Charles Taylor, amongst others, has tried hard to articulate the different forms of understanding that all exist within the umbrella discipline of 'psychology'. The aim of this is to promote a 'peaceful coexistence'. The idea is that, once we recognise that our colleagues are set about doing different jobs with different (conceptual, theoretical) tools, we will not accuse them of doing our own job badly with the wrong tools.
I would like to add my own tuppenceworth. Here are the two forms of understanding I frequently see confused in psychology:
1. The first aims to provide forms of understanding. It uses examples - case studies for example - to illustrate the intelligible possibilities it discerns. The aim of the enquiry is the discernment, the elucidation, of these possibilities. It hardly matters to this form of understanding whether there is more than one instance of the intelligible mode of personal being that it uncovers, although it would as a matter of fact often be pretty pointless in broadcasting the understanding if it were not one one suspected was encounterable elsewhere.
2. The second does not aim to elucidate new forms of understanding, does not intend to render intelligible or comprehensible what was previously barely visible or, if visible, not understood. Instead it aims to chart whether what is already understood actually happens, and if so, with what frequency, and in what circumstances. Psychoepidemiology is one way to understand it, with the populations in question being fairly restricted.
1. The first form of understanding has much obviously closer links with its philosophical forefathers. It is exemplified by both phenomenological and psychoanalytical approaches. A completely inappropriate reaction to an enquiry carried out in its spirit would be 'But what is your actual evidence for saying that?!'. Yet that is the reaction that is often unwittingly solicited from the pundits of (2.) empiricist psychology. (A case of general irrelevance may of course stick, depending on the case.) Misunderstanding itself, it may also offer empirical generalisations for which it has no evidence (e.g., Freud: Paranoia is due to repressed homosexuality. He should have said: I will show you how paranoia can intelligibly be understood as due to repressed homosexuality.)
2. The second form of understanding has much obviously closer links with the natural sciences. Its procedures and products may sometimes be unhelpfully criticised as unenlightening by pundits of an intelligibility-demonstrating psychology (1). Unhelpfully, because the research is not best understood as aiming at illumination, at the finding of meaning - but only at, say, the cataloguing of such meanings as are already understood. (A charge of boringness may still stick of course, depending on the case.)
In what follows I want to consider the question of the viability of causal explanation in psychology. I think it is often taken for granted that we can have causal explanation so long as we see it only as a type (2) understanding. But this, it seems to me, is a mistake; not because it can operate more widely, but because it is often not properly applicable in either case. Attempts at causal explanation in psychology seeem to me to be frequently born out of a particular conflation of the two forms of understanding.
Let me be clear, there do seem to be some good instances of causal explanation in the ambit of psychology. Certain neuropsychological explanations for example are clearly, and to my mind unproblematically, causal in nature. (The reason John can no longer respond to visual stimuli is because of this cerebral insult which damaged this part of his brain which prevents 'information' from getting from the retinae to the striate cortex, etc.) I don't feel I have any securer grip on the notion of causality than I do on the idea that these kinds of explanations are paradigmatically causal. (It is however interesting to consider whether the causality doesn't belong primarily to the neurological, rather than the psychological, side of the explanation.)
Consider the kind of ethological aetiologies provided to us by attachment theory. John has personality (problems) X in adulthood because when he was an infant and toddler he developed a certain attachment style Z to a mother who provoked this attachment style by her own personality style Y. Is X causally related to Z? Is Z causally related to Y? These may seem obviously truistic, but I wonder.
To take them in order: Z seems to be primarily related to X through its sharing of X's structure. The pattern is an intelligible development. Take a small spiral shape which is oval in general shape. Now take a larger oval spiral with further loops to it. We see how the latter embodies the form of the former. But do we say that the former shape caused the latter shape? Imagine that it had not changed - that the person had remained with an undeveloped, identical, personality. Would we say that the former small spiral (to pursue the analogy) has caused the later small spiral? To take an analogy from Roger Squires, would we say that the curtains' being red today causes the curtains' being red tomorrow? (Er, the answer is: No we (I at least) wouldn't.) Causation has to do with change, not with stasis, and what caused the change is the developmental millieu.
Is Z (child's attachment style) causally related to the mother's personality (Y)? Here I think we need to distinguish two different factors. On the one hand, what is important is just the sheer comprehensible intelligibility (1) of one given the other. But as well as this, there is surely a causal story. The causal story has to do with whatever it is that makes responses sediment into the personality. A causal story in terms of conditioning may be appropriate here.
So the moral is not that causal explanation is irrelevant, only that it comes in at certain junctures as does not by any means provide the typical form of understanding offered to us in psychological contexts. Now I want to turn briefly to some assumptions that seem to come along all too readily in discussions of linear regression analysis.
Such analyses, often used in empirical psychology, enable us to examine correlation between two or more variables (e.g. smoking behaviour and lung cancer) by random sampling from a population. The warranted criticism that correlation does not necessarily indicate causality is well known, and I shan't be making it here. Instead I want to suggest that (a typical maker of) that criticism often (him- or herself) buys into further assumptions which are not warranted. For example, we do of course do well to avoid (what is nevertheless prevalent) talking prematurely of 'dependent' and 'independent' variables, since this just inclines us to presuppose what, from the statistics alone, is completely unwarranted, that the independent variable is a cause of the dependent ones. But we must also, I believe, be wary of the implicit idea lying behind the making of the criticism: that the suggested assignment of causes and effects is not the only one, since either we may have them the wrong way round, or they may both be effects of a common cause.
What seems misguided to me about this is just the implicit idea that the relationship we are considering must be considered in causal terms at all. Let me give what is hopefully a sufficiently ridiculous example. Consider the relationship between smiling and reports of happiness. Bigger smiles, it is discovered, tend (albeit with fairly large so-called 'estimated error' values) to correlate with greater amounts of self-reported happiness. It would clearly be a mistake to suggest that this necessarily shows that happiness causes smiling, or (in a James-Langian moment) that smiling causes happiness. The two variables here are not 'smiling' and 'happiness', but 'smiling' and certain 'self reports'. And these, we may imagine, do not seem to cause one another.
The temptation I wish to diagnose, however, is that of saying that 'smiling' and 'self reports of glee' are therefore to be understood as effects of a common cause: happiness itself. Happiness is thereby reified, denatured by being hived off from its constituting behaviours and expressions, and posited as a cause of what is now seen as extrinsic to it (smiling etc.).
What might explain the correlation between smiling and the self-reports, though, if not a common cause? Well first I have not denied that they have a common cause (perhaps there is a neurological one); only denied that they have a psychological common cause ('happiness'). Second, that such expressive behaviours fall together as they do is a condition of the possibility of the meaningful deployment of terms such as 'happy'. This, however, is a logical, not a causal, condition!
I said above that I felt that attempts at causal explanation in psychology were often the product of conflating and misunderstanding two types of psychological understanding. Regression and factor analytic treatments of psychological topics often seem to perpetrate just such conflations and misunderstandings. There is of course nothing wrong with the techniques, just something wrong with the interpretation of their findings. Such interpretation seems to suggest that correlations are to be understood in terms of causality - either immediate or derived (from a common cause). A psychoepidemiological-style investigation (2) shows the frequencies of certain behaviours as they occur in certain situations. An intelligibility-finding-style investigation (1) shows, say, how one psychological phenomenon can be understood as an intelligible (not a causal!) function of another. Mistake the latter as providing causal explanation, and take the former as correlation, merge them together, and it seems as if we have causal psychological explanation.