Editors:
VOLUME 14
MICHAEL J. LOUX
The University of Notre Dame, Notre Dame, Indiana
SUBSTANCE AND
ATTRIBUTE
A Study in Ontology
Loux, Michael J.
Substance and attribute
FOREWORD ix
EPILOGUE 181
in ontology. The third has taught me the importance of the historical dimen-
sion of the philosophical enterprise and the need for a systematic approach to
philosophical problems. He will most certainly disagree with almost every-
thing I say here; but I think he will see his own influence on the book and
will recognize that the views I argue for here are merely his own inverted.
M.J.L.
PART ONE
ATTRIBUTES
CHAPTER ONE
shirts exhibit not different yet resembling attributes, but rather the same
attribute. Thus (3) suggests that attribute-agreement involves different objects
exhibiting numerically one and the same attribute.
A consideration of each of (I)-(3), then, gives rise to three different con-
ceptions of what is involved in attribute-agreement; nor is the inclination to
generalize from (1 )-(3) off the mark. In confronting other cases of attribute-
agreement, we typically find that we can express the agreement by means of
sentences exhibiting the general form of (I )-(3), so that the question natural-
ly arises: which of these forms of sentence presents the most perspicuous
representation of attribute-agreement? To raise this question is to pose the
age-old problem of universals, and to answer it by appeal to sentences of one
of the three forms in question is to align oneself with one of three important
traditions in the dispute over universals. To accept sentences of form-(3) as
the paradigmatic devices for expressing attribute-agreement is to align oneself
with a tradition which, following standard philosophical usage, I shall call
metaphysical realism or Platonism. To adopt sentences of form-(2) as paradig-
matic, on the other hand, is to embrace a tradition I shall call nominalism,
and, finally, to hold that sentences of form-(!) present the most perspicuous
representation of attribute-agreement is to adopt a view I shall call extreme
nominalism. 1
As I shall understand it, then, metaphysical realism is the view that where
objects agree in attribute, there is at least one attribute which those objects
"have in common". Proponents of this view have employed a number of
different locutions for expressing this "commonality". Some have said that
objects participate, share in, or partake of a single attribute. Thus, in the
Parmenides, we find the younger Socrates agreeing that "there exist certain
Forms of which these other things come to partake and so to be called after
their names; by coming to partake of Likeness or Largeness or Beauty or
Justice, they become like or large or beautiful or just;"2 and the Russell of
the Problems of Philosophy uses similar language when he tells us that all the
things that agree in being just "partake of a common nature which will be
found in whatever is just and nothing else. This common nature, in virtue of
which they are all just, will be justice itself ... ,,3 Other realists, however, have
invoked a less picturesque terminology here. P. F. Strawson, for example,
tells us that objects agreeing in attribute all instantiate a single attribute;4
whereas philosophers like Gustav Bergmann and Alan Donagan prefer to
speak of the exemplification of attributes. S If we stick with the last of these
expressions, we can say that for the metaphysical realist attribute-agreement
is grounded in the multiple exemplification of attributes.
ATTRIBUTE-AGREEMENT 5
do not "have the same shape" in the sense in which two children "have the same father"
or two streets have the same manhole in the middle of their intersection, or two college
boys "wear the same tuxedo" (and so can't go to dances together).! 3
Finally, philosophers I shall call extreme nominalists deny the existence of
attributes, whether construed in Platonistic or nominalistic terms. As they
interpret it, the phenomenon of attribute-agreement presupposes the exist-
ence of no entities over and above those that are correctly said to agree. I
have said that extreme nominalists want to take sentences of form-(l) (e.g.,
'Socrates and Plato are both men' and 'Socrates and Plato are both wise') as
the paradigms for expressing attribute-agreement. This is correct, but it can
be misleading; for it might seem to suggest that for the extreme nominalist
to say that objects agree in attribute is to say that some sentence of form-( 1)
is true. Stated in these terms, however, the extreme nominalist's account
would be doomed from the start; for surely attribute-agreement does not
depend upon language. Objects can agree in their attributes even where lan-
guage lacks the predicate resources for expressing their agreement in terms of
sentences of form-(I).
In fact, not many extreme nominalists have stated their position in a way
that makes attribute-agreement depend upon the predicate-resources at our
disposal. On the contrary, most extreme nominalists have expressed their
view by saying that two objects agree in property because both are wise, both
are round, both are red, or both are ... ; and they have contended that the
fact that objects are wise, round, red, or ... is fundamental in that it does not
presuppose the existence of any additional entities. Likewise, they have said
that two objects agree in kind because both are men, both are dogs, both are
gerania, or both are ... ; and they have denied that objects are dogs, men,
gerania, or ... in virtue of any further entities. Finally, they have claimed
that pairs of objects agree in relation because in the case of each pair, one
object is the teacher of the other, the father of the other, the sibling of the
other, or ... ; and they have denied that it is necessary to appeal to additional
entities in explaining how one object can be the father of, the teacher of, or
the sibling of another.
Now, when we express the extreme nominalist's view in these terms, it
becomes clear that his view does not presuppose that language incorporates
the predicate resources for expressing every case of attribute-agreement in
terms of a sentence of form-(1). What he is saying is simply that objects agree
in attribute because of how they ar;;, what they are, and how they are related
to other things and not because they exhibit or exemplify abstract entities of
one sort or another; and he wants to deny that how objects are, what they
A TT RIB UTE-AG R EEM ENT 7
are, and how they are related to each other are merely linguistic facts. On
the contrary, he tells us that these are things that are expressed in language;
and while he concedes that language may not always incorporate sufficient
predicate devices for specifying how things are, what they are, and how
they are related, he insists that where it does, agreement with respect to
these various attributes is most perspicuously represented by sentences of
form-(I).
Historically extreme nominalism can be traced at least as far back as the
writings of William of Ockham; for although Ockham accepts a nominalistic
interpretation of qualitative attribute-agreement, he is unwilling to generalize
here and claim that every instance of attribute-agreement presupposes the
existence of attributes. While conceding that there are qualitative attributes -
things like the Aristotelian proper sensibles, the various virtues, and the vices,
he denies that quantitative attribute-agreement or relational attribute-agree-
ment is grounded ih uniquely quantitative or relational attributes. 14 Where a
pair of objects agree in shape, for example, Ockham insists that their agree-
ment consists exclusively in the fact that both are triangular, both are circu-
lar, both are rectangular, or ...... ; and he denies that their being triangular,
their being circular, or their being rectangular presupposes their exemplifying
abstract entities of any sort. As he repeatedly tells us, all that is required of
an object that it be triangular, circular, or rectangular is that its various parts
be arranged in the requisite way. Ockham's account of attribute-agreement,
then, represents a blend of what I have called nominalism and extreme nomi-
nalism. Purer versions of extreme nominalism can be found in some ofW. V.
Quine's early papers and in the writings of Wilfrid Sellars. In "On What There
Is," for example, Quine says:
One may admit that there are red houses, roses, and sunsets, but deny except as a
popular and misleading manner of speaking, that they have anything in common. The
words 'house', 'rose', and 'sunset' are true of sundry individual entities which are
houses and roses and sunsets, and the word 'red' or 'red object' is true of each of sundry
individual entities which are red houses, red roses, and red sunsets; but there is not,
in addition, any entity whatever, individual or otherwise, which is named by the
word 'redness' nor, for that matter, by the words 'househood', 'rosehood', 'sunsethood'.
That houses, roses and sunsets are all of them red may be taken as ultimate and irre-
ducible. ls
Sellars, on the other hand, insists that all talk about attributes can be reduced
to talk about persons as language-users; and he tells us that the use of a
predicate-expression over numerically different objects or n-tuples of objects
has its ontological ground simply in "how objects are.,,16
8 CHAPTER ONE
As I have indicated, metaphysical realists have insisted that there are; and
while they have appealed to a wide variety of facts in arguing this point (in-
cluding facts about the meaningfulness of language, the inter-subjectivity of
conceptual thinking, and the lawlikeness of the universe), the most popular as
well as the most powerful defenses of realism have hinged on facts of three
sorts - facts about predication, facts about resemblance, and facts about
abstract reference. Realists have argued, first, that we can account for the
truth of subject-predicate discourse only if we suppose that predicate-expres-
sions are referentially tied to multiply exemplifiable entities. Second, they
have argued that the applicability of the concept of similarity or resemblance
presupposes the existence of universals. Finally, they have contended that
since linguistic devices for referring to universals play essential roles in true
sentences, we have no option but to suppose that realism is true.
Now, the phenomena the realist appeals to in defense of his position are
carefully chosen. We have already seen how the extreme nominalist takes
sentences like 'Socrates and Plato are both wise' and 'Socrates and Plato are
both men' as paradigmatic representations of the ontological structure of
attribute-agreement. Such sentences, however, essentially involve the subject-
predicate nexus. In arguing that the truth of subject-predicate discourse
presupposes a realistic ontology, then, the realist seeks to undercut the ex-
treme nominalist's framework for understanding attribute-agreement. So far
from showing us that attribute-agreement fails to commit us to an ontology
of abstract entities, the realist is claiming, sentences of the extreme nominal-
ist's favored form - form-(l) - are such that their truth presupposes the
existence of universals. Likewise by arguing that the phenomenon of resem-
blance presupposes a realistic ontology, the realist attempts to undercut the
nominalist's analysis of attribute-agreement. The nominalist takes sentences
like 'The color of this shirt is just like the color of that one' to provide the
most perspicuous representation of attribute-agreement; but in arguing that
the truth of such sentences presupposes the existence of multiply exemplifi-
able entities, the realist attempts to show that the sentences the nominalist
takes as paradigmatic involve a commitment to the very universals he is seek-
ing to avoid.
Finally, in appealing to the phenomenon of abstract reference, the realist
seeks to show the incompleteness of the accounts provided by his opponents;
for even if we suppose them to be right in assuming that the phenomena of
predication and similarity fail to have the consequences the realist claims for
them, both the nominalist and extreme nominalist have to confront the fact
that sentences of the realist's favored form - sentences of form-(3) - enable
10 CHAPTER ONE
us to express the fact that objects agree in attribute. But such sentences incor-
porate expressions which, to all appearances, are devices for referring to
universals. What the realist argues is that abstract referring devices occur
essentially in these sentences; he holds that it is impossible to paraphrase
these sentences in such a way that the apparent reference to universals is
eliminated, and so he concludes that the truth of the sentences he takes to
provide the most perspicuous representation of attribute-agreement presup-
poses the existence of universals. 19
What the realist wants to claim, then, is that attribute-agreement is a phe-
nomenon which, viewed from any of its various perspectives, ultimately com-
mits us to the existence of multiply exemplifiable entities and, hence, to the
ontological framework he recommends. In the following three chapters, I
want to examine the realist's claim here in detail. In Chapter Two, I shall
examine the realist's contention that an ontology of universals can be defend-
ed by an analysis of predication; in Chapter Three, I shall examine arguments
that seek to establish the existence of universals on the basis of an analysis
of the concept of resemblance; and in Chapter Four, I shall examine the way
in which the phenomenon of abstract reference might be thought to ground
an ontology of universals.
NOTES
lOne might object that what has traditionally been called conceptualism has no place
in this scheme. Actually, traditional conceptualistic approaches to the problem of univer-
sals turn out to be versions of either nominalism or extreme nominalism on my account.
2 Parmenides, 130 E-131 A, translated by F. M. Cornford, p. 925 in Plato: The Collected
Dialogues, edited by Hamilton and Cairns, (New York: Pantheon Books), 1961. Here, I
must apologize to Plato-experts. I am fully aware that Plato does not use the expression
TO KCX(Jc'x"AOV; I am also aware there is a debate as to whether there is a doctrine of
"universals" in Plato; but however that debate is finally resolved, I think that the issues
Plato is concerned with in his discussion of Forms are quite close to (indeed, are the
ancestors 00 the issues at work in subsequent controversies over universals and that the
way he resolves those issues is at least a near relative of the way later metaphysical
realists tried to resolve them. So, realizing that there may be something of an anachron-
ism here, I shall classify Plato as a realist and take his theory of Forms (at least as it is
presented in the middle dialogues) to be a theory of universals.
3 Problems of Philosophy, (Oxford: Clarendon Press), 1912, p. 143.
4 'Particular and General,' Proceedings of the Aristotelian Society, 1953-4; reprinted in
my Universals and Particulars (Notre Dame, Indiana: University of Notre Dame Press),
1976; see especially pp. 85-86. Hereafter, where an article is included in this anthology,
I shall cite the page references as they are found there.
S See Donagan's 'Universals and Metaphysical Realism,' Monist, 1963; reprinted in
ATTRIBUTE-AGREEMENT 11
Loux, Universals and Particulars, p. 135. See also Bergmann's 'Logical Positivism, Lan-
guage, and the Reconstruction of Metaphysics,' The Metaphysics of Logical Positivism
(Madison: University of Wisconsin Press), 1954, p. 52.
6 I do not mean to suggest that these are the only categories of attributes that meta-
physical realists have insisted on recognizing. Realists have sometimes argued for the
irreducible status of other types of multiply exemplifiable attributes including things like
actions, processes, and states or conditions. They may be right, but since the three cate-
gories of attributes I have listed here provide us with enough problems, I shall not con-
cern myself in what follows with the question whether the domain of attributes has to
include more categories than these three.
7 As with realism, there are problems about the number of distinct categories of attri-
butes the nominalist will recognize. Some have recognized only properties and have
insisted that other forms of attribute-agreement are to be analyzed in terms of the
phenomenon of property-agreement; whereas, other nominalists have admitted the
existence of both properties and relations. The notion of kind-agreement will turn out to
cause special problems for the nominalist. Indeed, it will be a central contention of
Chapter Four that no nominalist can adequately handle this notion.
8 The terms 'abstract particular' and 'trope' are used by D. C. Williams in 'On the Ele-
ments of Being, I,' Review of Metaphysics, 1953, p. 7.
9 The terms 'aspect' and 'case' are coined by Nicholas Wolterstorff. See On Universals
(Chicago: University of Chicago Press), 1970, p. 89 and pp. 133-134.
10 In some places (e.g., Metaphysics Z.16 (1040 b 25-27), Aristotle suggests that this
sort of approach is correct; in other places, however, he seems to accept a more Platonis-
tic interpretation of attribute-agreement. Anyone who has tried to become clear on
Aristotle's theory of universals will agree that it is difficult, if not impossible, to identify
a single account in all of Aristotle's remarks on the topic.
11 Or at least for most of the contents of this category. With regard to figure and form,
Ockham accepts an account of attribute-agreement that is pretty clearly of the extreme
nominalist sort. I do not find this too surprising since predicates expressing figure and
form are more closely related to quantitative predicates than they are to your run-of-
the-mill qualitative predicates. See Chapters 5 and 55 in the Summa Logicae, I, pp.
56-58 and pp. 178-180 in my Ockham's Theory of Terms (Notre Dame, Indiana:
University of Notre Dame Press), 1974.
12 Theory of Knowledge (London: Hutchinson University Library), 1949, p. 95.
13 'On the Elements of Being, I,' pp. 5-6.
14 See Chapters 6-8 of Part I of the Summa Logicae, pp. 50-64 in my Ockham's
Theory of Terms.
15 'On What There Is' in From a Logical Point of View (Cambridge, Mass.: Harvard
University Press), 1954, p. 10.
16 See, e.g., Science and Metaphysics (London: Routledge and Kegan Paul), 1967,
p.l07.
17 A familiar claim here is that the notion of multiple exemplification is incoherent.
This claim is first made in Plato's Parmenides (131 A-E), where Parmenides argues that
the multiple exemplification of or participation in Forms can be understood in either
of two ways. (1) The realist can say that the different objects exemplifying a single Form
each partake of a distinct part of one and the same Form; or (2) he can claim that each
of them partakes of the Form wholly and completely. Parmenides rightly objects to the
12 CHAPTER ONE
first interpretation of multiple exemplification on the grounds that Forms are not the
sorts of things that have parts; but he also claims that the second account is unsatis-
factory telling us that if a Form is wholly and completely present in distinct objects, it
will be divided from itself. But will it? Well, only if it is impossible for a single object to
be wholly and completely present in numerically distinct objects. But it should be clear
that Parmenides cannot invoke this principle in arguing against the realist; for to suppose
that realism is true is just to suppose that the principle is false.
18 Most contemporary philosophers would agree with me here. The only exception I
know of is Nicholas Wolterstorff. While he seeme to think that all three of the views I
have outlined here are coherent, he denies that the principle of simplicity serves to
adjudicate the dispute between proponents of those views. See, e.g., p. 127 of his On
Universals.
19 The realist will, in the end, reject the claim that talk about sentences is rich enough
to enable us to make sense of semantical properties like truth, falsity, analyticity, and
syntheticity. He will opt for the framework of propositions. But since what is at issue in
the debate is the existence of abstract entities of any sort, he can hardly employ that
framework in establishing his position. He must, then, make his case by talking about
sentences. I shall follow him in this. Thus, the first four chapters of this book are all
couched in the language of sentences. In Chapter Five, however, I adopt the framework
of propositions and employ it for the remainder of the book, except for a brief discus-
sion of subject-predicate discourse in Chapter Seven. There, the philosophical literature
on the bundle theory forces me to employ the framework of sentences in handling a
certain difficulty confronting that theory of substance.
CHAPTER TWO
13
14 CHAPTER TWO
world. And certainly this consideration is weighty. If the ultimate non-logical and non-
formal constituents of true propositions refer to nothing in the world, in what can the
truth of such propositions consist?8
account of the matter. They would, I think, all agree that unless we construe
predicates as devices for referring to universals, we cannot explain "how
propositions containing ... predicates"lO enable us to "state facts about the
world." This is a strong claim;' for what the realist is claiming is that he and
he alone has the resources for adequately explaining the truth of subject-
predicate discourse.
Before we can evaluate this claim, however, we have to get clearer on just
what is involved in a realistic interpretation of predication. According Dona-
gan, the realist is claiming that a predicate-term has a single universal as its
referent in the various subject-predicate sentences in which it is functioning
predicatively; and he is telling us that where one of these sentences is true,
the relevant universal is an attribute which the referent of the sentence's
subject-term exemplifies. But if this is what the realist is claiming, then his
interpretation of predication incorporates two theses. The first (I shall call it
(I) is the view that where a predicate-term 'F', can be truly applied to all and
only the objects, a . .. n, there is some universal, U, which all and only a ... n
exhibit or exemplify; the second (II) is the view that when a predicate-term
functions predicatively in a true subject-predicate sentence, it serves to pick
out or refer to the universal, U, which is exhibited by all and only the objects
of which it is truly predicable. Since (II) presupposes the truth of (I), let us
begin by examining (I).
(I), I have said, is the view that where a predicate-term is truly predicable of
all and only the objects, a . .... n, there is some universal exemplified by all
and only a . .... n. Two preliminary remarks are necessary here. First, as I
have stated it, (I) is a general claim that applies to all predicate-terms; but the
fact is that some proponents of a realistic interpretation of predication (e.g.,
Russell, Bergmann, and possibly Donagan) have insisted that (I) be restricted
to the case of primitive as opposed to defined predicates. 11 I am sticking with
the generalized version of (I) partly because I have doubts about the semantic
atomism that underlies the proposed restriction on the thesis. I do not fmd it
obvious that predicate-expressions are susceptible of a neat division into those
that are primitive and those that are defined or that if such a division is
possible, there is just one way of drawing it so that the division is any more
than system-relative. But even if it were to turn out that there is a sharp and
absolute distinction between the primitive and the defined, I am inclined to
think that it would be perfectly harmless to speak of universals as correspond-
16 CHAPTER TWO
ing to defined predicates. If we were to suppose, for example, that 'red' and
'circular' are primitive' predicates, we could, I think, still speak of the univer-
sal that is common to all and only the things to which the defined predicate
'red and circular' applies. We might, of course, want to claim that that univer-
sal is, in some sense, reducible to the universals corresponding to the predi-
cates 'red' and 'circular'; but I find this perfectly consistent with the general-
ized version of (I); for that thesis does not tell us that there must be an
irreducible or unanalyzable universal corresponding to every predicate-term,
but only that there must be some universal or other. 12
Second, whether (I) be taken in a general or restricted form, it should not
be confused with a quite different claim about predicate-terms, the claim that
a speaker's ability to apply predicate-terms correctly is grounded in his ability
to recognize in objects the presence of the universal that (I) tells us is exhibit-
ed by all and only the things of which that expression is truly predicable. This
claim is an epistemological thesis about speaker-competence, and it is likely
false; for it is plausible to think that, in general at least, the ability to classify
objects according to a predicate-term is prior to the ability to identify the
universal that all and only those objects exemplify. Proponents of (I) have, of
course, sometimes conflated (I) with this claim about speaker-competence;
but as I am understanding it, (I) makes no claim at all about the kind of
knowledge involved in a speaker's ability to use predicate-terms. (I) is a meta-
physical rather than an epistemological claim, the claim that where 'F' is a
predicate-term, there is a universal exemplified by all and only the objects
that are F. 13
But even when it is disentangled from the epistemological claim about
speaker competence, (I) is likely to appear false. It might be thought, for
example, that the phenomenon of predicate-ambiguity tells against (I). Thus,
the term 'bat', can be predicated of things of two quite different sorts - cer-
tain flying rodents and the wooden sticks used in playing baseball; but there
is no single universal that underlies the applicability of the term to entities of
both sorts. Likewise, the term 'mole' is truly predicable of all the members of
a certain species of burrowing mammals; it is also truly predicable of certain
congenital protuberances of the skin, but no single universal grounds the use
of the term in the two cases.
But do ambiguous predicates like these really tell against (I)? That depends
on whether we want to characterize predicate-ambiguity by saying that one
and the same term has several meanings or by speaking of different expres-
sions as embedded in one and the same phoneme or string of phonemes. It
seems to me that the question calls for a decision rather than an answer; but
PREDICATION AND UNIVERSALS 17
even if we decide upon the first alternative, we can reformulate (I) in such a
way that the phenomenon of predicate-ambiguity poses no problem for the
realist. We can express the insight at work in (I) by saying that where a predi-
cate-term applies in one and the same sense to all and only the objects, a . .. n,
there is a single universal, U, such that all and only a ... n exhibit U.
A rather different problem for (I) is posed by a predicate-term suggested
by some remarks of Alvin Plantinga. 14 Plantinga suggests that we might
introduce into English the predicate 'sizeable' and define it as follows:
_ _ _ _ is sizeable = df. ' _ _ _ _ ' has more than six letters.
Given the definition, Richard Milhouse Nixon is sizeable since 'Richard
Milhouse Nixon' has more than six letters; but Pele is not since 'Pele' does not
have more than six letters. Now, when we who know the ex-president only as
a public figure employ Plantinga's predicate in
(2) Gerald Ford is sizeable,
what we say is true; but when Mrs. Ford invokes the predicate in
(3) Jerry is sizeable,
what she says is false. The difficulty, however, is that the person Mrs. Ford
refers to in (2) is the same person we are referring to in (3); but, then, if we
stick with (I), we are forced to hold what is patently contradictory - that the
universal associated with 'sizeable' is simultaneously both exemplified and
not exemplified by one and the same person.
Now, it should be clear how this apparent counter-example to (I) is to be
handled. As Plantinga points out, the contradiction in question arises only if
we suppose that his definition of 'sizeable' has the effect of creating a predi-
cate genuinely applicable to non-linguistic objects;15 but it should be apparent
that it does not. All the definition does is provide us with a convention for
rewriting sentences like
(4) 'Gerald Ford' has more than six letters
without explicitly invoking the convention of quotation. The predicate 'size-
able', then, is only apparently predicated of the same objects in (2) and (3).
In fact, the use of the term in the two sentences has the effect of quoting the
terms ('Gerald Ford' and 'Jerry') with which it appears; and the predicate-
term which it abbreviates, quite compatibly with (I), expresses a single ortho-
graphical universal which 'Gerald Ford', but not 'Jerry" exemplifies.
Some remarks of Quine point to a family of predicates which appear to be
18 CHAPTER TWO
like 'sizeable' but cannot be handled in precisely the same way.16 Quine's
example is the predicate 'so called because of his size'. Now, the Italian
Barbarelli was apparently of diminutive stature and, consequently, was called
by the diminutive 'Giorgione', so that
(5) Giorgione was so-called because of his size
is true, but
(6) Barbarelli was so-called because of his size
is false. However, since Barbarelli was Giorgione, the proponent of (I) once
again appears to be committed to endorsing a contradiction. This time, how-
ever, the strategy of taking the problematic predicate to be metalinguistic will
not work; for in some sense (5) and (6) are about non-linguistic entities. The
contrast with the case of 'sizeable' comes not when we note that since (2) is
true just in case (4) is true, (2) comes out true even in the case where there is
no one named Gerald Ford. (5), on the other hand, can be true only if some-
one was actually called Giorgione. But while (5) and (6) do involve a reference
to one and the same non-linguistic object, the Italian in question, it would be
wrong to think that the predicate 'so called because of his size' is being predi-
cated of that individual taken by himself. The fact is that 'so called because
of his size' is a relational predicate which applies to objects taken in pairs.
One of the objects from the pair is a non-linguistic entity; the other is a lin-
guistic expression. What makes the sentences in which the predicate appears
so puzzling is that we use one and the same linguistic expression to refer to
the entities, both linguistic and non-linguistic, which make up these pairs; but
it is easy enough to paraphrase these sentences in such a way that their depth
grammar becomes perspicuous. Thus, (5) becomes
(7) Giorgione was called by the name 'Giorgione', and he was called
by that name because of his size;
whereas, (6) becomes
(8) Barbarelli was called by the name 'Barbarelli', and he was called
by that name because of his size.
But, then, the fact that (5) is true while (6) is false provides no consolation
for the opponent of {I); for the common predicate here is only apparently
predicated the same entity in the two cases. In fact, in (5) it is predicated
of Giorgione and 'Giorgione' taken as a pair; and in (6), it is predicated of
Giorgione and 'Barbarelli' taken as a pair. While the predicate truly applies in
the case of the first pair, it does not apply in the case of the second.
PREDICATION AND UNIVERSALS 19
So far our discussion of (I) has focused on the exceptional case where a
predicate-term is ambiguous and on exotic predicates like 'sizeable' and 'so
called because .. .'; but in the writings of the later Wittgenstein, we meet with
the contention that (I) is incapable of handling even the most ordinary predi-
cate-expressions. Thus, in the Philosop~ical Investigations, Wittgenstein insists
that if we approach the issue with no antecedent philosophical prejudices,
we find that there is no single universal or set of universals exhibited by
all and only the objects of which the expression 'game' is truly predicable.
Wittgenstein says:
Consider for example the proceedings that we call "games". I mean board-games, card-
games, Olympic games and so on. What is common to them all? Don't say: "There must
be something common, or they would not be called 'games' " - but look and see whe-
ther there is anything common to all. - For if you look at them you will not see some-
thing that is common to all, but similarities, relationships, and a whole series of them at
that. To repeat: don't think but look! Look, for example, at board-games, with their
multifarious relationships. Now, pass to card-games; here you find many correspondences
with the first group, but many common features drop out, and others appear. When we
pass next to ball-games, much that is common is retained, but much is lost. -- Are they
all 'amusing'? Compare chess with noughts and crosses. Or is there always winning and
losing, or competition between players? Think of patience. In ball-games there is winning
and losing; but when a child throws his ball at the wall and catches it again, this feature
has disappeared. Look at the parts played by skill and luck; and at the difference be-
tween skill in chess and skill at tennis. Think now of games like ring-a-ring-a-roses; here
is the element of amusement, but how many other characteristic features have dis-
appeared! and we can go through the many, many other groups of games in the same
way; can see how similarities crop up and disappear. 17
The claim, then, is that when we examine the set of objects correctly called
games, we find no single universal or group of universals which cuts across the
set. What we find, on the contrary, is a "complicated network of similarities
overlapping and criss-crossing" (what Wittgenstein goes on to call "family
resemblances"); it is this network of similarities, Wittgenstein is claiming,
rather than a single universal that grounds the use of the term 'game'.
Confronted with these remarks, one might think that the best course for
the proponent of a realistic interpretation of predication is simply to concede
Wittgenstein his counter-example to (I) and to try to find some way of re-
stricting the scope of (I). Indeed, since 'game' is pretty clearly not a primitive
predicate, it might be thought that a plausible strategy here is to follow
philosophers like Bergmann and Russell in distinguishing between primitive
and defined predicates and to claim that while Wittgenstein has succeeded in
showing that there need be no single universal underlying the use of predicates
20 CHAPTER TWO
of the latter sort, predicates of the former sort are immune to the Wittgen-
steinian critique. Such a strategy is, however, misguided on two counts. First,
part of what Wittgenstein means to challenge by these remarks is the sort of
atomism that Russell and Bergmann embrace - the view that every predicate
is either primitive or capable of being defined in terms of primitives. Second
and more importantly, a line of argument parallel to that presented in the
case of 'game' could be invoked to call into question the applicability of
(I) to the very terms that philosophers like Bergmann and Russell take to
be primitive. Color-words like 'red' and 'blue' are usually construed as para-
'digmatic examples of primitive or undefined expressions; but as Renford
Bambrough points out in criticism of a similar move on the part of A. J. Ayer,
Wittgenstein would deny that there is any single thing which all and only red
objects have in common. IS Bambrough's point is that since 'red' marks off a
fairly wide range of the spectrum, red objects can literally differ in color. Nor
will it do to insist that it is the predicates marking the different shades of red
rather than 'red' that count as primitive predicates; for the same form of argu-
ment applied in their case. Terms like 'crimson', 'scarlet', 'vermillion', and
'burgundy' do not mark off a discrete point on the spectrum; like 'red', they
apply within a range. Now, the point here is not that it is impossible for dif-
ferent objects to exhibit a color at one and the same point on the spectrum.
Bambrough's point I take it, is rather that given the limited stock of color-
predicates in our language, it is extremely unlikely that the proponent of even
a very restricted version of (I) will be able to identify predicates that corres-
pond exactly with the cases where this happens.
Thus, if one finds Wittgenstein's remarks about the predicate-term 'game'
impressive evidence against (I), he would be wrong to think that he could
preserve a realistic interpretation of predication by restricting that principle
to the case of primitive predicates. But do Wittgenstein's remarks about
'game' and Bambrough's remarks about color-words really count as evidence
against (I)? I am not convinced that they do; for while I think that Bambrough
is right in pointing to the divergences among red objects, I am, nonetheless,
inclined to think that the realist could plausibly argue that there is a property
common to all and only red objects, the property of having a color within a
certain range at the lower end of the spectrum. This, I take it, is just what the
property of being red is. Likewise, it seems to me that despite the veryexhaus-
tive inventory of attributes that might be common to all games, Wittgenstein
has overlooked the most obvious candidate here - the property of being a
game; for it is surely plausible to suppose that all and only the objects of which
'game' is truly predicable exemplify this property. But perhaps Wittgenstein
PREDICATION AND UNIVERSALS 21
would ask us what this property amounts to. One plausible answer, I take
it, is that the property of being a game is a disjunctive property, a property
formed from the disjunction of all of those properties his very detailed
reflections point to. Wittgenstein rejects this answer, telling us that in sug-
gesting this account of the property of being a game, we "are only playing
with words ;,,19 but just how are we playing with words here? If the property
of being a game actually is a disjunctive property of the sort just specified,
then what grounds for complaint are there if the proponent of (I) points
this out?
seems, I suppose, to be a single character which is the same when you look at them all;
hence you think that Largeness is a single thing.
Socrates: True.
Parmenides: But now take Largeness itself and the other things which are large. Suppose
you look at all these in the same way in your mind's eye, will not yet another unity
make its appearance - a Largeness by virtue of which they all appear large?
Socrates: So it would seem.
Parmenides: If so, a second Form of Largeness will present itself over and above the
things that share in it; and again covering all these, yet another, which will make them
large. So each of your Forms will no longer be one, but an indefinite number. 24
This passage has given rise to more critical commentary than any text in
Plato. Since I have no intention of adding to the very extensive literature on
this passage, I shall limit myself to a few comments about the relationship
between this argument and what I have called (I).
In this interchange, Socrates serves as the spokesman for realism and
Parmenides, its critic. Parmenides' opening remarks make it clear that some-
thing like (I) is at issue here; and the application of (I) that is envisioned bears
on the predicate-term 'large'. What the realist is supposed to be claiming is
that ordinary objects are called large because they one and all exemplify
Largeness. Parmenides, however, presumably wants to claim that Largeness
is itself large; and he takes this to be a reason for supposing that the realist
has to appeal to a second Largeness, a Largeness in virtue of which all ordi-
nary large things as well as the Largeness they exemplify can be said to be
large. But if the first Largeness is large, the second Largeness will also be
large, so that the realist must appeal to a third Largeness if he is to explain
how it is that objects can satisfy the predicate 'large'; and obviously there will
be no end to the Largenesses he must appeal to here; consequently, his use of
(I) can never do what it is supposed to do - give us a final explanation of
why objects are called large; and, of course, the argument here is perfectly
general, so that no application of (I) can accomplish what it is supposed to
accomplish.
It might seem that this argument can be blocked if we deny that the first
Largeness the realist appeals to is itself large. It might seem, that is, that it is
only if we suppose what is not the case - that universals are selfpredicable
(i.e., are such that the predicates whose use they serve to explain are truly
predicable of them) - that the infinite regress Parmenides points to actually
arises. Now, for many universals, it is true that they are not in the sense just
indicated self-predicable. If there is such a thing as Largeness, then, whatever
24 CHAPTER TWO
Plato may have thought, it is not itself large; neither is courage, courageous;
trangularity, triangular; nor mankind, a man. Nevertheless, some universals
are self-predicable. Thus, the property of being self-identical is self-identical;
the property of having no color has no color; and the property of being
colored if green is itself colored if green. Thus, while Parmenides may be
wrong in supposing that the infinite regress he points to arises for all appli-
cations of (I), there surely appear to be applications of (I) that give rise to the
difficulties he mistakenly reads into the application of (I) to the predicate-
term 'large'.
Even here, however, the infinite regress Parmenides points to arises only if
we assume that where a predicate is applicable to all and only the entities,
a . .. n, there is some entity outside the series a . .. n which all of those ob-
jects exemplify. Now, that assumption surely holds in the case where the
predicate-term whose applicability is to be explained is not predicable of the
universal which underlies its applicability; but can the realist not claim that
this assumption fails to hold in the case of a self-predicable universal? Can he
not say, that is, that where a universal, U, is self-predicable, the predicate-
term in question applies to U in virtue of the fact that U exemplifies itself?
It seems to me that he can. 1 find it perfectly plausible to think that the
predicate 'has no color' applies to the property of not having any color not
because that property exemplifies some universal distinct from itself, but
simply because it exemplifies itself. Likewise, 1 find it plausible to suppose
that the predicate 'cQlored if green' is truly predicable of the property of
being colored if green not because that property exemplifies something else
but simply because it is self-exemplifying. But if it is plausible to suppose that
self-predication is grounded in self-exemplification, then the proponent of (I)
can avoid the regress Parmenides points to even in the case of universals that
are self-predicable. 25
There are, however, other ways in which (I) might be thought to land its
proponent in an infinite regress. Thus, (I) tells us that the applicability of a
predicate-term is grounded in the exemplification of a universal. Let us sup-
pose (I) to be applied in the case of some arbitrary predicate-term 'F'. (I) tells
us that the objects, a . .. n, of which 'F' is truly predicable all exhibit some
universal. Call that universal F-ness. Now, if i~ is true that each of a . .. n
exemplifies F-ness, then the predicate 'exemplifies F-ness' is truly predicable
of each of a ... n; but given (I), this presupposes that a ... n exemplify some
new universal, that of exemplifying F-ness. Of course, the exemplification by
each of a ... n of this new universal allows us to affirm of each of a ... n yet
another predicate-term; and that this third term is truly predicable of each of
PREDICATION AND UNIVERSALS 2S
a . . . n presupposes, in turn, that each of them exemplifies yet another univer-
sal; and so on ad infinitum. The analysis to which (I) gives rise never ends;
each universal introduces a new predicate-term; and that, in turn, requires the
postulation of still another universal. Conclusion? (I) must be false since it
cannot be applied without landing its proponent in an infinite regress.
The realist can, however, respond to this infinite regress argument in either
of two ways. First, he can concede that this regress is real but claiming that
it does not render his explanation of the applicability of any particular predi-
cate-term impossible, he can deny that the regress is vicious. The point here is
that the regress would be vicious if it were impossible to explain the appli-
cability of a predicate-term without pointing to every member in the series
of intrusive universals; but the realist can deny that anything of the sort is
necessary here. He can concede that when we ground the applicability of a
particular predicate-term by appeal to a universal, we do introduce a new
predicate-expression, but he can deny that our failure to ground the appli-
cability of this new predicate-term does anything to threaten our account of
the applicability of the original predicate-expression. We can, he will agree,
go on and explain the applicability of this new predicate-expression; but he
will claim that we can also discontinue our account without invalidating what
has gone before.
But if, on grounds of theoretical simplicity, he finds this response unsatis-
factory, the realist can deny that the regress in question is real. While agreeing
that at each successive stage in the explanation we can formulate what
appears to be a new predicate-term, he can insist that each new expression
differs only syntactically or grammatically from those predicate-expressions
which precede it in the series. He can say, that is, that while the predicate-
term 'exemplifies wisdom' is structurally different from the predicate 'is
wise', the two are semantically indistinguishable; and, then, he can claim that
their applicability is grounded in one and the same universal, thereby stopping
the regress before it gets started. 26
Before concluding this section, I want to consider one last regress that
might be thought to be implied by (I). What (I) tells us is that a predicate-
term 'F' is truly predicable of an object, a, only if a and the universal under-
lying the use of 'F' (call it F-ness) are related in a certain way; but presum-
ably the proponent of (I) wants to deny that a and F-ness are related of and
by themselves; for what he tells us is that it is only because of a's exemplifi-
cation of F-ness that 'F' is truly predicable of a. It should be clear, however,
that exemplification can bring it about that a and F-ness are related in the
requisite way only if it is itself related to each of a and F-ness. But if a and
26 CHAPTER TWO
F-ness can be related only through the mediation of some third entity, then
the fact that a, F-ness, and exemplification are related presupposes the exist-
ence of some fourth entity; and this entity (call it exemplification2) can bring
it about that a, F-ness, and exemplification are related in the requisite way
only if it is related to those objects in a certain way. That, in turn, requires
the existence of exemplification3, which can do its job only in virtue of
exemplification4; and so on ad inifinitum. But obviously the regress here is
vicious; and that means that our original objects, a and F-ness, will never turn
out to be related in the way required, so that, given the assumptions built
into it, (I) cannot be employed in explaining why any arbitrary predicate-
term is truly predicable of any arbitrary object. 27
Now, initially one might want to agree that this third infinite regress is
vicious; but Nicholas Wolterstorffhas recently given reasons for thinking that
it is not. Focusing on the case of relational universals, he compares this in-
finite regress argument with Zeno's famous arguments for the impossibility
of motion. Wolterstorff says:
Zeno already noticed that the movement from one place to another can also be made
to look mysterious. Before one can go to B, one must go half the distance to B; but to
do this, one must first go half that distance; and so on. But of course there is no incom-
patibility here. One can consistently hold both that space is infinitely divisible and that
we sometimes move. One need not deny one or the other of these. So too John can love
Mary, even though in so doing, he stands in the relation of loving to Mary, and he and
Mary stand in the relation of R to loving, and he and Mary and loving stand in the rela-
tion of R' to R, and so on ad infi1litum. In short, I see no incompatibility between the
claim that things are related, and the principle that for every relation, if some entities are
to be in that relation, those entities must be in a certain relation. 28
Of course, some proponents of (I) may not find Wolterstorffs analogy con-
vincing; but if they do not, then they can respond to this third infinite regress
argument by denying that exemplification is a relation. While agreeing that
there is such a thing as exemplification serving to bind objects to universals,
the proponent of (I) can deny that an object's exemplifying a property by
possessing it, an object's exemplifying a kind by belonging to it, or an n-tuple
of objects' exemplifying a relation by entering into it is'a relational fact. It
turns out that most realists have taken just his line. Claiming that exemplifi-
cation does serve to link, connect, or bind, realists have insisted that the link
effected by exemplification differs from relational links in that the latter,
but not the former, are capable of linking objects only by the mediation of
some additional entity, Gustav Bergmann has marked this contrast by distin-
guishing between relations and what he calls nexus, exemplification being a
PREDICATION AND UNIVERSALS 27
There are doubtless other objections which could be raised against (I); but its
success in meeting the broad range of objections I have considered suggests
that (I) would withstand further criticism. (I), it is reasonable to conclude,
provides a coherent account of the applicability of predicate-terms. Let us,
then, examine (II). (II) tells us that the predicates of subject-predicate sen-
tences refer to the universals that underline their applicability; but just how
are we to understand this? According to Gustav Bergmann, we are to suppose
that predicates refer to universals in the sense that they name them. As we
have seen, Bergmann wants to restrict (II) to the case of primitive or unde-
fined predicates; and he tells us that ordinary color-predicates count as unde-
fined predicates. Thus, Bergmann tells us that where
(9) This is red
is a true subject-predicate sentence specifying the color of some object, say, a
spot, the sentence must be construed as incorporating two names. 'This', he
tells us, serves to name the particular spot in question; whereas, 'red' names
the universal it exhibits. The copula 'is', he tells us, is not a name; but with-
out naming, it expresses the fact that the particular in question exemplifies
the universal named by 'red'. 31
Now, in case of the predicate 'red', it is not implausible to suppose that we
have the name of a universal, but when we tum to sentences like
(10) This is triangular,
Bergmann's interpretation of (II) appears less satisfactory. The difficulty here
is that where an expression names an object, the expression can play the role
of logical subject and, in that role, serves to pick out the object it names. But
while 'red' might seem to conform to this general condition on naming (e.g.,
'Red is a color'), 'triangular' does not. 'Triangular' is not syntactically suited
to play the role of logical subject. The related term, 'triangularity' can, of
course, play that role; and when it does, it serves to pick out the universal, if
28 CHAPTER TWO
referring to universals are abstract nouns like 'circularity' and 'wisdom', and
he contends that if predicates are devices for referring to universals, they
ought to be intersubstitutable salva veritate with the abstract nouns that
correspond with them. 32 My remarks about predicate-terms and names show
that predicate-terms cannot be substituted for abstract nouns when these are
functioning as subject-terms. Searle argues the converse, that abstract nouns
cannot be substituted for predicate-expressions. He considers only the view
that it is 'is circular', for example, that is the predicate-term of
(11) This is circular
and points out that the result of substituting 'circularity' for this expression
in (II) is not a true or false sentence, but rather the mere list
(12) This circularity. 33
Of course, we have a corresponding failure of substitutivity on the Bergmann-
ian analysis of (11) according to which it is 'circular' taken by itself that
functions as the predicate of (11). On the analysis, the substitution of 'cir-
cularity' for the term 'circular' yields
(13) This is circularity.
Now, unlike (12), (13) is a sentence with a truth value; but if we assume that
(11) is true, the substitution resulting in (13) does not preserve truth. (13) is
not a subject-predicate sentence; it is a sentence expressing identity; and if
the referent of 'this' is invariant over (11) and (13), then (11) is true only if
(13) is false. Since the universal circularity is pretty clearly not self-predicable,
(11) can be true only if the referent of 'this' as it appears there is something
other than circularity; whereas, (13) can be true only if the referent of 'this'
as it appears in that sentence is identical with the referent of 'circularity'.
Searle takes the failure of substitutivity here to be a serious difficulty for
the proponent of (II); and he tells us that the defender of that thesis can
respond to the difficulty in one of two ways, neither of which Searle finds
satisfactory. He can deny that the referent of a predicate-term can ever be the
referent of a singular term, or he can claim that "the sense of 'refer' ..... is
different for predicates from what it is for singular referring expressions.,,34 I
think that we can agree with Searle that the first strategy is unsatisfactory.
Frege is, of course, the foremost proponent of this strategy. He tells us that
there is a categorial difference between the things that can serve as the refer-
ents of predicate-terms and the things that can function as the referents of
Singular terms or names. He calls predicable entities concepts and nameable
30 CHAPTER TWO
term has the effect of introducing into discourse. The latter, he wants to say,
is a universal. 39
All three of these philosophers are, I think, proposing that we accept
something like the second strategy Searle points to. While endorsing (II), all
three want to claim that the referential force of predicates differs in impor-
tant ways from the referential force of the abstract referring devices corres-
ponding to them. For these philosophers, an abstract singular term like
'wisdom' enters into a single referential relation; 'wisdom' is referentially tied
to wisdom and nothing else; they want to claim, however, that the predicate-
term 'wise' /,is wise' must be construed as entering into two different kinds of
referential ties; one of these ties links the term with the various objects of
which it can be truly predicated - the various wise men; whereas, the other
links 'wise'/'is wise' with wisdom; and presumably all three would want to
claim that it is because of this difference in referential force that 'wise' /'is
wise' and 'wisdom' are not intersubstitutable.
Searle, however, claims that this sort of view leaves the referential force of
predicates unexplicated. The fact that these philosophers say as much as they
do suggests that perhaps Searle's charge of obscurity is a bit exaggerated; but
obviously Searle will deny that they have been satisfactorily cleared of the
charge unless we can give sense to the idea that predicates stand in some sort
of referential relation with universals. I am inclined to think that .we can.
Consider
(IS) Socrates is courageous
(15) is a subject-predicate sentence; but it can be paraphrased in a perfectly
natural way as
(16) Socrates possesses courage.
Ukewise, the subject-predicate sentence
(17) Socrates is a man
can be paraphrased as
(18) Socrates belongs to the kind man,
and
(19) Socrates is the teacher of Plato
can be paraphrased as
(20) Socrates bears the relation of being the teacher of with respect to
Plato.
32 CHAPTER TWO
sion incorporating singular terms which, to all appearances, are devices for
referring to universals. 41
Thus, the proponent of (II) can make the point that predicates are referential-
ly tied to universals by saying that while they are satisfied by or true of the
objects to which they apply, they also express universals. (II), then, is subject
to a plausible formulation, but since (I) appears to be a plausible account of
the applicability of predicate-terms, the conjunction of (I) and (II) seems to
provide a plausible account of predication. As we have seen, however, the
realist wants to make a stronger claim for his account. He claims that the
analysis of predication provided by (I) and (II) is the only adequate account
of subject-predicate discourse. He wants to insist that truth is grounded in a
correspondence between language and the world. Since he contends that the
relevant correspondence presupposes that the non-logical constitutents of
true sentences be referentially tied to objects out in the world, he claims that
we cannot explain how subject-predicate sentences can be true unless we take
predicate-expressions to be referring devices; and he contends that the only
possible referents for predicate-expressions are universals. Some might object
to this talk of correspondence; but since I am inclined to think that at least
the truth of empirical discourse has to be anchored in non-lingUistic fact, I
shall assume that the realist is right here. Furthermore, I shall assume that he
is right in thinking that the correspondence that grounds truth presupposes
that the non-logical elements of language be referentially tied to objects out
in the world. What I want to ask is whether the realist is right in his conten-
tion that his interpretation of the referential force of predicate-expressions is
the only possible way of explaining how subject-predicate truth is grounded
in non-linguistic fact. To answer this question, we have to examine alternative
accounts of predication. Let us begin by considering the account proposed by
the nominalist.
The nominalist, as we have seen, wants to interpret agreement in attribute
in terms of Similarity of individual attributes. But, then, we can expect him
to explain the truth of subject-predicate discourse by an appeal to the notion
of an individual attribute. In fact, this is what the nominalist typically does.
Ockham and D. C. Williams are good examples here. As we have seen, Ock-
ham wants to construe only a limited number of predicates in nominalistic
terms; those predicates, he tells us, are true of (or as he puts it, are capable
of suppositing personally for) all and only the objects of which they are truly
34 CHAPTER TWO
predicable; but he also wants to claim that they consignify individual attri-
butes. Indeed, he claims that a qualitative predicate, 'F', just means 'object
possessing an F-ness'. Thus, on Ockham's account 'wise' is analyzed as 'object
possessing a wisdom' and 'courageous' as 'object possessing a courage'. Invoking
this analysis of qualitative predicate-terms, Ockham tells us that where 'F' is
a qualitative predicate, what grounds the truth of a subject-predicate sentence
of the form 'a is F' is simply that the referent (what Ockham calls the sup-
positum) of 'a' possesses one of the qualitative attributes consignified by 'F',
so that if 'Plato is courageous' is true, it is true because Plato possesses a
courage and if 'Socrates is wise' is true, it is true because Socrates possesses a
wisdom. 42 Although D. C. Williams, a more recent proponent of a nominalist
theory of predication, does not speak of predicates consignifying attributes,
he seems to think that individual attributes are correlated with predicate-
terms. Thus, all the rednesses that there are are correlated with the term 'red'
and all the courages that there are are correlated with the expression 'coura-
geous.' According to Williams, the effect of predicating a predicate-expression
of an ordinary object is to assert that the object is characterized by one of the
individual attributes correlated with that predicate-term. 43 But, then, despite
the difference in formulation, Williams' account of subject-predicate sentences
is of a piece with Ockham's account of predications within the Aristotelian
category of quality; for on Williams' account what makes the sentence 'Plato
is courageous' true is the fact that Plato is characterized by a courage and
what makes 'Socrates is wise' true is the fact that Socrates is characterized by
a wisdom.
The subject-predicate sentences we have used as examples in outlining the
accounts of predication presented by Ockham and Williams are all subject-
predicate sentences whose subject-terms refer to ordinary concrete objects.
Now, it is important to note that on the accounts presented by both, such a
subject-predicate sentence can be true only if some other subject-predicate
sentence is true. Thus, both Ockham and Williams are committed to the idea
that 'Socrates is wise' for example, is true only if a sentence which tells us
that some individual attribute of Socrates (call it ex) is a wisdom is true. How
exactly is the nominalist to explain the truth of this sentence? One strategy
here is to insist that the account just presented is perfectly general, so that
the truth of 'ex is a wisdom' is grounded in the fact that ex, in turn, possesses
some individual attribute (presumably one of a higher level). It should be
clear that this strategy commits the nominalist to the idea that an infinity of
attributes underlies every ordinary true subject-predicate sentence; but it
should be clear that there is nothing vicious in these infinitely long hierarchies
PREDICATION AND UNIVERSALS 3S
of individual attributes; for their infinity does nothing to threaten the nomi-
nalist's use of his theory of subject-predicate truth in any particular case.
While it is true that on his account every explanation of the truth of a subject-
predicate sentence introduces a new true subject-predicate sentence, the
nominalist's explanation of the truth of any particular subject-predicate sen-
tence does not require that he explain the truth of the new subject-predicate
sentence which that explanation brings upon the scene. 44
Nevertheless, the alleged theoretical Simplicity of nominalism is more than
a little compromised by these infinite hierarchies of attributes, so that the
nominalist is likely to eschew the strategy just suggested. A seemingly more
promising strategy here (one which both Ockham and Williams endorse) is to
insist that the account of subject-predicate truth we have been considering is
to be restricted to the case where subject-predicate sentences bear on ordinary
concrete objects. As regards the attributes underlying these predications, on
the other hand, the nominalist can claim that they are what they are not in
virtue of additional entities, but simply of and by themselves. On this ac-
count, then, while Socrates can be wise only if he possesses a wisdom, a's
being a wisdom is an ultimate fact, one involving no further entities.
This second strategy is not incoherent, but it implies an asymmetry among
subject-predicate sentences that is likely to leave us uneasy. It leads us to ask
why, if the truth of subject-predicate sentences like 'a is a wisdom' can be
accounted for without an appeal to further entities, a similar account cannot
be invoked at the outset for sentences like 'Socrates is wise'. It is, of course,
just such a move that is proposed by the extreme nominalist. He tells us that
the fact that Socrates is wise is ontologically basic; it presupposes the exist-
ence of Socrates and nothing else.
The central question, then, is whether the extreme nominalist is able to
mobilize this contention in providing an account of the truth of ordinary
subject-predicate sentences. Can he, to use Donagan's expression, explain how
subject-predicate sentences enable us to "state facts about the world"? The
extreme nominalist claims he can; and the account he presents here is dis-
armingly straightforward. He tells us that a subject-predicate sentence of the
form 'a is F' is true simply because the referent of 'a' is, in fact, F. 45 On his
account, then, if the sentence 'Socrates is wise' is true, it is true because
Socrates, the referent of 'Socrates', is wise; and likewise, if the sentence
'Plato is courageous' is true, it is true because Plato, the referent of 'Plato', is
courageous.
From a formal perspective, this account of subject-predicate truth is
unexceptionable: it provides an account of the truth of every true subject-
36 CHAPTER TWO
into
What makes it appear that the truth of (16) commits us to the existence of
a universal is the fact that the sentence incorporates the abstract term 'cour-
age'. Now, we are familiar with other sentences into which the term enters,
e.g.,
(21) Courage is a virtue
and
(22) Courage is admired by Plato;
and since we are inclined to think that as it occurs in these sentences, 'cour-
age' is functioning as a device for referring to a universal, we suppose that it is
playing the same role in (16). The extreme nominalist, however, will insist
that 'courage' is only apparently a device for referring to a universal in (21)
and (22). He will typically claim that the term is really just a device for
abbreviating discourse about familiar concrete objects, all the courageous
individuals that there are and that sentences like (21) and (22) can be re-
placed, without loss of content, by sentences in which the abstract term
'courage' gives way to the non-problematic predicate-term 'courageous'; and
he will conclude that it is simply wrong to suppose that as it appears in (16),
'courage' is a genuinely referring Singular term. He will grant, then, that sen-
tences like (15) are synonymous with sentences like (16), but he will claim
that what this shows is not that the truth of sentences like (IS) commits us to
the existence of universals, but rather that sentences like (16) are really just
very elaborate ways of making claims about familiar concrete objects and
nothing else. Thus, if the realist is to show that the paraphrases in question do
justify his analysis of predication, what he must do is counter the attempts of
the extreme nominalist (as well as the nominalist) to "analyze away" the
reference to universals ingredient in sentences like (21) and (22); but to do
this, he must go beyond any argument we have met in our discussion of (I)
and (II). He must appeal to a different line of argument - an argument for
the existence of universals based on what I earlier called the phenomenon of
abstract reference. If he is successful in formulating that line of argument,
then the availability of the paraphrases which convert ordinary subject-
predicate sentences into sentences of the form 'a possesses F-ness', 'a belongs
to K-kind', and 'a stands in the relation of being R with respect to b' could be
exploited in vindicating the analysis of predication expressed in (I) and (II).
Thus, given the paraphrastic equivalence between sentences like (15) and
(16), (I) and (II) might turn out to provide the only acceptable account of
predication; but an examination of the phenomenon of predication by itself
40 CHAPTER TWO
will not show this. Taken in isolation, predication appears to have the on-
tological neutriality which the extreme nominalist claims for it. We must
conclude, then, that the long and impressive line of philosophers mentioned
earlier are wrong in thinking that the phenomenon of predication alone is
sufficient to establish our commitment to a realistic ontology.
NOTES
1 De Interpretation 3 (16b 10), translated by E. M. Edgehill, in Richard Mckeon, Basic
Works of Aristotle (New York: Random House), 1946, p. 41.
2 Ibid., 7 (17a 38-39), p. 43.
3 'Concept and Object' in Philosophical Writings of Gottlob Frege, translated by Peter
Geach and Max Black (Oxford: Blackwell), 1952, p. 43 and p. 51. Frege refuses, of
course, to call the referents of predicate-terms "objects". I discuss this issue in Section V
of this chapter.
4 Problems of Philosophy, p. 145. In this passage, Russell contrasts substantives with
proper names, so what he has in mind here are common nouns.
5 Two Types of Linguistic Philosophy,' The Metaphysics of Logical Positivism, p. 122.
6 See Chapter V of Strawson's Individuals (London: Methuen), 1959. Strawson refuses
to use the term 'refer' in conjunction with predicate-expressions; but this is, I think, only
a terminological point. His use of 'term-introduction' is generic, covering the word-world
ties involved in the use of both general terms and singular terms. The former, he tells us,
are used to predicate one thing of another; the latter, to refer. I discuss this issue in
Section V of this chapter.
7 I shall use the term 'predicate-expression' in such a way that any term that can func-
tion predicatively in a subject-predicate sentence is a predicate-expression; where a predi-
cate-expression is functioning predicatively in a subject-predicate sentence, S, I shall call
it the predicate-expression of S.
8 'Universals and Metaphysical Realism,' p. 130.
9 Ibid., p. 127.
10 Ibid., p. 130.
11 See, e.g., p. 122 of Bergmann's 'Two Types of Linguistic Philosophy' and pp. 128-
129 of Donagan's 'Universals and Metaphysical Realism.'
12 Bergmann disagrees here. He so uses the term 'existent' that only what he calls "sim-
ples" count as existents; and as he explains it, a simple is the referent of a primitive
descriptive term. It should be clear, however, that Bergmann's use of the term 'existent'
is deviant; for on his use of the term, we would have to deny that automobiles, trees, and
persons are existents on the grounds that they have parts.
13 Thus, even defenders of the so-called Causal Theory of kind-words could accept my
(I). If they are right, then although (I) is true, it might turn out that all of the speakers
of a language are completely ignorant of just whkh universal (or to use Putnam's term,
just which "essential nature") is shared by all and only the objects to which a predicate-
term having general currency in the language correctly applies. See, e.g., Putnam's papers
'Is Semantics Possible?' in Language, Belief, and Metaphysics, edited by Kiefer and
Munitz (New York: State University of New York Press), 1970, pp. 50-63 and 'Meaning
and Reference,' Journal of Philosophy, 1973, pp. 699-711. Both of these papers are
PREDICATION AND UNIVERSALS 41
reprinted in Stephen Schwartz, Naming, Necessity, and Natural Kinds (Ithaca, New
York: Cornell University Press, 1977.
14 The Nature of Necessity (Oxford: Clarendon Press), 1974, pp. 225-227.
15 Ibid., pp. 2215-227.
16 'Notes on Existence and Necessity' in Leonard Linsky, Semantics and the Philosophy
of Language (Urbana: University of Illinois Press), 1963, pp. 77-78.
17 Philosophical Investigations, translated by G. E. M. Anscombe (London: MacMillan),
1953, p. 66.
18 'Universals and Family Resemblances,' Proceedings of the Aristotelian Society,
1960-1961; reprinted in Loux, Universals and Particulars, p. 113.
19 Philosophical Investigations, p. 67.
20 'Universals,' Philosophical Quarterly, 1951. Reprinted in Lo'ux, Universals and Parti-
culars, pp. 44-58.
21 Ibid., p. 47.
22 'Universals and Metaphysical Realism,' pp. 147-149.
23 This ambiguity is brought out in Wilfrid Sellars' 'Naming and Saying' in his Science,
Perception, and Reality (London, Routledge and Kegan Paul), 1963, pp. 242-246.
24 Parmenides, 131 E-132 B; pp. 926 in Hamilton and Cairns, Plato.
25 It turns out that my treatment of the so-called Third Man Argument invokes the two
assumptions that contemporary scholars have isolated as central to the argument. See,
e.g., G. E. L. Owen's 'The Platonism of Aristotle' in Studies in the Philosophy of Thought
and Action, edited by P. F. Strawson (London: Oxford University Press), 1968, pp.
147-174.
26 Strawson takes this line; but his reason is, I think, the wrong one; he takes the regress
here to be vicious. See Individuals, p. 178.
27 This third infinite regress argument is the one discussed by Donagan in 'Universals
and Metaphysical Realism,' pp. 136-139. The argument is essentially that found in
Gilbert Ryle's 'Plato's Parmenides (I),' Mind, 1939, pp. 137-138; it is a near relative of
Bradley's famous argument.
28 On Universals, p. 102.
29 See, e.g., 'Meaning' in Bergmann's Logic and Reality (Madison: University of Wiscon-
sin Press), 1964, pp. 87.-88.
30 See e.g., Individuals, p. 169.
31 See e.g., 'The Philosophy of Malebranche' in Bergmann's Meaning and Existence
(Madison: University of Wisconsin Press), 1959, pp. 190-191.
32 Speech Acts (Cambridge: Cambridge University Press), 1969, p. 102.
33 Ibid., p. 103.
34 Ibid.
35 'Concept and Object,' p. 103.
36 Speech Acts, ibid.
37 Metaphysics r. 4 (1006a 12-17).
38 On Universals, p. 85.
39 See once again Chapter V of Individuals.
40 For a confumation of this diagnosis, see Physics A.3 (186 a 25-32), pp. 25-26 of
On Universals, and p. 173 of Individuals.
41 Schematically, we can say that a predicate-term, 'F' is true of or satisfied by all and
only F-objects and expresses the universal F-ness. Searle objects to this sort of account;
42 CHAPTER TWO
the objection seems to be that since the use of predicate-expressions is typically mas-
tered before the use of their abstract counterparts, sentences like (16), (18), and (20)
cannot serve as paraphrases of sentences like (15), (17), and (19). See pp. 119-121 of
Speech Acts. I fmd this a bad objection; one might as well argue that 'male sibling'
cannot serve as an analysis of 'brother' since the use of 'brother' is typically learned
before the use of 'sibling'.
42 See, e.g., Summa Logicae 11.11, p. 281 in Summa Logicae, edited by Boehner,
Gal, and Brown (St. Bonaventure, New York: Franciscan Institute Publications), 1974.
The student of Ockham will want to make two objections here. The fll'st is that I fail
to deal with the distinction between signijicatio and personal suppositio. This is cor-
rect, but the point to keep in mind here is that for Ockham, primary signijicatio (as
opposed to secondary signijicatio or consignijicatio) just is the capacity to have a cer-
tain kind of personal suppositio. Second, one might object that Ockham gives a general
account of truth-<:onditions for sentences of the form 'a is F' and that this account
makes no reference to individual attributes. Again, the objection is correct. Ockham
wants a generalized theory of truth-<:onditions for subject-predicate discourse; and
since he is an extreme nominalist with respect to predications outside the category
of quality, any reference to individual attributes in a general theory of truth-<:ondi-
tions would be out of the question; nonetheless, as the passage in Summa 11.11 indi-
cates, where 'F' is a predicate-term from the category of quality (not expressive of
figure or form), the sentence 'a is F' can be true only if a possesses some individual
F-ness.
43 See 'On the Elements of Being, I,' pp. 11-12.
44 For a different approach to these issues, see David Armstrong, 'Infinite Regress
Arguments and the Problem of Universals,' Australasian Journal of Philosophy, 1974,
pp. 191-201.
45 See, e.g., the account of predication suggested in Quine's 'On What There Is' in his
From a Logical Point of View (Cambridge, Mass.: Harvard University Press), 1953, esp.
pp. 10-13. See also the theory of subject-predicate truth outlined in Wilfred Sellars
'Naming and Saying,' Science, Perception, and Reality, pp. 225-246.
46 See Panayot Butchvarov's Resemblance and Identity (Bloomington, Indiana: Indiana
University Press), 1966, Chapter One.
47 Donagan, 'Universals and Metaphysical Realism,' p. 153.
48 Quine, of course, uses the notion of satisfaction in identifying the word-world rela-
tions that tie predicates to the objects of which they are predicable. For a number of
reasons, Sellars refuses to construe the word-world relations in question as sernantical.
They are, he tells us, matter-of-factual or causal relations; nonetheless, on his account,
we have the consequence that it is in virtue of the relevant word-world relations that
subject-predicate sentences come out true. See, e.g., Sellars' 'Hochberg on Mapping,
Meaning, and Metaphysics,' Midwest Studies in Philosophy, Vol. II (Morris, Minnesota:
University of Minnesota at Morris), 1977, p. 222.
49 See, e.g., Chapters III and IV of Science and Metaphysics for an elaboration of
this theme. Pretty clearly, for a theory of predication that is at once nominalistic and
intensionalistic to succeed, it must be possible to provide analyses of notions like signi-
ficance and necessity/possibility that show the applicability of these concepts to pre-
suppose the existence of no entities (e.g., meanings or possible worlds) that would be re-
pugnant to an extreme nominalist. Obviously, I lack the space to discuss these issues here.
PREDICATION AND UNIVERSALS 43
Suffice it to say that Sellars thinks that meaning-talk can be analyzed as a theoretically
neutral form of classificatory discourse and that modal idioms can be analyzed meta-
linguistically in terms of the notion of linguistic rules and the linguistic commitments of
a language-using community.
CHAPTER THREE
Although the phenomenon of resemblance does not play the central role in
the history of realism enjoyed by the phenomenon of predication, distin-
guished realists have contended that the analysis of resemblance provides us
with the resources for constructing an argument for the existence of univer-
sals. In fact, we find a number of such arguments in the writings of realists.
In this chapter, I want to examine three of them. The first argument I shall
consider played an important role in the debate over universals in the early
part of this century; the locus classicus for this argument is Chapter IX of
Russell's Problems of Philosophy.
To get at the argument presented there, we have to recall that both nomi-
nalists and extreme nominalists want to understand the phenomenon of attri-
bute-agreement in terms of the idea of resembling particulars. This is explicit
in the nominalist's contention that ordinary objects agree in attribute if and
only if they "exemplify" distinct yet resembling attributes; but the extreme
nominalist too will agree that attribute-agreement is to be understood in
terms of similarity; for he will say that objects agree in attribute if and only
they resemble each other as men, as wise, as trees, and so on. In both cases,
then, the contention is that we can explain the phenomenon of attribute-
agreement by reference to resembling particulars rather than multiply exem-
plifiable entities.
Russell's argument is meant to establish the futility of these attempts at
understanding attribute-agreement. According to Russell, they succeed in
eliminating ordinary universals like whiteness and triangularity only by intro-
ducing a new universal - that of similarity or resemblance. The nominalist
and extreme nominalist will agree that there are many cases of attribute-
agreement; but, then, they must agree that "resemblance ... holds between
many pairs of particulars; and this," Russell tells us, "is the characteristic of
a universal." He concludes the argument by saying:
The relation of resemblance, therefore, must be a true universal. And having been forced
to admit this universal, we fmd that it is no longer worth while to invent difficult and im-
plausible theories to avoid the admission of such universals as whiteness and triangularity. 1
44
RESEMBLANCE AND UNIVERSALS 45
It will be useless to say that there is a different resemblance for each pair; for then we
shall have to say that the resemblances resemble each other and thus at last we shall be
forced to admit resemblance as a universal. 2
One might think, however, that while the argument just examined may not
establish the existence of universals, it contains an important insight. The
insight is simply that objects cannot resemble each other unless they literally
have some one thing in common; and that would seem to be possible only if
there actually are multiply exemplifiable entities. This contention is surely at
work in the following interchange in Plato's Parmenides:
Parmenides: If a thing is like must it not be like something that is like it?
Socrates: It must.
Parmenides: And must not the thing which is like share with the thing that is like it in
one and the same thing?
Socrates: Yes.
Parmenides: And will not that in which the like things share, so as to be like, be just the
Form itself that you spoke of?
Socrates: Certainly. 8
The suggestion here is that objects can resemble each other only if they
"share" some Form, only if they jointly exemplify one and the same univer-
sal. A similar claim is made by Charles Baylis:
The denial by extreme nominalists that particulars have common characteristics leads
them to affirm that no two things can be precisely alike in any respect. This has the
prima facie absurd consequence that, for example, no two books can be precisely alike in
containing exactly 232 pages. 9
Baylis evidently agrees with Plato, then, in supposing that the phenomenon
of resemblance commits us to the existence of universals; for he tells us that
it is only if they jOintly exemplify one and the same characteristic that ob-
jects can be alike.
But why should anyone hold this? Well, presumably the source of this
contention is the view that if resemblance-claims are to be true, to state facts
about the world, there must be some objective ground for their truth; and
this, the realist wants to claim, is guaranteed only if we suppose that the ob-
jects truly asserted to resemble each other actually agree in some one thing,
exemplify some one entity. It is the capacity for truth, then, which resem-
blance claims enjoy that supposedly forces us to grant the existence of
universals.
Now, the insight at work here is reminiscent of the contention underlying
so CHAPTER THREE
abstract, the truth of the claim that they resemble each other is grounded in
the exemplification by those objects of resembling attributes. Now, if he
generalizes the original account in this way, the nominalist is committed to
the idea that there are infinitely long chains of attributes underlying the truth
of any ordinary resemblance-claim. It should be obvious by now that there
is nothing vicious in these infinities; but it should be equally obvious that
these infinitely long strings of attributes tell against the nominalist's conten-
tion that his account has the special virtue of theoretical simplicity.
But, of course, the nominalist need not generalize his account in this way.
On the contrary, he can insist that the account just provided be restricted to
the case where ordinary, concrete objects are said to resemble each other. In
fact, this seems to be the line that Williams takes. ll Conceding that the truth
of ordinary resemblance-claims is grounded in the exemplification of resem-
bling attributes, Williams denies that the truth of resemblance-claims about
attributes presupposes any additional entities. Attributes, he seems to hold,
resemble each other not in virtue of exemplifying further attributes but
simply in virtue of what they are - wisdoms, triangularities, or rednesses.
This strategy is not incoherent; it should be clear, however, that it implies
an asymmetry among resemblance-claims that plays directly into the hands of
the extreme nominalist; for the extreme nominalist will demand to know
why, if we need no additional entities to account for the supposed truth of
sentences like
(1) The wisdom of Socrates resembles the wisdom of Plato,
we need special entities to ground the truth of sentences like
(2) Socrates resembles Plato;
and he will go on to argue that we can ground the truth of ordinary resem-
blance-claims simply by pointing to how ordinary objects are, what they are,
and how they are related to each other. He will grant that resembling objects
agree; they agree in being wise, red, triangular, human beings, dogs, or trees;
and he will concede as well that resembling n-tuples of objects agree in being
composed of objects that are teacher and student, father and son, mother
and daughter, and so on. The extreme nominalist, then, will insist that his
account meets the requirement set out by the realist - that of grounding the
truth of resemblance-claims in non-linguistic fact; and he will argue that since
his account is theoretically simpler than that of the nominalist or realist, it is
preferable.
In confronting the issue of predication, we found the extreme nominalst's
52 CHAPTER THREE
NOTES
5 Inquiry, p. 327.
6 Interestingly, this argument is not, in precisely this form, found in the writings of
realists. The argument can, however, be found in the writings of the opponents of real-
ism, where the argument is presented to be criticized. See, e.g., Price's Thinking and
Experience, pp. 19-23. See also Bambrough's 'Universals and Family Resemblances,'
p.123.
7 See, e.g., Pears' 'Universals,' p. 50.
s Parmenides, 132 D-E; p. 927 in Hamilton and Cairns, Plato.
9 'Universals, Communicable Knowledge, and Metaphysics,' Journal of Philosophy,
1951, p. 638.
10 'On the Elements of Being, I,' pp. 4-5.
11 Ibid.
CHAPTER FOUR
The upshot of the previous three chapters is that the realist cannot establish
the existence of universals merely by an appeal to the phenomena of predica-
tion and resemblance. Of the traditional approaches to universals, then, we
are left with that based on the phenomenon of abstract reference. If the realist
is to convince us that we should embrace the ontological framework he recom-
mends, he will have to base his case on the claim that the truth of sentences
which appear to involve devices for referring to universals actually presup-
poses the existence of those universals. Our discussion so far suggests that the
most plausible sentences to examine here are sentences like 'Triangularity is a
shape' and 'Red is a color' which incorporate what have been called abstract
singular terms; but before we confront the issue of the referential role of
abstract terms, we ought to consider an argument for the existence of univer-
sals which is found in the later writings of W. V. Quine.
Quine agrees that the phenomenon of abstract reference provides us with
grounds for concluding to the existence of what appear to be universals; but
he denies that the realist should focus his attention on sentences incorporat-
ing abstract terms. Quine wants to claim that the sentences that are to serve
as the basis for an argument for realism are those that involve the apparatus
of quantification. From his earliest writings, Quine has contended that the
apparatus of quantification is a referential framework which enables us to
determine the ontological presuppositions of what we say. This contention is
encapsulated in Quine's famous slogan "To be is to be the value of a bound
variable."} Towards filling out this slogan, Quine tells us that if we want to
determine the ontological commitments associated with the acceptance of a
certain body of discourse, e, we must translate the sentences, S} ... Sn, of
e into the language of quantification. Call the results of this translation
S' \ ... S'n. Next, we are to specify the truth-conditions for S'\ ... S'n. Ac-
cording to Quine, the acceptance of e commits us to the existence of an
object, x, if and only if there is a sentence, S'i, from S'\ ... S'n such that x
must be construed as !he value of variable bound by a quantifier in S'i if S'i
is to come out true.
54
ABSTRACT REFERENCE AND UNIVERSALS 55
He (i.e., the philosopher who denies the existence of abstract entities) is going to have to
accommodate his natural sciences unaided by mathematics; for mathematics except for
some trivial portions of very elementary arithmetic is irredeemably committed to quanti-
fication over abstract objects. 2
Quine wants to claim, then, that when we translate the sentences of mathe-
matics into quantificational notation, we find that many of those sentences
involve the quantification of variables whose values are classes, so that unless
we are willing to admit the existence of classes, we are forced to deny the
truth of large bodies of mathematical discourse. But among the things that
are classes are objects (like the class of all men and the class of all oak trees)
which are "ones" to which many different things can belong. Such one-many
relations (or quaSi-relations) would appear to involve us in an ontology of
universals, so that Quine's argument would appear to establish the existence
of universals.
Now, Quine's criterion of ontological commitment bears on the use of
both the universal ('if _ ) and the so-called existential (3 _ ) quantifiers.
The quantifiers are, however, interdefinable; and since the 3-quantifier is
more blatant in its apparent ontolOgical force, I shall focus my attention on
its use. Effectively what Quine is arguing is that the use of the 3-quantifier
with respect to variables other than individual variables commits us to the
existence of entities over and above concrete particulars. Quine himself is
willing to endorse only higher-level quantifications like
(1) (3 a) (Socrates a).
but his thesis is that if we are willing to endorse quantifications like
(2) (3 F) (F-Socrates),
we are committed to the existence of things like properties; and if we are
willing to endorse quantifications like
56 CHAPTER FOUR
which seems to commit its proponent to the existence of at least one abstract
entity, vis., some class to which Socrates belongs. But here one might claim
that the expressions which function as substituends for class variables, expres-
sions like '(x) (Man-x), (i.e., 'the class of all men') and '(x) (Oak tree-x), (i.e.,
'the class of all oak trees'), are only apparently singular referring devices. A
possible move here is to suggest that they are expressions contextually intro-
duced in terms of the following schema:
Invoking this account of class-"names", one could claim that a sentence like
(1) is really just equivalent to a sentence like (2) and, consequently, has the
same ontological neutrality that the 'some' reading attributes to (2).
If we are willing to accept this sort of account of sentences like (1), then
we seem forced to concede that our original willingness to accept Quine's
views about the ontic force of higher level quantification was too hasty. It
would seem that if we take an alternative informal reading of the 3-quantifier,
we can provide an acceptable reading of sentences like (1), (2), and (3) which
does not show them to involve assertions of existence. Quine, however, would
argue that the ontic force he finds in (1 )-(3) does not depend upon the infor-
mal reading we provide for sentences involving the 3-quantifier. What forces
us to take (I)-(3) as involving a commitment to abstract entities is the logi-
cian's account of the truth-conditions for formulae involving the 3-quanti-
fier. 4 What Quine has in mind here is the fact that the logician typically tells
us that an 3-quantification of the form (3 v) (..... v ..... ) is true if and
only if there is an object which satisfies the open sentence follOwing the
quantifier. On this account, (4) comes out true if and only if there is at least
one object that satisfies the open sentence (Man-x). Regardless of the infor-
mal reading we provide of (4), then, (4) would seem to come out true if and
only if there is at least one human being. But when we apply this account of
the truth-<:onditions of 3-quantifications to the case of(2) and (3), we seem
to get the conclusion that Quine wants - that endorsing such sentences com-
mits us to the existence of abstract entities; for regardless of how we read
these sentences, (2) comes out true if and only if some object satisfies the
open sentence (F-Socrates); and (3) comes out true if and only if some object
satisfies the open sentence (R-Socrates, Plato). The requisite objects, it would
seem, could only be abstract entities, so that to endorse (2) and (3) would,
after all, appear to involve a commitment to abstract entities; and if sentences
58 CHAPTER FOUR
like (2) and (3) commit their proponents to the existence of abstract entities,
talk about the contextual definition of classes becomes futile; for as we have
seen, the strategy of contextual definition does no more than reduce a sen-
tence like (l) to a sentence like (2).
The account of the truth-conditions for 3-quantification that we have
been discussing has been called the referential or objectual interpretation of
quantification; and although it provides the standard way of specifying truth-
conditions for quantified sentences, Quine himself points to an alternative
interpretation, what he and others have called the substitutional interpreta-
tion of quantification. s On this account, an 3-quantification of the form
(.3 II) (. .... II ...) comes out true if and only if there is a linguistic ex-
pression which, when substituted uniformly for the variable bound by the
3 -quantifier makes the open sentence following that quantifier come out
true. On this account, (4) comes out true because there is a linguistic expres-
sion - e.g., 'Richard M. Nixon' - which, when substituted for the 'x' in
(Man-x), yields a true sentence, e.g., 'Richard M. Nixon is a man'. Now, as
regards sentences like (4) the substitutional account provides us with an
account which, from the perspective of ontology, has roughly the same
consequences as the referential interpretation; for presumably 'Richard M.
Nixon is a man' comes out true only if 'Richard M. Nixon' names some exist-
ing human being; but when we turn to sentences like (2) and (3), we see how
this interpretation of quantification represents a radical departure from the
referential account of 3-quantification. On the substitutional interpretation,
(2) comes out true if and only if there is a linguistic expression which can be
substituted for the 'F' in (F-Socrates) to yield a true sentence; but the expres-
sion 'is wise' does the job here; and if we were right in our discussion of predi-
cation, the truth of a sentence like 'Socrates is wise' does not by itself commit
us to the existence of abstract entities. Thus, on the substitutional account, a
sentence like (2) does not have the ontological force that the referentialist
reads into it. Likewise, (3) comes out true on the substitutionalist's account
just in case there is a linguistic expression which, when substituted for 'R' in
(R-Socrates, Plato) converts (R-Socrates, Plato) into a true sentence; but 'is
the teacher of is such an expression; and its use does not by itself commit us
to the existence of an abstract entity, so that on the substitutionalist's ac-
count, endorSing (3) does not have the ontological consequences Quine attri-
butes to it; and, of course, if sentences like (2) and (3) lack the ontological
force Quine attributes to them, the strategy of taking classes to be contex-
tually defined becomes once again a live option for the philosopher who
wants to endorse sentences like (1) without embracing an ontology of classes.
ABSTRACT REFERENCE AND UNIVERSALS S9
fies the attribute being human and vice versa; nonetheless, these are different
attributes. Ukewise, every object exemplifying the property of triangularity
exemplifies the property of trilaterality and vice versa; yet we have two attri-
butes here and not one.
I am inclined to conclude, then, that Quine's account does not provide us
with an affirmative answer to the age-old question, "Do universals exist?"
Nonetheless, since the referential interpretation of quantification is correct,
sentences like (2) and (3) commit their proponents to the existence of attri-
butes. Quine, however, denies that (2) and (3) are true. While embracing
quantification over classes, he eschews quantification over attributes, whether
multiply exemplifiable or not. Given Quine's account, though, we need to be
convinced that sentences like (2) and (3) are true if we are to be convinced
that what have traditionally been called universals exist. But how could we
find out that such sentences are true? Well, it seems to me that the only way
the realist could convince Quine of their truth is by convincing him that the
truth of sentences like 'Socrates possesses wisdom' and 'Socrates and Plato
enter into the teach-student relation' presupposes the existence of universals.
Having been corwinced of that, Quine would be forced to grant the truth of
the quantified counterparts of these sentences. This suggests, however, that
Quine's criterion will not by itself enable us to determine whether or not
universals exist. To determine whether this is so, we must examine the phe-
nomenon of abstract reference as it is exhibited in sentences involving not
the apparatus of quantification but the use of what I earlier called abstract
singular terms. Let us, then, try to determine whether the apparently referring
uses of terms like 'triangulatiry' and 'wisdom' commit us to the existence of
universals.
roles are many and varied. They appear, as we have noted, in what might be
called exemplification contexts, contexts in which, to all appearances, we
pick out some object and say that it exemplifies or exhibits some universal.
Thus,
(8) Socrates possesses wisdom.
(9) The scalene exemplifies triangularity,
(10) The American Flag instantiates redness.
Abstract singulat terms also frequently function in what we can call inten-
tionality contexts, contexts in which we specify the objects of a person's
mental states or acts. When they function in such contexts, abstract singular
terms appear to serve as devices for identifying the abstract entity that a
person's mental act or state is of, for, or about. Examples of this use of
abstract singular terms are
and
At its most general level, the classificatory context engages the various ontol-
ogical categories. When abstract terms are coupled with category-words, we
appear to be picking out universals and subsuming them under their most
general kinds. Thus,
(1 7) Wisdom is a property,
(18) Redness is a property,
(19) Triangularity is a property,
(20) Animal is a kind,
(21) Paternity is a relation,
(22) Murder is an action. 9
ABSTRACT REFERENCE AND UNIVERSALS 63
Now, it should be clear that sentences like (8)-(22) can be used to make
true claims. What the realist contends is that the abstract singular terms ap-
pearing in those sentences are playing just the roles they appear to be playing
and that, consequently, the truth of sentences like (8)-(22) presupposes the
existence of universals. Effectively, then, the realist can be viewed as issuing a
challenge to the nominalist and the extreme nominalist, a challenge to provide
analyses of sentences like (8)-(22) according to which the truth of these
sentences does not presuppose the existence of universals. He is challenging
the opponents of realism to come up with an account of sentences like
(8)-(22) which shows the reference to universals involved in the use of these
sentences to be only apparent.
The realist, however, wants to lay down a condition which the accounts of
the nominalist and extreme nominalist must meet. We can call it the Condition
of Semantic Uniformity. In rough terms, this amounts to the claim that to be
acceptable an analysis of sentences incorporating abstract singular terms must
show any given abstract term to have one and the same role in the various
contexts I have pointed to. The rationale for this condition comes out when
we examine the ways in which the terms 'wisdom', 'triangularity', and 'red-
ness' function in the various sentences I have used as examples. Beginning
with 'wisdom', it should be clear that whatever role it is playing in (8), (11),
(14), and (17), that term is playing a single role in all four sentences. To see
this, we need only reflect on the fact that the conjunction of (8), (11), (14),
and (l7) entails
(23) A virtue Alcibiades aspires to is a property that Socrates possesses.
This entailment would not hold unless 'wisdom' were functioning as one and
the same linguistic expression (i.e., unless it were playing the same linguistic
role) in all four cases. The same point comes out in the case of 'triangularity'.
It has the same force in each of (9), (13), (16), and (19); for those sentences
taken together entail
(24) A property mathematicians think about is a shape that the scalene
exemplifies.
This entailment would not hold were 'triangularity' playing different roles in
the four cases. Finally, 'redness' is playing a single role, whatever that role
may be, in (10), (12), (15), and (18) since the conjunction of those sentences
entails
(25) A color Quine prefers to whiteness is a property instantiated by
the American flag;
64 CHAPTER FOUR
and this entailment presupposes that the force of 'redness' is invariant across
(10), (12), (15), and (18).
The challenge of the realist, then, is that the nominalist or extreme nomi-
nalist provide an analysis of sentences like (8)-(22) when shows the truth of
those sentences not to presuppose the existence of universals, but also shows
a given abstract term to have a single function or role in the different sen-
tences into which it enters. Nominalists and extreme nominalists have been
quick to take up this challenge. Indeed, although the issue of abstract singular
terms has been pivotal throughout the history of the controversy over univer-
sals, realists themselves have had little to say about the matter. They have
seldom even bothered to spell out the challenge I have just outlined. The
challenge, it would seem, is so obvious that it does not need to be formally
issued. The nominalist and extreme nominalist, on the other hand, have both
recognized the burden the challenge places upon them; for it is in their writ-
ings that we find detailed accounts of the linguistic role of abstract singular
terms. Evidently, then, the various parties to the dispute agree that the meta-
physical theories proposed by the opponents of realism cannot claim to be
adequate unless they incorporate an account of sentences like (8)-(22) which
shows the abstract singular terms embedded in these sentences to have a role
other than that they appear to have.
But while most philosophers who have wirtten on the problem ofuniver-
sals have taken the issue of abstract singular terms to be pivotal, there is one
dissenting voice here. Strangely enough, it comes from the mouth of a realist,
Nicholas Wolterstorff. Wolterstorff denies that an examination of the phe-
nomenon of abstract reference will provide us with the resources for resolving
the perennial debate between what I have called realists, nominalists, and
extreme nominalists. He claims that even if the nominalist or extreme nomi-
nalist were to come up with a satisfactory alternative to the realist's inter-
pretation of abstract reference, this would not have the effect of showing that
universals do not exist and, conversely, that if the realist were to succeed in
showing the inadequacy of all existing alternatives to his interpretation, this
would not serve as a definitive proof that realism is true. to I am inclined to
agree with Wolterstorff here. Suppose, for example, that we had compelling
independent reasons for thinking that universals exist (e.g., reasons based on
the nature of predication, resemblance, or higher order quantification) and
were to be confronted with some non-realistic account of abstract reference
equally as powerful as that provided by the realist. We would rightly con-
clude, I think, that the existence of that account is a curious but irrelevant
fact. And, if in the absence of independent grounds for endorsing realism, we
ABSTRACT REFERENCE AND UNIVERSALS 6S
Although most extreme nominalists have wanted to deny that abstract singu-
lar terms are genuinely singular referring expressions and to claim that they
are eliminable in favor of concrete general terms, some extreme nominalists
66 CHAPTER FOUR
have tried to accommodate the intuition that expressions like 'wisdom' and
'triangularity' are singular referring devices by identifying objects which,
while confirming to the rigors of an ontology that recognizes no attributes,
could legitimately be construed as referents of abstract singular terms. One
strategy here is to claim that an abstract singular term 'F-ness', functions
as the name of the set or class whose members are all and only those objects
that satisfy the concrete term, 'F'. On this view, 'redness' names the set
composed of all and only those objects that are red, 'mankind' names the
set whose members are all and only the individuals that are human beings,
and 'wisdom' names the set consisting of all and only the individuals that
are wise. Of course, if he is to avoid the conclusion that abstract terms are
constantly changing their referents, the proponent of this strategy must take
the 'are' here tenselessly, so that objects are, for example, red just in case
it is now true that they are red, it was true in the past that they are red, or
it will be true in the future that they are red. But even if he interprets the
notion of set-membership in this way, the extreme nominalist fails to provide
a satisfactory account of abstract singular reference. What is attractive about
this view is that it enables us to substitute objects with straightforward
identity-conditions for things that lack what Quine has called a "crystal clear
identity concept."ll But as we have already seen, it is just the feature of
sets or classes that makes this view attractive - their extensionality - which
is the downfall of this approach. Given the extensionality of sets, this view
forces us to say that two abstract terms name the same object just in case all
the objects satisfying the concrete term associated with one of these terms
satisfy the concrete term associated with the other and vice versa. Thus, since
all the objects that tenselessly are triangular tenselessly are trilateral and vice
versa, this account forces us to conclude that 'triangularity' and 'trilaterality'
name the same object; and since it is plausible to think that the future will
not deviate markedly from the past, it is plausible to suppose that all the
objects satisfying the concrete term 'human' satisfy the expression 'feather-
less biped' and vice versa, so that on this account we seem forced to say
that 'being human' and 'being a featherless biped' name the same object. We
know, however, that triangularity and trilaterality are different things and
that being human is something different from being a featherless biped, so
that we can conclude that the extreme nominalist's appeal to sets here is
unsatisfactory .
But if sets of concrete objects fail to provide us with satisfactory referents
for abstract singular terms, W. V. Quine points to objects that might appear
to do the trick. Quine says:
ABSTRACT REFERENCE AND UNIVERSALS 67
The regions to which 'red' applies are indeed not continuous with one another as those
are to which 'Cayester' (a river in Lydia) applies, but this is surely an irrelevant detail;
'red' surely is not opposed to 'Cayester', as abstract to concrete, merely because of dis-
continuity in geometrical shape. The territory of the United States including Alaska is
discontinuous, but it is none the less a single concrete object; and so is a bedroom suite
or a scattered deck of cards. Indeed every physical object that is not subatomic is, ac-
cording to physics, made up of spatially separated parts. So why not view 'red' quite on
a par with 'Cayester', as naming a single object extended in space and time? From this
point of view, to say that a certain shape is red is to affirm a simple spatiotemporal rela-
tion between two concrete objects; the one object, the drop, is a spatiotemporal part of
the other, red, just as a certain waterfall is a spatiotemporal part ofCayester. 12
The suggestion here is that the philosopher who wants to deny the existence
of attributes could take 'red' (Le., 'redness') to name a single concrete object
- the spatiotemporally discontinuous region of the world that is red. If we
generalize this account of 'red', we come up with the view that an abstract
term 'F-ness' is a Singular term naming that spatiotemporally discontinuous
region of the world that is F. Quine himself, however, presents this account
only to reject it; for while he concedes that the account provides a satisfac-
tory treatment of color-words, he is unwilling to extend it to cover all abstract
expressions. Part of the difficulty with this account is that it works only in
the case where the concrete term associated with an abstract term refers
cumulatively (Le., in such a way that the term applies to any sum of things
to which it applies). But, then, the account fails to handle terms like 'man-
kind' and 'triangularity' since clearly two individual men do not compose a
sector of the world that is itself a man; nor do two spatially discontinuous
triangular objects constitute some third thing that is triangular. But even if
the account could be extended to accommodate what Quine calls the phe-
nomenon of divided as opposed to cumulative reference, it fails to provide an
account of those abstract terms whose concrete counterparts cannot be con-
strued as applying to objects with spatial location. Thus, there is no spatio-
temporally discontinuous region of the world that is prime or composite, so
that the account fails to provide us with an account of the arithmetical terms
'being prime' and 'being composite'.
But, then, this approach (sometimes called the "exploded object" theory)
is no more satisfactory than the approach which identifies the referent of an
abstract singular term with a set of concrete objects. It is, of course, because
of the difficulties associated with theories like these two that most extreme
nominalists have claimed that abstract singular terms are only apparently sin-
gular referring devices. What they have claimed is that abstract Singular terms
are devices for abbreviating discourse about the various objects satisfying the
68 CHAPTER FOUR
concrete general terms out of which they are constructed. This strategy can be
traced back to the work of William of Ockham who seems to have thought
that in the case of many abstract terms, it is possible to replace sentences
incorporating those terms by sentences incorporating their concrete counter-
parts. Thus, on Ockham's account, sentences incorporating the abstract term
'triangularity' are synonymous with sentences in which this word does not
occur but the concrete form 'triangular' does; and sentences incorporating the
abstract term 'trilaterality' are synonymous with sentences in which this
expression does not occur but the concrete term 'trilateral' does. 13 Generaliz-
ing the account Ockham suggests in these cases, we have the view that abstract
singular terms are eliminable from discourse, that given any abstract term,
'F-ness', sentences incorporating 'F-ness' can be analyzed in terms of sentences
in which the term 'F-ness' does not occur but its concrete counterpart, 'F',
does.
Proponents of this account of abstract singular terms have not always been
clear on just how these analyses are to go; but in discussions of the issue, one
frequently meets with the claim that the use of an abstract singular term
signals the fact that what is being made is some necessarily true claim about
all the objects satisfying the corresponding concrete general term. The sugges-
tion, then, is that a sentential context of the form
(..... F-ness ..... )
can be paraphrased by a sentence of the form
Necessarily ( ..... every F-object .....).
Now, it should be clear that this suggestion does not tell us everything that
we must know if we are to paraphrase sentences incorporating abstract sin-
gular terms by means of sentences incorporating their concrete counterparts;
for since expressions syntactically suited for combination with an abstract
term will not, in general, be susceptible of combination salvo sensu with the
concrete counterparts of those terms, the application of the schema
(..... F-ness ..... ) -+ Necessarily (..... every F-object ..... )
will necessitate transformations over and above those explicitly indicated by
the schema. But if we allow those additional transformations to be dictated
by context, we can get a general idea of what this suggestion amounts to. It
would have us read (IS) as
(I Sa) Necessarily, every red object is colored
ABSTRACT REFERENCE AND UNIVERSALS 69
and (16) as
(16a) Necessarily, every triangular object is shaped.
But while it yields what appear to be satisfactory results for (15) and (16),
the schema in question is too strong to serve as a general recipe for paraphras-
ing sentences incorporating abstract Singular terms. Many true sentences of
the form ( ..... F-ness ..... ) fail to support necessarily true sentences about
all the relevant F-objects. Consider (12) and (14). The schema we are con-
sidering suggests the following readings for these sentences:
(12a) Necessarily, for every red object and every white object, Quine
prefers the red object to the white object,
(14a) Necessarily, every wise man is virtuous. 14
Now, it should be obvious that (12a) fails to provide a satisfactory paraphrase
of (12). For all I know (12) is true, but I do not need to know anything about
Quine's view on colors to know that (12a) is false. Quine is only a contingent
being, and his preference for red objects over white objects is likewise merely
a matter of contingency, so that
(12b) For every red object and every white object, Quine prefers the
red object to the white object
is not necessarily true. But not even (12b) taken by itself serves as an ade-
quate paraphrase of (12); for it is quite consistent with Quine's preference of
redness over whiteness that there be an occasional pair of red and white
objects such that, because of features those objects have over and above their
color, Quine prefers the white object to the red object. In the same way,
(14a) fails to capture the force of (14). (14) is true, but (14a) is false; for it is
possible that there be wise men who, since they lack other virutes essential to
a flourishing moral life, fail to count as virtuous; and as we reflect on the
various philosophers we have known - the so-called paragons of wisdom, we
are likely to conclude that even the non-modal
(14b) Every wise man is virtuous
is false.
What all of this suggests is that the extreme nominalist will have to settle
for the watered-down
(12c) Quine prefers red things to white things
and
(14c) Wise men are virtuous
70 CHAPTER FOUR
as readings of (l2) and (14). Now, (12c) and (l4c) may appear to be satis-
factory paraphrases of (12) and (14); but to be sure that they are, we need
some account of the force of the expressions 'red objects', 'white objects',
and 'wise men' as they figure in these sentences. We have already seen that
they cannot be understood as stand-ins for 'every red object', 'every white
object', and 'every wise man'; but perhaps the extreme nominalist will want
to claim that they have the force of 'most red things', 'most white things',
and 'most wise men'. Unfortunately, if this is what those expressions mean,
then (12c) and (14c) fail to provide us with satisfactory paraphrases of our
original (12) and (l4). Just as it is consistent with Quine's preference for the
color red over the color white that there be an occasional red and white
object such that Quine prefers the white object to the red object, it is pos-
sible that (l2) be true and that many, perhaps most, red objects have proper-
ties Quine abhors, so that on balance he prefers white objects to red objects.
Likewise, just as the truth of (14) is consistent with an occasional wise man
who fails to measure up to our standards of the virtuous man, it is possible
that only a few of the men who are wise succeed in measuring up to those
standards. And since both of these things are consistent with the truth of (12)
and (14), it is clear that the extreme nominalist cannot claim that the un-
quantified expressions 'red things', 'white things', and 'wise men' that appear
in (l2c) and (14c) are to be understood as 'every typical red thing', 'every
typical white thing', and 'every typical wise man'. If it is consistent with the
truth of (12) that Quine generally prefers white objects to red objects, then
it is also consistent with truth of (l2) that he prefer many, perhaps most,
typical white objects to typical red objects; and if it is consistent with the
truth of (14) that there be few truly virtuous wise men, then it is also consist-
ent with the truth of (14) that many, perhaps most, of your run-of-the-mill
wise men not be virtuous.
But it might be claimed that we are being unfair to the extreme nominalist
here; for he may want to claim that (l2c) and (l4c) are to be understood as
(l2d) Other things being equal, given any red object and any white
object, Quine prefers the red object to the white object
and
(14d) Other things being equ'!l, every man that is wise is virtuous.
Now, (l2d) and (l4d) may indeed be satisfactory paraphrases of (l2) and
(l4); but the question we have to ask is whether the extreme nominalist has
the right to invoke them. Pretty clearly, he does only if he can provide some
ABSTRACT REFERENCE AND UNIVERSALS 71
sort account of the force of the ceteris paribus clauses that occur in these
sentences. Now, for the realist and the nominalist, there is no problem here.
The realist, for example, can say that in (12d) the force of the ceteris paribus
clause is something like 'supposing a pair of objects to agree in all their
attributes other than color' and that in (14d) the clause is to be understood
to mean 'where they have all the other virtues'. But what can the extreme
nominalist say here? Since he refuses to recognize attributes, he has to
resort to talk about predicates and their satisfaction. He has to say that in
(12d) the ceteris paribus clause has the force of 'supposing a pair of objects
to satisfy all and only the same predicates except .. .' and that in (14d) it has
the force of 'where men satisfy all other virtue-predicates'. But in neither case
does he succeed in giving us satisfactory readings of these clauses; for since
there is no guarantee that there will be predicates corresponding to every
feature with respect to which red and white objects can agree, it could turn
out that while a pair of white and red objects agree with respect to all predi-
cates other than their color-predicates, they, nevertheless, differ in says that
might prove significant for Quine's preferences. Likewise, since there is no
guarantee that every dimension of the virtuous life is captured by the totality
of virtue-predicates, it could happen that someone satisfying all the virtue
predicates fails to count as virtuous.
We seem to have reached a deadend here; but a remark ofOckham's points
to a way of handling (12) and (14) that avoids all of the difficulties we have
been considering. IS Ockham suggests that the analysis of an abstract term
might involve an appeal to expressions like 'qua' and 'insofar as', what he and
other medieval logicians called reduplicative expressions. Thus, he tells us that
the abstract term 'humanity' is synonymous with the expression 'man qua
man' or 'man insofar as he is man'. Now, if the extreme nominalist were to
take the reduplication-operator to be implicit in (12) and (14), he would have
an adequate analysis of these sentences. Thus, (12) becomes
(12e) Quine prefers red things qua red to white things qua white,
But, then, in (11 a) and (8a) the reduplicative expression 'wise qua wise' is
replaceable without loss of content by the concrete term 'wise' taken by it-
self; and in (15c) 'red things qua red' is synonymous with the expression 'red
things'. The difficulty, however, is that as it occurs in (14c) 'wise men qua
wise' cannot be replaced by the non-reduplicative 'wise men' or 'wise'; and
'red things qua red' as it occurs in (12c) cannot be replaced by the non-redu-
plicative 'red things'. But, then, the extreme nominalist is forced to say that
'wisdom' as it occurs in (14) is a different term from 'wisdom' as it occurs in
(8) and (11) and that 'redness' as it occurs in (12) is a different expression
from the 'redness' which appears in (15). We have already seen, however, that
'wisdom' is one and the same term in (8), (11), and (14) and that 'redness' is
the same expression in (12) and (15). The entailments pointed to in the pre-
vious section indicate this; they would not hold were these abstract terms to
have varying forces over these different contexts. But, then, wliile the appeal
to reduplication may enable us to handle the use of abstract singular terms in
sentences like (12) and (14), it does so at the cost of violating the Condition
of Semantic Uniformity.
It is, of course, possible that the extreme nominalist will come up with a new
account here, but the difficulties we have run up against suggest that his
attempt to establish the eliminability of abstract singular terms is no more
promising than the views that identify the referents of abstract singular terms
with sets of concrete objects or with spatially discontinuous objects. Initially,
it would appear that the nominalist's prospects for providing a satisfactory
account of abstract Singular terms are more promising. He has all of the
resources available to the extreme nominalist; but in addition he can appeal
to the notion of an individual attribute. The availability of this concept has
led some nominalists to suggest that sets are, after all, the referents of abstract
singular terms. D. C. Williams explicitly tells us that an abstract singular term,
'F-ness', names the set composed of all and only the objects that tenselessly
are individual F-nesses; and G. F. Stout frequently suggests that he accepts
this interpretation as wel1. 16 On the Stout-Williams view, then, 'redness' names
the set of all rednesses; 'wisdom', the set of all wisdoms; and 'triangularity',
the set of all triangularities. This sort of account represents a definite advance
over the extreme nominalist's appeal to sets as referents of abstract terms. On
that account, whenever concrete general terms are coextensional (Le., satisfied
by all and only the same objects), the referents of their abstract counterparts
74 CHAPTER FOUR
turn out to be identical. Thus, 'triangularity' and 'trilaterality' name one and
the same object on the extreme nominalist's account. On the Stout-Williams
approach, however, these two terms have different referents; for, on their
account, 'triangularity' names the set composed of this triangularity, that
triangularity, and ... ; whereas, 'trilaterality' names the set composed of this
trilaterality, that trilaterality, ... ; and these are two different sets.
It might seem that there remains one case where the coextensionality
of non-synonymous concrete terms forces the proponent of the Stout-
Williams approach to identify the referents of their abstract counterparts. I
am thinking of the case where the concrete terms are "empty", i.e., are satis-
fied by nothing at all. Here, there simply are no individual attributes, so that
the referent of an abstract term would appear to be the empty set. The
difficulty, however, is that given the identity-conditions for sets there is
just one empty set, so that the abstract counterparts of all empty predicate-
terms would appear to name one and the same thing. But, then, 'being a
unicorn' and 'being a griffin' would name the same thing; and obviously they
do not.
I do not know how Stout would handle this criticism; but Williams antici-
pates it and responds by denying that there are any unexemplified attributes. 17
He denies, that is, that terms like 'being a unicorn' and 'being a griffin' name
anything at all and concludes that there can be no problem about the numeri-
cal identity of their referents. He is, as a consequence, committed to the view
that sentences incorporating referring uses of expression like 'being a unicorn'
and 'being a griffin' are all false; and I am inclined to think that this is not the
case. But I am also inclined to think that the realist need not assume that
there are unexemplified attributes in developing his case against this nomi-
nalistic interpretation of abstract reference; for as Nicholas Walterstorff has
shown, the realist can indicate the shortcomings of the view by pointing out
that it is an essential or necessary feature of a set that it have just the mem-
bers that it does. IS Given any set, it is impossible that it have members other
than those it does; its very identity as an object is tied to its membership.
But, then, the nominalist of the Stout-Williams persuasion is committed to
the view that it is an essential or necessary feature of triangularity, for exam-
ple, that it have as its members just those individual attributes it does. But
since on his view, objects are triangular in virtue of being characterized by a
triangularity, this entails that it is impossible that there be either more or
fewer triangular objects than there tenselessly are. The realist can demonstrate
the inadequacy of this account by pointing out that neither of these things
are impossible. Since there could have been more, as well as fewer, triangUlar
ABSTRACT REFERENCE AND UNIVERSALS 7S
objects than there actually are, we can be sure that abstract singular terms are
not names of sets of individual attributes.
But if the appeal to sets is unsatisfactory, perhaps the nominalist can take
a reductionist position and claim that terms like 'triangularity', 'wisdom', and
'redness' are only superfically singular terms. He can say that abstract terms
are merely devices for abbreviating discourse about individual attributes.
Ockham embraced this sort of account for abstract terms from the Aristotel-
ian category of quality.19 On his view, 'wisdom' is a device enabling us to
make claims about the various individual wisdoms and 'redness', a device for
making claims about individual rednesses. If we generalize Ockham's inter-
pretation of these abstract terms, we get the contention that any sentence
incorporating one or more abstract singular terms can be paraphrased by
means of a sentence in which no abstract singular terms appear but what we
might call abstract general terms do appear. We have already seen how the
requirement that sentences incorporating abstract singular terms are to be
paraphrased by means of sentences that are both necessary and universal runs
into difficulties. Let us, then, suppose that on the eliminationist approach
proposed by the nominalist, (14) is to be paraphrased as
(l4) Wisdoms are virtues;
(8) as
(8c) Socrates exemplifies a wisdom;
and (17) as
(17a) Wisdoms are properties.
These paraphrases are unexceptionable; and for many other sentences incor-
porating abstract singular terms, this strategy gives us just the results we want.
It runs into trouble, however, when we confront sentences like
(27) Man (Le., mankind) is a substance-kind.
The treatment provided in the case of (17) suggests that (27) be read as
(27a) Humanities are substance-kinds;
but obviously (27a) is unsatisfactory; for to be a kind just is to be a universal,
so that the claim expressed by (l7a) is effectively contradictory, vis., that
certain abstract particulars are universals. Perhaps, however, the nominalist
would want to claim that sentences involving abstract singular terms like 'man'
are to be paraphrased in terms of sentences mentioning individual substances.
76 CHAPTER FOUR
Given this strategy, then, (27) is implicitly a claim about individual men.
Which claim? Well, clearly not the claim that men are substance-kinds; but
perhaps
(27b) Men agree in substance-kind
is a more promising candidate. As it stands, however, (27b) is inadequate for
the nominalist's purposes since it is just a disguised way of saying that men
belong to one and the same kind; and the disguise is not even very good. What
the nominalist needs, then, is an analysis of the notion of kind-agreement
which is expressed in (l7b) that does not involve the notion of a multiply
exemplifiable entity.
It turns out, however, that providing such an analysis is no easy task.
Clearly, agreement in kind cannot be understood merely in terms of member-
ship in a common set; for apart from the obvious extensionality-difficulties,
there are sets whose members are things of radically different sorts or kinds.
Neither is kind-agreement simply a matter of similarity; for while similarity
may be a necessary condition of kind-agreement, it is certainly not a suffi-
cient condition. Nor will it do to bring these two notions together and explain
kind-agreement in terms of similarity-classes or Similarity-sets (where objects
constitute a similarity-set just in case, given any pair of objects in the set,
there is no object outside the set which more closely resembles one of the
chosen objects than they resemble each other); for as Goodman's discussion
of the problem of imperfect community shows, this approach fails to gen-
erate anything like intuitively plausible partitions among kinds. The difficulty
here is nicely spelled out by W. V. Quine:
A final and more plausible suggestion here is that the nominalist explain
kind-agreement in terms of the possession of individual properties. 21 He can
say that substances agree in kind if they possess certain substance-determining
properties. If he is to provide anything like a genuine analysis of kind-agree-
ment, however, the nominalist who wants to follow out this suggestion is
ABSTRACT REFERENCE AND UNIVERSALS 77
going to have to tell us more; for kind-agreement does not consist in the pos-
session of just any individual properties; the properties that underlie any
particular case of kind-agreement among substances (e.g., that consisting in all
men's being the same kind of thing) must be such that in virtue of possessing
them substances do, in fact, belong to a single kind; but pretty clearly a mini-
mal requirement here is that the attributes in question all be things of the
same kind, all agree in kind. But, then, the nominalist succeeds in "analyzing
away" one instance of kind-agreement only at the expense of introducing
another; and if the argument just presented is sound, to eliminate this second
case of kind-agreement, he must invoke a third case of kind-agreement; and
so on ad infinitum.
Now, earlier when we examined the phenomena of predication and resem-
blance, we found the nominalist confronted with what might seem to be
similar regresses. Those regresses, I argued, are not vicious. I want to claim,
however, that the present infmite regress is vicious. In the earlier cases, the
nominalist was attempting to provide an account according to which it is
pOSSible, given any true subject-predicate sentence or any true resemblance-
claim, to point to the non-linguistic ground of the truth involved. There was
no claim to be analyzing away, eliminating, or translating out either the
subject-predicate form of discourse or the concept of resemblance. Conse-
quently, the fact that each attempt to ground the truth of a subject-predicate
sentence or a resemblance-claim presents us with a new subject-predicate
sentence or a new resemblance-claim does not threaten the success of the
accounts in question. The nominalist can successfully do what he wants to
do, vis., specify the grounds of true subject-predicate sentences or true res em-
blance-claims. In the present case, however, the aim is explicitly reductionis-
tic; the goal is that of showing the in principle eliminability of all Platonistic
locutions; but since the attempt to explain away any given case of kind-agree-
ment forces us to confront yet another case of kind-agreement, the nominalist
can never complete the translations required if his reductionistic program is
to be successful. This infinite regress, then, is vicious; and, consequently, this
final strategy for eliminating the notion of kind-agreement fails.
My discussion so far tempts one to conclude that neither the nominalist nor
the extreme nominalist has the resources for accommodating the truth of
sentences incorporating abstract referring devices. Such a conclusion would,
78 CHAPTER FOUR
Frenchman assertively utters the French translations of (27), (28), (14), (8),
and (17), he is talking about something quite different from what we are
talking about when we assertively utter their English counterparts.
This difficulty might lead some to abandon the metalinguistic approach
to abstract reference altogether; but a more sanguine response is to attempt
to preserve the insight underlying the Carnapian approach while accomodat-
ing the fact that abstract reference is not invariably language-bound. This,
at least, is the response we meet with in Wilfrid Sellars' treatment of abstract
reference. Sellars agrees with Carnap that the use of abstract singular terms
involves metalinguistic reference, but he denies that the metalinguistic force
of abstract terms is that captured by standard mention-quotation. Using
abstract singular terms, Sellars contends, we achieve a special kind of abstract
reference, one that cuts across language-barriers. 23 Central here is Sellars'
view that different languages can incorporate linguistic expressions that come
to much the same point. He wants to claim, for example, that the French
'homme', the Spanish 'hombre', the German 'Mensch', and the Italian 'uomo'
all agree in being subject to roughly the same set of linguistic rules, those
which govern the use of the English 'man'. All these terms function in much
the same way as responses to perceptual situations; they have roughly the
same force in inferential contexts; and they are used in much the same way
in action-guiding contexts. To capture the similarities here, Sellars suggests
that we adopt a special form of quotation - what he calls dot quotation. In
rough terms, we can say that the application of dot quotation to a term 'T',
has the effect of creating a metalinguistic common noun, T, which is true of
all those expressions (in the sense of tokens rather than types) which in their
own languages are subject to the same set of linguistic rules which govern the
use of 'T's in the quoting language. On Sellars' account, then, 'man' is a term
satisfied by all those expressions, regardless of their language, which are
subject to the same set of rwes governing the use of English 'man's.
Now, what Sellars wants to claim is that the use of abstract singular terms
has the effect of invoking the sort of metalinguistic reference involved in dot
quotation; and he wants to claim that by appealing to this special quoting
device, we can make perspicuous what is involved in the use of such sentences.
Thus, on Sellars' account, (27) is to be read as
(8), as
(8f) ,Wise's can be truly predicated of Socrates;
and (17), as
(17c) ,Wise's are adjectives.
It is difficult not to be impressed by this account. My own view is that
it is the most sophisticated alternative to the realistic interpretation of abstract
reference that has yet been presented; and the view becomes all the more im-
pressive when one notes that on Sellars' interpretation of the phenomenon of
language, one is not even committed to an ontology of individual attributes;
for Sellars wants to insist that talk about linguistic expressions can be recast
in terms of talk about persons as speaking, inscribing, and the like. I shall not,
however, concern myself with this claim; for even if Sellars is wrong in his
contention that his account is consistent with the rigors of extreme nominal-
ism, it should be clear that it represents a genuine alternative to the account
of the realist. What I want to ask is whether it provides a viable alternative.
Now, it seems to me that Sellars' account handles most of the sentences in
which abstract singular terms appear. The fact is, however, that those are not
the only sentences by means of which we make claims about the objects,
whatever they may turn out to be, that are the apparent referents of abstract
singular terms. To see this, let us focus on
(17) Wisdom is a property.
(17) is a category-assertion; but if we suppose what is likely true - that wis-
dom is the property most frequently ascribed to Socrates, then the category-
subsumption effected by (17) can also be effected by
(29) The attribute most frequently ascribed to Socrates is a property.
Now, (29), no less than (17), appears to involve a reference to an abstract
entity, a universal; and if Sellars is to provide a satisfactory alternative to the
realist's interpretation of abstract reference, he must come up with an account
of sentences like (29). What Sellars will claim is that (29) makes a claim about
certain linguistic expressions. It should be clear, however, that he cannot say
that (29) is about the linguistic expressions that are the attribute most
frequently ascribed to Socrates's; for while (29) is true,
(29a) 'The attribute most frequently ascribed to Socrates's are adjectives
is false inasmuch as the attribute most frequently ascribed to Socrates's are
noun-phrases and not adjectives.
82 CHAPTER FOUR
What Sellars will argue, of course, is that (29) is a claim about wises.
Towards making this clear, he might insist that the definite description that is
the grammatical subject of (29) is really just a stand-in for 'the general term
most frequently predicated of Socrates', so that (29) is to be read as
(29b) The general term most frequently ascribed to Socrates is an
adjective.
Unfortunately, (29b) fails as a translation of (29). To see this, we need only
reflect on the fact that (29b) and (29) do not have the same truth conditions.
(29b) can serve as an analysis of (29) only if the general term to which it
points turns out to be wise. But is it, in fact, true that wisdom can be the
attribute most frequently ascribed to Socrates only if wises are predicated
of him more frequently than general terms of any other functional sort? I
think not; for wises do not exhaust our resources for ascribing wisdom to
Socrates. We can ascribe wisdom to him by saying that he is wise; but we can
also do this by saying that he exemplifies the virtue to which A1cibiades
aspires abov~ all others, that he exemplifies the property Quine prefers to all
others, that he exemplifies the virtue mentioned on the third line of page 157
of Jowett's translation of the Republic, etc. But, then, it could turn out that
while no attribute is ascribed to Socrates as frequently as wisdom, some ex-
pression other than wise (say, a common noun like man) is predicated of
him more frequently than any other general term.
Sellars might, however, deny that any kind of metalinguistic interpretation
of sentences like (29) is required for the success of his project. He might
argue that the opponent of realism merely needs metalinguistic readings of
sentences which contain what the realist takes to be proper names of univer-
sals, abstract singular terms. Having given such readings for these sentences,
he could provide a schema which shows the truth of sentences which (like
(29)) use definite descriptions as devices for achieving abstract reference to be
parasitic on the truth of sentences where abstract reference is secured by the
use of abstract Singular terms. Such a schema would be required for every
sentential context where we appear to be talking about universals; but we get
the drift of the proposal if we focus on the kind of schema required to handle
(29). Here, the context to be handled is ' is a property'.
The schema might go as follows:
An object x is a property if and only if either
(a) ys are adjectives (where 'x' is an abstract singular term and 'y',
its concrete counterpart); or
ABSTRACT REFERENCE AND UNIVERSALS 83
quoted term T bears to a dot quoted term 'T" just in case 'T's are coexten-
sional with T's. Thus, 'Socrates' is materially equivalent to the teacher of
Plato' and featherless biped', to man. Material equivalence, then, is a relation
that obtains between dot quoted expressions which are true of general terms
as well as those which are true of Singular terms. But if Sellars were to limit
himself to the case where we have material equivalence between dot quoted
terms satisfied exclusively by Singular terms and ifhe were to mark this more
specialized use of the notion by the symbol 'ME*', then he could read (30) as
(30b) 'The attribute most frequently ascribed to Socrates' ME*
wisdom.
But would even this very exotic reading of (30) do? I think not. The first diffi-
culty with (30b) bears on the term 'wisdom' that appears there. We need to
know how we are to understand this term. What Sellars tells us is that when
we apply dot quotation to an abstract singular term, the result must be recon-
structed in terms of double dot quotation. Thus, (30b) must be understood as
(30c) The attribute most frequently ascribed to Socrates. ME* "wise";
but, then, we have the result that the term 'wisdom' in (30) is a different
expression from the term 'wisdom' as it appears in (8), (14), and (17); and so
Sellars' appeal to the notion of material equivalence works only at the ex-
pense of violating the Condition of Semantic Uniformity.
But there is a second difficulty here. The notion of material equivalence is
simply too weak to serve as a reconstruction of our notion of identity. Ac-
cording to the proposal, we are to read sentences of the form 'x=y' as "x'
ME* 'Y"; but the fact is that a sentence of the latter form can be true even
where a sentence of the former form is false. Thus, 'Pegasus' is materially
equivalent to the winged horse of Bellaraphon'; but since 'Pegasus' is a fic-
tional expression,
(31) Pegausus is identical with the winged horse of Bellaraphon
is false; and even more paradoxically, since Pegasuss and 'Paul Bunyan's are
empty terms,
(32) 'Pegasus' ME* Paul Bunyan'
is true; but by no stretch of the Meinongian imagination is
(33) Pegasus is identical with Paul Bunyan
true.
ABSTRACT REFERENCE AND UNIVERSALS 8S
VI. CONCLUSION
NOTES
13 See, e.g., Summa Logicae, I. 6-8; pp. 58-68 in Loux, Ockham's Theory of Terms.
14 The difficulties I point to here are discussed in Arthur Pap's 'Nominalism, Empiric-
ism, and Universals, I,' Philosophical Quarterly, 1959, pp. 334-338 and in Chapter 9 of
Wolterstorffs On Universals. Pap agrees with me in seeing these difficulties as significant.
As I have already indicated, Wolterstorff does not. I suspect that the divergence of
opinion here is traceable to the fact that whereas Pap and I accept the principle of
theoretical simplicity as a methodological guide in ontology, Wolterstorff does not. See
once again his remarks about 'Ockham's Razor' on p. 127 of On Universals.
15 Summa Logicae I. 8; p. 65 in Loux, Ockham 's Theory of Terms.
16 See once again 'On the Elements of Being, I,' and Stout's 'On the Nature of Univer-
sals and Propositions', British Academy lecture, 1914, in J. N. Findlay's Studies in
Philosophy (Oxford), 1966, pp. 5-24. A more elaborate version of the account, which
identifies the referent of an abstract singular term with the class composed of the objects
satisfying its concrete counterpart as well as the various individual attributes those
objects exemplify, is found in Wolterstorffs 'Qualities,' Philosophical Quarterly, 1960;
reprinted in Loux, Universals and Particulars, pp. 87-105. See, especially pp. 95-99.
The reason for the hedge on Stout's view is that it is unclear whether he takes abstract
particulars to form extensional or intensional wholes. He indiscriminately calls the
wholes "classes" and "sorts or kinds".
17 See Williams' Principles of Empirical Realism (Springfield Illinois: Charles Thomas),
1966, p. 239.
18 On Universals, pp. 178-181.
19 See, e.g., Summa Logicae I. 55, pp. 178-180 in Loux, Ockham's Theory of Terms.
20 'Natural Kinds,' Ontological Relativity, pp. 120-121; for Goodman's original state-
ment of the difficulty, see The Structure of Appearance, 3rd edition (Dordrecht: Reidel),
1977, pp. 117-119.
21 This kind of view is suggested by D. C. Williams in 'On the Elements of Being, I,'
p.I0.
22 For Ockham's remarks here, see Summa Logicae, I. 72; p. 210 in Loux, Ockham's
Theory of Terms and Summa Logicae I. 18-25; pp. 88-104 in Loux, where Ockham
deals with the predicables; for Carnap's view, see The Logical Syntax of Language
(Patterson, New Jersey: Littlefield, Adams, and Co.), 1959, pp. 284-314.
23 See, e.g., 'Abstract Entities,' Review of Metaphysics, 1963; reprinted in Loux,
Universals and Particulars, pp. 156-205.
24 See Science and Metaphysics, pp. 84-87 for a discussion of this relation.
25 I am indebted to my colleogue, Richard Foley, for this point.
CHAPTER FIVE
So far I have given the impression that realists stand as a united front ready to
do battle with their common foes, the nominalists and extreme nominalists.
The fact is, however, that realists have argued amongst themselves. Perhaps
the central issue in such arguments has been the relationship between univer-
sals and the concrete objects that exemplify them. This issue was certainly
central in the philosophical relationship between Plato and Aristotle. In his
middle dialogues at least, Plato held that there is an asymmetry between
universals and particulars; he held that while concrete objects are what they
are and so exist only because they exemplify universals, the existence of
universals is independent of the existence of the various particulars that
happen to exhibit them. This view is metaphorically expressed in Plato's
picture of universals as entities that exist in a separate "world". 1 Aristotle,
on the other hand, seems to have denied that any such asymmetry exists;
for while conceding that the existence of concrete objects (what in the
Categories he calls "first substances") presupposes their exemplification of
universals, he tells us that the existence of universals is logically tied to the
existence of the concrete objects exemplifying them. Thus, in the Categories,
Aristotle says that "if these last (i.e., first substances) did not exist, it would
be impossible for anything else to exist.,,2 Just how this is to be understood
is not altogether clear in the Categories; but in other works, Aristotle seems
to imply that for a given universal, U, there is no time at which U fails to be
exemplified.
Controversy over this issue was also a dominant issue in medieval discus-
sions of universals, where defenders of the so-called ante rem ("before the
thing") theory of universals took the Platonic view, and defenders of the in re
("in the thing") theory sided with Aristotle. Even in our own day, realists
have argued both sides of the issue. Alan Donagan seems to be siding with
Plato when he argues that there is nothing incoherent in the notion of a uni-
versal that is never exemplified;3 whereas, Gustav Bergmann, who insists that
"every character is at least once exemplified,,,4 takes a mild form of the
Aristotelian view.
Which side is right here? Is the existence of universals independent of the
existence of objects which exemplify them, or is it a necessary truth that
every universal is exemplified? Towards answering this question, I want to
invoke the metaphysical framework of possible worlds ontology. Some will
doubtless object that this is to explain the obscure by the more obscure. I
disagree, but those who find talk about possible worlds objectionable can
TOWARDS A REALISTIC ONTOLOGY 93
A possible world is a state of affairs of some kind - one which could have obtained if it
does not. Hubert Horatio Humphrey's having run a mile in four minutes, for example, is
a state of affairs that is clearly possible in the relevant sense; his having had a brother
who never had a sibling is not. Furthermore, a possible world must be what we may call
a fully determinate state of affairs. Humphrey's having run a four-minute mile is a pos-
sible state of affairs, as, perhaps, is Paul X. Zwier's being a good basketball player. Nei-
ther of these, however, is fully determinate in that either of them could have obtained
whether or not the other had. A fully determinate state of affairs, S, let us say, is one
such that for any state of affairs, S', either S includes S' (that is, could not have obtained
unless S' had also obtained) or S precludes S' (that is, could not have obtained if S' had
obtained). 5
exemplified by every object in every possible world. But, then, for any possi-
ble world, W, if one of these universals exists in W, then there is at least one
object in W such that it exemplifies that universal in W - the universal in
question; and, of course, if there are any other objects in W, they too will
exemplify that universal in W.
Transcendental universals, then conform to the Aristotelian account. The
Aristotelian account also works for universals which, while not being tran-
scendental, are like transcendental universals in being necessarily self-exempli-
fied. Take, for example, the property of being a non-number or the property
of being non-human. They are properties such that they are self-exemplified
in every possible world in which they exist; consequently, there is no possible
world, W, such that either of these properties exists in Wand fails to be exem-
plified in W;.if either exists in a possible world, then there is at least one
object in that world which exemplifies the property - itself.
The Aristotelian account holds as well for some universals that are not self-
exemplified. Take, for example, the property of being prime. Being prime is
a property that is exemplified in every possible world in which it exists. To
see this, we need mereiy note that propositions like those expressed by
(1) Three is prime,
(2) Five is prime,
and
(3) Seven is prime,
are necessary truths. They are propositions true in every possible world. They
are, moreover, propositions to the effect that certain objects - the numbers
in question - exemplify the property of being prime. Consequently, their
truth presupposes the existence of the numbers in question. Put in the lan-
guage of possible worlds, there is no possible world in which these proposi-
tions are true and the numbers in question do not exist. But since they are
true in every possible world, the numbers in question exist in every possible
world. They are, we might say, necessary beings. The numbers, three, five,
and seven, then, exist in every possible world; and in every possible world,
they exemplify the property of being prime. Consequently, there is no world
in which the property of being prime exists unexemplified. And the argument
used to show that being prime conforms to the Aristotelian picture applies
in the case of many other universals. Indeed, we can say that every universal,
U, such that there is a necessarily true proposition to the effect that some
object exemplifies U conforms to the Aristotelian account. Thus, being a
TOWARDS A REALISTIC ONTOLOGY 95
color is a universal that cannot exist unexemplified; for the proposition ex-
pressed by
(4) Red is a color
is necessarily true; it is true in every possible world; but since its truth in a
possible world, W, presupposes that red exists in W, we can be certain that the
universal, being a color, is exemplified in every possible world. The same
holds for the universal, being a virtue; for since the proposition expressed by
(5) Courage is a virtue
is necessarily true, being a virtue is exemplified in every possible world; and
likewise, for many other universals.
There are, then, many universals that conform to the Aristotelian account;
nor would Platonists disagree here; but while conceding that many universals
cannot exist exist unexemplified, Platonists have insisted that some universals
fail to conform to the Aristotelian account. I think that they are right here.
To see this, we need only consider a universal like red. We know already that
red exists in every possible world. But is red exemplified in every possible
world? I think not. Clearly it is possible that there be no red objects; there
are, that is, possible worlds in which there are no objects that are colored red.
Aristotle would perhaps disagree; but I think that it is intuitively obvious that
he would be wrong in so disagreeing. Likewise, since the proposition express-
ed by
(6) Mankind is a substance-kind
is a necessary truth, mankind exists in every possible world; but clearly in
some possible worlds it fails to be exemplified; for the existence of individual
human beings is merely contingent; it is clearly possible that there be no
human beings at all.
We can generalize here. Every universal which is exemplified exclUSively by
beings whose existence is merely contingent (Le., beings that actually exist,
but whose non-existence is possible) fails to conform to the Aristotelian
account. To see this, we need merely reflect on the fact that every universal
is a necessary being; for given any universal, U, there is a necessarily true pro-
position to the effect that U is an attribute. Such a proposition is true in
every world, so that U must exist in every possible world; but where U is a
universal exemplified only by contingent beings, there are possible worlds in
which U exists, but no objects exemplifying U exist. But, then, while the
Aristotelian account holds for many universals, it fails as a general account of
96 CHAPTER FIVE
I have agreed, then, with Plato in holding that not all universals are necessarily
exemplified. Every universal exists in every possible world; but the universals
that are exemplifiable exclusively by contingent beings are unexemplified in
TOWARDS A REALISTIC ONTOLOGY 97
some possible worlds. Plato tells us other things about universals. He suggests
that universals are ingenerable and incorruptible. 7 Intuitively, the idea here is
that it is impossible for a universal either to come into existence or pass out
of existence, that the existence of universals is necessarily without beginning
and end. We can formulate these intuitions in terms of the framework ofpos-
sible worlds if we say, first, that an object x is ungenerated in a world W just
in case x exists in W but does not come into existence in Wand, second, that
an object, x, is uncorrupted in W just in case x exists in W but does not pass
out of existence in W. Thus, wherever W is a properly temporal possible world
(i.e., a world with some sort of temporal framework), an object, x, is both
ungenerated and uncorrupted in W if and only if there is no moment, t, in
the history of W such that x does not exist at t. Now, let us say that an object,
x, is ingenerable just in case x actually exists and, for every world, W, if x
exists in W, then x is ungenerated in Wand that an object, x, is incorruptible
just in case x actually exists and for every world, W, if x exists in W, then x is
uncorrupted in W.
I have so far argued that every universal is a necessary being; but as I have
understood the notion of a necessary being, this thesis does not entail that
every universal is both ingenerable and incorruptible; for as I have used the
term, a being is necessary merely if it exists in every possible world. To be
ingenerable and incorruptible, an object must exist at every moment in the
history of each properly temporal world in which it exists. But while the
concepts here may be different, the general line of argument used to establish
the necessary existence of universals establishes their ingenerability and in-
corruptibility as well. We have already seen that for every universal, U, there
is a necessarily true proposition to the effect that U is an attribute. Let us
focus on one of these propositions - the proposition that red is an attribute.
This proposition is necessary; but this is not just to say that it is true in every
possible world. If this proposition is necessary, it must as well be the case that
for every properly temporal world, W, the proposition is true at every mo-
ment in the history of W. Furthermore, for any properly temporal world, W,
and any moment, t, in the history of W, this proposition can be true at t only
if red exists at t. But, then, for every properly temporal world, W, there is no
moment, t, in the history of W such that red does not exist at t. Red, then,
exists in every possible world and at every moment in the history of each pro-
perly temporal world. But, then, red is both ungenerated and uncorrupted in
every properly temporal world in which it exists. If there are possible worlds
that have nothing corresponding to a temporal framework, then presumably,
it is incoherent to speak of anything's happening in those worlds, to speak of
98 CHAPTER FIVE
any change in those worlds. Thus, if there are any such worlds, red does not
come into existence or pass out of existence in them either. But, then, red is
ungenerated and uncorrupted in every world in which it exists; but since it
exists in the actual world, it is both ingenerable and incorruptible; and this is
a perfectly general argument, one holding for any universal. Now, although
my use of 'necessary being' is not such that to say that a being is necessary is
to imply that it is ingenerable and incorruptible, most metaphysicians have
used the term in this stronger sense; but, then, the argument just presented
shows that universals are necessary beings not just in my weak sense but also
in the strong sense that has been dominant in the history of metaphysics.
Plato also tells us that universals are unchangeable or immutable. 8 The
suggestion here is that universals are not only not subject to those changes
which are generations and corruptions, but that they are not subject to
changes of any kind. Is Plato right here? Well, that depends on what one
counts as a change. If one says that an object, x, undergoes a change in a
possible world, W, just in case there is a time, t, in the history of W such that
x possesses (or lacks) a property Pat t and a later time, t+n, in the history of
W such that x lacks (or possesses) P at t+n, then he must conclude that Plato
is wrong here; for surely universals can exemplify different properties at dif-
ferent times. Suppose, for example, that at a time, t, in a world, W, Socrates
exemplifies a property, P, and that at a later time, t+n, in W, Socrates comes
to lack P; then, at t, P exemplifies the property of being exemplified by
Socrates and at t+n, Placks this property. Again, suppose that there is a time,
t, in a world, W, such that Quine thinks of a certain relation, R, at t and that
there is a later time, t+n, in W such that at t+n Quine is engrossed in reflec-
tions on some other subject; then, at t, the relation R has a property it lacks
at t+n - the property of being thought of by Quine.
One might object, however, that the properties I have pointed to are all
relational (or quasi-relational) properties - those involving a universal's,being
exemplified by something else or its being the object of someone's mental
states or acts. Now, one might argue that changes with respect to properties
like these are not "real" changes on the part of universals. The point here
would be that a universal's exemplifying or failing to exemplify these proper-
ties depends not on how it is with the universal but on how it is with some
other object. It is, one might insist, Socrates and Quine who actually undergo
the relevant changes here; it is only in a derivative sense that the universals in
question can be said to change. Now, while I agree that the point being made
here is a bit vague, I am inclined to think that it involves a genuine insight. As
far as I can tell, universals can change only with respect to their relational (or
TOWARDS A REALISTIC ONTOLOGY 99
and vice versa. That this account of their identity-conditions is correct be-
comes obvious when we reflect on the fact that for every universal, U, there is
an attribute which is exemplified by U in every possible world and exemplified
by something distinct from U in no world at all - the attribute of being U. 13
The Quinean may object that the appeal to attributes like that of being red-
ness and that of being triangularity are useless in enabling us to determine whe-
ther a universal, U, and a universal, U', are the same or different; for to know
for any attribute of the form, being U, whether a universal, U', exemplifies
that attribute, we would first have to know whether U and U' are the same or
different. I fail, however, to see how this objection is relevant. The task was
one of specifying the conditions under which a universal, U, and a universal,
U', are the same or different; but the objection here is that we could never
apply the criterion presented for making judgments about the identity of uni-
versals. The response, I take it, is that my criterion of identity was not meant
to handle epistemological difficulties concerning our judgments about the
identity and difference of universals; it was meant to be an ontological account,
an account about how things are and not about we find out how they are.
But if the critic persists, insisting that we provide such an epistemological
account, my counter is to demand that he provide an airtight set of necessary
and sufficient conditions for judgments of identity about things other than
universals. The fact is, however, that it is notoriously difficult to specify such
conditions even for things as ontologically benign as material bodies and
persons. It may even turn out that there are no such conditions, that in some
cases, it is just impossible for us to say whether we have one and the same
material object/person or two different material objects/persons. But whether
this is so or not, it is surely not the case that the difficulty of providing neces-
sary and sufficient conditions for judgments of identity about material bodies
and persons calls into question the legitimacy of the ontological framework
of which they are a part. What I want to suggest is that the same is true of
universals. Just because it may be difficult for us to determine for a pair of
universals, U and U', whether U and U' are the same or different or to speci-
fy, in a non-question-begging way, how we know that U and U' are the same
or different, it does not follow that there is anything suspect about the ontol-
ogical framework of which universals are a part.
But while there may be a problem about specifying criteria for judgments
about the identity of universals, we can make some headway towards deter-
102 CHAPTER FIVE
mining how many universals there are. It seems to me that we can easily show
that there is at least a denumerable infinity of universals, i.e., as many as
there are natural numbers. To show this, we need merely reflect on a univer-
sal like mankind which is exemplifiable exclusively by particular substances
(material bodies and persons). Let us call such a universal a first order univer-
sal. Mankind, we can say, exemplifies the property of being a first order
universal; and if we say that a universal is a second order universal just in case
it is exemplifiable exclusively by universals which are first order universals,
then we can say that the property of being a first order universal has the pro-
perty of being a second order universal. This property, in turn, exemplifies
the property of being a third order universal; and so on ad infinitum. Be-
ginning with universals like mankind, then, we can sketch out a procedure for
generating an infinite hierarchy of universals, each member of which corre-
sponds with one of the natural numbers.
I think, however, that if we make a couple of fairly plausible assumptions,
we can establish an even stronger thesis - that there is a non-denumerable
infinity of universals, that there are as many universals as there are real num-
bers. The assumptions in question are, first, that the various real numbers are
existent objects and, second, that wherever an object, x, enters into a relation,
R, with an object, y, there is a property which x exemplifies - that of enter-
ing into R with y. Armed with these assumptions, it is an easy task to show
that there is a non-denumerable infinity of universals. Take the number one;
it is greater than the fraction ~; consequently, it has the property of being
greater than ~. Furthermore, for any number, n, such that n is greater than
~ but less than one, one has the property of being greater than n; but there
is a non-denumerable infinity of such numbers; therefore, there is a non-
denumerable infinity of such properties; and every one of them is exemplified
by the number one.
VI. CONCLUSION
and their role in aesthetic judgments. It could also lead us into the areas of
ethics and political philosophy where questions about the role of universals
as moral standards have vexed philosophers since at least the time of Plato.
In what remains of this book, however, I want to leave such interesting
questions aside and focus on some more specifically ontological questions
about universals. The questions I want to consider bear on the way in which
the concept of a universal interacts with another central concept in ontology,
the concept of substance. The issues I want to consider all follow from the
question of how universals figure in the ontological structure of substances.
This questIon, it turns out, forces us to examine a variety of perennial meta-
physical problems, problems about the individuation of substances, their
identity-conditions, problems about change, problems about the concept of
essence, and so on. Such problems have never been far from the surface when
the issue of universals has been debated. To these problems I now want to
turn.
NOTES
edited by Richmond Thomason, (New Haven: Yale University Press), 1974, especially,
pp.152-l60.
13 It might be thought that one could provide a more substantial account of the identity
conditions for universals as follows: for any universal, U, and any universal, U', U is
identical with U' just in case necessarily, for any person, P, and any object, x, U is be-
lieved by P to be exemplified by x if and only if U' is believed by P to be exemplified by
x. But while this sort of account appears promising, it fails since people can have incom-
patible de re beliefs. On this account, then, we would have to deny that wisdom is
identical with wisdom since it is clearly possible that there be a person, P, such that it is
the case both that wisdom is believed by P to be exemplified by an object x and that
wisdom is believed by P not to be exemplified by x.
PART TWO
SUBSTANCES
CHAPTER SIX
I. BARE SUBSTRATA
having some texture or other which we also associate with the ball. Whatever
is the literal bearer of the properties associated with the ball has that property
as well. But, then, the property-independence we must read into this object
forces us to say that whatever possesses the properties associated with the ball
is something which of itself has no texture at all. Likewise, this object is
something which in itself lacks the spherical shape associated with the ball;
but neither is it something which in itself is of some other shape; for it also
possesses the generic property of being of some shape or other. In its intrinsic
nature, then, the possessor of the properties associated with the ball lacks
that property too. Again, the possessor of the various properties associated
with the ball is something which in itself is not two inches in diameter; nor
since we associate with the ball the generic property of being of some size or
other, is it a thing which of itself has some other size. It is also something
which in itself lacks the odor characteristic of rubber and the property of
weighing five ounces; and its lacking these properties does not entail its hav-
ing other odor-properties or weight-properties which can compete with these;
for it is the literal possessor of the generic properties of being of some odor
and being of some weight and, so, of itself, it must be something without an
odor and without a weight.
But just what is this object that is the literal possessor of these properties?
We want to say that it is the ball, the piece of rubber before us, that possesses
these properties; but now we seem forced to think of that piece of rubber in
extraordinary terms - as something which in itself is of no color, texture,
size, shape, weight, or odor. It is astonishing that our ball should have turned
out to be something that can be apprehended independently of all these kinds
of properties. But even the piece of rubber thOUght of in these extraordinary
terms disappears when we realize that we associate with it the property of
being made of rubber. That too is a property we associate with the ball; and
the assumption about property-independence forces us to say that whatever is
the literal possessor of the properties associated with the ball is in itself some-
thing that does not involve this property. It is something which in itselflacks
even the property of being made of rubber. Nor again is that object something
which in itself has some other material characterization; for there is the gen-
eric property of being made of some stuff or other; and the object in question
must be the literal possessor of that property as well. In itself, then, the
possessor of the properties associated with the ball is something which has no
property in virtue of which it is any recognizable kind of stuff at all.
What the assumption drives us to, then, is the view that the ordinary way
of thinking about the ball is wrong. The ball does not literally possess the
110 CHAPTER SIX
properties associated with it; something else does. This something else is just
that - a something. It is an object which in itself lacks all the properties
associated with the ball. Nor is it something which in itself has properties
other than those associated with the ball; for if there are any properties
possessed by that something which are not, in ordinary parlance, attributed
to the ball, the argument which so relentlessly stripped the properties "of'
the ball from their literal possessor applies in the case of these properties too.
In itself, that object lacks these properties as well as any generic properties to
which they are subordinated. What the assumption about property-independ-
ence forces us to say, then, is that the literal possessor of the properties
ordinarily attributed to the ball is something which underlies those properties
but in itself is lacking in all properties. Philosophers who have accepted this
assumption have appropriately called this something bare substratum.
Now, the argument presented in the case of the red ball is perfectly general;
it can be applied in the case of any ordinary object. Thus, the assumption
about property-indepenrlence leads us to the general view that what I have
called substances - ordinary objects like material bodies, plants, animals, and
human beings - are simply complexes or wholes whose constitutents are,
first, the various properties associated with them and, second, a bare substra-
tum. This substratum is the literal possessor of the properties with which it
is co-present; and it is because they are all possessed by it that they are pro-
perties which, in ordinary parlance, we think of as properties "of' one and
the same substance.
To one unacquainted with the history of metaphysics, this general line of
argument and the view to which it gives rise are likely to appear shocking.
Indeed, one might fmd it hard to believe that any philosopher would ever
have argued in this way to the doctrine of bare substratum. In fact, this line
of argument or something closely resembling it can be found in the writings
of some of the greatest thinkers in the history of ontology. In the third chap-
ter of Metaphysics Z, for example, we find Aristotle saying:
When all else is stripped off evidently nothing but matter remains. For while the rest are
affections, products, and potencies of bodies, length, breadth, and depth are quantities,
and not substances ... , but the substance is rather that to which these belong primarily.
But when length and breadth and depth are taken away we see nothing left unless there
is something that is bounded by these; so that to those who consider the question thus
matter alone must be substance. By matter I mean that which in itself is neither a parti-
cular thing nor of a certain quantity nor assigned to any other of the categories by which
being is determined. For there is something of which each of these is predicated, whose
being is different from that of each of the predicates (for the predicates other than sub-
stance are predicated of substance, while substance is predicated of matter). Therefore,
TWO THEORIES OF SUBSTANCE 111
the ultimate substratum is of itself neither a particular thing nor of a particular quantity
nor otherwise positively characterized nor yet is it the negations of these, for negations
will belong to it only by accident. 2
Aristotle wants to claim, then, that every substance incorporates what he calls
matter, something which "of itself is neither a particular thing nor a particu-
lar quantity nor otherwise positively characterized." He deviates from the line
of argument I have just set out in denying that this must be construed as the
possessor of all the properties ("predicates") associated with an ordinary
substance. Most of those properties, he wants to claim, can be thought of as
possessed by the substance itself, where the substance is thought of as some-
thing which in itself has a being distinct from and independent of the being of
those properties. One property, however, vis., that of being the kind of sub-
stance in question, cannot be thought of as possessed by that object; its pos-
sessor must be something "whose being is different from" the property in
question; and here what I have called bare substratum and what Aristotle calls
matter enters the account.
Aristotle, then, seems to endorse the line of argument I have presented,
but wants to deny that it forces us to construe matter or bare substrata as the
subject of all the properties associated with a substance. John Locke, on the
other hand, presents us with an account of property-possession which con-
strues all the properties associated with a particular substance as having a
single possessor or subject:
Though, in the meantime, it be manifest, and everyone, upon inquiry into his own
thoughts, will find that he has no idea of any substance, e.g., let it be gold, horse, iron,
man, vitriol, bread, but what he has barely of those sensible qualities, which he supposes
to inhere; with a supposition of such a substratum as gives, as it were, support to those
qualities or ideas, which he has observed to exist united together. 3
Locke wants to deny, then, that the appeal to substratum involves any modi-
fication of our pre-philosophical thinking about substances. This notion is
presumably implicit in ordinary thought and talk about things like "gold,
horse, iron, man, vitriol, bread." Nonetheless, Locke's account of how this
notion enters our thinking about substances and his characterization of the
notion corresponds with the line of argument presented in the case of the red
ball. Essentially, Locke wants to say, we get the notion of underlying substra-
tum by a process of conceptually stripping off the various properties (what he
calls qualities) from the ordinary objects with which we associate them. This
comes out in the following famous passage from the Essay where Locke sub-
stitutes the term 'substance' for 'substratum':
112 CHAPTER SIX
So that if anyone will examine himself concerning his notion of pure substance in
general, he will fmd he has no other idea of it at all, but only a supposition of he knows
not what support of such qualities which are capable of producing simple ideas in us;
which qualities are commonly called accidents. If anyone should be asked, what is the
subject wherein colour or weight inheres, he would have nothing to say, but the solid
extended parts; and if were demanded what is it that solidity and extension adhere in,
he would not be in a much better position than the Indian before mentioned who, saying
that the world was supported by a great elephant, was asked what the elephant rested
on; to which his answer was - a great tortoise; but being pressed to know what gave
support to the broad-backed tortoise replied - something, he knew not what . ... The
idea that we have, to which we give the general name substance, being nothing but the
supposed, but unknown support of those qualities we find existing, which we imagine
cannot subsist since re sub stante , without something to support them, we call that sup-
port substantia, which according to the true import of the word is, in plain English,
standing under or upholding. 4
Thus, Locke wants to claim that underlying the properties associated with
an ordinary object is something which, because it lacks all properties in virtue
of which it mi~t be characterized, is unknowable. Locke's notion of the
underlying subject of properties has its proponents in our own century. Thus,
in retrospect, the later Russell tells us that the theory of substance developed
in his famous paper, "On the Relations of Universals and Particulars," involves
an implicit appeal to the notion of bare substrata;S and more recently Gustav
Bergmann and his followers have attempted to resurrect this notion of an
underlying possessor of properties. Bergmann insists that ordinary objects
have among their constituents what he calls bare particulars. He tells us that
while these entities are the literal exemplifiers of the properties entering into
the constitution of ordinary objects, they are things which "neither are nor
have natures; any two of them are not intrinsically but only numerically
different."6
But despite its long and distinguished history, the notion of bare substratum
has come in for some hard knocks. Its most relentless critics have been
philosophers of an empiricist temperament. They have contended that since
bare substrata are objects which in themselves lack all properties, they could
never be the objects of empirical acquaintance, whether perceptual or intro-
spective. In general, such philosophers are suspicious of any kind of appeal to
entities which transcend both perceptual and introspective experience; but
they have found the contention that it is only bare substrata that can proper-
ly function as the subjects of property-attributions positively outrageous. The
TWO THEORIES OF SUBST ANCE 113
appeal to substrata as the literal bearers of properties, they insist, has the
consequence that we are precluded from ever making correct property-
attributions; for to succeed here, we would have to identify the subjects of
such attributions; but given the substratum theorist's characterization of
these entities, this would presuppose our picking out objects that are empiri-
cally inaccessible to us.
Empiricist critics of bare substrata have typically insisted that we can pro-
vide a perfectly coherent account of ordinary objects without an appeal to
entities distinct from the properties we associate with them. Frequently,
critics of substratum-ontologies have expressed this view by way of a meta-
phor, claiming that ordinary objects are nothing more than collections, bun-
dles, or clusters of properties. Nor, they have argued, does the phenomenon
of property-attribution prove problematic for this sort of account of the
ontological structure of substances. When, in ordinary parlance, we say that
a substance has a property, we are not implicitly referring to some constituent
of the substance that is categorically different from the property we attribute
to that substance. We are merely asserting the existence of a relation between
that property and the totality of properties making up the substance; we are
saying that the former belongs to or is an element in the latter.
Historically, the empiricist scepticism which underlies this sort of rejection
of a substratum ontology in favor of what I shall call a "bundle" or "cluster"
ontology has its roots in Locke's own account of underlying substrata. As my
second quote from Locke's Essay indicates, the unknowability he feels forced
to read into the concept of bare substratum did not rest altogether easy with
his own empiricist predilections; nonetheless, he took it to be a notion central
to our pre-philosophical thinking about physical or material substances; and
he seems to have endorsed the notion, contending that "we cannot conceive
how they (Le., the properties associated with physical objects) should subsist
alone nor one in another;,,7 and Locke insisted that the same problem arises
in the case of the mind:
The same thing happens concerning the operations of the mind, vis., thinking, reasoning,
fearing, etc., which we concluding not to subsist of themselves, nor apprehending how
they can belong to body or be produced by it, we are apt to think them the actions of
some other substance which we call spirit. 8
and they denied as well that the concept of an object of this sort should enter
into the philosopher's thinking about substances. Berkeley was willing to
press these two contentions for the case of physical substances at any rate.
Thus, he tells us that in pre philosophical thinking, physical objects are sim-
ply collections of properties; he construes these properties in psychological
terms as sense-ideas and says that "a certain color, taste, smell, figure, and
consistence having been observed to go together are accounted one distinct
thing, signified by the name 'apple'. Other collections of ideas constitute a
stone, a tree, a book, and the like sensible things;,,9 and Berkeley explicitly
denies that our pre-philosophical view of physical objects as collections of
properties requires any sort of modification at the hands of the philosopher.
It is notorious, however, that Berkeley is unwilling to do awaY.completely
with Locke's substratum; for he insists that while physical objects are mere
collections of ideas, those ideas themselves need a mental substratum, a "thing
entirely distinct from them wherein they exist ;,,10 and despite Berkeley's
relentless criticism of Locke's material substratum, his characterization of this
"thing" involves an appeal to the unknowability explicit in Locke's charac-
terization of substratum; for while insisting that "the objects of human
knowledge" are exclusively ideas, Berkeley denies that we have any idea of
mental substratum. 11
As is well known, Hume pointed out the inconsistency in Berkeley's ac-
count here, insisting that the appeal to mental substratum is no more legiti-
mate than the appeal to material substratum. If the ordinary man attends to
his mental life, Hume tells us, he finds "nothing but a bundle or collection of
distinct perceptions which succeed each other with unconceivable rapidity
and are as in a perpetual flux and movement.,,12 Our pre-philosophical no-
tions of both material and mental substance, then, conform to Berkeley's
characterization of physical objects:
The idea of a substance is nothing but a collection of simple ideas that are united by the
imagination and have a particular name assigned them by which we are able to recall,
either to ourselves or others, that collection;13
and Hume plainly thought the philosopher's attempt to supplement these pre-
philosophical notions by the introduction of bare substrata to be illegitimate.
Like Berkeley, Hume framed his account of substance in the psychological
idiom which interprets ordinary properties as ideas; but Hume's rejection of
bare substratum and his ensuing account of substances as "collections" of
properties have been endorsed by succeeding generations of empiricists, in-
cluding many who would be unwilling to accept his puzzling interpretation of
TWO THEORIES OF SUBSTANCE 115
properties. The Russell of Inquiry into Meaning and Truth, for example,
argues that since "the introduction of an unknowable should be avoided
wherever possible," the notion of an underlying subject should be rejected in
favor of the view that ordinary objects are "bundles of qualities." 14 Likewise,
A. J. Ayer has argued that the notion of "the 'unknown somewhat' in which
properties adhere" is a mere "metaphysical invention" and has insisted that
we should adopt the view that things are "only 'bundles of qualities' ".15
More recently, Herbert Hochberg has taken a similar line, rejecting Bergmann's
bare particulars on the grounds that their introduction violates the basic tenet
of empiricism, what he calls the Principle of Acquaintance, the view that
"any primitive descriptive term of 'the language of ontology' must refer to
something which is presented in experience:,,16
But despite its initial appeal, the bundle theory has been subjected to
severe criticism; and while critics have pointed to a variety of difficulties in
the bundle theory, most objections to this theory of substance take one of
four forms. Towards clarifying the issues at stake in the debate between sub-
stratum and bundle theorists, I want to outline these four forms of objection.
Objection I; The bundle theorist cannot account for the con-
tingency of substances.
The bundle theorist holds that ordinary objects are nothing more than the
properties we associate with them. It would seem, then, that he is committed
to the view that to say that a substance exists is simply to say that the pro-
perties associated with the substance' all exist. Properties are, however, neces-
sary beings, so that in identifying a substance with its associated properties,
the bundle theorist is committed to holding that ordinary objects are all
necessary beings. IS But, then, he must hold that every substance is both in-
generable and incorruptible; and, furthermore, he must hold that given any
list of properties sufficiently determinate to constitute a possible substance, it
is a necessary truth that there exists a substance satisfying the properties in
the list. And all of this is plainly false; ordinary objects are both generable
and corruptible; and there are descriptions of possible substances that are not,
in fact, satisfied.
Objection II; The bundle theorist cannot make sense of identity
through change.
It is a commonplace that substances undergo change. As a result of change,
they become qualitatively different: they come to possess properties they did
not possess or they cease to possess properties they previously did possess.
But even though they are qualitatively different as a result of change, sub-
stances can and frequently do persist or remain numerically identical through
change. Indeed, identity through change is a pre-philosophical truism. It is a
truism, however, that the bundle theory of substance cannot accomodate.
Change, we have seen, always involves an alteration in the properties asso-
ciated with a substance; but since he construes a substance as a collection
of properties, the bundle theorist must say that what emerges from a change
is invariably a new collection of properties and, consequently, a different
substance.
Objection III; The bundle theorist must construe all true state-
ments ascribing properties to substances as tautologous.
TWO THEORIES OF SUBSTANCE 117
tion of the substance in which it is present, this entity can neither be repeat-
able nor incorporate repeatable entities among its "ingredients"; it must, that
is, be independent of all properties; it must be bare substratum_
Now, while we have used a single label - "bare substratum" - for the
entities introduced in response to each of the four objections, the various
objections could be handled by an appeal to different entities; for clearly,
there is no a priori reason why there must be a single entity that functions as
the ground of contingency in a substance, the principle of permanence in a
substance, the literal possessor of the properties associated with a substance,
and the principle of individuation in a substance. Nonetheless, the description
of what is required of an entity if it is to play anyone of these roles suggests
that one and the same entity could play all three roles, so that given some
principle of theoretical simplicity, it is only natural to think that one and the
same entity does; and that entity, it is only plausible to think, is the underly-
ing subject of a substratum ontology.
These four objections, then, call into question our initial intuitions that
the bundle the or/is preferable to a substratum ontology. They point to fea-
tures central to our thinking about ordinary objects which appear to have a
plaUSible explanation within the context of a substratum ontology but seem
to be inconsistent with the basic tenets of a bundle theory of substance. In
the next chapter, I want to consider these objections in greater detail to
determine whether the apparent inconsistencies they point to are real.
NOTES
1 Of course, it is not the case that throughout this tradition the term used in contrast
with 'substance' is 'attribute'. Frequently, the expression employed here is 'accident',
where accidents are predicamental rather than predicable universals, so that to speak of
something as an accident is not to imply that it is exemplified only contingently.
2 1029a 10-25, translated by W. D. Ross in McKeon, Basic Works of Aristotle, p. 785.
There are interpretative difficulties with this passage; it might be doubted, for example,
whether Aristotle is presenting this account of predication as his own. For the grounds
for these doubts see Chapter Nine, where I discuss what I call substance-theories of
substance.
3 Essay Concerning Human Understanding II.xxiii.6. I am simply assuming the tradi-
tional interpretation of Locke's view here. For a quite different interpretation, see, e.g.,
Martha Bolton's 'Substances, Substrata, and Names of Substances in Locke's Essay',
Philosophical Review, 1976, pp. 488-513. In general, I am aware that my remarks on
the empiricists in this section tend toward the procrustean. The point, obviously, is not
to provide a novel interpretation of classical empiricist thinking on substance, but simply
to introduce the traditional dialectic on the issue.
120 CHAPTER SIX
4 Ibid., 2.
5 See, e.g., My Philosophical Development (New York: Simon and Schuster), 1959,
pp.160-162.
6 Realism (Madison: University of Wisconsin Press), 1967, p. 24.
7 Essay, II.xxiii.4.
8 Ibid.,5.
9 A Treatise Concerning the Principles of Human Knowledge, paragraph 1; p. 22 in
Principles, Dialogues, and Correspondence, ed. by C. M. Turbayne, (Indianapolis: Bobbs-
Merrill), 1965, p. 22.
10 Ibid., paragraph 2; page 23 in Turbayne.
11 Ibid., paragraph 1 (p. 22 in Turbayne) and paragraph 27 (p. 34 in Turbayne).
12 A Treatise of Human Nature, Book I, Part IV, section vi; p. 302 in A Treatise of
Human Nature, ed. by D. G. C. Macnabb (London: Collins), 1962.
13 Ibid., Part I, section vi; pp. 59-60 in Macnabb.
14 Ibquiry, p. 93.
IS 'The Identity of Indiscernibles' in Loux, Universals and Particulars, p. 267.
16 'On Being and Being Presented,' Philosophy of Science, 1965, p. 123.
17 'Ontology and Acquaintance,' Philosophical Studies, 1966, p. 53.
18 Here and in the following chapter, I use the term 'necessary being' in the stronger of
the two senses distinguished in Chapter Five.
CHAPTER SEVEN
The first objection, we have seen, insists that since he identifies substances
with the properties associated with them, the bundle theorist is committed to
the view that substances are necessary beings. The claim here is that since the
bundle theorist takes all of the constituents of ordinary objects to be neces-
sary beings, he is forced to construe the wholes which they comprise as
necessary. Now, when it is expressed in these terms, it becomes clear that the
first objection involves the fallacy of composition, the fallacy of assuming
that if each object entering into the constitution of a complex exhibits a cer-
tain property, the resulting complex must exhibit that property as well. But
while the argument underlying this objection may be fallacious, there clearly
are ways of parsing the "bundling" metaphor at work in traditional formula-
tions of the view which have just the consequences the first objection reads
into the bundle theory. If, for example, the bundle theorist tells us that an
ordinary object is a bundle of properties in the sense that it is to be identified
with the set of properties we associate with it, then it seems plausible to sup-
pose that his account forces us to construe substances as necessary existents.
It is unclear whether all sets are necessary beings; there are genuine grounds
for doubting whether a set at least one of whose members is a contingent
being is itself a necessary being. However, where a set is composed exclusively
of necessary beings, it is plaUSible to think that the set itself is necessary. But
if this is true, then the view that identifies an ordinary object with the set of
properties associated with it commits us to holding that ordinary objects are,
one and all, necessary beings. Ordinary objects, then, tum out to be both
ingenerable and incorruptible; and, furthermore, given any complete descrip-
tion of a possible substance in terms of its properties, the existence of a sub-
stance meeting that description would be a matter of necessity.l
Ukewise, if the bundle theorist tells us that ordinary objects are bundles
of properties in the sense that they are simply conjunctions of all the proper-
ties associated with them, he is forced to construe ordinary objects as neces-
sary beings. On this view, an ordinary object just is a property, an extremely
complex property formed from the conjunction of all the properties associated
121
122 CHAPTER SEVEN
with it. But clearly such a conjunctive property is no less necessary than the
properties that are its conjuncts. It is as ingenerable and incorruptible as each
of them is; and given any list of properties sufficiently determinate to gen-
erate the notion of a possible substance, it is necessarily true that there exists
some conjunctive property whose conjuncts are all and only the properties
entering into the list.
There are, then, ways of parsing the metaphor of "bundling" that have just
the consequences Objection I points to. They agree in that they explain the
metaphorical notion of a bundle of properties by reference to relations which
the properties supposedly constitutive of a substance necessarily bear to one
another. It is a necessary truth that the properties which, on the bundle
theory, constitute an ordinary object form a set or enter into a conjunctive
property. The difficulty, however, with the refutation of the bundle theory
proposed by Objection I is that no bundle theorists (at least none that I know
of) have explained the "bundling" metaphor in terms of relations of this sort.
Invariably, bundle theorists have explained the metaphor by reference to
some relation between properties which those properties only contingently
enter into. The later Russell, for example, appeals to the notion of compTe-
sence in parsing the "bundling" metaphor. As he explains it, compresence
is a relation which n-tuples of objects only contingently exemplify. Roughly
speaking, it is the relation of "occurring together;,,2 and as Russell charac-
terizes it, compresence is a relation which any two entities entering into the
constitution of one and the same substance bear to each other. But while he
says this, Russell refuses to provide a formal definition of compresence,
claiming that its fundamentality defies definition. It is a metaphysically
primitive relation which can, nonetheless, take on a variety of forms. In the
case of material objects, compresence takes the form of spatio-temporal coin-
cidence; whereas, in the case of mental substances, it appears as a psychologi-
cal relation logically independent of the concept of space. But regardless of
how it appears, compresence has an invariant consequence - that of making
a group of entities constituents of a single substance. 3
Where Russell speaks of compresence, D. C. Williams, another bundle
theorist, speaks of collocation, which he tells us is "the unique congress in the
same volume, which we call 'belonging to' (or inhering in, or characterizing)
the same thing.,,4 According to Williams, collocation is "the limiting value"
of a more general relation which he calls location; and location, he insists, is
"external" in the sense that a property "per se does not entail or determine
its location with respect to other" properties. s Herbert Hochberg, to mention
a third bundle theorist, parses the "bundling" metaphor in terms of the notion
THE BUNDLE THEOR Y 123
cally identical if for some time, t, a and b have different constituents at t. Put
formally, this point is expressed as follows:
(2) Necessarily, for any substance, a, and any substance, b, a is iden-
tical with b only if for any time, t, and any entity, c, c is a con-
stituent of a at t if and only if c is a constituent of b at t.
It should be obvious, however, that if we reject (1) in favor of (2), Objection
II collapses; for so far from entailing the impossibility of identity through
change on the bundle theory, (2) provides the bundle theorist with a frame-
work for coherently characterizing the phenomenon of change.
Given (2), then, it is difficult to see why there should be any need to
appeal to bare substrata to explain identity through change. Indeed, I can
think of only one way that bare substrata might be thought to play an ex-
planatory role here. If one held, first, that it is possible for a substance to
remain identical through a change in all of its properties and, second, that it
is impossible for a substance to remain numerically identical over a period of
time unless there is at least one entity which persists as a constituent of that
substance throughout the period of time in question, then one might appeal
to bare substrata as the guarantors of substantial identity through the radical
changes just described. Nonetheless, it should be obvious that no substance
could, in fact, remain identical through a radical change of this sort. A sub-
stance existing at one time can be identical with a substance existing at an-
other time only if there is some minimal continuity in the properties asso-
ciated with each. Consequently, the single context in which bare substrata
might play a genuinely explanatory role here is one that never confronts the
ontologist.
I have argued that since (2) rather than (1) provides the correct account of
the relationship between constituent-identity and numerical identity, the
bundle theorist can handle identity through change. Ontologists favoring (1)
might, however, object to (2) on the grounds that it robs the constituent-
whole relation of any real substance. They might argue that since substances
just are complexes of their constituents, the constituent-whole relation can-
not be a merely contingent relation that is variable over time. That relation,
they might insist, must be such that where a is a substance and b, one of its
constituents, it is necessarily true that if a exists, b belongs to a.
I think that they would be wrong in so insisting; nonetheless, I want to
conclude this section by arguing that even if he accepts this "necessitarian"
interpretation of the constituent-whole relation, the bundle theorist is not
committed to maintaining any proposition that experience controverts. The
126 CHAPTER SEVEN
bundle theorist could agree that identity through change is impossible but
argue that there is a relation weaker than that of strict identity such that it is
possible for a bundle of properties entering into a change to bear that relation
to the bundle of properties emerging from the change. He might dub this
relation "sameness", for example, and insist that it is this relation rather than
that of strict identity which is really at work in the pre-philosophical truism
that objects can persist through change. Needless to say, he would owe us an
account of the difference between sameness and identity; but assuming that
such an account is forthcoming, the pre-philosophical truism could be sal-
vaged even on the bundle theory.
Thus, even if one wants (as I do not) to reject (2) in favor of the tougher
conception of the constituent-whole relation at work in (1), he need not fear
that that conception forces the bundle theorist to reject the pre-philosophical
notion of persistence through change; but, then, not only does Objection II
fail to establish the inescapability of a substratum ontology; it does not even
refute the bundle theory.
sentence of the form 'a is P', it is impossible to know which substance is the
referent of 'a' without knowing that the property expressed by 'P' can be
truly ascribed to that substance; and that is to say that all true sentences of
the form 'a is P' come out tautological on the bundle theory.
But just how is the defender of bare substrata able to avoid this difficulty?
Well, one way of characterizing his response to the objection is to think of
him as proposing a distinction between two different senses in which an
entity can be said to possess a property. There is the "loose and popular"
sense of property-possession; and in this sense, substances can be said to
possess properties; but there is also a "strict and philosophical" sense of
property-possession; and in this sense, it is only bare substrata that can func-
tion as the possessors of properties. Now, if we interpret the substratum theo-
rist's view in this way, we must say that on his account the general form 'a is
P' splits into two quite different forms. In the one case, 'a. is a placeholder
for the proper names of substances; in the other, 'a' can be supplanted exclu-
sively by names of bare substrata and 'is P' points to the "strict and philo-
sophical" sense in which a property can be ascribed to an object.
Now, it should be clear that the assumption which appears to make Objec-
tion III so devastating for the bundle theorist's interpretation of subject-
predicate discourse has precisely the same consequence for the substratum
theorist's interpretation of sentences in which we pick out a substance by its
proper name and invoke the "loose and popular" sense of property-possession
in ascribing a property to it. That assumption tells us that we can know which
substance a proper name takes as its referent only if we know in advance all
the constituents of that substance; but as we have already noted, even the
substratum theorist wants to number the properties of a substance among its
constituents, so that, given the assumption in question, he too is committed
to construing every true subject-predicate sentence whose subject-term refers
to a substance as tautologous.
The substratum theorist, of course, will grant this point. What he wants to
insist is that in those contexts where we have the "strict and philosophical"
sense of property-possession, we can have subject-predicate sentences that are
both true and non-tautologous. He holds that by postulating a constituent
within substance which neither is a property nor is constitued by properties,
he can accept the assumption in question while denying that the ascription of
properties to objects is invariably tautological.
But does the appeal the "strict and philosophical" sense of property-pos-
session really provide the substratum theorist with the resources for escaping
Objection III? Well, if it preserves the non-tautologous nature of true subject-
128 CHAPTER SEVEN
which, taken together, are sufficient to pick out that object uniquely, there is
no a priori reason for supposing that the set of constituents underlying the
use of a proper name should be more than a proper sub-set of the entities
which constitute the bearer of that name. Quite consistently with (Q), then,
the bundle theorist could maintain that not all true sentences of the form 'a
is P' are tautologous. He could, in conformity with the assumption just noted,
agree that where a sentence of this form has as its predicate a term expressing
some property from the set of constituents required for identifying the sub-
stance which is the referent of the subject-term in question, the result is a
tautologous sentence; but he could hold that there are properties (those not
belonging to that set) which can be truly yet non-tautologously predicated
of the substance.
But is it true, as the proponent of Objection III insists, that (Q) commits
one to the view that none of the properties involved in fixing the reference
of a proper name can be truly yet non-tautologously predicated in subject-
predicate sentences whose subject-term is the proper name in question?
Well, one might argue that this is so only if it is true that for each proper
name, there is a single set of properties to which anyone capable of using
the name must appeal in fixing its reference. But the bundle theorist who
accepts (Q) could deny that this is the case. He could argue that for each
substance, there is a variety of such individuating constituent-sets. Anyone
of those sets, he could hold, would be sufficient for fIXing the reference of
the proper name of a substance; but, then, he could deny that a property's
belonging to one of the individuating constituent-sets for a particular proper
name entails that a predicate-term expressing that property can only be
predicated tautologously in a subject-predicate sentence whose subject-term
is the proper name in question. The point would be that while the property,
in conjunction with other properties, is sufficient for fixing the reference of a
proper name, the reference of that name could be fixed independently of the
appeal to that property - in terms of an individuating constituent-set not
including the property. Indeed, a bundle theorist accepting this version of
(Q) could maintain that only those properties which must enter into every
individuating constituent-set for a substance are predicated tautologously ,)f
that substance.
In any case, some bundle theorists have denied that (Q) is true. Thus,
the later Russell holds that as proper names function in human languages,
it is possible to apprehend a whole, to give that whole a name, and to use
that name successfully in referring to the whole without knowing any of
the entities that are its constituents.ll Russell's view that the use of proper
130 CHAPTER SEVEN
Necessarily, for any substance, a, and any substance, b, if, for any
entity, c, c is a constituent of a if and only if c is a constituent of
b, then a is identical with b. 13
Now, the bundle theorist wants to claim that it is necessarily true that the
constituents of substances are exhausted by their properties; but this claim, in
conjunction with the Principle of Constituent Identity, entails the Identity of
Indiscernibles, the claim that it is impossible for distinct substances to have
all and only the same properties. Expressed formally, the Identity of Indis-
cernibles goes as follows:
(II) Necessarily, for any substance, a, and any substance, b, if for any
property, P, P is an attribute of a if and only if P is an attribute of
b, then a is identical with b.
The proponent of Objection IV, however, insists that this principle is false;
and so concludes that the conjunction of the bundle theory of substance and
the Principle of Constituent Identity is false; he insists, however, that the
Principle of Constituent Identity is true and so concludes that objects are not
constituted exclusively by their properties.
Now, there are bundle theorists who can accept the Identity of Indiscerni-
bles with the greatest equanimity. I am thinking of bundle theorists who, like
Berkeley, Hume, and D. C. Williams, interpret properties in strictly nomi-
nalistic terms. 14 They maintain that it is impossible for different substances
(neither of which is a constituent of the other) to have literally one and the
same constituent. On their view, while different substances can have proper-
ties that are exactly similar, it is necessarily true that no two substances have
what is literally a Single property. But since they hold that no two substances
can literally share even a single property, bundle theorists of this persuasion
132 CHAPTER SEVEN
can hardly be expected to find the claim that different substances cannot
share all their properties problematic.
Effectively, these nominalistic bundle theorists want to claim that where
substances have what in ordinary parlance we call "the same property", they
do not incorporate a property that is literally identical; consequently, they
respond to Objection IV by claiming that if it is, in fact, possible that there
be distinct substances which, in ordinary parlance, agree in all their properties,
this would not falsify the Identity of Indiscernibles. We have seen, however,
that a nominalistic interpretation of property-agreement is unacceptable.
Where objects exhibit the phenomenon of property-agreement, they exem-
plify what is literally the same property. But, then, since a realistic inter-
pretation of the pre-philosophical fact of property-agreement is correct, the
possibility of distinct substances agreeing in all their properties would falsify
the Identity of Indiscernibles. The proponent of Objection IV, however,
argues that it is possible for distinct objects to agree in all their properties and
concludes that since the Principle of Constituent Identity is true, the bundle
theorist's interpretation of substances as bundles of compresent or consub-
stantial properties is false.
The difficulty with this attempted refutation of the bundle theory, how-
ever, is that (II) is arguably true. Thus, one might argue that given any sub-
stance, a, there is a property ex.emplified by a and by nothing other than a,
vis., the property of being identical with a; but, if this is correct, then it
trivially follows that associated with each substance there is a set of properties
such that it is impossible for any substance distinct from it to exemplify all
of the properties in the set. But even if one is suspicious of Leibnizian proper-
ties like those of being identical with Plato and being identical with Socrates,
one could still make a plausible case for the truth of (II) by appealing to pro-
perties which determine the spatio-temporal location of substances; for it
plausible to think that it is a necessary truth that one and only one material
body or person can wholly and completely occupy a given region of space at
a given time. IS But, then, in virtue of occupying a particular region of space
at a particular time, each substance would seem to have a set of properties
exemplified by no other substance; so that contrary to Objection IV, the
truth of (II) would seem to be non-problematic for the bundle theorist.
But is (II) as strong a principle as the bundle theorist is committed to? I
think not. To appreciate this, we need only reflect on the distinction between
what I shall call impure and pure properties. Impure properties are properties
(like being married to Henry VIII and being a student of Socrates) which
"incorporate" at least one determinate substance; whereas, pure properties
THE BUNDLE THEORY 133
are properties (like redness and wisdom) which do not. This distinction corre-
sponds to Strawson's distinction between "universals-cum-particulars" and OT-
dinary universals; and it is roughly what other philosophers have had in mind
when they have constrasted relational properties with non-relational proper-
ties. I have so far explained the distinction in metaphorical terms, speaking of
properties as "incorporating" or not "incorporating" substances. We can ex-
plain the distinction more precisely if we say that a property, P, is impure just
in case there is some relation, R, and some substance, s, such that necessarily,
for any object, x, x exemplifies P if and only if x enters into R with s and that
a property, P, is pure just in case P is not impure. We can see how this account
of the distinction operates if we consider the properties I have used as exam-
ples. Corresponding to the property, being married to Henry VIII, there is the
relation, being married to, and the substance, Henry VIII; and it is necessarily
true that an object exemplifies being married to Henry VIII if and only if she
bears that relation to that substance; consequently, being married to Henry
VIII is an impure property. Likewise, corresponding to the property, being a
student of Socrates, there is the relation, being a student of, and a substance,
Socrates, and since it is necessarily the case that an object exemplifies being a
student of Socrates if and only if he/she bears that relation to that substance,
being a student of Socrates is an impure property. But in the case of neither
wisdom nor redness is it possible to identify a relation R, and a substance, s,
such that necessarily, an object exemplifies either redness or wisdom if and
only if it bears R to s, so that redness and wisdom are both pure properties.
Invoking this distinction, we can say that the bundle theorist is committed to
the claim that no two substances can share all their pure properties.
The bundle theorist wants to maintain that substances are derived entities.
On his account, particular substances do not comprise the basic entities or
"simples" in our ontology; the basic items, he is saying, are properties. Sub-
stances are complexes "built out of" these primitives. But if this reductionistic
approach to substance is not to be a sham, the constituents out of which
substances are supposed to be constructed or built cannot already incorporate
the complexes they constitute. 16 Impure properties, however, all incorporate
determinate particulars and so they cannot be numbered among the "building
blocks" out of which particular substances are constituted. But, then, the
bundle theorist is not just committed to (II), but to the stronger principle
(II) Necessarily, for any substance, a, and any substance, b, if for any
pure property, P, P is an attribute of a if and only if P is an attri-
bute of b, then a is identical with b.
134 CHAPTER SEVEN
Now, bundle theorists have been sensitive to this version of Objection IV.
Indeed, in My Philosophical Development, the later Russell, a paradigmatic
bundle theorist, tells us that it was just this objection that had earlier con-
vinced him of the inadequacy of the bundle theory;lS and even a quick read-
ing of his early paper "On the Relations of Universals and Particulars" con-
firms the centrality of this objection in Russell's early theory of substance.
It is, he tells there, "obviously logically possible that they (sc. numerically
different substances occupying different regions of space-time) should have
no intrinsic differences whatsoever;" they can be "wholly indistinguishable as
to predicates;" consequently, "the bundle of coexisting qualities in the same
place is not an admissible substitute for the thing.,,19
In his later writings, however, Russell contends that this line of argument
fails as a refutation of the bundle theory. The contention there is that what I
have called Objection IV goes wrong in failing to distinguish between physical
space and physical time, on the one hand, and perceptual space and percep-
tual time, on the other. 2o The space and time of the physicist, he argues, are
THE BUNDLE THEORY 135
As regards the sensible world, it is clear on reflection that position in experienced space
is not relative, like the space of physics. In my momentary visual field, position is de-
rmed by qualities. What is in the centre of the field of vision has a quality that we may
call 'centrality'. Everything else that I am seeing at the moment has various degrees of
two qualities: up-and-down and right-and left. 22
I applied similar considerations to order in time. Suppose that some quality occurs twice
over in one man's experience, as, for example, when a clock is striking the hour. What is
it that makes you recognize two strokes as two, and not as one thing repeated? I came
to the conclusion that this recognition depends upon a quality which we may call 'sub-
jective pastness'. The contents of my mind, in so far as they are concerned with experi-
enced occurrences, can be arranged in a series beginning with sensation, going on to
akoluthic sensation, thence to immediate memory, and thence to memories having a
quality of more or less distance from present sensation. In this way, a subjective time
series is generated consisting of items which, from an objective point of view, are all
now. 23
NOTES
BARE SUBSTRATA
Objection IV, then, presents us with a powerful argument against the bundle
theory of substance. Since it is possible for different substances to be qualita-
tively indiscernible, the bundle theorist's attempt to construe substances as
wholes whose constituents are exclusively properties breaks down. But Objec-
tion IV is not meant merely as a refutation of the bundle theory; as we saw in
Chapter Six, it presents itself as well as an argument establishing the inescapa-
bility of a substratum ontology. It should now be clear how the objection can
play that role. The Principle of Constituent Identity tells us that indiscerni-
bility with respect to constituents entails numerical identity; but, then, since
it is possible for different substances to be indiscernible with respect to their
pure properties, each substance must incorporate a constituent over and
above its pure properties. Since, however, the aim is to specify the constitu-
ents out of which substances are composed, the additional entity cannot be
anything which presupposes the wholes that are substances; it cannot, then,
be an impure property. But since we have set the properties of a substance
(both pure and impure) on the one side and this additional entity on the
other, the constituent of a substance that guarantees its individuation must
be something such that its existence is independent of the properties with
which it is co-present; its being whatever it is cannot involve any of those pro-
perties. But this would seem to be just the characterization that holds of bare
substrata. Given the Principle of Constituent Identity, then, the possibility of
qualitatively indiscernible substances seems to force us to hold that in addi-
tion to their properties substances are constituted by pure individuators, bare
substrata.
Now, if we look at the work of contemporary defenders of substratum
ontologies, we find that the above line of argument accurately represents the
way in which bare substrata are introduced into the ontological arena. Thus,
we find Gustav Bergmann arguing as follows:
Let a and b be two spots of exactly the same shape and exactly the same color; tal & a2l
and [b 1 & b 2l the two classes (pairs) of qualities "in" them. Call the nexus again V. Assume
again that each spot has exactly three constituents. Then the assay of a yields al and a2
140
BARE SUBSTRATA 141
connected by V and nothing else; and that of b, bl and b2 connected by V and nothing
else. That presents us with the following alternative. (1) If we choose realism, the qualities
in the case are universals. Hence, the properties being exactly the same, al is literally the
same as b l ; a2 literally the same as b 2. Hence, by the fundamental principle, if each spot
had only the three constituents, a and b would be one and not two. Each spot, therefore,
must have at least one further constituent; and these further constituents must not be
literally the same. (2) If we choose nominalism, then al and bl> though perhaps exactly
alike, whatever that may mean, are not literally the same. Similarly, for a2 and b 2. In
strict logic, a nominalist is therefore ... not forced to search for further constituents. I
Thus, Bergmann attempts to establish the need for what I have called bare
substrata and what he calls bare particulars by appealing to a line of reasoning
essentially the same as my Objection IV. We have seen, however, that Objec-
142 CHAPTER EIGHT
tions I-III also attempt to specify ontological jobs which only bare substrata
can perform. Objecton I argues that we must appeal to bare substrata in ex-
plaining the contingency of ordinary objects. Bergmann, however, seems
unwilling to attribute this role to bare particulars. He tells us that there is a
nexus that ties together the various constituents of an object; and he insists
that this nexus is one into which those constituents might have failed to
enter. Presumably, then, it is this nexus rather than the bare particular which
enters into it that grounds the contingency of an ordinary object. 5
Objection II, on the other hand, points to the need for bare substrata as
the principles of continuity or permanence in changing substances. Bergmann,
however, denies that bare particulars are continuants. 6 He wants to deny that
each of the things I have called substances - familiar material bodies and
persons - incorporates a single bare particular. On the contrary, he argues
that bare particulars enter into the constitution of more primitive entities -
things like phenomenal spots and cubes. Ordinary material bodies and persons
are constructions out of these more primitive entities; and the primitive en-
tities out of which they are composed are, for Bergmann, momentary entities
that cannot persist through change. Any change in a momentary object like
one of the spots referred to above involves the "corruption" of that entity;
its bare particular is replaced by another, and we have a new object.'
But while Bergmann is unwilling to construe his bare particulars as princi-
ples of contingency and permanence, he does see them as playing the role
assigned them by Objection III. Bare particulars are the literal exemplifiers
or possessors of the properties with which they are co-present. Thus, it is the
bare particular that is "in" a phenomenal spot which is the literal possessor of
the properties which, together with it, constitute that spot. As Bergmann puts
it, bare particulars are the referents of the subject-terms and the properties
co-present with them, the referents of the predicate-terms of true subject-
predicate sentences. Bergmann, then, wants to attribute two roles to bare
particulars: they are the guarantors of the numerical diversity of objects as
well as the literal exemplifiers of the properties we associate with those ob-
jects. Now, it should be clear that these are two distinct roles. 8 Sometimes
Bergmann fails to keep this fact in mind. Consider, for example, the following
passage:
I look at two cubes. Assume that they are alike in all non-relational respects. For in-
stance, they are both (the same shade ot) green. Assume, furthermore, that while looking
at these two cubes, I make two judgments, 'This is green' and 'That is green'. In one way
these judgments differ; in another, they agree. In the two statements, the difference is
expressed by the occurrence of different words, 'this' and 'that', in the subject-place; the
BARE SUBSTRATA 143
agreement, by the occurrence of the same word 'green' in the predicate-place. For the
judgments to be grounded, both the difference and the agreement must be grounded in
the two states of affairs that are judged. How, then, are they grounded? One possible
answer has four parts. (1) Each of the two situations contains two constituents. (2)
There are altogether three constituents named by 'this', 'that', and 'green' respectively.
The constituents named by 'this' and 'that' are called particulars; the one named by
'green' a universal. (3) The universal, which is the constituent common to both situations
accounts for my judging both cubes to be green. (4) The two particulars' being different
accounts for my knowing the cubes from each other. 9
But whatever we may think of the subsequent claim that bare particulars are
the literal exemplifiers of the properties with which they are co-present,
reflection on the phenomenon of indiscernible yet diverse objects clearly
makes Bergmann's appeal to bare substrata appear attractive. What we must
determine, then, is whether the persistent criticisms of substratum ontologies
have any real force. As we have seen, the most relentless critics of bare sub-
strata have been philosophers of an empiricist termperament. Now, Bergmann
and his followers all claim to subscribe to the basic tenets of empiricist epis-
temology and, accordingly, are anxious to show that the appeal to bare sub-
strata does not violate any restrictions imposed on the ontologist by the
144 CHAPTER EIGHT
First, there is the use of 'know' in which to know something is to be acquainted with it.
Second, there is the use in which to know something means to be able to recognize it. IO
Now, Allaire concedes that in the second sense of 'to know', bare particulars
are unknowable. Since bare particulars do not differ intrinsically, but only
numerically, it is impossible, when presented with a bare particular, to deter-
mine whether it is a bare particular encountered on some earlier occasion.
But towards showing us that bare particulars are knowable in the first sense,
Allaire asks us to reflect on two discs that are qualitatively indiscernible:
Consider once more the two discs. When presented together, they are presented as
numerically different That difference is presented as is their sameness with respect to
shape, (shade 00 color, and so on. What accounts for that difference are the numerically
different individuals. No character nor group of characters can do that. Thus, to say that
they are individuals is to say that things may be merely numerically different. No matter
what description one proposes, the numerical difference of two things which are alike in
all (nonrelational) respects must be accounted for ... To claim that both discs are collec-
tions of literally the same universals does not account for the thimess and thatness which
are implicitly referred to in speaking of them as two collections. That is, the two collec-
tions of characters - if one persists in speaking that way - are, as presented, numerically
different. Clearly, therefore, something other than a character must also be presented.
That something is what proponents of the realistic analysis call a bare particular. I I
The suggestion here, then, is that in being acquainted with the numerically
different, yet indiscernible objects, we are eo ipso acquainted with the en-
tities which guarantee their numerical diversity. But, then, in the first sense of
'to know', bare particulars are knowable. It turns out, however, that this is
just the sense of 'to know' that is at work in the empiricist injunction against
introducing unknowables; for according to Allaire, the empiricist approach to
ontology is expressed in what he (like Herbert Hochberg) calls the Principle
of Acquaintance (PA). He formulates this prinCiple as follows:
The PA states that the undefinable terms of any "ontological" description must refer to
entities with which one is directly acquainted. 12
Sellars' contention here is that the claim that bare particulars are the posses-
sors of properties is inconsistent. Why? Well, because to claim that something
is a bare particular is to claim that it has no properties; but, then, to go on
and say that bare particulars are the literal possessors of the properties asso-
ciated with the objects into whose constitution they enter is to say that
things which have no properties at all do, in fact, have properties.
Now, what is so striking about this criticism of the Bergmannian account
of predication is its disarming simplicity. Indeed, it is difficult to understand
how, if the objection is sound, the bare particular theorist could have failed
to see its force. The fact is, however, that the bare particular theorist is not
committed to the contradiction Sellars reads into his view. The bare particular
theorist does not explain the notion of a bare particular by saying that it is a
thing which has no properties; what he says, on the contrary, is that bare
particulars or bare substrata are things which in themselves have no proper-
ties. But just what does this mean? Well, Bergmann tells us that his particulars
are bare in the sense that they have no natures. Presumably, what Bergmann
BARE SUBSTRATA 147
means to deny here is not that bare particulars have or possess properties but
rather that they have any properties essentially. Now, the idea that objects
have some of their properties essentially or necessarily is one that we shall
examine later. For the present, we can say that an object has a property
essentially just in case it has that property and it is impossible that it both
exist and fail to have that property.16 But, then, in saying that bare particu-
lars are the literal possessors of properties, what Bergmann and the whole line
of substratum ontolOgists have been saying is not that the possessors of pro-
perties are not the possessors of properties, but only that the literal possessors
of properties possess none of their properties essentially. They are claiming,
that is, that the things which possess properties are such that there is no pro-
perty they possess that they might not have failed to possess; and whether
this claim is true, it clearly does not involve the contradiction Sellars points
to. 17
But while I think Hat he is wrong in claiming that the substratum theorist
both asserts and denies that bare particulars possess properties, I am inclined
to think that Sellars is right in claiming that the notion of a bare particular is
implicitly contradictory. The defender of substrata tells us that an entity is
bare just in case it exemplifies no properties essentially. But it seems unlikely
that it should be a merely contingent fact about an entity that it is bare. If a
thing is bare, it would seem, then it is necessarily bare. But, then, every entity
that is bare has at least one property essentially, the property of having no
properties essentially; and this is to say that the notion of a bare particular is
inconsistent. To qualify as bare, an object can have no properties essentially;
but the property of having no properties essentially is itself a property that is
essential to anything that has it; consequently, it is impossible that there be
any bare entities, whether particulars or not.
Now, the defender of bare substrata will doubtless deny that there is any
property of the sort I suggest - the property of having no properties essen-
tially. I am not sure how he could make this point convincing; but even ifhe
could, his characterization of bare substrata would still implicitly involve a
reference to properties which substrata exemplify essentially. To appreciate
this, we need only reflect on Bergmann's claim that it is impossible for a bare
particular to be "in" more than one object at a given time. Surely this claim
entails that every bare particular has at least one property essentially, vis., the
property of being the constituent of at most one object at a given time. No
148 CHAPTER EIGHT
bare particular could lack this property, so that once again the contention
that there are bare substrata turns out to involve the contradictory claim that
what has no essential properties has at least one property essentially.
The defender of the Bergmannian account might retrench here and claim
that what he calls bare particulars have essentially none of the properties
which, in ordinary parlance, we associate with the ordinary objects into
whose constitution they enter. Even this weaker claim, however, is incorrect;
for we associate with each ordinary object all of the transcendental proper-
ties, i.e., all of those properties exemplified by every object in every possible
world; but if there are any entities of the kind that have been called bare
particulars, they surely exemplify all of these properties essentially. It is im-
possible that there be anything which fails to possess the properties of being
self-identical, being human or non-human, and being colored if green. Nor is
it merely transcendental properties that are problematic here. If an ordinary
object is green, for example, then we associate with that object the property
of being green or a bare particular; but this disjunctive property is one that
would have to be exemplified essentially by the bare particular entering into
its constitution. Likewise, if some ordinary object is human, then there is the
property of being human or incapable of entering into the constitution of
more than one object at a given time; and that property is one that is asso-
ciated with the ordinary object in quesiton and would as well have to be
exemplified essentially by its bare particular. 18
Now, if we think back to the beginning of our discussion of the problem
of substance we will recall that my introduction of the notion of substratum
was grounded in an assumption about the relationship between an object and
its properties. The assumption (found in both Aristotle and Locke) was that
the possessor of a property, P, is always something such that whatever it is, its
being that does not presuppose its possessing P. Effectively that assumption
is just the one underlying Bergmann's account of predication, the assumption
that the literal possessors of properties exemplify no properties essentially.
What we must now conclude is that this assumption is false; but while our
original discussion of the view that substances incorporate a constituent over
and above their properties was grounded exclUSively in that false assumption,
we have since discovered that the view can be justified by the appeal to other
considerations - those bearing on the possibility of indiscernible yet diverse
substances. What all of this suggests is that we reject one of the themes tradi-
tionally associated with substratum ontologies, the view that the constituent
of a substance which is distinct from the properties associated with it is bare;
if there is any such constituent, it has innumerable properties essentially.
BARE SUBSTRATA 149
My modified substrata are simply individuators; they are the entities which
ground the possibility of diverse, yet indiscernible objects. Now, when asked
whether they succeed in performing their assigned task as individuators, one
is inclined to ask how they could fail; for what I have salvaged (indeed, what
seems worthy of salvaging) from the traditional substratum view is simply the
claim that substances incorporate individuating entities. Thus, all that remains
of the original substrata is their individuating role, so that the question "Do
they individuate?" has to answered affirmatively. They just are individuators;
obviously they individuate!
Indeed, one is likely to think that they perform their task as individuators
a little too well. To see how one might have suspicious here, one oUght to
consider the following case. Suppose that there is some phenomenon cp; now,
150 CHAPTER EIGHT
so resolve once and for all the problem of indiscernible yet diverse objects.
Unfortunately, however, individuators cannot be bare, and so they do not
enable us to revolve the problematic stemming from Objection IV. Despite
initial appearances to the contrary, then, substratum ontologies are no more
capable of providing a solution to the problem of individuation than are
bundle theories of substance, so that the only argument we have to justify
the appeal to a substratum ontology, even of the modified sort suggested
above, collapses.
NOTES
For the past three chapters, we have been examining two accounts of the
ontological structure of substances. According to one, substances are wholes
whose constituents are simply properties; whereas, the other account takes
substances to incorporate an additional constituent, a substratum. Now, for
epistemological reasons, we were initially inclined to accept the bundle theo-
rist's account of substance; but we found the substratum theorist pointing to
four supposed flaws in the bundle theory, each of which, he claimed, could
be overcome only by the appeal to a substratum ontology. The first difficulty
was that since substances are contingent beings, they cannot be constituted
exclusively by necessary beings. According to the substratum theorist, the
contingency of ordinary objects is accounted for only if we admit among
their constituents entities that are themselves contingent - substrata. The
second difficulty was that the bundle theorist cannot account for identity
through change; the possibility of identity through change, we were told,
presupposes that each substance incorporates among its constituents a princi-
ple of permanence - again, a substratum. The third difficulty was that the
bundle theorist is forced to construe all true sentences in which we attribute
properties to substances as tautological; to explain the non-tautological truth
of such sentences, the substratum theorist insisted that we take substances to
incoporate a constituent which, while capable of serving as the possessor of
properties, is ontologically independent of the properties which can be predi-
cated of it; and once again the additional constituent was identified with
substratum. Finally, the substratum theorist argued that the bundle theorist
cannot account for the possibility of diverse yet qualitatively indiscernible
substances and contended that it is necessary to postulate as "ingredients" in
substances carriers of numerical diversity or pure individuators; and, as in the
case of the first three difficulties, these additional "ingredients" were said to
be substrata.
Now, we have seen how the first three difficulties are only apparent. A
bundle theorist can explain the contingency of substances without appeal-
ing to a constituent that is itself contingent by holding that the properties
153
154 CHAPTER NINE
but since (2) and (3) are true, (1) is false. Objects are not constituted exclu-
sively by their properties; they must include among their "ingredients" purely
individuating constituents; and only bare particulars will do the job here; for
it is only if the individuating constituents of substances are in themselves un-
characterized that they provide a final answer to the question of individua-
tion. Unfotunately, the notion of a bare particular is contradictory.
The choice, then, appears to be between a false principle - (II*) - and an
inconsistent notion - that of bare substrata or bare particulars. We could, of
course, escape this dilemma if we could show that the falsity of (I) does not
entail an ontology of bare substrata or that (2) is false or, finally, that (3) is
false. Philosophers have tried to escape our dilemma in each of these three
ways. In a couple of places, the later Russell, for example, tells us that it is
extremely improbable that two substances agree in all their pure properties;
and he seems to think that the empirical improbability of diverse yet qualita-
tively indiscernible objects is sufficient for the truth of the bundle theory?
Presumably, then, Russell thinks that the bundle theorist is committed not to
(11*) but to the weaker principle
(11**) For any substance, a, and any substance, b, if for any pure pro-
perty, P, P is an attribute of a if and only if P is an attribute of b,
then a is identical with b.
It need not be impossible, then, that there be no numerically different yet
qualitatively indiscernible substances; it is sufficient for this to be a merely
contingent fact. Obviously, Russell can substitute (11**) for (II*) only if he
rejects one of (I )-(3); and I am inclined to think that he wants to reject (I);
but in rejecting (1), he does not mean to be endorsing a substratum ontology;
he merely means to be denying that the bundle theory is to be construed as a
necessary truth; he is saying that it is contingently true that objects are con-
stituted exclusively by their properties.
I am not convinced, however, that a proposal of the sort I read into
Russell's remarks is a satisfactory way out of our dilemma. For one thing, the
bundle theorist who takes this line will have to confront the persistent claims
of Bergmann and his followers that the numerical diversity of qualitatively
indiscernible objects is not a merely abstract possibility, but a phenomenolog-
ical fact. Now, it may be that defenders of bare substrata are wrong in their
descriptions of things like phenomenal spots and cubes; but it is hard to see
how a bundle theorist of Russell's persuasion could show this in a non-ques-
tion-begging way. But even if he could, his proposal to treat the bundle theory
as a merely contingent truth only puts off the evil day when he must confront
TOWARDS A SUBSTANCE-THEORY IS7
the dilemma of individuation; for while it may be true that no two objects in
our world are qualitatively indiscernible, this remains a possibility. To use the
jargon introduced in Chapter Five, there are possible worlds where diverse
substances agree in all their pure properties; and the bundle theorist has to
provide us with an account of the ontolOgical structure of substances in those
worlds; but, then, in describing those worlds, he runs up against the very
dilemma he seeks to escape in characterizing the actual world. In those
worlds, substances are diverse yet indiscernible in their properties and so
cannot be characterized in bundle theoretic terms; the only way of explaining
their structure is by appealing to bare individuators; but, as we have seen,
there are no possible worlds whose substances incorporate bare individuators
among their constituents.
A rather different response to our dilemma is found in the writings of
Herbert Hochberg, who rejects (2), the Principle of Constituent Identity. 3 It
is not just that Hochberg thinks that it is possible for different objects to
agree in all their constituents; he seems to agree with Bergmann and his
followers that we actually confront diverse yet indiscernible objects; but since
he denies that identity of constituents entails numerical identity, he fmds this
fact non-problematic for the bundle theorist. Unfortunately, Hochberg pre-
sents no arguments for the falsity of the Principle of Constituent Identity.
Indeed, he seems to reject the principle merely because of its lethal conse-
quences for the bundle theory. Nor is it surprising that he provides no reasons
for rejecting the principle that are independent of its consequences in the
dialectic we are considering. The principle is clearly a necessary truth, one
that defines for the ontologist the very notion of the constituent-whole rela-
tion. Here, Bergmann is right; for the philosopher who takes the analysis of
substance to be the task of specifying the constituents of certain wholes or
complexes, (2) is a fundamental principle. To reject it is to pay mere lip ser-
vice to the basic contention of both the bundle and substratum theories, vis.,
that substances are complexes whose ontological structure is revealed by the
identification of those entities that enter into their constitution.
A third way out of our dilemma would involve the rejection of (3). One
could avoid both an ontology of bare individuators and (II*) by denying that
the entities involved in the constitution of substances must one and all be
"pure" entities. Without meaning to embrace any version of the bundle
theory, Nicholas Wolterstorff suggests this response to the problem of indivi-
duation. 4 He rejects Bergmann's contention that it is possible for numerically
different objects to agree in all their non-relational properties, claiming that,
given any pair of numerically different substances, a and b, a has a property
158 CHAPTER NINE
which b does not have - the property of being identical with a; and b has a
property that a does not have - the property of being identical with b. Earlier
we ran up against identity-properties of this sort, but we argued that since the
bundle theorist is engaged in the reductive enterprise of constructing sub-
stances from their constituents, such properties could not enter into his
"assay" of substances. In insisting that identity-properties enter into our
description of apparently indiscernible objects, Wolterstorff seems to be pro-
posing that we reject the reductionism implicit in both the bundle and sub-
stratum theorists' accounts of substance. He seems to be denying that we can
"get below" the concept of individual substances in presenting an ontological
account of their structure.
Now, while I am sympathetic with Wolterstorffs anti-reductionism here, I
am inclined to think that his outright rejection of the reductionistic frame-
work is a bit too abrupt. Neither the bundle theorist nor the defender of bare
substrata is likely to be convinced by Wolterstorffs mere stipulation that
impure properties must lie at the basis of our analysis of substance. A more
effective strategy in dealing with their theories of substance is to adopt the
reductionistic framework they take for granted and to use that framework
against itself. At any rate, that is the strategy I shall employ. I shall simply
accept the restriction bundle and substratum theorists impose upon the
ontologist - that of analyzing substances in terms of only "pure" entities;
and I shall try to show how this restriction ultimately calls into question the
reductionistic framework they both employ.
Constituent Identity in deriving (11*) from the bundle theory; and, similarly,
it is only if we accept this assumption that we are, given the falsity of (11*),
warranted in invoking the Principle of Constituent Identity in grounding a
substratum ontology.
Now, the universals that have been at the center of the stage for the past
three chapters have all been properties; for both the bundle theorist and the
substratum thorist tend to identify the pure universals of a substance with its
pure properties. Given this tendency, it is easy enough to see why the assump-
tion has never come into question; for in the case of the properties associated
with a substance, the assumption seems correct. To accept a realistic account
of attributes, it seems, is just to commit oneself to the view that the instantia-
tions of a given property are all numerically identical, that the red of this ball
is identical with the red of that ball, that they have the same shape, and so
on. But if it should Wrn out that not all of the pure universals associated with
substances are like properties in having numerically identical instantiations,
then perhaps we would have a way out of our dilemma. More, specifically,
what we need to isolate if we are to escape between the Scylla of bare parti-
culars and the Charybdis of (11*) is a set ot universals that meets four require-
ments. The set must be such that (A) the various instantiations of any univer-
sal from the set are all numerically different entities; (B) it must be plausible
to think that every substance exemplifies at least one universal from the set;
(C) none of the universals from the set are impure; and (0) none of the uni-
versals in the set are wholly and completely reducible to or analyzable into
universals which exhibit numerical identity in their instantiations.
Is there a set of universals which meets these requirements? If we can trust
some remarks in the Categories, we can conclude that Aristotle at least
thought so. s Indeed, Aristotle seems to have held that even the properties
associated with ordinary objects have numerically distinct instantiations. He
seems to hold that while all white objects exemplify some one entity, white-
ness, their jointly exemplifying that entity does not involve their incorporat-
ing numerically identical constituents. According to Aristotle, although the
property, whiteness, is equally exhibited by white objects, a, b, ... n, the in-
stantiations of whiteness in a, b, ... , n (Le., the whiteness of a, the whiteness
of b, ... , the whiteness of n) are all numerically different entities. As he con-
strues it, appending the expression 'of a', 'of b', etc. to the name of a property
does not yield further designations of that property, but designations of
numerically different particulars - the instantiations of the property. The
suggestion in the Categories is that this sort of account holds for all properties.
The consequence of this suggestion for our problem should be clear: since no
160 CHAPTER NINE
service of phenomenalism (the view that our concepts of material objects are
analyzable in terms of merely sensory properties) that philosophers have tried
to effect this reduction; and their attempts have, one and all, ended in dis-
aster. Invariably, it turned out that they could reduce substance-concepts to
the concepts of properties only if they illicitly smuggled in vestiges of the
kind-concepts they were trying to eliminate. Their failures, of course, do not
entail the impossibility of their task; but they do suggest that the burden of
proof lies not with the philosopher who takes substance-kinds to be irreduc-
ble to properties, but with his opponents. Given the lack of success attending
their enterprise, his view emerges as the plausible view. He must, of course, be
prepared to handle all future attempts to carry out the enterprise; but that is
quite different from proving in advance their impossibility.
To deny, however, that kinds can be eliminated in favor of properties is
not to deny the important connection between being a member of a kind and
exhibiting certain properties. That such a connection exists and that it is
more than merely contingent are both claims no one can doubt. Their indubi-
tability likely lies at the bottom of many attempts to reduce substance-kinds
to properties. But while granting the relevant connection, one can deny that
predicating a kind of an object is merely ascribing a set of properties to it.
This is what I am denying; and denying it, I am arguing, is plausible.
Now, if we accept this plausibility claim, we can conclude that substance-
kinds do meet the requirements set out in (A)-(O). Substance-kinds, then,
provide us with a resolution of our dilemma. If we hold that every ordinary
object belongs to a substance-kind, we need neither adopt (11*) nor accept
the view that objects incorporate bare substrata. Objects can exhibit precisely
the same universals without incorporating some additional substratum just
because each object exhibits at least one universal, a substance-kind, whose
instantiations are all numerically different.
The view I want to defend, then, is that each substance falls under a substance-
kind and that since substance-kinds are not reducible to universals of any
other type, each substance exemplifies a universal which guarantees its nu-
merical diversity from every other substance. Kinds are universals whose
instantiations are numerically different; but the instantiations of a substance-
kind just are the various substances which belong to or fall under it. Thus,
there is no need either to deny what is obvious - that it is possible for dif-
ferent objects to be indiscernible with respect to their pure universals or to
164 CHAPTER NINE
against reductive ontologies, one must not, in his zeal to avoid reductionism,
lose sight of the main consideratons that speak in favor of my analysis of
substance. The main justification I have provided for my account is that it
enables us to resolve a problem that is irresolvable on both the bundle and
substratum accounts - the problem of individuation. Since substance-kinds
of and by themselves diversify their members, by accepting the framework I
propose here, we can provide a coherent account of substance without endors-
ing either (II*) or a substratum ontology. (II*) is false not because substances
incorporate bare substrata among their constituents but simply because they
are substances, members of substance-kinds.
IV. ESSENTIALISM
our conventionalist critic would have us believe), I find this fact irrelevant to
the status of substance-kinds themselves; and it is their status and not the
status of our concepts of those kinds that is crucial to my proposed solution
to the problem of individuation.
A more serious criticism stems from my account's commitment to essen-
tialism. Indirectly we have discussed essentialism already; it is the view that
objects exhibit or exemplify some of their properties contingently or acciden-
tally and others, essentially or necessarily. Now, obviously substances exem-
plify some of their attributes only contingently; a red ball might have had
some other color, and a wise man might have failed to achieve the wisdom he
possesses. Substances do not, however, belong to their proper substance-kinds
contingently; they are essentially or necessarily members of the substance-
kinds they exemplify. Dogs could not have been substances of other kinds;
and Similarly for cats, gerania, and human beings. My claim that substances
are irreducibly kinded, then, commits me to the view that at least in the case
of substances, it is possible to distinguish between attributes essential to their
exemplifiers and those exemplified only contingently.
Just how my commitment to essentialism might be thought problematic is
as follows. Critics of essentialism argue that the distinction between essential
and accidental attributes is obscure ~d needs to be clarified. Typically, essen-
tialists have held that one can explain the distinction between necessary and
non-necessary attributes (the so-called de re modalities) in terms of the more
familiar notion of de dicto modality, the notion of modality as applied to
propositions. One might think that an explanation of this sort is straight-
forward, that one can simply say that an object exemplifies an attribute
essentially just in case the proposition that it exemplifies that attribute is
necessarily true and that an object exemplifies an attribute contingently just
in case the proposition that it exemplifies that attribute is true, but not neces-
sarily true; but although some essentialists have suggested such an account,lO
a moment's reflection shows it to be inadequate. A proposition can be neces-
sarily true only if it is impossible that it be false. Furthermore, the proposition
that an object exemplifies an attribute comes out false in the case where the
object does not exist; but, then, the explanation of the de re modalities just
proposed has the consequence that no contingent being (no existing being
whose non-existence is possible) can exemplify even one attribute essentially
or necessarily. But obviously the objects that are of concern to us - material
bodies and persons - are all contingent beings, so that the account just pro-
posed can hardly help us in clarifying what is involved in a substance's belong-
ing to its proper substance-kind.
168 CHAPTER NINE
Central to my account is the idea that substances are irreducible unities, that
they are not complexes reducible to more basic entities. But there is an
obvious difficulty here; for the things I have so far taken to be paradigmatic
substances - artifacts, plants, animals, and human beings - all have parts; and
this suggests that what I have called substances are mere complexes after all,
not complexes of metaphysical entities like properties or substrata, but com-
plexes of things the non-philosopher takes to be their parts. This difficulty
poses itself on two fronts; for there are the things that the layman takes to be
the parts of substances, their "commonsense" parts, and there are the things
the physicist tells us are constituents of every ordinary object - the various
micro-particles of contemporary physical theory. But whether one focuses on
the things the layman calls the parts of substances or on the micro-particles
of scientific theory, the result would appear to be the same: ordinary objects
fail to be the unities required by a substance-theory of substance; they are
rather complexes of more basic things.
This difficulty is on Aristotle's mind in the central books of the Meta-
TOWARDS A SUBSTANCE-THEORY 171
separated from the substances whose parts they are, they come to be actual
substances, they are no longer the kinds of things they were when they were
parts of ordinary objects. This sort of account works well enough in dispelling
the idea that ordinary objects are mere complexes of what the layman calls
their parts; but it is unclear how it can be extended to handle the things the
physicist calls the parts of substances; for the micro-objects he points to -
things like molecules, atoms, and quanta - are objects of kinds whose nature
can be apprehended independently of any reference to the wholes they con-
stitute. Indeed, the very point of positing such objects is to provide an ex-
planation of how it is that the stuffs they are supposed to constitute have the
natures they do. How can a substance-theorist accomodate the scientist's
account of ordinary macro-objects as complexes of micro-objects?
Well, one strategy here is simply to deny that the claims of the scientist
have any ontological force. While granting that the scientist's account plays
an important role in enabling him to move from a set of observational pre-
mises to a set of observational conclusions, one can insist that his micro-
objects are mere instrumental fictions. One can deny, that is, that the physic-
ist's talk of micro-objects forces us to hold that ordinary objects are literally
composed of things like molecules, atoms, and quanta; one can hold that the
scientist's reference to these "entities" is a mere piece of calculational ma-
chinery invoked to expedite scientific predictions, retrodictions, and the like.
The sort of instrumentalism that underlies this response is an approach
with which. I have some sympathy; but it is one that few other philosophers
nowadays take seriously. Most philosophers who confront the framework of
scientific explanation want to interpret the theoretical entities of science
realistically and to hold that to have good reasons for accepting a scientific
theory is eo ipso to have good reasons for thinking that the micro-objects it
postulates actually exist. But even the substance-theorist who accepts this
realistic interpretation of scientific theories has the resources for defending
his account of substance. Indeed, there are at least two different ways in
which he might accomodate the idea that ordinary objects are complexes of
micro-objects within a substance-theory of substance.
First, he can invoke a modified version of Aristotle's account of the parts
of living substances. While agreeing that ordinary objects are literally com-
posed of the micro-objects postulated by contemporary physics, he can hold
that, qua constituting ordinary objects, these micro-objects are only virtual
substances. He can insist that it is only when they exist in separation from the
wholes they constitute that they are truly self-subsistent entities. Then, he
can go on and argue that while ordinary objects are made up of micro-objects,
TOWARDS A SUBSTANCE-THEORY 173
kind, that of substance. There is, he held, no kind more general than sub-
stance to which individual substances belong. Substance, then, is a summum
genus or highest level kind. On Aristotle's view, terms like 'object' and 'entity'
which are more general than 'substance' do not express kinds. 17 Indeed,
Aristotle wants to claim that these terms (the so<alled transcendental or uni-
versally applicable terms) are not even univocally applied to objects that are
and objects that are not substances. 18
I shall not develop Aristotle's views on the impossibility of transcendental
genera; but I want to point to a misunderstanding that the phenomenon of
kind-subordination is likely to engender. Confronted with the fact that every
infima species or lowest level substance-kind is subordinated to more general
kinds, one might think that substance-kinds like mankind, geranium, and
amoeba are merely derivative universals. One might think that to be a gera-
nium, for example, is simply to be a plant with certain properties and that to
be a plant, in turn, is simply to be a substance with certain properties. The
Aristotelian picture of definition only reinforces the temptation of think of
infimae species in these reductive terms; for Aristotle tells us that to define a
substance-kind, K, one points to the kind, K', to which it is immediately
subordinated and specifies, along with K', some property that marks the
members of K off from those members of K' that do not belong to K. The
suggestion to which this account gives rise is that it is only when we have
reached the kind substance that we have a substance-kind that stands on its
own two feet; it and it alone is a basic or non-derivative substance-kind; all
other substance-kinds are really dispensible in an account of what there is.
Now, Aristotle is aware that his account of definition suggests this picture
of the derivative status of lower level substance-kinds, and he is anxious to
counter the picture; for it presents us once again with a threat to the supposed
unity of individual substances. Substances, on this picture, turn out to be
complexes of the very abstract universal substance and a series of differentiat-
ing properties. In countering this picture, Aristotle employs what I take to be
a sound strategy. He tells us that while definition plays an important epis-
temological role in providing us w;ch an analytic tool for understanding sub-
stance-kinds, it is easy to misuse that tool; for while it may be true that from
the epistemological perspective of definition generic kinds are prior to infimae
species, ontologically it is infimae species that are basic or fundamental.
The possibility of defining lowest level kinds by reference to generic kinds
suggests that the ontolOgical structure of substances involves the exemplifica-
tion, first, of a very general kind and, second, of a series of ever more specific
differentiating properties. But, according to Aristotle, this sort of account has
TOWARDS A SUBSTANCE-THEORY 175
things backwards; it takes higher level kinds as basic and construes infimae
species as universals formed from generic universals by the supplementation
of differentiating properties. For Aristotle, however, objects are first and
foremost members of a variety of infimae species and because they agree in
various ways, they can be classified according to higher level kinds. Aristotle
wants to claim, then, that we go wrong if we suppose that it is in virtue of
belonging to generic kinds that objects fall under specific kinds. On the con-
trary, he holds that it is because objects from different infimae species agree
in various ways that they fall under the generic kinds to which we speak of
their infimae species as subordinated.
Aristotle makes this point by saying that no genus exists apart from its
species;19 and what he means here is that nothing is just a member of a genus.
To be a member of a genus is to be a member of one of the infimae species
subordinated to it. Thus, there is no such thing as an animal simpliciter; to be
an animal is to be a cat, a dog, a human being, or .... On Aristotle's view,
then, ordinary objects do not fall under their infunae species because they fall
under the genefa to which they are subordinated; it is, on the contrary, only
because they fall under infimae species that they are members of the relevant
generic kinds. Thus, although in definition, we analyze infimae species in
terms of the higher level kinds to which they are immediately subordinated,
infimae species are ontologically prior to their genera; generic kinds are an
abstraction from the ontologically more basic lowest level kinds.
Now, this account of the ontolOgical primacy of infimae species among sub-
stance-kinds raises the question of whether lowest level substance-kinds are
the ontologically most basic attributes that substances exemplify essentially.
We have said that to be an animal is to be a cat, a dog, or a human being and
that it is impossible to exemplify the kind animal without exemplifying one
of these more specific kinds; but, then, can we stop with the claim that
infimae species are the least general attributes which substances exemplify
essentially? Is it not plausible to think that there are even less general attri-
butes exemplified essentially by substances which limit or determine lowest
level substance-kinds in just the way that those kinds limit or determine
generic substance-kinds? Traditionally, Aristotelians have denied that there
are any such attributes; they have held that the attributes exemplified essen-
tially by a given substance are exemplified essentially by all the other sub-
stances which belong to the same lowest level substance-kind. On this view,
176 CHAPTER NINE
essences are kind-invariant; and the various features in terms of which objects
of the same infima species differ are exemplified only contingently. But
opposed to this Aristotelian notion of general essences is the view (held most
prominently by Leibniz)20 that each substance has an individual essence, an
attribute which is exemplified essentially by that substance and which no
other object could possibly exemplify. The paradigms of individual essences
on this view are the various identity-attributes (e.g., being identical with
Socrates and being identical with Plato) that we have already encountered on
several occasions. On this view, each substance has its own identity attribute;
it could not exist and fail to exemplify that attribute and, fmally, it is impos-
sible for an entity distinct from that substance to exemplify that attribute
either essentially or contingently. Frequently, defenders of individual essences
invoke the framework of possible worlds here, telling us that each existing
object exemplifies its individual essence in the actual world and in every pos-
sible world in which it exists and that no object distinct from it exemplifies
that attribute in any possible world.
Now, it might be thOUght that this view (I shall call it Leibnizian essen-
tialism) is incompatible with the approach to substance that I have sketched
out in this chapter. In fact, this is at most only partially true. Even if it
turns out that each substance exemplifies an individual essence of the sort
I have described, it remains the case that substances fall under kinds, so
that one can still appeal to substance-kinds in escaping the dilemma of
individuation. If there are such essences, one can, of course, also circumvent
the dilemma by invoking them (in the way that Wolterstorff does); but as
we have seen, the outright appeal to attributes like the property of being
Socrates and the property of being Plato represents such a radical departure
from the reductionistic framework underlying the dilemma that it is not
likely to prove very convincing to philosophers who, like the traditional
bundle and substratum theorist, simply take that framework for granted.
Thus, even if the defender of Leibnizian essentialism is right, my strategy
for dealing with the dilemma of individuation remains the most effective
way of pointing out the inadequacies of the views leading to (I1*) and bare
substrata. But while the truth of Leibnizian essentialism would not call
into question my response to the dilemma of individuation, it would seem
to point to an incompleteness in my version of the substance-theory of
substance; for if each substance exemplifies an individual essence, then my
claim that lowest level substance-kinds furnish the ontologist with concepts
of full-fledged substances would seem to be wrong. If the Leibnizian view
is correct, then the ontologist would seem to be able to provide a complete
TOWARDS A SUBSTANCE-THEORY 177
of" a substance "sees at the same time the basis and the reason of all the
predicates that can be truly uttered regarding,,21 that substance.
I shall not bother to evaluate this very strong claim; for even if individual
essences have the richly individuating conceptual content Leibniz claimed for
them, the fact that that content is hidden from us has the consequence that
reflection on individual essences will not contribute much to the ontologist's
understanding of substance. But, then, while the Aristotelian is wrong in
denying ontological status to individual essences, the fact that we cannot non-
circularly identify just what it is that they add to the content of the infimae
species they limit or determine shows that for all practical purposes the
Aristotelian framework is rich enough for understanding the ontological
structure of substances. Infimae species may not be the least general attri-
butes that substances exemplify essentially; but from the perspective of the
finite, human ontologist, this is a difference that makes no difference. 22
NOTES
1 Pretty clearly, this extension of the distinction between what is pure and impure
requires a definition slightly different from that outlined in Chapter Seven. The defIni-
tion, I suggest, goes as follows: an entity, a, is impure just in case there is some relation,
R, and some substance, s, such that necessarily, for any object, x, a is a constituent of x
if and only if x enters into R with s; whereas, an entity, a, is pure just in case a is not
impure.
2 See, e.g., Hurruzn Knowledge, p. 295. In the light of Russell's very elaborate attempt
(discussed at the end of Chapter Seven) to show the truth of (II*), this is a shocking
claim. The reader of Russell's writings, however, soon learns to expect such inconsis-
tencies in Russell's work.
3
'Universals, Particulars, and Predication,' pp. 89-9l.
4 'Bergmann's Constituent Ontology,' Noux, 1970, pp. 109-134.
5 'This (not very novel) interpretation of Aristotle's view is suggested by his definition
of "present in" at la 23-24. For an alternative interpretation, see G. E. L. Owen's'In-
herence,' Phronesis, 1965, pp. 97-109.
6 See, e.g., Metaphysics Z.l (l028a 34-1028b 1).
7 In fact, I shall not discuss Aristotle's interpretation of universals outside the category
of substance. Once we recognize the individuating role of substance-kinds, the question
of whether or not properties and relations are identical in their instatiations becomes, I
am inclined to think, a decision-question. I am inclined to take a non-Aristotelian analy-
sis of properties; but I have no knock-down argument against the Aristotelian view which
interprets properties as kinds whose members are their individual instantiations. My
only reasons for accepting the approach I do are first, that the Aristotelian account
seems to me to place us on the "slippery slope" back to nominalism: and second, that I
find it redundant to recognize both properties construed as universals and their indivi-
dual instantiations when it is possible to recognize only properties construed as universals.
TOWARDS A SUBSTANCE-THEORY 179
universals whose disjuncts are K I'S, K2'S, ... , Kn-I'S. For a more detailed explanation
of each of these conditions, and for the rationale behind them, see 'The Concept of a
Kind,' Philosophical Studies, 1976, pp. 53-61.
17 See, e.g., Metaphysics B.3 (99Sb 21-2S).
18 See, e.g., Metaphysics r.2 (l003a 33-1003b 11) and Z.1 (l02Sa lO-30).
19 See, eg., Metaphysics Z.14 for a discussion of this thesis. (See also Metaphysics I.S,
especially lO5Sa I-S).
20 And in our own day by Alvin Plantinga. See, e.g., The Nature of Necessity, Chap-
ter V.
21 Discourse on Metaphysics (La Salle, illinois: Opencourt), 1945, p. 14.
22 In this respect, I think that the "individual" essences of attributes differ from those
of substances. The former have, whereas the latter do not, conceptual content which is
accessible to us. It is this difference which makes it possible to provide a non-trivial
account of the identity-conditions of universals by saying, as I did in Chapter V, that a
universal, U, is identical with a universal, U', just in case for every possible world, W,
every attribute exemplified by U in W is exemplified by U' in Wand vice versa.
EPILOGUE
183
184 INDEX OF NAMES
185
186 INDEX OF SUBJECTS
Editors:
WILFRID SELLARS, Univ. of Pittsburgh and KEITH LEHRER, Univ. of Arizona