to the usual definition;
£
then, if £n + (p1 ^ p2) is true,£n + p1 and £n + p2 are both true since £p1 ^ p2 entails both p1
and p2. Conversely, let n fulfil the condition in (13b), and let p1 entail p2. Then, p1 is equivalent to
£
£p1 ^ p2 . Therefore, if £n + p1 is true,£n + (p1 ^ p2) is also true, whence by (13b) £n + p1 and
£
£n + p2 are true.
17
According to Barwise and Cooper, an NP is ‘‘definite’’ if it denotes a principal filter.
It can be shown that for any referential NP that fulfils the four
criteria, there is exactly one individual corresponding to it (Löbner
2000: 252), whence the NP refers uniquely.18
For a positive test of unique reference, all four criteria have to be
positive; for a negative test, it is often convenient to just test
consistency. The tests are most probably universally applicable. They
are formulated in a way that does not presuppose grammatical means of
Boolean sentence negation or Boolean sentence conjunction. Even
Boolean predicate negation (which is probably universal) and Boolean
predicate conjunction need not be available as general grammatical

means. All one needs for the negation test is pairs of predicate
expressions p and p’ that are Boolean opposites. The conjunction test
can be performed with triples of predicate expressions p, p1 and p2 such
that p is the Boolean conjunction of p1 and p2, for example, for
‘woman’, ‘person’, ‘female’.
There is a second class of NPs that fulfils the negation condition and
the conjunction condition: characterizing NPs such as English generic
bare plurals and mass nouns and simple singular indefinites (Löbner
2000: §4). The fact that the criterion is positive for both referential
definites and characterizing NPs is of little harm since the distinction
of these two types of NPs is easy enough. The common logical
behaviour of characterizing NPs and definite NPs might be considered
a motivation why in many languages, for example, French,
characterizing NPs receive singular definiteness marking (Lyons 1999:
51).
Contrastive demonstratives, uncontroversially recognized as defi-
nite, seem to violate the consistency condition: for example, the
sentence in (16) is not contradictory:
(16) THIS man is dumb and THIS man isn’t.
The sentence is not contradictory if one assumes that the two
occurrences of this man refer to different men. If this is assumed, the
context of utterance on which the reference depends changes in
midsentence. After reference of the first occurrence is fixed, the
context is changed to admit a new demonstration by which the
reference for the second occurrence is fixed. The notions of logical
entailment and logical equivalence are defined for one fixed context of
utterance for the sentences involved: A entails B iff in every context of
utterance, B is true if A is true. Thus, sentences like (16) do not show
18
Therefore, definite NPs denote principal ultrafilters, rather than just principal filters (cf. Löbner
1987: 183 ff.)
that demonstrative NPs violate the consistency condition. Note

that—if the reference of two or more occurrences of one de-
monstrative NP is fixed for the whole sentence—the criterion holds,
along with the other definiteness criteria.
3.5 Definite determination and unique reference

Up to this point, it was assumed that a referential NP with definite
determination refers uniquely. There are, however, apparent excep-
tions: indefinite NPs used for unique reference and NPs with definite

article, but nonunique reference. One such case is German NPs with
the superlatives meist- (‘most’) and wenigst- (‘least’), which obligatorily
contain a definite article, unlike their English equivalents.19
(17) Die meisten / die wenigsten Kinder sind infiziert. (German)
the most / the least children are infected
‘Most / only very few children are infected.’
Another example is provided by German constructions denoting
£
£
a half: £die Ha¨lfte + possessor NP or £definite article + halbA + N
(18) a. Die Hälfte der Besucher ist eingeschlafen.
the half of the visitors has fallen asleep
‘Half of the visitors have fallen asleep.’
b. Das halbe Brot ist schon aufgegessen.
the half bread is already consumed
‘Half of the bread is already consumed.’
The half-constructions are inconsistent and hence do not refer
uniquely: die Ha¨lfte der Besucher ist eingeschlafen und die Ha¨lfte ist wach
geblieben ‘half of the visitors fell asleep and half of the visitors stayed
£
awake’ is not a contradiction. NPs of the form £die meisten N ‘most N’

are consistent, but violate the conjunctivity condition. £Die wenigsten
£
CNP ‘only very few CNP’ is not upward monotone: if there are only
£
very few cases for which £p1 ^ p2 is true, it does not follow that p1 and
p2 each are true for only very few cases.
The discrepancy between what appears to be definite determination
and actual logical properties can be explained away, if one observes that
these constructions carry idiomatic meanings. German £die meisten
£
CNP constructions regularly receive the same interpretation as their

£
English counterparts £most CNPplur/CNPmass . Likewise, the half-

constructions in (18) receive the indefinite partitive interpretation of
19
For a recent analysis, and a comparison with English, see Hackl (2009).
£
English £half of NP . Hence, these constructions simply are not definite.
What these cases show is not that definite NPs may be used for
nonunique reference; rather that idiomatic NP constructions contain-
ing a definite article need not carry definite determination.
As to the converse case, there is a way of using specific indefinites to
the effect of unique reference. There is a special method of avoiding
literal unique reference, but leaving it to a conversational implicature
(‘Oh, somebody has been indiscreet!’ said in the presence of the one
concerned, where everybody knows who is meant). Such uses of
indefinites do not present mismatches between determination and

reference, either. The indefinite NPs in these cases are used as for
nonunique reference. Grice’s maxim of quantity is violated, but overtly,
for example, by looking at the person meant, and this triggers the
conversational implicature that the speaker could as well have used
a more precise, that is, unique description. Note that the special effect
of insinuation only arises because the reference is literally nonunique.
In ‘‘configurational uses’’ (Löbner 1985: 304 ff.), definites are used
although ‘‘objectively’’ there may be more than one single case:
(19) a. She put her hand on his knee.
b. He was the son of a poor farmer.
c. I do not want my daughter to marry the lover of her sister.
In such uses, relational, but not necessarily functional, nouns are
used with unique determination—hand, knee, son, daughter, lover,
sister—although in the situation given there may be, or even necessarily
is (cf. the two daughters in c), more than one case to which the concept
applies. In these uses an abstract type of situation, a configuration, is
described, and reference by the relational nouns is unique relative to
the abstract configuration.20
When we turn to possessive chains in section 4.3, we will encounter
a different type of possible mismatch: possessive NPs with a definite
article and an indefinite possessor specification, such as the father of
a girl. Such NPs are indefinite although the Det element of the head
NP is the definite article. As will be shown below, the compositional
rules for such constructions regularly yield indefinite determination
of the whole NP. Thus, again, there is no mismatch between
determination and the type of reference.
As a result, we will assume that for referential NPs (in)definite
determination does match with (non)unique reference.
20
For alternative analyses of ‘‘weak definites’’ and ‘‘weak possessives’’, see Poesio (1994) and
Barker (2004).
4 TYPES OF NOUNS AND TYPES OF DETERMINATION (2):

RELATIONAL AND ABSOLUTE
4.1 Basic correspondences (2): relational and possessive
Relational determination links the referent of the noun to a possessor.
For example, my cousin in argument use links the cousin to the speaker.
Again, we distinguish between the I and the O characteristics of reference
types. [+R] CNPs are natural with I-relational, [–R] CNPs with I-
absolute determination. An NP with a possessor specification, such as my
cousin in (20a), has O-absolute reference since the specification of the

possessor by the possessive pronoun saturates the possessor argument.
(20) a. I met my cousin.
b. I met a cousin.
c. I met a logician.
Therefore, possessive pronouns are I-relational and O-absolute. By
contrast, the determination of a cousin in (20b) is O-absolute and I-
absolute. [+R] cousin is used absolutely; the lexical possessor argument
of cousin must be saturated pragmatically; note that it need not be the
speaker’s cousin. Example (20c) is a case of I-absolute and O-absolute
determination for a [–R] noun.
In the following, we need to distinguish between NPs and maximal
NPs: a maximal NP is an NP that is not a proper part of another NP. In
(21), the father of a student is a maximal NP, but neither the father nor
a student is.
(21) The father of a student came to my office hours.
For maximal NPs in referential use, a basic constraint applies:
(22) Absoluteness constraint on maximal NPs
Referential maximal NPs carry absolute determination.
The constraint derives from the simple fact that the utterance
meaning of an NP cannot be determined as long as there is an open
possessor argument. If the possessor is not explicitly specified, the
possessor argument must be taken care of by some way of coercion.
In English there are three constructions that are I-relational and
hence natural with [+R] CNPs: (i) possessive pronouns £pronounposs
£
CNP , (ii) the Saxon genitive £NPposs’s CNP and (iii) of-PP
complements £CNP of NPposs .21 We will refer to the first two
£
21
In addition, there is the ‘‘double genitive’’ construction (cf. books of John’s or books of mine). I
follow Barker (1998) in assuming that these are not possessive but partitive constructions.
constructions as left possessives, and the third as right possessive. While

the pronominal construction is O-absolute, the two other construc-
tions have an open NPposs position. Whether or not the result of filling
this position is an absolute NP depends on the possessor NP: if it is
relational, another possessor specification may be added to it, and so
on.
(23) [R] characteristics of relational determinations in English
determination input output
possessive pronoun [+R] / [–R]
Saxon genitive [+R]

of-PP [+R]
The subtypes of definite and indefinite determinations discussed in
section 3 are indifferent to the [R] features. Maximal definite and
indefinite NPs are absolute, but as nonfinal components22 of possessive
chains they are neutral. For example, the father as a maximal NP is [–R],
while in the father of a student it is [+R].
4.2 Explicit and implicit relationality

Due to the absoluteness constraint, the matter of implicitness is
different for the [R] feature. An NP carries relative determination if it
denotes the possessum in an explicit possessor specification; else, its
determination is absolute. Most languages do not mark absolute use by
explicit functional elements (for exceptions, see section 6.2.4). But if
we consider the lack of an embedding possessive construction as an
overt, hence explicit, expression of absolute determination, it follows
that along with relative and possessive determination, absolute
determination is always explicit.
If a [+R] CNP is used referentially without a possessor specification,
a shift to [–R] is coerced. This is not to be mistaken as implicit
relationality because such NPs are used for absolute reference. [+R] NPs
are systematically used without possessor specification in at least three
cases. One is associative anaphora (see sections 5.2 and 5.3), the others are
predicative possessive constructions (24) and external possessor con-
structions. The issue of implicitness is open for predicative possession:
(24) a. verbal: This word has a special meaning / has the
meaning ‘rat’.
b. prepositional: a word with a special meaning / a word with the
meaning ‘rat’
22
These ‘‘components’’ are not necessarily constituents (cf. the end of section 4.3).
If these constructions are used with [+R] CNPs, the verb have or
the preposition with relate possessor and referent with the relation
semantically defined. The underlying relational meaning is not
suspended by coercion, although the NP comes with absolute
determination.
In external possessor constructions, the possessor specification is not
part of the NP referring to the possessum; hence, the possessum NP is
absolute, but saturation of the possessor argument nevertheless makes
use of an explicit possessor specification. In (25), the dative is construed
as possessor specification.

(25) (German)
Sie schüttelte ihm die Hand.
she shook him.DAT the hand
‘she shook his hand’
4.3 Possessive chains

A possessor specification can be expressed by a determining element
such as possessor affixes in Hungarian or possessive pronouns in
English.
(26) Hungarian23
a. a könyv- em
DEF book- POSS.1SG
‘my book’
b. egy könyv- em
INDF book- POSS.1SG
‘a book of mine’
Alternatively, possessive constructions provide a possessor specifica-
tion by a full NP. The two open possessive constructions available in
English can be recursively iterated (cf. my father’s wife’s mother’s car or
the car of the mother of the wife of my father). They can also be mixed:
the car of my father’s wife’s mother or the car of the mother of my father’s
wife. In all these variants, the logical order of the chain is:
car—mother—wife—father—(the speaker). Right possessives leave the
determiner position of the head noun open; it is therefore available for
definite or indefinite determination. This is not the case for left
possessives in English.
23
Glossing follows the Leipzig glossing rules, unless indicated otherwise (cf. http://
www.eva.mpg.de/lingua/resources/glossing-rules.php)
A ‘‘possessive chain’’ consists of a head (denoting the possessum),

a possessor specification and possibly recursively embedded further
possessor specifications. We will investigate the question what the
concept type is for a complete possessive chain, separately for the [U]
feature and the [R] feature.
All nonfinal elements of the chain are [+R] because they carry
a possessor specification. The last element in the chain must be [–R] in
order for the maximal chain to fulfil the absoluteness constraint. If the
[R] feature of any member of the chain is not semantically fulfilled, it is
coerced. Obviously, all initial sections of a chain, except the maximal

chain, are [+R], while the complete chain is [–R]. The compositional
mechanism is this: when a possessor specification is added, its referential
argument fills the possessor argument of the head (see the Appendix).
The [U] features of the NPs in a possessive chain can be chosen
independently; there is no a priori value for any of the elements. Let us
see what happens if a [6U] possessor NP is combined with a [6U]
CNP. Of the possible four constellations, three yield the following
picture:
(27) possessum possessor total
the father [+U] of the girl [+U] the father of the girl [+U]
a friend [–U] of the girl [+U] a friend of the girl [–U]
a friend [–U] of a girl [–U] a friend of a girl [–U]
The father of the girl is definite, a friend of the girl is indefinite and so,
of course, is a friend of a girl. These cases do not reveal, however,
whether the determination of the total phrase results from forming the
minimum of the two [U] values, or from applying the first determiner
to the whole. Does the first determiner apply to the CNP with or
without possessor specification? The critical case is a definite possessum
CNP with indefinite possessor specification, like the father of a girl. The
question can be decided by applying the consistency test: such NPs are
inconsistent and hence indefinite. Consider, for example, (28) in the
context of a teacher at a girls’ school meeting parents during office
hours; the sentence is not contradictory:
(28) The father of a girl left and the father of a girl came in.
As a consequence, the [+U] determination of a functional noun
with possessor specification cannot be related to the whole NP, but
only to the head noun itself. The definite article in ‘the CNPfunctional of
NP’ only indicates that the referent of the functional CNP is
determined uniquely relative to the possessor. Consequently, the value
of the [U] feature of an initial or complete possessive chain is the
minimum of all [U] values. Determination in the sense of (6) is

strictly local: it applies only to the CNP immediately determined. It
therefore makes sense to say that in a right possessive chain the first
Det + CNP component is an NP that is combined with a possessor
NP and so on, even if the first component is not a syntactic
constituent of the whole.24
The considerations on the [U] and [R] features of possessive chains
can be summarized as follows:
(29) determination of possessive chains
For any possessive chain, initial or maximal,

(i) the total [R] value is the minimum of the [R] values of the
members of the chain, where initial chains are [+R] and
maximal chains [–R] and
(ii) the total [U] value is the minimum of the [U] values of the
members of the chain.
The rules in (29) can be considered to hold also for [+R] nouns in
absolute use, where a possessor specification is coerced. If such NPs
carry definite reference, the implicit possessor must be [+U] in order to
admit explicit [+U] determination for the NP. Thus, ‘‘the father’’ is
always the father of somebody particular, given by an individual
concept available in the context. By contrast, [–U] a mother, calls for
possessor specification by a [–U] concept, for example, by existential
saturation. With a [+U] possessor as in a mother [of Liza], inherent
uniqueness of ‘mother’ would have to be abandoned to yield [–U] for
the whole; this would be a possible but pragmatically disfavoured
option.
It follows from the rules in (29) that the combination of
a [+U][+R], that is, functional, CNP with an [aU][bR] possessor
NP is always [aU][bR]. In this sense, functional concepts are
‘‘transparent’’.
(30) transparency of functional NPs
The type of a functional [+U][+R] NP with a possessor
specification is the same as the type of the possessor specification.
We will see in the next subsection that (29) may interfere with
a further constraint that applies to referential left possessives.
24
Compare Barker (2004) for a proposal for a compositional treatment of this aspect of possessor
constructions.
4.4 Determiner possessives v. attributive possessives

Lyons (1999: §3.3) distinguishes between ‘‘determiner genitives’’ and
‘‘adjectival genitives’’, for which I prefer to use the terms d(eterminer) and
a(ttributive) possessives, respectively. As a cross-linguistic generalization,
Lyons observes that d-possessives are in determiner position and impose
definiteness on the possessum noun, while a-possessives are open to
definite or indefinite determination (133 f.). Haspelmath (1999: 229 f.)
shows that the crucial condition is in fact that the possessor specification be
in complementary distribution with definiteness marking. An a-possessive
construction is illustrated for Hungarian in (26). English left possessives

are d-possessives; right possessives are a-possessives. For d-possessives, one
has to distinguish between referential and predicative use.
The definiteness test proves that English referential left possessives
are definite, if the possessor NP is definite. Possessive pronouns yield
the same result.
(31) a. conjunction test:
The girl’s brother is blind and the girl’s brother is a spy.
5 The girl’s brother is a blind spy.
b. negation test
The girl’s brother didn’t stay. 5 The girl’s brother left.
If the possessor NP is indefinite, the whole is indefinite. Replacing
the right possessive in (28) by a Saxon genitive yields the same result:
the sentence is not contradictory.
(32) A girl’s father left and a girl’s father came in.
These findings can be accounted for if we assume that, for some
reason, referential left possessives impose [+U] on the possessum CNP
and after this follow the rules in (29). This yields the following analyses
for the girl’s brother, a girl’s father and my book, respectively:
(33) possessum left possessive possessor
a. brother [–U][+R] ’s brother [+U][+R] the girl [+U][–R]
b. father [+U][+R] ’s father [+U][+R] a girl [–U][–R]
c. book [–U][–R] poss.pron. [+U][+R] 1SG [+U][–R]
total
a. the girl’s brother [+U][–R]
b. a girl’s father [–U][–R]
c. my book [+U][–R]
The case is different for d-possessives in predicative use (see
Holmberg 1993 for discussion and a syntactic analysis). Here, we can
apply a test similar to the consistency test. If the predicative NP is [+U],

that is, of logical type e, a copula construction with a definite subject
NP amounts to an equational statement. If the predicative NP is [–U],
that is, of type Æe,tæ, the same construction amounts to a simple
predication. The same predication may be true for two different
arguments, but not the same equational statement. Applying the test to
[–U] brother and [+U] father, we obtain the following contrast:
(34) a. Art is my brother and Bart is my brother.
b. Art is my father and Bart is my father.

If father in both occurrences is construed in the standard meaning,
(34b) is contradictory, while (34a) is not. Obviously, in the predicative
constructions left possessives do not impose [+U] on the possessum
CNP; the [U] and [R] features of the whole chain derive immediately
in accordance with (29). Note that the observation also applies to [–R]
nouns:
(35) a. THIS is my book and THIS is my book.
b. Béla is Hungary’s greatest composer and György is Hungary’s
greatest composer.
The [–U][–R] noun book is coerced to [–U][+R], which according
to (29) yields [–U][–R] with any possessor specification; (35a) is not
contradictory. The superlative greatest shifts sortal composer to an
individual concept [+U][–R]; the possessor specification imposes a shift
to a functional [+U][+R] concept (‘greatest composer of ’); with the
[+U][–R] possessor specification, the result is [+U][–R]; that is why
(35b) is contradictory.
The data for predicative d-possessives suggest that d-possessives as
such should not be considered as imposing [+U] on the possessum
CNP. The definiteness effect of referential d-possessives must be due to
the fact that the whole NP is referential and the possessor specification
blocks definiteness marking. Haspelmath (1999: 227) argues that the
phenomenon is ‘ECONOMICALLY MOTIVATED [. . .]: the definite article
can be omitted because possessed NPs have a very high chance of being
definite, for semantic and pragmatic reasons’. The effect can be
captured by the following constraint:
(36) definiteness constraint on referential determiner possessives
In a referential possessive chain, a CNP element with a possessor
specification in complementary distribution with definiteness
marking obtains [+U] determination.
The constraint has the effect that the feature combination [+U] is
imposed on the CNP together with [+R] by virtue of the possessive
construction itself. Hence, the CNP is coerced to a functional
concept.25 The rules in (29) work out correctly, if the [+U] feature
imposition according to the definiteness constraint applies first [cf.
example (33)].
The findings in sections 3 and 4 show that there is an asymmetry
among the four types of nouns and CNPs: there often are determiners
for simple indefinite, definite and personal possessive determinations.
Assuming the absoluteness constraint (22) and the definiteness

constraint (36), these three elementary types of determination are in
harmony with sortal [–U][–R], individual [+U][–R] and functional
[+U][+R] nouns, respectively. There is, however, no simple type of
determination in harmony with relational [–U][+R] nouns proper.26
5 CONGRUENT V. INCONGRUENT DETERMINATION

5.1 (In)congruence of CNP type and determination type
We have established in sections 3 and 4 that certain modes of
determination require certain types of CNPs. We will now introduce
the distinction between ‘‘congruent’’ and ‘‘incongruent’’ determina-
tion. One-step determination, for example, simple indefinite and
simple definite determination, is congruent if the determination does
not change the concept type of the CNP; multiple-step determination
is congruent if the CNP satisfies the input type requirement without
coercion.
Table 1 shows which modes of determination are congruent, or
incongruent, with which types of nouns; the results carry over to
CNPs. The table refers to nouns since the determinational properties
add to the characterization of the four noun types. Congruent
determinations are marked with U, incongruent ones with l. For the
sake of brevity ‘‘Indef.’’ is used for subsuming simple and unspecific
indefinite, free choice, negative and interrogative; ‘‘Plural’’ is short for
plural, numerical and quantitative [cf. (4) above].
Incongruent determination always involves coercion. Since the
coercion operations always consist in a change of conceptual type, they
will be referred to as ‘‘type shifts’’. Coercion causes higher semantic
25
For an elaborate discussion of alternative compositional accounts for determiner possessives, see
Dobrovie-Sorin (2004); also Jensen & Vikner (2004).
26
The asymmetry parallels the system of three main cognitive-ontological classes discussed in
Fraurud (1996), viz. individuals, functionals and instances (sortals).
[–U] Inherently unique [+U]
[–R] SORTAL NOUNS INDIVIDUAL NOUNS

stone book adjective water moon weather date Maria
U Indef., Plural, quantif., dem. l Indef., Plural, quantif., dem.
l singular definite U singular definite
U absolute U absolute
l relational, possessive l relational, possessive
[+R] inherently RELATIONAL NOUNS FUNCTIONAL NOUNS

relational sister leg part attribute father head age subject (gramm.)
U Indef., Plural, quantif., dem. l Indef., Plural, quantif., dem.

l singular definite U singular definite
l absolute l absolute
U relational, possessive U relational, possessive
Table 1 Four types of nouns and their respective congruence with modes of determination
complexity; consequently, the theory presented here predicts that

incongruent determination will be distinguished in several ways.
(37) general tendencies of incongruent determinations
a. Incongruent determinations are less frequent.
b. Incongruent determinations receive more salient expression,
such as strong v. weak marking, marking v. nonmarking,
additional morphemes.
c. Incongruent determinations require contextual support; they
often draw on nonlexical information.
Evidence for (37a) is presented in section 6.1. For (37b), there are
cross-linguistic data that will be discussed in section 6.2. Example (37c)
has been touched on in sections 2.3 and 2.4, and will be concretely
illustrated in the discussion of the example to follow.
In Löbner (1985), I introduced the terms ‘semantic’ and ‘pragmatic
definites’ for the distinction between congruent and incongruent
definites. In the present framework, I prefer to use the terms ‘semantic’
v. ‘pragmatic uniqueness’ for NPs that are [+U] for semantic reasons v.
being coerced by determination.
5.2 Example
An example will illustrate the interaction of CNP type, determination
and context. Referential NPs are in italics; incongruent determination
is in addition marked by bold type.
(38) There she read about a man who had been discovered after
spending two thousand years in a peat bog in the Netherlands. A
photograph appeared on one page. The man’s brow was furrowed,
but there was a serene expression to his face.27
The interpretation of she cannot be reconstructed here because the
preceding text is not included. It is used anaphorically, referring to the
main character of the story, a woman named Sandy. Thus, the lexical
concept ‘she’, essentially ‘ix (person(x) ^ female(x))’, is substantially
enriched by the information compiled about the character in the
preceding part of the story. There in the first sentence refers to a certain

place in a book with the title Mysteries of the Past, which Sandy has
picked up and opened to this place.
The noun man, ‘male adult human’, is sortal, that is, [–U][–R]; the
indefinite article with the first occurrence indicates congruent
determination; the absolute use is congruent, too. Equally congruent
is the cardinal, absolute determination of sortal year in two thousand
years, and the singular indefinite, absolute determination of sortal peat
bog. The Netherlands is a proper name, [+U][–R], in harmony with
definite, absolute use.
By the end of the first sentence, an individual concept is established
of the content:
(39) ‘man that Sandy read about there in that book with the title
Mysteries of the Past who had been discovered after spending two
thousand years in a peat bog in the Netherlands’
It uniquely determines a particular individual. Its inherent
uniqueness is due to the way in which the referent is linked to the
situation as it is described: in the given context, only one such man can
be meant.
The relevant reading of photograph is [–U][+R] ‘photographic
picture of ’. Singular indefinite determination is congruent, while the
incongruent absolute use coerces saturation of the possessor argument.
According to the type rules for relational nouns in (29), the possessor
specification might be [–U] or [+U] because both would result in
indefinite determination for the NP. The obvious choice is a [+U]
possessor specification retrieved anaphorically from the given context:
the possessor is construed as the unique referent of the concept in (39).
Application of ‘photograph of ’ to this possessor yields the concept:
27
From Carver, Raymond (2003), ‘Preservation’. In Cathedral. Vintage. London. 32.
(40) ‘photograph of the man whom Sandy read about there in that
book with the title Mysteries of the Past who had been discovered
after spending two thousand years in a peat bog in the
Netherlands’
Similarly, the relational noun page is related to the book, yielding
the sortal concept:
(41) ‘page of the book with the title Mysteries of the Past where Sandy
read about the man who had been discovered after spending two
thousand years in a peat bog in the Netherlands’

Numerical indefinite determination of one page is congruent with
the lexical type [–U][+R], the [–R] feature of absolute determination is
satisfied by retrieving the possessor argument from context. The two
NP tokens, a photograph and one page, are instances of indefinite
associative anaphora: relational nouns in absolute use where the
possessor required is construed as something uniquely determined in
the preceding discourse.
The man in the third sentence carries incongruent [+U] de-
termination, while the absolute use is congruent. A classical case of
anaphora, the [+U] requirement is met by enriching the sortal concept
‘man’ with all the information accumulated for the referent of a man
two sentences back, yielding the individual concept:
(42) ‘man whom Sandy read about there in that book with the title
Mysteries of the Past who had been discovered after spending two
thousand years in a peat bog in the Netherlands and a photograph
of whom appeared on one page of that book’
Brow is a body-part term of type [+U][+R]; its determination here
is congruent. The type of the total possessive chain is [+U][–R], in
accordance with (29) and the transparency of functional concepts, since
the man is [+U][–R]. The same holds for his face. Expression, too, in the
sense relevant here, is a functional noun, [+U][+R]. Its determination is
incongruent [–U][–R]. Crucial here is the modification with the
adjective serene; the adjective implicitly contrasts the actual value of the
functional concept ‘expression’ with possible alternative values. The
concept ‘expression’ is thereby shifted to a concept for the possible
values of the function. This concept is sortal. In general, functional
nouns can be shifted to sortal concepts that denote the values that the
functions may take. This was observed by Guarino (1992) who states
that these terms can systematically be used for ‘‘attributes’’ (functional
concepts) and the associated ‘‘concepts’’ (sortal concepts in our
terminology). I would therefore like to dub this shift from functional to

sortal concepts ‘‘Guarino shift’’. The Guarino shift derelativizes the
given functional concept.28
5.3 Levels and types of shifts

5.3.1 Levels of shifts Type shifts coerced by incongruent determi-
nation can be located at three levels of processing:
(43) levels of type shifts
level 0 semantic interpretation

level 1 adjustment at type level: application of general lexical
meaning shifts
level 2 adjustment at token level: application of shifts that draw
on particular contextual information
Level 0 is trivial for nouns as such; many polysemous nouns have
meaning variants of different concept types. Thus, the first option of
meeting type requirements is choosing a lexical meaning of appropriate
type. For example, for my children one would preferably choose the
relational meaning of child rather than the sortal.
Forming CNPs from nouns, shifts at level 0 can be accomplished by
explicit modification. These shifts will not be discussed here; they
include modifiers such as adjectival superlatives, ordinal numerals,
‘‘last’’ and ‘‘next’’, individual noun appositions ([the] year 2003) and
‘‘establishing relative clauses’’ in the terminology of Hawkins (1978:
138 ff.) ([the] woman he went out with last night), which all shift to [+U].
Another level 0 shift is possessor specification.29
For many semantic classes of nouns, systematic meaning shifts are
available, such as metonymy, metaphor, mass-to-count shifts and
others. These belong to level 1; Bierwisch (1982) introduced them as
‘‘conceptual shifts’’. They are available for larger classes of semantically
similar nouns; they produce meaning variation that is to a certain
degree systematic, even across languages, since the shifts represent
general cognitive patterns of concept modification. Some of these shifts
change the conceptual type. The shifts with artefact terms mentioned
in section 2 are level 1.
28
Note that even though the NP a serene expression is used here for expressing a sortal concept,
£
the embedding verbal locative-possessive construction £NP was to his face exploits the lexical
relation inherent to the functional noun expression.
29
Some compositional level 0 shifts, in particular those triggered by favourite, former and new have
been discussed in detail (Partee & Borschev 2002; Vikner & Jensen 2002).
Level 1 shifts are components of a dynamic lexicon. They enable

a division of labour between storing lexical meanings and deriving
further meanings by general procedures. If the same shift occurs
repeatedly with a lexical item, storing its result may become more
economic than repeated shifting. As a result, there is no clear borderline
between polysemy and level 1 shifts. And this is what the literature on
polysemy, as well as actual lexicography, reflects.30
Adjustment at level 2 is what we encountered with the anaphoric
NPs in (38): the shifts are accomplished by enriching the semantic
concept with information retrieved from the context. Level 2 shifts

trigger the merge of level 1 output with extralinguistic information. If
an utterance is interpreted, this merge inevitably takes place; one
mechanism is the choice of referents and the consequent merge of the
linguistic description with the knowledge independently available. The
analysis of the man in (38) gives one example of the process. Shifts at
level 2 thus serve another division of labour: the interaction of
linguistic information with extralinguistic knowledge. As the in-
formation exploited for a level 2 shift is contingent on the given
context of utterance, the results will not get stored in the lexicon.
5.3.2 Types of shifts Of the various level 1 and level 2 shifts
applicable, I want to mention only the most frequent ones. Most of
them concern the triangle of sortal, functional and individual concepts.
Sortal to individual: The most frequent shifts of this type are the
level 2 shifts involved in direct anaphoric and deictic uses of definites.
Sortal to functional: When sortal CNPs are used with possessive
determination (dubbed ‘‘extrinsic’’ possession by Partee 1997), a shift
applies either at level 1 or level 2. Jensen and Vikner (Vikner & Jensen
2002; Jensen & Vikner 2004) discuss ‘‘lexical’’ (i.e. level 1) shifts based
on the qualia information in the possessum concept; these apply
frequently to CNPs that denote artefacts exclusively used by one user,
such as car, flat and toothbrush (cf. section 2.4). Apart from these uses,
possessive constructions with [–U] heads may receive ‘‘pragmatic’’
(Jensen & Vikner 2004) or ‘‘free’’ (Storto 2004) interpretations such as
for your book, ‘book you mentioned’, ‘book you stole’ and ‘book you’d
like to borrow’. These are the result of level 2 shifts that exploit
particular contextual information. One subclass is constituted by
‘‘control’’ interpretations of possessives, including cases of ordinary
ownership (Storto 2004).
30
Bierwisch (1982) would not count systematic meaning variation due to level 1 shifts as
polysemy, while Tyler & Evans (2003) do; dictionaries usually list some products of level 1 shifts as
meaning variants, but not generally so.
Sortal to relational: These shifts are triggered if a sortal CNP is

used with relational indefinite determination, for example, in the plural
with right possessor specification (books of . . .): the [–U][–R] sortal
noun is shifted to [–U][+R]. There are obvious parallels to the previous
type: level 1 shifts for artefact concepts that specify a certain type of use
but such that more than one such artefact can be used by the same user.
Analogously, there are level 2 shifts of this type for cases where the
relationship between the objects and a possessor is to be retrieved from
context.
Functional to individual: These shifts are most naturally achieved

by combining a functional CNP with an individual concept possessor
argument. As a level 2 shift, this occurs with definite associative
anaphors: the functional concept with absolute determination is related
to a possessor introduced before. The door in (44) is an example:
(44) (in a hotel) She walked into the foyer and sat in a chair near the
door.31
The [–R] requirement for the door is met by choosing the foyer as
the possessor of the door; the [+U] requirement is met by contextual
restriction to the one door relevant here, the entrance door.
Functional to sortal: Functional CNPs can be shifted to sortal
concepts by any [–U] type saturation of the possessor argument. One
option is level 1 existential saturation of the possessor argument, as in
she is a late mother. The Guarino shift is another level 1 instance of an
FC to SC shift.
Relational to sortal: Saturation of the possessor argument of
a relational noun leads to a sortal concept both with [–U] and [+U]
possessor concepts. If the possessor is retrieved from the given context,
the shift is level 2.
5.4 The system of conceptual types and determinations

Determinations constitute operations in a space of conceptual types.
Figure 1 displays the system for the one-step modes of determination:
(simple) indefinite, singular definite, rel(ational), poss(essor specifica-
tion) with a [+U][–R] unique possessor, the combination rel(ational)-
def(inite) for d-possessives and abs(olute).
More generally, Figure 1 supplies a scheme for type shifts among the
four basic terms. The type shifting determination operations are just
grammaticalized instances of type shifts.
31
From Enright, Anne (2007) The Gathering. Black Cat. New York. 13. The quotation is
shortened.
indefinite sg definite
sortal sg definite indiv.

indefinite
rel rel-def rel/rel-def
poss poss
abs abs
sg definite
relat. funct.

indefinite
indefinite sg definite
Figure 1 System of types of nominal concepts and types of determination.
5.5 Determination and redundancy

Despite the semantic redundancy of congruent determination, there
are several motivations for marking it. First, explicit congruent
determination may serve disambiguation of the CNP and its type.
Second, determiners may carry functions beyond the mere specifica-
tion of concept type: case marking, deictic differentiation, quantity
specification and so on. Third, if there is explicit incongruent
determination, the use of congruent determination leads to a uniform
syntax: there is an analogical pressure towards the inclusion of
congruent determination (see Lyons 1999: 322 ff.).
As to incongruent determination, its motivation is obvious. It
serves as an explicit trigger for the type shifts to be applied. It might
be asked, conversely, how languages can do without marking
incongruent determination. The answer is twofold. First, there are
those other means of indicating the type of reference, for example,
information structure or nonnominal possessive constructions.
Second, the basic concept types are not only types of noun and
CNP meanings but more generally the basic types of concepts that
our cognition works with anyway. Nonquantificational, referential
full NPs are either construed as sortal concepts or as individual
concepts. These are the two fundamental types of mental representa-
tions, for categories and individual objects, respectively. Thus,
choosing a determination type means choosing between the two
most general concept types available.
5.6 Reflection on the analysis of determination

The theory presented here—let us call it the concept type theory
of determination (CTD)—is different from other approaches, in
particular, from other approaches to definiteness, in two regards. It is
a theory of determination in general, not only of definites or
possessives, and it is claimed to apply to all languages. It predicts
phenomena such as cross-linguistic splits in determination marking
(see section 6.2). Apart from its wider scope, CTD offers an analysis
not in terms of truth-conditions but in terms of the cognitive level of
interpretation.

In its underlying understanding of linguistic communication and the
role of context, CTD is close to relevance theory (Sperber & Wilson
1995). Wilson (1991) discusses cases of definites in a similar vein as in
section 5.2. Rather than starting from the assumption of a common
ground, CTD puts emphasis on the way in which utterances are
formulated: choosing a referential NP with definite determination, the
speaker declares the referent of the NP as being uniquely determined in
the relevant context by the chosen description. By the respective
contextualization cues (Gumperz 1982), the recipients receive an
indication as to what the relevant context is like. The final
interpretation of an NP token can then be modelled according to
relevance theory. To give a simple example, Raymond Carver, in using
the man in the third sentence in (38), conveys to the reader by the use of
the definite article that the context is such that the referent is a ‘‘man’’
uniquely determined. Given that the context essentially is the actual
interpretation of the story up to this point, the interpretation of the NP
is predictable as the most relevant one.
For elaboration, CTD requires a theory of concepts that is able to
model the level of lexical meanings and NP token interpretations. A
promising candidate for a theory of concepts is currently being
developed on the basis of Barsalou’s (1992) theory of frames (cf.
Petersen 2007). The theory of concepts-as-frames, that is, recursive
attribute-value structures with constraints, is able to model the four
types of concepts discussed here as frames of different architecture.
Concluding this section, I will briefly address the relation of CTD
to other theoretical approaches to definiteness and possession.
Possession: The approach taken to relational and possessive uses is
in accordance with Partee & Borschev (2002, and later work). The idea
that the relational use of [–R] nouns is enabled by a variety of type shifts
is widespread in formal semantics (see Barker forthcoming for an
overview).
Definiteness: The [–U] account of indefinites appears to be

uncontroversial. Except for [+R] nouns and proper names and
pronouns, nouns are usually considered sortal, of type Æe,tæ. For
a recent treatment, see Landman (2004).
Some accounts of definiteness have raised the question as to whether
there is a uniform notion of definiteness (Lyons 1999; Schwarz 2009;
Abbott 2010). CTD provides a positive answer. The more recent
discussion of definiteness focuses on the questions whether definiteness
is essentially a matter of familiarity or of uniqueness, or neither;
whether uniqueness, if it is actually given, is presupposed or

conversationally implicated; and whether uniqueness is ‘‘semantic’’
(in the sense assumed by Russell) or the result of pragmatic ‘‘domain
restriction’’.32
In CTD, uniqueness is not objective uniqueness in the given
universe of discourse, as is expressed for ‘‘N’’ by Russell’s formula:
(45) dx((N(x) ^ P(x)) ^ "y(N(y)/y¼x))
Roberts (2003) argues that this conception of uniqueness does not
apply to uses of definite NPs such as in an example discussed in Heim
(1982: 28), in a context, where more than one wine glass broke:
(46) A wine glass broke last night. The glass was very expensive.
According to CTD, there are two sorts of unique NP concepts.
First, there are those which are unique at level 0 or level 1: lexically
unique CNPs, CNPs with level 0 shifting modifiers that turn them
[+U] or CNPs that are unique due to a general context-independent
level 1 type shift. These are the cases that I call ‘‘semantically unique’’.
The second type of unique concepts is the utterance meanings of [–U]
CNPs with incongruent definite determination, for example, ana-
phoric NPs such as the glass in (46). In these cases, a level 2 shift
modifies the given [–U] concept to yield a [+U] concept in the relevant
context. We have seen in (39)–(42) what these modifications are like.
They capture the ‘‘pragmatic enrichment’’ that is made responsible in
the literature for the uniqueness of ‘‘familiar’’ definites. CTD, however,
rejects the idea suggested by Heim (forthcoming: §1.5) that the level 2
shift from [–U] to [+U] can be attributed to ‘a covert predicate which
in the semantic computation conjoins with the overt NP’. Conjunction
32
See Birner & Ward (1994) and Roberts (2003) on uniqueness and familiarity and the nature of
uniqueness, and Stanley & Szabó (2000) on domain restriction.
of a [–U] CNP meaning with another [–U] (type Æe,tæ) predicate

cannot yield a [+U] NP meaning however narrow the restriction be.33
According to CTD, uniqueness of an NP, more generally the
concept type of an NP, has the status of a presupposition. The concept
type is a nondetachable part of the token meaning of the NP. It is not
part of the content of the predication that the NP specifies an argument
of, and hence neutral with respect to the truth-value of the predication.
Definite NPs presuppose unique reference; indefinite NPs presuppose
the possibility that reference is not unique, a condition weaker than
nonuniqueness; and relational NPs presuppose a relation of the referent

to some possessor.
As for languages with no explicit marking of either simple definites
or simple indefinites, CTD disagrees with the ‘‘competition’’ analysis of
(in)definiteness (Heim forthcoming: §1.7). Heim locates definites and
indefinites on a scale of competing alternatives where definites entail
the corresponding indefinites (the dog barked would entail a dog barked).
Heim concludes that, in languages without (in)definiteness marking,
simple NPs are just indefinite. This view is rejected here since CTD
considers indefinite and definite reference conceptually incompatible;
definites do not entail the corresponding indefinites because they have
contrary presuppositions. Other authors favour a type shift approach
congenial with CTD for implicitly definite NPs with [–U] heads, for
example, Bittner & Hale (1995), and Chierchia (1998) who elaborates
on Partee’s (1987) type shifting approach.
As a semantic theory, CTD is not the framework to provide an
account of the conceptual mechanisms at work in the case of level 2 shifts.
A theory of these mechanisms would have to answer questions such as
follows: What qualifies an NP token in a discourse as an antecedent for
subsequent anaphors? What are the requirements on nonlinguistic
context for deictic and other nonanaphoric uses of definites? A further
fundamental question is the role of salience in text interpretation, an issue
that has received much attention in analyses of anaphora resolution and
definite NPs in general (cf. e.g. Gundel et al. 2001).
6 EVIDENCE
In this section, evidence for CTD is provided from three perspectives:
statistical investigations into the relative frequencies of congruent v.
incongruent determination, typological data on (in)congruency and
33
Compare note 5 above as to the crucial difference between the stronger notion of uniqueness
relevant here, and accidental uniqueness, which might be achieved by additional predicative
restriction on a sortal concept.
historical evidence showing that concept types are relevant for the
development of definiteness marking.
6.1 Statistical evidence

Fraurud (1990) investigates a Swedish corpus of nonfictional written
texts with 3877 NP tokens. Distinguishing determination modes of
NPs, she counted which occurrences were first mention (referring to
a discourse-new referent) and which ones subsequent-mention. One
subclass of NPs is DefNP: definite NPs with a head noun with the
definite suffix –en/-et. DefNP excludes all definite NPs without definite

suffix: those with a definite article only, NPs with demonstrative
determiners and determiner possessives, as well as proper names and
personal pronouns. Fraurud counted a proportion of 60.9% of first-
mention NPs within DefNP.34 In nonfictional written texts, there will
be no situational definites such as in ‘Open the door, please!’ Since
incongruent definites are either situational or anaphoric (Hawkins
1978), first-mention definites must be due to CNPs that are inherently
unique [+U]. Among the subsequent-mention uses of definite NPs
were another 20.8% with a (first-mention) DefNP as antecedent; these,
too, must be [+U], yielding a [+U] proportion within DefNP of more
than 80%. Of determiner possessives, 85% are first mention. Since
uniqueness of this subclass may be due to the definiteness constraint
(36), the proportion of [+U] CNPs among them cannot be
determined. The ratio of DefNP to determiner possessives is about
4.3. Two further classes of NPs can be added to definite uses of [+U]
nouns: personal pronouns and proper names, which together out-
number the DefNP tokens in the corpus. It follows that the proportion
of congruent definites in the corpus is around 90%.
In a corpus study of Wall Street Journal articles, Vieira (1998)
determines a similar result for English definite descriptions (i.e. NPs
with definite article): a proportion of about 48% percent discourse-new
uses. Again, these must be [+U]; adding definites that are coreferent
with first-mention definites, cases of associative anaphora, proper
names and personal pronouns will result in a clear majority of
congruent definites.
The two studies may not be fully representative. Narrative texts
would probably contain more anaphoric NPs; but then, most
anaphoric NPs are [+U] third-person pronouns. I would expect that
34
Fraurud defines ‘‘first-mention’’ definites not just as definites with a head noun used for the first
time in the text but as definite NPs that are not coreferential with any NP occurring earlier in the
text. Thus, first-mention definites are not anaphoric.
the proportion of incongruent anaphoric NPs will be somewhat higher

but will not outweigh congruent uses. In oral texts or written texts
containing dialogues, there will be a minor proportion of deictic
definites.
Nissim (2004) investigates whether [+R] and [–R] nouns preferably
occur with its or with the in the British National Corpus. According to
her study, there are nouns that occur significantly more often with its
than with the, and vice versa; among the former is a proportion of 90%
[+R] nouns; among the latter, the proportion is 8%. The same
tendency is observed by Jensen & Vikner (2004). They present a count

of 2333 English prenominal possessives in fictional and nonfictional
text; for more than 71%, the possessor relation is ‘‘inherent’’, that is, the
head noun is [+R].
6.2 Typological evidence
The theory predicts tendencies of grammatical asymmetries between
congruent and incongruent determination. What one would expect
are phenomena such as
implicit congruent determination v. explicit incongruent de-
termination, with morphemes corresponding to the type shifts
coerced by incongruent determination and
strong marking for incongruent determination v. weak marking for
congruent determination.
6.2.1 Incongruent definites Evidence for the difference between

congruent and incongruent definites is very strong in Germanic
languages. In many German dialects, a strong definite article marks
incongruent definites, while a weak article is used for congruent
definites, including proper names.35 The strong and weak forms of
definite articles in Ripuarian (Rhineland) dialects are as follows; the
weak forms exhibit reduced vowels:
(47) sg masc sg fem sg neuter plural
strong dä [de] die [di] dat [dat] die [di]
weak dr [d ] de [d*] et [*t] de [d*]
We would have (48a) as an announcement of the carnival parade in
Düsseldorf, where the ‘‘zoch’’ (German Zug, ‘parade’) is an institution
and therefore zoch, in this reading, an individual noun. In a railroad
35
See Hartmann (1982) and Himmelmann (1997: 54 ff.) for Rhineland dialects; for Swiss
German, see Studler (2004). Gerland & Horn (2010) present further discussion of these phenomena,
in the framework of CTD.
station context, waiting for a train (German Zug, dialect ‘zoch’) to

come, where ‘zoch’ would represent just a sortal notion, one would use
the incongruent strong article da¨.
(48) a. Dr Zoch kütt
DEFARTWEAK parade comes
b. Dä Zoch kütt.
DEFARTSTRONG train comes
Dutch has a similar distinction between strong die (SG.M/F, P) and dat
(SG.N) v. weak de and het, respectively. Die and dat are restricted to

incongruent definiteness, while the weak forms can be used for all
definites. The strong forms are demonstrative determiners, but stripped
of specific deictic distinctions, unlike the proximal demonstrative deze.
Albert Ortmann (p.c.) gives the following quote; de warme hand can be
considered a functional NP,36 as opposed to the anaphoric expression
die man with a sortal head.
(49) En zo voelde Anton het ook: de [weak] warme hand van die
[strong] man, twintig jaar ouder dan hij, die hem meenam.37
‘And so Anton felt it, too: the warm hand of the man, twenty
years older than him, who took him with [him].’
In German, there are contractions of prepositions with definite
articles that result in reductions of the articles to the final consonantal
case morphemes. The contractions are used only when definite
determination is congruent. Restricted by phonotactic rules, possible
contractions are for example im ¼ in dem, am ¼ an dem, vom ¼ von dem,
ans ¼ an das, vors ¼ vor das, zur ¼ zu der.38
Swedish marks congruent definiteness with the suffixes -(e)t or
-(e)n; for example, universitetet Stockholm (universitet ‘university’)
would be ‘the university Stockholm’, where ‘university Stockholm’ is
an individual concept. Incongruent definites are marked with an article
den/det, with or without additional suffixal marking on the noun. The
White House (individual concept) is vita huset ‘white house’ (hus
‘house’). Det vita huset with additional article is ‘the white house’
referring to some white house (sortal concept), and definiteness
therefore incongruent.39
36
Strictly speaking, ‘hand’ is relational; here, an RN to FN shift intervenes in a ‘‘configurational’’
use of hand (cf. section 3.5).
37
From Mulisch, Harry (1982) De anslaag. De Bezige Bij. Amsterdam. 50.
38
For recent in-depth work on this split in standard German, see Schwarz (2009).
39
Quoted from Schroeder (2006: 564) who gives a more detailed account of definiteness marking
in Scandinavian and other European languages.
In the case of Fering, a North Frisian dialect spoken in the German

island of Föhr, Ebert (1970) reports a weak definite ‘‘A-article’’ (a, at)
and a strong definite ‘‘D-article’’ (di, det); again, the strong article is used
for incongruent definites, and the weak one for congruent definites.
The type of split with distinguished marking for incongruent
definites is not restricted to Germanic languages. Lyons (1999: 53 f.)
mentions languages that have a special definite article (different from
demonstratives) for marking anaphoric reference: Ewe and Hidatsa
(Lyons 1999: 158). Lakhota is discussed as a language with a special
definite article k’ for anaphoric use and a general definite article k

(Lyons 1999: 54).
The asymmetries of definiteness marking can be placed on a scale
that reaches from pragmatic to semantic uniqueness, that is, from [–U]
to [+U] CNPs/NPs, with a variety of subtypes (Figure 2). Deictic and
anaphoric uses of definite NPs are at the [–U] end of the scale; they
roughly coincide with the range of demonstrative determiners. The
[+U] end of the scale forms an implicative scale with [+U] CNPs <
proper names < personal pronouns. In the middle of the scale are [–U]
CNPs with ‘‘establishing relative clause’’ (Hawkins 1978), such as
student who wanted to see you and functional nouns with unique
possessor, including definite associative anaphors. These middle types
sometimes go with strong definiteness marking, probably owing to the
contextual information contained within the complex CNP. In other
cases, they receive weak marking, apparently due to the fact that the
whole NP is [+U].
Explicit definiteness marking usually cuts off at some point on the
scale. In standard written German and in English, proper names are
used without a definite article. In colloquial German or Modern
Greek, proper names do take definite articles. Very rarely, even personal
pronouns receive definiteness marking. Maori personal pronouns in
nontopical use and proper names receive a definite article a (Bauer
pragmatic deictic with sortal CNPs

uniqueness < anaphoric with sortal CNPs
< sortal CNPs with establishing relative clause,
functional CNPs with explicit definite possessor
< definite associative anaphors
< individual CNPs
< proper names
< 3 personal pronouns semantic
< 1, 2 personal pronouns uniqueness
Figure 2 Scale of uniqueness.

1993: 108 ff.). One reason why personal pronouns are rarely used with
definiteness marking might be that [–U] uses of them are very
infrequent. Hungarian, in addition to marking definiteness with
a definite article a, has ‘‘objective conjugation’’ for transitive verbs with
definite objects (Havas 2004). It is used with all types of definite NPs
except first- and second-person pronouns: this indication of definite-
ness cuts off after third-person pronouns.
Languages that develop definite articles from demonstratives start
out with [–U] cases, the home base of demonstratives, passing through
a stage where only incongruent definiteness is marked. When

a language starts to use an originally demonstrative determiner for
[+U] CNPs, a definite determiner is achieved.40 This is presently to be
observed with Upper Sorbian, a Slavic variety spoken in Eastern
Germany with a long record of German language contact. Dwo´rnisˇcó
‘station’ in (50a) is an individual noun. It therefore has no article,
although it carries definite reference. Breu quotes the example as an
instance of what he calls ‘‘situative Unika’’; these are individual nouns
in CTD terminology, that is, terms for (local) institutions. The train
referred to by cá ‘train’ in the second sentence in (50b) is not referred to
as a local institution but as uniquely identified in the given context of
discourse; cá here is anaphoric. In the first sentence of (50b), cáom, the
instrumental case of cá, can be taken either as an individual noun,
similar to the use of train in the English translation, or as a sortal noun
with indefinite determination.
(50) Upper Sorbian (Western Slavic, Breu 2004: 30, 39)
a. Tame jo dwórnišćo. [IN]
there is station
‘there’s the station’
b. Mĕrko jo s ća-om šijo1.
M. is with train-INS come.PTCP.
‘Mirko came by train.
Tón ća [SN] jo dźewećich Kamencu bó1.
DEFART train ...
‘The train arrived in Kamenz at nine o’clock.’
The data show that there are many languages exhibiting splits in the
marking of definites which can be located consistently on the scale in
Figure 2. A critical point is the one between incongruent uses of
definites, with [–U] CNPs, and congruent uses, with [+U] CNPs.
40
For a comprehensive typological study of the grammaticalization of definite articles from
demonstratives, see Himmelmann (1997).
Thus, the data confirm the significance of the distinction between

inherently unique and inherently nonunique CNPs and nouns.
6.2.2 Incongruent indefinites Incongruent indefinites are [+U] CNPs

with indefinite determination. One would expect them to exhibit
a stronger marking of indefiniteness. One relevant observation is the
fact that proper names in indefinite use are often accompanied by
additional modal adjectives [cf. a certain XY, ein gewisser XY (German),
een zekere XY (Dutch)].
A second observation is the behaviour of indefinite determiners in

quantificational use. These take semantically [+U] CNPs that refer to
the domain of quantification as a whole into NPs that refer to
a nonunique part of the whole [cf. (9)]. Consider the contrast in (51):
(51) a. What’s that noise?—Oh, there are some kids playing in the
street.
b. What are the kids doing?—SOME kids are playing in the street.
In (51a), the indefinite NP in the ‘sm’ reading is nonquantifica-
tional, that is, not partitive. The plural kids is [–U] and the indefinite
determiner some is congruent. Some is semantically almost redundant,
not only for its congruency but also for the vague quantity predication;
it is syntactically not obligatory; it does not carry stress and can be
phonetically reduced to [sm]. By contrast, given the context in (51b),
some kids will receive a partitive quantificational interpretation,
equivalent to ‘some of the kids’. In this interpretation, [–U] kids is
shifted to [+U] [the] kids; some then shifts the [+U] concept for the
whole set of kids to a [–U] concept for some part of it. In this case, some
is incongruent; as such it is not redundant. It cannot be omitted. It
carries stress and cannot be phonetically reduced.41
6.2.3 Incongruent relatives Many languages grammatically distinguish

different types of possessive constructions.42 The distinction is usually
referred to as ‘‘(in)alienability splits’’.43 As a general tendency,
inalienable possession receives less salient expression than alienable
possession. Inalienability essentially coincides with relationality: in-
alienable possession constructions occur with [+R] CNPs and alienable
41
See Löbner (1987: §3) for further discussion. Diesing (1992) offers a syntactic account of the
two uses of indefinites.
42
References and data in this section and the next are based on talks held by Albert Ortmann as
well as on Ortmann (2002: 52 f.). For further discussion of incongruent relatives and absolutes in the
framework of CTD, see Gerland & Horn (2010).
43
For general information, see Nichols (1992: 116 ff.); Seiler (1983) speaks of ‘inherent’ v.
‘established’ possession, a terminology essentially in accordance with the approach taken here.
constructions with [–R] CNPs.44 Languages differ, however, as to

which [+R] CNPs enter the inalienable possessive construction
(Chapell & McGregor 1996: 8 f.).
The rich data reported in the literature support the type shift
approach. For example, [+R] nouns in Lakhota are immediately
prefixed with a possessor marking, while [–R] nouns require an extra
REL (relativizer) morpheme that enables the noun to take a possessor
prefix:
(52) Lakhota (Siouan, Robert D. Van Valin p.c.)
a. ni-nági ki

2SG-spirit DEF
‘your spirit’
b. ni-thá-wowapi ki
2SG-REL-book DEF
‘your book’
The [+R] noun na´gi ‘spirit’ takes the congruent prefixal possessor
specification ni- that shifts it to [–R]. [–R] wowapi ‘book’ is shifted to
[+R] by the relativizing prefix REL tha`-, in order to take the possessor
specification ni- that shifts it back to [–R]. The relativizer tha´ is used iff
there is a type shift from [–R] to [+R]. Analogous data are reported by
Lehmann (1988: 52, 56) on Yucatec, and by Thompson (1996: 654 ff.)
for Koyukon.
(53) Yucatec (Mayan, Mexico, Lehmann 1998: 52; 56)
a. in la’ak
1SG.POSS friend
‘my friend’
b. le h-o’
DEF house-DISTAL
‘the house’
c. in nah-il
1SG.POSS house-REL
‘my house’
In Koyukon, inalienably possessed nouns, that is, [+R] nouns in our
theory, can only occur with a possessor affix, unlike [–R] nouns
(Thompson 1996: 651). If the latter are used with a possessor
specification, they require a relativizing REL suffix –e’ (with certain
morphophonological effects on the noun), or the prefix k’e, here
glossed as K’E because it also serves as a derelativizer [cf. (58) below]:
44
Barker (1995: 67) arrives at the same conclusion.
(54) Koyukon (Athabaskan, Thompson 1996: 655, 667, 659)

a. hutl (sortal) be-ghudl-e’
sled 3SG-sled-REL
‘his/her sled’
b. gguh (sortal) se-k’e-gguh
rabbit 1SG-K’E-rabbit
‘my rabbit’
c. *to’ (functional) be-to’
father 3SG-father
‘his/her father’

In other languages, alienable possession receives a syntactic
treatment, whereas inalienable possession is expressed by morpholog-
ical means:
(55) Georgian (South Caucasian, Khizanishvili 2006: 12 f.)
a. deda-čem-i
mother-1SG.POSS.-NOM
‘my mother’
b. čem-i c_ ign-i
1SG-POSS book-NOM
‘my book’
Ewe combines [+R] CNPs immediately with a possessor specifi-
cation, unlike [–R] CNPs:
(56) Ewe (Niger-Congo, Ameka 1996: 790 ff.)
a. inalienable
)evı́-á-wó t bé
child-DEF-PL grandfather
‘the children’s grandfather’
b. alienable
kofı́ Ué awu
Kofi REL GARMENT
‘Kofi’s garment’
The data show that, in these languages, either incongruent
possession is marked with a corresponding relativizing morpheme
(Lakhota, Yucatec, Koyukon, Ewe) or the connection between the
noun and the possessor expression is less close (Georgian) and therefore
more salient.
6.2.4 Incongruent absolutes Incongruent absolute use should tend to

be more saliently marked than congruent use. This kind of distinction
is to be expected only with languages that distinguish [+R] CNPs

grammatically. In Mam, according to England (1983), [+R] nouns in
absolute use carry an extra derelativizing morpheme.
(57) Mam (Mayan, Guatemala, England 1983: 69)
a. relative
n-yaa’¼ya
1.SG.ERG-grandmother¼NON3
‘my grandmother’
b. absolute
yaa-b’aj

grandmother-DEREL
‘grandmother’
Parallel data are attested by Lehmann (1998: 70 ff.) for the Mayan
language Yucatec. Koyukon is another such case. The prefix k’e-
occurring as relativizer in (54b) can also function as a derelativizer in
incongruent absolutes with nonhuman possessors. In (58a), the prefix
serves existential saturation of the possessor argument; (58c) shows that
it also serves both functions in the same expression.
(58) Koyukon (Thompson 1996: 666 f.)
a. *kkaa’ [+R] k’e-kkaa’
foot K’E-foot
‘a/the foot [of some animal]’
b. gguh kkaa’ ne-kkaa’
rabbit foot 2SG-foot
‘rabbit’s foot’ ‘your foot’
c. ne-k’e-gguh kkaa’ ne-k’e-k’e-kkaa
2SG-K’E-rabbit foot 2SG-K’E-K’E-foot
‘your rabbit foot’ ‘your [animal’s] foot’
The expressions in (58a and b) involve a single switch from [+R] to
[–R]. The expressions in (58c) involve a series of three switches [+R]
to [–R] to [+R] to [–R]: in ne-k’e-gguh kkaa’, [+R] kkaa’ is switched to
[–R] by the inalienable possessor specification gguh; k’e switches back
to [+R] to make alienable possession possible, which is then achieved
by adding ne- to yield [–R]. The three steps are analogous with ne-k’e-
k’e-kkaa: first derelativizer k’e expressing existential saturation switches
to [–R]; next, the result is taken to [+R] by relativizing k’e; the ultimate
possessor specification (ne- 2SG) reduces [+R] to [–R].
Obviously, the prefix k’e indicates just a [6R] switch. The double
use of k’e is not redundant; direct affixation with ne- would have
yielded ‘your [own] foot’. English your foot is ambiguous between the
reading ‘your [own] foot’ and the reading ‘your foot [of someone or
some animal]’. The latter reading would be the result of the same shifts
as are involved with ne-k’e-k’e-kkaa’, the only difference being that
Koyukon makes these shifts explicit, while English does not. Other
languages with a strict distinction of relational and nonrelational nouns
use just a relativizer on a relational noun in these cases. According to
Haude (2006), in Movima, an isolate language in Bolivia, as [ART] bo:sa
is ‘my arm’ (the absence of a possessor affix on bo:sa is understood as
first-person singular possession, p. 233). The noun bo:sa also means
‘sleeve’; it can take the affix -e which ‘[i]n nouns seem[s] to indicate

that an entity is possessed which is either normally unpossessed or
possessed by a different possessor than usual’ (p. 239). Hence, this suffix
serves as a relativizer on [–R] nouns or as a re-relativizer on [+R]
nouns, that is, the combined effect of first derelativizing and then
relativizing. It can be analysed as a shifter from [–R] to [+R], which,
applied to [+R] nouns, coerces derelativization before shifting to [+R].
With this affix we get as [ART] bosa:-ne ‘my sleeve’ (p. 242). In (59),
the shifts are depicted from right to left in order to preserve the word
order of the examples:
(59) shift 3 shift 2 shift 1
poss.1 rel. derel.
[–R] ) [+R] ) [–R] ) [+R]
Koyukon ne k’e k’e kkaa’
Movima impl. ‘my’ ne impl. bo:sa
English my impl. impl. foot
6.3 Evidence from the evolution of definiteness marking

The theory predicts the well-known fact that demonstrative deter-
miners are a possible source of definite articles. Demonstratives shift
[–U] CNPs to [+U] NPs, that is, unique reference. In the course of the
development into definiteness marking, they undergo two extensions
of usage: first, detachment from their specific deictic function, for
example, the neutralization of a proximal/distal distinction—this step
extends their use along the uniqueness scale from deictic to anaphoric
use. Then, the deictic distinction and contrastivity is dropped; this in
turn makes the [–U] input requirement void; it is dropped, and thereby
the application of the determiner is extended to [+U] nouns. Typically,
there are remnants of semantic definiteness with no explicit definiteness
marking, for example, personal pronouns, proper names and institution
terms. After extension to [+U] nouns, we may eventually observe
a third stage, the reduction to weak forms of definiteness marking with

[+U] nouns.45
The approach predicts a second possible source of definiteness
marking, namely possessive pronouns. These take nouns of whatever
type to [+U][–R] reference [cf. (4.3)]. Such cases do, in fact, exist.
According to Schroeder (2006: 587), in Uralic languages such as Komi,
Nganasan and Selkup, third- and second-person possessive suffixes in
nonrelational use mark [+U][–R] nouns (‘‘nouns with unique
referents’’) and sortal nouns as definite.
(60) Nganasan (Uralic, Tereščenko 1979: 95):

nonrelational possessive marking with individual nouns
m*u- du
earth- POSS3SG
‘the earth’
(61) Selkup (Uralic): nonrelational possessive marking with anaphoric
sortal nouns (Nikolaeva 2003)
tyt qanyqqyn anty totta, anty-ty apyk*:l e:sa.
river bank.on boat stands boat-POSS3SG oar.without is
‘A boat stands on the riverbank, the boat has no oar.’
Himmelmann (1997: §5.3.2) mentions Indonesian and Amharic as
further languages exhibiting this phenomenon. Fraurud (2001)
discusses the phenomenon extensively, including also Turkish, Yucatec
and other languages. She posits several stages of the development of
possessive marking into definiteness marking. The development starts
from congruent use with [+R][+U] nouns, for example, in associative
anaphoric use. Since the possessor specified by possessive affixes is
definite, the result is definite. The use of possessive suffixes is
eventually extended to individual nouns, and to those sortal nouns for
which unique reference is grounded in contextual givenness. The
semantic steps are similar to the ones of demonstratives developing
into definite articles: first, the specific content, that is, the specification
of a first-, second- or third-person possessor, is neutralized; with it, the
function of relating the operand to a ‘possessor’ extinguishes. What
remains is a general indexical quality and [+U] determination. Once
the suffixes are stripped of their possessive quality (in certain uses), they
can be applied regularly to [–R] nouns for the indication of definite
reference.
45
See Gelderen (2007) for more on the historical cycle of definiteness marking in Germanic
languages.
6.4 Conclusions
The statistical, typological and historical data corroborate the concept
theory of determination types and related type shifts. The clearest
evidence is provided by cases where certain suffixes serve the direct and
exclusive indication of the type shifts postulated, such as relativizing or
derelativizing affixes. Other evidence lies in the fact that definiteness
splits and (in)alienability splits align with the distinction between [–U]
and [+U], and [–R] and [+R]. I know of no splits that contradict this
alignment. This article offers only a few illustrative examples, but the
evidence in the literature appears to be robust. More comprehensive

studies are currently in progress.
APPENDIX: POSSESSIVE COMPOSITION

Table A1 gives the composition rules for the combination of [+R] nouns
with a possessor NP of each of the four types. S, R, i and f represent
constants corresponding to sortal, relational, individual and functional
nouns, of logical type Æe,tæ, Æe,Æe,tææ, e and Æe,eæ, respectively. Due to the
syntactic design of predicate logic, individual and functional constants do
not exhibit a referential argument. Thus, the natural language feature [+U]
corresponds with the fact that these constants are not of a predicate type.
For predicate constants, the first argument r/r’ is the referential one; the
second p/p’ the possessor. The compositional mechanism is basically
functional application and relational or functional composition. Indefinite
NPs are analysed in terms of an g operator, as it offers the possibility of
direct argument insertion for indefinite possessor arguments.
Table A2 gives the types of head-with-possessor constructions for
natural language. Note that functional heads with possessor inherit the
type from the possessor (‘‘transparency’’).
Head Type Possessor Type Head with possessor Type
krkp R(r, p) Æe,Æe,tææ kr’ S(r’) Æe,tæ kr R(r, gr’ S(r’)) Æe,tæ
krkp R(r, p) Æe,Æe,tææ kr’kp’ R’(r’,p’) Æe,Æe,tææ krkp’ R(r, gr’ R’(r’, p’)) Æe,Æe,tææ
krkp R(r, p) Æe,Æe,tææ i e kr R(r, i) Æe,tæ
krkp R(r, p) Æe,Æe,tææ kp’ f ’(p’) Æe,eæ krkp’ R(r, f ’(p’)) Æe,Æe,tææ
kp f(p) Æe,eæ kr’ S(r’) Æe,tæ kr r¼f(gr’ S(r’)) Æe,tæ
kp f(p)) Æe,eæ kr’kp’ S(r’,p’) Æe,Æe,tææ krkp’ r¼f(gr’ S(r’, p’)) Æe,Æe,tææ
kp f(p) Æe,eæ i e f(i) e
kp f(p) Æe,eæ kp’ f ’(p’) Æe,eæ kp’ f(f ’(p’)) Æe,eæ
Table A1 Logical composition for combinations of head and possessor specification

Head Type Possessor Type Head with possessor Type
RC [–U][+R] SC [–U][–R] SC [–U][–R]

RC [–U][+R] RC [–U][+R] RC [–U][+R]
RC [–U][+R] IC [+U][–R] SC [–U][–R]
RC [–U][+R] FC [+U][+R] RC [–U][+R]
FC [+U][+R] SC [–U][–R] SC [–U][–R]
FC [+U][+R] RC [–U][+R] RC [–U][+R]
FC [+U][+R] IC [+U][–R] IC [+U][–R]
FC [+U][+R] FC [+U][+R] FC [+U][+R]

Table A2 Type composition of for combinations of head and possessor specification
Acknowledgements
The research was supported by the German Science Foundation (DGF) grant FOR 600,
Research Unit ‘Functional Concepts and Frames’. I am indebted to the fellow researchers
in the unit, in particular to Albert Ortmann, Doris Gerland and Christian Horn.
SEBASTIAN LÖBNER
Institut für Sprache und Information
Heinrich-Heine-Universita¨t Düsseldorf
Universita¨tsstr. 1
40225 Düsseldorf
Germany
e-mail: loebner@phil.uni-duesseldorf.de
REFERENCES
Abbott, Barbara. (2010), Reference. Ox- and B.H. Partee (eds.), Possessives and
ford University Press. Oxford. beyond: semantics and syntax. University
Ameka, Felix. (1996), Body parts in Ewe of Massachusetts Occasional Papers in
grammar. In H. Chappell and W. Linguistics 29. GLSA Publ. Amherst.
McGregor (eds.), The Grammar of 89–113.
Inalienability: a typological perspective Barker, Chris (to appear, 2011), Posses-
on body part terms and the part whole sives and relational nouns. In Claudia
relation. Mouton de Gruyter. Berlin. Maienborn, Klaus von Heusinger, and
783–840. Paul Portner (eds.), Semantics: An
Barker, Chris. (1995), Possessive Descrip- International Handbook of Natural Lan-
tions. CSLI. Stanford, CA. guage Meaning. Volume II. De
Barker, Chris. (1998), ‘Partitives, double Gruyter. Berlin, New York. Ch. 45.
genitives and anti-uniqueness’. Natural Barsalou, Lawrence W. (1992), ‘Frames,
Language & Linguistic Theory concepts, and conceptual fields’. In A.
16:679–717. Lehrer and E.F. Kittay (eds.), Frames,
Barker, Chris. (2004), ‘Possessive weak Fields, and Contrasts: New Essays in
definites’. In J.Y. Kim, Y.A. Lander, Semantic and Lexical Organization.
Lawrence Erlbaum. Hillsdale, NJ. Language Structures Online. Max Planck

21–74. Digital Library. München. Chapter 41.
Barwise, Jon & Robin Cooper. (1981), Available online at http://wals.info/
‘Generalized quantifiers and natural feature/41. Accessed on 2010-07-29.
language’. Linguistics and Philosophy Dobrovie-Sorin, Carmen. (2004), ‘Gen-
4:159–219. itives and determiners’ In J.Y. Kim,
Bauer, Winifred. (1993), Maori. Routledge. Y.A. Lander, and B.H. Partee (eds.),
London. Possessives and beyond: semantics and
syntax. University of Massachusetts
Bierwisch, Manfred. (1982), ‘Formal and
Occasional Papers in Linguistics 29.
lexical semantics’. Linguistische Berichte
GLSA Publ. Amherst. 115–132.
80:3–17.
Dryer, Matthew S. (2008a), ‘Definite

Birner, Betty & Gregory Ward. (1994),
articles’. In M. Haspelmath, M. Dryer,
‘Uniqueness, familiarity, and the def-
D. Gil, and B. Comrie (eds.), The
inite article in English’. Proceedings of
World Atlas of Language Structures
the Annual Meeting of the Berkeley
Online. Max Planck Digital Library.
Linguistic Society 93–102.
München. Chapter 37. Available on-
Bittner, Maria & Kenneth Hale. (1995), line at http://wals.info/feature/37.
‘Remarks on definiteness in Warlpiri’. Accessed on 2010-07-29.
In E. Bach, E. Jelinek, A. Kratzer, and Dryer, Matthew S. (2008b), ‘Indefinite
B.H. Partee (eds.), Quantification in articles’. In M. Haspelmath, M. Dryer,
Natural Languages. Kluwer. Dor- D. Gil, and B. Comrie (eds.), The
drecht. 81–105. World Atlas of Language Structures
Breu, Walter. (2004), ‘Der definite Online. Max Planck Digital Library.
Artikel in der obersorbischen München. Chapter 38. Available on-
Umgangssprache’. In M. Krause and line at http://wals.info/feature/38.
C. Sappok (eds.), Slavistische Linguis- Accessed on 2010-07-29.
tik 2002. Sagner. München. 9–57. Ebert, Karen H. (1970), Referenz,
Chapell, Hilary & William McGregor. Sprechsituation und die bestimmten Arti-
(1996), ‘Prolegomena to a theory of kel in einem nordfriesischen Dialekt
inalienability’. In H. Chappell and W. [Fering]. Ph.d. thesis, University of
McGregor (eds.), The Grammar of Kiel, Nordfriisk Institut, Bredstedt.
Inalienability: a typological perspective England, Nora C. (1983), A Grammar of
on body part terms and the part whole Mam, a Mayan Language. University
relation. Mouton de Gruyter. Berlin. of Texas Press. Austin.
3–30.
Fraurud, Kari. (1990), ‘Definiteness and
Chierchia, Gennaro. (1995), Dynamics of the processing of noun phrases in
Meaning. University of Chicago Press. natural discourse’. Journal of Semantics
Chicago. 7:395–433.
Chierchia, Gennaro. (1998), ‘Reference Fraurud, Kari. (1996), ‘Cognitive ontol-
to kinds across languages’. Natural ogy and NP form’. In T. Fretheim and
Language Semantics 6:339–405. J. K. Gundel (eds.), Reference and
Diesing, Molly. (1992), Indefinites. MIT Referent Accessibility. Benjamins.
Press. Cambridge, MA. Amsterdam. 65–87.
Diessel, Holger. (2008), ‘Distance Fraurud, Kari. (2001), ‘Possessives with
Contrasts in Demonstratives’. In M. extensive use. A source of definite
Haspelmath, M. Dryer, D. Gil, and B. articles?’ In I. Baron, M. Herslund,
Comrie (eds.), The World Atlas of and F. Sørensen (eds.), Dimensions of
Possession. Benjamins. Amsterdam. motivation in noun phrase syntax’.

243–67. Language 75:227–43.
Gelderen, & Elly van. (2007), ‘The Haspelmath, Martin. (2008), ‘Occur-
definiteness cycle in Germanic’. Jour- rence of Nominal Plurality’. In M.
nal of Germanic Linguistics Haspelmath, M. Dryer, D. Gil, and B.
19:275–308. Comrie (eds.), The World Atlas of
Gerland, Doris & Christian Horn. Language Structures Online. Max
(2010), ‘Referential properties of Planck Digital Library. München.
nouns across languages’. In D.-H. Chapter 34. Available online at
Choi, J.-S. Hong, H.-K. Kang, Y.-S. http://wals.info/feature/34. Accessed
Kang, K.-H. Kim, K.-A. Kim, J.-Y. on 2010-07-29.
Yoon, S.-H. Rhee, and J.-S. Wu

Haude, Katarina. (2006), A Grammar of
(eds.), Universal Grammar and Individ- Movima. Ph.D. thesis, University of
ual Languages. Proceedings of SICoL Nijmegen. (http://webdoc.ubn.ru.nl/
2010. Korea University Press. Seoul. mono/h/haude_k/gramofmo.pdf).
Groenendijk, Jeroen & Martin Stokhof. Havas, Ferenc. (2004), ‘Objective con-
(1991), ‘Dynamic predicate logic’. jugation and medialisation’. Acta Lin-
Linguistics and Philosophy 14:39–100. guistica Hungarica 51:95–141.
Guarino, Nicola. (1992), ‘Concepts, Hawkins, John A. (1978), Definiteness
attributes, and arbitrary relations: and Indefiniteness. Croom Helm.
some linguistic and ontological crite- London.
ria for structuring knowledge bases’.
Data & Knowledge Engineering Heim, Irene. (1982), The Semantics of
8:249–61. Definite and Indefinite Noun Phrases.
Gumperz, J. J. (1982), Discourse Strate- Ph.D. thesis, University of Massachusetts,
gies. Cambridge University Press. Amherst, MA.
Cambridge. Heim, Irene. (to appear, 2011), ‘Defi-
Gundel, Jeanette K., Nancy Hedberg & niteness’. In C. Maienborn, K. von
Ron Zacharsky. (2001), ‘Definite Heusinger, and P. Portner (eds.),
descriptions and cognitive status in Semantics: An International Handbook
English: why accommodation is un- of Natural Language Meaning. Volume
necessary’. English Language and Lin- II. De Gruyter. Berlin, New York.
guistics 5:273–95. Ch. 41.
Hackl, Martin. (2009), ‘On the grammar Himmelmann, Nikolaus P. (1997),
and processing of proportional quan- Deiktikon, Artikel, Nominalphrase.
tifiers: most versus more than half ’. Zur Emergenz syntaktischer Struktur.
Natural Language Semantics 17:63–98. Niemeyer. Tübingen.
Hartmann, Dietrich. (1982), ‘Deixis and Holmberg, Anders. (1993), ‘On the
anaphora in German dialects: the structure of predicative NP’. Studia
semantics and pragmatics of two Linguistica 47:126–38.
definite articles in dialectal varieties‘. Jensen, Per Anker & Carl Vikner.
In J. Weissenborn and W. Klein (eds.), (2004), ‘The English pre-nominal
Here and There. Cross-linguistic Studies genitive and lexical semantics’. In J.Y.
on Deixis and Demonstration. Benjamins. Kim, Y.A. Lander, and B.H. Partee
Amsterdam. 187–207. (eds.), Possessives and beyond: semantics
Haspelmath, Martin. (1999), ‘Article- and syntax. University of Massachu-
possessor complementarity: economic setts.
Kamp, Hans & Uwe Reyle. (1993), From Nikolaeva, Irina. (2003), ‘Possessive af-
Discourse to Logic. Kluwer. Dordrecht. fixes as markers of information struc-
Khizanishvili, Tamar. (2006), ‘Posses- turing: Evidence from Uralic’. In P.
sion im Georgischen’. In T. Stolz and Suihkonen and B. Comrie (eds.), In-
C. Stroh (eds.), Possession, Quan- ternational Symposium on Deictic Systems
titative Typologie und Semiotik. and Quantification in Languages spoken in
Brockmeyer. Bochum. 1–76. Europe and North and Central Asia.
Collection of papers. Udmurt State
Krifka, Manfred, Francis Jeffry Pelletier, University. Izhevsk. Max Planck In-
Gregory N. Carlson, Alice ter Meu- stitute of Evolutionary Anthropology.
len, Godehard Link & Gennaro Leipzig. 130–145.
Chierchia. (1995), ‘Genericity: an
Nissim, Malvina. (2004), ‘Lexical in-

introduction’. In G. N. Carlson and
F. J. Pelletier (eds.), The Generic Book. formation and choice of determiners’.
University of Chicago Press. In J.Y. Kim, Y.A. Lander, and B.H.
Chicago. 1–124. Partee (eds.), Possessives and beyond:
semantics and syntax. University of
Landman, Fred. (2004), Indefinites and Massachusetts Occasional Papers in
the Type of Sets. Blackwell. Oxford. Linguistics 29. GLSA Publ. Amherst.
Lehmann, Christian. (1998), Possession 133–52.
in Yucatec Maya. Lincom Europa. Ortmann, Albert. (2002), Kategorien des
München. Nomens: Schnittstellen und Ökonomie.
Löbner, Sebastian. (1985), ‘Definites’. Niemeyer. Tübingen.
Journal of Semantics 4:279–326. Partee, Barbara H. (1987), ‘Noun phrase
Löbner, Sebastian. (1987), ‘Natural lan- interpretation and type-shifting prin-
guage and generalized quantifier the- ciples’. In J. Groenendijk, M. Stokhof,
ory’. In P. Gärdenfors (ed.), Generalized and D. de Jongh (eds.), Studies in
Quantifiers: Linguistic and Logical Ap- Discourse Representation Theory and the
proaches. Reidel. Dordrecht. 181–201. Theory of Generalized Quantifiers.
Löbner, Sebastian. (1990), Wahr neben Foris. Dordrecht. 115–43.
Falsch. Duale Operatoren als die Quan- Partee, Barbara H. (1997), ‘Genitives.
toren natürlicher Sprache. Niemeyer. A case study. Appendix to Theo M. V.
Tübingen. Janssen, Compositionality’. In J. F. A.
K. van Benthem and A. ter Meulen
Löbner, Sebastian. (2000), ‘Polarity in
(eds.), Handbook of Logic and Linguis-
natural language: predication, quan-
tics. Elsevier. Amsterdam. 464–70.
tification and negation in particular
and characterizing sentences’. Lin- Partee, Barbara H & Vladimir Borschev.
guistics and Philosophy 23:213–308. (2002), ‘Integrating lexical and formal
semantics: genitives, relational nouns,
Lyons, Christopher. (1999), Definite- and type-shifting’. In R. Cooper and
ness. Cambridge University Press. T. Gamkrelidze (eds.), Proceedings of
Cambridge. the 2nd Tbilisi Symposium on Lan-
Matushansky, Ora. (2008), ‘On the guage, Logic and Computation. Tbilisi
linguistic complexity of proper State University. Tbilisi. 229–41.
names’. Linguistic and Philosophy Petersen, Wiebke. (2007), ‘Represen-
31:573–627. tation of concepts as frames’. In J.
Nichols, Johanna. (1992), Linguistic Di- Skilters, F. Toccafondi, and G.
versity in Space and Time. University Stemberger (eds.), Complex Cogni-
of Chicago Press. Chicago. tion and Qualitative Science. The
Baltic International Yearbook of Cog- Studler, Rebekka. (2004), ‘Voller und

nition, Logic and Communication, reduzierter Artikel in der
vol. 2. University of Latvia. Riga schweizerdeutschen DP’. In S. Bračič,
151–70. D. Čuden, S. Podgoršek, and V.
Poesio, Massimo. (1994), ‘Weak defi- Pogačnik (eds.), Linguistische Studien
nites’. In Proceedings of the 4th Confer- im Europäischen Jahr der Sprachen.
ence on Semantics and Linguistic Theory. Akten des 36. Linguistischen Kolloquiums
Cornell University Press. Ithaca, NY. in Ljubljana. Lang. Frankfurt. 625–35.
Pustejovsky, James. (1995), The Generative Tereščenko, Natalja. (1979), Nganasan-
Lexicon. MIT Press. Cambridge, MA. skij jazyk [The Nganasan Language].
Nauka. St. Petersburg.
Roberts, Craige. (2003), ‘Uniqueness in
Thompson, Chad. (1996), ‘On the

definite noun phrases’. Linguistics and
Philosophy 26:287–350. grammar of body parts in Koyukon
Athabaskan’. In H. Chappell and W.
Schroeder, Christoph. (2006), ‘Articles McGregor (eds.), The Grammar of
and article systems in some areas of Inalienability: a typological perspective
Europe’. In G. Bernini and M.L. on body part terms and the part whole
Schwartz (eds.), Pragmatic Organiza- relation. Mouton de Gruyter. Berlin.
tion of Discourse in the Languages of 651–78.
Europe. Mouton de Gruyter. Berlin.
545–611. Tyler, Andrea & Vyvyan Evans. (2003), The
Semantics of English Prepositions. Cam-
Schwarz, Florian. (2009), Two Types bridge University Press. Cambridge.
of Definites in Natural Language. De-
partment of Linguistics, University of Vieira, Renata. (1998), Definite Descrip-
Massachusetts. Amherst, MA. tion Processing in Unrestricted Text.
Ph.D. thesis, University of Edinburgh,
Seiler, Hansjakob. (1983), Possession as an Edinburgh.
Operational Dimension of Language.
Vikner, Carl & Per Anker Jensen. (2002),
Narr. Tübingen.
‘A semantic analysis of the English
Sperber, Dan & Deirdre Wilson. (1995), genitive. Interaction of lexical and
Relevance. Communication and Cogni- formal semantics’. Studia Linguistica
tion, 2nd edition. Blackwell. Oxford. 56:191–226.
Stanley, Jason & Zoltán Gendler Szabó. Wilson, Deirdre. (1991), ‘Reference and
(2000), ‘On quantifier domain restric- relevance’. University College London
tion’. Mind & Language 15:219–61. Working Papers in Linguistics
Storto, Gianluca. (2004), ‘Possessives in 4:167–91.
Context’. In J.Y. Kim, Y.A. Lander,
and B.H. Partee (eds.), Possessives and
beyond: semantics and syntax. Univer- First version received: 12.04.2010
sity of Massachusetts Occasional Pa- Second version received: 17.09.2010
pers in Linguistics 29. GLSA Publ. Third version received: 28.10.2010
Amherst. 59–86. Accepted: 31.10.2010

Lob Ner 2011

Uploaded by

Copyright:

Available Formats

Lob Ner 2011

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lob Ner 2011

Uploaded by

Copyright:

Available Formats

Journal of Semantics, 28, 2011: 279–333

Concept Types and Determination

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

2 LOGICAL TYPES OF NOUNS: BASIC DISTINCTIONS

Sortal nouns are unary predicate terms, of type Æe,tæ. They

complements such as possessor specifications. Sortal nouns include

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

Relational nouns are binary predicate terms, of type Æe,Æe,tææ.6

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

2.2 Types are speciﬁed in the lexicon

There are only two ways to technically evade fixing a lexical

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

Individual nouns other than proper names and pronouns can be

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

with indefinite or quantificational determination, they undergo a type

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

2.4 Relationality is a semantic property of relational and functional

If a noun meaning provides for an additional argument, this is not

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

2.5 Meaning and context

fixed and it takes on utterance meaning. The utterance meaning of I am

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

3 TYPES OF NOUNS AND TYPES OF DETERMINATION (1):

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

plural definite no yes

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

Using the O-characteristics of modes of determinations, we can set up

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

arbitrary individual cases and so on. The total series of mappings

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

3.3 Explicit and implicit (in)deﬁniteness

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

3.4 A criterion for deﬁnite determination

(10) a. subject of existential there constructions

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

definition, let £n + p represent the syntactic combination of an NP

£n + p is false iff £n + (: p) is true

If £n + p is true, necessarily £n + (:p) is false.

Or: £n + p and £n + (:p) cannot both be true.

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

utterance both £n + p and £n + (:p) may be true. They are also

incomplete: £n + p and £n + (:p) may both be false.

£n + (p1 ^ p2) is true iff £n + p1 and £n + p2 are both

If £n + (p1 ^ p2) is true, then necessarily £n + p1 and

£n + p2 are both true.

If £n + p1 and £n + p2 are both true, then necessarily

£n + (p1 ^ p2) is true.

(14) conjunction test

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

that demonstrative NPs violate the consistency condition. Note

3.5 Deﬁnite determination and unique reference

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

awake’ is not a contradiction. NPs of the form £die meisten N ‘most N’

CNP constructions regularly receive the same interpretation as their

English counterparts £most CNPplur/CNPmass . Likewise, the half-

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015

4 TYPES OF NOUNS AND TYPES OF DETERMINATION (2):

Downloaded from http://jos.oxfordjournals.org/ at EPFL Lausanne on March 23, 2015