Thursday, November 10, 2005

Bulletin of the British Psychological Society 25 (1972), pp. 15-18


A PSYCHOLINGUISTIC ACCOUNT OF CAUSALITY






JOHN J. RAY

University of New South Wales

Although the layman might reasonably expect that the work of the psychologist should lead to an increase in our knowledge about the causes of behaviour, psychologists themselves at the present time are most likely to plump for the more modest goal of trying to predict behaviour. Statements of the form 'x causes y' are viewed with what is very often a fully justified degree of suspicion. For all that, few would want to deny that 'finding the causes of things' is a large part of the job that, in principle, science purports to do. There does then seem to be something of a gap between the theoretical goals of scientific inquiry and the actual acknowledged goal of many particular scientific inquiries. Part of the reason for this gap is no doubt some feeling that the standards of proof needed to back up a statement of causation are much higher than the standards needed to back up a statement about constant conjunction.

And yet the influential analysis of Hume asserted that a statement about causality is nothing but a statement about constant conjunction (Hume, 1777). In a modified form this account is still widely accepted among philosophers (e.g. Reichenbach, 1951). There are, of course, many ways of stating the case which do require us to acknowledge that the goal the psychologist appears to have when he says he only wants to predict behaviour is in fact more modest than the goal of finding the causes of behaviour. Nonetheless, it seems at least possible that the classical Humean attack upon 'necessary connection' as a feature of causation is one of the reasons for our reluctance to talk about causes. Some clarification of what talk about causation entails therefore seems timely -- doubly so, because work in psychology itself has enabled a fuller account than has heretofore been possible.

To recapitulate, then, Hume may be summarized as analysing the notion of cause into the two notions of temporal priority and necessary connection. Contiguity, too, he detects to be in some sense a feature of cause. Although apparently willing to grant that we have a notion of necessary connection, he is unwilling to grant that there is a necessary connection between an effect and its cause. On this view, then, it is sufficient grounds to say X is the cause of Y if X always precedes Y and if X and Y are, in some sense, contiguous. A notion of necessary connection between X and Y is, on the other hand, just a little something we add on when speaking or thinking of the conjunction.

This rejection of necessary connection by Hume stems from a real difficulty in finding out what it might consist of. The connection is either an event, a relationship or an object. If it is an object it is either physical or, perhaps, non-physical. Whichever is the case it is difficult to see what could be so versatile as to fit in 'between' each and every instance of cause and effect. Fundamentally, it is the purpose of this paper to suggest wherein this 'connection' might reside.

Some initial groundwork needs to be laid first in developmental psychology. William James is often quoted as saying that the world of the newborn child is a 'blooming, buzzing confusion' -- a chaos out of which particular things are only slowly differentiated. In varying ways, this view is widely accepted by psychologists in their treatment of perceptual development-psychologists as diverse as Hebb (1949) and Piaget (1926).

The second consideration needed here is the work of Whorf (1966) on linguistic relativity. Whorf shows that the way we differentiate our world is more arbitrary than had been supposed. Different environmental pressures will result in a cultural heritage of differentiations between objects and between events that is adaptive for the particular group of people concerned. These differentiations will be reflected in the language of that tribe or culture. The colour spectrum is a popular example of this process. We break up what is essentially a continuum into categories of red, blue, green and yellow, etc. Some primitive tribes have only one word for what we call blue and green. They just do not attend to differences between these wavelengths. It has never been adaptively important for them to do so.

It is easy to extend the above analysis to events. We speak of events such as my striking the golf ball and the golf ball flying through the air, but is this differentiation in fact anything other than convenient? Surely it would be just as correct (though perhaps less convenient) to speak of a single process of motion which was at first exhibited by the golf club and later in the golf ball. Pursuing the point even further, what is necessary about the categories of striking and flying ? It would be just as correct (though less useful) to say: 'There was an event of physical objects in motion.' We cut up the happening going on around us in non-necessary ways. We could cut up our club-ball happening into the downward motion of the object and the upward motion of the object. If we did this we would say that the downward motion of the object plus some other centripetal vector X caused the upward motion of the object. It might be of no interest that in the upward motion the object (as attended to) was first the club and then the ball.

What is asserted then is that we cut up the 'happening', the 'goings on' or the 'process-stream' about us in non-necessary ways. That part before the cut we call the cause and that part after the cut we call the effect. The very notion of cause is itself an arbitrary supercategory. Whorf reports that some Indian tribes use no such category.

Let it be stressed at this point that the above account is not dependent upon that extreme version of the process-stream argument which asserts that there is nothing about the ontological given which predisposes people to use one categorization rather than another. It is simply asserted that pragmatic usefulness must be taken into account in explaining any particular categorization. It would seem apparent, moreover, that (insofar as there is any distinction between objects and events) the categorization of objects might be much more dependent on the ontological given than the categorization of events (which could, indeed, be almost entirely pragmatically based). Note, however, that the whole distinction is only an approximate one. In the Hopi Indian language the word for table -- one of the philosopher's prime candidates for an example of an object -- is conjugated as a verb! This example should also make us less assertive about the degree to which the ontological given impels certain categorizations. Whatever the differential impact of the raw ontological given, it does seem at least necessary on Whorf's evidence to assert that pragmatic considerations must always have a potentially overriding influence in determining what categorizations are actually used by a particular tribe, culture or linguistic group.

The above does not, of course, imply that two people from different cultures must fail to agree on what object or event is being attended to, though each might be puzzled by an apparent tendency in the other to include 'irrelevancies' under the one term. Eventually, of course, they will be able to compromise to the point where they will agree entirely. We can be sure this is possible because the ontological given really is there. The categorizations we apply do not in any way alter what is given.

We could now, if desired, replace 'necessary connection' by 'essential connectedness' where 'essential' merely means 'in the essence' (i.e. the ontological given). This formulation is, however, still not entirely satisfactory (except perhaps as an obeisance to Hume). Any use of 'connect' implies discreteness. 'Essential continuity' would be better. It is, in short, denied that discreteness is an inevitable part of an accurate and useful description. Paradigmatically, to say that X is the cause of Y is to say that X is the or a temporally prior category of what is also perceived as one event. Psychologically, our imputation of connectedness lies then in our ability to see that two events are in fact one or, more exactly, that one event has been made into two. To perceive X as causing Y is to recognize the larger event F. There is, of course, nothing sacred about the (often unnamed) larger category called here F It might, for instance, be characteristically more useful to group Y and Z as G rather than X and Y as F.

This explanation clears up rather parsimoniously some of what would be difficulties for a simple temporal priority account of cause and effect. Take the example of a water pump. The cause of the water rising (effect) is the upward motion of the piston in the cylinder. But surely both of these, in fact, go on at once. Do we have here an instance of cause and effect being concurrent? The very fact that we can ask such a question shows some weakness in a simple temporal priority account of cause. If, however, a cause is seen as simply a convenient category within one event, there is, of course, no difficulty.

Note that this example also shows clearly how in this culture we tend to divide up single events in terms of what we have already decided their physical object membership to be.

It is, of course, assumed that a pragmatic and hence changeable base for any categorization does not prevent that categorization from being repeatable in the case where a similar ontological given is again present. Although offering no particular difficulty, this assumption about human abilities is mentioned because it is one not required by the naive view that our categorization of objects and events is in some sense 'natural'. We could not, of course, speak of a cause if the particular category were not repeatable. To recognize a causal relationship then is to recognize that a particular recurrent stretch of the process-stream is such as to be easily seen as a single event.

It will be seen, then, that this paper agrees with Hume in denying that there are any necessary connections. The justification offered here (that there is in fact no necessary discreteness) is the point of difference. Along the way, of course, this paper also eliminates Hume's difficulty of explaining from whence it is that we get the idea or feeling of necessary connection. Such an idea is shown to be in fact the very idea of a thing or event having a separate identity of its own. It is not desired to assert that we always get the idea of necessary connection from the idea of an event or object having particularity but the idea of necessary connection is the idea of what could also be perceived as elements being in fact perceived as a whole. The idea of necessary connection is then an idea of a wholeness or of particular identity.

In conclusion, let us try out the above account on a possible criticism of it.* Someone might say: 'I have no more difficulty regarding the falling of the rain outside my window and the growth of plants inside on the windowsill as one event including them both than I have in regarding the falling of the rain and the continued growth of the lawns on which the rain falls. But I regard the former pair as not causally connected, and the latter pair as causally connected. Therefore is not at least the account of the idea of causal connection inadequate ?'

The answer here is relatively simple: if the falling of the rain outside did continue to go with the growing of the plants inside, we would probably tend to believe that they were part of one larger event. The account here does not dispense with the need for constant conjunction. It merely makes constant conjunction a necessary member of a (variable) set of conditions for perceiving what had been priorly thought separate events to be in fact part of a larger event.

If the constant conjunction were in fact established, then we might, in fact, assume that there was no essential discreteness between what had apparently been two events. We might, in fact, look for some intervening process to 'join' them (e.g. seepage of moisture and the nutrient qualities of my woodwork).

Perhaps another point inherent in the above criticism is a simple denial that either of the two sets of events can in any way be 'seen' as one event. The point to be made in reply is that any event can in principle be infinitely subdivided and we do not have (or need) words for every possible category. If we are dealing with a category for which there is no word, it is not surprising that we might tend to feel that that category is different in type from those for which we do have single words. There is, of course, a reason why we have words for only a few possible categories-a pragmatic reason having at least a great deal to do with frequency of usefulness.

On this line of reasoning any 'strangeness' we feel about regarding something as one event is an outcome of our not having one word for it. It also follows that where there is a very common or frequent cause-effect sequence we might well expect that there would be a word for it. There are, of course, many such instances.

When we speak of a man firing a rifle the same event might just as accurately be described as: the man pulled the trigger -- the firing pin struck the percussion cap of the round -- the percussion cap did whatever percussion caps do --gases were produced inside the round -- the bullet was expelled through the barrel, because we can see this common set of descriptively discrete events as one continuous event we also have one word for it in recognition of the fact. We might not know all the above details but at least we know that the man did something to the rifle which caused (sic) the bullet to be expelled.

We seek causes, therefore, by looking for connections between one event and another. These connections are not the metaphysical necessary connections treated by Hume but simply other, more minute, intervening events. We stop looking when we have found enough to convince ourselves that we are in fact dealing with one continuous event.

The implication of the foregoing for psychology then is that insofar as they concentrate their inquiries upon 'input-output relationships' and other types of constant conjunction, psychologists will never succeed in justifying causal inferences. More reductionistic types of inquiry (e.g. psychophysiology) will be needed before the 'other', more minute, intervening events' mentioned above can be identified. Insofar as they have not had such knowledge, it is little wonder that psychologists have so far eschewed much talk of causes. It was evident to them that despite the legitimation furnished by the Humean account for a concentration on mere conjunctions of events, this could not enable more than prediction.

Disputes over determinism also take on a new look when the above account of causality is used. The fundamental question 'Do all events have causes?' becomes equivalent to 'Are all events part of some larger event?' If our categorization of raw 'happening' into events is done partly on pragmatic grounds, the answer to this question does not lie entirely in the field of objective inquiry.


Acknowledgements

*I owe this to Max Deutscher.

Earlier versions of this paper were criticized by Dr Robert McLaughlin and Professor Max Deutscher. I would like to thank them most sincerely for their help, but all responsibility for errors and deficiencies remains my own.

References

HEBB, D. 0. (1949). The Organization of Behaviour. New York: Wiley.

HUME, D. (1777). Enquiries Concerning the Human Understanding and Concerning the Principles of Morals. Oxford: Clarendon Press, 1963.

PIAGET, J. (1926). The Language and Thought of the Child. New York: Harcourt, Brace.

REICHENBACH, H. (1951). The Rise of Scientific Philosophy. Berkeley: Univ. of California Press.

WHORF, B. L. (1966). Science and linguistics. In E. Maccoby et al. (eds.), Readings in Social Psychology. London: Methuen.


FINIS

No comments: