20th WCP: Habits and Explanation

If you do something often and in a way which appears settled, you are normally described as having a habit. We assume we have habits as parts of skills, habits of social behaviour, and also deeper ones, like habits of thought. Even those who are good at using Ockham's razor confess sometimes their acceptance of habits. Hume admits that we all have a habit of jumping from constant conjunction to necessary connection. Quine alludes to the fact that we all have a habit of disturbing our set of accepted beliefs as little as possible. Moreover, the notion of practice as used by sociologists and historians is often assumed to be composed of habits: either as a summation of habits of many individuals, or as a collective habit of social entities like groups or institutions.

So habits form a crucial part of our everyday conceptual scheme used to explain normal human activity. They have however been neglected in debates concerning folk-psychology. These debates have concentrated on propositional attitudes, like beliefs. Using these attitudes as examples, radical eliminativists argue that folk-psychological explanation would have merit only if it reduces smoothly to neurophysiological explanation. But, according to them, folk-psychology exhibits explanatory failures on an epic scale (e.g. Churchland, 1981; Stich, 1983). The way our best theories in cognitive science carve up our mental activity is, or will turn out to be, so different from the way folk-psychology does it that there will be no possibility of identifying an adequate correspondence between the two. When we talk of beliefs, we are missing the joints: we are not referring to natural kinds that could ever be reduced to neurophysiological states of affairs. Smooth reduction is impossible. So we ought to abandon all folk-psychological explanations and the ontology they presuppose. But opponents have come to the rescue of folk-psychology. Among others, we find those who assure us that it enjoys a certain kind of healthy autonomy from neurophysiology precisely because there is no smooth reduction (Terrence and Woodward, 1985). Moreover, taking folk-psychology to be a body of theory of the same kind as a scientific theory is a mistaken assumption: folk-psychology is not so much a theory as a craft (Dennett 1991).

All these arguments however show some lack of generality because the propositional attitudes do not exhaust all kinds of mental states. Mental states that are not propositional attitudes have been neglected. In this paper I propose to add generality to the debate by considering habits. I will argue that the case for the autonomy and plausibility of folk-psychological explanation becomes stronger when one takes into consideration one example from the non-propositional-attitude mental states, namely habits.

First I must secure the claim that habits are in fact mental states of this kind. Since there is no consensus about the dividing line between the mental and the physical, I will start by examining the similarity between habits and typical mental states like beliefs. Our normal understanding suggests that habits need time to set in. For example, you used to do P in situation Q consciously and intentionally. Then, at a later time after repeated performance, you do P in situation Q without the full consciousness and full intention you used to have before. Your action becomes part of your set-up, and you just follow the rails leading you on: you proceed on automatic pilot. But the important point is that you retain some control. Even when you have a set habit of doing P in situation Q, you still retain the power to readjust your behaviour: the power to consciously, and sometimes with considerable effort, not do P in situation Q. Lacking this possible self-control, you will not be described as having a habit but a compulsion.

If we accept this much about habits, we see that they are dispositions of which we are not always conscious. This does not imply that they cannot be mental states. Even beliefs are not always accompanied by the awareness that we have them. You certainly had the belief yesterday, say, that foxes do not play the clarinet even though you never thought of it before. The fact that we are not always aware of our beliefs shows that beliefs are dispositions to have one thought rather than another. Just as the dispositional character of beliefs does not preclude calling them mental, so also with habits. The dispositional character of habits is however of a certain kind. If we accept that, in general, properties are dispositions, we can say, for instance, that logs have dispositions. But logs cannot wilfully go against their dispositions: they cannot not burn when close to a fire. Humans have dispositions which they share with logs: humans also cannot refrain from burning if close to a fire. But humans also have dispositions of another kind: dispositions over which they retain control, namely habits. A person's habit cannot be one of his physical properties: if his regular behaviour becomes one of his physical properties, that shows that his habit has turned into a compulsion. Habits carry with them the requirement that their bearers can consciously go against them. Hence, they are not physical but mental dispositions. Moreover, they are not mental dispositions of the same kind as beliefs. People do not have a habit that p, but normally of doing P in situation Q. Habits are thus mental states of the same kind as fear and love, which are not attitudes towards propositions but mental states dealing with other entities like persons or situations. In sum, habits are in that special category of non-propositional-attitude mental states.

To deal with habits, radical eliminativists could react in a number of ways. I will mention two possible arguments which are in line with their project outlined above.

(A1) They could argue that talk involving habits should be abandoned because it blinds us with vacuous explanations. Suppose two persons show the same outward behaviour. Folk-psychology would have us explain the similarity by alluding to the same habit the persons have. This move however is hopelessly circular. No proper explanatory role is being played by habits. In fact, to say that the behaviour patterns of two people are the same because they have the same habit is as useless an explanation as saying that a certain potion makes you sleep because it has a dormitive virtue. Conclusion: eliminate folk-psychological habit-talk.

(A2) We may perhaps want to save the word. In line with the radical eliminativist program, we may therefore attempt to reduce the folk-psychological notion of habit to the causal chain of the observed behaviour pattern, as is sometimes done in social theory (Turner 1994, 100). However, this is not a smooth reduction. One major property we usually associate with habits falls. Folk-psychology would have us believe that habits can be shared among many people. But, once neatly reduced to causal chains, habits cannot be so shared anymore. Reducing habits to causal chains obliges us to hold that there are only individual habits. If many people show the same kind of outward behaviour, they cannot be described as possessing a similar kind of entity we call habit. That's old-fashioned folk-psychology and it's wrong: it involves the same kind of error one falls into when thinking that two persons who speak the same language must have learnt it in the same way. The mastery of a language may have very different causal ancestry for different persons. So one should forget about the old-style shareable habits. We should carve at the joints and stick to individual causal chains only.

It seems at this point that extending the folk-psychology debate by including non-propositional attitudes like habits seems to strengthen the eliminativist program. There is however a problem with each of the above arguments.

As regards the first argument, I will start with some general reflections concerning explanation. Higher-order properties can be relevant to explain a certain effect (Pettit, 1993, 32-42; Pettit and Jackson, 1990). It is not wrong to say that the bending of an eraser is explained by its elasticity. The elasticity of the eraser is relevant to its bending because elasticity is related to lower-order properties which are causally relevant, namely to the molecular structure of the eraser. Admittedly, the elasticity brings about the bending in so far as the molecular structure brings it about, and not in virtue of an autonomous power. So the question arises: if the molecular structure has done the explanatory work, why does one need the elasticity? What is the specific contribution of elasticity in the explanation? The contribution is most evident when we are unsure of the molecular structure. The elasticity of the eraser, as a dispositional state, is distinct from the molecular structure: the latter is just one particular realiser state. The explanation in terms of the dispositional state is acceptable because it 'programs' for the bending: it is saying that the eraser has a certain kind of molecular structure and not another.

Consider another example: a glass container cracks when boiling water is poured into it. Why did the flask crack? One simple explanation is: because of the boiling water. It was not because the supporting clip was too tight; it was not because the glass gets more brittle as years go by and this flask was quite old. It cracked because of the boiling water. A lower-order property can also supply an explanation: the flask cracked because of the momentum of this particular water molecule was enough for the breaking of the first bonds in the molecules of the glass surface of the container. Both these answers are relevant. Both give proper explanations. Neither can replace the other completely and neither is vacuous. Saying that the water is in the boiling state is a claim about, among other things, a certain kind of distribution of moving molecules, a distribution that guarantees the fact that there will be some molecule with a momentum sufficient to crack a molecular bond in the surface. The explanation in terms of the dispositional state is not vacuous because boiling is about that particular distribution which ensures the presence of a causal chain leading to the explanandum.

This model of explanatory relevance shows how using habits in explanations of regular behaviour is not vacuous. Replace the dispositional states in the previous examples by the habit, and replace the lower-order state by one of the causal chains leading to the specific behaviour pattern we are considering. Hence, instead of elasticity and instead of the boiling water we now have a habit. Instead of the particular molecular structure of the eraser material and instead of the momentum of this particular water molecule hitting this particular surface molecular bond we now have one of the possible causal chains leading to the observed behaviour pattern. Consider an element of behaviour in need of explanation. Suppose we ask: why does this student always start the day by doing stretching exercises? One straightforward and simple explanation is: because he has a habit. It is not because he is worried about anything in particular; it is not because he thinks his neighbour is using a telescope to watch him to check that he is regular. He does it because he has a habit of doing it. A lower-order property can also supply an explanation: the student behaves this way because he has gone through a causal chain, C1, consisting of, say, the way his mother used to make him wake up early and go to bed early, and the way this was reinforced by the kind of boarding school he used to attend as a teenager, and so on. In line with the program model discussed above, it is plausible to hold that both of these answers are relevant explanations. Neither can replace the other completely and neither is vacuous. When we say that he has a habit, we are saying that one of a particular set of causal chains is operative. We are saying effectively that either C1 or C2 or C3 or ... is in operation.

This shows that it is possible to use habits in proper explanations. In some situations, explanations using higher-order properties like habits are not only possible but even advantageous. I have two main reasons for this claim, especially as regards the use of habits.

First, given the complexity of the causal ancestry of behaviour patterns, it seems plausible to hold that dispositional-state explanations play a more significant role in the habit case than in the boiling water case. When it is behaviour patterns that are to be explained, the intricate causal chain leading to the effect we see in the person's way of acting is difficult to determine, if it can be determined at all. The interaction that life in society involves makes it very difficult to conceive of a determinate causal chain leading to a behaviour pattern. When it is particular empirical events that are to be explained, our scientific models sometimes allow us to consider ideal situations where intruding causes are minimised. Hence, we can easily picture to ourselves a molecule with a certain momentum hitting a particular surface molecular bond. But it is more difficult to isolate a similar picture consisting of the causal ancestry of the element of behaviour consisting of doing stretching exercises on waking up. Hence explanations using higher-order properties become more and more acceptable as the complexity of the explanandum becomes more and more difficult to handle.

Secondly, physics is often seen as a model of what proper explanation should be like. But even explanations in physics do not always give preference to explanation in terms of lower-order properties over explanation in terms of higher-order properties. It is true that most explanations in physics do involve an attempt at reaching an explanation in terms of a lower-order property. Such explanations seem more convincing, and therefore more worthwhile. However, we should recall the interesting and significant change in the status of explanation from classical to modern physics. On one interpretation of quantum mechanics, not all changes require explanation. Discontinuous action, annihilation of elementary particles and the radioactive decay of nuclei are all taken to be basic. They need no explanation. In classical mechanics, higher-order relations could be reduced to lower-order relations. Many statistical relations, like temperature in the kinetic theory, were shown to be equivalent to the causally connected propositions of mechanics. In this case, a higher-order property is simply a term which is used when we are ignorant of all the micro-properties that give rise to the effect we want to study. For the Copenhagen interpretation of quantum mechanics however, higher-order properties are not reducible in this way. Higher-order properties are not terms used when we are ignorant of hidden causal chains that would, if known, give the definite explanation. Higher-order properties are all there is to say: they supply a complete description. What is relevant to say about a radioactive substance is that, after a specific number of years, half of its atoms disintegrate. The question why this atom rather than that atom disintegrates first is considered irrelevant. This shows that explanations in terms of lower-order properties are no longer needed in certain domains. All the explanatory work is done by higher-order properties. The moral here is that even explanations in physics, a discipline which is often seen as a model of what proper explanation should be like, do not always give preference to explanation in terms of lower-order properties over explanation in terms of higher-order properties.

On the strength of this response to the first argument A1, the second argument loses its effectiveness. The autonomy and legitimacy of habit-explanation shows that reducing habits to individual causal chains is wrong. Although two persons may have arrived at their similar behaviour patterns by different routes, their disposition is the same. A multiplicity of different routes of acquisition does not necessarily mean different thing acquired. If, as the above discussion demands, 'habit' refers to the acquired disposition, it makes no sense saying that a token behaviour pattern is caused by the habit. In general, dispositions do not cause anything: the solubility of salt, say, does not cause, but is manifested by, the disappearance of salt crystals in water. We should hence say that the habit is manifested by a token behaviour pattern. Just like all other dispositions, a habit has activating conditions that render its manifestations possible. Hence, to manifest the habit of walking the dog in the evenings, one must, among other things, have a dog. The correct way of talking about habits is in terms of manifestations and activating conditions not in terms of causal chains.

If eliminativists argue in a strategy involving A1 or A2 they cannot succeed. To this extent, by the added generality gained through the consideration of habits, the case for folk-psychology has been strengthened. If you do something often and in a way which appears settled, your folks shouldn't feel old-fashioned when describing you as having a habit.*