FRANCIS JEFFRY PELLETIER

ON SOME PROPOSALS FOR THE SEMANTICS

OF MASS NOUNS*

Simple mass nouns are words like water, furniture and gold. We can
form complex mass noun phrases such as dirty water, leaded gold and
green grass. I do not propose to discuss the problems in giving a
characterization of the words that are mass versus those that are not. For
the purposes of this paper I shall make the following decrees: (a) nothing
that is not a noun or noun phrase can be mass, (b) no abstract noun
phrases are considered mass, (c) words like thing, entity and object are
not mass, (d) I shall not consider such words as stuff, substance or
matter, (e) measures on mass nouns (like gallon of gasoline, blade of
grass, etc.) are not considered, (f) plurals of count terms are not con-
sidered mass. Within these limitations, we can say generally that mass
noun phrases are those phrases that much can be prefexed to, by many
cannot be prefexed to, without an0maly.l Semantically, such phrases
usually have the property of coZZectiveness- they are true of any sum of
things of which they are true ; and of divisiveness - they are true of any
part (down to a certain limit) of things of which they are true. All of this,
however, is only generally speaking - I shall mostly use only the simple
examples given above and ignore the problems in giving a complete
characterization of mass nouns.
In the paper I want to discuss some problems involved in casting
English sentences containing mass nouns into some artificial language;
but in order to do this we should have some anchoring framework on
which to justify or reject a given proposal. The problem of finding an
adequate language can be viewed as a case of translation (from English
to the artificial language), where the translation relation must meet
certain requirements. I shall suggest five such requirements; others could
be added. Let Si be a sentence of English and SF be its translation into
an artificial language, then

1. If S has a truth value, then S* must have that truth value

2. If S is analytic, then S* must be analytic

Journul of Philosophical Logic 3 (1974) 87408. AN Rights Reserved

Copyright 0 1974 by D. Reidel Publishing Company, Dordrecht-Holland
88 FRANCIS JEFFRY PELLETIER

3. If S is deducible from S1, S, . . . S,,, then S* must be de-

ducible from S, S*, . . . S,*
4. S and S* must be about the same entities
5. S, is a paraphrase of S, if and only if St is a paraphrase of S,*.

I state condition 3 syntactically; the reason is not so much that I object to

the semantical is a consequence of as it is that I insist on the relationship
being obvious. One could cook up all kinds of consequence relations
on an ad hoc basis one for each mass term (see Section I below on corre-
sponds to). But given some standard uninterpreted artificial language,
it is rather more difficult to cook up ad hoc deducibility relations. For
paraphrase in condition 5, I mean synonymy (to the extent that it
applies to artificial languages - see Lewis on meaning). Doubtless
these conditions could be clarified more, but their intent should be clear
(except perhaps for #4, which will be discussed later).

I. CLASSICAL QUANTIFICATION THEORY

The classical view is presented in Quines Word and Object (henceforth

W & 0). I shall argue that this view of the translation relation violates
condition 3 and condition 2.
Classical quantification theory allows for two distinct uses of mass
nouns. In the predicative use (like water in this puddle is water) it is
treated as a general term, which is true of each portion of the stuff in
question, excluding only the parts too small to count (W & 0, p. 98).
General terms, of course, are translated by a predicate letter. On the
other hand, when a mass noun occurs in subject position (like water
in water is wet), it is treated as a singular term ( W & 0, pp. 97-98). So
the English argument This puddle is water, Water is wet, ergo This
puddle is wet would be translated

Ft, Gw k Gt

(where F: is water, G: is wet, t: this puddle, w: water) which is obviously

not deducible. But since the English argument of which it is the translation
obviously is deducible, this constitutes a violation of condition 3.
Furthermore, the English sentence water is water is obviously
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 89

analytic, yet its translation into classical quantification theory would be

Fw
which (equally obviously) is not analytic, thus violating condition2.
Perhaps, though, Quine would want to claim that in certain circumstances
we should treat is as identity, so that we would get the analytic
w=w
for this sentence. But then we would be at a loss to account for the
analyticity of sentences like Dirty water is water. Another possible
adjustment might be to claim If a is a singular term and F is the general
term corresponding to a, then Fa is analytically true. In addition to
the difficulties in giving a syntactical explication of the corresponding to
relation, I still fail to see how this could account for the analyticity of
Dirty water is water. For we would have
Fd
(where d: dirty water), which, while true, is not analytic.

II. A MEREOLOGICAL INTERPRETATION

One might have doubts about the rationale of this dual treatment of
mass terms by Quine and also about his assertion that this is the
simplest plan, but he has good reasons both for giving this dual treat-
ment and for the assertion. He does consider the possibility that mass
terms be treated as singular terms regardless of their grammatical role,
but rejects this proposal for two reasons. First, it requires that the copula
be ambiguous. In sentences like Socrates is a man it would have to be
taken as the is of predication (or, if one prefers a set-theoretic inter-
pretation, as set membership), while in sentences like this puddle is
water it would have to be taken as is a part of. Secondly, conditions 1
and 5 fail since the representations of x is water and x is a part of
water would be identical and yet the English sentences are not paraphrases
of one another and havent even the same truth-value (since not all parts
of water are themselves water - e.g., the atoms).
The problem with the dual approach seems to be that it cannot
exhibit sufficient intra-sentential logical form to abide by our first three
90 FRANCIS JEFFRY PELLETIER

conditions. However, Julius Moravcsik (1973) claims that there is a way

to exhibit sufficient intra-sentential logical form. That way is to take the
pure mereological approach rejected by Quine, and to use the Leonard-
Goodman calculus of individuals to exhibit the form.
Obviously, in order for Moravcsik to succeed he must dispose of
Quines objections to the pure mereological approach. He does this
by giving a re-interpretation of is a part of and tries to show that this
re-interpretation does not violate the paraphrase condition. What we
need here, claims Moravcsik, is some general scheme by which we can
indicate what the smallest part of any given mereological entity are.
Any part of such an entity, e.g., Water, can be specified by saying it is
a part (in the ordinary sense) of Water and it has certain empirically
describable structural properties (abbreviated 2%). Thus the proposal is
to translate x is 44 where M is a mass term as x is a part (with the
required SP) of the individual M.
Doing this, of course, is to give up Quines simplest plan; but,
Moravcsik says, we do not want this simplicity anyway - we want to be
able to illustrate the difference between mass and count terms. Here this
is done by exhibiting the internal structure of a sentence in different
ways: with count terms the copula is construed as ordinary predication
(or set-membership), with mass terms the copula is construed as is a
part (with SP) of. And in support of this proposal we might note that
it does meet the objections to Quines dual approach. (First though,
it should be noted that Moravcsik allows certain adjectives (e.g., wet)
to also be treated mereologically. And while I decreed earlier that I shall
not consider adjectives to be either mass or count, for the purposes of
the discussion here lets agree to concede Moravcsik this point.) The
argument whose deducibility Qume could not demonstrate was Water
is wet, This puddle is water, ergo This puddle is wet. Moravcsik
could translate it
W<,,T, P<,,W kP<,,T
where W: the individual Water, T: the individual Wet, P: this puddle, and
<*p is interpreted as is a part of (with SP). Given the transitivity of
< sp, this argument is obviously valid. Quine also could not account
for the analyticity of Dirty water is water. But since for Moravcsik a
noun phrase with an adjective plus mass noun denotes the overlap of
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 91

the individual denoted by the adjective and the individual denoted by

the noun, he could demonstrate the analyticity of this sentence as:

where D: the individual Dirty, W: the individual Water, and my is

interpreted as the overlap of x and y. This is analytic in the pure calculus
of individuals (i.e., without the sp subscripts), and intuitively says The
overlap of the dirty-parts and water -parts is a water-part.
Unfortunately, it seems clear that this pure mereological approach will
not work. There are two objections, either of which being sufficient to
show this. The first has to do with the interpretation given to cs,,. As
Moravcsik makes clear, it is not simply to be interpreted as is a part of,
since there are to be smallest parts of (say) water - for instance, the
water-molecules. What Moravcsik has in mind is an intensional inter-
pretation: in order to determine the truth of ~<~ry we must look at
the description given by y. Thus we could have two descriptions, a and
b, which denote the same thing (i.e., the same mereological whole), and
yet x cspa and x < spb might have different truth-values, since the SP-set
relevant to a might be different from the SP-set relevant to b. Moravcsik
in fact makes such a claim when he says (p. 281), F and G are distinct if
and only if there is a part of one, say F, that is an F-part but though it may
be a part of G is not a G-part. Moravcsik does not find this objectionable,
but he should, since we then would not meet one of his own conditions
of adequacy of a semantics; namely, we would not be able to compute
the denotations of complexes containing mass terms on the basis of the
denotations of the contained terms. And furthermore, such an inter-
pretation effectively denies transitivity to xsp, since the SP relevant to
y in xc spy might be different from the SP relevant to z in y < spz,
so that we cannot conclude x < spz. But Moravcsik needs the transitivity
of csp to demonstrate the deducibility of the argument given above. I
believe Moravcsiks only alternative here would be to retain < of the
pure calculus of individuals and make SP be a predicate (a large number
of them, one for each mass term). Thus This puddle is water would be
translated
(P< w> & SP(P)
where P: this puddle, W: the individual Water, SP: has the structural
properties of water. But if this is what is meant, then surely it would be
92 FRANCIS JEFFRY PELLETIER

easier to just drop the superfluous calculus of individuals and say merely
SP(P).

So we see that Moravcsiks attempt to add SP as a covering scheme to

avoid Quines objection merely brushes all the difficulties under the rug.
The second objection is aimed at any attempt to introduce mereo-
logical entities (whether Quines partial approach or Moravcsiks pure
approach), and can be found in T. Parsons (1970), about which we shall
talk later. Given a suitable state of affairs, it might be the case that all
the wood in the world was, is, and will be made into furniture and that
all the furniture in the world was, is, and will be made of wood. So all
parts of the individual Wood are parts of the individual Furniture and
vice versa; and thus the two mereological entities are identical. But
some pieces of wood are not pieces of furniture, so the mereological
approach is false. Moravcsik recognizes the applicability of this objection
to his position and says (p. 282), in addition to the principle quoted above :
. . . we have two choicesfor the denotation of a massterm. It is either the classwhose
membersare the mereologicalunits from each possibleworld, or the mereologicalunit
(super-individual) that is made up of all the mereological units from the various
possibleworlds.
Moravcsik opts for the former alternative for good reason, for the latter
involves the conceptual absurdity of one individual being simultaneously
in all possible worlds. But this first alternative also strikes me as too facile.
Suppose we have two mereological units F and G which are identical in
some world (all F-parts are G-parts and vice versa). How now are we to
tell if the F (which is G in this world) is still F (when it is not G) in some
other possible world? To do this we surely need some criterion for distin-
guishing F from G. And if we do have such a criterion, what could it be
but that some predicate applies to F but not to G? And if this is so, the
mereological interpretation is otiose - we should rather let F be a pre-
dicate (the one indicated by our criterion). Thus, Moravcsiks proposal
destroys the very position it is designed to salvage.

III : SOME SET-THEORETIC INTERPRETATIONS

There have been suggestions e.g., by Strawson in Individuals, and perhaps

also (sometimes) by Quine in W & 0 and his review of Geachs Reference
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 93

and Generality, and by Clarke (1970), to the effect that uses of mass terms
should be understood as elliptical for some more complex phrase in
which there is an explicit individuating standard (or count phrase) by
means of which we can give sense to there being a certain number of
things of which the mass term is true. Thus, is water might be elliptical
(in some circumstances) to is a body of water or (in other circumstances)
is a kind of water. Is sugar might be rendered is a shipment of sugar;
is gold might be short for is a vein of gold or as is a nugget of gold,
depending on the context.2
It is instructive to see that this position is an improvement over the
mereological interpretation, for it at least avoids the problem where all
furniture is, was, and will be wooden, and all wood is, was, and will be
furniture. Under the present interpretation, we need not bring into play
any such objectionable entities as other possible worlds, for in the
actual world there are individuating standards applicable to wood
which are not applicable to furniture. E.g., the set of pieces of wood is
distinct from the set of pieces of furniture. The leg of a chair is an element
of the set of pieces of wood but not of the set of pieces of furniture.
Thus wood and furniture do not denote the same sets, and the case
does not pose a problem for this interpretation.
This view has been elucidated and criticized in Helen Cartwright (1965)
(henceforth H). I shall briefly mention a few of the difficulties to be
found in it. First, if we were to incorporate such a view into our transla-
tion relation, we would find that certain sentences would have no
representation. Consider the sentence What Jeff spilled is the same coffee
as what he wiped up. This sentence is an identity claim, so what is on
one side of the equal sign must be identical with what is on the other
side; and that implies that the same individuating standard must be
applicable to both sides. But what could it be? It cannot be puddle of
coffee for that cannot be spilled. It cannot be cup of coffee for that is
not the kind of thing which one can wipe up. Second, certain sentences
violate condition 1, the truth-value sameness condition. Consider the
sentence The sugar here is the same sugar as that which was on the boat
when the sugar which was on the boat was melted before it came here.
The extension of this sugar here contains no lumps, grains, etc., so there
no longer is a set of these things. Thus the purported set-equality fails,
yet the English sentence may very well be true. And thirdly, there can be
94 FRANCIS JEFFRY PELLETIER

no general translation procedure, for given a particular case, we may have

no way to tell which of the individuating standards is to be used (of
all the ones that truly apply). For instance, sameness of shipment of
sugar is also sameness of lumps of sugar and sameness of grains of sugar.
How, in any particular case, do we know which one to use? And finally,
a sentence like This shipment of sugar is constituted by the grains of
sugar is true, but it is quite unclear just what individuating standard
will apply here - it looks as if we need a use of sugar that does not
depend on the individuating standards of shipment, lump, or grain.
Thus, this method of calling into play context-dependent individuating
standards is seen to be unsatisfactory. It seems that if one wishes to
call sets into play at all, what is required is the same set for all occurrences
of that mass term. Perhaps the simplest plan would be to allow mass
terms to denote the set of the smallest entities of which it is true - the set
denoted by mass noun M would be all and only those things which are
M and of which no part is M. For example, furniture would denote the
set of individual pieces of furniture, the set denoted by water might be
the set of water molecules. Note that this proposal also avoids Parsons
objection: just because all furniture is wood, and all wood furniture,
still it does not follow that furniture is (identical with) wood, because the
minimal parts of each are distinct.2
However, it is obvious that this proposal cannot work; it implies
that for every mass term there will be some count term with the same
denotation. Suppose that water molecules are the minimum parts of
water. It would then follow that water molecules (a count term) and
water (a mass term) would have the same denotation. But this is
obviously false, since (a) it invokes an empirical claim about the meanings
of mass terms (that there always is such a count term), and (b) the
extensions of water and water molecules would still have to be distinct
since water is true of this puddle but water molecules is not. But this is
impossible, since by hypothesis they have the same denotation. And
finally, (c) the copula would have to be regarded as ambiguous: sentences
like This molecule is water would be translated as

but to translate This puddle is water we would have to graft something

like the calculus of individuals onto our set-theoretic base. What is
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 95

required is some sentence like All parts of this puddle that are exactly
the right size are (elements of) water. So we need to add something
(Moravcsiks SP?) which will say which ones are big enough to count.
And we have already seen that this involves many insuperable difficulties.
These objections seem to show that any position which would have a
mass term M denote some, but not all, of the things that are M, cannot
be maintained. The obvious way to proceed then, would be to have M
denote the set of all things that are M. Such a position is the one advanced
by Helen Cartwright in H and also in Quantities (henceforth Q).
I think this position is also wrong, but tirst it needs to be laid out; thus, I
shall state the central thread in H and Q. (Not in detail of course,
but enough to give the proposal a fair run and to bring out what seems
to me wrong).
In English there is a partitive quantifier, some. This quantifier can
be used with count terms in both the singular and plural, and with mass
terms but only so long as they are understood as kind of plus mass term.
There is another word in English, a word that can be used only with
count terms in the plural and with mass terms understood normally.
This word happens to be spelled the same as the partitive quantifier, but
is pronounced with weak stress, and for typographical convenience I
shall indicate it by sm. (The distinction was first made in the philosophical
literature, I believe, in ?-I.) Perhaps some examples will bring out the
difference.
1. Give John sm water.
2. Some man wants water.
3. Sm water would taste good now.
4. Some water tastes worse than L.A. water.
In 1 and 2 it is clear enough what is going on. If the request in 1 is carried
out, John will receive some quantity or other of water - some indeter-
minate amount is asked for (within certain contextual limits). In 2
some is a quantifier: there is a man who wants water. In 3, sm functions
as in 1: it is not that a kind of water (say mineral water) would taste
good now, but simply that the having of (any amount of) water would
taste good. In 4, we have the quantifier some together with an apparent
mass term. This means that water must be understood as kind of
water: and that is precisely what the sentence asserts - the water which
96 FRANCIS JEFPRY PELLETIER

tastes worse than L.A. water is a kind of water, say water from Badwater,
Death Valley. In the cases where sm is used with mass terms, its function
is similar to that of a in
5. A river is good to bathe in.
6. John is a man.
and might be called the indefinite article appropriate to mass nouns. It is
the presence of sm that makes
7. For some X, x is sm water, and Hera&us bathed in x yester-
day, and Heraclitus bathed in x today.
rather than
8. For some x, x is water, and Heraclitus bathed in x yesterday,
and Heraclitus bathed in x today.
be the correct analysis of
9. Heraclitus bathed in water yesterday and bathed in the same
water today.
In the sense in which we would normally say 9 was false, it is because
Hera&us bathed in some water x yesterday and did not bathe in that
same water x today - the same state of affairs which falsifies 7. But to
falsify 8 we need to suppose in addition that what Heraclitus bathed in
today was (say) milk. But now we need to find out what the permissible
values of x are in x is sm water.
In H p. 485, Cartwright suggests that the permissible values are
quantities (of water). In Q she explicates the notion of quantity, but
seems to give a different answer to what the permissible values are. Lets
start with a brief indication of Cartwrights notion of quantity. The first
caution is to avoid identifying it with an amount. We could have the same
amount of water but not have the same quantities of water; non-identical
quantities may be the same amount, and to bring this out Cartwright
adopts the terminology of saying that a quantity contains a certain
amount of it (rather than that it is that amount). It is also important to
mention that Cartwrights notion of the amount contained in a quantity
is not dependent upon a choice of measure, and is not dependent upon
the conditions of measurement (for justification, see Q).
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 97

Now, with these preliminaries granted, Cartwright says

the sense in which a quantity of something contains an amount of it is just analogous

to the sense in which a set of things contains a number of them. A set of things [deter-
mined by a count term like cat or apple] may be defined by
Dl : x is a set of A if and only if, for some y, x and y are comparable with respect
to the number of A each contains, and x contains nothing other than A... .
. . . we my put Dl by saying that a set of things is anything which may be numerically
equal to something with which it need not be identical.. .
D2: x is a quantity of B if and only if, for some y, x and y are comparable with
respect to the amount of B each contains, and x contains nothing other than B.

Cartwrights strategy now is to make an analogy which goes as follows:

we (think we) understand quantification in the set (count) case. If we do,
it is because we understand what a set is and what quantification over the
entities contained in a set comes to. If we are to understand quantification
in the mass case, we must find something which performs the role
performed by the entities collected into a set. The point of Dl and D2,
then, is to evoke this analogy.
But to explain quantification we must explain how one can satisfy
open sentences like x is a man or x is sm coffee. In the former case, it is
clear enough what Cartwright has in mind: man denotes the set of
men and to satisfy the open sentence under some interpretation is for
that interpretation to assign something contained in this set to x. In H
p. 485, Cartwright claims that the values of x which would satisfy x is
sm coffee under an interpretation are quantities of coffee. But this would
be analogous to saying that it is sets of men which satisfy x is a man
under an interpretation. And since this last is false, I see no reason to
hold that the former is true. In Q p. 39, Cartwright has changed her
mind: here it is what is contained in a quantity of coffee which satisfies x
is sm coffee under an interpretation. But what is contained in a quantity
is an amount; and that is not what we want to quantify over. The amount
of coffee is just irrelevant (as Cartwright herself noted in H). Further,
every quantity contains exactly one amount (Q p. 33), yet sets contain
ever so many elements. So, we are not quantifying over the quantity, for
that corresponds to the set; and we are not quantifying over the amounts,
for that corresponds to the number (measure) of the set. What is it that
corresponds to elements of a set? Those are what we want to quantify
98 FRANCIS JEFFRY PELLETIER

over. Nothing in the account will tell us, and so I conclude that the analogy
is not drawn closely enough to justify accepting either of Cartwrights
recommendations.
A charitable thing to say here is that Cartwright wants a mass noun
M to denote the set of quantities of M. But even this seems not to be
sufhcient, for there is an objection to having mass terms denote any
physical object or set of physical objects: the mass term may not be true
of any actual object. Consider two never-to-be-realized (but realizable
and describable) substances called Kaplanite and Suppesite. The two
sets would then be identical, but Kaplanite is a liquid might be true
while Suppesite is a liquid is false,3 thus violating condition 1.
Perhaps the answer here is that the charitable formulation of
Cartwrights view (if it could be made out clearly) gives us the extension
(that is, the things of which it is true) of the mass terms over which we
quantify, but that it does not give us their denotation (that is, the entity
assigned by the semantics to that term). And perhaps this is all that
Cartwright intended. But if so, it still leaves open the question of what
the denotation of mass terms are, and how one can state their semantics.

IV. MASS NOUNS AS DENOTING ABSTRACT SUBSTANCES

Faced with the failures of the physical interpretations-the mereological

and set-theoretic interpretations, perhaps the most plausible thing to do
would be to allow occurrences of mass nouns to name abstract entities.
It is important to note here that under such a proposal, mass terms are
names and not, e.g., predicates or other general terms. Such is the approach
of Parsons (1970).
Let us give the simple mass terms - water, gold, and the like - simple
names, such as w and g respectively. Under Parsons interpretation,
these name substances, a word which is to be taken in the chemists
sense, to stand for any material. We also introduce the relational
constant C to be understood as is constituted of. Thus a sentence
like My ring is gold is translated as

So C is a relation between objects and substances which is true just in

case the matter of the object is a quantity of the substance. The relation
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 99

is a quantity of, symbolized Q, is a primitive in Parsons analysis, but

(p. 367)
I can explain it roughly as follows: A substancelike gold, is found scatteredaround the
universe in various places.Wherever it %ccurs we have a bit of matter which is a
quantity of gold .. . . If it is true to say of an object (a physicalobject) that it is gold,
then the matter makii it up will be a quantity of gold.

Of course, Parsons is aware of Quines objections to analyzing terms after

the copula as singular terms and construing the copula as is a part of ;
but he claims that all this shows is that his is a quantity of does not
mean the same as Qumes is a part of: What parts of x are quantities
of x depends on x, and not some abstract notion of part (p. 366fn).
This suggestions obviously has some connection with Moravcsiks SP,
but we shall investigate that after we look at the rest of the proposal.
The relata of Q are abstract substances on the one hand and bits of
matter on the other. The latter are characterized as, first, not necessarily
being identical with any object of which such a bit comprises all its
matter, and secondly, as being Goodman individuals - i.e., obeying
the laws of the calculus of individuals. After giving this characterization
of bits of matter, Parsons now claims that we are in a position to
quantify over them. We can symbolize Water is wet as

where w: the substance water, W: is wet. And taken together with the
symbolization of This puddle is water as
PQW
where p: this puddle, we obviously can deduce

WP,

This puddle is wet - a feat that Quine could not perform. We can also
demonstrate the analyticity of Dirty water is water as

(4 ((xQw & W =xQw)

another feat beyond classical quantification theory.
Parsons wants to introduce a substance abstraction operation,
Ox [. .x. . 1, which is to be on a par with the set abstraction operator,
2 [. .x. .]. The set operator refers to the set of those xs which satisfy the
100 FRANCIS JEFFRY PELLETIER

formula inside the brackets; the substance operator is to refer to the

substance of those xs which satisfy the formula inside the brackets. Now
that we (supposedly) know the permissible values of variables of quantifi-
cation, we are in a position to represent complex mass terms by means of
this operator. For example, the substance Dirty Water is referred to by
Ox [Dx & XQW]
i.e., the substance whose quantities are dirty quantities of water. (Of
course, the substance Water could be referred to by

as well as simply by w.)

The difficulties with this view can be broken into three categories
formal difficulties, difficulties with the interpretation of Q, and difficulties
with mass terms denoting abstract entities.
I said before that Parsons Q is similar to Moravcsiks < sp. We saw
above that under Moravcsiks interpretation of such sentences as X-Cspy
we had to look to the descriptions x and u in order to determine the
truth of such sentences - i.e., the positions occupied by x in y there are
not purely referential. And so it is in Parsons proposal: in xQy we have
to look to the descriptions x and v in order to determine the sentences
truth value (a fact implied by Parsons p. 366fn - quoted above - but not
remarked upon). And surely Moravcsiks restriction on adequate seman-
tics for mass terms is correct: we must be able to compute the denotation
of complex expressions from the denotation of the simpler expressions
contained therein (e.g., we must be able to give Tarski-type truth defini-
tions for sentences). It is not obvious that this can be done in Parsons
non-extensional language.
Secondly, and related to the first, how is Q to be understood? We
saw above that attempts to introduce a whole collection of predicates
like batch of furniture, vein of gold, nugget of gold, shipment of
sugar, etc., led to problems of explaining how terms with a mass-like
extension can combine with terms that have sets as their denotation. It
is not clear tbat we can sidestep this problem by introducing Q as a
primitive. Does Q have different meanings depending upon whether we
are talking of gold, sugar or furniture? Does it have different mean-
ings depending on whether we are talking of batches, veins, shipments
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 101

or nuggets? It seems that what we have gained in ability to meet the

analyticity and deducibility conditions we have lost in explanatory
power.4
Thirdly, the criticisms of the interpretation of mass terms as denoting
abstract entities. What are we to say of such sentences as gold is yellow?
We cannot attribute yellowness to an abstract entity - we must instead
construe this as an assertion about quantities of Gold. Generally, what
seem to be direct attributions to a thing (viz., the thing denoted by a mass
name) turn out to be indirect attributions to quantities of that thing.
Moravcsik says that, intuitively, the substance Red Ink should be a part
of the substance Ink; but since these substances are abstract entities in
Parsons system, this cannot be so. The relation in question will have to
be put: all quantities of one are quantities of the other. This is different
from the case where we have two predicates: in that case it is permissible
to look to the physical objects and show that all things that satisfy the
one predicate also satisfy the other (e.g., white man and man). But
here we do not have properties - we have things, and the relation should
be that one is a part of the other. If Red Ink is not part of Ink, it is
hard to see why Red ink is ink is analytic. That is, I grant that All
quantities of red ink are ink is symbolized by Parsons in such a way as
to be analytic; still the intuitive understanding of the semantics does
not make Red Ink a part of Ink. Hence, the semantics gives us no
reason to believe that Red Ink is Ink is necessarily true; so it seems
improper for it to be symbolized in such a way as to be analytic. I.e., it
would seem that Red Ink is Ink cannot be adequately characterized
as All quantities of red ink are ink. (This argument is adapted from
Aristotles criticism of Platos theory of forms, Meta. B6). We could put
this objection: in sentences like Red ink is ink we are not talking about
quantities, and so Parsons symbolism involves a violation of our condition
4, the aboutness condition. Parsons himself points out that in sentences
like Most gold is unmined we are not talking about quantities (pp. 371-
373). Its not clear that Most quantities of gold are unmined even makes
sense, given an indefinite number of ways of counting quantities. And
if we do agree on a way, Most gold is unmined can be true while Most
quantities of gold is false - as if all the unmined gold is in one big nugget.
Parsons avoids the problem here by (a) claiming that he is not interested
in most quantifiers, only all and some, and (b) with all and some the
102 FRANCIS JEFFRY PELLETIER

truth-values are the same, and (c) he is only interested in the truth-value
condition, not the aboutness condition. It seems to me, though, that
the aboutness condition is an important one, both in its own right and
in its implications for the analyticity condition (see above). An
artificial language just cannot be adequate if it talks about one group of
things while the natural language talks about a different group.
Finally, whatever advantages Parsons language has, it is not the basic
explanation of the denotation of mass terms. In his informal explanation
of Q (quoted above) he had said: If it is true to say of an object.. . that
it is gold, then the matter making it up will be a quantify of gold.
For all the world, it looks as if we here have a notion more basic than
xQg - namely, that x is gold is true. But if this is so, then there is a
more straightforward explanation of mass terms than the one given by
Q. This one will be studied in the next section.

V. MASS NOUNS AS PREDICATES

When confronted with a sentence like Water is wet, the first impulse of
a student who has completed an elementary course in logic is to translate
it as
(x) (Water x 3 Wet X)
- i.e., to translate water as a predicate. Presumably, this impulse stems
from the instructors recommendation to translate Men are mortal
as
(x) (Man) XX Mortal x).
It is, however, easy to dissuade the student. First one points to the
dissimilarities in the two cases. Men is obviously plural and takes the
plural are, while water seems non-plural since it takes the singular
is. We can give a clear sense to the phrase For all X, if x is a man.. .,
because we have an understanding of what it is to be one man (or a
distinct man, or the same man, etc.); but in the case of water we do not
have an analogous understanding. In For all x, if x is water . . . it is
difficult to give a clear meaning to what x is. And if we do attempt to
give a clear meaning to such locutions, it seems to involve a change in
sense. For example, we might try to restate the quantifier phrase as
For all x, if x is a water.. .. But under the most normal understanding,
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 103

a water is a certain kind of water; and then the sentence would be

equivalent to All kinds of water are wet. And surely this should not
be so.
Once that student is in doubt about the adequacy of his translat-
ion, the teacher can give some arguments to show that the general
attempt to force mass nouns into the mold of predicates is mistaken.6
(1) If we treat mass nouns as predicates, then it is not clear what these
predicates should be true of (there may be many answers: My ring is
gold suggests physical objects, The element with atomic number 79 is
gold suggests elements in the chemists sense, and The particular bit of
matter which makes up my ring is gold suggests matter). (2) Demon-
strating the deducibility of arguments like X is made of gold, Gold is
the element with atomic number 79, ergo X is made of the element with
atomic number 79 depends on gold being a name in the canonical
notation, not a predicate.
By now the bright student is puzzled. If water is not a predicate but
rather a name, what sense are we to make of the phrase all water?
Further, if water is a name what is dirty water? If it is a name does
it have internal structure? The various attempts to construe mass terms
as names (of mereological entities, of sets, or of abstract entities) have
all met with serious difficulties. It seems that the only option left is that
the student is right: mass nouns are predicates. Hence, I shall attempt
to answer the difficulties raised by this identification.
Lets start by adopting the notion of property given by Kripke (1963).
It is a function on possible worlds to classes: the property indicated by
the predicate 4 is the function from possible worlds to the set of 4s in
each world.6 In the non-mass case this is fairly clear: pig indicates a
function from each possible world to the set of pigs in that world. But
things are not so clear in the mass case. Water does not pick out the
class consisting of water in each possible world, unless we give it a special
sense. For, as it is, it seems to suggest water as opposed to milk, honey
etc.
Consider sentences like Pigs are pink. In such sentences we can
always paraphrase the (implicit) quantification as Anything that is a
pig.... The analogue of this in the mass case - as in Water is wet - is
Anything that is sm water.. ., where we have a use of sm, the indefinite
article appropriate to mass nouns. In any case when we wish to speak
104 FRANCIS JEFFRY PELLETIER

of some physical object (say this puddle) of which the predicate water is
true, we can paraphrase it as This puddle is sm water. Generally, when
we are speaking of the extension of the mass term M, we can paraphrase
it sm M.
Recall now the fist argument against treating mass terms as predicates -
that we cannot specify what it is that (say) is sm water if true of, other
than simply to say that it is true of whatever is sm water. However,
when put this way, the objection loses whatever force it once had. Compare
it with: we cannot specify that (say) is a man is true of other than
simply to say it is true of whatever is a man. And surely this objection
is off base - perhaps it is an interesting philosophical mater to find out
what being a man amount to, but it is absolutely clear that the philosopher
of language need not decide such a matter before he says is a man is
true of whatever is a man. And is merely this last that we need do in
giving a semantics for a language. This is perhaps a critical mark of
predicates as opposed to names - with a name it is essential for the seman-
tics to assign it a denotation. With a predicate, however, we need merely
indicate what things it is true of. It was in the inadequacies of satis-
factorily explaining what mass nouns allegedly named, that enabled us
to show the deficiencies in the previously-discussed proposals.
But still, our detractor might continue, is sm gold is true of so
many different sorts of things - nuggets, flakes, veins, watches, rings,
etc. - that it must be the case that we need some further information.
The tist answer to this is to point out that watches and rings are made of
gold; nuggets, veins, etc., are not. Secondly, an analogous objection
could be made in the non-mass case: is an animal is true of many
different things - species (The camel is an animal), breeds, individuals -
that we must need further information. But surely it is pointless to make
this objection here: animal individuates its reference into individuals;
the fact that other things can also be called animal is irrelevant. Gold
individuates in its own way (picks out a certain stuff), and the fact
that other things (nuggets, veins) can be called gold is irrelevant.
There are other advantages in interpreting certain terms as predicates
rather than as names. Aristotle long ago pointed out that if we interpreted
such terms as man and animal as naming objects (Forms) which are
distinct and not part of one another, then the most that can be said of the
relation between the two is that anything which partakes of the one
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 105

also partakes of the other. There is nothing we can add which will
make All men are animals be necessarily true. This is because when
we talk of two objects, X and Y, we have to have the relation be part of
in order for AI1 X is Y to be necessarily true. Aristotle points out that
such is not the case with predicates - here one looks to other criteria,
such as whether the extension of X must be included in the extension
of Y. And the same seems to be the case with mass nouns: we want to
avoid treating them as names so that All red ink is ink can straight-
forwardly be shown analytic by appeal to the extensions of ink and
red ink.
We want to give a rough-and-ready method of distinguishing exten-
sional from non-extensional uses of mass terms. If M is a mass term and
is used extensionally when not in subject position or when after an explicit
quantifier (but then grammatical considerations may force a change in
the form of the quantifier: all to any, etc.), then M can usually be
paraphrased by (is) 8m 44. Thus This puddle is water becomes This
puddle is sm water, John is eating cake becomes John is eating sm
cake which in turn becomes There is something which John is eating
and that is sm cake, All water is liquid becomes Anything which is
sm water is liquid (there are other ways to paraphrase this). When M
occurs without a quantifier in subject position, it most often is para-
phrased by Anything which is sm M, as when Water is wet becomes
Anything which is sm water is wet. There are cases though, where an
unquantified mass term in subject position is existential in its meaning:
Water is leaking through the crack becomes Something which is sm
water is leaking through the crack or when Water is found on Mars
becomes Something which is sm water is found on Mars (note the inter-
play at the quantifier some with the article s&).7
When these attempted paraphrases are not correct, we are not talking
extensionally. The second argument given above against identifying
mass nouns with predicates contains such a use. In The element with
atomic number 79 is gold, we cannot correctly paraphrase it as The
element with atomic number 79 is sm gold, for the former is true but the
latter false (or meaningless). Rather what is being asserted is that in the
actual world, two properties are true of the same entities. A paraphrase
might be Anything which is (entirely) made of the element with atomic
number 79 is sm gold. Thus we could demonstrate the deducibility
106 FRANCIS JBFFRY PELLETIER

of Xis made of gold, The element with atomic number 79 is gold, ergo x is
made of the element with atomic number 79 without requiring mass
nouns to be names. However, this paraphrase has the disadvantage that
it does not imply Gold is the element with atomic number 79 as the
original did. Perhaps we would want to handle this as a case of predicate
identity, but only identity in this possible world. Even if this line is taken,
it is not the same as introducing abstract objects in Parsons sense;
rather it is resorting to second order logic.
Complex mass terms can be handled in the same way: Dirty water is
bad to drink can be paraphrased Anything which is sm dirty water is
bad to drink. Now consider sentences such as Water is a liquid: this is
clearly a case of predicating the second-level predicate is a liquid of
an ordinary predicate water. (This is justified by noting the anomaly of
Anything which is sm water is a liquid. Of course one must distinguish
is a liquid from is liquid). We can also form such sentences as Dirty
water is a liquid, which again is a case of second-level predication. The
proposal put forth by Montague (1973), which is in many ways similar
to the one advocated here, differs on this point. As Montague notes, his
theory implies that complex mass phrases in subject positions must be
taken extensionally; i.e., it is always the case that we are talking about
the (physical) things of which (say) dirty water is true. Or, as we can
put it, it is alwars permitted to put is sm dirty water in its place. But
this makes such sentences as Dirty water is a liquid, Salt water is a
liquid, etc., ill-formed. Surely this is false: in English such sentences are
true; to preserve our truth-value sameness condition their translations
must also be true, not meaningless, in any adequate artificial language.

VI. EPILOGUE

I think I have adequately answered the difficulties raised by the identilica-

tion of mass terms with predicates, and also have shown what is wrong
with the other alternatives. But there may still be the question: Just how
different are these proposals? Especially, is there really all that much
difference between Cartwrights proposal and the extensional part of
yours? Is there really all that much difference between Parsons proposal
and yours? The answer is no. If one understands Cartwright in the
charitable way mentioned, that just is to understand mass terms as
 PROPOSALS FOR THE SEMANTICS OF MASS NOUNS 107

predicates restricted to one (the actual) possible world. The difference is

that predicates are not to be restricted to the actual world: they express
functions whose domain is the set of all possible worlds.8 The difference
with Parsons view is merely this: predicates as opposed to names. Of
course Parsons xQ& (where m is a mass term) will be true of exactly
the same things as x is sm M is. The problem with names is that a
semantics must state what they name; and if one is going to represent
predication by Q one is obliged to give an account of Q.
These are the points of difference. One might think that there are
many more points of similarity than difference, especially considering
how different the proposals appear on the surface. But, then, insofar as
we are all correct to some degree or other, that there are large areas of
agreement was to be expected.

University of Alberta
NOTES
* This paper is a summary of (a portion of) my dissertation Same Problems of Non-
Singular Reference: A Logic for Mass, Sortal and Adverbial Terms (UCLA, 1971). I
would like to thank the members of my committee for their help: Keith Donnellan,
Montgomery Furth, Barbara Hall Partee, and John Perry. This paper was also read at
the Canadian Philosophical Association meetings in Montreal, 1972. I would lie to
thank my commentator, Henry Laycock, for his insights. The nonexistence of many
mistakes is due to the (necessarily anonymous) referee. I thank him/her.
Added in proof. This paper was completed before the publication of J. Hintikka,
J. M. E. Moravcsik, and P. Suppes (eds.), Approaches to Natural Language, and thus
before the publication of Moravcsiks and Montagues articles on mass nouns (see
bibliography). In the preparation of this paper I had access only to the versions of
their papers as presented at the workshop in Stanford, 1970. There are a number of
differences between the two versions, especially in Moravcsik; however, the main
objections I wish to level are applicable to both versions. One should also see, in this
volume, the responses to Moravcsik by Cheng and Grandy (who gives an account
which, in some ways, resembles mine), and also Moravcsiks reply.
l Actually it is senses of nouns or noun phrases (or something like that) which are mass.
Consider chicken.
2 In W & 0 see pp. 97-98 and p. 101. In his review of Geach see p. 102. In Strawson
see p. 242. Of course, Quine, unlike Clarke and Strawson, only recommends this for
predicate occurrences and for complexes formed by a demonstrative plus mass terms,
not for subject occurrences.
28 H. Laycock Some Questions of Ontology, Phil. Rev. 81(1972), 3, and J. Bacon Do
Generic Descriptions Denote?, Mind 82 (1973), 331; both appear to subscribe to this
view of mass terms.
3 The example is from Montague (1973).
* The criticism is from Moravcsik.
108 FRANCIS JEFFRY PELLETIER

s These arguments are from Parsons, p. 364. Presumably Moravcsik and possibly
Qume would also assentto their correctness.
s For a further development of this notion of property, see the works of Richard
Montague.
7 Apparent definite descriptionsare discussedin Cartwright H, p. 481.
s And anyway the charitable view may not be Cartwrights at all.

BIBLIOGRAPHY
Cartwright, Helen; Hera&us and the Bath Water, Phil. Rev. 74 (1965) 466.
Cartwright, Helen: Quantities, Phil. Rev. 79 (1970) 25.
Clarke, D. S.: Mass Terms as Subjects,Phil. Studies 21 (1970) 25.
Rripke, Saul: SemanticalConsiderationson Modal Logic, Actu Philosophica Fennica
16 (1963) 83.
Lewis, David: General Semantics,Synthese 22 (1970).
Montague, Richard: English as a Formal Language, Linguaggi Nell0 Societo e Nella
Tecnica, Milan, Italy, 1970.
Montague, Richard: Universal Grammar, i%eoria 26 (1971) 68.
Montague, Richard: Comment on Moravcsik, in J. Hintikka, J. M. E. Moravcsik,
and P. Suppes(eds.) Approaches to Natural Language, Reidel, Dordrecht, Holland,
p. 289.
Moravcsik, Julius: MassTerms in English,in J. Hintikka, J. M. E. Moravcsik, and P.
Suppes(eds.),Approaches to Natural Language, Reidel, Dordrecht, Holland, p. 263.
Parsons,Terrance: An Analysisof Mass and Amount Terms, FoundationsofLanguage
6 (1970) 363.
Quine, W. V. : Word and Object, MIT Press,1960.
Quine, W. V. : Review of Geach Reference and Generality, Phil. Rev. 73 (1964) 100.
Strawson, P. F.: Individuals, Methuen 8c Co., 1959.