Buddhist reductionism

·期刊原文

Buddhist reductionism

Mark Siderits

Philosophy East and West

Vol.47 No.4 Oct 1997, Pp.455-478

Copyright by University of Hawaii Press

--------------------------------------------------------------------------------

There has been much recent discussion concerning the relation
between Parfit's Reductionism and the view of persons to be
found in early Buddhism and Abhidharma. Some have claimed
that Parfit is wrong to see in the Buddhist position an
important anticipation of his own view-since the Buddhist
view is not, as he supposes, Reductionist, but rather
Eliminativist.[1] Others suggest that while certain of the
Abhidharma schools might have held a Reductionist position,
early Buddhism may defensibly be interpreted as
Non-Reductionist in character.[2] It is something of a cliche
among scholars of Buddhist and comparative philosophy that
contemporary philosophical discussions would be enriched
through greater familiarity with the Buddhist philosophical
tradition. If such claims are to be taken seriously, then we
should at least be clear about how the Buddhist philosophical
problematic maps onto contemporary discussions. Further, we
need to demonstrate, through concrete case studies, that
Buddhist resources can facilitate genuine progress on current
disputes. I shall here propose a taxonomy of views on persons
and personal identity that I believe helps clarify the
relations among the various Buddhist and contemporary
positions on these issues. I shall also seek to show that
when we develop a taxonomy that accommodates both traditions,
certain of the currently held positions come to appear more
plausible.

We might do well to begin with a general characterization of
reductionism.[3] This is widely held to be, first and
foremost, a view about what belongs in our ontology. To be a
reductionist about things of kind K is, on this view, to hold
that while it is not wholly false to claim that there are Ks,
the existence of Ks just consists in the existence of certain
other sorts of things, things that can be described without
asserting or presupposing that Ks exist. Thus a reductionist
about mobs would maintain that while mobs may be said to
exist in a sense, the existence of a mob is really nothing
over and above the existence of certain particular persons
behaving in certain ways at a certain place and time. The
characteristic reductionist "just consists in" clause is
often explicated in terms of the "complete description" test:
that we could give a complete description of all the facts
about reality without ever mentioning Ks. Thus, suppose that
last night a mob set fire to government buildings in the
capital. Our reductionist about mobs would claim that we can
exhaustively describe this event just by describing certain
particular facts about the particular people in that mob, and
without ever mentioning the mob. Since the mob's action of
setting fire to those buildings just consists in certain
individual actions, nothing is omitted from our description
when we describe those individual actions and say nothing of
a mob engaging in arson.[4] And, claims the reductionist
about mobs, if we can give a complete description of reality
without either asserting or implying that mobs exist, mobs
are thereby shown to be ontologically superfluous, and thus
have no place in our ultimate ontology.

When reductionism is seen in this way, non-reductionism is
then readily characterized as the view that things of kind K
do belong in our ultimate ontology--that there is nothing
that Ks "just consist in," since Ks are ontologically
primitive. (Hence a non-circular analysis of "K" cannot be
given.) But the third possible view about Ks, the
eliminativist view, is not so easily captured on a purely
ontological approach to the issue. For then the eliminativist
about mobs would seem to be one who proposes that we
eliminate mobs from our ontology altogether, and it is far
from clear how this position would differ from that of the
reductionist about mobs. It is only when we bring in the
semantic dimension of the issue that these two types of
positions can be clearly distinguished. To say, for instance,
that organic chemistry may be reduced to quantum mechanics is
to make a certain claim about the relation between two
theories: talk of covalent bonds may be systematically
replaced by talk of certain quantum mechanical states, so
that the latter theory is thought to explain the predictive
success of the former. By contrast, the demonic-possession
theory of disease does not reduce to the germ theory, for
there is no way systematically to translate talk of being
possessed by a certain demon into talk of bacterial or viral
infection.[5]

Thus it is that we take a reductionist stance toward the
covalent bond, but an eliminativist stance toward
disease-causing demons. The term "covalent bond" is now
revealed to refer to certain distinctive sorts of quantum
mechanical phenomena. So while a complete description of
reality need not mention covalent bonds (whereas it would
have to mention such things as quantum shifts), we may
tolerate talk of such things just because the term is a
useful way to refer to a certain class of quantum phenomena
in which we take an interest. With demons, though, things are
quite different. When we come to accept the germ theory of
disease, it becomes apparent that our former talk of being
possessed by demons cannot be seen as just a rough-and-ready
way of referring to bacterial or viral infection. For one
thing, demons are thought to have properties, such as
malicious intent, that could not be explained on the
assumption that they just consist in masses of microscopic
replicators. For another, a given kind of demon was thought
to be responsible for what we now consider to be essentially
unrelated diseases (the viral versus bacterial pneumonia
problem). Whereas the covalent bond is a posit of a useful
though subsumed theory, the demon is a posit of a discredited
theory; hence all talk of demons is to be eliminated.

Distinguishing between reductionism and eliminativism
requires introduction of the semantic dimension. But this
complicates matters significantly. For success at translating
between one theory and another is something that admits of
degrees. Consider the terms "sunrise" and "sunset," which are
intermediate between the case of the covalent bond and that
of the disease-causing demon. We might have thought, when we
transferred allegiance from the geocentric model to the
heliocentric model of the solar system, that these terms were
ripe for elimination. Yet they survive. Had we expected
otherwise, this would have derived from the notion that their
meanings seemed inextricably bound up with the now
discredited geocentric theory. Instead, these terms exhibited
sufficient semantic flexibility that we could retain them
while suppressing the implication that the astronomical
phenomena are explained by the sun's motion around the earth.
This semantic shift was not accompanied by a shift in
supposed referent: we take the Ptolemaic astronomers to have
been referring to the same thing we refer to with these
terms. We can imagine circumstances under which something
similar might have occurred with our talk of demons (and as
did happen with "humor") . What this suggests is that
reductionism and eliminativism represent the ends of a
continuum, with a middle range of cases in which it may be
indeterminate whether the entities of the old theory are
being reduced to, or eliminated in favor of, the entities of
the new theory. But as is often the case with sorites
phenomena, the existence of such an intermediate gray area
need not count against there being a real distinction to be
drawn between reductionism and eliminativism.

Characterized semantically, then, non-reductionism about Ks
will be the claim that Ks will be mentioned in our final
theory about the ultimate nature of reality. Both
reductionists and eliminativists deny this claim, but they
disagree over whether continued talk of Ks will have any
utility in the light of our final theory. The eliminativist,
of course, proposes eliminating all talk of Ks, both in our
final theory and in its ordinary language adjuncts. It might
be thought, then, that the reductionist sees matters thus:
while the term "K" is in principle eliminable from our
language (since we can give a complete description of reality
without mentioning Ks), its continued use is both tolerable
(because of translatability), and of some utility given our
interests.

Now I believe this is an accurate portrayal of the
reductionist position, but there are those who would dispute
this. Some draw a distinction between strong and weak
reductionism. The strong reductionist (or "conventionalist")
is said to subscribe to the eliminability thesis, while the
weak reductionist does not. A weak reductionist about Ks
maintains that Ks have a kind of "conceptual priority" such
that, although a noncircular, nontrivial analysis of the term
"K" can be given, some of the concepts occurring in the
analysis will be adjectival on "K," in the way in which
"citizen" is adjectival on "nation."[6] Thus the weak
reductionist denies what the strong reductionist affirms:
that Ks are ontologically parasitic on, or logically
constructed out of, more particular things; that we believe
Ks to exist only because of the way in which we talk (hence
the label "conventionalist" for strong reductionism). I have
already indicated that I find this distinction problematic.
It is not clear to me that so-called weak reductionism is
properly characterized as a kind of reductionism at all. But
further discussion of this point must await an examination of
the reductionist, non-reductionist, and eliminativist
approaches to the concepts of a person and personal identity,
to which we now turn.

Parfit's general account of Reductionism (1984, pp. 210-214)
provides a convenient starting point for this examination.
Parfit characterizes Reductionism about persons as the view
that the existence of a person just consists in the existence
of a brain and body, and the occurrence of a series of
interrelated physical and psychological events. Given the
reductionist force of this "just consists in," all of the
facts to which the existence of a person is here being
reduced--the existence of a particular brain, the occurrence
of a certain psychological event, the holding of causal
relations between a particular psychological event and
certain other physical and psychological events, and so
forth--will allow a completely impersonal description, that
is, a description that neither asserts nor presupposes that
this person exists. Hence the Reductionist view of the
identity over time of persons: that this just consists in
more particular facts, facts that can be described in a
thoroughly impersonal way. (This general schematism is
fleshed out by neo-Lockeans with purely psychological facts,
by others with facts about the body or the brain; hence there
are two possible Reductionist approaches to personal
identity.)

There are also two possible Reductionist views about what a
person is: that a person just is a particular brain and body
and a series of interrelated events, and that a person is an
entity that is distinct from a brain and body and such a
series of events. To those familiar with the Buddhist
problematic, this will sound rather like the question whether
a person is identical with or distinct from the five skandhas
(bodily constituents, feelings, perceptions, volitions, and
consciousnesses). And since the standard Buddhist response to
this question is that the person is neither, one might be
inclined to wonder how much affinity there is between
Reductionism and any Buddhist view. Our puzzlement is
heightened when Parfit also tells us that on the "distinct
entity" view that he prefers, not only may persons be said to
exist, they may also be said to be the subjects of
experiences, the owners of their bodies, and the like. And
yet all Reductionists, we are told, deny the Non-Reductionist
thesis that a person is a separately existing entity, that
is, an entity whose existence is distinct from that of a
brain and body and a series of interrelated events.

Is Parfit's preferred form of Reductionism coherent? Indeed
it is. The key to understanding the view may be found in this
passage:

Even Reductionists do not deny that people exist. And, on our
concept of a person, people are not thoughts and acts. They
are thinkers and agents. I am not a series of experiences,
but the person who has these experiences. A Reductionist can
admit that, in this sense, a person is what has experiences,
or the subject of experiences. This is true because of the
way in which we talk. (1984, p. 223)

The implication, of course, is that we conceive of persons as
existing distinct from their experiences, and so forth, only
because of the grammar of our language. A person is not in
fact an entity that exists separate from a particular body,
and so forth; our belief that they are is merely an artifact
of our language, yet it is not entirely wrong to continue to
speak of persons in this way. What is at work here is the
tension characteristic of all reductionisms between a
deflationary tendency that seeks to diminish the significance
accorded the things being reduced, and a felt need not to
sever all connections with existing practices. Since persons
are not to be found in our ultimate ontology, the continued
existence of a person cannot have the kind of importance it
is commonly thought to have. Yet since our ordinary ways of
conceiving of persons are not wholly mistaken, we may expect
that at least some of the normative weight that we now invest
in persons will be preserved in the more particular facts to
which the existence of persons is to be reduced. Reductionism
is a middle path between the extremes of Non-Reductionism and
Eliminativism.

A Non-Reductionist need not maintain that persons are
separately existing entities. Parfit distinguishes between
two versions of Non-Reductionism: the view that persons are
separately existing entities (such as Cartesian Egos) and the
view that while we are not separately existing entities, the
existence of a person involves a further fact, over and above
the "more particular facts" of the existence of a brain and
body and the occurrence of a series of interrelated physical
and psychological events. The classic expression of the first
version in the Western tradition is, of course, Descartes'
conception of the "1" as a thinking substance; Swinburne
(1984) represents a more recent formulation. (This type of
view is also well represented in the Indian tradition, for
example in the Nyaya theory of the atman and the Samkhya
theory of purusa.) The basic idea here is that the human
being (and possibly other life-forms as well) is a complex
system consisting not only of those things, such as body
parts and mental events, that are ordinarily observable
through sense perception and introspection, but also
containing some one entity that constitutes the essence of
the system, that one part the presence of which is required
in order for the system to exhibit the properties that we
ascribe to persons; it is the continued existence of this
entity that constitutes personal identity over time.

While this type of view is relatively familiar, the "further
fact" version of Non-Reductionism seems rather more puzzling.
Here we appear to have a reluctance like that of the
Reductionist to posit extra entities, together with the
insistence that certain key facts about persons cannot be
accounted for in terms of more particular, wholly impersonal
facts. In the Western tradition, Reid, Butler, and more
recently Chisholm have put forward views that might be
interpreted as of this sort. But in each of these cases there
are also elements that seem to suggest a belief in persons as
separately existing entities.[7] And it is not difficult to
see why clear-cut instances of this position might be hard to
come by, since if it is true that persons exist and it is
also the case that the concepts of person and personal
identity are simple and unanalyzable, we should expect there
to be some distinct entity the existence and persistence of
which explain these singular facts about persons and their
identity over time. The Vatsiputriyas or Pudgalavadins of the
Buddhist tradition do, though, represent a clear instance of
this type of Non-Reductionism; this is precisely the force of
their claim that the person (pudgala) is, while existent,
neither identical with nor distinct from the skandhas.[8]

Parfit does not himself discuss Eliminativism, so to see
where this fits on the philosophical terrain we need to turn
to those authors, namely Giles and Stone, who advance this
view as the Buddhist response to Parfit's Reductionism.
Unfortunately, Giles is not especially helpful, apparently
falling prey to some serious terminological confusion. He
begins by distinguishing between reductionism and
eliminativism through the use of an example from philosophy
of mind, the dispute between identity theorists and
eliminative materialists. And he notes the characteristically
eliminativist tenor of the latter's claim that folk
psychology is so hopelessly confused as to defy translation
into any scientific discourse, and so should be abandoned.
Yet he goes on to propose that we call Eliminativist the
early Buddhist anatman theory (as well as Hume's Treatise
position), despite the fact that on that view we will, "on
pragmatic grounds, continue to permit the use of the language
of personal identity" (1993, p. 176). But if talk of persons
and their identity over time has some utility (as all
Buddhists in fact maintain), then our theory of persons
cannot be a candidate for utter elimination. He also takes
the Milindapanha denial that the person is identical with the
psychophysical complex as a whole to be a rejection of
Reductionism. But Reductionists do not claim that a person is
identical with a certain sum of impersonal elements; instead
they claim that the existence of a person just is the
occurrence of certain impersonal elements. While Giles sees
the importance of the Buddhist theory of two truths, he fails
to see how it may be used to mark the difference between "is
identical with" and "just is."

Stone (1988) is more helpful. Eliminativism is to be
distinguished from Reductionism not in terms of the denial of
a self (both agree that we are not "something extra") or in
terms of the denial that persons are to be found in our
ultimate ontology (both deny that persons have this
privileged status), but rather in terms of the question
whether the attitudes we ordinarily take toward ourselves and
others are at all coherent. The Non-Reductionist claims that
such things as prudential concern, anticipation, regret,
responsibility for past deeds, merit, and the like all
require that there be something extra, over and above body
and brain, and so forth. The Reductionist denies that there
is this something extra, but holds that such attitudes may
still be rational (even if their scope is somewhat altered
when we come to accept Reductionism). The Eliminativist
agrees with the Reductionist that we are not something extra,
but also agrees with the Non-Reductionist that our attitudes
toward persons are coherent only if we are something extra.
Prudential concern, hopes, fears and regrets, judgments of
responsibility, merit, and praise and blame--all these are
irrational. And since Locke is right to see the forensic
elements as central to the concept of a person, it follows
that all talk of persons is deeply incoherent. In place of
the mildly dismissive Reductionist attitude toward
persons--as "mere constructions" out of more fundamental
entities--we find in Eliminativism an outright rejection of
all that persons are thought to be. Like the demons believed
in by our ancestors, persons are posits of an utterly
misguided theory.

Like Giles, Stone identifies the Buddha as an Eliminativist.
As will already be evident, I believe this is mistaken. But
it may prove worthwhile to examine why some might see early
Abhidharma as Eliminativist rather than Reductionist. First
we need to see how the taxonomy we have developed so far
might need adjustment in order to accommodate the Buddhist
problematic, and so we turn to an examination of some
representative early Buddhist and Abhidharma texts. When this
is done we will return to the question of reading Buddhism as
Eliminativist.

One passage in the early Buddhist text Milindapanha[9]
clearly exhibits signs of Reductionism. The text as a whole
is in the form of a dialogue between the Buddhist monk
Nagasena and a king, Milinda by name, who seeks to enhance
his understanding of Buddhism by asking Nagasena a series of
probing questions concerning the system. The passage in
question opens with Milinda asking whether adult and infant
are the same person or distinct persons. Nagasena replies
that they are neither the same person nor distinct
persons,[10] and asks Milinda's view. The king replies that
adult and infant are distinct persons. This answer is rather
surprising, since throughout the work Milinda tends to
represent the commonsense view of things, and most people
would judge the adult Milinda to be the same person as the
infant Milinda. But it appears from the context as if in this
case the king is not to serve as representative of our
commonsense intuitions, but is rather expressing a certain
understanding of the Buddha's teachings on persons. The king
has already learned through his discussions with Nagasena
that the person is made up of five skandhas, none of which
exists continuously throughout the course of a lifetime, and
thus that the person is devoid of a self. He now reasons that
since none of the skandhas that made up the infant is present
in the adult, adult and infant must be distinct persons. He
supposes, that is, that personal identity over time requires
the continued existence of some one entity through the
distinct stages in the life of a person; since the skandha
analysis reveals the absence of any such entity, it follows
that personal identity does not extend over any substantial
portion of a lifetime.

Anyone who has taught the early Buddhist argument for nonself
(anatman) from the impermanence of the skandhas will
recognize this response, since it is quite common for
students to understand the Buddhist teaching in just this
way. But Nagasena makes it clear through a series of
reductios that Milinda has reasoned incorrectly. By the same
reasoning, it would follow that there are no mothers or
fathers, no educated persons, and no one who deserves
punishment for past crimes. A mother, for instance, is a
person who conceives, bears to term, and then gives birth.
But the skandhas making up the woman who conceives are no
longer present, for example, in the woman carrying a
second-trimester fetus. By Milinda's reasoning, then, the
woman who gives birth is not the same person as the woman who
conceived, or the woman who bore the fetus, and so is not a
mother. Milinda agrees that these results are unacceptable,
and asks Nagasena how he views the matter. Nagasena replies
that adult and infant are the same person. This would appear
to contradict his earlier response, that adult and infant are
neither the same person nor distinct persons. But Nagasena
goes on to explain that those skandhas making up the adult
have as their causal antecedents the skandhas that made up
the infant; impermanent elements existing at distinct times
are collected together--that is, make up a person--when they
bear the right sorts of causal relations to one another.

This is illustrated with the example of the one light that
shone all night. If a lamp were to be lit in the evening and
burned continuously until the morning, we would agree that
there was one light--one source of illumination--that shone
all night. Yet it is agreed that that which actually
illuminates at any one moment during the night, namely a
flame, is numerically distinct from that which illuminates at
any other moment. This follows from the physics of flames,
for a flame is a collection of fire atoms (by classical
Indian physics) or incandescent hydrocarbon molecules (by our
physics), these entities undergoing constant replacement. Yet
we do not, for all that, say that there were many distinct
lights illuminating the room over the course of the night; it
is not incorrect to say that there was just one light shining
all night. This, explains Nagasena, is because the flame at
4:00 A.M. has as its remote causal antecedent the very first
flame in the series, that which occurred when the lamp was
lit at 9:00 P.M. As long as the right conditions obtain, a
given collection of fire atoms will, in going out of
existence, cause a new collection to come into existence.
Whenever distinct flames form such a causal series, this may
be referred to as one continuously existing light.

Earlier, Nagasena told Milinda that the name "Nagasena" is a
mere convenient designation for a causal series of sets of
impermanent skandhas. So here he might say that the
expression "one light that shone all night" is a convenient
designation for a causal series of collections of fire atoms.
In the Nikayas, the canonical literature of early Buddhism,
"chariot" was the stock example of a convenient designation.
In the Abhidharma literature we find such other instances as
"forest," "village," "oven," "head hair," and "army." To call
"forest" a convenient designation is to say approximately the
following: given the regular occurrence in the world of
certain sorts of clusterings of trees, and the nature of the
interests we have in such clusterings, it has proven
advantageous that we have a single term that may be used to
refer to such clusterings, and thus avoid the prolixity
involved in referring to the individual trees and their
relations to one another.

The concept of a convenient designation plays a key role in a
two-part strategy for undermining belief in a self. First it
is argued, through the use of the skandha analysis, that
there is no empirical evidence for the existence of an entity
having the properties of a self: continued existence
throughout a lifetime, being the subject of experience,
performing the executive function, and the like. Then it is
argued that our belief in a self is generated by our use of
such convenient designations as "person, " the personal
pronouns, and personal names, in conjunction with the
misguided acceptance of a naive semantic realism that takes
the meaningfulness of a word to require the existence of some
entity bearing that word as its name. Thus, since all of our
ways of conceptualizing persons may be accounted for without
supposing there to be anything more to the existence of a
person than just a complex causal series of impermanent
collections of skandhas, and there is no empirical evidence
for the existence of anything other than the skandhas, we
have no reason to believe that the existence of a person
involves anything other than impersonal phenomena in a
complex causal series.

It is possible to see the strategy that Parfit employs in
defense of Reductionism as proceeding along similar lines.
For Parfit first argues (1984, pp. 223-228) that we do not
have the sort of evidence that would support the claim that
we are separately existing entities. He then proceeds to try
to show, through the use of various puzzle cases, that
personal identity over time is subject to the same sorts of
sorites difficulties as are heaps, clubs, and nations. And
since most of us are willing, when confronted with such
difficulties, to say that heaps, clubs, and nations are said
to exist only because of the way that we talk, this suggests
that Parfit sees "person" as a kind of convenient
designation. If such a reading is correct, then this parallel
counts as evidence in support of the claim that early
Buddhism and Abhidharma are Reductionist. But Nagasena's use
of the notion of a convenient designation brings out
something else as well. While the Milindapanha does not
employ the doctrine of the two truths, this passage nicely
illustrates the tensions that led to the development and
articulation of this device in the Abhidharma.

The doctrine of the two truths distinguishes between what is
called conventional (samvrti) truth and ultimate (paramartha)
truth. A statement is said to be conventionally true if it
conforms to common sense, that is, if it is in accordance
with conventionally accepted linguistic and epistemic
practices. If we use "conceptual fiction" to refer to
whatever is thought to exist only because of our use of a
convenient designation, then "ultimate truth" may be defined
as follows: a statement is ultimately true if and only if it
corresponds to the facts and neither asserts nor presupposes
that conceptual fictions exist. Given a sufficiently
restrictive ultimate ontology, it will then turn out that
most statements that are conventionally true are ultimately
false. It thus becomes necessary to explain why most such
conventionally true but ultimately false statements appear to
have utility for human practice. The explanation is simply
that while certain of the entities quantified over in such a
statement do not exist, it is possible systematically to
replace all talk of such entities with talk of entities that
do ultimately exist, thereby arriving at a statement that is
ultimately true, that is, that does correspond to facts the
constituents of which belong to our ultimate ontology. Thus,
most conventionally true but ultimately false statements are
amenable to full translation[11] into ultimately true
statements; all that is lost in such translation is the
misleading implication that conceptual fictions exist.[12]

This device allows us to resolve the seeming contradictions
in Nagasena's position. When he claims that adult and infant
are neither the same person nor distinct persons, he is
stating what he takes to be an ultimate truth. Since persons
are conceptual fictions, any claim concerning the identity
over time of a person must be ultimately false. His assertion
that adult and infant are the same person he takes to be only
conventionally true. Likewise Milinda's assertion that adult
and infant are distinct persons Nagasena claims to be
conventionally false. For it is our conventional practice to
refer to infancy and adulthood as merely two stages in the
life of one continuous person. And a whole host of other
customary practices involves this as presupposition, for
instance the notion that as adults we are obligated to care
for our aged parents because they cared for us as infants.
But the conventional truth that I am the same person as the
child pictured in a certain old photograph may be accounted
for in terms of facts about a complex causal series of
collections of skandhas, just as the conventional truth that
there was one light that shone all night may be accounted for
in terms of the facts about a causal series of collections of
fire atoms. We may eliminate all talk of such conceptual
fictions as persons and enduring lights, yet preserve the
underlying truths that such statements are attempting, in
their rough-and-ready way, to assert.

I think we can now see why early Buddhism and Abhidharma have
often mistakenly been seen as Eliminativist. The error arises
through attending solely to what is said at the ultimate
level of truth, and failing to appreciate the relation
between the ultimate and conventional levels of truth.
Eliminativism is not simply the view that talk of persons may
in principle be eliminated. Both Reductionist and
Eliminativist maintain that ultimately there are no persons.
But the Eliminativist urges in addition that the claim that
there are persons be seen as conventionally false as well,
since the Eliminativist maintains that our commonsense theory
of persons is incoherent, or at least so misleading as to be
more troubling and confusing than theoretically useful.[13]
By contrast, the Reductionist holds that while unquestioning
adherence to the commonsense theory of persons does result in
misguided views about how we should live our lives, the
theory does have its uses, which fact requires explanation;
hence it is conventionally true, though ultimately false,
that there are persons. Like Milinda's wrong view about adult
and infant, the reading of early Buddhism as Eliminativist
results from the failure to consider the semantic dimension
of the dispute.

I claimed earlier that a reductionist about things of kind K
sees much of our talk of Ks as not wholly incorrect. The
doctrine of the two truths gives us one way of understanding
what this "not wholly incorrect" might come to: ultimately
false, since "K" is a mere convenient designation and Ks are
not in our ultimate ontology; but still conventionally true,
since talk of Ks both has utility given our interests and
customary practices, and may systematically be replaced with
talk of entities that are in our ultimate ontology. There are
those, however, who, while describing themselves as
reductionists, would resist this assimilation, since they
have serious reservations about the notion of an ultimate
ontology. This is clearly so for those so-called "weak"
Reductionists, such as Shoemaker and Brennan, who deny that
talk of persons is eliminable from our discourse, and deny
that persons are logically constructed out of more particular
and completely impersonal things. But this might also be true
of a "strong" Reductionist like Parfit, who seems somewhat
reluctant to commit himself to any one ultimate ontology.[14]

Now such reluctance may reflect nothing more than a
rhetorical strategy designed by Parfit to appeal to the
widest possible audience. But it might instead come from the
belief that any ontology must always be provisional, forever
open to revision in the light of future experience and future
changes in human interests and practices. I would maintain,
however, that reductionism requires minimally that the notion
of an ultimate ontology be held to be coherent, something to
whose attainment we may at least sensibly aspire. For the
normative force of reductionist claims requires ontological
backing.[15] The reductionist's strategy is to persuade us
that Ks have less importance than we are wont to believe, by
first convincing us that Ks are really just constructions on
more primitive entities. And the notion of relative
primitiveness at work here makes sense only on the assumption
that there is such a thing as the ultimate ontology. Parfit
first argues that we are not what we believe (separately
existing entities), and then uses this result to support his
claims concerning rationality and morality.

The effectiveness of such a strategy depends on the implicit
premise that persons would not have the sort of rational and
moral significance ordinarily ascribed to them unless they
were themselves ultimately real entities. Indeed, the theory
of two truths, and the view of ontology that that theory
implies, are admirably suited to express the points that
Parfit seeks to express--for instance, by claiming that the
facts of personal identity over time and the separateness of
persons are "less deep" on the Reductionist view, and so lack
the rational and moral significance we tend to give them
(e.g., 1984, p. 337). For the metaphor of depth may be
replaced along the following lines: a fact is "less deep"
just in case a statement expressing that fact is
conventionally but not ultimately true. But to say this is
just to say that some of the entities referred to in that
statement are not ultimately real, are not to be found in our
ultimate ontology.

That the account of persons developed in early Buddhism and
Abhidharma is meant to have normative force is clear.
Liberation from the cycle of rebirth is said to result from
the realization that there is no self, and that the continued
existence of a person just consists in a complex causal
series of collections of physical and psychological elements.
And liberation involves the rejection of that mode of life
typical of the "householder, " with its characteristic
attachments to home, family, and occupation as ongoing
enterprises, as well as its characteristic concern for one's
prospects upon rebirth. All such attachments and concerns are
to be replaced by a way of life marked by equanimity,
spontaneity, and the developed capacity to feel sympathetic
joy at the welfare of all. The sorts of significance that we
ordinarily attach to our life projects and our situation
beyond this life are undermined by the truth about what we
ultimately are.

Buddhism is often said to be a kind of middle path between
two extremes. A number of different pairs of opposing extreme
views are identified in the tradition, but one such pair is
frequently singled out for special emphasis, namely that of
eternalism and annihilationism. Eternalism is the view that
there is an eternal self, and thus that rebirth is
transmigration. Annihilationism is usually portrayed as the
view that the self ceases to exist at the end of a single
lifetime. Eternalism is said to have the normative
consequence that since one will deserve the karmic fruits
reaped in the next life from one's present deeds, we all have
a reason to act in accordance with the karmic moral rules.
Annihilationism, by contrast, is said to result in a radical
antinomianism: since there can be no karmic retribution
beyond this present life, one has no reason to act morally
where doing so involves sacrificing one's own immediate
gratification. Both views are, the Buddhist claims, false
because of their shared presupposition that a self exists.
Yet some of the normative consequences of eternalism must be
preserved, since belief in the karmic moral order is required
if persons are to progress toward enlightenment.

The middle path between eternalism and annihilationism thus
consists of a demonstration that rebirth is compatible with
the nonexistence of the self. Since the continued existence
of a person in one life just consists in the obtaining of
appropriate causal connections among various physical and
psychological events, the continued existence of a person
over several lives is likewise possible in the absence of an
enduring self, provided the right sorts of causal connections
obtain between lives. And our alleged ability to recall
events from past lives presumably shows that such connections
do obtain. Thus to the extent that one is justified in
feeling concern for what happens to oneself in the later
stages of this life, concern is equally justified with
respect to one's future lives. Annihilationism wrongly
assumes that only the continued existence of a self could
give one a reason for self-interested concern, yet this is
clearly false in the case of a single lifetime. As the life
of the enlightened person demonstrates, one can know that
there is no self yet not lose oneself in a "solipsism of the
present moment."[16] While enlightened persons do not exhibit
self-interested concern in the same way and to the same
extent as the unenlightened, they do appear to be motivated
by considerations concerning how their present acts will
affect them in the future.

Parfit describes three possible views about what it is
rational for an agent to seek: (1) the classical
self-interest theory (S), according to which we as rational
agents should ultimately be governed by a temporally neutral
bias in our own favor; (2) the extreme claim (E), that if
Reductionism is true then we have no reason to be concerned
about our own futures; and (3) the moderate claim (M), that
if Reductionism is true, then the causal connections
obtaining between different stages in our lives give us some
reason to be concerned about our own futures. Parfit claims
that since Reductionism is true, S is false; but he also
asserts that both E and M are defensible.[17] Clearly, this
dispute does not map perfectly onto the Buddhist problematic
of eternalism, annihilationism, and a middle path between
them. I would claim, however (though I cannot argue at any
length for this here), that the Buddhist discussion of
eternalism and annihilationism suggests that M should be
accepted and E rejected. Like annihilationism, E appears to
be motivated not by Reductionism but by Eliminativism. The
Reductionist would agree that since there are ultimately no
persons, such anticipatory attitudes as dread and such
retrospective attitudes as regret cannot rationally be
justified at the ultimate level of truth.[18] But persons are
conventionally real: the practice of speaking of ourselves as
persons has greater overall utility than the available
alternatives.[19] And so, certain person-involving attitudes
may turn out to be rationally justifiable at the conventional
level of truth.

Consider this analogy. Most of us would agree that a city
just consists in certain buildings, streets, persons, and so
forth, arranged in certain characteristic ways; strictly
speaking there are no cities. Still, given the utility of the
convenient designation "city," a certain degree of civic
pride may be justifiable. The overweening pride of civic
chauvinism is ruled out, since it seems to require that one
think of the city as a separately existing entity. But urban
aggregations themselves have some degree of utility, and this
is enhanced by the behavior that results when their
inhabitants exhibit some appreciation for the character of
the particular aggregate in which they reside. Thus the
practice of encouraging civic pride has a consequentialist
justification. By the same token, the practice of thinking of
ourselves as persons can be expected to have significant
utility. Much of this stems from its facilitating such
person-involving attitudes as anticipation and regret.
Suppose that I, realizing that the action I now contemplate
will result in future pain for me, am deterred through
anticipating that I shall experience pain. Now there is no
further fact that makes it the case that the person who will
feel that pain will be me; this fact just consists in the
obtaining of certain relations among certain purely
impersonal present and future entities and events. I do
nonetheless have a special reason for refraining from the
action, namely the fact that this is (typically) the best way
to insure that that future pain does not occur. Because pain
is bad, we all have a reason to try to prevent its
occurrence; and in general I am better positioned than anyone
else to prevent my own future pains. Existential dread may be
unjustifiable, but a moderate degree of concern over one's
anticipated future pain does have considerable utility.

Buddhist Reductionism thus has significant normative
consequences. I claimed above that such consequences require
ontological grounding, hence that reductionism is to be
construed as a thesis about our ultimate ontology. It would
be fruitful to examine how the Buddhist Reductionist proposes
to determine the contents of our ultimate ontology. It is
striking, for instance, that while sorites-induced
difficulties are crucial to Parfit's argument against
further-fact Non-Reductionism, sorites paradoxes play no role
whatever in the arguments used to support Buddhist
Reductionism. Indeed, the Indian tradition overall has no
Chariot of Devadatta problem to answer the Ship of Theseus
puzzle. Instead, in the texts of early Buddhism there is the
expectation that we will simply agree that chariots are
ultimately unreal, once it is pointed out that a chariot is
an assemblage of parts. In later Abhidharma texts we can
discern an argument against the existence of wholes in
general, but this argument makes no use of the
boundary-setting difficulties that the admission of partite
entities presents us with. I would suggest, though, that this
is not because Indian philosophers were simply ignorant of
sorites phenomena. Rather, those Indian philosophers, the
Naiyayikas, who championed the existence of wholes went to
extraordinary lengths precisely to insulate their doctrine of
complex substances from problems of indeterminacy.

Thus, Nyaya maintains that the addition or subtraction of a
part from a given substance results in the destruction of
that substance and the coming into existence of a new
substance. And likewise for change in the arrangement of the
parts, and for such qualitative changes as change of color,
taste, odor, texture, and the like.[20] This position
obviously rules out many of the sorts of spectra of
indeterminacy that would otherwise plague the champion of
complex substances. I know of no Nyaya text explicitly
linking this position with the intention to avoid such
difficulties. But I find it hard to imagine what else might
motivate the adoption of such a counterintuitive view,
particularly on the part of the eminently sensible
Naiyayikas. In any event, the Buddhist Reductionist was not
given the opportunity to exploit sorites difficulties in
arguing against persons and other such wholes; those
difficulties had been anticipated by the opponent.

Now it would still be open to the Buddhist Reductionist to
use sorites difficulties to argue against the existence of
persons, since if these are to explain personal identity over
time then they must be thought of as complex substances that
endure through replacement of parts, qualitative change, and
the like. (Nyaya accounts for personal identity over time by
means of the continued existence of an impartite self, not
through the continued existence of a person.) But rather than
pursuing this strategy, they seek instead to show that
partite entities in general must be thought of as mental
constructions and are thus not ultimately real. But then the
ultimate point of the sorites strategy as wielded by the
reductionist about Ks is precisely to show that the Ks in
question are just the result of human conventions, are a mere
mental fabrication.[21] For genuine vagueness induces
bivalence failure at many different levels: not only is there
a gray area where one may say of a given number of grains of
sand neither that it is nor that it is not a heap; it is also
indeterminate just where this zone of indeterminacy begins
and ends in the heap spectrum.

Genuine vagueness is thus not the result of mere ignorance,
something that could always be overcome by one sort or
another of precisification. And for the metaphysical realist,
it is inadmissible that mind-independent reality should in
itself be the source of such indeterminacies. If
mind-independent reality is to be the final arbiter of the
truth of propositions, then reality itself must be fully
determinate; any failures of bivalence in our theory about
the world can only result from that theory's not having
carved up reality at its joints--from our having used
concepts and categories not wholly derived from the nature of
mind-independent reality. To the metaphysical realist,
sorites phenomena can only indicate an element of mental
construction; it is this that makes such phenomena useful for
reductionist purposes. The Buddhist Reductionist simply cuts
to the chase--avoids discussing sorites phenomena and
proceeds directly to the notion of mental construction.

The Buddhist Reductionist argument against the existence of
real wholes--that is, for the conclusion that the partite is
mentally constructed-is relatively simple and
straightforward. Suppose we agree that the parts of, for
example, the chariot are themselves real. If, in addition to
the chariot parts, the chariot itself is thought to be real,
then it cannot be said to exist distinct from the parts given
that these are related to one another in the manner that
results from their assembly. For there is no evidence for the
existence of a chariot that is not just evidence for the
existence of one or more chariot parts and their assembly
relations; a distinct chariot is a superfluous posit. But
neither can it be said that a real chariot is identical with
the assembled chariot parts. For the chariot may be said to
have n parts, while the assembled parts may not be said to
have n parts--they can only be said to be n in number. And if
both chariot and parts are real, then the chariot must be
either identical with or distinct from its parts. Those such
as the Pudgalavadins who claim that the whole is neither
identical with nor distinct from the parts are easily
convicted of logical incoherence.[22] There still remains the
possibility that the chariot is real while its parts are
unreal. But the chariot is itself a part of a larger whole,
namely the universe; if only wholes are real, there can only
be one real thing. There would then arise seemingly
insuperable difficulties in trying to account for the
apparent utility of the myriad distinctions we routinely
draw. We must conclude that only impartite entities are
ultimately real; the partite is mentally constructed out of
those impartite entities that regularly co-occur in ways that
have a high degree of saliency for sentient systems like
ourselves.

This is the reasoning behind the development of the various
dharma theories of the Abhidharma schools. The dharmas are
those entities that are contained in our ultimate ontology;
and the argument tells us that they must be impartite. But
this does not result in the sort of simpleminded Democritean
atomism that one might expect. The universally agreed upon
definition of a dharma is: that which bears its own essential
nature (svabhava). The basic thinking here is that all
partite entities must borrow their essential properties from
parts: the characteristic shape of the chariot is a function
of the shapes of its parts, its utility as a means of
transport is a function of various interrelations among its
parts, and so forth. Thus, whatever does not borrow its
essential properties from other things, but instead bears
that nature as its own, must be ultimately real and not a
mere mental construction. This approach has some important
advantages. It gives a useful way to extend the
partite/impartite distinction beyond what is spatially
extended to those psychological events that make up the
so-called nama skandhas. It thus becomes possible to think of
a particular pain sensation, for instance, as impartite, and
a complex emotional episode like an occurrence of jealousy as
partite. It likewise opens the way to analyzing physical
objects not into aggregates of indivisible atoms with
determinate size, but rather into bundles of simple property
particulars.[23]

Most importantly though, the svabhava criterion of
dharma-hood yields a uniform way of telling whether something
is ultimately real or merely a mental construction: is it
analytically findable, or does it dissolve upon analysis?
Thus Buddhaghosha tells us that head hairs are mere
conceptual fictions since they can be analyzed into color,
shape, solidity, and smell; a particular smell, on the other
hand, presumably remains as the terminus of any analysis, and
is thus a dharma.[24] To borrow a Russellian distinction,
conceptual fictions can be known by description, whereas
dharmas can only be known by acquaintance. A description, of
course, utilizes a combination of two or more concepts. Thus
an element of aggregation must enter into the constitution of
that which is represented as one thing yet can be known by
description. And combination or aggregation is a mental
contribution; all that is ever actually given in experience
is co-occurrence. In pure acquaintance, on the other hand,
the mind is wholly passive and receptive to the given. That
which bears its own essential nature is free of all taint of
mental construction; that which borrows its essential nature
from what is other can only be a conceptual fiction, the
product of an inveterate tendency of the mind to construct
unreal aggregates.

This view depends, in the end, on the soundness of the
Abhidharma argument described above for the conclusion that
wholes are unreal. And some find that argument unpersuasive.
It should be recalled, though, that sorites arguments can
always be constructed to try to show that a given sort of
partite entity is conceptually constructed. And there is some
reason to believe that such attempts will generally be
successful. It is, for instance, relatively easy to show that
complex organisms cannot be ultimately real, given the many
indeterminacies that arise in connection with their continued
existence as their parts are replaced. There is also a
problem in supposing that a given organism has a determinate
species membership, given that our species concepts are
themselves fluid and responsive to a variety of pragmatic
pressures. Indeed, philosophical discussions of "natural
kind" terms tend to overlook the actual practice of
biologists, chemists, and physicists in their use of these
classificatory devices--not to describe unalterable
regularities in nature, but to express those idealizations of
observed tendencies that are necessary for purposes of theory
construction. The actual data of the natural sciences provide
fuel for countless sorites arguments against the ultimate
reality of the various entities posited by mature sciences.
Possible-worlds machinery yields another rich vein of sorites
arguments, since this allows us to construct spectra of
indeterminacy across worlds for any partite entity.

Thus there appears to be some reason for the Reductionist to
embrace the sort of ultimate ontology described in the
Abhidharma dharma theories. There is, though, one final step
in the development of Buddhist Reductionist thinking about
our ultimate ontology. This step seems not to have been taken
by any of the classical Abhidharma schools, but only appears
with the rise of Yogacara-Sautrantika in the work of Dinnaga.
It is the claim that what is ultimately real is just the
svalaksana, the ineffable pure particular. This radical
nominalism results from the realization that the same
considerations that militate against the existence of partite
entities apply with equal force to allegedly real universals
and resemblances. The relation between a universal and its
instances turns out to be just as problematic as that between
a whole and its parts. And the universal is just as much the
product of the mind's tendency to posit a one when it has
collected together the many. Of course the Buddhist
nominalist is faced with the daunting task of explaining the
efficacy of our discursive practices, given that at least
some of the terms of a language must be general.[25] But the
Yogacara-Sautrantika responds to this challenge with the
theory of apoha, according to which the meaning of "cow" is
"not non-cow." Here the "not" is to be read as exclusion
negation and the "non" as choice negation. The resulting
formal model, and the psychological machinery that is said to
instantiate the model, are together designed to explain how a
speaker can learn to use "cow" to refer to just those
particulars that other speakers of the language agree in
calling cows, given that there is no cowness universal
inhering in, or real resemblance shared by, those
particulars.

At this point, those with Reductionist inclinations might
begin to balk at the price being asked of them: an ultimate
ontology of ineffably unique pure particulars, a two-level
truth theory, and a radically nominalistic semantics. It must
be borne in mind, though, that Reductionism is espoused
because of its normative consequences, which require
ontological grounding. And the notion of ontological
grounding at work here is in turn based on metaphysical
realist presuppositions. Coming to know the truth about what
we are is supposed to change our views about how to live our
lives precisely because we take "truth" to be correspondence
to mind-independent reality. We will be inclined to transform
our habitual modes of conduct only to the extent that we find
them to stem from beliefs contaminated by elements of mental
construction--by styles of thought that leave the mind free
to project its desires onto the world. Reductionists must be
metaphysical realists, and as metaphysical realists must seek
to purge all elements of mental construction from their
ultimate ontology. Buddhist Reductionism is consistent and
complete Reductionism.

NOTES

An earlier version of this essay was read at the East-West
Philosophers' Conference of the Australasian Society for
Asian and Comparative Philosophy, held at Massey University,
Palmerston North, New Zealand, 12-15 August 1994. I wish to
thank the following for helpful comments on prior drafts:
David L. Anderson, Kenton Machina, Roy Perrett, C.
Ram-Prasad, and Liane Stillwell.

1.See Stone 1988 and Giles 1993. Parfit now calls the
Buddha's view "Eliminative Reductionism, " which he
distinguishes from his own view, now called "Constitutive
Reductionism" (1995, pp. 16f). For reasons that will become
evident below, although I once described the view of early
Buddhism and Abhidharma as a form of eliminative
reductionism, I now see this as a mistake.

2.Duerlinger (1993) represents both Vatsiputriya and
Prasangika Madhyamaka as holding Non-Reductionist
positions. His evident sympathy for Candrakirti's views
would seem to suggest that he takes at least the Prasangika
position on persons to be a defensible interpretation of
the Buddha's view.

3.I shall use "Reductionism" to refer to the view of persons
and personal identity developed in Parfit 1984, and
"reductionism" to refer to any theory that attempts to
reduce entities of one sort to entities of a distinct sort.
As shall soon become clear, not every reductionist about
persons is a Reductionist.

4.It might be thought that this commitment to the possibility
of a completely "im-mob-ish" description places a severe
constraint on the reductionist, in that such a description
will necessarily omit just those elements that distinguish
mob behavior from the behavior of individuals and small
groups. But this is not so. Once we know all the facts
about all the thoughts, feelings, and actions of each
individual, we will have captured all the ways in which the
actions of each affected the behavior of other individuals
in the mob.

5.At least not on any known version of the demonic-possession
theory. If, however, such a theory had posited distinct
types of demons for vital and bacterial pneumonia
independently of the development of the germ theory, we
might be more inclined to look for translation rules.

6.Noncircularity is obtained through the use of a "Ramsey
sentence" approach. See Shoemaker and Swinburne 1984, pp.
99 f.

7.Swinburne (Shoemaker and Swinburne 1984, p. 27) represents
all three authors as further-fact theorists. Noonan (1989,
p. 19) takes them as representative of what he calls the
Simple View, which is a form of the separately-existing-entity
thesis.

8.See Kathavatthu, pp. 1-71, Abhidharmakosabhasya 9, pp.
462463.

9.While the Milindapanha appears to have been written during
the period of Buddhist scholasticism (the Abhidharma
period), and thus not during that phase of the Buddhist
tradition commonly identified as the era of early Buddhism,
the text appears sufficiently free of commitment to
partisan Abhidharma scholastic positions as to seem better
classified as early Buddhist. The episode in question is at
11.2.1.

10.This is the most natural reading of the Pali, namely as
denying both numerical identity and numerical
distinctness, qua persons, of adult and infant. This and
similar passages are, however, sometimes translated more
simply by the phrase "neither the same nor different,"
which is then interpreted (e.g., by Collins [1982])as
involving an equivocation: neither qualitatively identical
nor numerically distinct. Such a reading misses the
connection between these passages and the Buddha's
treatment of the "indeterminate questions" (avyakrta),
which he compares to such questions as, when a fire has
gone out, in which of the four possible directions it has
gone. One may sensibly deny that the fire has gone in any
of these directions without equivocating on "gone to the
north, " and so forth. On the avyakrta as involving
bivalence failure, see Ruegg 1977.

11. The notion of full translatability is somewhat
problematic, since the full expression, in the privileged
discourse, of certain conventional truths would seem to
require strings of indefinite if not infinite length, and
it is far from clear that such a string could count as a
translation. In defense of the claim of full
translatability, it could be said that our inability to
produce or comprehend indefinitely long strings is a
"mere practical difficulty, " that is, only reflects
human limitations and so does not represent a shortcoming
in the analysis itself.

12. It might appear unduly harsh to call all statements
employing convenient designations false, rather than just
potentially misleading. The Buddhist reductionist agrees
that such convenient designations as "house" and "forest"
are relatively innocuous. Our use of the term "person,"
though, regularly leads to the most dire consequences.
Parfit makes a similar point about the term: "Though we
need concepts to discuss reality, we sometimes confuse
the two. We mistake conceptual facts for facts about
reality. And in the case of certain concepts, those that
are most loaded with emotional or moral significance, we
can be led seriously astray. Of these loaded concepts,
that of our own identity is, perhaps, the most
misleading" (1995, p. 45). Parfit merely refrains from
taking the next step: declaring all employment of the
concept useful but false.

13. Of course the Eliminativist cannot claim that it is
conventionally false that there are persons, since the
theory of persons is, as things now stand, accepted by
common sense. Eliminativists can only advocate that the
"folk theory" of persons be excised from our common
sense, and that our language be correspondingly revised.
They do so on the grounds that acceptance of the theory
leads to beliefs, desires, and actions that presuppose we
are something extra; since we are not something extra,
practical reason informed by this theory will inevitably
encounter insuperable difficulties. See Stone 1988, pp.
525-530.

14. This reluctance might be another factor that has led some
to see sufficiently great differences between Parfit's
Reductionism and the early Buddhist position as to
classify the latter as Eliminativist. Early Buddhism (as
well as Abhidharma) seems quite prepared to say what
should go in our ultimate ontology--and that persons are
not to be found there. This is not tantamount to
Eliminativism, but the apparent contrast with Parfit's
view might have led some to see it as such.

15. It might be wondered whether, for example, the reduction
of organic chemistry to quantum mechanics actually
carries any normative weight. But the relative status of
the disciplines of physics and chemistry seem to derive
at least in part from the possibility of such reduction.
The prestige of physics appears to stem in some measure
from this notion that it approaches more closely to a
grasp of reality undistorted by the limitations imposed
by such factors as the scale of our sensory apparatus or
our ability to track only so many particulars at once.

16. Here is further evidence that the Extreme view E (see
below) is inconsistent with Reductionism: the Extremist
seems to hold that while I have no reason for concern
about my future, I do have reason for concern about my
present state. Presumably this is because that future
person will not be me, but the present person is me. But
it is ultimately false that that future person is not me
and that the present person is me. Ultimately there are
no persons. If there are any facts that explain the
rationality of my concern for persons, these facts are
only visible at the conventional level of truth.

17. Parfit's view may have changed. He writes (1986, pp.
836-837) that while it is, on the Reductionist view,
irrational to view personal identity as what matters, it
is rational to care about the causal connections
obtaining between, say, a person and that person's
Replica. This would appear to rule out E as well as S. He
does not, however, give any argument against E.

18. Stone's (1988) argument for Eliminativism depends
crucially on the claim that such attitudes are incoherent
in the absence of the further fact that Non-Reductionism
requires, hence that our theory of persons is incoherent
given that there is no such further fact.

19. For the Buddhist Reductionist the utility of the
conceptual fiction of persons is, in the final analysis,
soteriological: it is because we think of ourselves as
persons that we seek to avoid suffering and thus are
disposed to enter the Path to nirvana. See Collins 1982,
p. 152. In this respect Buddhist Reductionism resembles
those Idian Non-Reductionist systems (such as Samkhya and
Advaita Vedanta) that take liberation from suffering as
the primary goal. Such systems commonly posit knowledge
of self as necessary for liberation, but then worry that
the self, as subject of knowledge, cannot take itself as
object of knowledge. The paradox is resolved by claiming
that the self comes to know itself indirectly, namely by
overcoming the error of misidentifying with some
inappropriate category. Ignorance thus plays a crucial
role in the soteriologies of these systems as well.

20. See Ramaiah 1978, pp. 61-90.

21. So Horgan (1994), for instance, after arguing that
genuine vagueness is logically incoherent and hence
impossible, seeks to reconcile this with the evidence for
widespread vagueness in the empirical world, by claiming
that while vagueness is impossible in mind-independent
reality, it actually occurs in thought and language,
where it possesses some utility and where the deleterious
effects of its logical incoherence may be contained
through various insulating strategies. This position
seems to stem from a view that has long been advanced by
Kenton Machina.

22. See Katthavatthu, pp. 3-71. Also see Ruegg 1977, pp.
34-36.

23. This would appear to have been the tack taken by certain
Sautrantikas in their account of the rupa dharmas. See
Vijnaptimatratasiddhi 14, where the opponent is a realist
who rejects atomism yet holds that there are rupa
dharmas. Vasubandhu's arguments against this position in
verse 15 make clear that the view in question analyzes
physical objects into bundles of property-particulars
devoid of substrate.

24. Visuddhimagga XI.88.

25. Indeed given that the svalaksana or pure particular is
radically momentary, all the terms of a language will
have to be general. For then a svalaksana is simply too
evanescent to be dubbed. What we actually succeed in
referring to through, for example, the use of a
demonstrative, is always a series of "resembling"
particulars. There can be no "logically proper names."

REFERENCES

Abhidharmakosabhasyam of Vasubandhu. 1975. Edited by Prahlad
Pradhan. Patna: K. P. Jayaswal Research Institute.

Brennan, Andrew. 1988. Conditions of Identity: A Study of
Identity and Survival New York: Oxford University Press.

Collins, Steven. 1982. Selfless Persons. Cambridge: Cambridge
University Press.

Duerlinger, James. 1993. "Reductionist and Nonreductionist
Theories of Persons in Indian Buddhist Philosophy." Journal
of Indian Philosophy 21: 79-101.

Giles, James. 1993. "The No-Self Theory: Hume, Buddhism, and
Personal Identity." Philosophy East and West 43: 175-200.

Horgan, Terrence. 1994. "Robust Vagueness and the
Forced-March Sorites Paradox." Philosophical Perspectives 8,
edited by James E. Tomberlin (Atascadero, California:
Ridgeview): 159-188.

Kathavatthu. 1961. Edited by Kassapa, Bhikkhu Jagadosa. Bihar
Government: Pall Publication Board. Translated as Points of
Controversy by T. W. Rhys Davids. London: Pall Text Society,
1915.

Milindapanho. 1972. Edited by R. D. Vadekar. Bombay: Bombay
University Publications. Translated as The Questions of King
Milinda by T. W. Rhys Davids. Sacred Books of the East, vols.
35, 36. Oxford: Oxford University Press, 1890.

Noonan, Harold. 1989. Personal Identity. London: Routledge.

Parfit, Derek. 1984. Reasons and Persons. New York: Oxford
University Press.

-----. 1986. "Comments." Ethics 96: 832-872.

-----. 1995. "The Unimportance of Identity." In Identity,
edited by Henry Harris, pp. 13-45. Oxford: Oxford
University Press.

Ramaiah, C. 1978. The Problem of Change and Identity in
Indian Philosophy. Tirupati: Sri Venkateswara University
Press.

Ruegg, D. Seyfort. 1977. "The Uses of the Four Positions of
the Catuskoti and the Problem of the Description of Reality
in Mahayana Buddhism." Journal of Indian Philosophy 5: 1-71.

Shoemaker, Sydney, and Richard Swinburne. 1984. Personal
Identity. Oxford: Basil Blackwell.

Siderits, Mark. 1987. "Beyond Compatibilism: A Buddhist
Approach to Freedom and Determinism." American Philosophical
Quarterly 24: 149-159.

Stone, Jim. 1988. "Parfit and the Buddha: Why There Are No
People." Philosophy and Phenomenological Research 48:519-532.

Unger, Peter. 1979. "Why There Are No People." Midwest
Studies in Philosophy 4, edited by Peter French, Theodore
Uehling, and Howard Wettstein (Minneapolis: University of
Minnesota Press): 177-222.

Vijnaptimatratasiddhi of Vasubandhu. 1980. Edited and
translated by K. N. Chatterjee. Varanasi: Kishor Vidya
Niketan.

Visuddhimagga of Buddhaghosha. 1950. Edited by Kosambi and
Warren. Harvard Oriental Series, vol. 41. Cambridge: Harvard
University Press. Translated as The Path of Purification by
Bhikkhu Nanamoli. Berkeley: Shambala, 1976.