From Free Life, Issue 23, August 1995
ISSN: 0260 5112

We may describe as philosophical reductionism any view that holds that concepts of one class may be "reduced to", i.e. analysed completely in terms of, concepts of another class. For example, some philosophers have held that all descriptions of mental states can in principle be dispensed with (without loss of content) in favour of descriptions of behaviour. In this book, John Carroll argues against philosophical reductionism with respect to nomic concepts, i.e. causality, laws of nature and other closely allied concepts like chance, physical necessity, causal explanation, dispositions and counterfactual conditionals (i.e. "if it were the case that p, then it would be the case that q").

Attempts to analyse away the nomic concepts have their provenance in Hume. Hume supposed that all our concepts must be derivable from our sensory experiences, but his analysis of our sensory experience could find no basis for any concept of necessary connection. Since necessary connection is integral to the concept of cause, that concept had to be rejected, and our talk of causes was then explained away as a projection on to the world of a psychological compulsion.

In this century, logical positivists and others in the Humean tradition have tried to rehabilitate the nomic concepts by analysing them in terms of non-nomic concepts, i.e. in terms of a combination of purely logical notions and concepts which are plainly derivable from our sensory experience. The problem that has preoccupied these philosophers has been that of providing an analysis of laws of nature (e.g. no signal travels faster than the speed of light) that distinguishes them from accidental generalisations (e.g. no gold spheres are more than ten metres in diameter) without relying on nomic notions like physical necessity.

The difficulties encountered in trying to analyse the nomic concepts in non-nomic terms have led some philosophers of this bent to substitute a weaker kind of reduction for analysis: they claim that the nomic concepts supervene upon non-nomic concepts. Roughly, this means that any two possible worlds which agree on their non-nomic facts will also agree on their nomic facts (so there must be some deep metaphysical dependence of the nomic on the non-nomic). As Mr Carroll explains it, what this claim amounts to is the claim that no two possible worlds have propositions that agree on all their non-nomic features and also disagree on their status as laws.

Mr Carroll makes the following distinctions. The nomic concepts are those which directly involve generality and some sort of (non-logical) necessity (see above for examples). Within this class he distinguishes the causal concepts from the rest, as being those that have very direct and obvious connections with causation. Within the class of non-nomic concepts he distinguishes concepts with nomic commitments, which are those which could not be instantiated unless some nomic concepts were also instantiated. Examples include fundamental concepts like perception, persistence and materiality, as well as a wide range of very ordinary everyday concepts like being a table. For instance, it is a conceptual truth that one cannot perceive something unless it is causally instrumental in bringing about oneþs perceptual state; and it is a conceptual truth that something cannot be a table unless it is in some way capable of supporting objects (i.e. it must have a disposition). Finally, the concepts free of nomic commitments, make up a pretty barren class, consisting of the concepts of logic and mathematics, perhaps concepts of spatial and temporal relations and, although he doesnþt mention these, I think he must also include concepts of pure sensory qualities like sounds and colour appearances ("Humean" concepts). The fact that this class of concepts is pretty barren, shows that the nomic concepts are central to our thought about the world.

Mr Carroll argues for the following conclusions:

(a) nomic concepts cannot be analysed in terms of concepts free of nomic commitments (this rules out the many kinds of regularity analysis of lawhood);

(b) nomic concepts do not supervene on non-nomic concepts (so, a fortiori, the former cannot be analysed in terms of the latter), even if we include in the latter class concepts which have nomic commitments;

(c) causal concepts do not supervene upon non-causal concepts (so, a fortiori, the former cannot be analysed in terms of the latter), even if we include in the latter the non-causal nomic concepts.

As a consequence, he also concludes that the business of analytical philosophy is not so much the search for analyses as the drawing of distinctions and the mapping out of connections, or, as he calls it, conceptual geography. I think he is dead right in all of these conclusions.

In arguing for these conclusions, Mr Carroll employs the techniques and, to a much lesser extent, the symbolism, of modern symbolic modal logic. Thus, to show that two classes of concepts are distinct, and that the one class does not supervene upon the other, he carefully describes possible worlds which, it becomes clear, agree so far as the concepts of the one class are concerned but disagree as to whether the concepts of the other class are instantiated. For example, he describes a possible world in which, to put it schematically, it is a law that the state of affairs, a, causes the state of affairs, b, and in which it is a law that a causes the state of affairs c. In addition, it is also a law that nothing but a could cause either b or c. In other words, it is a law that a if and only if b; and it is a law that a if and only if c. From this it follows that it is a law that b if and only if c; yet b does not cause c, and c does not cause b. This shows the distinctness of causality from lawhood. He then describes another possible world just like the first, except that a causes b and b causes c, and only a could cause b and only b could cause c. So once again, it is a law that a if and only if b, and it is a law that b if and only if c, so it follows that it is also a law that a if and only if c. In other words, all the laws are the same as in the previous world, but the causal facts are different. It follows that causality does not supervene upon the non-causal concepts, including lawhood.

Of course, with this type of argument the appeal is, ultimately, to intuitions (as it always has to be, ultimately); so there is always the possibility (nay, likelihood) that there will be philosophers of a Humean frame of mind who just reject the carefully constructed counter-examples as unintelligible or question-begging (as drawing distinctions where none exist). Thatþs life.

Mr Carroll represents his arguments as an attack on empiricism. While this is understandable, I think it is also misleading, as I shall now explain. For Humean empiricists, our sensory experience puts us in touch with our sensory states, and this seems to provide a basis only for concepts free of nomic commitment. Mr Carrollþs demonstration of the distinctness and non-supervenience of the nomic concepts therefore raises the Kantian question: with what right do we employ the nomic concepts in our thought about the world? Unless an answer can be given to this question, we seem to be vulnerable to a general scepticism concerning lawhood and causality. Mr Carroll himself admits to a limited scepticism about lawhood and causality in that he admits the possibility of certain situations in which we would never be able to tell whether a given regularity is a law, or whether a given law is causal. However, he side- steps the Kantian question and, avoids the threat of general scepticism about lawhood and causality, by means of two devices. First, he espouses a direct realism according to which we have direct perceptual knowledge of causation. Second, he posits a special kind of non-deductive inference which leads legitimately from perceptual knowledge to knowledge of lawhood. This, it seems to me, is tantamount to regarding the nomic concepts as either empirical concepts or concepts in some way derivable from our sensory experience. So he is still a kind of empiricist, but he disagrees with the Humeans over what our sensory experience encompasses and what kinds of inferences are valid. To me, on the other hand, direct realism seems to be plainly false and Mr Carrollþs posited inferences seem highly suspect. So, given that I accept his conclusions about distinctness and non-supervenience, I have to take the Kantian question seriously. (A kind of Kantian "transcendental argument" in answer to this question is suggested by Mr Carroll himself at pp88-94, though I am very doubtful as to its success).

Mr Carrollþs book is somewhat technical in that it presupposes familiarity with the methods of modern modal logic and also with some of the history of the discussion of the issues it addresses (although he outlines the various positions that the participants in this discussion have taken, I do not think that these summaries could be understood by someone coming to the topic for the first time). This is certainly not a book for the philosophical tyro. However, I think it is an excellent book, and I recommend it to all serious students of analytical philosophy.

Danny Frederick