Show Summary Details
Page of

What does rationality have to do with psychological causation? Propositional attitudes as mechanisms and as control variables 

What does rationality have to do with psychological causation? Propositional attitudes as mechanisms and as control variables
What does rationality have to do with psychological causation? Propositional attitudes as mechanisms and as control variables

John Campbell

Page of

PRINTED FROM OXFORD MEDICINE ONLINE ( © Oxford University Press, 2016. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Medicine Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: null; date: 18 August 2018

7.1 Introduction

I contrast two causal roles for propositional attitudes. One is that they play a role in our conception of a psychological mechanism: we think of rationally connected propositional attitudes as constituting psychological mechanisms. The second, more fundamental role for propositional attitudes is that they function as control variables in the psychological lives of subjects. These two roles for propositional attitudes come apart dramatically in the case of deluded subjects. We have propositional attitudes as control variables. But they do not figure in rational connections.

Separating these two roles for propositional attitudes makes a difference to how we think of the relation between ordinary propositional-attitude psychology and cognitive-neuroscience approaches to mental illness. Suppose the only causal connections between propositional attitudes are rational connections. Then consider what happens in a case, such as a case of delusion, in which we do not seem to have ordinary rationality. Then, one possibility is that, as Young suggests, the subject’s propositional states are after all broadly rational reactions to sensory states whose occurrence can be explained by cognitive neuroscience (de Clérambault, 1924/2002; Maher, 1974; Young, 2000). Or else we could abandon the appeal to propositional attitudes altogether, and explain the subject’s reactions purely in terms of cognitive neuroscience, as Churchland (2002) recommends.

Suppose, however, we think that there can be causal relations between propositional attitudes that are not rational connections. We can think there are causal relations between propositional states in this sense – that an intervention on one state would have made a difference to the other. A causal connection in this sense need not be a rational connection. There is still a question, though, about the mechanism linking the two states. And here we might find a role for cognitive neuroscience, in explaining what that mechanism is. This is a role for cognitive neuroscience in explaining the mediation between two propositional attitudes that are not themselves rationally linked. In this chapter I will try only to set out the issue a bit more fully.

7.2 Rational connections as mechanisms

Suppose you are an epidemiologist investigating the impact of environmental factors on the risk of contracting some disease. Suppose, for example, that you are looking at whether contaminated water is a risk factor for cholera. You may do observational studies that suggest it is a risk factor, as John Snow observed that cholera cases in Soho were centred round the water pump (Snow, 1855). The critical test for a causal link comes when you intervene on the water supply, for example, by removing the handle on the pump. If that makes a difference to the contraction of cholera, then there is a causal link between the water supply and cholera. Even though that is all it takes to establish a causal connection, there is a further question – that is, how to characterize the mechanism that links the water supply to the disease? It would for a detailed description of the link between water contamination and cholera. But it is not easy to give a description of what it takes to describe a ‘mechanism’ in general. Consider the case of poverty and mental illness. There has long been known to be a correlation between the two. But what is the direction of causation, and what is the mechanism by which causal influence is transmitted? Recently, there was a ‘natural experiment’ in a population of American Indians and others living side by side. A casino opened on the reservation, and there was an income supplement for the American Indians which increased annually. Psychiatrists were at the time engaged in a longitudinal study of the whole population. So, they could look at the impact of this intervention on poverty on levels of mental illness in the American Indian population, using the rest of population as controls. They found that poverty was indeed a cause of mental illness; as a result of the intervention on income, levels of certain types of mental illness fell in the American Indian population. But, the compelling question is, what was the mechanism by which this happened? The question here is formally the same as in the case of cholera; but here the germ theory of disease hardly seems to the point. The investigators proposed that a principal mechanism here – not the only one, no doubt – was that the relief of poverty made it possible for there to be more parental supervision of children, in particular teenage children (Costello, et al., 2003). Of course, this is not a definitive answer, but it is a reasonable conjecture. What is the general notion of ‘mechanism’ that covers both the water–cholera case and this case? It is not obvious that we know how to characterize this notion, essential though it is.

One approach is to do without a fully general description of the notion of mechanism, and instead to try to characterize certain prototypical or basic cases of mechanisms. We might then think that insofar as we have a fully general notion of ‘mechanism’ it is achieved only by a kind of extension or analogy with these basic cases. In the physical case, we might think about examples like the following. Suppose you flip a switch and the light comes on. What is the mechanism? There must be one, even if you are not very sure what it is. Suppose, in contrast, that one billiard ball hits another and the second one moves on. What is the mechanism by which movement was transmitted from one billiard ball to the other? There is no immediate need here to postulate hidden strings or pulleys. The transmission of motion by impulse is one of our prototypical mechanisms. We do not need to postulate some mechanism other than the impact of one ball on the other to explain how it happens that the movement of the first ball causes the second to move. When you move to the level of microphysical analysis, you might give a further breakdown of how this mechanism works. But you are not giving a description of some different mechanism; rather, you are re-describing the communication of motion by impulse.

There is a still simpler case. Suppose I light an oil heater while outdoors. Then, a bit later, I carry the heater indoors. The lighting of the heater happened outdoors. It is now hotter indoors. How did that happen? What is the mechanism by which causal influence was transmitted from place to place? We obviously do not need to postulate hidden strings and pulleys here. The movement of the heater from one place to another was the mechanism by which causal influence was transmitted from the one place to the other. Once the movement of the object is reckoned into account, there is no need to look for further mechanisms. This is a quite general point: the movement of an object from one place to another is the principal mechanism by which causal influence is transmitted from one place to another. This is one of the simplest cases of ‘mechanism’, because unlike the case of billiard-ball collisions, there is only one object here. (Of course I come in as lighting the heater and as carrying the heater, but these are, as it were, supporting roles; the mechanism for the transmission of causal influence from place to place is the movement of the object.)

There is much to be said about these cases, but in this chapter I want to focus on the psychological case. What can we make of the notion of a psychological mechanism? We have already glimpsed the kind of question and answer that might be given here, in the case of the casino on the reservation. But let us look at some much simpler cases. Suppose you say to me, ‘I think Sally would be a good person for this post.’ I say, ‘What makes you think that?’ Your reply is, ‘I saw a sparrow on the lawn last night.’ It cannot stop there. How did you get from one to the other? What was the mechanism? How did seeing a sparrow on the lawn cause you to have this belief about Sally? If there is no answer, if the causation here was genuinely without any further rational filling-in, then we have here a kind of spooky ‘action-at-a-distance’ between your belief and your perception. Suppose, in contrast, you say, ‘I saw Martha, whose judgement I believe to be flawless, and she said we should give Sally the post. So I think we should give Sally the post.’ Here there is no causal question as to how your seeing Martha caused your belief. We do not here have spooky ‘action-at-a-distance’. You and I might of course have an argument about normative questions – whether you should give such weight to Martha’s views, and so on. But there is, on the face of it, nothing similarly puzzling about the causation here. It is like the case of movement of the heater from one place to another. To grasp the rationality of the transition is already to grasp the cognitive mechanism by which one psychological state caused the other.

Let me sum up. The notion of a mechanism is essential to much, perhaps most scientific research. At the same time, it is hard to see how to give a general characterization of it. It is hard because in particular, there are both psychological and physical mechanisms. A cognitive psychologist asking whether traumatic experiences cause depression, for example, might naturally want to pursue the cognitive mechanisms involved. But it is hard to see how to give a characterization of ‘mechanism’ that would cover both this case and the cholera case. So we can instead try to characterize the prototypical psychological and physical mechanisms. I am proposing that we can use the notion of rationality to characterize the prototypical psychological mechanisms. When we have a rational causal sequence of psychological states, we do not need to look for the cognitive mechanism by which our subject got from one state to the other: we have it already. You might say that there must be brain sequences here, which we can try to characterize. But that is not a matter of trying to find a different mechanism here, some different strings and pulleys. Just so, when I move the heater from outdoors to indoors, the mechanism by which causal influence is transmitted from place to place is evident. You might indeed look at the microstructural analysis of the heater. But that is not a matter of you providing a different mechanism, another route by which causal influence is transmitted from outside to inside. You might say that you feel uneasy about the weight that is being put on these judgements about when we do and do not need to find a mechanism. I have a certain sympathy with that uneasiness, but I have to point out to you that you are deeply out of tune with the science of the past 400 years. Science is driven by these judgements about when we have found a mechanism and when we have still to find a mechanism. All I have done so far is suggest that we look at the application of the notion of mechanism to the very simplest cases, and see how it works there. This is a cautious step. Throwing out the notion of mechanism altogether is a radical move we might yet make – but not yet.

The role I have proposed for the notion of rationality is related to, but not the same as, the place that is usually accorded to the notion in analyses of propositional-attitude psychology. The classical view in philosophy of mind is that the ascription of propositional attitudes depends on an assumption that the subject is broadly rational. Davidson thought that the ascription of propositional attitudes depends on a background assumption of the rationality of the agent. Ascribing propositional attitudes is a way of making sense of the agent, and making sense of the agent requires that we discern rationality in the agent. In fact, there is something paradoxical about the very idea that we might find an agent without rationality. We cannot find propositional attitudes without finding rationality.

When thought takes thought as its subject matter, the observer can only identify what he is studying by finding it rational – that is, in accord with his own standards of rationality.

(Davidson, 2004, p. 98)

Now it seems to me that this view has some difficulty with cases of mental illness. Consider the following report of the beliefs of a schizophrenic patient:

A 22-year-old woman had the delusion that thoughts and feelings emanating from her mother’s unconscious were being carried in raindrops that fell on her air conditioner. When the raindrops hit the air conditioner they made a noise, and simultaneously these thoughts and feelings merged with her own unconscious. This merging had resulted in her own mental illness.

(Spitzer et al., 1993, p. 882)

It is possible to argue that even the schizophrenic displaying symptoms is fully rational, and simply illustrates Davidson’s point. But it does stretch credulity that the figure of the schizophrenic could be used to explain what Davidson meant by ‘rational’. The trouble is not even that patient is not rational. We have no idea what a rational way of going on would be, once one has accepted that thoughts are being inserted into one’s mind. How must the world be, for that to happen? Would it make sense to argue with this patient that, by her own lights, it is not the raindrops and the air conditioning that should be blamed, but rather the electrical sockets all around? We have departed so far from the ordinary world that we have no idea what stands fast and what has to go. We do not know what a reasonable way of arguing would look like. (For further argument that deluded subjects cannot be regarded as rational, see Bortolotti, 2005a, b.)

On the other hand, you could argue that the schizophrenic does not have propositional attitudes. (For a sophisticated form of this view, see Currie, 2000. For resistance see Bayne and Pacherie, 2005.) The attitude usually expressed by the phrase, ‘a crazy person’, is something like: we do not have to reckon this person’s propositional attitudes into account. In the present context, though, we do (see Bayne and Pacherie again). The patient does have propositional attitudes, even though there is little rationality exhibited. That ought to be impossible, in Davidson’s view. Dennett provides a blunter expression of the same idea. Would you like to know how to ascribe beliefs and desires to people?

Here is how it works: first you decide to treat the object whose behavior is to be predicted as a rational agent; then you figure out what beliefs that agent ought to have, given its place in the world and its purpose. Then you figure out what desires it ought to have, on the same considerations, and finally you predict that this rational agent will act to further its goals in the light of its beliefs. A little practical reasoning from the chosen set of beliefs and desires will in most instances yield a decision about what the agent ought to do; that is what you predict the agent will do.

(Dennett, 1987, p. 17)

This strategy would not allow us to ascribe propositional attitudes to many of the mentally ill. The trouble is that despite the deficiencies in rationality, many mental patients seem plainly to have propositional attitudes. I will pursue this point in a moment.

On the approach I am recommending, though, there is something highly puzzling about these patients. They do have propositional attitudes, and there are causal relations among their propositional attitudes. But so long as we lack insight into how these relations among propositional attitudes can be seen as rational, we lack insight into the psychological mechanisms involved in the formation of their delusions. There is no contradiction, though, in the idea that the causal patterns in the world may be such that there are no mechanisms to be found to underpin them.

On this analysis, there is an analogy between:

  1. 1 the idea that propositional-attitude ascriptions depends on the ascription of rationality to the subject, and

  2. 2 the idea that all causal interactions between pieces of matter must be comprehensible in mechanistic terms.

Both ideas express an insight – that we find it extremely puzzling when we encounter causal relations among propositional attitudes that are not broadly rational, just as we find it extremely puzzling when we encounter causal interactions between physical objects that are not mechanistic, and that involve spooky ‘action-at-a-distance’. Both ideas express a natural impulse of philosophers – to elevate this kind of point into a kind of synthetic a priori demand that reason makes on the world. This impulse has to be resisted. Delusional patients provide a dramatic example where it looks as though we have causal relations among propositional attitudes that are not rational. Quantum-mechanical phenomena provide dramatic cases where it looks as though we have causal interactions between pieces of matter that require spooky action-at-a-distance.

In the case of delusion as in the case of quantum mechanics, the detailed analysis of the phenomena is not a matter to be settled quickly. I make no attempt to settle either quickly here. The point I am making is that we should not try to settle these cases in advance by appealing to a priori constraints on how the world must be. We should not appeal to the idea that there are a priori constraints on causal relations among propositional attitudes. We have to accept that the propositional attitudes are one thing and the causal relations among them are another. If the propositional attitudes do not conform to rationality, that is puzzling. But we cannot legislate in advance that this cannot happen. Similarly in the case of matter, we should not appeal to the idea that there are a priori constraints on how matter must be. If matter does not conform to mechanistic constraints, that is puzzling. But we cannot legislate in advance that the world cannot be so, that spooky action-at-a-distance is impossible.

7.3 Interventions

One role for propositional attitudes in psychological causation is through their implication in rational sequences, the psychological mechanisms by which causal influence is transmitted. But there is another element to causal role. In an interventionist analysis of causation, for X to be a cause of Y is for intervening on X to be a way of manipulating Y. (An ‘intervention’ on X with respect to Y will have to meet the conditions of an ideal experiment, such as a randomized controlled trial (RCT). For discussion and details see Woodward, 2003; Woodward and Hitchcock, 2003; Campbell, 2007.)

This analysis makes no appeal to any notion of mechanism. It says only that for X to be a cause of Y is for the values of X and Y to be correlated under interventions on X. So in this analysis, for one belief to be a cause of another is for intervening on the first belief to be a way of changing the second. Suppose you believe:

  1. 1 that this man is stroking his chin, and

  2. 2 that this man believes you need to shave.

(For more on such ‘delusions of reference’ see Startup and Startup, 2005.) What is it for the first belief to be a cause of the second? On the interventionist analysis, it is for intervention on the first belief to be a way of changing whether you have the second belief. So if some external force changed your belief that this man is stroking his chin, you would no longer believe that he believes you need to shave. There is no appeal to rationality here, no appeal to mechanisms.

Someone who suffers from delusions of reference in this way is not rational, and we naturally look for some explanation as to why it can happen that belief (1) causes belief (2). This is where cognitive neuroscience accounts might be expected to play some role; but the role is to explain a causal connection that we know already to exist. We want to find the background against which these kinds of causal connections can occur. Let me give another example:

A 22-year-old Rastafarian man of Jamaican parents was admitted from casualty, having superficially stabbed himself in the chest with broken glass. He had become acutely distressed over the past 2-3 days, feeling anxious and depressed and believing that his movements were watched by TV cameras, that signals about him were passed between shopkeepers and that people in shops were talking about him. In addition, he was particularly distressed by the scaly appearance of his skin, which he believed was caused by a lizard growing inside his body, the lizard’s skin being evident on his arms and legs. He gave the growth of the lizard inside his chest as the reason for stabbing himself. He related this to an incident 10 years before when, in Jamaica, a lizard had run across his face. He believed the lizard had ‘left its mark’ and that a curse then had produced his skin lesions.

(Browning and Jones, 1988, p. 766)

You might try to argue that we have here a ‘background of rationality’ against which the delusions make sense. The trouble is that the delusion is so far-reaching in its implications for the patient’s other beliefs that it is hard to see what significant ‘background of rationality’ there is left, once we subtract the delusion. It is sometimes suggested that delusions are broadly rational reactions to unusual experiences that the patient has. This kind of approach is typically proposed only for monothematic delusions, and the present case brings out something of the reason for the restriction (cf. Davies and Coltheart, 2000). The problem is that in this case we seem to know what experience the patient is reacting to. He is reacting to the scaly appearance of his skin. It is hard to see how his reaction can be regarded as broadly rational. This remains true even if we emphasize that the reaction is being said to be only ‘broadly’ rational, and that there may be reasoning biases. This patient is not just unusually suspicious or inclined to jump to conclusions. Nevertheless, there are causal links here between his propositional attitudes. It may well be that this patient is predisposed to form some delusion or other anyhow. But there is a question about the causes of the specific content of the delusion the patient actually forms. What is causing the patient to form a belief about lizard invasion, rather than any other kind of delusion, is his belief that his skin is scaly, and his memory of a lizard running across his face 10 years ago.

Whatever procedure we are using when we ascribe propositional attitudes to this subject, it does not seem that it can be this procedure: ‘first you decide to treat the object whose behavior is to be predicted as a rational agent; then you figure out what beliefs that agent ought to have, given its place in the world and its purpose. Then you figure out what desires it ought to have, on the same considerations, and finally you predict that this rational agent will act to further its goals in the light of its beliefs’ (Dennett, 1987, p. 17). It is difficult to see how use of this procedure could allow you to predict that this patient would stab himself.

I am suggesting that what it comes to that there is a causal link between the propositional attitudes is this: that were there to be an intervention on the cause, there would be a difference in the outcome (cf. e.g. Woodward, 2003). In this account, the existence of causal links between propositional attitudes does not depend on the existence of rational connections between them, and it does not depend on the existence of any mechanism at all to connect them. Scientists will always search for a mechanism, and that is what cognitive neuroscience aims to do. But the causal link is one thing, and the mechanism is another. Suppose we consider the epidemiological case again. Suppose you are asking whether some environment risk factor is a cause of a particular disease. Suppose, as before, that you are asking whether infected water is a cause of cholera. The decisive answer in this kind of case would usually be taken to be provided by a RCT. Here, there is an intervention on the level of consumption of infected water, and we observe whether consumption and cholera are correlated under the intervention. In any particular case, you might ask whether the RCT was carried out correctly. But it hardly makes sense to suppose that the RCT was methodologically unproblematic but there was no causal connection. The causal link is established without any need to find mechanisms linking cause and effect. Whether there is a causal link is one thing and whether there is a mechanism linking cause and effect is another. Similarly, to establish that there is a causal connection between one propositional attitude and another is one thing, and to establish that there is a rational connection or mechanism linking the two is another.

7.4 Control variables

In the case of psychological states and mental illness, there is a question that comes up with some urgency. This is the question whether we are identifying candidate causes and effects ‘at the right level’. After all, you might say, are not all the phenomena of psychiatry fundamentally biological? Are not psychological variables ‘at the wrong level’ to identify the causes of mental illness and its symptoms? One motivation for such an approach is provided by the idea that it is a priori that the ‘mechanisms’ of causation here must be biological and that the ‘right level’ of variable must always be found to relate to the ‘biological mechanism’. There are problems with this idea. It is not a priori that the mechanisms of causation here must be entirely biological. Recall the case I mentioned earlier, of the impact of poverty on mental illness. It is by no means obvious that there is a biological mechanism to be found that links poverty to mental illness. As I remarked, the researchers in this case conjectured that one mechanism linking poverty to illness is that relief from poverty provides the possibility of greater parental supervision. This is not a biological mechanism. There may nonetheless be a causal link between poverty and mental illness. Second, even in cases in which there is a ‘biological mechanism’ linking cause and effect, it does not follow that cause and effect must themselves be identified in biological terms. Smoking is a cause of cancer. The mechanism is biological. ‘Smoking’ is not itself a variable defined at the biological level. It does not follow from this that we are using a variable ‘at the wrong level’ when we say that smoking is a cause of cancer. Similarly, propositional-attitude variables could be causes of psychological outcomes even if the mechanisms linking cause and effect were biological.

How are we to determine the ‘right level’ of variable to use in characterizing mental illnesses? Suppose we had a scanner that would give a complete microphysical description of a human body. This is not even a description at the level of cell biology, it is much finer-grained than that. And suppose we scan thousands of subjects at various stages of their lives, over a period of years. And we observe which of those subjects develop schizophrenia and which do not. We will be able to form a big disjunctive characterization of total microphysical states that are nomically sufficient for schizophrenia. This will not be an exhaustive disjunction – there may well be total microphysical states not on the list that would lead to the onset of the disease. But for all that, it will be informative. Would such a list tell us anything at all about the causes of schizophrenia?

There are a number of problems with this strategy for finding the causes of schizophrenia. In general, when we characterize the correlation between potential interventions on a ‘cause’ variable and the values of an outcome variable, we are describing a function from one state space to another. What we have in the present case is a function from one complex variable – total microphysical condition – to another: risk of schizophrenia. Notice first that this is likely to be a ‘function’ only in the formal sense of ‘a set of ordered pairs’, the first element of which specifies total microphysical conditions and the second element of which specifies risk of schizophrenia. However, the function here is likely to be ‘pathological’ in the sense that it defies concise mathematical expression in familiar terms. In effect, we will have a complex or ‘gerrymandered’ function from one state space to the other. In general, when we look for the causes of a phenomenon, we are looking for variables which do have a ‘dose–response’ relation to the outcome, in that we can find a relatively concise mathematical expression, in familiar terms, of the relation between the state space of the cause variable and the state space of the outcome variable. Similarly, when we try to track the impact of particular genetic variants on risk of schizophrenia, we are looking at relatively simple functions from genetic structure to risk. Even complex models on which multiple genes have weak interlinked effects on risk for schizophrenia have nothing of the unsurveyable complexity of a microphysical approach.

Relatedly, we want our characterization of causes to be a characterization of variables that can be affected by local processes. The total microphysical state of the individual is not a variable that can be systematically affected by any local process. In contrast, variables such as those relating to genetic structure can be affected by local processes. They are, therefore, better candidates for cause variables.

Second, we want changes in the ‘cause’ variable to be correlated with changes in the ‘effect’ variable. To take a simple example, suppose our outcome variable is ‘the state of H2O’, as solid/liquid/gas, within ordinary temperature ranges. You might suppose that the cause variable here is temperature. But that seems to give us too fine-grained a way of characterizing the cause. If we ask why H2O is frozen, one answer is ‘because its temperature is –23°C’. But it does not matter whether H2O is at that particular temperature. The variable we want here is simpler. It has just three values: at or below 0°C, between 0°C and 100°C, and at or over 100°C. Then we say that the cause of the particular state of H2O is the variable having some particular one of those three values. In this sense, we want the ‘cause’ variable to be ‘specific’ to the outcome variable. The sense of ‘specificity’ here is not that the cause variable has an impact only on the outcome variable; any variable will participate in multiple causal processes and have multiple effects. Rather, the point is that we want changes in the cause variable to be correlated, as much as possible, with changes in the outcome variable. And we also have to describe the state space of the cause variable in a sufficiently fine-grained way that we can draw all the distinctions that do make a difference to the outcome variable. So we do not want to say that the cause of the state of H2O is ‘whether the substance is above or below 0°C’, because that does not allow us to draw all the distinctions we need to explain the various states of the substance (cf. Woodward, in press).

Suppose we now go back to the examples I was discussing earlier. These are cases in which there are causal connections between propositional attitudes even though those propositional attitudes are not rationally connected. For example, we have delusions of reference, where a belief that someone is stroking his chin causes me to believe that he thinks I need to shave, or the case in which perception of a skin disorder causes me to believe that I am being invaded by a lizard. For the propositional attitudes to be ‘cause’ variables in these cases, I suggest there should be: (1) systematic relations between the cause variables and the subsequent delusions. Relatedly, local action on the cause variables should be possible, and (2) correlation between change in the cause variables and change in the outcome variables. That is, an intervention on whether one sees the other person scratch his chin, or on whether one believes one has the skin disorder, should make a systematic difference to the content of the delusion.

It seems evident that these conditions for causation could be met, whether or not there are rational connections between the propositional-attitude variables. But still, you might say, the role I am envisaging here for causal relations among propositional-attitude variables is peripheral. It has to do only with the specific content of one’s delusions. It does not explain why one has a tendency to form delusions in the first place.

That is of course correct, and I am not trying to construct a detailed general theory of the causation of psychosis. My point is rather about the role of ordinary beliefs and desires and so on in such an account. The classical philosophical approach has been to regard propositional attitudes as part of a ‘conceptual scheme’ that we bring to bear in describing the ordinary world. This conceptual scheme is taken to have strong a priori constraints on its applicability. In particular, as we have seen, rationality is taken to be a norm with which the scheme has to comply. In this view, we are at ‘at the right level’ to use psychological predicates only when we can ascribe propositional attitudes in such as way as to find the other person ‘consistent, a believer of truth, and a lover of the good (all by our own lights, it goes without saying)’ (Davidson, 2001). In this view, cognitive neuroscience steps in when we cannot find rationality, and eliminates the propositional attitudes in favour of notions internal to cognitive neuroscience.

The appeal I have just been making to the notion of a control variable is intended to replace this invocation of rationality. The question is not whether the subject is rational; the question is whether propositional attitudes figure as control variables in the subject’s mental life. That is, the question is whether the propositional attitudes of the subject meet the conditions I have just indicated, of being locally manipulable and systematically mapped onto specific psychological outcomes. If the propositional attitudes function as control variables in this sense, then we do have a causally functioning mental life, whether or not the subject is rational. Of course, it is true that in the mental life of a broadly rational subject, propositional attitudes function as control variables. But it is the fact that we have control variables, not the fact that we have rationality, which means that we are ‘at the right level’ to talk of beliefs and desires.

We have to acknowledge, as I began by acknowledging, that when we find psychological causation without rationality, we ‘look for a mechanism’, since we usually think of rational connections as psychological mechanisms. When we find psychological causation with no semblance of rationality, we have a kind of ‘action-at-a-distance’ across psychological states that we would like to have explained in terms of some kind of mechanism, even if we are not sure exactly what kind of ‘mechanism’ we are looking for. One recent proposal is that there is early in development a key pathology of the prefrontal cortex.

Disruption of prefrontal control would lead to a pathological increase in emotional responsivity of the subject, which in turn could be mediated via the amygdala’s influence over the hypothalamic-pituitary-adrenal (HPA) axis…

(Broome et al., 2005, p. 30)

This model:

proposes that transition into psychosis is a consequence of primary prefrontal dysfunction leading to secondary enhanced subcortical stress response and dopamine transmission…

(Broome et al., 2005, p. 30)

Dopamine dysregulation in turn may be implicated in difficulties of rational engagement with the world. Centrally, it may affect the regulation of attention and the perception of salience, so that there are problems in integrating information, registering the correct significance of events, and finding whether there are meaningful connections between distinct episodes.

… theories implicating impaired contextual integration and abnormal appraisal on the one hand and dopamine dysregulation on the other may be attempts at explaining the same processes at the different levels of information processing and neurochemistry, respectively.

(Broome et al., 2005, p. 30)

What I have been proposing is that we can think of this kind of account as explaining how it comes about that the rationality of the subject can be undermined. In this reading, the role of cognitive neuroscience is not to displace the appeal to propositional attitudes, conceived as governed by an a priori constraint of rationality. Nor is it to supplement the appeal to propositional attitudes, conceived as causally related to one another only by rational mechanisms. The role of cognitive neuroscience is, rather, to describe the mechanisms that make it possible for there to be causal connections between the propositional attitudes in the absence of rationality.


Thanks to Ken Kendler for many many hours discussing these topics. Thanks also to Lisa Bortolotti and Matthew Broome for their comments on an earlier draft.


Bayne, T. and Pacherie, E. (2005). In defence of the doxastic conception of delusions. Mind and Language, 20, 163–188.Find this resource:

Bortolotti, L. (2005a). Intentionality without rationality. Proceedings of the Aristotelian Society, CV, 385–392.Find this resource:

Bortolotti, L. (2005b). Delusions and the background of rationality. Mind and Language, 20, 189–208.Find this resource:

Broome M. R., Woolley, J. C., Tabraham, P., et al., (2005). What causes the onset of psychosis? Schizophrenia Research, 79, 23–34.Find this resource:

Browning, S. M. and Jones, S. (1988). Ichthyosis and delusions of lizard invasion. Acta Psychiatrica Scandanavica, 78, 766–767.Find this resource:

Campbell, J. (2007). An interventionist approach to causation in psychology. In Causal Learning: Psychology, Philosophy and Computation (eds. A. Gopnik, and L. Schulz), pp. 58–66. Oxford, Oxford University Press.Find this resource:

Churchland, P. (2002). Eliminative materialism and the propositional attitudes. In Philosophy of Mind: Classical and Contemporary Readings (ed. D. Chalmers). Oxford, Oxford University Press.Find this resource:

    Costello, E. J., Compton, S. N., Keeler, G., and Angold, A. (2003). Relationships between poverty and psychopathology: a natural experiment. Journal of the American Medical Association, 290(15), 2023–2029.Find this resource:

    Currie, G. (2000). Imagination, delusion and hallucinations. In Pathologies of Belief (eds. M. Coltheart, and M. Davies), pp. 167–182. Oxford, Blackwell.Find this resource:

      Davidson, D. (2001). Mental events. In Essays on Actions and Events (ed. D. Davidson). Oxford, Oxford University Press.Find this resource:

      Davidson, D. (2004). Representation and interpretation. In Problems of Rationality (ed. D. Davidson), pp. 87–99. Oxford, Oxford University Press.Find this resource:

      Davies, M. and Coltheart, M. (2000). Introduction: pathologies of belief. In Pathologies of Belief (eds. M. Coltheart, and M. Davies). Oxford, Blackwell.Find this resource:

        de Clérambault, G. G. (1924/2002). Psychoses with base of automatism (Part 1). In Mental Automatisms: A Conceptual Journey into Psychosis Commentaries and Translation of the Work of Gaëtan Gatian de Clerambault (ed. P. Hriso, (2002)). Bayonne, NJ, Hermes Whispers Press.Find this resource:

          Dennett, D. (1987). True believers. In The Intentional Stance (ed. D. Dennett). Cambridge, MA, MIT Press.Find this resource:

            Maher, B. (1974). Delusional thinking and perceptual disorder. Journal of Individual Psychology, 30, 98–113.Find this resource:

            Snow, J. (1855). On the Mode of Communicaton of Cholera. London, John Churchill.Find this resource:

              Spitzer, R. L., First, M. B., Kendler, K. S., and Stein, D. J. (1993). The reliability of three definitions of bizarre delusions. American Journal of Psychiatry, 150, 880–884.Find this resource:

              Startup, M. and Startup, S. (2005). On two kinds of delusion of reference. Psychiatry Research, 137, 87–92.Find this resource:

              Woodward, J. (2003). Making Things Happen: A Theory of Causal Explanation. Oxford, Oxford University Press.Find this resource:

                Woodward, J. and Hitchcock, C. (2003). Explanatory generalizations, part 1: counterfactual account. Nous, 37, 1–24.Find this resource:

                Woodward, J. (In press). Cause and explanation in psychiatry: an interventionist perspective. In Philosophical Issues in Psychiatry (eds. K. Kendler, and J. Parnas). Baltimore, MD, Johns Hopkins University Press.Find this resource:

                  Young, A. (2000). Wondrous strange: the neuropsychology of abnormal beliefs. In Pathologies of Belief (eds. M. Coltheart, and M. Davies). Oxford, Blackwell.Find this resource: