13
Mereology
Lucas Champollion and Manfred Krifka
Mereology
513
13.1 Notions of parthood
Mereology (from the Greek μέρος, ‘part’) is the study of parthood, an
important notion in ontology, natural-language semantics and cognitive
sciences. Examples include rice as part of paella, a thumb as part of
a hand, John Lennon as part of the Beatles, the Adriatic Sea as part
of the Mediterranean, the battle of Issos as part of Alexander’s war
against Persia, November 1989 as part of the year 1989, or hood as
part of the word parthood. These examples illustrate that the relation
may hold between masses, objects, objects and groups, locations, events,
times, and abstract entities (see Simons, 1987; Winston et al., 1987, for
classification of parthood relations).
One important distinction is between what we may call structured
and unstructured parthood. Structured parts are cognitively salient
parts of the whole. The thumb is a structured part of a hand: it is one of
the parts of the hand, besides the fingers and the palm. The hand is itself
a cognitively salient, integrated whole and not just a random collection
of parts. Similarly, hood is a structured part of the word parthood : it is
a suffix with a meaning of its own, attached to the stem part, and contributing to the overall meaning of the word. The notion of unstructured
parthood, in contrast, can be illustrated with the water in the left half of
a cup as part of the water in the entire cup, or the letter sequence artho
as part of the sequence parthood. Unstructured parts need not be cognitively salient parts of the whole, but may slice up the whole in arbitrary
ways. (For discussion of related philosophical issues, see Simons, 1987;
Moltmann, 1997; Varzi, 2010. For linguistic applications of a structured
notion of parthood to the semantics of mass nouns, plurals, and events,
see Moltmann, 1997, 1998, 2005. A summary and critical discussion of
Moltmann, 1997 is provided in Pianesi, 2002).
The two part relations have different structural properties. Structured
parthood is not transitive, as has often been noted (e.g., Cruse, 1979;
Moltmann, 1997). The thumb is a structured part of the hand, and the
hand is a structured part of the arm, but the thumb is not a structured
part of the arm. Linguistic evidence for this comes from the fact that we
can say the thumb of this hand, or a hand without thumb, in contrast to
*the thumb of this arm, or *an arm without thumb. In contrast, unstructured parthood is transitive: e.g., the sequence rth is part of artho, and
artho is part of parthood, hence rth is part of parthood. For unstructured
parts, we can use part as a mass noun (x is part of y); for structured
parts, as a count noun (x is a part of y). For unstructured parts of
514
rigid entities that hang together, we may also use the word piece (cf.
Cruse, 1986). As a consequence of transitivity, unstructured parts may
overlap, in contrast to structured parts. For example, art and tho are
both unstructured parts of parthood. (For discussion on the transitivity
of parthood, see Varzi, 2006 and references therein.)
In this article, we will focus on unstructured part relations. This
is not to deny that structured part relations are of great importance
in natural-language semantics. There are two such structured part relations: meronymy (the relation between meronyms like hand and
holonyms like arm), and hyponymy (the relation between hyponyms
like poodle and hyperonyms like dog, in English expressed by words like
kind or type). These relations structure large parts of the lexicons of natural languages, and they have certain general properties (cf., e.g., Kay,
1975 for taxonomies, based on the hyponymy relation, and Cruse, 1986
for a textbook treatment). In this, we make no claim about the semantics of the English word part and related expressions like a part. (See
Moltmann, 1997, 1998 for discussion of the semantics of the expressions
part and a part.)
We will take the parthood relation to be reflexive, which is at variance with how English part is used. We will write ≤ for parthood, and
distinguish it from irreflexive < , which will be called proper parthood.
If we consider parthood for entities extended in space or time, we
may restrict the notion of part to contiguous parts, e.g. strings like
art or ood, or we may allow for non-contiguous parts, such as discontinuous strings like pa—od or p–r–h–o, as parts of parthood. Contiguity
is a topological notion, and from topology we also can derive notions
like borders or interior parts of objects. The formal notion of parthood typically is understood in a way as to allow for non-contiguous
parts. Even structured parts can be non-contiguous; for example, the
fingers are considered part of the hand, and circumfixes and infixes lead
to non-contiguous morphological parts of words. (For combinations of
mereology and topology, see Smith, 1996; Casati and Varzi, 1999; Forrest, 2010).
Parthood relations have played an important role in philosophy, especially ontology (cf. Simons, 1987; Varzi, 2010, for overviews). Yet the
formal study of parthood relations took off in the foundations of mathematics, as an alternative to set theory. Set theory is based on two
parthood relations, element ∈ and subset ⊆, and a corresponding typetheoretic distinction (if α ⊆ β, then α and β must be sets; if α ∈ β, then
Mereology
515
just β is required to be a set). Consequently, set theory distinguishes between singleton sets and their elements (a 6= {a}), and assumes an empty
set. Mereology was proposed by Leśniewski (1916), see Simons (2011),
and Leonard and Goodman (1940) as a simpler alternative without such
assumptions. It does not distinguish between elementhood and subsethood, and it does not assume abstract entities like sets. Consequently,
it does not distinguish between singleton sets and their elements, and
does not entertain the notion of the empty set.
This article will be concerned specifically with some of the applications
that mereology has found in natural-language semantics. After a section
on essential formal properties of the (unstructured) parthood relations
and their axiomatization we will deal with linguistic applications in the
nominal domain, in the expression of measurement functions and in the
verbal domain.
13.2 Axiomatizations of Mereology
The notion of parthood has been captured with various axioms. The
most congenial approach to mereology is to take the notion of part
as central; this gives us an order-theoretic perspective. There is another
approach that starts with the notion of sum; this is the lattice-theoretic
perspective.
13.2.1 The order-theoretic perspective
We present an axiomatization of the parthood relation known as classical
extensional mereology (CEM). This system is the one most commonly
used in natural language semantics, and comes closest to a standard
in mereology.1 However, there are alternatives: see for example Rescher
(1955) for an alternative axiomatization of the parthood relation. We
take reflexive, unstructured parthood (≤) as the primitive relation and
formulate axioms that impose constraints on it. The following axioms
constrain parthood to be a partial order:
Axiom (Reflexivity)
∀x(x ≤ x)
1
(13.1)
The discussion in this section is based on Champollion (2010) and on the
excellent surveys of mereology in Simons (1987); Casati and Varzi (1999); Varzi
(2010).
516
Axiom (Transitivity)
∀x∀y∀z((x ≤ y ∧ y ≤ z) → x ≤ z)
(13.2)
Axiom (Antisymmetry)
∀x∀y((x ≤ y ∧ y ≤ x) → x = y)
(13.3)
We often want to talk of objects that share parts. For this purpose,
we introduce the auxiliary concept of overlap, which we write as ◦,
following Simons (1987) and Link (1998).
Definition (Overlap)
x ◦ y =def ∃z(z ≤ x ∧ z ≤ y)
(13.4)
With the part relation, we can define the notion of sum, also called
fusion. Sums formally capture the pretheoretical concept of collection
– that which you get when you put several parts together. In naturallanguage semantics, two common applications of sums are conjoined
terms and definite descriptions. For example, Link (1983) proposes to
represent the denotation of the conjoined term John and Mary as the
sum of the individual John with the individual Mary.2 Sharvy (1980)
suggests representing the denotation of a definite description like the
water as the sum of all water. We come back to definite descriptions in
Section 13.3.2.
As discussed in Hovda (2009), there are several ways to define sum,
and while they are equivalent given the CEM axiom system, they are
not logically equivalent. This means that when putting together the
axioms, it is important to choose the definition carefully to make sure
that the axioms that use it lead to CEM as intended. The definitions
below are metalanguage statements about shorthand expansions, that is,
they indicate what formula “sum(x, P )” is a shorthand for. Intuitively,
this shorthand stands for “x is a sum of (the things in) P ”.3
2
3
See Winter (2001) for a dissenting view. For Winter, and is always translated as
generalized conjunction, in this case as conjunction of two generalized
quantifiers, and the equivalent of sum formation enters the system through a
collectivization operator on the resulting quantifier.
If we formulate our theory in first-order logic, whichever one of these two
definitions we pick must be interpreted as an axiom schema, or a list of axioms in
each of which P (y) is instantiated by an arbitrary first-order predicate that may
contain y as a free variable. If we formulate our theory in second-order logic, we
can instead understand the definition we pick as quantifying over a predicate
variable P , which may be interpreted as ranging over sets, as suggested by our
paraphrases. The set interpretation makes the theory more powerful (Pontow
and Schubert, 2006; Varzi, 2010). The main issue here is that there are only
countably many formulas, but uncountably many sets. In the following, we will
talk about sets rather than axiom schemas.
Mereology
517
The first definition of sum is due to Tarski (1929).
Definition (Sum (1))
sum(x, P ) =def ∀y(P (y) → y ≤ x)∧
∀z(z ≤ x → ∃z ′ (P (z ′ ) ∧ z ◦ z ′ ))
(13.5)
A sum of a set P is a thing that consists of everything in P and whose
parts each overlap with something in P .
The second definition appears in Simons (1987, p. 37) and Casati and
Varzi (1999, p. 46):
Definition (Sum (2))
sum(x, P ) =def ∀y(y ◦ x ↔ ∃z(y ◦ z ∧ P (z)))
(13.6)
A sum of a set P is a thing such that everything which overlaps with it
also overlaps with something in P , and vice versa.
The following fact follows from either definition of sum:
Fact
∀x(sum(x, {x}))
(13.7)
The proof is immediate if sum is defined as in (13.6). If sum is defined
as in (13.5), we can rewrite (13.7) as x ≤ x ∧ ∀z(z ≤ x → z ◦ x). The first
conjunct follows from reflexivity (13.1); the second follows from the fact
that parthood is a special case of overlap, which in turn follows from
reflexivity.
Different mereology systems disagree on what kinds of collections have
a sum, and whether it is possible for one and the same collection to have
more than one sum. In CEM, sums are unique, therefore two things composed of the same parts are identical. This is expressed by the following
axiom:
Axiom (Uniqueness of sums)
∀P (P 6= ∅ → ∃!z sum(z, P ))
(13.8)
The operators in (13.9) and (13.10) give us a way to refer explicitly
to the sum of two things, and to the sum of an arbitrary set. In the
following, ιxP (x) is only defined if P holds of exactly one individual,
and when defined, it denotes that individual.
518
Definition (Binary sum)
x ⊕ y =def ιz sum(z, {x, y}).
(13.9)
For example, the meaning of the term John and Mary can be written as
j ⊕ m.
Definition (Generalized sum)
For any nonempty set P , its sum ⊕P is defined as ιz sum(z, P ).
(13.10)
For example, the meaning of the term the water can be written as
⊕water.4
If Tarski’s definition of sum (13.5) is used, CEM is defined by the
axioms of reflexivity (13.1), transitivity (13.2) and antisymmetry (13.3)
taken together with uniqueness of sums (13.8). In fact, this setup makes
axioms (13.1) and (13.3) redundant, because any transitive relation that
satisfies axiom (13.8) is provably reflexive and antisymmetric. If the
definition of sum (13.6) is used, transitivity and uniqueness of sums are
not sufficient to define CEM, as there is a model of these axioms in
which we have one element that is not part of itself. Adding reflexivity
and transitivity rules out this model but still does not exclude certain
more complex models which are not models of CEM. Tarski’s definition
of sum is therefore to be preferred. (For discussion and proofs of these
facts, see Hovda, 2009).
The properties of parthood described by CEM are very similar to
those of subsethood in standard set theory. More specifically, one can
prove that the powerset of any given set, with the empty set removed,
and with the partial order given by the subset relation, satisfies the
axioms of CEM. The empty set must be removed because it is a subset
of every other set (a “bottom element”), but CEM precludes anything
from being part of everything. For practical purposes one can therefore
often regard sums as sets, parthood as subsethood, and sum formation
as union. Some correspondences between CEM and set theory are listed
in Table 13.1. Readers who are unfamiliar with mereology might find
this table useful to strengthen their intuitions about the properties of
the parthood relation and of the other operations in CEM.
4
The semantic literature following (Sharvy, 1980) often uses the equivalent
notation σxP (x) rather than ⊕P . We use ⊕ to make the connection between
binary and generalized sum clearer.
Mereology
519
Property
CEM
1.
Reflexivity
x≤x
x⊆x
2.
Transitivity
x≤y∧y ≤z →x≤z
x⊆y∧y ⊆z →z ⊆z
3.
Antisymmetry
x≤y∧y ≤x→x=y
x⊆y∧y ⊆x→x=y
4.
Interdefinability
x≤y ⇔x⊕y =y
x⊆y ⇔x∪y =y
5.
Unique sum/union
P 6= ∅ → ∃!z sum(z, P )
∃!z(z = ∪P )
6.
Associativity
x ⊕ (y ⊕ z) = (x ⊕ y) ⊕ z
x ∪ (y ∪ z) = (x ∪ y) ∪ z
7.
Commutativity
x⊕y =y⊕x
x∪y =y∪x
8.
Idempotence
x⊕x=x
x∪x=x
9.
Unique separation
x < y → ∃!z(x ⊕ z = y ∧ ¬(x ◦ z))
x ⊂ y → ∃!z(z = y − x)
Set Theory
Table 13.1 Correspondences between CEM and set theory
13.2.2 The algebraic (lattice-theoretical) perspective
Above, we have developed the notion of mereology starting out from
the notion of parthood, deriving from it the notion of sum, or fusion. In
the most succinct formulation, Properties 2 (transitivity) and 5 (unique
sum) in Table 13.1 are considered axioms, and the other properties follow
from them as theorems. Alternatively, we may start out with the sum
operation, and derive from that the part relation.
That is, we may start out with the notion of a lattice, or to be specific,
a join semi-lattice. This is defined as a structure hL, ⊕i, where L is a
set, and ⊕ is a two-place operation on this set called join (full lattices
also have an operation of meet, ⊗). For each two elements x, y ∈ L, the
join x ⊕ y is defined, and x ⊕ y ∈ L. By definition, the join operation
of a lattice meets properties 6 (associativity), 7 (commutativity) and
8 (idempotence). (For full lattices with a meet operation, the law of
absorption holds: x ⊕ (x ⊗ y) = x, x ⊗ (x ⊕ y) = x). If we furthermore
impose property 9 (unique separation), the join operation has the same
properties as the sum operation in mereology. A lattice in which unique
separation holds is called complemented. Given unique separation, if
we define parthood in terms of join as in property 4, then the properties
of reflexivity (13.1), transitivity (13.2) and antisymmetry (13.3) follow
as theorems.
In order to qualify as a mereology, the join semi-lattice must not have
a bottom element. This entails that there is no general meet operation.
But we can still define a restricted meet of x ⊗ y = z as the largest
element such that z ≤ x and z ≤ y, provided that there is such an
element.
Unique separation plays an important role in excluding non-intended
520
c
d
b
a
c
b
a
d
(a)
(b)
g
d
a
c
d
e
f
b
a
b
c
(c)
(d)
Figure 13.1
models of parthood such as in Structure 13.1(a) below. (The models in
Figure 13.1 show the parthood relation exhaustively. The elements that
form the leaves of these models — a, b, c in Structure 13.1(a), for example
— are atoms, and they are pairwise disjoint.) Intuitively, we want a ⊕ b
to consist of a and b and nothing else which is not already part of a
and b. To see that unique separation is violated in Structure 13.1(a),
notice that a is a proper part of d, and that there is more than one thing
x disjoint from a such that a ⊕ x = d. Specifically, a ⊕ b = a ⊕ c = d.
Structures like 13.1(a) have “too many” parts. Structures like 13.1(b) and
13.1(c), on the other hand, have “too few” parts; in both of them, b is a
proper part of c, but there is no other immediate proper part of c. Such
structures are excluded by unique separation as well, as we have b < c,
but there is no x disjoint from b such that b ⊕ x = c. If we have three
atomic elements, a, b, and c, then the only structure with three atomic
elements a, b, c satisfying commutativity, associativity, idempotence, and
unique separation is Structure 13.1(d). Note that Structure 13.1(d) is
isomorphic to the powerset of the set {a, b, c} minus the empty set,
where parthood corresponds to the subset relation. This isomorphism is
the reason for the correspondences in Table 13.1 above.
Mereology
521
The role that unique separation plays in eliminating Structures
13.1(a), 13.1(b) and 13.1(c) is played by distributivity in full lattices,
i.e., lattices with a bottom element (where distributivity, here for sets,
states that x ∪ (y ∩ z) = (x ∪ y) ∩ (x ∪ z)).
Structure 13.1(d) would also be identified as the lattice that is generated by the atoms a, b, c, where the join operation satisfies commutativity, associativity, and idempotence, does not express equalities that are
not required by these laws (such as a ⊕ b = a ⊕ c) in 13.1(a)) and does
not entertain any differences not required by these laws (such as c 6= b
in 13.1(b) and 13.1(c)). Such structures are called free lattices that are
generated by a set of atoms (see Landman, 1991).
The notion of join as introduced above entails that the join of any finite
subset of L is uniquely defined. That is, for any non-empty finite subset
L′ ⊆ L, the elements of L′ form a unique join, which is the minimal
upper bound of L′ (the smallest x such that for all a ∈ L, a ≤ x).
This is because by commutativity, associativity, and idempotence, the
order of joining two elements, of more than two elements, and of joining
the same element repeatedly does not matter. We can write ⊕L′ for this
minimal upper bound, corresponding to the generalized sum operation in
the previous section. As a consequence, every finite lattice is bounded,
in the sense that there is an y ∈ L such that for all x ∈ L it holds that
x ≤ y. However, this does not guarantee that join is defined for infinite
sets. The existence of a unique minimal upper bound for infinite sets
would have to be stipulated explicitly. The resulting structure is often
called a complete join semilattice, where completeness implies that
all subsets of L have a join; but recall that the empty set has no join, as
this would require a bottom element. The structure can also be described
as a complete lattice with the bottom element removed, or equivalently,
as a complete Boolean algebra with the bottom element removed.
The algebraic conception of a lattice generates similar structures as
the order-theoretic conception; a fundamental theorem of lattice theory
shows that the two coincide (cf. Grätzer, 1998; Landman, 1991). The
algebraic perspective is motivated by the fact that natural language allows for a conjunction of terms, expressed by and, which is to be modeled
by the join operation. As this conjunction can only generate phrases of
finite length, we might not even want to require that it is infinite. However, if the join operation is used to define the part relation, and the
part relation in turn is used to render the meaning of definite descriptions, then a finite join operation is not sufficient to cover terms like the
natural numbers, whose nouns apply to an infinite number of entities.
522
Other comparable systems are the logic of plurality defined by Link
(1983, 1998), the part-of structures of Landman (1989, 1991, 2000), and
the lattice sorts and part structures of Krifka (1990a, 1998). These systems are clearly intended to describe CEM. For example, Link (1983)
and Landman (1989) explicitly argue that modeling reference to plurals
requires systems with the power of a complete join semilattice. However, careful review shows that many of these axiomatizations contain
errors or ambiguities and therefore fail to characterize CEM as intended
(Hovda, 2009).
13.2.3 Mereology versus set theory, and the issue of
atomicity
In the wake of Massey (1976) and Link (1983), most theories of plurals and mass nouns have been formulated in a mereological framework.
Given the correspondences in Table 13.1, working directly with set theory instead of mereology might look like a better choice. After all, set
theory is better known and more generally accepted than mereology. Indeed, early approaches to plural semantics adopted set theory (Hausser,
1974; Bennett, 1974).
So what are the reasons for preferring mereology? Some authors, such
as Link (1983), have advanced philosophical reasons: Following the philosophical doctrine of “composition as identity”, they take an ontological
commitment to sums to be less substantive than a commitment to sets,
because a sum is taken to be nothing over and above its parts, as opposed to a set. Related to this, sets are abstract, but sums are concrete
if their parts are concrete. This corresponds to how we conceive of sum
individuals: the plural individual consisting of these three apples has a
certain weight, and is situated in space – properties that sets, as abstract
entities, arguably lack.
However, others, like Landman (1989), have argued that these reasons
are independent of linguistic considerations, and that sets can be used
as a model of sum individuals. In order to do so, mereological atoms
need to be identified with the singleton sets that contain them (Scha,
1981; Schwarzschild, 1996).
With ordinary set theory, we are forced to assume that everything is
ultimately composed of atoms, represented by singleton sets:5
5
Alternatively, one can provide room for non-atomic structures in set theory by
modifying the standard axioms so as to allow infinitely descending sequences of
sets, as in the ensemble theory of Bunt (1985).
Mereology
523
Definition (Atom)
atom(x) =def ¬∃y(y < x)
(13.11)
An atom is something which has no proper parts.
However, atomistic mereologies may cause problems for the modeling
of events, mass entities, and spatiotemporal intervals, where one does
not want to be forced to assume the existence of atoms.
By themselves, the axioms of CEM do not specify whether atoms
exist or not: among the models they describe, there are some in which
everything is made up of atoms (in particular, this includes all models
that contain finitely many elements), some in which there are no atoms
at all, and intermediate cases. Either (13.12) or (13.13) can be added to
CEM to constrain it to one of the two limiting cases. For example, many
authors assume that the count domain is constrained to be atomic. The
referents of proper names and the entities in the denotations of singular
count nouns are then taken to be mereological atoms.
Axiom (Optional axiom Atomicity)
∀y∃x(atom(x) ∧ x ≤ y)
(13.12)
All things are made up of atoms.
Axiom (Optional axiom Atomlessness)
∀x∃y(y < x)
(13.13)
Everything is infinitely divisible.
When neither of these axioms is added, the system remains underdetermined with respect to whether or not atoms exist. This underdetermination is one of the advantages of mereology: When we describe the
domains of space, time, and mass substances, we do not need to decide
whether atomic events and atomic instants of time exist or whether mass
substances can be infinitely subdivided.
A different kind of atoms, called impure atoms or groups, is used
in Landman (1989, 2000) and elsewhere to model certain instances of
collective predication and group nouns like committee or team. Groups
are atomic entities which are derived from sums via a special group
formation relation or operator. Group formation introduces a distinction
between the sum a ⊕ b whose proper parts are the individuals a and b,
and the group that corresponds to this sum. This group has a and b as
524
“members”, but it has no proper parts. Groups can be seen as adding a
notion of structured parthood into mereology that is otherwise absent,
and formally speaking, the group formation operation is not a part of
CEM but added on to it. Landman’s view on group nouns is influential
but not uncontested; other authors assume that group nouns involve
reference to set- or sum-like pluralities (Bennett, 1974), or to integrated
wholes (e.g. Moltmann, 1997, 2005), that they involve sum formation
of discourse referents instead of objects (Krifka, 1991), or that it is not
possible to recover the membership of a group from the denotation of
a group noun (see Barker, 1992). Pearson (2011) points out that group
nouns as a class are heterogeneous: some of them (e.g. bunch, collection)
behave less like committee and more like measure nouns (Section 13.3.2).
13.3 Linguistic Applications
13.3.1 Nominal Domain
One of the principal applications of mereology in semantics is the characterization of oppositions such as count-mass and singular-plural
in terms of higher-order properties. For example, various authors have
identified the properties of mass and plural terms like gold and horses
with the notions of cumulative reference: if two things are gold then
their sum is also gold (see Quine, 1960), and if you add some horses to
some other horses then you again get some horses (see Link, 1983).
Definition (Cumulative reference)
CUM(P ) =def ∀x∀y(P (x) ∧ P (y) → P (x ⊕ y))
(13.14)
A predicate P is cumulative if and only if whenever it holds of two things,
it also holds of their sum.
As an example, consider structure 13.2(a); the gray circles form a
cumulative set.
Cumulativity can be generalized for infinite sets as follows:
Definition (Cumulative reference)
CUM(P ) =def ∀P ′ (P ′ 6= ∅ ∧ P ′ ⊆ P → ⊕P ′ ∈ P )
(13.15)
A set is cumulative if and only if it contains the sums of all of its
nonempty subsets.
Mereology
a
c
b
525
a
d
(a)
c
b
d
(b)
Figure 13.2
The dual of cumulative reference, divisive reference, has also been
proposed to describe mass terms (see Cheng, 1973). While cumulative
reference “looks upward” from the part to the sum, divisive reference
“looks downward” from the sum to the parts. On the view that mass term
like gold have divisive reference, this means that any part of anything
which is gold is also gold.
Definition (Divisive reference)
DIV(P ) =def ∀x(P (x) → ∀y(y < x → P (y)))
(13.16)
A predicate P is divisive if and only if whenever it holds of something,
it also holds of each of its proper parts.
The assumption that divisive reference holds of mass denotations runs
into the minimal-parts problem. If we want it to apply to mass nouns
like water, divisivity is at odds with chemistry, as a hydrogen atom as
part of an H2 O molecule does not count as water. The problem of divisivity is even more obvious with heterogeneous mass terms like fruit
cake (e.g. Taylor, 1977): a portion of fruit cake may contain sultanas, but
these sultanas do not themselves qualify as fruit cake. Most semanticists
accept that mass nouns do not in general have divisive reference (see
Gillon, 1992). Many authors have modified the definition of divisive reference to avoid the minimal-parts problem, e.g., by adding a “granularity
parameter” that prevents it from applying from parts that are lower than
a certain threshold (see Champollion, 2010). Landman (2011) proposed
a further distinction beyond atomicity: mess nouns like lemonade have
overlapping atoms, while the atoms of neat nouns like furniture do not
overlap; this cuts across the grammatical mass/count distinction, as e.g.
fence is a messy count noun.
There is another property that has been proposed to model semantic
properties of nominals, quantized reference (see Krifka, 1989):
526
Definition (Quantized reference)
QUA(P ) =def ∀x(P (x) → ∀y(y < x → ¬P (y)))
(13.17)
A predicate P is quantized if and only if whenever it holds of something,
it does not hold of any of its proper parts.
For example, the fat circles in Structure 13.2(b) form a quantized set,
and so do the gray circles. Singular count nouns are interpreted as quantized sets. If we consider count nouns as applying to atoms, quantization
follows vacuously; if not, quantization seems to be an essential property
(as e.g. a proper part of a chair is not a chair). However, there are count
nouns like twig or sequence for which quantization is a problematic assumption (Zucchi and White, 2001): A part of a twig may also be a
twig, and a part of a sequence may also be a sequence. One possible
line of response is that the meaning of these nouns is partially specified
by context, and that when the context is fixed, each of them denotes a
quantized set (see Chierchia, 2010; Rothstein, 2010).
Given the denotation of a singular count noun, the denotation of the
corresponding plural count noun can be described by algebraic closure, often represented by the star operator (see Link, 1983). Algebraic
closure extends a predicate P so that whenever it applies to a set of
things individually, it also applies to their sum.
Definition (Algebraic closure)
∗
If set P is nonempty, then
P =def {x | ∃P ′ P ′ 6= ∅ ∧ P ′ ⊆ P ∧ x = ⊕P ′ }.
(13.18)
The algebraic closure of a set P is the set that contains any sum of things
taken from P.
As an example, suppose the set C = {a, b, c} is the set of all cats. Then
C represents the meaning of the singular count noun cat. The algebraic
closure of C is written ∗C and is the set {a, b, c, a⊕b, b⊕c, a⊕c, a⊕b⊕c}.
This set contains everything that is a cat or a sum consisting of two or
more cats, so it is a superset of C. Also, it is a cumulative set; whenever
x, y ∈ C, then x ⊕ y ∈ C. In Structure 13.2(a), the set of gray circles is
the algebraic closure of the set of fat circles.
Other definitions of algebraic closure are also sometimes found in the
literature. For example, Sternefeld (1998) renders Link’s definition as
follows:
Mereology
Definition (Alternative definition Algebraic closure)
P, ∗P is the smallest set such that
(i)
P ⊆ ∗P ;
(ii)
If a ∈ ∗P and b ∈ ∗P, then a ⊕ b ∈ ∗P.
527
For any set
(13.19)
This reformulation is equivalent to the one in Definition 13.18 for
finite P , but the equivalence breaks down when P has countably infinite cardinality. Definition 13.18 amounts to powerset formation; by
Cantor’s diagonal argument, the powerset of a countably infinite set is
uncountably infinite. Definition 13.19, by contrast, amounts to forming
the union of countably many countably infinite sets. Given the axiom of
choice, this union is itself only countably infinite. So given a countably
infinite P such as the set of natural numbers, the cardinality of ∗P will
vary depending on the definition.
In most mereological approaches to the semantics of count nouns, a
plural count noun denotes roughly the algebraic closure of the set denoted by its singular form. There is disagreement on whether the denotation of a plural noun also contains entities denoted by its singular
form (see Farkas and de Swart, 2010). On the exclusive view, singular
and plural forms of a count noun denote disjoint sets (see Link, 1983;
Chierchia, 1998a). For example, if in Structure 13.2(a) the set {a, b, c}
forms the denotation of cat, then the four gray circles without bold lining
form the denotation of cats. The plural form cats essentially means two
or more cats. On the inclusive view, the plural form of a count noun
denotes its algebraic closure (see Krifka, 1986; Sauerland, 2003; Sauerland et al., 2005; Chierchia, 2010). In 13.2(a) this is the set of all gray
circles. The plural form essentially means one or more cats. The inclusive view assumes that when singular reference is intended, singular and
plural forms are in pragmatic competition, and the more specific singular form blocks the plural form. There are good arguments against the
exclusive view (see Schwarzschild, 1996, page 5). For example, suppose
that there is only one doctor. On the exclusive view, the plural form doctors denotes the empty set. Therefore, the exclusive view cannot derive
the meanings of sentences like (1) compositionally without resorting to
devices such as intensionality.
(1)
No doctors are in the room.
A more tricky problem is presented by dependent plurals, i.e., bare plurals which are c-commanded by a plural coargument (see de Mey, 1981).
528
(2)
Boys / Some boys / Several boys / Five boys flew kites.
These sentences have a reading which entails that each of the boys
in question flew one or more kites (which is a problem for the exclusive
view), and that the total number of kites flown was two or more (which
is unaccounted for by the inclusive view). A possible solution may be to
take the inclusive view and to treat the inference that at least two kites
were flown in total as a grammaticalized scalar implicature (Spector,
2007; Zweig, 2009, see also Sauerland, 2003; Sauerland et al., 2005 for a
similar proposal).
Mass nouns generally do not show a singular/plural contrast, except
if coerced to count nouns, as in we ordered three beers. In English, they
typically occur in the singular, although some plural nouns have been
argued to be mass (e.g. clothes, see Ojeda, 2005). If mass nouns are
intrinsically cumulative, then plural formation has no function, as it
does not change their denotation — if CUM(P ), then P = ∗P (Chierchia,
1998b).
Proof Assume CUM(P); as it always holds that P ⊆ ∗P we just have
to show that ∗P ⊆ P . Assume to the contrary that ∗P 6⊆ P , that is,
there is an a, a ∈ ∗P and a 6∈ P . Then there is, following (13.18), a P ′
with P ′ ⊆ P and a = ⊕P ′ . But as P is cumulative, we have ⊕P ′ ∈ P ,
following (13.15). Hence a ∈ P , contrary to assumption.
However, there are cases in which mass nouns can be pluralized with
special meaning effects, e.g., in Greek for expressing large quantities (see
Tsoulas, 2008).
We end this section by pointing out that bare plural and mass terms
also can be used to refer to kinds, as in Potatoes were first cultivated
in South America, or Nylon was invented in 1935. (For further reading,
see Krifka et al., 1995; Chierchia, 1998b; Krifka, 2004; Delfitto, 2005;
Cohen, 2007; Lasersohn, 2011; Doetjes, 2012).
13.3.2 Measure constructions
We now turn to constructions like three liters of milk or two cats. Such
expressions refer to quantized sets; no proper part of an entity that falls
under three liters of milk also falls under three liters of milk, and no
proper part of an entity that falls under two cats should fall under two
cats. We can derive the quantized status of two cats from the properties
of its parts. We introduce the following operator:
Mereology
529
Definition (Atomic number)
atoms(x) =def card({y | y ≤ x ∧ atom(y)})
(13.20)
The atomic number of x is the cardinality of the set of atoms that are
part of x.
With this, we can interpret two cats as {x | ∗cat(x) ∧ atoms(x) = 2},
which can be shown to be a quantized set, assuming that cat only applies to atoms. This set is represented by the gray circles in Structure
13.2(b), if the set of cats is {a, b, c}. However, this kind of representation is not useful for pseudopartitive expressions based on measure
nouns and mass nouns, like three liters of water. Here, we assume a
measure function that stands in a systematic relation to parthood.
Measure functions can be thought as mapping substances to positive
real numbers (see Krifka, 1989), in which case three liters of water may
be represented as {x | water(x) ∧ liters(x) = 3}. Alternatively, measure
functions map substances to degrees which are in turn mapped to numbers by what we may call unit functions, in which case three liters
of water may be represented as {x | water(x) ∧ liters(volume(x)) = 3}
(Lønning, 1987). While we will use the former representation for convenience, the latter one has advantages: for example, three inches of water
may be represented either using inches(depth(x)) or inches(diameter(x))
depending on context. For other advantages, (see Schwarzschild, 2002).
Not every measure function is permissible in a pseudopartitive; while
three liters of water is admissible, *three degrees Celsius of water is
not. This constraint is related to the distinction between extensive and
intensive measure functions (see Krifka, 1998). In physics and measurement theory, an extensive measure function is one whose magnitude
is additive (see Krantz et al., 1971; Cohen et al., 2007). For example,
when one considers the water in a tank, liters is an extensive measure
function because the water as a whole measures more liters than the
volume of any of its proper parts. But degrees is not extensive because
the temperature in degrees of the water as a whole is no different from
the temperature of its proper parts.
One way of making this notion precise is the following:
Definition (Measure function)
a set S ⊆ dom(µ) iff:
A measure function µ is extensive on
∀x(x ∈ S → ∀y(y < x ∧ y ∈ dom(µ) → µ(y) < µ(x)))
(13.21)
530
For every proper part of every element in the set, the function returns
a lower value than for that element.
Not all measure functions used in pseudopartitives are extensive in this
strict sense. If two feet of snow fell on West Berlin and two feet of snow
fell on East Berlin, then two feet, and not four feet, fell on Berlin as a
whole. So the notion of extensivity needs to be appropriately relativized.
Intuitively, the underlying measure function height is still extensive as
long as we restrict ourselves to layers of snow piled on one another rather
than adjacent layers (see Schwarzschild, 2006; Champollion, 2010).
Given that the measure function liters is extensive on the set of all
water quantities, we can show that the denotation of three liters of water,
{x | water(x) ∧ liters(x) = 3}, denotes a quantized set.
Proof Take an x in the set and a proper part y, y < x. Then either
¬water(y), in which case y cannot be in the set; or water(y), in which
case, since liters is extensive, we have liters(y) < 3, so y cannot be in
the set.
We find the definite article in expressions like the two cats, or the three
liters of water. It has been suggested to interpret the definite article as
generalized sum (e.g., Link, 1983). This works well for cumulative sets;
e.g., in the model above, the cats would refer to ⊕cat = a ⊕ b ⊕ c, the
sum of all cats, represented by the black circle in Structure 13.2(b).
However, this would not work in the case of quantized sets; for example, the two cats should not be defined in the model above, but
⊕{x | x ∈ *cat ∧ atoms(x) = 2} is defined, and would also refer to
the black circle in Structure 13.2(b). Following Montague (1973) (see
also Sharvy, 1980), we can alternatively interpret the definite article as
referring to the supremum of a set (see Krifka, 1986):
Definition (Supremum of a set P)
sup(P ) =def ιz(P (z) ∧ ∀x(P (x) → x ≤ z))
(13.22)
The supremum of a set P is the single element z in P that is greater
than or equal to any of its elements.
Many sets will not have a supremum. For a cumulative set like *C, the
supremum sup(*C) exists, and is identical to ⊕*C. For a non-cumulative
set like {x | x ∈ *C ∧ atoms(x) = 2}, the supremum exists only if it is
a singleton set, and then refers to the element in this set – hence, only
if there are exactly two cats. This captures the intuitive meaning of the
Mereology
531
two cats, and the proposal can be extended to the cat as well, if cat
applies to atomic cat entities; the expression will be defined only if there
is exactly one cat, and will refer to that cat.
The notion of measure function in a mereological setting has been
fruitful to explain interesting linguistic phenomena. Krifka (1986) has
suggested that count nouns in English like cat do not refer to atoms,
but contain a built-in measure function, called “natural unit”. In this
they differ from count nouns in languages like Chinese, which require a
classifier. Krifka (1990b) and Doetjes and Honcoop (1997) have analyzed
sentences like 4000 ships passed through the lock, on the reading where
any given ship may pass through the lock many times, as involving the
construction of a complex additive measure function for events.
13.3.3 Verbal Domain
Mereological notions have been applied to predicates to model certain
cases of conjunction, as in the girls sang and danced, which is true if some
of the girls sang and the other girls danced. Link (1983) proposed to lift
sum formation to predicates. Take *S and *D to be the set of singers
and dancers; then sang and danced denotes the set {x ⊕ y | x ∈ *S ∧ y ∈
*D}, the set of individuals consisting of singers x and dancers y. Krifka
(1990b) has extended this to other types, for example to prepositions,
to handle cases like the planes flew above and below the clouds.
The notion of parthood has been used to formally reconstruct the
telic-atelic opposition. Mereology makes it possible to formally relate
this opposition to the singular-plural and count-mass oppositions (e.g.,
Krifka, 1986; Bach, 1986; Moltmann, 1997; Champollion, 2010). Stating generalizations across nouns and verbs is easiest if these categories
both denote sets, so it is convenient to assume that verbs denote sets of
events (the neo-Davidsonian position, Parsons, 1990). A reformulation
of mereological applications to the verbal domain into eventless semantics is possible in most cases, though not always conspicuous in practice
(see Bayer, 1997).
Divisive reference has been used to formally capture the notion of
atelicity. A verbal predicate like run is atelic, or according to Vendler
(1957), an activity; it has the property that “any part of the process is
of the same nature as the whole”. In a neo-Davidsonian event semantics,
run will be interpreted as a set of events R. Now Vendler’s characterization can be restated by claiming that activities, and other atelic
predicates like states (e.g., be happy), have divisive reference: Let e ∈ R
532
be an event of John running from his house to the store. Now e will have
parts, e.g., a part e′ ≤ e in which John runs from his house halfway to
the store, and a part e′′ ≤ e in which he runs from the halfway point all
the way to the store. These parts e′ and e′′ will themselves be running
events, and hence be elements of R. As with mass nouns, divisiveness is
limited by the minimal parts problem; there may be certain parts of e
that are too small to count as running. A notion that is similar to divisive reference is the subinterval property, which holds of any predicate
P just in case whenever P holds at a temporal interval, it also holds
at every subinterval of it (Bennett and Partee, 1978; Dowty, 1979). Cumulativity and nonquantization are also occasionally invoked to model
atelicity. The differences between these various notions are investigated
in (Champollion, 2010).
Constructions such as for two hours can be seen as the verbal equivalent of pseudopartitives (cf. two hours of running). Restricting our attention to events that cannot be executed simultaneously, like runnings
by one agent, we find that {e | e ∈ R ∧ hours(e) = 2} is a quantized
set, if hours is an extensive measure function. The verbal predicate run
for two hours then is telic, an accomplishment; other examples of telic
predicates are accomplishments like run a mile and build a house and
achievements like reach the summit. Telic predicates imply, according to
Vendler (1957), “definite and unique time periods” and may be characterized using quantized reference and related notions (Krifka, 1998). Filip
(2008) analyses perfectivity as a grammatical operation that expresses
telicity, and involves maximalization, similar to the definite article in the
nominal domain.
There is a correspondence between the parts of an event and the parts
of the participants of an event, which has been expressed in (Krifka, 1986,
1992) in terms of a homomorphism-like relation. For example, let e be
the event in which John (j) lifts a certain box b, hence j is the agent
and b is the theme of e, which we note as AG(e) = j and TH(e) = b.
Similarly, assume that e′ is the event in which Mary (m) lifts a certain
table t. We assume that AG and TH have the homomorphic property
called “summativity” or “cumulativity” in (Krifka, 1986, 1989; Landman,
2000):
Definition (Cumulative thematic relation)
Θ(e ⊕ e′ ) =def Θ(e) ⊕ Θ(e′ )
(13.23)
Mereology
533
The Θ-participant of the sum of two events is the sum of the Θparticipants of the two events.
Let L be the set of lifting events; we then have e ∈ L ∧ AG(e) =
j ∧ TH(e) = b (John lift the box ) and e′ ∈ L ∧ AG(e′ ) = m ∧ TH(e′ ) = t
(Mary lift the table). We then also have e ⊕ e′ ∈ L ∧ AG(e ⊕ e′ ) = j ⊕ m ∧
TH(e⊕e′ ) = b⊕t, i.e., we can derive the truth of John and Mary lifted the
box and the table. One condition, of course, is that L is cumulative. Such
inferences motivate the assumption of lexical cumulativity (called
“lexical” because it is taken to apply to all verbs, but not to all verb
phrases): Whenever two events are in the denotation of a verb, so is
their sum (see Scha, 1981; Schein, 1986, 1993; Lasersohn, 1989; Krifka,
1986, 1992; Landman, 1996, 2000; Kratzer, 2007). In this property, verbs
are similar to mass nouns or bare plurals.
But there are problems with cumulativity of thematic relations. Suppose that there are three events e1 , e2 , e3 in which Al dug a hole, Bill
inserted a rosebush in it, and Carl covered the rosebush with soil. Then
one can say that there is also an event e4 in which Al, Bill, and Carl
planted a rosebush. Do we consider e4 equal to the proper sum event
e1 ⊕ e2 ⊕ e3 ? If we do, this scenario is a counterexample to the cumulativity assumption (Kratzer, 2003). The themes of e1 , e2 , e3 are the hole,
the rosebush, and the soil, and the theme of e4 is just the rosebush. The
theme of e4 is not the sum of the themes of e1 , e2 , and e3 , violating
cumulativity. A possible objection is that e4 is not actually the sum of
e1 , e2 , and e3 . Even though the existence of e4 can be traced back to
the occurrence of e1 , e2 , and e3 , nothing forces us to assume that these
three events are mereological parts of e4 (for general discussion and other
solutions, see also Williams, 2009; Piñón, 2011).
It has been observed (e.g., Verkuyl, 1972) that mereological properties
of the arguments or adjuncts of verbs have an impact on mereological
properties of complex verbal constructions. For example, drink milk is
atelic, whereas drink two liters of milk is telic. This can be explained
on the basis of general properties of thematic relations. We show that
drink milk is cumulative, and therefore atelic. Assume that D and M
are cumulative; then the set {e | D(e) ∧ ∃x(M (x) ∧ TH(e) = x)} is
cumulative, too.
Proof Assume that e1 , e2 are two events in this set; this means that
there are x1 , x2 such that M (x1 ), M (x2 ) and TH(e1 ) = x1 , TH(e2 ) = x2 .
As M, D and TH are cumulative, e1 ⊕ e2 is also in this set. Hence drink
milk denotes a cumulative set.
534
To show that drink two liters of milk is quantized, and therefore telic,
we have to assume that certain thematic roles have the additional property we may call Distinctiveness (cf. also the notion of incremental theme
in Dowty, 1991):
Definition (Distinctiveness of Θ)
If e 6= e′ , then Θ(e) 6= Θ(e′ ).
(13.24)
Distinct events have distinct Θ-participants.
This applies, for example, to the theme of drink : Two distinct drinking
events will have two distinct entities that are drunk.
Proof of quantization of ‘drink two liters of milk’ Assume that 2LM is
quantized, and assume that e1 is an event in {e | D(e) ∧ ∃x(2LM(x) ∧
TH (e) = x)}. Hence there is an x1 such that 2LM(x1 ) and TH(e1 ) = x1 .
Assume now a proper part e2 of e1 , that is, e2 < e1 , hence e2 6= e1 . By
distinctiveness, there is an x2 such that TH(e2 ) = x2 and x2 6= x1 , and
by cumulativity, x2 ≤ x1 , hence x2 < x1 . As 2LM is quantized, we have
¬2LM(x2 ), but then also e2 6∈ e1 ; hence the set {e | D(e) ∧ ∃x(2LM(x) ∧
TH(e) = x)} is quantized, too.
The notion of homomorphism inherent in lexical cumulativity, Definition 13.23, also applies to trace functions, which map events to those
entities (e.g. intervals) that represent their temporal and spatial locations (Krifka, 1986). Trace functions have been used for many purposes;
among others, they relate event semantics to interval semantics; they
are involved in the denotation of adverbials like for an hour and to the
store; and they play the same role as measure functions in the analysis
of pseudopartitives such as three hours of running. Commonly, two such
functions are assumed: temporal trace or runtime (τ ) and spatial trace
or location (σ). For example, if e1 and e2 are two events, the runtime
of their sum, τ (e1 ⊕ e2 ), is the sum of their runtimes, τ (e1 ) ⊕ τ (e2 ).
(For more on temporal and spatial trace functions, see Hinrichs, 1985;
Lasersohn, 1988; Krifka, 1998; Link, 1998, as well as Zwarts, 2005, 2006,
where the spatial and temporal trace functions are combined into one.)
The telic-atelic opposition, as well as the singular-plural and countmass oppositions, can be formally related to the collective-distributive
opposition (Champollion, 2010). Distributive predicates require any
event in their denotation to be divisible into events that are atomic
with respect to the appropriate thematic role. For example, any plural event in the denotation of the distributive predicate smile must be
Mereology
535
divisible into parts that have atomic agents and that belong to the denotation of the same predicate. This captures the distributive entailment
from John and Mary smiled to John smiled. By contrast, the subjects
of collective predicates like be numerous or gather can be plural entities whose parts do not themselves satisfy the predicate. Formally, the
subinterval property can be modified and generalized so that it covers
both the property of being atelic and the property of being distributive.
Instead of requiring a predicate that holds at a certain interval to hold
at every subinterval, we can require a predicate that holds of a certain
plural individual to hold of every atomic part of that plural individual.
Acknowledgements We thank Ivano Ciardelli and Paul Hovda for valuable advice on mathematical questions, and Friederike Moltmann for
discussing her work with us. All errors remain ours. Parts of this paper are based on Lucas Champollion’s doctoral research, for which he
owes special thanks to his advisor, Cleo Condoravdi; to the members of
his dissertation committee, Aravind Joshi, Maribel Romero and Florian
Schwarz; and to the following institutions: University of Pennsylvania,
Palo Alto Research Center Inc., Stanford University, Eberhard Karls
Universität of Tübingen, New York University. Manfred Krifka gratefully acknowledges support by the Bundesministerium für Bildung und
Forschung to the Zentrum für Allgemeine Sprachwissenschaft, Berlin,
Grant 01UG0711.
References
Bach, Emmon. 1986. The algebra of events. Linguistics and Philosophy, 15,
5–16.
Barker, Chris. 1992. Group terms in English: representing groups as atoms.
Journal of Semantics, 9(1), 69–93.
Bayer, Samuel L. 1997. Confessions of a Lapsed Neo-Davidsonian: Events and
Arguments in Compositional Semantics. New York, NY: Garland.
Bennett, Michael R. 1974. Some extensions of a Montague fragment of English.
Ph.D. thesis, University of California, Los Angeles.
Bennett, Michael R., and Partee, Barbara H. 1978. Toward the logic of tense
and aspect in English. Bloomington: Indiana University Linguistics Club.
Reprinted in Barbara H. Partee. 2004. Compositionality in Formal Semantics. Oxford: Blackwell. 122-152.
Bunt, Harry C. 1985. Mass terms and model-theoretic semantics. Cambridge,
UK: Cambridge University Press.
Casati, Roberto, and Varzi, Achille C. 1999. Parts and places: the structures
of spatial representation. Cambridge, MA: MIT Press.
536
Champollion, Lucas. 2010. Parts of a whole: Distributivity as a bridge between
aspect and measurement. Ph.D. thesis, University of Pennsylvania.
Cheng, Chung-Ying. 1973. Comments on Moravcsik’s paper. Pages 286–
288 of: Hintikka, Jaakko, Moravcsik, Julius, and Suppes, Patrick (eds),
Approaches to Natural Language. Dordrecht, The Netherlands: Reidel.
Chierchia, Gennaro. 1998a. Plurality of mass nouns and the notion of ‘Semantic Parameter’. Pages 53–104 of: Rothstein, Susan (ed), Events and
Grammar. Dordrecht, The Netherlands: Kluwer.
Chierchia, Gennaro. 1998b. Reference to kinds across language. Natural Language Semantics, 6(4), 339–405.
Chierchia, Gennaro. 2010. Mass nouns, vagueness and semantic variation.
Synthese, 174, 99–149.
Cohen, Ariel. 2007. Between Kinds and Properties: Bare Plurals across Languages. Pages 53–70 of: Friedman, Tova, and Gibson, Masayuki (eds),
Proceedings of the 17th Semantics and Linguistics Theory (SALT) conference. Ithaca, NY: Cornell University.
Cohen, E. Richard, Cvitas, Tom, Frey, Jeremy G., Holstrm, Bertil, Kuchitsu,
Kozo, Marquardt, Roberto, Mills, Ian, Pavese, Franco, Quack, Martin,
Stohner, Jergen, Strauss, Herbert L., Takami, Michio, Thor, Anders J.,
and Renner, Terry (eds). 2007. Quantities, units and symbols in physical
chemistry. Oxford, UK: Blackwell Science.
Cruse, D. Alan. 1979. On the transitivity of the part-whole relation. Journal
of Linguistics, 15, 29–38.
Cruse, D. Alan. 1986. Lexical Semantics. Cambridge, UK: Cambridge University Press.
de Mey, Sjaap. 1981. The dependent plural and the analysis of tense. Pages 58–
78 of: Burke, Victoria A., and Pustejovsky, James (eds), Proceedings of
the Eleventh Conference of the North Eastern Linguistics Society (NELS
11). Amherst, MA: Graduate Linguistic Student Association (GLSA) for
University of Massachusetts.
Delfitto, Denis. 2005. Bare plurals. Pages 214–259 of: Everaert, Martin, and
van Riemsdijk, Henk (eds), The Blackwell Companion to Syntax. Malden,
MA: Blackwell Publishing.
Doetjes, Jenny S. 2012. Count/mass distinctions across languages. Pages
2559–2580 of: Maienborn, Claudia, von Heusinger, Klaus, and Portner,
Paul (eds), Semantics: An International Handbook of Natural Language
Meaning, Vol. 3. Handbooks of Linguistics and Communication Science.
Berlin: Mouton de Gruyter.
Doetjes, Jenny S., and Honcoop, Martin. 1997. The semantics of event-related
readings: A case for pair-quantification. Pages 263–310 of: Szabolcsi,
Anna (ed), Ways of scope taking. Studies in Linguistics and Philosophy,
vol. 65. Dordrecht, The Netherlands: Kluwer.
Dowty, David R. 1979. Word meaning and Montague grammar. Dordrecht,
The Netherlands: Reidel.
Dowty, David R. 1991. Thematic proto-roles and argument selection. Language, 67(3), 547–619.
Mereology
537
Farkas, Donka F., and de Swart, Henriëtte. 2010. The semantics and pragmatics of plurals. Semantics and pragmatics, 3(6), 1–54.
Filip, Hana. 2008. Events and maximization: The case of telicity and perfectivity. Pages 217–256 of: Rothstein, Susan (ed), Theoretical and crosslinguistic approaches to the semantics of aspect. Amsterdam: John Benjamins.
Forrest, Peter. 2010. Mereotopology without Mereology. Journal of Philosophical Logic, 39(3), 229–254.
Gillon, Brendan S. 1992. Towards a common semantics for English mass and
count nouns. Linguistics and Philosophy, 15, 597–640.
Grätzer, George A. 1998. General Lattice Theory (second edition). Basel:
Birkhäuser Verlag.
Hausser, Roland. 1974. Quantification in an extended Montague grammar.
Ph.D. thesis, University of Texas, Austin.
Hinrichs, Erhard. 1985. A Compositional Semantics for Aktionsarten and NP
Reference in English. Ph.D. thesis, Ohio State University.
Hovda, Peter. 2009. What is classical mereology? Journal of Philosophical
Logic, 38(1), 55–82.
Kay, Paul. 1975. A model-theoretic approach to folk taxonomy. Social Science
Information, 14, 151–166.
Krantz, David H., Luce, R. Duncan, Suppes, Patrick, and Tversky, Amos.
1971. Foundations of measurement. New York, NY: Academic Press.
Kratzer, Angelika. 2003. The Event Argument and the Semantics of Verbs,
Chapter 3. http://semanticsarchive.net/Archive/GU1NWM4Z/.
Kratzer, Angelika. 2007. On the plurality of verbs. Pages 269–300 of: Dölling,
Johannes, Heyde-Zybatow, Tatjana, and Schäfer, Martin (eds), Event
structures in linguistic form and interpretation. Berlin, Germany: Mouton
de Gruyter.
Krifka, Manfred. 1986. Nominalreferenz und Zeitkonstitution. Zur Semantik
von Massentermen, Pluraltermen und Aspektklassen. München: Fink.
published 1989.
Krifka, Manfred. 1989. Nominal reference, temporal constitution and quantification in event semantics. Pages 75–115 of: Bartsch, Renate, van Benthem, Johan, and van Emde Boas, P. (eds), Semantics and contextual
expression. Groningen-Amsterdam Studies in Semantics, vol. 11. Dordrecht, The Netherlands: Foris.
Krifka, Manfred. 1990a. Boolean and non-boolean ‘and’. Pages 161–188 of:
Kálmán, László, and Pólos, László (eds), Papers from the Second Symposium on Logic and Language. Budapest: Akadémiai Kiadó.
Krifka, Manfred. 1990b. Four thousand ships passed through the lock: Object
induced measure functions on events. Linguistics and Philosophy, 13,
487–520.
Krifka, Manfred. 1991. How to Get Rid of Groups, Using DRT: A Case for
Discourse-Oriented Semantics. Texas Linguistic Forum, 32, 71–110.
Krifka, Manfred. 1992. Thematic relations as links between nominal reference
and temporal constitution. Pages 29–53 of: Sag, Ivan A., , and Szabolcsi,
Anna (eds), Lexical Matters. Stanford, CA: CSLI Publications.
538
Krifka, Manfred. 1998. The origins of telicity. Pages 197–235 of: Rothstein,
Susan (ed), Events and Grammar. Dordrecht, The Netherlands: Kluwer.
Krifka, Manfred. 2004. Bare NPs: Kind-referring, Indefinites, Both, or Neither? Pages 180–203 of: Young, Robert B., and Zhou, Yuping (eds),
Proceedings of the Thirteenth Conference on Semantics and Linguistic
Theory (SALT XIII). Ithaca, NY: CLC Publications, for Cornell University.
Krifka, Manfred, Pelletier, Francis J., Carlson, Gregory N., ter Meulen, Alice,
Chierchia, Gennaro, and Link, Godehard. 1995. Genericity: An Introduction. Pages 1–124 of: Carlson, Gregory N., and Pelletier, Francis J. (eds),
The Generic Book. Chicago, IL: University of Chicago Press.
Landman, Fred. 1989. Groups, I. Linguistics and Philosophy, 12, 559–605.
Landman, Fred. 1991. Structures for semantics. Dordrecht, The Netherlands:
Kluwer.
Landman, Fred. 1996. Plurality. Pages 425–457 of: Lappin, Shalom (ed),
Handbook of Contemporary Semantics. Oxford, UK: Blackwell Publishing.
Landman, Fred. 2000. Events and plurality: The Jerusalem lectures. Dordrecht, The Netherlands: Kluwer.
Landman, Fred. 2011. Count Nouns – Mass Nouns, Neat Nouns – Mess Nouns.
The Baltic International Yearbook of Cognition, Logic and Communication, 6, 1–67.
Lasersohn, Peter. 1988. A semantics for groups and events. Ph.D. thesis, Ohio
State University.
Lasersohn, Peter. 1989. On the readings of plural noun phrases. Linguistic
Inquiry, 20, 130–134.
Lasersohn, Peter. 2011. Mass Nouns and Plurals. Pages 1131–1153 of: Maienborn, Claudia, von Heusinger, Klaus, and Portner, Paul (eds), Semantics:
An International Handbook of Natural Language Meaning, Vol. 2. Handbooks of Linguistics and Communication Science. Berlin: Mouton de
Gruyter.
Leonard, Henry S., and Goodman, Nelson. 1940. The calculus of individuals
and its uses. The Journal of Symbolic Logic, 5(2), 45–55.
Leśniewski, Stanisław. 1916. Podstawy ogólnejteorii mnogości I [Foundations of the General Theory of Sets, I] (No further parts appeared.).
Moscow: Popławski. (Prace Polskiego Koła Naukowego w Moskwie. Sekcya matematyczno-przyrodnicza, No.2.).
Link, Godehard. 1983. The logical analysis of plurals and mass terms: A
lattice-theoretical approach. Pages 303–323 of: Bäuerle, Reiner, Schwarze,
Christoph, and von Stechow, Arnim (eds), Meaning, Use and Interpretation of Language. Berlin, Germany: de Gruyter.
Link, Godehard. 1998. Algebraic semantics in language and philosophy. Stanford, CA: CSLI Publications.
Lønning, Jan Tore. 1987. Mass terms and quantification. Linguistics and
Philosophy, 10, 1–52.
Massey, Gerald J. 1976. Tom, Dick, and Harry, and all the king’s men. American Philosophical Quarterly, 13(2), 89–107.
Mereology
539
Moltmann, Friederike. 1997. Parts and wholes in semantics. Oxford, UK:
Oxford University Press.
Moltmann, Friederike. 1998. Part structures, integrity, and the mass-count
distinction. Synthese, 116(1), 75–111.
Moltmann, Friederike. 2005. Part structures in situations: The semantics of
individual and whole. Linguistics and Philosophy, 28, 599–641.
Montague, Richard. 1973. The proper treatment of mass terms in English.
Pages 173–178 of: Pelletier, Francis J. (ed), Mass terms. Dordrecht, The
Netherlands: Reidel.
Ojeda, Almerindo. 2005. The paradox of mass plurals. Pages 389–410
of: Mufwene, Salikoko (ed), Polymorphous linguistics: Jim McCawley’s
Legacy. Cambridge, MA: MIT Press.
Parsons, Terence. 1990. Events in the semantics of English. Cambridge, MA:
MIT Press.
Pearson, Hazel. 2011. A new semantics for group nouns. Pages 160–168 of:
Washburn, Mary B., McKinney-Bock, Katherine, Varis, Erika, Sawyer,
Ann, and Tomaszewicz, Barbara (eds), Proceedings of the 28th West Coast
Conference of Formal Linguistics (WCCFL). Somerville, MA: Cascadilla
Proceedings Project, for University of Southern California.
Piñón, Christopher. 2011. Themes, cumulativity, and event descriptions.
Handout of a talk at Dimensions of Grammar workshop, Konstanz, Germany.
Pianesi, Fabio. 2002. Book review: Friederike Moltmann – Parts and wholes
in semantics. Linguistics and Philosophy, 25(1), 97–120.
Pontow, Carsten, and Schubert, Rainer. 2006. A mathematical analysis of
theories of parthood. Data & Knowledge Engineering, 59(1), 107–138.
Quine, Willard van Orman. 1960. Word and Object. Cambridge, MA: MIT
Press.
Rescher, Nicholas. 1955. Axioms for the part relation. Philosophical Studies,
6(1), 8–1.
Rothstein, Susan. 2010. Counting and the mass/count distinction. Journal of
Semantics, 27(3), 343–397.
Sauerland, Uli. 2003. A new semantics for number. Pages 258–275 of: Young,
Robert B., and Zhou, Yuping (eds), Proceedings of the Thirteenth Conference on Semantics and Linguistic Theory (SALT XIII). Ithaca, NY:
CLC Publications, for University of Washington.
Sauerland, Uli, Anderssen, Jan, and Yatsushiro, Kazuko. 2005. The plural is
semantically unmarked. Pages 413–434 of: Kepser, Stephan, and Reis,
Marga (eds), Linguistic Evidence: Empirical, Theoretical and Computational Perspectives. Berlin, Germany: Mouton de Gruyter.
Scha, Remko. 1981. Distributive, collective and cumulative quantification.
Pages 483–512 of: Groenendijk, Jeroen, Janssen, Theo, and Stokhof, Martin (eds), Formal methods in the study of language, Part 2. Amsterdam,
The Netherlands: Mathematisch Centrum. Reprinted in Groenendijk,
Jeroen and Janssen, Theo and Stokhof, Martin (eds.) Truth, intrpretation and information. GRASS 3. Dordrecht, The Netherlands: Foris, 1984,
pp. 131–158.
540
Schein, Barry. 1986. Event logic and the interpretation of plurals. Ph.D. thesis,
Massachusetts Institute of Technology.
Schein, Barry. 1993. Plurals and events. Cambridge, MA: MIT Press.
Schwarzschild, Roger. 1996. Pluralities. Dordrecht, Netherlands: Kluwer.
Schwarzschild, Roger. 2002. The grammar of measurement. Pages 225–245 of:
Jackson, Brendan (ed), Proceedings of the Twelfth Conference on Semantics and Linguistic Theory (SALT XII). Ithaca, NY: CLC Publications,
for University of California, San Diego and San Diego State University.
Schwarzschild, Roger. 2006. The role of dimensions in the syntax of noun
phrases. Syntax, 9(1), 67–110.
Sharvy, Richard. 1980. A more general theory of definite descriptions. The
Philosophical Review, 89(4), 607–624.
Simons, Peter. 1987. Parts: a study in ontology. Oxford, UK: Clarendon Press.
Simons, Peter. 2011. Stanisław Leśniewski. In: Zalta, Edward N. (ed), The
Stanford Encyclopedia of Philosophy, winter 2011 edn.
Smith, Barry. 1996. Mereotopology: A Theory of Parts and Boundaries. Data
and Knowledge Engeneering, 20, 287–303.
Spector, Benjamin. 2007. Aspects of the Pragmatics of Plural Morphology: On
Higher-Order Implicatures. In: Sauerland, Uli, and Stateva, Penka (eds),
Presuppositions and Implicature in Compositional Semantics. Hampshire,
U.K: Palgrave-Macmillan.
Sternefeld, Wolfgang. 1998. Reciprocity and cumulative predication. Natural
Language Semantics, 6, 303–337.
Tarski, Alfred. 1929. Les fondements de la gómétrie des corps (Foundations of
the geometry of solids). Ksiega Pamiątkowa Pierwszego Polskiego Zjazdu
Matematycznego, suppl. to Annales de la Société Polonaise de Mathématique, 7, 29–33.
Taylor, Barry. 1977. Tense and continuity. Linguistics and Philosophy. Linguistics and Philosophy, 1, 199–220.
Tsoulas, George. 2008. On the grammar of number and mass terms in Greek.
Pages 131–146 of: Proceedings of the 2007 Workshop in Greek Syntax.
Massachusetts: MIT Working Papers in Linguistics.
Varzi, Achille. 2006. A note on the transitivity of parthood. Applied ontology,
1(2), 141–146.
Varzi, Achille C. 2010. Mereology. In: Zalta, Edward N. (ed), The Stanford
Encyclopedia of Philosophy, spring 2010 edn.
Vendler, Zeno. 1957. Verbs and times. The Philosophical Review, 66, 143–160.
Verkuyl, Henk J. 1972. Om the Compositional Nature of the Aspects. Foundations of Language Supplement Series, vol. 15. Dordrecht, The Netherlands: Reidel.
Williams, Alexander. 2009. Themes, cumulativity, and resultatives: Comments
on Kratzer 2003. Linguistic Inquiry, 40(4), 686–700.
Winston, Morton E., Chaffin, Roger, and Douglas, Hermann. 1987. A taxonomy of part-whole relations. Cognitive Science, 11, 417–444.
Winter, Yoad. 2001. Flexible principles in Boolean semantics. Cambridge,
MA: MIT Press.
Mereology
541
Zucchi, Sandro, and White, Michael. 2001. Twigs, sequences and the temporal
constitution of predicates. Linguistics and Philosophy, 24, 187–222.
Zwarts, Joost. 2005. Prepositional Aspect and the Algebra of Paths. Linguistics and Philosophy, 28, 696–740.
Zwarts, Joost. 2006. Event shape: Paths in the semantics of verbs. Manuscript.
Radboud University Nijmegen and Utrecht University.
Zweig, Eytan. 2009. Number-neutral bare plurals and the multiplicity implicature. Linguistics and Philosophy, 32, 353–407.