Calling all analytical Buddhists and other metaphysicians interested in reductionism, wholes and parts, and cognate topics. This is a draft nearing completion, and I would like to solicit some comments and criticisms before submitting it to a journal. Who knows, you may convince me that it is not worthy of (hard) publication! What's in it for you? Well, you may learn something or else teach me something; you earn a free Mexican dinner complete with adequate cerveza con tequila enabling if ever our paths should cross; you will be acknowledged when and if this draft appears in a philosophical periodical.
1. Introduction
A reductionist about nonpersons, whether artifacts or naturally occurrent things, need not be a reductionist about persons. But the two forms of reductionism are often jointly asserted. Employing the terminology of Mark Siderits, a Buddhist reductionist is a Buddhist who is a reductionist about both persons and nonpersons. (Personal Identity and Buddhist Philosophy, p. 7, note 9. Reviews here and here.) The aim of this paper is to lend support to my contention that Buddhist reductionism (BR) is false. My concern will be much more with the reductionism of BR than with the Buddhism of BR. Since BR is a conjunctive claim, refutation of one of its conjuncts suffices to refute it. Here I do not discuss persons, except in passing, but attempt to refute reductionism about nonpersons. Since Siderits’ book is a rigorous, state-of-the-art discussion of the issues, I will advance my case via a discussion of some of his theses and arguments.
2. Reductionism, Non-Reductionism, and Eliminativism
At issue is ontological reductionism. This being noted, I will henceforth omit the qualifier ‘ontological.’ Consider a partite entity, any entity that is composed of parts, a bicycle for example, or one of its proper parts such as the chain. The reductionist holds that the existence of a partite entity E is nothing more than the existence of the more basic entities of which E is composed. The reductionist thus reduces wholes to their parts. ‘Parts’ is elliptical for ‘proper parts’ unless otherwise noted. Reductionism implies that in ultimate reality there are no such partite entities as bicycles or pools of water or bolts of lightning. A pool of water is just a collection of H2O molecules; but since each molecule is itself a partite entity, each molecule reduces to its constituent atoms, and so on, until we arrive (if we do arrive) at impartite entities that alone would make up ultimate reality. This is the essence of ontological analysis: we start with complex wholes, resolve them into their parts, and ultimately into their simple (not further
analyzable) parts. Such analysis is guided by a conception of ultimate reality according to which the ultimately real is the ultimately simple. Siderits makes what I take to be an equivalent point when he writes, “...only that which bears its own essential nature (that is, has an intrinsic nature)is ultimately real.” (96) Later we will have to ask how this ‘self-nature’ criterion of reality (X is real if and only if X has self-nature) comports with another relevant criterion, namely, the ‘causal’ criterion: X is real if and only if X is causally active/passive.
Reductionism is a ‘middle path’ between non-reductionism and eliminativism. The non-reductionist holds that things like bicycles are irreducibly real, and that their having parts has no tendency to show that they reduce to these parts. The eliminativist, on the other hand, holds that not only are partite things not ultimately real, they are not real in any sense. Eliminativism about every partite entity is of course absurd. What is not absurd is eliminativism relativized to a kind K of partite entities. For some values of ‘K,’ eliminativism is reasonable. Thus it is reasonable to be an eliminativist about flying horses, demons, caloric, phlogiston, and the ether that the Michaelson-Morley experiment failed to detect. The reductionist about Ks thus occupies a middle position between the non-reductionist, for whom Ks are irreducibly real, and the eliminativist, for whom Ks have no reality whatsoever. For the reductionist, the reality of Ks reduces to the reality of their parts.
Applying this to persons, we can say that, for the reductionist, the reality of persons consists in their being composed of impersonal or subpersonal psychophysical constituents such as the five skandhas. Thus there are persons, contrary to the eliminativist claim, but they are not irreducibly or ultimately real, contrary to the non-reductionist claim. In Buddhist terms, persons are empty or devoid of selfhood. There are persons, but they are selfless, anatta (Pali), anatman (Sanskrit).
3. The Conventionally Real and the Ultimately Real
Despite the apparently natural distinction between reductionism and eliminativism, one may wonder whether it can stand up to close scrutiny: Does not reductionism ultimately collapse into eliminativism? If X reduces to Y, or to some collection of sub-X items (a, b, c . . . ), how does that differ from saying that X does not exist? Suppose that a person is just a collection or bundle
of impersonal constituents such as thoughts, feelings, and so on. For the reductionist, the existence of the person is nothing other than the existence of these impersonal constituents. But this seems to imply that the person as person, the person as the diachronic and/or synchronic unity of these constituents, does not exist. Why then doesn’t reductionism about persons amount to eliminativism about persons? If one person reduces to many nonpersons or subpersons, many
psychophysical constituents, then it would seem that the one person does not exist, and the reduction is in effect an elimination. To put it another way, if persons qua persons are ultimately unreal – since they reduce to nonpersons – then what is the difference between a person such as you, dear reader, and Santa Claus or Pegasus? Clearly, human persons have a sort of reality that flying horses lack. To solve this problem, Siderits invokes the old Buddhist distinction between
what is conventionally real and what is ultimately real. (7) Partite entities such as chariots and persons, though not ultimately real, are conventionally real. This yields three categories: (i) the
ultimately real, (ii) the conventionally real which is ultimately unreal, and (iii) the ultimately unreal. Chariots and persons are on this account useful fictions. The difference between a useful
fiction such as myself and a fiction such as Santa Claus is that in the case of the former there are certain impersonal entities that undergird the fiction, while in the latter case there are not. To put
it another way, a chariot is an aggregate mentally constructed out of sub-chariot entities, while in the case of Santa Claus there are no sub-Santa Claus entities to be mentally aggregated.
If “all aggregation involves mental construction,” (Siderits 7) then we may have a way of distinguishing an assembled, and thus functional, chariot from the corresponding set or sum of disassembled chariot parts. There is clearly a difference here that any adequate theory must be able to account for. This is obvious from the fact that the parts can exist without the chariot existing, but not vice versa. Hence there is more to the chariot than its parts: there is the
connectedness of the parts, which, Siderits seems to be suggesting, derives from mental construction. We shall return to this a bit farther down.
4. An Argument for Reductionism
So much for explaining the reductionist thesis, and its difference both from non-reductionism and from eliminativism. The next step is to examine the grounds for it. One way to argue for reductionism is by eliminating its competitors. The following is my paraphrase of Siderits’ argument. Consider:
1. A whole and its parts are equally real in one of two ways. Either (1a) the whole is identical to its parts, or (1b) the whole is distinct from its parts.
2. A whole and its parts are equally unreal.
3. A whole is real, while its parts are unreal.
4. A whole is unreal, while its parts are real.
We may exclude the nihilistic thesis (2) right away inasmuch as it implies, contrary to fact, that nothing exists. We may also eliminate (3) due to its monistic consequences. For if only wholes are real, and no parts of wholes, then any whole that is a part of a wider whole is also unreal, whence it follows that there is only one whole, the world-whole, and all distinctions go by the board. If this strikes the reader as too quick, it doesn’t matter for the purposes of this paper since I am not disagreeing with Siderits on this point. Siderits and I agree in rejecting (3). As for (1a), it too is eliminable on the ground that there are properties that wholes and their parts cannot share. Thus each whole is one, while its parts are many. (1b) is also easily eliminated. A whole is not something above and beyond its parts as it would have to be if it were wholly distinct from its parts. A chariot, for example, is not something entirely distinct from wheels, axle, etc. After all, it is composed of them, and indeed exhausted by them taken collectively. There is no empirical
evidence of a chariot apart from empirical evidence of wheels, axle, etc. When one looks at a chariot, one does not see something distinct from wheels, axle, etc. If a chariot has n gross parts, it seems strange indeed to conclude that the chariot itself is an n + 1 th entity. (A gross part is a part uncovered at the first stage of analysis. Thus a wheel is a gross part despite the fact that it
has parts, and its parts have parts, etc.)
Allow me to expand on this point using a simpler example. My smoking pipe has two gross parts, a bowl and a stem. It sounds distinctly odd to say that these add up to three entities: bowl, stem, pipe, all equally real. And if I charge $100 for the bowl, and $5 for the stem, you would consider it either a bad joke or a moral outrage were I to charge you $210 for the assembled pipe: $100 + $5 + $105. But why isn’t that the fair price if you are purchasing three entities all equally real?
Having excluded (1a), (1b), (2), and (3) using arguments that I accept, Siderits concludes that the reductionist thesis (4) is true. In ultimate reality, there are no wholes; there are only simples that
may or may not come to form wholes. “Strictly speaking, there are no chariots...” (77) or anything else partite either. Nevertheless, partite entities can be said to be conventionally real as explained above. Since partite entities are accorded some reality – conventional reality – the absurdity of eliminativism about all partite entities is avoided.
5. Refutation of the Foregoing Argument by Elimination
There is, however, a third way one might construe (1). Instead of being forced to choose between saying that (1a) a whole is identical to its parts or (1b) distinct from them, one might say,quite commonsensically, that (1c) a whole is its parts properly arranged, its parts connected in the right way. A chariot is neither one of its parts, nor the mere sum of its (disconnected) parts, nor something entirely distinct from its parts: a chariot is its parts so connected as to given them the functionality of a chariot. To take a simpler example, suppose whole W has two proper parts, a and b. Then we ought to agree that W is not a; W is not b; W is not the set {a, b} or the
mereological sum (a + b); and W is not something entirely distinct from a and b. What W is is Rab, where R is the relation that connects a and b . (For example, a smoking pipe is not its bowl,its stem, the set or sum consisting of the two, or something wholly distinct from bowl and stem; a pipe is a bowl with its stem properly inserted into it.) If so, (1), construed as (1c), cannot be eliminated, and the reductionist thesis (4) remains unproven.
Some may object to the syntax of sentences like ‘This pipe is its parts properly connected.’ For ‘This pipe’ is a singular term while ‘its parts’ is plural, thus violating the syntactical requirement
that the blanks of the open sentence ‘___ is identical to —' be filled either by two singular terms or two plural terms. My response is that the above sentence does satisfy the requirement since ‘its parts properly connected’ is a singular term: it refers to exactly one entity, the same entity designated by ‘This pipe.’ In other words, I am not saying that a pipe is identical to its parts, but identical to its parts properly connected. Now let us think about this connectedness, which is not to be confused with the pipe itself.
The connectedness of a and b is not nothing; it is a reality that stands in the way of the reduction of W to its parts. Bear in mind that this reduction is supposed to show that, in ultimate reality,
there is no W, but only a and b. The fact of the matter, however, is that this connectedness is just as real as a and b. To appreciate that the connectedness is real, consider that it confers causal powers and causal liabilities upon the whole that are not possessed by the parts. An assembled chariot has the power to convey its driver into battle – a power its parts taken as a mere collection do not have.
The point I have just made presupposes the second criterion of reality mentioned above in section 2, namely, the ‘causal’ criterion according to which X is ultimately real if and only if X is causally active/passive. Does this criterion conflict with the ‘self-nature’ criterion? On the latter, X is ultimately real if and only if X has self-nature. This amounts to saying that the ultimately real is
the ultimately simple. This is because any contingent whole W is asymmetrically dependent on items (its parts) that could exist even if W does not exist – which is just to say that W lacks self-nature or own-being: its being is not intrinsic to it, but depends on factors that are what they are whether or not W exists. Now there need be no conflict between the two criteria of reality if everything that ultimately exists is simple. For there is nothing to prevent these simples from being causally active/passive. But if some wholes have causal powers not possessed by their parts, or causal powers not fully explainable by the causal powers of their parts, then the two criteria will diverge.
That would obviously cause trouble for the reductionist. The reductionist must therefore show that there are no wholes whose causal powers are not fully explainable in terms of the causal powers of their parts.
Siderits is aware of this problem and tackles it by saying that the relation R which establishes the connectedness of a and b “is decomposable into a set of properties...” of a and b “so that when all
the facts about the parts are taken into account, the effect of the chariot is wholly accounted for.” (77) One way of construing this idea is as follows. Relational facts like Rab decompose into
conjunctions of monadic facts, facts involving an individual’s exemplification of a nonrelational property. To give an example of my own, Al’s being heavier than Bill reduces to the conjunction
of Al weighs 210 lbs. & Bill weighs 200 lbs. In the case of a wheel’s being attached to an axle, there is a monadic property P1 of the wheel and a monadic property P2 of the axle such that the wheel’s being attached to the axle is the conjunctive fact, P1w & P2a. Let us call this doctrine foundationism. On foundationism, relations reduce to their monadic foundations. Accordingly, a relation is not a third thing ‘between’ its relata. This implies that there are no relations in ultimate reality; what there are are individuals exemplifying monadic properties, some of which are the foundations of their relatedness. On this foundationist scheme, the fact of relatedness can be explained without recourse to relations. The relatedness (connectedness) of a and b does not require anything that ‘refers beyond’ a and b. Thus the whole W = Rab reduces without remainder to a and b, when the latter are taken together with their monadic properties.
Even if foundationism is defensible, it is not clear that adopting it solves the problem. The problem is that (1c) must be eliminated if we are to arrive at (4). (1c) states that a whole is its parts properly arranged, connected in the right way. The whole W has a and b as parts related by R. Thus W = Rab. On Siderits’ view, Rab reduces to P1a & P2b. But a and b have the properties P1 and P2 respectively if and only if a and b are connected; they do not have these properties when a and b are disconnected. How then can it be said that W reduces to a and b? A reduction cannot be circular: if a whole reduces to its parts such that the whole is ultimately unreal and the parts ultimately real, then the parts cannot have monadic properties that entail that the parts are connected into a whole. Otherwise, the whole has not been reduced but presupposed. To reduce a whole to parts that cannot have the properties they have except as parts of that very whole is not to effect a reduction at all.
Equivalently, a reduction of a whole to its parts, if it is to be a genuine reduction which does away with the irreducible reality of the whole qua whole, must reduce the whole to its disconnected (unrelated) parts.
To put it still another way, if the parts have monadic properties that entail that they form a whole, then those parts are distinct from the parts to which the whole could be noncircularly reduced. Reduction is asymmetrical: if a whole reduces to its parts, then its parts do not reduce to the whole. But if the parts are taken together with the monadic foundations of their relatedness, then the parts depend on the whole to be what they are. We have a situation of symmetry rather than
asymmetry. But then the parts of a whole are as dependent on it, as it is on them, which implies that wholes and their parts are equally real.
Let us not forget that the wholes we are concerned with are contingent wholes. A contingent whole is one that may or may not exist given the existence of its parts: the existence of the parts does not entail the existence of the whole. An example of a noncontingent whole would be a set: the set {a, b} ‘automatically’ exists given the existence of a and b. Properly formulated, the reductionist thesis is that contingent wholes reduce to entities whose parthood (in a given whole, and in any whole) is extrinsic to them. Although it is essential to every contingent whole that it have some parts or other, if reductionism is true, then it is accidental to the parts of a whole that it be parts of any whole. From this one can see that reductionism faces a dilemma. If a whole W reduces to parts whose parthood is extrinsic to them, then the entities to which W reduces are what they are whether or not they are connected so as to form W. But this is to say that W
reduces to its disconnected parts. This, however, is absurd, since a whole is a unity, or connectedness, of parts. On the other hand, if a whole W reduces to parts whose parthood is intrinsic to them, then the asymmetry of reduction is violated, and there is no reduction.
Our interim conclusion is that contingent wholes are irreducible to their parts. For although a whole is dependent on its parts (in that it cannot exist unless they exist), a whole is more than its parts inasmuch as it is their connectedness. Now this connectedness can neither be a further connecting part – on pain of a Bradley-type regress – nor can it be reduced to monadic properties
of the parts, as we have just seen. Since contingent wholes are irreducible to their parts, the argument for reductionism presented in the preceding section is unsound. But this requires further support.
6. Refining my Refutation of the Argument by Elimination
The reductionist denies that wholes and their parts are equally ultimately real, maintaining instead that only the parts are ultimately real. But a whole is more than its parts, being their
connectedness, and this connectedness is in some sense real. As in some sense real, it simply must be accounted for. The Buddhist reductionist, as Siderits represents him, holds that wholes are
conventionally real, where this conventional reality derives from mental aggregation on our part. This way of accounting for connectedness will be discussed in the following section. What I want
to do now, however, is to consider in more detail Siderits’ rejection of the idea that wholes have non-conventional, or ultimate, reality. A reason to embrace this idea is that wholes have causal powers not possessed by their parts, together with the natural assumption that causal efficacy is a mark of reality. If reductionism is to be viable, this idea must be rejected.
Against the notion, defended by Jaegwon Kim, that the mereologically supervenient possesses distinctive causal powers, Siderits says this:
...it is of course true that water dissolves salt (something that neither hydrogen nor oxygen can do), just as it is true that the chariot (but not the wheels, not the axle, etc.) transports the driver. But the dissolving of the salt is nothing over and above
the ionization of sodium and chlorine, something accomplished by hydrogen and oxygen atoms when suitably related – just as the transporting of the driver just is the result of the lateral motion of the chariot body caused by forces transmitted from the axle, etc. (p. 93)
I fail to see how this refutes the idea that wholes have causal powers above and beyond those of their parts. Note that it is the hydrogen and oxygen atoms when suitably related that has the salt-dissolving power. It is not the hydrogen by itself, or the oxygen by itself, or the mere set of the two, but the two related so as to form a whole. But this is just to concede Kim’s point about the mereologically supervenient having distinctive causal powers. For when we consider the suitable relatedness of the hydrogen and oxygen atoms, we realize that we have a whole whose causal powers cannot be accounted for in terms of the causal powers of the parts when taken separately.
7. Can Mental Construction Account for Connectedness?
It is self-evident that a contingent whole such as a chariot is a connectedness of parts, and not a
mere collection of (disconnected) parts. But we have just seen that the ground of this connectedness cannot be internal to a whole either as a further part or as a set of monadic properties of its parts. Call the ground of connectedness the connector. If the connector is a
special part in addition to the primary parts, then the problem arises as to what connects the connector to what it connects. That way lies Bradley’s regress. If, on the other hand, we think of the connector R as reducing to monadic properties of the primary parts, then, as we saw in section 5, no genuine reduction is achieved. I conclude from this that the reductionist, to remain a reductionist and thus to avoid saying that a whole and its parts are equally real, must appeal to
something external to a contingent whole to account for the connectedness of parts that makes it a whole. There is need for an external unifier. Only with an external unifier or connector can a
reductionist remain a reductionist in the teeth of the arguments I have presented. It is clear that a reductionist cannot say that a whole unifies itself – in the sense that it is a unity distinct from its parts – for the simple reason that this amounts to abandoning reductionism.
Siderits appears to provide two competing accounts of the difference between a chariot and the corresponding collection of unassembled chariot parts. One of them -- the foundationist account -- we have just criticized. The other appeals to the notion that “all aggregation involves mental construction.” (7) Could mental aggregation be the ontological ground of connectedness? I
submit that this notion of mental construction as the ground of connectedness is untenable. Ask a simple question: Whose mind is doing the aggregating? Does the chariot-driver assemble his
chariot by ‘thinking together’ its parts, starting at ontological rock-bottom with impartite parts? Must he continuously ‘think them together’ to maintain the chariot’s functionality? That would
be absurd. The unity of parts that bestows upon them chariot-functionality is logically and ontologically antecedent to any act of mental constructing or aggregating by any individual. This is painfully obvious in the case of very small parts such as molecular and atomic parts. The charioteer can drive his chariot only because it is already (logically speaking) a full-fledged unity of parts. Furthermore, the mind that does the aggregating is itself a partite entity, and therefore one that cannot be ultimately real if Buddhist reductionism is true. If so, the aggregating mind M is itself in need of an aggregator to account for the difference between M and its parts. The appeal to partite minds as aggregators pretty clearly leads nowhere. Is there another option?
If it cannot be individual minds that do the aggregating, is it language that does it, or language together with social practices? Siderits points out that we have the name ‘chariot,’ but no name
for the corresponding collection of disassembled parts. (7) He says that this is because we have an institutionalized use – as a means of transportation – for the assembled parts, but no such use for the parts in their disassembled state. Siderits claims that it is our “institutionally arranged interests” (8) that bring it about that we view the chariot parts as a “single entity” when in ultimate reality there is no single entity. The suggestion is that what makes the chariot a single entity, a unity of parts having chariot functionality, is something social or institutional or linguistic.
I would say, however, that this puts the cart (or the chariot!) before the horse. It is because the chariot is a “single entity” that we can ride it and to name it ‘chariot.’ Granted, it has a name because of its human usefulness, but it is humanly useful because of the functionality that derives from the connectedness of its parts. The connectedness, therefore, cannot derive from our applying of ‘chariot’ to a bunch of otherwise disconnected chariot parts. The connectedness
whereby it is a single substantial entity is logically and ontologically prior to any mental act of constructing or any linguistic act of naming. This is not to say that the chariot is ultimately real; it is to say that the reality of the chariot is not a merely linguistically or mentally created reality. The chariot may not be ultimately real, but its reality is greater than that of any mentally constructed object.
8. Summary
There is no denying that reductionism is attractive. It allows us to avoid monism (wholes alone are real), nihilism (neither wholes nor parts are real), and dualism (both wholes and parts are real). Unfortunately, reductionism flies in the face of the obvious fact that wholes are in some sense real. This is why Siderits supplements straight reductionism -- which threatens to collapse into eliminativism -- with a doctrine of conventional reality as brought about by mental aggregation on our part. But conventional reality is not reality enough. The salt-dissolving power of water is a reality more substantial than mere conventional reality. Water has this power
independently of us, our institutions, our language, our interests, and any mental aggregating we may engage in.
Where does this leave us? Siderits’ tetralemma is exhaustive: we must choose among monism, nihilism, dualism, or reductionism, where reductionism, if it is not to collapse into eliminativism,
must be fortified by recourse to an external unifier, connector, aggregator. But this aggregator cannot be me or you or any partite mind. After all, any partite mind will itself require a unifier to
insure its unity. So if we wish to maintain reductionism, we must have recourse to an impartite mind, an ontologically simple mind. Or is that too much to swallow?