Psychologists such as Uta Frith and John Morton have recently attempted to clarify the, or at least a, framework constraining the reasoning deployed in certain of the models and theories found at the psychological end of today's cognitive neurosciences. They are interested, in particular, in cognitive development and developmental disorders. My interest here is, however, more epistemological than developmental or psychological. What I want to do is to start to elucidate a distinction between two types of explanation in psychology, and then suggest that what goes by the name of 'causal modelling' typically conflates these two types of explanation. (Personal background: irritation with the way in which psychologists mistake their ability to conceptually distinguish between cause and correlation as the resting point for the requisite reflective understanding of causation...)
Two Types of Explanation
We are looking out of a small window. We see a branch break off the tree in front of us. But we don't see what caused this to happen. Perhaps it was the wind; perhaps it was someone sawing out of sight near the trunk; perhaps it was a woodpecker; perhaps it was a big fat squirrel sitting out of sight on the end away from the trunk. Perhaps it was all of these at once. Let's call the search for the reason for the branch breaking: "looking for the efficient cause of a happening".
We are looking out of another small window. And we see some letters from the end of a banner moving past. We can't see the whole of the banner; just the end of it. The letters are:
And we wonder what the banner is saying. What is the pattern that completes? What will we see when we go outside? In a Robert Graves moment, stuck in our room, we have two hypnagogic hallucinations. The first reads:
Unlike Graves, we don't have to base our hypothesis on the strength of the imagery: We go outside and discover a large processional advert for Morton's book. So, let's call the search for the particular type of explanation required here the search for the "pattern that completes".
So what are the differences between these two types of explanation?
And here is what I want to say about the second: It tells us more about what it is that we are seeing. We come to an understanding of what it is, an understanding that makes sense of something in terms of something which already makes sense to us. We might not know why someone is carrying this banner, why it happened that the banner went past the window, or how the letters got printed on it. There is then at least one sense in which coming to see what else is on the banner does not explain why the letters went past the window. We are told more about the phenomenon itself, and not more about its origination. The phenomenon is now one which we find intelligible: we now find the particular letters intelligible in terms of the kinds of things we already have experience of. If it had turned out that there was just nonsense written on the banner, with these letters at the end of the nonsense words, our finding out this extra information would not have made what we saw intelligible in the way that our finding out that it was part of an advert for Morton's book made it intelligible.
I hope this is enough to be getting on with for now. The one makes an event intelligible by informing us of its precipitants or 'efficient causes'. Call this clearly 'causal'. The other makes for intelligibility by situating a particular within an intelligible context. We now 'see what' it was that was there. In the first, we know how something happened; in the latter, we understand better what it was that happened. (Of course, events may be redescribed in terms of their causes, and I am relying on your not finding it just obvious that this is just what an intelligibility-conferring, pattern-completing, type two, explanation amounts to.)
And I hope now that others will find it acceptable enough when I claim that to be told that it was 'explaining developmental disorders' rather than 'campaigning animal protesters' that went past the window is not to be given a causal explanation of the occurrence of the letters I saw. In this latter case, my puzzlement and my inquiry just was not such as to inspire the provision of a cause by way of an answer. What I needed was more phenomenology, and not more aetiology.Causal Modelling in Psychology
So what I shall be claiming is that 'causal models' in psychology can sometimes run these two senses together. That is, they portray different sorts of explanation as causal. And this can lead to epistemological confusion. Morton tells us that his own strategy is a 'causal modelling approach'. He is interested in the way in which developmental disorders develop, and his general strategy is to posit cognitive processes as causal intermediaries between biological abnormalities and phenotypic behavioural expressions.
So let's look at an example from his book (well worth a read by the way, and in many ways a model of clarity).
Here are my objections. First, the model is presented as if the relation between the 'biological' and the 'cognitive' were the same kind of relation as that between the 'cognitive' and the 'behavioural'. Second, and relatedly: the suggestion is made that the cognitive factors are 'inferred entities', and that the form of our understanding here is, accordingly, inference to the best causal explanation.
Here is what Morton says regarding his three levels of description (p. 22):
The one substantive assumption in the framework that I will present to you is that what I call the cognitive level has a major role to play in the causal chains of interest to us. This is implicit in many of the diagnostic descriptions that we see. In the framework, the cognitive is made explicit. The reality of cognition makes it clear why people get so confused when they try to map biology straight on to behaviour.And this is what I am objecting to: That the significance of the cognitive level, and the confusions that result when we try to map biology straight onto behaviour, are a function of the cognitive factors occupying a 'causal intermediary' function between, in this case, brain and behaviour.
For this is how things seem to me: When I want to understand relations between cognitive factors, or when I want to understand 'the reason for' some bit of behaviour, what I am searching for is far more akin to a Type 2 explanation than a Type 1. Morton tells us that a cognitive factor is (p. 21) 'far more than a redescription of the behaviour from which the idea (of the cognitive factor in question) sprang.' This seems doubtful to me, although much will turn on what we mean by 'redescription'.
Consider the genetics to brain relations in Figure 3.14. They are surely causal, Type 1, relations. So too, I would say, are the effects of the brain differences on the resulting cognitive structures. It is not that the inner intelligibility of the cognitive factors is elucidated by the brain dysfunctions. Rather, we see how they originated, what the efficient causes of the cognitive factors are, when we appreciate their putative biological origins. But the relations between theory of mind and unlearned social conventions, or between theory of mind deficits and imaginative deficits, is to my mind quite different. So too are the relations between, say, lack of imagination and no pretend play.
Consider this latter example. We see a child who is not pretend playing, in a situation which we might otherwise expect a child to be pretend playing, and we wonder why. Perhaps they are very anxious. Perhaps they have no imagination. Perhaps they are glued to their seat, or are lazy. According to Morton, these are to be considered 'causal explanations' of the lack of pretend play. But to my mind, they are far more analogous to Type 2 explanations. We discover what is going on for someone, for example, not when we independently identify a lack of imagination quite independently from all of its alleged manifestations. Rather, we discover it when we come to see the pattern which obtains in their life; when we see a whole host of their behaviours under the aspect of 'imaginal impoverishment'. When we see that they are otherwise acting without anxiety or boredom or glue. This is no straightforward redescription of the one particular behaviour to be sure. It tells us a lot more. It provides a pattern which explains through the particular kind of intelligibility - and not efficient causality - which pertains to the cognitive. We see similarities, aspects, instantiations, immanence. What we don't see are relations between isolable causes and isolable effects. The 'inference' to lack of imagination is not an inference to something which completely transcends the behavioural data (lack of pretend play) confronting us, nor of course is it simply reducible to it. What we see is rather the meaningful pattern which connects.
The case is surely even more obvious when we consider the relation between 'low general ability' and 'low IQ'. In what way could a low IQ be said to be 'caused by' a 'low general ability'? The concept of an 'ability' is obviously dispositional - obviously logically and not merely empirically related to the behavioural. If someone has a low general ability then it follows deductively that they will have a low IQ (or am I missing something?). Now this deductive character cannot be said to obtain for most psychological phenomena (to the behaviourist's dismay, no doubt) - and if this were not the case it wouldn't really be obvious why we would even have a psychological vocabulary in the first place.
A pecularity of Morton's book - of pretty much his entire approach - as it seems to me is the purely 'downward arrow' trajectory from biology, to cognition, to behaviour. Leaving aside my above epistemological qualms, I'm surprised by the lack of a dynamic systems perspective, given the way in which such theories have typically displaced the linear approach in the last few years (but cf ch. 11). Furthermore, development does not normally occur in social isolation from others (it is not simply maturation). Building these two factors in will lead us to models which are dynamic (i.e. with feedback loops within them, within and between levels) and systemic (involving interaction back and forth with the environment), rather than static and linear. Morton's approach would only seem to work for disorders which stem in an isolable way from biologically generated disturbances in cognitive function. (cf Stuart Shanker (2004) Autism and the Dynamic Developmental Model of Emotions, Philosophy, Psychiatry, & Psychology, 11, 3, 219-233 as an example of a different approach).