You are on page 1of 20

The Persistence of Counterexample: Re-examining the Debate Over Leibniz Law

Author(s): Gregory Landini and Thomas R. Foster


Reviewed work(s):
Source: Noûs, Vol. 25, No. 1 (Mar., 1991), pp. 43-61
Published by: Blackwell Publishing
Stable URL: http://www.jstor.org/stable/2216092 .
Accessed: 02/06/2012 20:04

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Blackwell Publishing is collaborating with JSTOR to digitize, preserve and extend access to Noûs.

http://www.jstor.org
The Persistenceof Counterexample:
Re-examiningthe Debate Over Leibniz Law
GREGORY LANDINI
UNIVERSITY
OF IOWA
THOMAS R. FOSTER
BALLSTATEUNIVERSITY

1. INTRODUCTION

Are there good reasons to reject the principle of the Identity of In-
discernibles as a necessary truth? Refutations seem to abound, the
most common of which appeal to considerations of symmetry. But
just as numerous are defenses of the principle. Of course, any argu-
ment that the identity of indiscernibles is (or fails to be) necessarily
true must be examined against an assumed background theory of
universals.' A change in the theory may well result in a different
assessment of whether indiscernibility with respect to those univer-
sals is sufficient for identity. Opponents in the debate have expressed
much concern over what univerals there are-one admitting only
a subset of the other's universals. This paper contends that a more
fundamental issue concerns what the opponents take a "universal"
to be. We distinguish two non-overlapping theories of universals,
Logical Realism and Natural Realism, and show that much of the
debate conflates these frameworks.

2. LOGICAL REALISM
Logical Realism holds that properties and relations subsist in a way
that is not only independent of language, but also of the capacity
that humans have for thought and representation. Universals trans-
cend the world and its objects, subsisting independently of whether
there is a world at all. Their subsistence, however, provides the
semantic ground for the application of predicate expressions. In this
way, Logical Realism provides a direct account of predication in
both language and thought in the form of the universals it posits.
NOUS 25 (1991) 43-61
? 1991 by Nou's Publications 43
44 NOUS

Couched in a standard second-order predicate logic,2 the com-


prehension axiom for Logical Realism is as follows:
(CP) (21P) (VZ 1) . . . (VZ J (F(z1 * Zn)-v5)
where Fn is not free in 4.
Here any open well-formed-formula is held to correspond to a univer-
sal, including those containing free variables, constants, bound
predicate variables, as well as those that are logically true or logically
false of any individual whatsoever.
In Logical Realism the Principle of the Identity of Indiscernibles
(hereafter "the PI") is expressed as:
(Vx)(Vy)((VF)(Fx Fy) - x = y).

There is no need for an equivalence between "Fx" and "Fy", for


by (CP) the complement of any predicate which stands for a universal
will also stand for a universal. There appears to be a simple proof
of the P11 in Logical Realism. Suppose (VF) (Fx -Fy). By (CP)
we have (AG) (Vz) (Gz z = x). Hence, since (x = x) we know
that Gx; and by Universal Instantiation of 'F' we have (Gx - Gy).
By Modus Ponens we get Gy, from which we get (y = x) by the
comprehension axiom for the property G.
There have been a number of traditional objections to this proof.
For instance, it is claimed that the proof defines identity in terms of
itself. That is, the proof allows the property 'being identical with
x' to be in the range of the predicate quantifier, and this is held
to be circular. (Hereafter, we use Church's lambda-notation to
generate abstract singular terms for universals.) Now it is true that
the identity symbol need not be taken as a logical primitive once it
is granted that the PII is logically true. But the alleged "circular-
ity" here can be generated only by assuming that "impredicative
universals" are inadmissible. (A universal is said to be "im-
predicative" if its definition involves reference to a totality of univer-
sals to which it belongs.) In the framework of Logical Realism, the
notion of a predicative or impredicative universal is entirely without
substance. Since universals are independent of thought and language
altogether, the positing of universals via (CP) is not to define or
construct them in any way.
Another objection is that the P11 should have more that a merely
formal content. Thus, it is held that a property such as 'fXz z =x]'
is purely formal and its use in the proof trivializes the principle.
But the universals of Logical Realism posited as the ontological
ground for the semantics of predication; they are intended to have
a purely formal or logical status insofar as they are independent
of the world and its objects. Statements that are true solely in virtue
LEIBNIZ LAW 45

of such universals are "true in all logically possible worlds"-i.e.,


they are logically true. This does not trivialize them. The P11 is
no more "trivial" as a logical truth than is any theorem derivable
from the laws of logic.
The rejection of purely logical universals sometimes surfaces in
the claim that the property '[Xz z =x]' is not "genuine". Univer-
sals must be able to be exemplified by more than one particular.
For example, Armstrong writes that "the mark of a universal is
the logical possibility that the class of particulars which have it, be
an infinite class" (Armstrong [1], p. 93). Again, such a criterion
is inappropriate in the framework of Logical Realism. Armstrong
does not have logically real universals in mind here at all. For even
allowing his criterion, it is a simple matter to find a logically real
universal which both meets the criterion and yet is coextensive with
the property '[Xz z = x]'. Consider the property '[Xz (z = x) v p]',
where p is some contingently false proposition. Since p is false the
two properties are coextensive. But p is contingent, so it is logically
possible that '[Xz (z = x) v p]' is shared by infinitely many particulars.
Armstrong might reply that this alleged "property" is unnatural
or merely linguistic and thus inadmissible. But such a reply would
only illustrate the point that the real source of the criticism, as with
all the objections so far considered, lies in a shift to a theory of
universals distinct from Logical Realism.
There is, however, a serious criticism based on considerations
within the theory of Logical Realism. For example, in arguing against
the necessity of the P11, Max Black maintains that we could have
two distinct spheres which are a distance apart and which, because
they are in a symmetric universe, would share all non-relational
properties (Black [3], p. 207). Attempts to show there is a contradic-
tion by appealing to "relational" properties such as '[Xz z =x]' or
'[Xz R(z,a)]' are disallowed, Black claims that the symmetry of the
world prevents not only forming a definite description which would
single out one of the spheres, but also establishing a connection
between one of the spheres and a proper name.
Formulated in this way, the objection confuses univerals with
the linguistic expressions used to comprehend them. The universals
of Logical Realism are independent of thought and representation
and thus linguistic or epistemic considerations are irrelevant to the
metaphysical issue of what universals there are. But a version of
Black's argument, hereafter called "the argument from dependence"
can (at least prime facie) be formulated so as to avoid confusing
epistemic and metaphysical issues. According to Logical Realism,
universals subsist independently of the world and its contingent ob-
jects. But certain linguistic expressions for "relational properties"
46 NOUS

involve relations to contingent particulars. This can happen in two


ways. The expression might involve a relation term all of whose relata
are bound variables. This occurs in the expression "[Xz (3x)R(z,x)]",
and it also occurs in fully descriptive expressions such as
"[Xz (5Ix) (Fx & (Vy) (Fy - x = y)] & R(z,x))". On the other hand,
the expression might involve both a relation term and a logically
proper name (or a variable free in every subwff), such as in
" [Xz R(z,a)] ", or " [Xz R(z,x)] ". The former two cases are unprob-
lematic. But the latter cases are held to conflict with the independence
of logically real universals. If there were relational properties cor-
responding to the latter expressions, then there would be universals
whose subsistence is dependent on the existence of contingent par-
ticulars. Hence, there must be no universals corresponding to ex-
pressions involving free variables (unbound in every subwff) or proper
names.
The Logical Realist would reply, however, that the argument
now confuses subsistence with exemplification. The comprehension
axiom of Logical Realism does not specify what universals subsist,
it specifies conditions under which a given universal is exemplified.
Universals subsist independently, but some univerals are such that
theyare exemplifiedjust in case a given relation to a unique contingent
particular is also exemplified. The argument from dependence must
show that the mere subsistenceof certain universals and not simply
their exemplification, involves the existence of some contingent par-
ticular. The occurrence of a proper name or free variable in the
linguistic expression for the universal does not show this.
Perhaps the most frequent attempt to show dependence is to
appeal to the intuition that a universal corresponding to a complex
linguistic expression is itself ontologically complex. An open for-
mula containing a proper name or a free variable (unbound in every
subwff) would, therefore, express a universal which itself "contains"
constituents which are ontic counterparts of the proper name or
variable. The argument is wanting. A given universal may have
a complex expression involving connectives, quantifiers, and the
like, and yet in another language it may have a simple expression.
Complexity of expression seems to reflect nothing of the ontological
nature of the universal itself. What expressive devices are used in
language to refer to universals is a purely epistemic and conceptual
matter according to Logical Realism. Universals do not have "con-
stituents"; they are not "complex" or "simple" in a sense that
corresponds to the complex or simple linguistic expressions for them.
Providing a non-linguistic formulation of the "dependence" of
a universal is, therefore, not easily accomplished. Nonetheless, one
might concede the case against ontic counterparts for quantifier ex-
LEIBNIZ LAW 47

pressions and logical connectives and make a special case for proper
names and free variables. However arbitrary some expressions for
universals are, the occurrence of proper names (or free variables)
admittedly appearsto be exceptional. But even if a relational property
did "contain" a contingent object as a constituent, it is not clear
that this would make its subsistence contingent. The objection must
add special assumptions about how the contingent existence of a
particular is to be interpreted in modal logic. In this regard, the
use of proper names raises the largely unsettled debate concerning
contingent existence, "trans-world identity", and the semantic in-
terpretation of quantified modal logic. For example, an advocate
of the dependence argument might claim that there is no trans-
world identity and hence, maintain that contingent existence is to
be construed as the contingency of a description being satisfied.
Modal Logical Realism is by no means committed to this view.3
In fact, the behavior of proper names in modal contexts seems to
challenge the description theory.
To discuss the many issues involved here would take us too far
afield. Our intent is simply to reveal what issues are substantive
within Logical Realism and which import intuitions ontologically
alien to that framework. We therefore grant that the dependence
argument can be articulated within the framework, but conclude
that without appeal to a particular philosophical interpretation of
modal logic, it does not provide a telling objection to the use of
free variables or proper names in the positing of a universal. This
allows the P11 to stand as logically true in the framework of Logical
Realism.

3. INDISCERNIBILITY IN PHENOMENALISM
The dependence argument is often associated with a criticism of rela-
tional properties based on the phenomenalistic view that the P11
must provide a solution to the so-called "Problem of Individua-
tion". Suppose that a is R to b and that all objects in the relation
R are distinct. It is held that the relational property '[Xz R(z,b)]'
cannot "account" for the numerical distinctness of a and b, since
it holds of a only insofar as a is distinct from b. If the Identity of
Indiscernibles is to give an "account" of distinctness, then rela-
tional properties are inadmissible.
The trouble with this argument is that "giving an account of
distinctness" must be distinguished from providing a principle which
assures distinctness. To see this, consider a theory of classes which
allows self-membership and adopts Extensionality. We may have
two classes, A and B, where A differs from B only in that A is
a member of A and A is not a member of B. By Extensionality,
48 NOUS

the classes are clearly different-they have different members. But


does Extensionality give us an "account" of the difference? Clearly
not! So if there is a legitimate issue over what is to "account" for
distinctness, it is separable from the issue of whether a given prin-
ciple assures distinctness. From within the framework of Logical
Realism there is no demand that indiscernibility be able to give
an "account" of numerical distinctness.4
Matters are different in Phenomenalism. While there are many
variants of Phenomenalism, traditionally it involves two main theses,
one conceptual and the other ontological. The conceptual thesis main-
tains that our knowledge of concrete particulars (or better, "contin-
uants", since sense-experiences are themselves momentary concrete
particulars) is constructed or built up from the "simple ideas" of
immediate experience. The meaning of statements about continuants
is, therefore, to be analyzed in terms of actual and possible ex-
periences. The ontological thesis accepts the conceptual thesis and
goes on to maintain that continuants themselves are "bundles" of
these actual and possible simple ideas (or their properties).
On both theses, the objects of direct sense-experience involve
only "simple" sensory properties such as 'blue', 'red', 'sweet',
'smooth', 'hard', 'round', and the like. No principle of individua-
tion is required to identify or re-establish the presence of such prop-
erties. Continuants, on the other hand, cannot be recognized without
employing some such principle. The phenomenalist holds that we
are only "mediately" acquainted with them. Our knowledge of con-
tinuants is formed by inference from the coherent structural rela-
tionships of immediate experience. (On a traditional account, the
"coherence" or "structure" of these relationships is grounded upon
the constant conjunction of the sense-experiences). Complex prop-
erties, in turn, are held to be constructs formed on a level higher
than (or at least equal with) that of continuants; they are (as it were)
abstracted from propositions about continuants.
On this view, the individuation of continuants is fused with the
issue of what "accounts" for their distinctness. This is because con-
tinuants are conceptual constructs out of an already determinate
class of sense-experiences involving only simple sensory properties.
Thus, the individuation of continuants must be based on indiscer-
nibility with respect to the simple properties involved in immediate
sense-experience. In short, the framework of Phenomenalism serves
to ground the view that expressions involving the identity symbol,
constants, free variables and quantifiers (for continuants), and prop-
ositional connectives such as 'and', 'or', 'if-then', do not stand for
simple properties. Appealing to relational properties to individuate
is inappropriate, since as complexes they depend upon the logically
LEIBNIZ LAW 49

prior construction of the very continuants they are called upon to'
individuate.
Phenomenalism, in this sense, cannot succeed. It fails for the
same reason that the reductive dogma of Empiricism fails (Quine
[14], p. 38ff). The meaning of a statement about a continuant can-
not be analyzed in terms of sense-experiences singly, but only col-
lectively in a body of statements about continuants. Therefore,
reference to continuants must always occur in the analysans. This
carries over to the so-called "bundle" thesis as well. After all, the
bundle view simply assumes that the conceptual thesis is successful,
and collects together in a bundle the actual and possible sense-
experiences (or the properties of these sense-experiences) referred
to in the analysans. Relational properties involving continuants must,
therefore, be included in the bundle. Yet according to
Phenomenalism, these properties are inadmissible; continuants would
be elevated to a status on a par with the immediate objects of
experience.
One might object that there are bundle views which are not
tied so closely with the conceptual thesis. But this can be so only
by abandoning the basic tenet of traditional Phenomenalism which
analyzes our knowledge of concrete particulars in terms of the "struc-
ture", "constant conjunction" or, as Russell once called it, the
"compresence", of the simple ideas of immediate sense-experience.
This is an improtant point which is lost in discussions of the bundle
view. It is the conceptual thesis of Phenomenalism that commits
the bundle view to the PIIP, i.e., a principle which says that shar-
ing all phenomenally simple properties is logically sufficient for iden-
tity.5 The logical sufficiency here is assured by the phenomenalist
construction of continuants. Black's two-sphere universe fails to be
a counterexample to the PI.P* He assumes that there are two con-
tinuants in a symmetric world, and challenges that the continuants
may yet share all non-relational properties. If the counterexample
is not to be question begging, it must be formulable in accordance
with a phenomenalistic analysis. The identity of continuants will
be analyzed in terms of actual and possible sense-experience. To
assume that there can be two continuants in a sense that goes beyond
this is to simply assume outright that Phenomenalism is false.
The tie between the bundle thesis and Phenomenalism's concep-
tual analysis of statements about continuants shows that the PIIP
is immune to Black's argument. It is, rather, the failure of the reduc-
tive dogma of Empiricism that refutes the traditional phenomenalistic
construction and its bundle view. It should not be concluded,
however, that since sharing all simple sensory properties is not suf-
ficient for identity, the "strong" form of the Identity of Indiscer-
50 NOUS

nibles fails in Logical Realism. The very distinction of a "strong"


form (where only simple sensory properties are in the range of the
predicate quantifiers) and a "weak" form (where complex proper-
ties are also allowed) is ontologically alien to Logical Realism. It
is just a confusion to import the phenomenalistic distinction between
simple and complex properties into Logical Realism.

4. "ATTRIBUTE REALISM"
The failure to distinguish the properties and relations of Logical
Realism and the constructivism of Phenomenalism may well have
its origin in a framework we shall call "Attribute Realism". Like
Logical Realism, Attribute Realism construes universals as logically
real and independent of the causal structure of the world. Unlike
Logical Realism, it places restrictions on what well-formed-formulas
shall count as standing for its logical universals.
The Attribute Realist owes a philosophical account which justifies
placing restrictions on the comprehension axiom while at the same
time holding that attributes are logically real. One historic source
is the framework of Logical Atomism, espoused by Wittgenstein
and later by Russell. The fundamental assumption of the framework
is that every atomic proposition is logically independent of every
other. The n-adic predicate expressions of the atomic propositions
must therefore respect the independence thesis. Hence, the framework
maintains that monadic wffs containing the identity symbol, proper
names or free variables, and logical connectives or quantifiers do
not stand for universals. If, for example, there were universals such
as '[Xz R(z,b)]' and '[Xz R(a,z)]', then the proposition '[Xz R(z,b)](a)'
would be atomic and logically equivalent with '[Xz R(a,z)](b)'. Ac-
cording to Wittgenstein, the logical connectives are a part of the
"syntax" which provides the "scaffolding" for the possibility of
empirically meaningful sentences (i.e., those that have conditions
of truth and falsehood). A logical "truth" is not a truth "in" every
possible world; it is the structure which establishes the "possibil-
ity" of a possible world. In later years, particularly due to Russell's
version of Atomism, the elementary sentences were taken as sentences
about sense-data, and the properties inhering in them were con-
strued as those that are phenomenally immediate.
There are variants of Attribute Realism. Some drop certain of
the philosophical assumptions of Atomism and, accordingly, soften
the restrictions on the comprehension axiom. Invariably the result
is a framework which takes some version of the phenomenalist's
"phenomenal property paradigm" together with features of Logical
Realism. The logically real properties and relations are thus pared
LEIBNIZ LAW 51

down to just those regarded as "primitive" in Phenomenalism. In


Attribute Realism, the Identity of Indiscernibles can be expressed
as follows:
PIPB (Vx) (Vy) ((V BF) (Fx - Fy) - x = y).
Here the superscript "B" indicates that the quantifier ranges over
just those properties and relations an Attribute Realist would allow.
Though this may vary, our intention is to parallel Black's restric-
tions. In particular, names and free variables (unbound in every
subwff) are excluded.
Unfortunately, much of the discussion of the Identity of Indiscer-
nibles is carried on without reference to the point that the background
philosophical framework is a variant of Attribute Realism. In Attri-
bute Realism, as in Logical Atomism, indiscernibility is not logically
sufficient for identity. Indeed, Black's symmetric two-sphere world
is a counterexample to PIIB. But we must be careful to recognize
that though this result holds for the constrained notion of "logical
necessity" involved in Logical Atomism (and in Attribute Realism),
it does not follow that it holds in Logical Realism itself. The universal
quantifier in the PIIB ranges over only a subclass of the logically
real univerals in the range of the universal quantifier of the PII.
The notions of "logical truth" (necessity) in each framework must
not be confused.
The logical possibility of a world in which two distinct objects
share all attributesdoes not show that the PII fails to be logically
true. Attribute Realism blurs the distinction between Attributes and
the universals of Logical Realism. It can do this in another way
as well. The universals of Attribute Realism are commonly viewed
in two ways. Universals ground a Realist theory predication-a
theory, incidentally, which need not assume a Phenomenalistic con-
struction continuants. At the same time, universals are regarded
as the fundamental building blocks of the natural world. This leads
many to confuse the universals of Attribute Realism with the material
properties and relations investigated in science.6
Consider, for instance, an otherwise insightful discussion by
Ronald Hoy. Hoy argues that indiscernibility is sufficient for iden-
tity and criticizes the often cited objections to invoking 'relational
properties' to show the logical necessity of the Identity of Indiscer-
nibles. Many of the so-called "non-relational" properties, he says,
are revealed by science to be relational after all.
Science goes beyond the monadicity of the predicates and says that
color, weight and hardness are complex dispositionalpropertieswhose
analyses require mentioning relations with other things (Hoy [121,
p. 286).
52 NOUS

Thus, he concludes that often the properties in imagined counterex-


amples of the Identity of Indiscernibles turn out to be complex and
"this shows how the principle can, (and should), be saved" ([12],
p. 288). But Hoy's point concerning the scientific investigation of
color, weight, and the like, applies only in a framework where univer-
sals are regarded as maternal properties and relations uncovered by
scientific inquiry. In the progress of science the "manifest image"
eventually gives way to the "real image"; nominal characteriza-
tions of material dispositional properties such as 'blue' and 'hard-
ness', are replaced by real characterizations and these may indeed
involve material relations. But scientific discovery has no bearing
at all on the view that attributes such as 'blue', 'hardness', and
'weight' are "simple" because phenomenally immediate. Indeed,
the ontological nature of the universals of Attribute Realism,
Phenomenalism, and Logical Realism are not the subject of scien-
tific analysis or empirical inquiry.
The conflation of the logically real universals of Attribute
Realism, and the material properties and relations investigated in
science creates the false impression that the possible worlds in At-
tribute Realism are physically possible worlds-i.e., worlds that are
possible relative to the laws of nature. But Attributes are not material
properties and relations. Material properties and relations are not
merely a subclass of the universals of Logical Realism, they are
different in kind. The philosophical framework of Natural Realism,
where universals are regarded as structures causally realizable within
the laws of nature, is fundamentally different from both Logical
Realism and Attribute Realism.

5. NATURAL REALISM

The universals of Natural Realism, unlike those of Logical Realism


and Attribute Realism, do not transcend the causal structure of the
world. Neither, however are they dependent in the phenomenalist
sense. Natural Realism holds that universals are physical (or
'material") properties and relations; they are causally realizable
physical structures and relations which are dependent upon the ac-
tual laws of nature. The dependence does not require that such struc-
tures be realized in the actual history of the world, but only that
they be causally realizable-i.e., realized in some world which is
possible relative to the laws of nature.
To be sure, some of the linguistic expressions for the univerals
of Attribute Realism are also held to express natural properties and
relations in Natural Realism. For example, the predicates 'being
red' and 'being blue' stand for logically real attributes according
LEIBNIZ LAW 53

to Attribute Realism. It is also a well confirmed empirical hypothesis


of Natural Realism that an object can have a material dispositional
property or capacity to absorb certain wave-lengths of light and reflect
others-a capacity which we call 'being red'; and similarly there
is a capacity which we call 'being blue'. Moreover, in Attribute
Realism there is no logically real attribute 'being red or blue'. Neither
has any empirical hypothesis of a single material dispositional capacity
for being red or blue ever been confirmed.
The similarity of the two frameworks is superficial, however.
The Attribute Realist rules out 'being red or blue' by placing restric-
tions on the wffs that stand for logical universals. (In particular,
disjunction signs are not allowed). The Natural Realist imposes no
such restrictions. Complex English predicates may well stand for
material properties. Whether there is a material property or rela-
tion is always an empirical question. Many complex wffs, such as
those defining "mass", and "gravitational attraction", do corres-
pond to material properties and relations. Natural Realism does
not validate (as a logical truth) any comprehension axiom regard-
ing the positing of naturally real universals. Any such posit amounts
to a scientific hypothesis which is always open to falsification
(Cocchiarella [8], p. 113). Thus, unlike frameworks which account
for predication in thought and language by positing a universal for
every open wff, Natural Realism does not attempt such an account
of predication in the form of the universals it posits.
Since the framework of Natural Realism will only recognize
universals as causally realizable material structures dependent on
natural laws, it would be inappropriate to add a Logical Realist
theory of predication. However, a Conceptualist theory of predica-
tion, according to which "concepts" are understood to be disposi-
tional capacities for thought formation, is an appropriate ally
(Cocchiarella [8], p. 114). Concepts are causally dependent upon
the natural capacities cognitive creatures have for thought and
representation-capacities which are realized by the material prop-
erties and relations that are components of the causal structure of
the world. For example, conceptual capacities might be explained
in terms of the functional organization of entities instantiating cer-
tain material properties and relations. The intent is to provide a
naturalistic account of thought, reference and predication. Concepts
are not anything like the universals of Logical Realism which whol-
ly transcend the causal laws of nature. Concepts are intersubjective
cognitive capacities whose realization in thought is what informs
our mental acts with a predicable or referential nature.
On a conceptualist account of predication, all open well-formed-
formulas, including those not representing material attributes, stand
54 NOUS

for concepts. Understood as cognitive capacities, the ontic nature


of concepts is to be explained in terms of naturally real properties
and relations. But they are not identical with them. Nor is it assumed
that each concept represents a single material property or relation.
For example, for every wff standing for a concept, the negation
of that wff also stands for a concept. But this is not always true
where expressions standing for material properties and relations are
concerned. The complement of an expression standing for a material
property or relation does not automatically stand for a material prop-
erty or relation. Moreover, disjunctions and conjunctions of
predicates standing for material properties or relations do not usually
stand for new material attributes. (This has nothing to do with the
complexity of the linguistic expressions for them, however.) Finally,
'identity', is not a material relation. 'Identity' is a conceptual (or
logical) relation allowing the full substitutivity of its relata in all
contexts, including those standing for concepts that do not repre-
sent material properties or relations.
Now if universals are construed as material properties and rela-
tions physically realizable within the actual laws of nature, then the
"Identity of Indiscernibles" concerns whether indiscernibility with
respect to materialproperties and relations is sufficient for identity.
The principle should be represented as:
PuIN (Vx) (Vy) ((VNF) (Fx - Fy) - x = y),
where (VNF) ranges over material (natural) properties and rela-
tions. Is the PIIN of Natural Realism logically necessary?
In Natural Realism, the difference between concepts and material
properties and relations carries with it a distinction between two
notions of "necessity", conceptual necessity and causal/physical (or
natural) necessity. It is conceptual necessity that represents the notion
of logical necessity found in Logical Realism.7 A sentence is con-
ceptually necessary in virtue of its logical form alone. It is true under
all interpretations of its descriptive terms; we are not restricted to
interpretations which map the predicates of the language onto
material properties or relations realizable relative to the laws of
nature. The notion of a conceptually possible world is much wider
than the notion of a physically possible world. What counts as a
conceptually possible world is limited by maximal consistency alone.
Conceptual necessity is truth in all possible worlds, and this includes
worlds that are not physically possible. Physical necessity, on the
other hand, is truth in all worlds possible relative to the laws of
nature.
Of course, conceptual necessity implies physical necessity. But
physical necessity does not imply conceptual necessity. Since the
LEIBNIZ LAW 55

PIIN limits indiscernibility to just those material properties


realizable under the laws of nature, it is not conceptually (logically)
necessary. In fact, the PIIN is an empirical hypothesis which may
not even be true, though the following principle:
(Vx) (Vy) ( EJN[(VNF) (Fx Fy)] - x = y),
where "[ ]N" represents material necessity, is at least quite
plausible.
This shows that it is a confusion to present physically possible
worlds in an argument against the conceptual (or logical) necessity
of the P11. Certainly for all p, if p logically necessary then p physi-
cally necessary. But any such argument is equivocal when p is taken
to be "the" principle of the Identity of Indiscernibles. As we have
seen, there is more than one principle involved. The predicate quan-
tifier in the PJJN ranges over material properties alone. But in-
discernibility with respect to material properties should not be con-
fused with indiscernibility with respect to all concepts. Thus, find-
ing a physically possible world in which the PIIN is false has no
bearing whatsoever on the logical necessity of the P11.

6. THE PERSISTENCE OF COUNTEREXAMPLE

It remains, however, to consider whether Max Black's symmetric


two-sphere world provides a counterexample to the physical necessity
of the PIIN. Ian Hacking has challenged that the Identity of In-
discernibles can be preserved in spite of examples such as Black's.
According to Hacking, "it is vain to contemplate possible spatio-
temporal worlds to refute or establish the Identity of Indiscernibles"
(Hacking [11], p. 249). The reason, he continues, is that "if we
consider only objects which . . . occur in space and time, there is
no possible universe that must be described in a manner incompati-
ble with I/I [the Identity of Indiscernibles]" ([1 1], p. 249). Black
describes a possible world that is a symmetric universe in which
two spheres are some distance apart, and then challenges that the
spheres apparently share all properties. But Hacking contends that
this description is question begging. Black's possible world could
be redescribed as involving only one sphere, for space in the world
could be a Riemannian space so tightly curved that a journey in
some direction leads back to the single sphere (and not to the allegedly
distinct sphere some "distance" away).
Hacking employs the fact that pure geometry is an abstract for-
mal calculus, and must be provided with an empirical interpreta-
tion for its application. There are no semantically privileged terms
such as "straight", "distance", "parallel", and the like. (For ex-
ample, it is only after interpreting "straight line" as the path of
56 NOUS

a light ray, that we can go on to measure the metric of space.)


One is entitled to conclude that the geometry of space is Euclidean
only if the empirical interpretation satisfies the calculus characteriz-
ing a Euclidean space. Applying Hacking's observation to Black's
argument, the point is that Black is not entitled to merely imagine
a symmetric possible world with a Euclidean geometry and with
two spheres a distance apart. Black must empirically establish that
there can be a physically possible symmetric world whose geometry
is Euclidean and whose space contains just two spheres. To merely
assume, per imagination, that there is such a physically possible
world is to beg the question.
Now it turns out that the metric of space in our actual universe
is non-Euclidean on the large scale, (especially, near massive bodies
producing a strong gravitational field). Space approximates a Eucli-
dean metric only in small localities. This discovery forges a concep-
tual unification of the concepts of inertia and gravitation. Inertial
and gravitational forces are "covariant" -i.e., they vary depend-
ing upon the- point of reference adopted. Covariance illustrates that
Newton's "thought experiments" arguing for absolute space were
mistaken (Reichenbach [15], p. 212). Newton had appealed to "possi-
ble worlds" to argue that the consideration of dynamics shows motion
in space to be absolute. It is important that we pinpoint Newton's
fallacy, for Hacking's plan is to derive important lessons concern-
ing appeals to possible worlds in arguing against the Identity of
Indiscernibles.
Newton considered a universe with nothing in it except a pail
half full of water which is spinning with respect to the water. After
a short time, the water will be dragged along with the pail and
centrifugal forces will cause the surface of the water to have a hollow
shape. When the pail is stopped the water will continue to rotate
and the hollow shape will remain. Newton concluded that the cen-
trifugal forces causing the water to rise up the sides of the pail can-
not be explained by relative motion. If the pail rotates and the water
is at rest, then there would be no centrifugal force and no hollow-
ing effect.
Ernst Mach was among the first to cast doubt on Newton's argu-
ment. Mach pointed out that Newton's argument imports the ac-
tual laws of nature into the imagined universe. But if there are ex-
actly two objects, the water and the pail, we can say nothing about
what forces would go into effect. The situation is unverifiable. On
the other hand, if the pail and its water are in our universe, then
the hollowing effect could be due to the gravitational effect of the
movement of the earth and fixed stars around the water. Mach sug-
gested that in our universe the two descriptions of the situation are
LEIBNIZ LAW 57

indistinguishable and that science could not choose one over the
other.
Einstein transformed Mach's verificationist objection into an em-
pirical refutation of absolute space. Einstein's theory of General
relativity grounds Mach's objection in an empirically testable Prin-
ciple of Equivalence. The centrifugal forces operative when the pail
is regarded at rest will be explained as gravitational effects when
the reference frame is shifted so that the water is taken at rest. The
Principle of Equivalence is empirically testable, because the precise
relationship of inertial and gravitational forces expressed in the law
of the equality of inertial and gravitational mass is held to be due
to the non-Euclidean metric of space-a metric whose deviation from
the Euclidean is measurable. For example, Newton held that the
revolution of the Earth must be described in terms of the gravita-
tional force exerted by the sun being counterbalanced by the rec-
tilinear inertia of the Earth in Euclidean space. But Einstein's unifica-
tion of inertial and gravitational forces reveals that the Earth's revolu-
tion can be viewed as the result of the Earth's inertia carrying it
along the straightest geodetic path in the non-Euclidean space sur-
rounding the sun.
What Hacking draws as a lesson is that considerations of physical-
ly possible spatio-temporal worlds involve more restrictions than
might at first be apparent. One cannot simply stipulate the number
of objects in such a world and expect the same forces and geometry
to obtain. The physical forces and the geometry of our universe
are not unrelated to the number of objects, their masses, and their
dispersion in space. Such stipulations involved in examining physical-
ly possible worlds would be question begging, just as they were in
Newton's argument for absolute space.
Hacking goes on to draw a further lesson which appeals to a
certain kind of equivalence that holds between Euclidean, and non-
Euclidean descriptions of the geometry of space. The equivalence
is brought out nicely by Carnap in his discussion of General Relativ-
ity and Poincare's Conventionalism (Carnap [4], p. 156ff). The for-
mulation of non-Euclidean geometries led Poincare to adopt a con-
ventionalist position with respect to the question as to the actual
geometry of space. The physicist can choose whatever geometry he
finds convenient for science. Einstein appears to have revealed this
to be false; the geometry of space, he claimed, is in fact non-
Euclidean. Carnap shows how to reconcile the two.
Carnap points out that we can preserve a Euclidean metric pro-
vided we are willing to make appropriate adjustments in certain
physical laws ([4], p. 157). That is, if we choose a Euclidean metric,
we adopt a certain language for doing science. In this language,
58 NNOUS

we no longer speak of a rod bending in the presence of a gravita-


tional field due to the non-Euclidean nature of space; instead, the
rod will be said to undergo a specific degree of contraction in the
presence of the field. Similarly, the laws of optics will require
modifications due to the presence of a gravitational field, for such
fields will deflect light. On the other hand, if a language involving
Einstein's non-Euclidean metric is chosen, the laws of mechanics
and optics remain the same as in pre-Einsteinian physics. Solid bodies
are rigid except for certain usual deformations, such as elastic ex-
pansions and contractions, thermal expansions, and so forth. These
outside factors may influence a rod's length and must be considered
in the mechanical laws governing rigid bodies, but gravity will never
be considered among the factors. With respect to light, the laws
of optics remain as usual, gravitational fields do not deflect light.
The path of a light ray is simply a geodetic line in the non-Euclidean
space surrounding a massive body.
Carnap concludes that Poincare was correct in his claim that
choice between the two descriptions is wholly pragmatic. It was,
rather, Poincare's prediction that scientists would always choose
Euclidean geometry that was upturned. The geometry of space does
indeed become an empirical matter, but only after the choice has
been made to preserve the usual mechanical and optical laws rather
than to adjust them for the presence of gravitational fields. The
lesson Hacking draws is that it is always possible to preserve the
Identity of Indiscernibles by appropriately modifying the geometry
and the physical laws. In a non-question begging consideration of
a physically possible world, equivalence of description will hold. Thus,
one is free to alter the description so that the Identity of Indiscer-
nibles is preserved-provided one is willing to make (perhaps ex-
tensive) adjustments in the physical laws. In Hacking's view,
adherence to the Identity of Indiscernibles reflects a pragmatic choice
among equivalent descriptions. His position is analogous to
Poincare's conventionalist views concerning geometry. We can always
redescribe any physically possible world in which the Identity of
Indiscernibles fails in such a way that it, is preserved.
In discussing Hacking's argument, we have taken pains to
distinguish physical possibility from logical possibility. Unfortunately,
Hacking does not do this. His points are undermined by his failure
to carefully distinguish logically possible worlds and the PII, from
physically possible worlds and the PIIN. "Like the principle of suf-
ficient reason", he concludes, 'the I/I is not true in each possible
world. It is true about all possible worlds" (Hacking [11], p. 255).
Hacking seems to think that his conclusion addresses the status of
the PII as a logical truth. But Hacking's points concerning
LEIBNIZ LAW 59

equivalence are valid only in physically possible worlds-i.e., worlds


based on the laws of nature. The Principle of Equivalence is irrele-
vant where the logical possibilities of Logical Realism are concerned.
The Principle of Equivalence is also irrelevant to the question
of the logical truth of the PIIB. The subsistence of attributes is
held to be a logical truth in Attribute Realism. Hence, Black is
not restricted to "physically possible worlds" in the same way as
is a Natural Realist. For Black to maintain that his world is physically
possible he would, indeed, have to have empirical evidence that
relative to our natural laws, there can be a symmetric and Eucli-
dean universe in which there are two spheres a distance apart. Only
then could he go on to claim that the spheres may well share all
material properties and yet be distinct. But set in the framework
of Attribute Realism, Black's argument is not against the physical
necessity of the PIIN. His argument is concerned to show that the
PIIB is not logically necessary. A sentence is logically necessary
when it remains true under all admissible interpretations of its
descriptive terms. There is no restriction to those interpretations
which map the predicates of the language onto material properties
and relations realizable under the laws of nature. No appeal to any
physical properties of space-time or the laws of nature need be made
in finding a logicallypossible world. Thus, Black's argument cannot fail
on the grounds that equivalent descriptions of a physically possible world
are available. Black need only maintain that his imagined symmetric
world containing only two spheres a distance apart is logically possible-
i.e., it is a Euclidean space so that "(Vx) (Vy) (D(x,y) - xty)"
holds of 'distance', and it satisfies the sentences, "(ax) (Hy)D(x,y)",
and "(3tx) (ay) ((VBF) (Fx -Fy) & (x * y))". Where logical truth
is concerned, imagination (logical consistency) is sufficient.
Imagination is insufficient to counterexample the PIIN,
however. Hacking has made a convincing argument that it is a con-
sequence of the Principle of Equivalence that if we are clever enough
we can describe any physically possible world in such a way that
the PIIN holds.8 There are then no absolute facts about whether
a physically possible world has a Euclidean space with two materially
indiscernible objects, or a non-Euclidean space with one object. Both
are equivalent descriptions. The PIIN is always salvageable, though
it might demand a radical modification in the natural laws and a
shift in what we regard as the geometry of the world. This is not
to say, however, that the PIIN will rise to the status of a metaprin-
ciple for doing science-a principle which, if we consider worlds
possible relative to the laws of physics, is "not true in each possible
world. . .", but "is true about all possible worlds". Hacking's pro-
phetic remark that science will always choose to salvage the PIIN
60 NOUS

remains to be seen. Science may confound Hacking just as swiftly


as it did Poincare.
It is essential, therefore, to separate the background assump-
tions concerning the nature of universals. Otherwise proponents of
the debate over "the" Identity of Indiscernibles talk past one another.
Objectors to the PIIB need not concern themselves with Hacking's
points about physical science. Defenders of the PII need not be con-
cerned with either Hacking or Black. This does not mean that the
issue is settled. There may well remain problems internal to each
framework that are significant to the status of their respective "Iden-
tity of Indiscernibles" principle. In certain frameworks, however,
the issue is settled only by science and not the philosophy of logic.

NOTES
'For a discussion of the background formal logics of alternative theories of universals
and their philosophical foundations, see Cocchiarella [8].
2Logical Realism admits of a "Platonic" variant which maintains that universals have
an individual as well as a predicable nature. Thus, it allows nominalized predicates. We
shall avoid this complication, here.
3See, for example, Kripke [13].
4See Castajieda [6] for a clarification of the "Problem of Individuation" and a state-
ment of criteria for its solution.
5Castanieda's "guise theory", it should be noted, is a form of the bundle view which
adopts a different version of the conceptual thesis. Continuants are constructed to be sure,
but not solely from phenomenally simple properties; the construction allows relational prop-
erties to be members of the "bundle" analyzing a continuant. See Castafieda [5].
6Armstrong's [1] and [2] share much with Attribute Realism. His intended framework,
however, would seem to be that of Natural (or Scientific) Realism.
7We allow a "holistic" form of Conceptualism where impredicative concept formation
is possible (Cocchiarella, [8]). See Cocchiarella [7] for a discussion of the primary semantics
of logical necessity as opposed to a secondary semantics which would be appropriate for
material necessity.
81t is interesting to consider whether Hacking's argument based on Einstein's Principle
of Equivalence would also be applicable to the particles which are subject to the laws of
quantum mechanics. Cortez [9] argues that photons show "the" Identity of Indiscernibles
to fail as a logical truth. Paul Teller [16] thinks the argument is inconclusive. But neither
carefully distinguish material from logical universals and physical from logical necessity.

REFERENCES
Armstrong, David
[1] 1978 Nominalismand Realism, (Cambridge: Cambridge University Press).
[2] 1978 A Theory of Universals: Universals and Scientific Realism, (Cambridge: Cambridge
University Press).
Black, Max
[3] 1954 "The Identity of Indiscernibles," Mind 61
Carnap, Rudolf
[4] 1966 The Philosophical Foundations of Physics, (Basic Books, Inc.)
Castafieda, Hector-Neri
[5] 1974 "Thinking and the Structure of the World", Philosophia4, 3-40.
LEIBNIZ LAW 61

[6] 1975 "Individuation and Non-Identity: A New Look," AmericanPhilosophicalQuarterly


12, 131-140.
Cocchiarella, Nino B.
[7] 1984 "Philosophical Perspectives on Quantification in Tense and Modal Logic," in
Vol. 11 of The Handbook of Philosophical Logic, D. Gabbay and F. Guenthner,
eds., Dordrecht, Reidel, Vol. 11, pp. 309-353.
[8] 1986 Predication Theory and the Problem of Universals, (Naples: Bibliopolis Press).
Cortez, Alberto
[9] 1976 "Leibniz's Principleof the Identity of Indiscernibles:A False Principle", Philosophy
of Science 43: 491-505.
Foster, Thomas R.
[10] 1982 "Symmetrical Universes and the Identity of Indiscernibles", Philosophical
Research
Archives 3: 169-183.
Hacking, Ian
[11] 1975 "The Identity of Indiscernibles," Journal of Philosophy72: 249-256.
Hoy, Ronald R.
[12] 1984 "Inquiry, Intrinsic Properties, and the Identity of Indiscernibles," Synthese61:
275-297.
Kripke, Saul
[13] 1980 Naming and Necessity,(Cambridge: Harvard University Press).
Quine, W.V.O
[14] 1951 "Two Dogmas of Empiricism", PhilosophicalReview60: 20-43. Reprinted in From
a LogicalPoint of View, (Cambridge: Harvard University Press), pp. 20-46.
Reichenbach Hans
[15] 1957 The Philosophy of Space and Time, (New York: Dover Publications, Inc).
Teller, Paul
[16] 1983 "Quantum Physics, The Identity of Indiscerniblesand Some Unanswered Ques-
tions", Philosophy of Science 50: 309-319.
**Special thanks to Professor Nino B. Cocchiarella for helpful suggestions on
an early draft of this article.

You might also like