Special Sciences (Or: The Disunity of Science as a Working Hypothesis)

Jerry A. Fodor
Massachusetts Institute of Technology

Original publication: Jerry A. Fodor, “Special Sciences (Or: The Disunity of Science as a Working Hypothesis),” Synthese 28, no. 2 (October 1974): 97–115. DOI · Springer · JSTOR · PhilPapers · archived copy

Author’s note: I wish to express my gratitude to Ned Block for having read a version of this paper and for the very useful comments he made.

Web-edition note: The text has been proofed against the supplied journal scan. The numbered formulae are reset in LaTeX, and schema (5) is redrawn as a responsive vector figure. Obvious printer and OCR errors are silently corrected; Fodor’s period spelling “guaranty” is retained. A notational inconsistency on p. 111 is normalized to match schema (5) and formula (6).

---

A typical thesis of positivistic philosophy of science is that all true theories in the special sciences should reduce to physical theories in the long run. This is intended to be an empirical thesis, and part of the evidence which supports it is provided by such scientific successes as the molecular theory of heat and the physical explanation of the chemical bond. But the philosophical popularity of the reductivist program cannot be explained by reference to these achievements alone. The development of science has witnessed the proliferation of specialized disciplines at least as often as it has witnessed their reduction to physics, so the widespread enthusiasm for reduction can hardly be a mere induction over its past successes.

I think that many philosophers who accept reductivism do so primarily because they wish to endorse the generality of physics vis-à-vis the special sciences: roughly, the view that all events which fall under the laws of any science are physical events and hence fall under the laws of physics.¹ For such philosophers, saying that physics is basic science and saying that theories in the special sciences must reduce to physical theories have seemed to be two ways of saying the same thing, so that the latter doctrine has come to be a standard construal of the former.

In what follows, I shall argue that this is a considerable confusion. What has traditionally been called ‘the unity of science’ is a much stronger, and much less plausible, thesis than the generality of physics. If this is true it is important. Though reductionism is an empirical doctrine, it is intended to play a regulative role in scientific practice. Reducibility to physics is taken to be a constraint upon the acceptability of theories in the special sciences, with the curious consequence that the more the special sciences succeed, the more they ought to disappear. Methodological problems about psychology, in particular, arise in just this way: the assumption that the subject-matter of psychology is part of the subject-matter of physics is taken to imply that psychological theories must reduce to physical theories, and it is this latter principle that makes the trouble. I want to avoid the trouble by challenging the inference.

I

Reductivism is the view that all the special sciences reduce to physics. The sense of ‘reduce to’ is, however, proprietary. It can be characterized as follows.²

Let

$$S_1x \to S_2x \tag{1}$$

be a law of the special science $S$. ((1) is intended to be read as something like “all $S_1$ situations bring about $S_2$ situations.” I assume that a science is individuated largely by reference to its typical predicates, hence that if $S$ is a special science, $S_1$ and $S_2$ are not predicates of basic physics. I also assume that the ‘all’ which quantifies laws of the special sciences needs to be taken with a grain of salt; such laws are typically not exceptionless. This is a point to which I shall return at length.) A necessary and sufficient condition of the reduction of (1) to a law of physics is that the formulae (2) and (3) be laws, and a necessary and sufficient condition of the reduction of $S$ to physics is that all its laws be so reducible.³

$$S_1x \leftrightarrow P_1x \tag{2a}$$ $$S_2x \leftrightarrow P_2x \tag{2b}$$ $$P_1x \to P_2x \tag{3}$$

$P_1$ and $P_2$ are supposed to be predicates of physics, and (3) is supposed to be a physical law. Formulae like (2) are often called ‘bridge’ laws. Their characteristic feature is that they contain predicates of both the reduced and the reducing science. Bridge laws like (2) are thus contrasted with ‘proper’ laws like (1) and (3). The upshot of the remarks so far is that the reduction of a science requires that any formula which appears as the antecedent or consequent of one of its proper laws must appear as the reduced formula in some bridge law or other.⁴

Several points about the connective $\to$ are in order. First, whatever other properties that connective may have, it is universally agreed that it must be transitive. This is important because it is usually assumed that the reduction of some of the special sciences proceeds via bridge laws which connect their predicates with those of intermediate reducing theories. Thus, psychology is presumed to reduce to physics via, say, neurology, biochemistry, and other local stops. The present point is that this makes no difference to the logic of the situation so long as the transitivity of $\to$ is assumed. Bridge laws which connect the predicates of $S$ to those of $S^*$ will satisfy the constraints upon the reduction of $S$ to physics so long as there are other bridge laws which, directly or indirectly, connect the predicates of $S^*$ to physical predicates.

There are, however, quite serious open questions about the interpretations of $\to$ in bridge laws. What turns on these questions is the respect in which reductivism is taken to be a physicalist thesis.

To begin with, if we read $\to$ as ‘brings about’ or ‘causes’ in proper laws, we will have to have some other connective for bridge laws, since bringing about and causing are presumably asymmetric, while bridge laws express symmetric relations. Moreover, if $\to$ in bridge laws is interpreted as any relation other than identity, the truth of reductivism will only guaranty the truth of a weak version of physicalism, and this would fail to express the underlying ontological bias of the reductivist program.

If bridge laws are not identity statements, then formulae like (2) claim at most that, by law, $x$’s satisfaction of a $P$ predicate and $x$’s satisfaction of an $S$ predicate are causally correlated. It follows from this that it is nomologically necessary that $S$ and $P$ predicates apply to the same things (i.e., that $S$ predicates apply to a subset of the things that $P$ predicates apply to). But, of course, this is compatible with a non-physicalist ontology since it is compatible with the possibility that $x$’s satisfying $S$ should not itself be a physical event. On this interpretation, the truth of reductivism does not guaranty the generality of physics vis-à-vis the special sciences since there are some events (satisfactions of $S$ predicates) which fall in the domains of a special science ($S$) but not in the domain of physics. (One could imagine, for example, a doctrine according to which physical and psychological predicates are both held to apply to organisms, but where it is denied that the event which consists of an organism’s satisfying a psychological predicate is, in any sense, a physical event. The up-shot would be a kind of psychophysical dualism of a non-Cartesian variety; a dualism of events and/or properties rather than substances.)

Given these sorts of considerations, many philosophers have held that bridge laws like (2) ought to be taken to express contingent event identities, so that one would read (2a) in some such fashion as “every event which consists of $x$’s satisfying $S_1$ is identical to some event which consists of $x$’s satisfying $P_1$ and vice versa.” On this reading, the truth of reductivism would entail that every event that falls under any scientific law is a physical event, thereby simultaneously expressing the ontological bias of reductivism and guaranteeing the generality of physics vis-à-vis the special sciences.

If the bridge laws express event identities, and if every event that falls under the proper laws of a special science falls under a bridge law, we get the truth of a doctrine that I shall call ‘token physicalism’. Token physicalism is simply the claim that all the events that the sciences talk about are physical events. There are three things to notice about token physicalism.

First, it is weaker than what is usually called ‘materialism’. Materialism claims both that token physicalism is true and that every event falls under the laws of some science or other. One could therefore be a token physicalist without being a materialist, though I don’t see why anyone would bother.

Second, token physicalism is weaker than what might be called ‘type physicalism’, the doctrine, roughly, that every property mentioned in the laws of any science is a physical property. Token physicalism does not entail type physicalism because the contingent identity of a pair of events presumably does not guarantee the identity of the properties whose instantiation constitutes the events; not even where the event identity is nomologically necessary. On the other hand, if every event is the instantiation of a property, then type physicalism does entail token physicalism: two events will be identical when they consist of the instantiation of the same property by the same individual at the same time.

Third, token physicalism is weaker than reductivism. Since this point is, in a certain sense, the burden of the argument to follow, I shan’t labour it here. But, as a first approximation, reductivism is the conjunction of token physicalism with the assumption that there are natural kind predicates in an ideally completed physics which correspond to each natural kind predicate in any ideally completed special science. It will be one of my morals that the truth of reductivism cannot be inferred from the assumption that token physicalism is true. Reductivism is a sufficient, but not a necessary, condition for token physicalism.

In what follows, I shall assume a reading of reductivism which entails token physicalism. Bridge laws thus state nomologically necessary contingent event identities, and a reduction of psychology to neurology would entail that any event which consists of the instantiation of a psychological property is identical with some event which consists of the instantiation of some neurological property.

Where we have got to is this: reductivism entails the generality of physics in at least the sense that any event which falls within the universe of discourse of a special science will also fall within the universe of discourse of physics. Moreover, any prediction which follows from the laws of a special science and a statement of initial conditions will also follow from a theory which consists of physics and the bridge laws, together with the statement of initial conditions. Finally, since ‘reduces to’ is supposed to be an asymmetric relation, it will also turn out that physics is the basic science; that is, if reductivism is true, physics is the only science that is general in the sense just specified. I now want to argue that reductivism is too strong a constraint upon the unity of science, but that the relevantly weaker doctrine will preserve the desired consequences of reductivism: token physicalism, the generality of physics, and its basic position among the sciences.

II

Every science implies a taxonomy of the events in its universe of discourse. In particular, every science employs a descriptive vocabulary of theoretical and observation predicates such that events fall under the laws of the science by virtue of satisfying those predicates. Patently, not every true description of an event is a description in such a vocabulary. For example, there are a large number of events which consist of things having been transported to a distance of less than three miles from the Eiffel Tower. I take it, however, that there is no science which contains ‘is transported to a distance of less than three miles from the Eiffel Tower’ as part of its descriptive vocabulary. Equivalently, I take it that there is no natural law which applies to events in virtue of their being instantiations of the property is transported to a distance of less than three miles from the Eiffel Tower (though I suppose it is conceivable that there is some law that applies to events in virtue of their being instantiations of some distinct but co-extensive property). By way of abbreviating these facts, I shall say that the property is transported … does not determine a natural kind, and that predicates which express that property are not natural kind predicates.

If I knew what a law is, and if I believed that scientific theories consist just of bodies of laws, then I could say that $P$ is a natural kind predicate relative to $S$ iff $S$ contains proper laws of the form $Px \to \alpha x$ or $\alpha x \to Px$; roughly, the natural kind predicates of a science are the ones whose terms are the bound variables in its proper laws. I am inclined to say this even in my present state of ignorance, accepting the consequence that it makes the murky notion of a natural kind viciously dependent on the equally murky notions law and theory. There is no firm footing here. If we disagree about what is a natural kind, we will probably also disagree about what is a law, and for the same reasons. I don’t know how to break out of this circle, but I think that there are interesting things to say about which circle we are in.

For example, we can now characterize the respect in which reductivism is too strong a construal of the doctrine of the unity of science. If reductivism is true, then every natural kind is, or is co-extensive with, a physical natural kind. (Every natural kind is a physical natural kind if bridge laws express property identities, and every natural kind is co-extensive with a physical natural kind if bridge laws express event identities.) This follows immediately from the reductivist premise that every predicate which appears as the antecedent or consequent of a law of the special sciences must appear as one of the reduced predicates in some bridge, together with the assumption that the natural kind predicates are the ones whose terms are the bound variables in proper laws. If, in short, some physical law is related to each law of a special science in the way that (3) is related to (1), then every natural kind predicate of a special science is related to a natural kind predicate of physics in the way that (2) relates $S_1$ and $S_2$ to $P_1$ and $P_2$.

I now want to suggest some reasons for believing that this consequence of reductivism is intolerable. These are not supposed to be knock-down reasons; they couldn’t be, given that the question whether reductivism is too strong is finally an empirical question. (The world could turn out to be such that every natural kind corresponds to a physical natural kind, just as it could turn out to be such that the property is transported to a distance of less than three miles from the Eiffel Tower determines a natural kind in, say, hydrodynamics. It’s just that, as things stand, it seems very unlikely that the world will turn out to be either of these ways.)

The reason it is unlikely that every natural kind corresponds to a physical natural kind is just that (a) interesting generalizations (e.g., counter-factual supporting generalizations) can often be made about events whose physical descriptions have nothing in common, (b) it is often the case that whether the physical descriptions of the events subsumed by these generalizations have anything in common is, in an obvious sense, entirely irrelevant to the truth of the generalizations, or to their interestingness, or to their degree of confirmation or, indeed, to any of their epistemologically important properties, and (c) the special sciences are very much in the business of making generalizations of this kind.

I take it that these remarks are obvious to the point of self-certification; they leap to the eye as soon as one makes the (apparently radical) move of taking the special sciences at all seriously. Suppose, for example, that Gresham’s ‘law’ really is true. (If one doesn’t like Gresham’s law, then any true generalization of any conceivable future economics will probably do as well.) Gresham’s law says something about what will happen in monetary exchanges under certain conditions. I am willing to believe that physics is general in the sense that it implies that any event which consists of a monetary exchange (hence any event which falls under Gresham’s law) has a true description in the vocabulary of physics and in virtue of which it falls under the laws of physics. But banal considerations suggest that a description which covers all such events must be wildly disjunctive. Some monetary exchanges involve strings of wampum. Some involve dollar bills. And some involve signing one’s name to a check. What are the chances that a disjunction of physical predicates which covers all these events (i.e., a disjunctive predicate which can form the right hand side of a bridge law of the form “$x$ is a monetary exchange $\leftrightarrow \ldots$”) expresses a physical natural kind? In particular, what are the chances that such a predicate forms the antecedent or consequent of some proper law of physics? The point is that monetary exchanges have interesting things in common; Gresham’s law, if true, says what one of these interesting things is. But what is interesting about monetary exchanges is surely not their commonalities under physical description. A natural kind like a monetary exchange could turn out to be co-extensive with a physical natural kind; but if it did, that would be an accident on a cosmic scale.

In fact, the situation for reductivism is still worse than the discussion thus far suggests. For, reductivism claims not only that all natural kinds are co-extensive with physical natural kinds, but that the co-extensions are nomologically necessary: bridge laws are laws. So, if Gresham’s law is true, it follows that there is a (bridge) law of nature such that “$x$ is a monetary exchange $\leftrightarrow x$ is $P$,” where $P$ is a term for a physical natural kind. But, surely, there is no such law. If there were, then $P$ would have to cover not only all the systems of monetary exchange that there are, but also all the systems of monetary exchange that there could be; a law must succeed with the counterfactuals. What physical predicate is a candidate for $P$ in “$x$ is a nomologically possible monetary exchange iff $Px$”?

To summarize: an immortal econophysicist might, when the whole show is over, find a predicate in physics that was, in brute fact, co-extensive with ‘is a monetary exchange’. If physics is general - if the ontological biases of reductivism are true - then there must be such a predicate. But (a) to paraphrase a remark Donald Davidson made in a slightly different context, nothing but brute enumeration could convince us of this brute co-extensivity, and (b) there would seem to be no chance at all that the physical predicate employed in stating the co-extensivity is a natural kind term, and (c) there is still less chance that the co-extension would be lawful (i.e., that it would hold not only for the nomologically possible world that turned out to be real, but for any nomologically possible world at all).

I take it that the preceding discussion strongly suggests that economics is not reducible to physics in the proprietary sense of reduction involved in claims for the unity of science. There is, I suspect, nothing special about economics in this respect; the reasons why economics is unlikely to reduce to physics are paralleled by those which suggest that psychology is unlikely to reduce to neurology.

If psychology is reducible to neurology, then for every psychological natural kind predicate there is a co-extensive neurological natural kind predicate, and the generalization which states this co-extension is a law. Clearly, many psychologists believe something of the sort. There are departments of ‘psycho-biology’ or ‘psychology and brain science’ in universities throughout the world whose very existence is an institutionalized gamble that such lawful co-extensions can be found. Yet, as has been frequently remarked in recent discussions of materialism, there are good grounds for hedging these bets. There are no firm data for any but the grossest correspondence between types of psychological states and types of neurological states, and it is entirely possible that the nervous system of higher organisms characteristically achieves a given psychological end by a wide variety of neurological means. If so, then the attempt to pair neurological structures with psychological functions is foredoomed. Physiological psychologists of the stature of Karl Lashley have held precisely this view.

The present point is that the reductivist program in psychology is, in any event, not to be defended on ontological grounds. Even if (token) psychological events are (token) neurological events, it does not follow that the natural kind predicates of psychology are co-extensive with the natural kind predicates of any other discipline (including physics). That is, the assumption that every psychological event is a physical event does not guaranty that physics (or, a fortiori, any other discipline more general than psychology) can provide an appropriate vocabulary for psychological theories. I emphasize this point because I am convinced that the make-or-break commitment of many physiological psychologists to the reductivist program stems precisely from having confused that program with (token) physicalism.

What I have been doubting is that there are neurological natural kinds co-extensive with psychological natural kinds. What seems increasingly clear is that, even if there is such a co-extension, it cannot be lawlike. For, it seems increasingly likely that there are nomologically possible systems other than organisms (namely, automata) which satisfy natural kind predicates in psychology, and which satisfy no neurological predicates at all. Now, as Putnam has emphasized, if there are any such systems, then there are probably vast numbers, since equivalent automata can be made out of practically anything. If this observation is correct, then there can be no serious hope that the class of automata whose psychology is effectively identical to that of some organism can be described by physical natural kind predicates (though, of course, if token physicalism is true, that class can be picked out by some physical predicate or other). The upshot is that the classical formulation of the unity of science is at the mercy of progress in the field of computer simulation. This is, of course, simply to say that that formulation was too strong. The unity of science was intended to be an empirical hypothesis, defeasible by possible scientific findings. But no one had it in mind that it should be defeated by Newell, Shaw and Simon.

I have thus far argued that psychological reductivism (the doctrine that every psychological natural kind is, or is co-extensive with, a neurological natural kind) is not equivalent to, and cannot be inferred from, token physicalism (the doctrine that every psychological event is a neurological event). It may, however, be argued that one might as well take the doctrines to be equivalent since the only possible evidence one could have for token physicalism would also be evidence for reductivism: namely, the discovery of type-to-type psychophysical correlations.

A moment’s consideration shows, however, that this argument is not well taken. If type-to-type psychophysical correlations would be evidence for token physicalism, so would correlations of other specifiable kinds.

We have type-to-type correlations where, for every $n$-tuple of events that are of the same psychological kind, there is a correlated $n$-tuple of events that are of the same neurological kind. Imagine a world in which such correlations are not forthcoming. What is found, instead, is that for every $n$-tuple of type identical psychological events, there is a spatiotemporally correlated $n$-tuple of type distinct neurological events. That is, every psychological event is paired with some neurological event or other, but psychological events of the same kind may be paired with neurological events of different kinds. My present point is that such pairings would provide as much support for token physicalism as type-to-type pairings do so long as we are able to show that the type distinct neurological events paired with a given kind of psychological event are identical in respect of whatever properties are relevant to type-identification in psychology. Suppose, for purposes of explication, that psychological events are type identified by reference to their behavioral consequences.⁵ Then what is required of all the neurological events paired with a class of type homogeneous psychological events is only that they be identical in respect of their behavioral consequences. To put it briefly, type identical events do not, of course, have all their properties in common, and type distinct events must nevertheless be identical in some of their properties. The empirical confirmation of token physicalism does not depend on showing that the neurological counterparts of type identical psychological events are themselves type identical. What needs to be shown is only that they are identical in respect of those properties which determine which kind of psychological event a given event is.

Could we have evidence that an otherwise heterogeneous set of neurological events have these kinds of properties in common? Of course we could. The neurological theory might itself explain why an $n$-tuple of neurologically type distinct events are identical in their behavioral consequences, or, indeed, in respect of any of indefinitely many other such relational properties. And, if the neurological theory failed to do so, some science more basic than neurology might succeed.

My point in all this is, once again, not that correlations between type homogeneous psychological states and type heterogeneous neurological states would prove that token physicalism is true. It is only that such correlations might give us as much reason to be token physicalists as type-to-type correlations would. If this is correct, then the epistemological arguments from token physicalism to reductivism must be wrong.

It seems to me (to put the point quite generally) that the classical construal of the unity of science has really misconstrued the goal of scientific reduction. The point of reduction is not primarily to find some natural kind predicate of physics co-extensive with each natural kind predicate of a reduced science. It is, rather, to explicate the physical mechanisms whereby events conform to the laws of the special sciences: I have been arguing that there is no logical or epistemological reason why success in the second of these projects should require success in the first, and that the two are likely to come apart in fact wherever the physical mechanisms whereby events conform to a law of the special sciences are heterogeneous.

III

I take it that the discussion thus far shows that reductivism is probably too strong a construal of the unity of science; on the one hand, it is incompatible with probable results in the special sciences, and, on the other, it is more than we need to assume if what we primarily want is just to be good token physicalists. In what follows, I shall try to sketch a liberalization of reductivism which seems to me to be just strong enough in these respects. I shall then give a couple of independent reasons for supposing that the revised doctrine may be the right one.

The problem all along has been that there is an open empirical possibility that what corresponds to the natural kind predicates of a reduced science may be a heterogeneous and unsystematic disjunction of predicates in the reducing science, and we do not want the unity of science to be prejudiced by this possibility. Suppose, then, that we allow that bridge statements may be of the form

$$Sx \leftrightarrow (P_1x \lor P_2x \lor \cdots \lor P_nx) \tag{4}$$

where $P_1 \lor P_2 \lor \cdots \lor P_n$ is not a natural kind predicate in the reducing science. I take it that this is tantamount to allowing that at least some ‘bridge laws’ may, in fact, not turn out to be laws, since I take it that a necessary condition on a universal generalization being lawlike is that the predicates which constitute its antecedent and consequent should pick out natural kinds. I am thus supposing that it is enough, for purposes of the unity of science, that every law of the special sciences should be reducible to physics by bridge statements which express true empirical generalizations. Bearing in mind that bridge statements are to be construed as a species of identity statements, (4) will be read as something like “every event which consists of $x$’s satisfying $S$ is identical with some event which consists of $x$’s satisfying some or other predicate belonging to the disjunction $P_1 \lor P_2 \lor \cdots \lor P_n$.”

Now, in cases of reduction where what corresponds to (2) is not a law, what corresponds to (3) will not be either, and for the same reason. Namely, the predicates appearing in the antecedent or consequent will, by hypothesis, not be natural kind predicates. Rather, what we will have is something that looks like (5) (see next page).

That is, the antecedent and consequent of the reduced law will each be connected with a disjunction of predicates in the reducing science, and, if the reduced law is exceptionless, there will be laws of the reducing science which connect the satisfaction of each member of the disjunction associated with the antecedent to the satisfaction of some member of the disjunction associated with the consequent. That is, if $S_1x \to S_2x$ is

exceptionless, then there must be some proper law of the reducing science which either states or entails that $P_1x \to P^*$ for some $P^*$, and similarly for $P_2x$ through $P_nx$. Since there must be such laws, it follows that each disjunct of $P_1 \lor P_2 \lor \cdots \lor P_n$ is a natural kind predicate, as is each disjunct of $P_1^* \lor P_2^* \lor \cdots \lor P_m^*$.

This, however, is where push comes to shove. For, it might be argued that if each disjunct of the $P$ disjunction is lawfully connected to some disjunct of the $P^*$ disjunction, it follows that (6) is itself a law.

$$(P_1x \lor P_2x \lor \cdots \lor P_nx) \to (P_1^*x \lor P_2^*x \lor \cdots \lor P_m^*x) \tag{6}$$

The point would be that (5) gives us $P_1x \to P_2^*x$, $P_2x \to P_m^*x$, etc., and the argument from a premise of the form $(P \supset R)$ and $(Q \supset S)$ to a conclusion of the form $(P \lor Q) \supset (R \lor S)$ is valid.

What I am inclined to say about this is that it just shows that “it’s a law that —” defines a non-truth-functional context (or, equivalently for these purposes, that not all truth functions of natural kind predicates are themselves natural kind predicates). In particular, that one may not argue from “it’s a law that $P$ brings about $R$” and “it’s a law that $Q$ brings about $S$” to “it’s a law that $P$ or $Q$ brings about $R$ or $S$.” (Though, of course, the argument from those premises to “$P$ or $Q$ brings about $R$ or $S$” simpliciter is fine.) I think, for example, that it is a law that the irradiation of green plants by sunlight causes carbohydrate synthesis, and I think that it is a law that friction causes heat, but I do not think that it is a law that (either the irradiation of green plants by sunlight or friction) causes (either carbohydrate synthesis or heat). Correspondingly, I doubt that “is either carbohydrate synthesis or heat” is plausibly taken to be a natural kind predicate.

It is not strictly mandatory that one should agree with all this, but one denies it at a price. In particular, if one allows the full range of truth functional arguments inside the context “it’s a law that —”, then one gives up the possibility of identifying the natural kind predicates of a science with those predicates which appear as the antecedents or the consequents of its proper laws. (Thus (6) would be a proper law of physics which fails to satisfy that condition.) One thus inherits the need for an alternative construal of the notion of a natural kind, and I don’t know what that alternative might be like.

The upshot seems to be this. If we do not require that bridge statements must be laws, then either some of the generalizations to which the laws of special sciences reduce are not themselves lawlike, or some laws are not formulable in terms of natural kinds. Whichever way one takes (5), the important point is that it is weaker than standard reductivism: it does not require correspondences between the natural kinds of the reduced and the reducing science. Yet it is physicalistic on the same assumption that makes standard reductivism physicalistic (namely, that the bridge statements express true token identities). But these are precisely the properties that we wanted a revised account of the unity of science to exhibit.

I now want to give two reasons for thinking that this construal of the unity of science is right. First, it allows us to see how the laws of the special sciences could reasonably have exceptions, and, second, it allows us to see why there are special sciences at all. These points in turn.

Consider, again, the model of reduction implicit in (2) and (3). I assume that the laws of basic science are strictly exceptionless, and I assume that it is common knowledge that the laws of the special sciences are not. But now we have a painful dilemma. Since $\to$ expresses a relation (or relations) which must be transitive, (1) can have exceptions only if the bridge laws do. But if the bridge laws have exceptions, reductivism loses its ontological bite, since we can no longer say that every event which consists of the instantiation of an $S$ predicate is identical with some event which consists of the instantiation of a $P$ predicate. In short, given the reductionist model, we cannot consistently assume that the bridge laws and the basic laws are exceptionless while assuming that the special laws are not. But we cannot accept the violation of the bridge laws unless we are willing to vitiate the ontological claim that is the main point of the reductivist program.

We can get out of this (salve the model) in one of two ways. We can give up the claim that the special laws have exceptions or we can give up the claim that the basic laws are exceptionless. I suggest that both alternatives are undesirable. The first because it flies in the face of fact. There is just no chance at all that the true, counter-factual supporting generalizations of, say, psychology, will turn out to hold in strictly each and every condition where their antecedents are satisfied. Even where the spirit is willing, the flesh is often weak. There are always going to be behavioral lapses which are physiologically explicable but which are uninteresting from the point of view of psychological theory. The second alternative is only slightly better. It may, after all, turn out that the laws of basic science have exceptions. But the question arises whether one wants the unity of science to depend upon the assumption that they do.

On the account summarized in (5), however, everything works out satisfactorily. A nomologically sufficient condition for an exception to $S_1x \to S_2x$ is that the bridge statements should identify some occurrence of the satisfaction of $S_1$ with an occurrence of the satisfaction of a $P$ predicate which is not itself lawfully connected to the satisfaction of any $P^*$ predicate. (I.e., suppose $S_1$ is connected to a $P'$ such that there is no law which connects $P'$ to any predicate which bridge statements associate with $S_2$. Then any instantiation of $S_1$ which is contingently identical to an instantiation of $P'$ will be an event which constitutes an exception to $S_1x \to S_2x$.) Notice that, in this case, we need assume no exceptions to the laws of the reducing science since, by hypothesis, (6) is not a law.

In fact, strictly speaking, (6) has no status in the reduction at all. It is simply what one gets when one universally quantifies a formula whose antecedent is the physical disjunction corresponding to $S_1$ and whose consequent is the physical disjunction corresponding to $S_2$. As such, it will be true when $S_1 \to S_2$ is exceptionless and false otherwise. What does the work of expressing the physical mechanisms whereby $n$-tuples of events conform, or fail to conform, to $S_1 \to S_2$ is not (6) but the laws which severally relate elements of the disjunction $P_1 \lor P_2 \lor \cdots \lor P_n$ to elements of the disjunction $P_1^* \lor P_2^* \lor \cdots \lor P_m^*$. When there is a law which relates an event that satisfies one of the $P$ disjuncts to an event which satisfies one of the $P^*$ disjuncts, the pair of events so related conforms to $S_1 \to S_2$. When an event which satisfies a $P$ predicate is not related by law to an event which satisfies a $P^*$ predicate, that event will constitute an exception to $S_1 \to S_2$. The point is that none of the laws which effect these several connections need themselves have exceptions in order that $S_1 \to S_2$ should do so.

To put this discussion less technically: we could, if we liked, require the taxonomies of the special sciences to correspond to the taxonomy of physics by insisting upon distinctions between the natural kinds postulated by the former wherever they turn out to correspond to distinct natural kinds in the latter. This would make the laws of the special sciences exceptionless if the laws of basic science are. But it would also lose us precisely the generalizations which we want the special sciences to express. (If economics were to posit as many kinds of monetary systems as there are kinds of physical realizations of monetary systems, then the generalizations of economics would be exceptionless. But, presumably, only vacuously so, since there would be no generalizations left to state. Gresham’s law, for example, would have to be formulated as a vast, open disjunction about what happens in $\mathrm{monetary\ system}_1$ or $\mathrm{monetary\ system}_n$ under conditions which would themselves defy uniform characterization. We would not be able to say what happens in monetary systems tout court since, by hypothesis, ‘is a monetary system’ corresponds to no natural kind predicate of physics.)

In fact, what we do is precisely the reverse. We allow the generalizations of the special sciences to have exceptions, thus preserving the natural kinds to which the generalizations apply. But since we know that the physical descriptions of the natural kinds may be quite heterogeneous, and since we know that the physical mechanisms which connect the satisfaction of the antecedents of such generalizations to the satisfaction of their consequents may be equally diverse, we expect both that there will be exceptions to the generalizations and that these exceptions will be ‘explained away’ at the level of the reducing science. This is one of the respects in which physics really is assumed to be bedrock science; exceptions to its generalizations (if there are any) had better be random, because there is nowhere ‘further down’ to go in explaining the mechanism whereby the exceptions occur.

This brings us to why there are special sciences at all. Reductivism, as we remarked at the outset, flies in the face of the facts about the scientific institution: the existence of a vast and interleaved conglomerate of special scientific disciplines which often appear to proceed with only the most token acknowledgment of the constraint that their theories must turn out to be physics ‘in the long run’. I mean that the acceptance of this constraint, in practice, often plays little or no role in the validation of theories. Why is this so? Presumably, the reductivist answer must be entirely epistemological. If only physical particles weren’t so small (if only brains were on the outside, where one can get a look at them), then we would do physics instead of paleontology (neurology instead of psychology; psychology instead of economics; and so on down). There is an epistemological reply; namely, that even if brains were out where they can be looked at, as things now stand, we wouldn’t know what to look for: we lack the appropriate theoretical apparatus for the psychological taxonomy of neurological events.

If it turns out that the functional decomposition of the nervous system corresponds to its neurological (anatomical, biochemical, physical) decomposition, then there are only epistemological reasons for studying the former instead of the latter. But suppose there is no such correspondence? Suppose the functional organization of the nervous system cross-cuts its neurological organization (so that quite different neurological structures can subserve identical psychological functions across times or across organisms). Then the existence of psychology depends not on the fact that neurons are so sadly small, but rather on the fact that neurology does not posit the natural kinds that psychology requires.

I am suggesting, roughly, that there are special sciences not because of the nature of our epistemic relation to the world, but because of the way the world is put together: not all natural kinds (not all the classes of things and events about which there are important, counterfactual supporting generalizations to make) are, or correspond to, physical natural kinds. A way of stating the classical reductionist view is that things which belong to different physical kinds ipso facto can have no projectible descriptions in common; that if $x$ and $y$ differ in those descriptions by virtue of which they fall under the proper laws of physics, they must differ in those descriptions by virtue of which they fall under any laws at all. But why should we believe that this is so? Any pair of entities, however different their physical structure, must nevertheless converge in indefinitely many of their properties. Why should there not be, among those convergent properties, some whose lawful inter-relations support the generalizations of the special sciences? Why, in short, should not the natural kind predicates of the special sciences cross-classify the physical natural kinds?⁶

Physics develops the taxonomy of its subject-matter which best suits its purposes: the formulation of exceptionless laws which are basic in the several senses discussed above. But this is not the only taxonomy which may be required if the purposes of science in general are to be served: e.g., if we are to state such true, counterfactual supporting generalizations as there are to state. So, there are special sciences, with their specialized taxonomies, in the business of stating some of these generalizations. If science is to be unified, then all such taxonomies must apply to the same things. If physics is to be basic science, then each of these things had better be a physical thing. But it is not further required that the taxonomies which the special sciences employ must themselves reduce to the taxonomy of physics. It is not required, and it is probably not true.

Bibliography

Block, Ned, and Jerry A. Fodor. 1972. “What Psychological States Are Not.” The Philosophical Review 81 (2): 159–181. JSTOR · PhilPapers

Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Open Library

Footnotes

1. I shall usually assume that sciences are about events, in at least the sense that it is the occurrence of events that makes the laws of a science true. But I shall be pretty free with the relation between events, states, things and properties. I shall even permit myself some latitude in construing the relation between properties and predicates. I realize that all these relations are problems, but they aren’t my problem in this paper. Explanation has to start somewhere, too. ↩

2. The version of reductionism I shall be concerned with is a stronger one than many philosophers of science hold; a point worth emphasizing since my argument will be precisely that it is too strong to get away with. Still, I think that what I shall be attacking is what many people have in mind when they refer to the unity of science, and I suspect (though I shan’t try to prove it) that many of the liberalized versions suffer from the same basic defect as what I take to be the classical form of the doctrine. ↩

3. There is an implicit assumption that a science simply is a formulation of a set of laws. I think this assumption is implausible, but it is usually made when the unity of science is discussed, and it is neutral so far as the main argument of this paper is concerned. ↩

4. I shall sometimes refer to “the predicate which constitutes the antecedent or consequent of a law.” This is shorthand for “the predicate such that the antecedent or consequent of a law consists of that predicate, together with its bound variables and the quantifiers which bind them.” (Truth functions of elementary predicates are, of course, themselves predicates in this usage.) ↩

5. I don’t think there is any chance at all that this is true. What is more likely is that type-identification for psychological states can be carried out in terms of the “total states” of an abstract automaton which models the organism. For discussion, see Block and Fodor (1972). ↩

6. As, by the way, the predicates of natural languages quite certainly do. For discussion, see Chomsky (1965). ↩

傑瑞・A・福多（Jerry A. Fodor）
麻省理工學院（Massachusetts Institute of Technology）

原文書目信息：Jerry A. Fodor, “Special Sciences (Or: The Disunity of Science as a Working Hypothesis),” Synthese 28, no. 2 (October 1974): 97–115. DOI · Springer · JSTOR · PhilPapers · 存檔全文

作者註：謹向內德・布洛克（Ned Block）致謝；他讀過本文的一個版本，並提出了極有助益的意見。

網頁版說明：本文依據所提供的期刊掃描本校勘。編號公式以 LaTeX 重排，圖式（5）則重繪為可隨版面縮放的向量圖。原刊顯然的排印錯誤及掃描辨識錯誤均逕行訂正；福多當時採用的拼法 “guaranty” 予以保留。原刊第 111 頁的一處記號不一致，依圖式（5）與公式（6）統一。

---

實證主義科學哲學(positivistic philosophy of science)的一項典型命題是：從長遠來看，特殊科學(special sciences)中一切為真的理論都應當還原(reduce)為物理理論。這原意是一項經驗命題；支持它的部分證據，來自熱的分子理論、化學鍵的物理解釋等科學成就。然而，還原論綱領(reductivist program)在哲學上的流行，不能只憑這些成就來解釋。科學發展見證專門學科增生的次數，至少不下於見證它們被還原為物理學的次數；因此，對還原的普遍熱衷很難只是從既往成功所作的歸納。

我認為，許多接受還原論(reductivism)的哲學家，主要是因為他們希望肯定物理學相對於特殊科學的普遍性(generality)；粗略說來，也就是：凡落在任何一門科學之定律下的事件，都是物理事件，因而也落在物理學定律之下。¹ 對這些哲學家來說，聲稱物理學是基礎科學，與聲稱特殊科學的理論必須還原為物理理論，似乎只是同一件事的兩種說法；於是後一學說便成為對前一學說的標準詮釋。

下文將論證，這是一項相當嚴重的混淆。傳統上所謂的科學統一性(unity of science)，是遠比物理學的普遍性更強、也遠欠可信的命題。若此說為真，便事關重大。還原論雖是一項經驗學說，卻意在對科學實踐發揮調節性作用(regulative role)。能否還原為物理學，被視為特殊科學理論可否接受的一項限制；其奇特後果是：特殊科學愈成功，就愈應當消失。尤其是心理學的方法論問題，正是如此產生的：人們把「心理學的研究對象是物理學研究對象的一部分」這一假定，理解為它蘊涵「心理理論必須還原為物理理論」；而造成麻煩的正是後一原則。我希望藉由質疑這項推論來避開麻煩。

I

還原論認為，一切特殊科學都能還原為物理學。然而，“還原為”在這裡有其特定含義，可作如下刻畫。²

令

$$S_1x \to S_2x \tag{1}$$

為特殊科學 $S$ 的一條定律。（（1）意在讀作類似「一切 $S_1$ 情形都會帶來 $S_2$ 情形」。我假定，一門科學主要是參照其典型謂詞來個別化的；因此，若 $S$ 是一門特殊科學，$S_1$ 與 $S_2$ 就不是基礎物理學的謂詞。我也假定，量化特殊科學定律的“一切”必須打些折扣；這類定律通常並非全無例外。稍後我會詳細回到這一點。）把（1）還原為一條物理定律的充分必要條件，是公式（2）與（3）皆為定律；而把 $S$ 還原為物理學的充分必要條件，則是 $S$ 的全部定律都能如此還原。³

$$S_1x \leftrightarrow P_1x \tag{2a}$$ $$S_2x \leftrightarrow P_2x \tag{2b}$$ $$P_1x \to P_2x \tag{3}$$

$P_1$ 與 $P_2$ 應是物理學的謂詞，而（3）應是一條物理定律。像（2）這樣的公式通常稱為橋接定律(bridge laws)；其特徵在於，同時包含被還原科學(reduced science)與還原端科學(reducing science)的謂詞。因此，像（2）這樣的橋接定律，與像（1）、（3）這樣的本門定律(proper laws)相對。以上所述的結論是：要還原一門科學，其任何本門定律的前件或後件中出現的公式，都必須在某一條橋接定律中作為被還原的公式出現。⁴

關於聯結詞 $\to$，有幾點需要說明。第一，不論它還具有哪些性質，眾所公認的是它必須具有傳遞性(transitivity)。這一點很重要，因為人們通常假定，某些特殊科學是經由把其謂詞連到中介還原理論之謂詞的橋接定律而得到還原的。例如，心理學被認為要途經神經學、生物化學以及其他沿途停靠站，方才還原為物理學。當前的要點是：只要假定 $\to$ 具有傳遞性，這對整個情形的邏輯毫無影響。只要另有橋接定律直接或間接地把 $S^*$ 的謂詞連到物理謂詞，把 $S$ 的謂詞連到 $S^*$ 謂詞的橋接定律，就會滿足把 $S$ 還原為物理學的限制。

不過，如何解釋橋接定律中的 $\to$，仍有相當嚴重而懸而未決的問題。還原論究竟在何種意義上是一項物理主義(physicalist)命題，取決於這些問題。

首先，若我們把本門定律中的 $\to$ 讀作“帶來”或“造成”，橋接定律便須使用另一個聯結詞，因為帶來與造成大概都是非對稱的，橋接定律表達的關係卻是對稱的。此外，若橋接定律中的 $\to$ 被解釋為同一性以外的任何關係，那麼還原論為真至多只能保證某種弱式物理主義為真，而這不足以表達還原論綱領內在的本體論取向。

若橋接定律不是同一性陳述，那麼像（2）這樣的公式至多聲稱：依定律，$x$ 滿足某個 $P$ 謂詞與 $x$ 滿足某個 $S$ 謂詞在因果上相關。由此可知，$S$ 與 $P$ 謂詞在律則上必然適用於同一些事物（也就是說，$S$ 謂詞所適用的事物是 $P$ 謂詞所適用事物的一個子集）。但這當然與非物理主義的本體論相容，因為它容許 $x$ 滿足 $S$ 這件事本身並非物理事件。依此解釋，還原論為真並不保證物理學相對於特殊科學具有普遍性，因為有些事件（即對 $S$ 謂詞的滿足）落在特殊科學 $S$ 的論域中，卻不落在物理學的論域中。（例如，我們可以設想一種學說：物理謂詞與心理謂詞都適用於有機體，但有機體滿足某個心理謂詞所構成的事件，在任何意義上都不是物理事件。其結果會是一種非笛卡爾式的心物二元論(psychophysical dualism)：它是事件和／或性質的二元論，而非實體的二元論。）

基於這類考量，許多哲學家認為，像（2）這樣的橋接定律應被理解為表達偶然的事件同一性(contingent event identities)；因此，（2a）可以讀成大致如下：“由 $x$ 滿足 $S_1$ 所構成的每一事件，都與某個由 $x$ 滿足 $P_1$ 所構成的事件同一，反之亦然。”依此讀法，還原論為真就會蘊涵：落在任何科學定律之下的每一事件都是物理事件；於是既表達了還原論的本體論取向，也保證物理學相對於特殊科學具有普遍性。

若橋接定律表達事件同一性，而落在某門特殊科學本門定律下的每一事件又都落在一條橋接定律下，我們便得到一項我將稱為個例物理主義(token physicalism)的學說。個例物理主義無非主張：諸科學所談論的一切事件都是物理事件。關於個例物理主義，有三點需要注意。

第一，它弱於通常所謂的唯物論(materialism)。唯物論既主張個例物理主義為真，也主張每一事件都落在某一門科學的定律之下。因此，一個人可以是個例物理主義者而不是唯物論者，儘管我看不出有誰會費心採取這一立場。

第二，個例物理主義弱於或可稱為類型物理主義(type physicalism)的學說；後者粗略地主張，任何科學定律中提及的每一項性質都是物理性質。個例物理主義並不蘊涵類型物理主義，因為一對事件偶然同一，大概並不保證構成這些事件的那些被例示性質亦同一；即使事件同一性在律則上是必然的，也仍然如此。另一方面，若每一事件都是某項性質的例示，類型物理主義便蘊涵個例物理主義：當兩個事件都是同一個體在同一時刻例示同一項性質時，兩者就是同一的。

第三，個例物理主義弱於還原論。這一點在某種意義上正是下文論證的重擔，我不在此多作鋪陳。但作為初步近似，還原論是個例物理主義與下述假定的合取：在理想完成的物理學中，存在一些自然類謂詞，逐一對應於任何理想完成的特殊科學中的每一個自然類謂詞。我的一項結論將是：不能從個例物理主義為真的假定推導出還原論為真。還原論是個例物理主義的充分條件，卻不是必要條件。

下文將假定一種蘊涵個例物理主義的還原論讀法。據此，橋接定律陳述的是律則上必然的偶然事件同一性；而把心理學還原為神經學，將蘊涵：由某項心理性質的例示所構成的任何事件，都與某個由某項神經性質的例示所構成的事件同一。

至此我們所得如下：還原論至少在這個意義上蘊涵物理學的普遍性——落在某門特殊科學論域內的任何事件，也會落在物理學的論域內。此外，凡能由一門特殊科學的定律與一項初始條件陳述推出的預測，也能由物理學、橋接定律及該項初始條件陳述所組成的理論推出。最後，由於“還原為”被認為是一種非對稱關係，物理學也將成為基礎科學；也就是說，若還原論為真，物理學便是唯一在上述意義上具有普遍性的科學。現在我要論證，還原論對科學統一性施加了過強的限制；但在相關意義上較弱的學說，仍會保存還原論所欲得到的後果：個例物理主義、物理學的普遍性，以及物理學在諸科學中的基礎地位。

II

每一門科學都蘊含一套關於其論域(universe of discourse)內事件的分類體系(taxonomy)。更具體地說，每一門科學都使用一套由理論謂詞(theoretical predicates)與觀察謂詞(observation predicates)組成的描述詞彙(descriptive vocabulary)；事件正因滿足這些謂詞而落在該科學的定律之下。顯然，對事件的真實描述並不全都屬於這種詞彙。例如，有大量事件的內容是某物被運送到距埃菲爾鐵塔不足三英里之處。然而，我認為，沒有任何科學會把“被運送到距埃菲爾鐵塔不足三英里之處”納入其描述詞彙。等價地說，我認為，並無任何自然定律會因事件本身是“被運送到距埃菲爾鐵塔不足三英里之處”這一性質的例示(instantiation)而適用於它們（儘管我想，可以設想有某條定律，會因事件本身是某個雖有別於此、卻與之同延(co-extensive)的性質之例示而適用於它們）。為簡稱這些事實，我將說，“被運送……”這一性質並不決定一個自然類(natural kind)，而表達此性質的謂詞也不是自然類謂詞。

若我知道定律究竟是什麼，並且相信科學理論僅由一組組定律構成，那麼我便可以說：相對於科學 $S$ 而言，$P$ 是自然類謂詞，若且唯若 $S$ 包含形式為 $Px \to \alpha x$ 或 $\alpha x \to Px$ 的本門定律；粗略地說，一門科學的自然類謂詞，就是那些其項在該科學的本門定律中作為約束變項(bound variables)出現的謂詞。即便處於目前這種無知狀態，我仍傾向於這樣說，並接受由此而來的後果：這會使“自然類”這個含混概念惡性地依賴同樣含混的“定律”與“理論”概念。這裡沒有任何穩固的立足點。若我們對什麼是自然類意見不一，那麼出於同樣的理由，我們很可能也會對什麼是定律意見不一。我不知道如何擺脫這個循環，但我認為，關於我們究竟身處哪一個循環，仍有一些有意思的話可說。

例如，現在我們可以刻畫還原論對科學統一論的詮釋究竟在哪一方面過強。若還原論為真，則每一自然類要麼就是物理自然類，要麼與某個物理自然類同延。（若橋接定律表達性質同一(property identities)，那麼每一自然類就是物理自然類；若橋接定律表達事件同一(event identities)，那麼每一自然類便與某個物理自然類同延。）這直接由還原論的以下前提得出：特殊科學定律中凡作為前件(antecedent)或後件(consequent)出現的謂詞，都必須在某一橋接式中作為被還原謂詞出現；再加上自然類謂詞就是那些其項在本門定律中作為約束變項出現的謂詞這一假定，結論便隨之成立。簡言之，若每一條特殊科學定律都與某條物理定律形成如（3）與（1）之間的關係，那麼特殊科學的每一個自然類謂詞，也都會與物理學的某個自然類謂詞形成如（2）將 $S_1$、$S_2$ 分別聯繫於 $P_1$、$P_2$ 的關係。

我現在要提出若干理由，說明為什麼應當相信還原論的這項後果不可接受。我並不把它們當作一錘定音的理由；鑒於還原論是否過強歸根結柢是一個經驗問題，它們也不可能如此。（世界可能竟是每一自然類都對應一個物理自然類，正如世界也可能竟是“被運送到距埃菲爾鐵塔不足三英里之處”這一性質在譬如流體動力學中決定了一個自然類。只不過就現狀而言，世界似乎極不可能是這兩種情形中的任何一種。）

每一自然類之所以不太可能都對應一個物理自然類，理由無非如下：（a）我們往往能對物理描述毫無共同之處的事件作出有意思的概括，例如支持反事實條件句的概括(counterfactual-supporting generalizations)；（b）被這些概括涵攝的事件，其物理描述是否具有任何共同之處，在一個顯而易見的意義上，往往與這些概括的真實性、有趣性、確證程度，乃至其任何認識論上重要的性質，全然無關；（c）特殊科學大量從事的，正是作出這類概括。

我認為，以上幾點明顯得近乎不證自明；只要作出那個（貌似激進的）轉變，開始認真看待特殊科學，它們便會立刻躍入眼簾。比如，假設格雷欣“定律”確實為真。（若有人不喜歡格雷欣定律，那麼換成任何可設想的未來經濟學中的任何一條真概括，大概也同樣有效。）格雷欣定律告訴我們，在特定條件下，貨幣交換中將會發生什麼。我願意相信，物理學在如下意義上具有普遍性：它蘊含任何一項貨幣交換事件（因而也就是任何落在格雷欣定律之下的事件）都可用物理學詞彙作出一項真實描述，且該事件正憑藉這項描述落在物理定律之下。然而，最尋常的考量便表明，能涵蓋所有這類事件的描述必定是一項極其析取式的描述。有些貨幣交換涉及貝殼串珠，有些涉及美元紙鈔，還有些涉及在支票上簽名。由物理謂詞構成、足以涵蓋所有這些事件的析取，究竟有多大可能表達一個物理自然類？換言之，一個可構成“$x$ 是一項貨幣交換 $\leftrightarrow \ldots$”這種橋接定律右側的析取謂詞(disjunctive predicate)，究竟有多大可能表達一個物理自然類？尤其是，這樣的謂詞究竟有多大可能成為某條物理學本門定律的前件或後件？問題在於，貨幣交換確實具有一些有意思的共同之處；若格雷欣定律為真，它所陳述的便是其中一項。然而，貨幣交換的有趣之處，肯定不在於它們在物理描述下的共同性。貨幣交換這樣的自然類，或許碰巧與某個物理自然類同延；但若果真如此，那將是宇宙尺度上的偶然。

事實上，還原論的處境比以上討論所顯示的更糟。因為還原論不僅主張所有自然類都與物理自然類同延，而且主張這些同延關係是律則上必然的(nomologically necessary)：橋接定律確實是定律。因此，若格雷欣定律為真，便可推出存在一條自然界的橋接定律，其形式為“$x$ 是一項貨幣交換 $\leftrightarrow x$ 是 $P$”，其中 $P$ 是一個物理自然類詞項。然而，肯定沒有這樣的定律。若有，$P$ 所涵蓋的便不僅必須包括所有現存的貨幣交換制度，還必須包括所有可能存在的貨幣交換制度；一條定律必須在反事實情形中也成立。在“$x$ 是一項律則上可能的貨幣交換 $\leftrightarrow Px$”中，哪一個物理謂詞堪作 $P$？

總而言之：當整場戲落幕時，一位不朽的經濟物理學家或許會在物理學中找到一個謂詞，作為一項純粹事實，它與“是一項貨幣交換”同延。若物理學具有普遍性——若還原論的本體論偏向為真——那就必定有這樣一個謂詞。然而，（a）借用唐納德・戴維森在一個稍有不同的語境中所作的評論，唯有純粹枚舉才能使我們相信這一純粹同延的事實；（b）用來陳述這項同延關係的物理謂詞，似乎完全不可能是一個自然類詞項；而且（c）這項同延關係具有定律性的可能更為渺茫，也就是說，它不僅要在那個最終成為現實的律則上可能世界中成立，還要在任何一個律則上可能世界中都成立。

我認為，以上討論強烈表明，在科學統一論主張所特有的那種還原意義上，經濟學不可還原為物理學。我猜，在這方面，經濟學並無任何特殊之處；經濟學不太可能還原為物理學的理由，與心理學不太可能還原為神經學的理由彼此平行。

若心理學可以還原為神經學，那麼對每一個心理自然類謂詞，都有一個與之同延的神經學自然類謂詞，而且陳述這項同延關係的概括本身是一條定律。 顯然，許多心理學家都相信某種近似的主張。世界各地的大學設有“心理生物學”或“心理學與腦科學”系所；這些系所的存在本身，就是一項制度化的賭注，押注人們能夠找到這種具有定律性的同延關係。然而，正如近來關於唯物論的討論反覆指出的，我們有充分理由對這些賭注有所保留。除了心理狀態類型與神經狀態類型之間最粗略的對應外，我們沒有任何確鑿資料；而且，高等生物的神經系統完全可能通常會以多種多樣的神經學手段達到某個既定的心理目的。若是如此，將神經結構與心理功能逐一配對的企圖便注定失敗。像卡爾・拉什利（Karl Lashley）這樣重要的生理心理學家，所持的正是這種觀點。

這裡的要點是，無論如何，心理學中的還原論綱領不能以本體論理由來辯護。即便每一個例心理事件都是一個例神經事件，也推不出心理學的自然類謂詞與任何其他學科（包括物理學）的自然類謂詞同延。換言之，每一心理事件都是物理事件這一假定，並不保證物理學——更遑論任何較心理學更一般的學科——能為心理理論提供適當的詞彙。我之所以強調這一點，是因為我相信，許多生理心理學家之所以把成敗都押在還原論綱領上，恰恰是因為他們把該綱領與個例物理主義混為一談。

我一直質疑的是，是否存在與心理自然類同延的神經學自然類。眼下愈來愈清楚的是，即便存在這種同延關係，它也不可能具有定律性。因為愈來愈可能的是：除了生物體之外，還有其他律則上可能的系統——即自動機(automata)——會滿足心理學中的自然類謂詞，卻完全不滿足任何神經學謂詞。正如普特南所強調的，只要存在任何這種系統，它們的數量大概就極其龐大，因為等價的自動機幾乎可以由任何東西製成。若這一觀察正確，那麼我們便沒有任何切實希望，能以物理自然類謂詞描述那一類其心理實際上與某種有機體的心理相同的自動機（當然，若個例物理主義為真，仍可用某個物理謂詞把這一類刻畫出來）。其結果是，科學統一論的經典表述只能任由電腦模擬(computer simulation)領域的進展擺佈。當然，這無非是說，那種表述過強了。科學統一論原本意在成為一項可由某些可能的科學發現所推翻的經驗假說；但沒有人原本設想，它竟會被 Newell、Shaw 與 Simon 推翻。

迄今為止，我所論證的是：心理還原論——即每一心理自然類本身就是某一神經學自然類，或與某一神經學自然類同延的學說——既不等同於個例物理主義——即每一心理事件都是神經事件的學說——也不能由後者推出。然而，有人或許會主張，不妨乾脆把兩種學說視為等價，因為我們對個例物理主義所能擁有的唯一可能證據，也會是還原論的證據；這種證據就是發現類型對類型的心物相關(type-to-type psychophysical correlations)。

然而，稍加思索即可看出，這項論證並不成立。若類型對類型的心物相關可以成為個例物理主義的證據，那麼其他可以明確刻畫的相關形式同樣可以。

所謂類型對類型的相關，是指：每一個由同一心理類型的事件組成的 $n$ 元組，都有一個與之相關、且由同一神經學類型的事件組成的 $n$ 元組。現在設想一個找不到這種相關的世界。人們反而發現，對每一個由類型同一(type-identical)的心理事件組成的 $n$ 元組，都有一個在時空上與之相關、卻由類型相異(type-distinct)的神經事件組成的 $n$ 元組。也就是說，每一心理事件都與某個神經事件配對，但同一心理類型的事件可能與不同神經類型的事件配對。我此處的要點是：只要我們能夠表明，與某一類心理事件配對的那些類型相異的神經事件，在所有與心理學中的類型辨識相關的性質方面皆彼此相同，這種配對為個例物理主義提供的支持，便與類型對類型的配對一樣充分。為便於說明，假設心理事件是參照其行為後果(behavioral consequences)來辨識類型的。⁵ 那麼，對所有與一組類型同質的心理事件配對的神經事件，所要求的僅僅是它們在行為後果方面相同。簡言之，類型同一的事件當然不會具有全部相同的性質，而類型相異的事件仍必然在某些性質方面相同。個例物理主義的經驗確證，並不取決於證明類型同一的心理事件所對應的神經事件本身也是類型同一的。需要證明的僅僅是：它們在那些決定一個給定事件屬於哪一類心理事件的性質方面彼此相同。

我們能否取得證據，表明一組在其他方面異質的神經事件共同具有這類性質？當然可以。神經學理論本身或許能解釋，為何由類型相異的神經事件構成的某個 $n$ 元組在行為後果方面相同；甚至能解釋，它們為何在數目不定的許多其他此類關係性質(relational properties)中的任何一個方面相同。而且，若神經學理論未能作出這種解釋，某門比神經學更基礎的科學或許能夠成功。

我在以上整段討論中的要點仍然不是：類型同質的心理狀態與類型異質的神經狀態之間的相關，會證明個例物理主義為真。我的要點只是：這種相關或許能給予我們與類型對類型相關同樣充分的理由，使我們成為個例物理主義者。若此說正確，那麼從個例物理主義推向還原論的認識論論證必定有誤。

在我看來——將這一點說得十分一般——科學統一論的經典詮釋其實誤解了科學還原(scientific reduction)的目標。還原的要旨並不主要在於，為被還原科學的每一個自然類謂詞找到某個與之同延的物理自然類謂詞；它毋寧在於闡明事件藉以遵循特殊科學定律的物理機制(physical mechanisms)。我一直論證的是：無論在邏輯上還是在認識論上，都沒有理由認為後一項工程的成功必須以先一項工程的成功為條件；而且，只要事件遵循某條特殊科學定律所憑藉的物理機制是異質的，這兩項工程事實上便很可能彼此分離。

III

我認為，至此的討論表明，還原論很可能是對科學統一性過強的詮釋：一方面，它與特殊科學中很可能出現的結果不相容；另一方面，若我們主要所求只是成為立場穩健的個例物理主義者，它又比我們所需假定的更多。下文將嘗試勾勒一種對還原論的放寬版本；在這些方面，它似乎恰好足夠強。然後，我會提出兩項相互獨立的理由，說明為何這項修正版學說或許才是正確的。

一直以來，問題都在於：存在一種尚未排除的經驗可能性——與被還原科學的自然類謂詞相對應的，可能是還原端科學中一組異質而缺乏系統性的謂詞析取；我們不希望因這種可能性而預先否定科學統一性。那麼，假定我們容許橋接陳述採取如下形式：

$$Sx \leftrightarrow (P_1x \lor P_2x \lor \cdots \lor P_nx) \tag{4}$$

其中，$P_1 \lor P_2 \lor \cdots \lor P_n$ 在還原端科學中並不是自然類謂詞。我認為，這等於容許至少某些“橋接定律”事實上最後可能並不是定律；因為我認為，一項全稱概括要具有定律性(lawlike)，有一項必要條件，就是構成其前件與後件的謂詞必須指稱自然類。因此，我假定，就科學統一性的目的而言，只要特殊科學的每一條定律都能藉由表達真實經驗概括的橋接陳述還原為物理學，就已經足夠。須記住，橋接陳述應理解為同一性陳述的一種；因此，（4）可以讀成大致如下：“由 $x$ 滿足 $S$ 所構成的每一事件，都與某個由 $x$ 滿足析取 $P_1 \lor P_2 \lor \cdots \lor P_n$ 中某一謂詞所構成的事件同一。”

現在，在與（2）相應的陳述不是定律的還原個案中，與（3）相應的陳述也不會是定律，理由相同：依假定，前件或後件中出現的謂詞不是自然類謂詞。我們所得的反而會是類似（5）的情形（見下圖）。

亦即，被還原定律的前件與後件將各自連到還原端科學中一組謂詞析取；而且，如果被還原定律全無例外，還原端科學中就會有若干定律，把同前件相聯的析取中每一成員的滿足，連到同後件相聯的析取中某一成員的滿足。也就是說，若 $S_1x \to S_2x$ 全無例外，就必須有還原端科學的某條本門定律，陳述或蘊涵：對某個 $P^*$ 而言，$P_1x \to P^*$；從 $P_2x$ 到 $P_nx$ 也同樣如此。既然必須有這樣的定律，便可推出 $P_1 \lor P_2 \lor \cdots \lor P_n$ 的每個析取支都是自然類謂詞；$P_1^* \lor P_2^* \lor \cdots \lor P_m^*$ 的每個析取支亦然。

不過，關鍵也正是在這裡。有人可能論證：若 $P$ 析取的每個析取支都依法連到 $P^*$ 析取的某個析取支，則（6）本身也是一條定律。

$$(P_1x \lor P_2x \lor \cdots \lor P_nx) \to (P_1^*x \lor P_2^*x \lor \cdots \lor P_m^*x) \tag{6}$$

其理由會是：（5）給出 $P_1x \to P_2^*x$、$P_2x \to P_m^*x$ 等；而從形式為 $(P \supset R)$ 與 $(Q \supset S)$ 的前提，推到形式為 $(P \lor Q) \supset (R \lor S)$ 的結論，是有效論證。

對此，我傾向於說，這只表明“——是一條定律”界定了一個非真值函數語境(non-truth-functional context)（或者，就此處的目的而言等價地說，並非自然類謂詞的一切真值函數本身都是自然類謂詞）。具體而言，不能由“$P$ 帶來 $R$ 是一條定律”與“$Q$ 帶來 $S$ 是一條定律”，推到“$P$ 或 $Q$ 帶來 $R$ 或 $S$ 是一條定律”。（當然，從這些前提推到不加限定的“$P$ 或 $Q$ 帶來 $R$ 或 $S$”，完全沒有問題。）例如，我認為陽光照射綠色植物造成碳水化合物合成是一條定律，也認為摩擦生熱是一條定律；但我不認為（陽光照射綠色植物或摩擦）造成（碳水化合物合成或熱）是一條定律。相應地，我懷疑能否合理地把“是碳水化合物合成或熱”視為自然類謂詞。

嚴格說來，並非人人非同意以上全部論述不可；但否認它要付出代價。尤其是，若容許在“——是一條定律”這一語境內進行所有真值函數式論證，就必須放棄這種可能：把一門科學的自然類謂詞等同於其本門定律中作為前件或後件出現的那些謂詞。（這樣一來，（6）便會是一條不符合該條件的物理學本門定律。）於是，我們便需要另行解釋自然類這一概念，而我不知道這種替代解釋會是什麼樣子。

結論似乎如下：若不要求橋接陳述必須是定律，那麼，要麼特殊科學定律所還原到的某些概括本身並不具有定律性，要麼有些定律不能以自然類來表述。不論如何理解（5），重點在於它弱於標準還原論：它不要求被還原科學與還原端科學的自然類之間存在對應。然而，在使標準還原論成為物理主義的同一項假定之下，它仍是物理主義的（亦即，橋接陳述表達真實的個例同一性）。而這些正是我們希望修正後的科學統一性說明所具有的性質。

現在，我要提出兩項理由，說明為何這種對科學統一性的詮釋是正確的。第一，它使我們看出特殊科學的定律如何可以合理地容有例外；第二，它使我們看出究竟為何會有特殊科學。以下依次討論。

再次考察（2）與（3）所蘊含的還原模型。我假定基礎科學的定律嚴格地全無例外，也假定特殊科學的定律並非如此乃是常識。但這樣一來，我們便面臨一個痛苦的兩難。由於 $\to$ 表達的是一種（或若干種）必須具有傳遞性的關係，（1）唯有在橋接定律也有例外時才能有例外。然而，若橋接定律有例外，還原論便失去其本體論上的力道，因為我們再也不能說：由某個 $S$ 謂詞的例示所構成的每一事件，都與某個由 $P$ 謂詞的例示所構成的事件同一。簡言之，在還原論模型下，我們不可能無矛盾地同時假定橋接定律與基礎定律全無例外，而特殊科學的定律卻有例外。但除非我們甘願削弱作為還原論綱領主旨的本體論主張，否則也不能容許橋接定律遭到違反。

要擺脫這一困境（挽救該模型），有兩條路：或者放棄特殊科學定律有例外的主張，或者放棄基礎定律全無例外的主張。我認為兩種方案都不可取。第一種方案公然違背事實。譬如心理學中真實而支持反事實條件句的概括，根本不可能最後在其前件得到滿足的每一個情形中都毫無例外地成立。縱使心靈固然願意，肉體卻往往軟弱。總會有一些行為失常可以由生理學解釋，卻從心理學理論的觀點看來毫無意趣。第二種方案只稍微好一些。畢竟，基礎科學的定律最後也可能被發現有例外；但問題是，我們是否願意讓科學統一性依賴於它們確有例外這項假定。

然而，依（5）所概括的說明，一切都能得到令人滿意的處理。$S_1x \to S_2x$ 出現一項例外的律則上充分條件(nomologically sufficient condition)是：橋接陳述把某次對 $S_1$ 的滿足，等同於某次對一個 $P$ 謂詞的滿足，而後者本身並未依法連到任何 $P^*$ 謂詞的滿足。（也就是說，假定 $S_1$ 連到某個 $P'$，卻沒有任何定律把 $P'$ 連到橋接陳述同 $S_2$ 相聯的任一謂詞。那麼，任何與 $P'$ 的一項例示偶然同一的 $S_1$ 例示，都會是一個構成 $S_1x \to S_2x$ 之例外的事件。）請注意，在這種情況下，我們無須假定還原端科學的定律有任何例外，因為依假設，（6）不是一條定律。

事實上，嚴格說來，（6）在還原中根本沒有任何地位。它只是對一個公式作全稱量化所得的結果；該公式以前件 $S_1$ 所對應的物理析取為前件，以後件 $S_2$ 所對應的物理析取為後件。因此，當 $S_1 \to S_2$ 全無例外時，它就為真；否則便為假。真正承擔表達物理機制之工作的並非（6），而是那些分別把析取 $P_1 \lor P_2 \lor \cdots \lor P_n$ 的成員，連到析取 $P_1^* \lor P_2^* \lor \cdots \lor P_m^*$ 之成員的定律；事件的 $n$ 元組正是藉由這些機制而符合或不符合 $S_1 \to S_2$。若有一條定律把滿足某一 $P$ 析取支的事件連到滿足某一 $P^*$ 析取支的事件，這對如此相聯的事件便符合 $S_1 \to S_2$。若一個滿足 $P$ 謂詞的事件並未依法連到某個滿足 $P^*$ 謂詞的事件，該事件就會構成 $S_1 \to S_2$ 的一項例外。要點在於：為使 $S_1 \to S_2$ 可以有例外，實現這些個別聯繫的定律本身無一需要有例外。

把這番討論說得不那麼技術化：若我們願意，大可以要求特殊科學的分類體系與物理學的分類體系相對應；辦法是，只要前者所設定的自然類在後者中對應於不同自然類，就堅持在前者之中也加以區分。若基礎科學的定律全無例外，這樣做便能使特殊科學的定律也全無例外。但它也恰恰會使我們失去希望由特殊科學表達的那些概括。（若經濟學設定的貨幣制度種類，與貨幣制度的物理實現種類一樣多，那麼經濟學的概括便會全無例外。然而，可以推想，這只會空洞地成立，因為此時根本沒有概括可供陳述。比如，格雷欣定律將不得不表述成一項龐大而開放的析取，述說 $\text{貨幣制度}_1$ 或 $\text{貨幣制度}_n$ 在某些條件下會發生什麼，而這些條件本身也無法得到統一刻畫。我們將不能述說不加限定的貨幣制度會發生什麼，因為依假設，“是一種貨幣制度”並不對應於物理學中的任何自然類謂詞。）

事實上，我們的做法恰好相反。我們容許特殊科學的概括有例外，從而保存這些概括所適用的自然類。但由於我們知道，這些自然類的物理描述可能相當異質；又知道，把這類概括之前件的滿足連到其後件之滿足的物理機制，也可能同樣多種多樣；所以我們既預期這些概括會有例外，也預期這些例外會在還原端科學的層次上得到“消解式解釋”。這正是物理學確實被假定為基底科學(bedrock science)的一個方面：其概括的例外（若確有例外）最好是隨機的，因為在解釋例外發生的機制時，已經無處可以“再向下”追究。

這便把我們帶到一個問題：究竟為何會有特殊科學。正如開篇所說，還原論公然違背有關科學建制的事實：現存的是一個由特殊科學學科組成、龐大而彼此交織的複合體；這些學科往往看似只對如下限制作了極其象徵性的承認——其理論“從長遠來看”必須最後成為物理學。我的意思是，接受這項限制，在實際工作中往往對理論的驗證幾乎不起作用，甚至全無作用。為什麼會如此？可以推想，還原論者的答案必然完全是認識論的：假如物理粒子不那麼微小（假如大腦長在外面，可以讓人看個清楚），我們便會研究物理學而非古生物學（研究神經學而非心理學，研究心理學而非經濟學，如此逐級向下）。這裡確有一個認識論上的回答：即使大腦真長在外面可供觀察，以我們目前的狀況，仍然不知道該尋找什麼；我們缺少用心理學分類體系來分類神經事件所需的適當理論裝置。

若神經系統的功能分解(functional decomposition)最後確實與其神經學分解（解剖學、生物化學、物理學分解）相對應，那麼，研究前者而非後者的理由就只是認識論上的。但假若沒有這種對應呢？假若神經系統的功能組織交叉切分(cross-cuts)其神經組織（從而在不同時刻或不同有機體中，相當不同的神經結構能夠實現同一種心理功能）呢？那麼，心理學之所以存在，就不取決於神經元可悲地過於微小，而取決於神經學並不設定心理學所需要的自然類。

粗略說來，我的主張是：特殊科學之所以存在，不是因為我們與世界之間認識關係的性質，而是因為世界本身的構造方式。並非所有自然類（亦即，並非一切有重要而支持反事實條件句之概括可作的事物與事件類別）都是物理自然類，或與物理自然類相對應。表述古典還原論觀點的一種方式是：屬於不同物理種類的事物，依此事實本身（ipso facto）便不可能共有任何可投射的描述(projectible descriptions)；若 $x$ 與 $y$ 在那些使其落在物理學本門定律下的描述上有別，它們就必定也在那些使其落在任何定律下的描述上有別。但我們為什麼應當相信如此？任何一對實體，不論物理結構多麼不同，仍必定在數量無定限的諸多性質上相合。為什麼這些相合的性質之中，不能有一些其依法成立的相互關係足以支持特殊科學的概括？簡言之，特殊科學的自然類謂詞，為什麼不能交叉分類(cross-classify)物理自然類？⁶

物理學發展出最適合其目的的研究對象分類體系；其目的，是表述在上述數種意義上皆屬基礎而全無例外的定律。但若要服務一般科學的目的——例如，要把一切可陳述的、真實而支持反事實條件句的概括陳述出來——所需的分類體系未必只有這一種。因此，特殊科學以各自專門的分類體系，負責陳述其中一些概括。若科學要獲得統一，所有這些分類體系就必須適用於同一些事物。若物理學要成為基礎科學，這些事物中的每一個最好都是物理事物。但除此之外，並不要求特殊科學所使用的分類體系本身必須還原為物理學的分類體系。這既非必要，也很可能不是事實。

參考文獻

Block, Ned, and Jerry A. Fodor. 1972. “What Psychological States Are Not.” The Philosophical Review 81 (2): 159–181. JSTOR · PhilPapers

Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Open Library

Footnotes

1. 我通常會假定，諸科學是關於事件的；至少在這個意義上如此：使一門科學的定律為真的，是事件的發生。不過，我會相當寬鬆地處理事件、狀態、事物與性質之間的關係，甚至也容許自己在理解性質與謂詞之間的關係時保留一些餘地。我知道所有這些關係都是問題，但它們不是我在本文中的問題。解釋也總得從某處開始。 ↩

2. 本文所討論的還原論版本，比許多科學哲學家所主張的更強；這一點值得強調，因為我的論證所要表明的，恰恰是它強得無法成立。不過，我認為我所批判的，正是許多人提及科學統一性時心中所想的學說；而且我懷疑（雖然不打算加以證明），許多放寬版本也帶有我所視為該學說古典形式之同一個根本缺陷。 ↩

3. 這裡隱含地假定，一門科學無非就是一組定律的表述。我認為這項假定並不可信；但討論科學統一性時，人們通常都採取它，而且就本文的主要論證而言，它是中立的。 ↩

4. 我有時會說“構成一條定律之前件或後件的謂詞”。這是下述說法的簡稱：“某一謂詞，連同其約束變項及約束這些變項的量詞，共同構成一條定律的前件或後件。”（依此用法，初等謂詞的真值函數當然本身也是謂詞。） ↩

5. 我認為這根本不可能為真。更可能的是，心理狀態的類型辨識可以參照一部模擬有機體的抽象自動機之“總體狀態”來進行。相關討論見 Block 與 Fodor（1972）。 ↩

6. 順帶一提，自然語言的謂詞確實正是如此。相關討論見 Chomsky（1965）。 ↩