Inconsistency and interpretation by Lisa Bortolotti | Papers by Lisa

This is the unedited, uncorrected final draft of a paper which was published in Philosophical Explorations VI(2), May 2003, 109-123. LISA BORTOLOTTI INCONSISTENCY AND INTERPRETATION 1. Introduction In this paper my purpose is to examine whether the case of inconsistent believers can offer a reason to object to theories of belief ascription that rely on a rationality constraint. I shall first illustrate how the possibility of inconsistent believers might be a challenge for the rationality constraint and then assess Davidson's influential reply to that challenge. The structure of the objection from inconsistency is the following. There is some consensus on the idea that believed inconsistencies are violations of rationality. The rationality constraint tells us, depending on formulations, that we have to assume that a system is rational if we want to ascribe beliefs to that system, or that we can ascribe to a system only rational beliefs. If the rationality constraint holds, then inconsistent systems cannot be belief systems, or we cannot ascribe inconsistent beliefs to any system. But we do seem to ascribe inconsistent beliefs at times. So there must be something wrong with the rationality constraint. According to Dennett's intentional system theory, for example, it is impossible to ascribe to a system any pair of inconsistent beliefs, or better, the phrase 'inconsistent beliefs' does not make any sense. We can talk about inconsistencies, but not about believed inconsistencies (Dennett, 1981, p. 95). Within Davidson's theory of belief ascription, the idea of a speaker entertaining inconsistent beliefs does make sense. His rationality constraint requires that no system be ascribed a large number of inconsistent beliefs or an obvious inconsistency. In this context, when I talk about a believer engaging in an obvious inconsistency, I follow Davidson and refer to the case of someone who explicitly endorses the conjunction of two contradictory statements and is aware of the 1 inconsistency as such. Given the greater flexibility of Davidson's views, in what follows I shall concentrate on his reply to the objection from inconsistencies. In spite of what interpretationists might say, it looks as if we happen both to hold and to ascribe to others pairs of beliefs that are mutually inconsistent. If we think that inconsistency is a failure of rationality, and that nonetheless there are inconsistent believers, then we might be tempted to deny that rationality is a necessary precondition on belief ascription. But it is far from easy to turn the case of inconsistent believers into a clear counterexample to the rationality constraint. In order to do so, one needs to show that there are inconsistent believers (section 2) and that inconsistent believers are nonrational believers, at least in the sense of 'rationality' that the belief ascription theorists employ (section 3). Finally, one needs to consider whether Davidson has a successful reply to the objection from inconsistency (sections 4 and 5). Davidson adopts a three-step strategy: 1. He holds that it is possible to ascribe non-rational beliefs to a system only if in that system there is a general background of rationality (background argument). 2. He provides a psychological explanation of how believers might be inconsistent by introducing the notion of a compartmentalised mind (compartmentalisation). 3. He claims that, even though believers might be inconsistent, they preserve the capacity to recover from their inconsistency (recovery). The upshot of Davidson's defence is that there is no challenge to the rationality constraint on belief ascription from the case of inconsistent believers. The real challenge - Davidson says - would be the case of a believer who is obviously inconsistent, or the case of a believer who fails to recover from an inconsistency when she is made aware of it. Within this framework, the presence of some inconsistencies in a belief system is not itself an anomaly, though their spreading and their persistence would signal a genuine departure from rationality. 2 Davidson's defence will be assessed according to the following criteria: how successful it is in denying that the rationality constraint is challenged by the case of inconsistent believers, and how it fits with the motivations behind an interpretationist theory of belief ascription. I shall argue that, as a consequence of Davidson's three-step strategy, the formulation of the rationality constraint that is at the core of his theory of belief ascription is given a new reading. The rationality constraint Davidson needs in order to resist the objection consists in assuming that a believer can recover from an inconsistency by revising her own beliefs if found to be inconsistent, rather than in assuming that the believer's behaviour generally conforms to the norm of consistency. I shall argue that this position allows Davidson to answer the objection from inconsistencies, but that it creates some tension in his account of radical interpretation and partially undermines the behavioural bases of his interpretationist theory of belief ascription. 2. Inconsistency and Hypocrisy The evidence I shall present here has been recently gathered by some psychologists in the attempt to support a version of the cognitive dissonance theory. According to this very influential theory, originated by Leon Festinger (1957), pairs of 'cognitions', if relevant to one another, can be either consonant or dissonant. They are consonant if one follows from the other and dissonant if the opposite of one cognition follows from the other. The thesis is that, in the presence of a dissonance, the subject feels psychologically uncomfortable and attempts to reduce it. Elliot Aronson is responsible for applying the theory to dissonant cognitive elements concerning conceptions of the self. He claims that one can make accurate predictions about dissonance reduction when one considers the behaviour of subjects who come to believe something about themselves that does not cohere with the idea they have about the kind of people they are. The prediction is that they will attempt to reduce the dissonance by revising their behaviour in accordance with their self-concept. My interest here is in the data about dissonant cognition and not in the adequacy of the predictions psychologists can make about dissonance reduction. I shall consider the case of dissonant 'cognitions' about the self as evidence that subjects can be at risk of 3 inconsistency. I shall present one example, where what the subjects believe seems to be at odds with what they do. Or, as Aronson puts it, they "do not practise what they preach". The case I am going to consider here concerns the use of condoms by college students in the States. Aronson became interested in the situation because it looked as if students were endorsing a general principle but not acting in accordance to it. Surveys and interviews had shown that sexually active young adults were in favour of people using condoms to prevent AIDS. Still, other evidence was available to the effect that the great majority of college students were not using condoms regularly. When asked about the reason for their behaviour, they replied that condoms were a 'nuisance'. The outcome of the surveys seemed to imply that students were concerned about the dangers of unprotected sex but did not think the dangers "applied to them in the same way as they applied to everyone else." (Aronson, 1999, page 114) Aronson predicted that, if the college students were alerted to the fact that they were in a state of hypocrisy, then they would attempt to change their behaviour, since the belief that they were hypocrites surely would have not matched their self-concept. To test this hypothesis, Aronson devised an experiment during which students were divided in two groups and had to make a video about the dangers of not using condoms. One group only was asked to think about the circumstances in which they found it particularly difficult or impossible to use condoms in the recent past. At a distance of a few months, those students who participated in Aronson's experiment and were made aware of their hypocrisy, reported to have used condoms at a higher percentage than the students who were in the other group. Also, 83% of the students in the first group took advantage of a condom sale in the college. Leaving aside the question whether the results of the experiment we have described confirm the thesis of dissonance reduction, the observed patterns of behaviour should not be a surprise. We do experience similar instances of inconsistent behaviour in our everyday life. For my purposes here, the interesting aspect of the case presented by Aronson is that the subjects in the experiments were having, and some will have continued to have, conflicting 'cognitions'. One might have ascribed to the students inconsistent beliefs. They seemed to believe that there was danger in unprotected sex but 4 that they did not need to worry about unprotected sex themselves. They endorsed condom use as a general principle while their own behaviour showed little concern for safe sex. The two pairs of beliefs we might have ascribed to the subjects are not plainly inconsistent, because there could be reasons why some general principle does not apply to certain subjects in certain situations. But given the context, we can confidently say that some of the subjects in the experiments devised by Aronson were at least at risk of being inconsistent, and almost certainly hypocritical. 3. Who's Afraid of Inconsistency? Is inconsistency always a sign of irrationality? While nobody denies that consistency is desirable, some believe that there are epistemic virtues that override it in specific circumstances, and that our striving to be consistent should not be unconditional. For instance, Rescher and Brandom are convinced that the goal of 'capturing the truth' often outweighs the negative consequences of inconsistency. Consistency is an epistemic desideratum - but not one that is absolute and unqualified. Man's mind does not thrive on consistency alone: the blockage to order that results from conflicting images is a crucial goad to enquiry and a pivotal motive for enlarging our information. (Rescher and Brandom, 1979, page 43) Rescher and Brandom argue against the view that inconsistency is always a violation of rationality. There are cases, they claim, in which a rich body of knowledge with some inconsistencies is to be preferred to a perfectly consistent but smaller body of knowledge. For instance, suppose the richer system offers answers to some of our key questions and has just a few localised areas of inconsistency. The smaller system, though perfectly consistent, might leave us in the dark about the things that really matter to us. It is not clear that it is more rational to prefer the more consistent system in this case. Moreover, richness and relevance of information are goals that we might refrain from pursuing for fear of inconsistencies. We should avoid adopting an epistemic attitude that leads to an extreme caution in the acquisition of new information, such as the following: Rather than risk contradiction within your overall set of acceptance commitments, accept nothing at all. For Rescher and Brandom, this attitude promotes 5 excessive conservatism. In those situations when the subject cannot decide which belief to give up, when there is no possibility of suspending the judgement until new data are available, or when suspending the judgement is simply not advisable, then the fear of inconsistencies can lead to excessively static epistemic policies. In psychology as well, the presence of inconsistencies in a subject's belief system is seen as a strong motivation for the subject to gain additional information in the attempt to solve the inconsistencies. In the following passage, Festinger explains why dissonance between two cognitive elements can encourage the search for new information. Adding information can help decide which belief of the inconsistent pair has to go. If dissonance exists between two cognitive elements or between two clusters of cognitive elements, this dissonance may be reduced by adding new cognitive elements which produce new consonant relationships. One would then expect that in the presence of dissonance, one would observe the seeking out of information which might reduce the existing dissonance. (Festinger 1957, page 126) Of course, these considerations about the possible positive consequences of holding inconsistent beliefs cannot be turned into a defence of inconsistency as such. In the context of interpretationist theories of belief ascription, inconsistency is not just a violation of rationality, but possibly the most serious one. The following passage will confirm that consistency is at the heart of the notion of rationality on which Davidson relies: It is only when beliefs are inconsistent with other beliefs according to principles held by the agent himself -in other words, only when there is an inner inconsistency- that there is a clear case of irrationality. Strictly speaking, then, the irrationality consists not in any particular belief but in inconsistency within a set of beliefs. (Davidson, 1985a, p. 348) For Davidson the norm of consistency is among the most fundamental norms of rationality. In his theory the possibility that believers engage in an obvious inconsistency is ruled out. That is, believers never state or endorse a plain contradiction. But for Davidson, as I anticipated, it is possible to engage in a non-obvious inconsistency and so the problem arises. If one can ascribe inconsistent pairs of beliefs, then the rationality 6 constraint seems to fail to achieve its target as a constraint, since the particular case it would most certainly rule out (that of inconsistent beliefs) is shown to be possible. For the radical interpreter, the rationality assumption consists in taking for granted that speakers (i) share the observance of the principle of consistency and possibly of other fundamental logical laws that govern relations among beliefs and (ii) converge on beliefcontents at a low level of conceptual sophistication. Both constraints concern the conformity of the speaker's beliefs to some standard - rationality or truth. The basic strategy must be to assume that by and large a speaker we do not yet understand is consistent and correct in his beliefs. (Davidson, 1974b, p. 238) From the assumption of rationality in radical interpretation, the rationality constraint and the truth requirement extend to familiar cases of belief ascription. A speaker's representational states gain the status of beliefs, only if they meet the requirements of consistency and truth. Now it should be clear why a failure of consistency in a belief system can create problems for the rationality constraint. How can we intelligibly ascribe inconsistent beliefs, when the assumption that the speaker is consistent is the most basic assumption the interpreter can make? Davidson (1985a) has a solution. He concedes that there can be local violations of rationality in a largely rational belief system, but he denies that believers can engage in obvious inconsistencies or fail to recover from an inconsistency that has been pointed out to them. If we adopt this view, it seems prima facie possible to reconcile the necessity of the rationality constraint on belief ascription with the presence of few inconsistent beliefs in a largely consistent belief system. The interpreter might ascribe to the speaker a pair of inconsistent beliefs if the speaker's belief system is otherwise largely consistent. While Davidson accepts that there can be inconsistent believers, he realises that the very possibility of a believed inconsistency is problematic. To identify some irrationalities with inner inconsistencies is not to explain, or even go far in describing, such psychological states; indeed, it makes the problems of description and explanation seem impossible. (Davidson, 1985a, p. 353) 7 According to Davidson, to explain why a speaker A holds a belief p is to make explicit the reasons for A to believe that p. The same process of elucidation of the speaker’s reasons has an effect of rationalisation. From the speaker’s point of view, beliefs can always be justified and they appear to be rational to an interpreter who attempts a truly charitable reconstruction. The only exception would occur when speakers believe they have reasons to violate one of the fundamental norms of rationality to which they subscribe. However, "nothing can be viewed as a good reason for failing to reason according to one’s best standards of rationality" (Davidson, 1985b, p. 92). A believer can believe p for a reason and believe not-p for a reason, and therefore be ascribed the inconsistency, but she cannot believe 'p and not-p' for a reason. The believer might find herself believing 'p and not-p', but she cannot have a reason to endorse the inconsistency. The claim is that, when the believer realises that she is violating the norm of consistency, she will do something to restore it. Interpreters recognise a set of potentially non-rational beliefs when they detect an inconsistency in the speaker’s belief system. As consistency is a fundamental norm to which speakers have to subscribe, its violation can never be grounded on good reasons. Davidson has to explain how a believer might be inconsistent without failing to satisfy the rationality requirement. To do so, he appeals to the simple fact that the inconsistency is not necessarily a sign that the believer fails to appreciate the principle of consistency, but rather it indicates that the believer is not aware that she is holding two mutually inconsistent beliefs. 4. Recovering from Inconsistencies In this section I shall take a closer look at Davidson's attempt to show that inconsistent believers do not constitute an exception to his theory of belief ascription. First, I shall focus on the thesis of compartmentalisation and see how a subject might be unaware of holding inconsistent beliefs. Then, I shall turn to the claim that a believer can always recover from inconsistency. This is the crucial move for Davidson. Being inconsistent is not necessarily a departure from rationality. The fact that there are believers who are inconsistent but do not engage in obvious inconsistencies and have the capacity to 8 recover from their inconsistencies is compatible with the dictates of the rationality constraint. The appeal to compartmentalisation is an interesting and plausible move. Why does a believer tolerate an inconsistency in her belief system? Because she doesn't realise there is one. Believers lack a general overview of the beliefs they hold at any one time and might not realise that there are mutually contradictory beliefs in separate compartments of their minds. Compartmentalisation fits with many things we know about the mind. As a consequence of our limited cognitive resources, and in particular our memory limitations, it is plausible to deny that human agents can constantly attend to all the beliefs they hold. Given that the mind is compartmentalised, subjects might not be aware of the tension between two of their beliefs. If they were aware of the fact that an inconsistency affects their belief systems, then, according to Davidson, they would solve the tension among their conflicting beliefs in order to re-establish consistency. The idea of recovery is that, even though temporarily inconsistent, believers never lose their ability to eliminate an inconsistency, once it is apparent to them. The capacity to recover from the inconsistency is a sign of the believers' subscription to a fundamental norm of rationality. In the context of interpretation, when the interpreter draws the speaker’s attention to the fact that there is a tension between two of her beliefs, then the speaker activates the beliefs in question and is able to recognise the risk of inconsistency. As a consequence of her subscribing to the norm of consistency, the speaker might give up one of her conflicting beliefs or suspend judgement. In Davidson's account, there are two important notions that need explaining, the notion of activation and that of subscription. I shall concentrate on activation here and turn to subscription in the next section. How can we make sense of the idea that the beliefs we hold are either activated or inactive? This is not the same distinction as that between explicit and implicit beliefs. Both the activated and the inactive beliefs are belief states the speaker has explicitly entertained at some point in the past, and that play a role in her cognitive life. Maybe an analogy with visual perception can be useful here. Suppose Jude is facing an open window and he is looking out. I ask him: "Have you seen that red car parked in front of the bakery?" He replies: "Yeah, it is a beautiful car." The 9 car has been there for a while and Jude had seen it before I mentioned it, when he looked out of the window in the first place. Then his attention drifted away and he 'forgot about' the car. When I mentioned it, he focused on it again. I believe Davidson has an analogous scenario in mind when he appeals to the notion of activated beliefs. Let's reconstruct a case of possible inconsistency, by reference to a famous example in the literature (Festinger, 1957, p. 1). In his ethics class, two days ago, when Mike was talking about equality, he said that blacks are just as good as whites. Then today, when he went home and noticed that his new neighbour is a black lady, he has complained about it with his flatmates. Suppose one of his flatmates, Tara, heard him talk about equality in class. She is puzzled and asks him: "What's the problem? Didn't you say that blacks are just as good as whites the other day?" It is not that Mike had been insincere in class. And he hasn't changed his mind now. It is just that when he complained about his new neighbour, the issue of equality didn't even cross his mind. His belief that blacks and whites are equal was not activated. Tara's remark activated it. And now Mike is aware that he is at risk of being inconsistent or of being a hypocrite, if he cannot offer other reasons to dislike his new neighbour other than her being black. The simultaneous activation of two conflicting belief states might not be sufficient for the speaker to recognise the risk of inconsistency. One can imagine situations of divided attention in which the speaker is aware of both the conflicting beliefs but fails to bring them together in one single thought (Mike's case could also be re-described that way). In order to violate consistency then, it seems as if the believer needs to endorse p and endorse not-p, to activate both p and not-p and to co-attend to p and not-p.i Let's suppose that the notion of inactivation does a good job in illustrating the unawareness involved in some everyday cases of inconsistency. Davidson should still avoid endorsing the view that each speaker's mind is divided in mini-minds, all internally consistent and that inconsistency can occur only between two or more mini-minds. Davidson himself seems to rule out the plausibility of such a view: The idea that the mind can be partitioned at all has often been held to be unintelligible, since it seems to require that thoughts and desires and even actions be attributed to 10 something less than and therefore distinct from, the whole person. But can we make sense of acts and attitudes that are not those of an agent? (Davidson, 1982, p. 291) Davidson doesn't even need to commit himself to the idea that belief systems are divided into thematically organised sub-systems. The issue with sub-systems is that, if we postulate their existence, we have to be able to account for the criteria according to which sub-systems are formed. What are the implications of such a thematic characterisation? If we don't specify the criteria for the sub-division of a belief system, then there is some element of indeterminacy. Does the belief about the undesirability of having a black neighbour fall under the theme 'black people' or 'desirable neighbours' or neither? If we introduce principles of thematic organisation, then we might obtain a rigid theory of belief grouping that can be easily disconfirmed by psychological evidence, and whose generality would be undermined by the presence of individual differences. For the purposes of allowing inconsistencies in a belief system, Davidson does not have to appeal to anything more than attention drifting. It is sufficient to establish that it is not feasible for the believer to activate all her beliefs at any one time. The believer might not realise whether two of her beliefs are inconsistent, if she does not attend to both of them together. This is not to deny that there can be more psychologically complex ways of explaining why our minds work as if they were compartmentalised. It is just to say that we can use the notion of belief activation without recurring to a more substantial notion of compartmentalisation, which would have serious implications for our cognitive architecture. Let's sum up the conditions under which inconsistent believers do not violate the rationality constraint. In order to be regarded as a non-rational creature, A must hold two active beliefs that are mutually inconsistent and co-attended to. A is not rational, not a legitimate believer, unless A can eliminates the inconsistency so-generated from her belief system. In order to respond to the argument from inconsistencies, Davidson has to emphasise one reading of the rationality constraint. In his classical papers on interpretation, Davidson views the rationality constraint as an assumption according to which the interpreter can ascribe beliefs to the speaker on the basis of the speaker's 11 behaviour. As we saw in section 3, the constraint consists in assuming that in a belief system some logical relations among belief states, and particularly consistency, are preserved. However, the assumption Davidson needs in order to make sense of the steps of activation and recovery seems not to concern straightforwardly the conformity of the speaker's behaviour to those logical relations. A speaker can fail to conform to consistency and still be rational. All that rationality seems to require is that the believer be able to revise her beliefs in accordance with some logical relations, if her beliefs were not already standing in those logical relations. In particular, the presence of an inconsistency would not be a violation of rationality, unless it was an obvious inconsistency from which the believer could not recover. Emphasising this reading of the rationality constraint, where the capacity to recover from failures of rationality becomes the central notion, allows Davidson to successfully reply to the challenge from inconsistencies. In fact, the only move open to his adversary now is to show that some believers can engage in obvious inconsistencies, and that they can fail to recover from it when challenged to provide reasons for it. Is there any chance to show that believers adopt obvious inconsistencies and that they maintain the inconsistency when challenged? I believe it is possible for to find plausible cases of this phenomenon, but it is very difficult to find uncontroversial cases. The take-home message is that more work needs to be done to devise a successful counterexample to the rationality constraint, and this cannot be done here. The simple observation that there are inconsistent believers seems not to seriously damage Davidson's theory. In the rest of the paper, I shall explore the implications of the reading of the rationality constraint that Davidson needs to emphasise in order to resist the challenge from inconsistency. 5. Assessing the Damage to the Rationality Constraint Here I shall concentrate on the effects of the threat posed by inconsistencies on the formulation of the rationality constraint. In particular, I shall take into consideration whether the qualifications added by Davidson to the rationality constraint are compatible with the motivations behind the interpretationist picture of belief ascription. I take it that 12 one of the main motivations for endorsing an interpretationist account is the convergence between considerations about the nature of beliefs and the practice of belief ascription. What is distinctive about the interpretationist account is the attempt to characterise a creature's beliefs by reference to the creature's behavioural patterns. There is a clear sense in which attributions of belief and desire […] are supervenient on behaviour more broadly described. (Davidson, 1975, page 159) If engaging in an obvious inconsistency and failing to restore consistency are what it takes to deviate from the standards of rationality, and genuine believers are never obviously inconsistent or impervious to recovery, then, according to Davidson, the rationality constraint is undefeated. The presence of cases of non-obvious inconsistencies, or even of inconsistencies that are abandoned when challenged, seems not to undermine the necessity of the rationality constraint on the ascription of beliefs. One can still hold true that there is a necessary connection between having beliefs and being rational, since non-obviously inconsistent believers can restore consistency, and thereby exercise their rational capacities, once the conflicting beliefs are both activated and co-attended to. The idea of recovery is straightforward, but the interesting question is what makes believers restore consistency. According to Davidson, believers recover from inconsistencies because they all subscribe to the principle of consistency. What does subscription amount to? It is not easy to reconstruct Davidson's account of subscription coherently. I shall start by discussing what subscription is not. Believers are not required to be able to formulate the fundamental principles of rationality, nor they are required to recognise such principles (and the violations of such principles) in all the relevant circumstances (Davidson, 1985a, page 352). Believers might even fail to conform to those principles. However, they must have some nonexplicit knowledge of the principles and such competence is supposed to manifest itself in the general conformity to those principles. The analogy is with speakers who may not be able to make explicit the grammar rules of their native language, in spite of their general observance of those rules. 13 For a person to ‘accept’ or have a principle like the requirement of total evidence mainly consists in that person’s pattern of thoughts being in accordance with the principle without being aware of it or able to articulate it. (Davidson, 1985b, page 81) There are some vagueness and possible sources of tension in Davidson's position. Davidson thinks that, in order to be rational, one needs to subscribe to norms of rationality. But it is not clear whether Davidson thinks that a believer's subscription to norms of rationality implies that her patterns of behaviour largely conform to such norms. In other words, is conformity to norms necessary for subscription? Davidson seems to maintain that what makes a creature A rational is not A's ability to articulate a norm, but the fact that A's thoughts and actions are in accordance with the norm. If we take seriously the view that subscription implies conformity, then we have to say that A is rational if A's thoughts are (largely) consistent, regardless of whether A is able to appreciate the principle of consistency. But, in the case of non-obvious inconsistencies, even conformity to the norm ceases to be a necessary condition for rationality. It is true that believers conform to consistency in the sense that they do not engage in an obvious inconsistency, but their conformity to the principle is not always necessary to their being rational. They seem to subscribe to consistency without conforming to it, if they can engage in a non-obvious inconsistency and then recover from it. Subscription to a norm is what allows the believer to recognise when the norm is violated. We are told that non-obviously inconsistent believers do not violate the requirements of rationality, because they have the capacity to recover from their inconsistency, when they are made aware of it. And we are also told that the reason why they restore consistency is that consistency is one of those norms to which they subscribe. If we take subscription seriously, when the interpreter expects believers to be rational, she expects them not to engage in obvious inconsistencies and to be disposed to eliminate non-obvious ones. She no longer regards the conformity of a believer's pattern of behaviour to the principle of consistency as a necessary condition for the ascription of beliefs, as a believer can entertain inconsistent beliefs if at least one of them is inactive or otherwise not attended to. The interpreter expects believers to subscribe to norms of rationality. 14 This focus on subscription affects the formulation of the rationality constraint on the attribution of beliefs. The principle of charity used in radical interpretation was formulated in terms of rationality defined as the conformity of a pattern of behaviour to given principles, as we have seen in the passages quoted in section 3. The principle was not formulated in terms of the believers' disposition to recover from deviations from norms of rationality. The relation between patterns of behaviour and norms could not be weaker than conformity, since it is by hearing a creature's utterances and observing its behaviour that the interpreter can find a large background of rationality and legitimately ascribe intentional states and action to that creature. In order to reply to the objection from inconsistencies, Davidson needs to partially detach the rationality constraint from claims about strict conformity to norms of rationality. What makes a creature rational now is the fact that the creature subscribes to the principle of consistency and thus is disposed to restore consistency in the appropriate circumstances. This gap between subscription and conformity is explicit in the following passages: I think everyone does subscribe to those principles [= principles of decision theory], whether he knows it or not. This does not imply, of course, that no one ever reasons, believes, chooses, or acts contrary to those principles, but only that if someone does go against those principles, he goes against his own principles. I would say the same about the basic principles of logic, the principle of total evidence for inductive reasoning, or the analogous principle of continence. These are principles shared by all creatures that have propositional attitudes or act intentionally. (Davidson, 1985a, page 351, my emphasis) The possibility of inconsistency depends on nothing more than this, that an agent, a creature with propositional attitudes, must show much consistency in his thought and action, and in this sense have the fundamental values of rationality; yet, he may depart from these, his own, norms. (Davidson, 1985a, page 353) Subscription to norms is not conscious and does not necessarily result in conformity to such norms. Nor does it always need to be manifested in the believer's capacity to recognise that the norms have been violated. The divorce between subscribing and conforming to principles of rationality is a move that helps to explain the apparent 15 irrationality of everyday believers at the level of conformity and, at the same time, to guarantee their rationality at the level of subscription. If a coherent account of the distinction between conformity and subscription is available, the interpretationist can explain how inconsistent beliefs can be ascribed to rational creatures. However, a coherent account of the distinction must also include an explanation of the relation between subscription and conformity. Davidson does not explicitly address this issue, but we can make a charitable suggestion by drawing on the background argument. Maybe the relation between subscription and conformity is the following. If a belief system largely conforms to the norms of rationality, then this counts as evidence that the believer subscribes to the norms. The suggestion here is that there could be an evidential relation between conformity and subscription. But when we take a closer look to Davidson's account of subscription, we find that he seems to oscillate between endorsing an evidential relation between conformity and subscription and denying that it is possible. Davidson (1985b) recognises that there might be impediments of various kind to the manifestation of rational behaviour: "conformity is more likely when there is more time for thought, less associated emotional investment in the conclusion or when explicit Socratic tutoring is provided." (page 83). It seems as if these could be useful guidelines for the interpreter, who has to account for an apparent failure of rationality. This passage seems to speak in favour of, rather than against, the existence of an evidential relation between conformity and subscription. But Davidson (1985a) insists that subscription to norms is not an empirical matter, and that there cannot be evidence for or against it. The question whether a creature "subscribes" to the principle of continence, or to the logic of the sentential calculus, or to the principle total evidence for inductive reasoning, is not an empirical question. For it is only by interpreting a creature as largely in accord with these principles that we can intelligibly attribute propositional attitudes to it, or that we can raise the question whether it is in some respect irrational. (Davidson, 1985a, page 352) 16 If Davidson is not able to relate his claims about subscription to claims about conformity, problems emerge for his theory of interpretation. He needs an assumption about conformity to get interpretation started (something such as: 'creatures with intentional states largely conform to the norms of rationality'). If he settles for an assumption about subscription ('creatures with intentional states subscribe to the norms of rationality'), where the role of conformity is not clearly delineated, then his theory of radical interpretation is worse off. It is time to consider the consequences of the appeal to subscription for the plausibility of the rationality constraint on belief ascription. In the spirit of the principle of charity, the interpreter must assume that a creature is rational in order to ascribe beliefs and other intentional states to that creature and be able to predict and explain its behaviour. If 'being rational' is a matter of subscription, and there is no requirement that subscription should be systematically reflected in a creature's actual thoughts and actions, then the rationality constraint can be of little help as a guide to the practice of interpretation, explanation and prediction. The appeal to rationality could still play an important theoretical role, but would not offer to the interpreter the resources to ascribe meaning to the speaker's utterances on the basis of the assumption of consistency and truth. Would an assumption about the speaker's dispositions be sufficient to ascribe determinate content to her beliefs and utterances? It seems as if the assumption of subscription to norms of rationality would work only in case it did imply that the speaker's behaviour largely conforms to such norms. From the point of view of the interpreter who observes the behaviour of a creature producing utterances in an unknown language, it does not really matter whether the creature subscribes to consistency, unless subscription translates into a systematic pattern of easily observable behaviour. It might be said that recovery from inconsistency is criterial for subscription. Suppose, for a moment, that this is so and suppose that a creature would be disposed to restore consistency, if its inconsistent beliefs were activated and co-attended to. Maybe the creature could be disposed to restore consistency, if its inconsistent beliefs were activated. Still, in order to 'activate' the speaker's inconsistent beliefs, the interpreter would need to engage in a fairly complex conversational exchange with the speaker and this, by hypothesis, is precluded in first 17 stages of radical interpretation. So the interpreter might not be able to determine whether the creature subscribes to the norm of consistency. Radical interpretation is the situation in which the interpreter has to understand the utterances of a speaker of an unknown language. Ordinary belief ascription, in contrast, is the situation in which interpreter and speaker share the same language. We have just seen that the proposal that a rationality constraint would guide the idealised project of radical interpretation is seriously compromised if we have no systematic account of how subscription to norms is manifested in behavioural dispositions. But this criticism does not hold in less exotic cases of ordinary belief ascription, where the shared language between interpreter and speaker makes it easier for the interpreter to elicit the speaker's dispositions and observe the speaker's response to the provided Socratic tutoring. I imagine that Davidson would want to defend the analogy between the practices of radical interpretation and ordinary belief ascription. One way of denying that the emphasis on subscription might bring a paralysis of radical interpretation, is to claim that there is nothing unobservable and mysterious in behavioural dispositions. I seem to have assumed in my previous discussion that, while the conformity of the speaker's behavioural patterns to consistency is something to which the radical interpreter has access, the speaker's disposition to restore consistency in some given circumstances cannot be observed. But at least according to one view of dispositions (Kenny, 1976, page 210), if all the external conditions for the manifestation of a disposition are present, then the disposition will necessarily be exercised. This indicates that dispositions are observable patterns of behaviour after all, i.e. they are patterns of behaviour one can observe when the external conditions for the manifestation of the dispositions are present. There are at least two problems with this response. First, the radical interpreter might not be able to bring it about that the conditions for manifesting the disposition to recover are present. Secondly, a pattern of behaviour that seems to manifest a disposition to recover from inconsistency might not be due to such a disposition. Let me briefly elaborate the first point. The behaviour of a genuine believer is the behaviour of a creature that restores consistency when alerted that an inconsistency might 18 occur among its beliefs. It seems as if the manifestation of this disposition is a perfectly observable instance of behaviour, and pace Davidson, could be regarded as evidence for the claim that the creature subscribes to the norm of consistency. However, for a disposition to be exercised, all the external conditions for its manifestation need to be present. One of the conditions for manifesting the disposition to recover from an inconsistency is that the believer recognise the inconsistency and she can do so only if her inconsistent beliefs are simultaneously activated and co-attended to. The typical way in which activation occurs is by means of Socratic tutoring. That is, the interpreter points out to the believer that she is committed to p and to not-p. But in the case of radical interpretation, as I have argued above, the interpreter might not be in the suitable conditions to elicit in the believer the disposition to recover from the inconsistency, as she might not know how to activate the speaker's beliefs. So it seems to me that if the conditions for eliciting the speaker's dispositions are not met, then the speaker's subscription to a norm of rationality might not turn out to be easily accessible by the interpreter. 6. Conclusion Davidson's reply to the objection from inconsistencies is a Pyrrhic victory, because it seems to compromise the philosophical role of the scenario of radical interpretation. On the one hand, Davidson argues convincingly that the possibility of inconsistent believers is compatible with his version of the rationality constraint on belief ascription by adopting his three-step strategy and introducing the notion of subscription to norms of rationality. On the other hand, his account is at best incomplete, as the notion of subscription to norms of rationality is underspecified and no coherent and unitary account is given of the relation between subscribing to a norm and conforming to it. Either apparent cases of believed inconsistencies are a counterexample to the rationality constraint or else, if the constraint is read in such a way as to block the generation of easy counterexamples, the revised constraint undermines the use of the scenario of radical interpretation in the ascription of content. 19 If subscription to norms of rationality implies general conformity to such norms, then the assumption of subscription can play a role in the practice of radical and everyday interpretation, but there is still a tension between the rationality constraint and the observation that believers engage in non-obvious inconsistencies. If subscription does not imply conformity, as the appeal to the notion of recovery suggests, then a paralysis of radical interpretation might ensue. The radical interpreter would find it extremely difficult to elicit the conditions in which the speaker's behavioural dispositions could be manifested. Conformity is what matters for the practice of belief ascription, and it has always played a key role in the interpretationist picture. Conformity seems to be neither necessary nor sufficient for subscription. Even though Davidson's appeal to subscription could work in the case of ordinary belief ascription, where the interpreter can gather evidence that is sufficient to determine whether the speaker subscribes to the norms, in the context of radical interpretation acute problems emerge. The idea that one can ascribe beliefs to a creature by reference to the creature's behavioural patterns, which is distinctive of the Davidsonian picture, loses its intuitive appeal if the assumption of rationality concerns the creature's reasoning competence rather than the manifestation of its actual behaviour.ii 20 References CHERNIAK, C. (1986). Minimal Rationality. Cambridge (Mass.): MIT Press. DAVIDSON, D. (1973). Radical Interpretation. In D. DAVIDSON (1984), Inquiries into Truth and Interpretation. Oxford: Clarendon Press. Davidson, D. (1974a). On the Very Idea of a Conceptual Scheme. In D. DAVIDSON (1984), Inquiries into Truth and Interpretation. Oxford: Clarendon Press. DAVIDSON, D. (1974b). Psychology as Philosophy. In D. DAVIDSON (1982), Essays on Actions and Events. Oxford: Oxford University Press. DAVIDSON, D. (1975). Thought and Talk. In D. DAVIDSON (1984), Inquiries into Truth and Interpretation. Oxford: Clarendon Press. DAVIDSON, D. (1982). Paradoxes of Irrationality. In R. WOLLHEIM (ed), Essays on Freud. London: Cambridge University Press. DAVIDSON, D. (1985a). Incoherence and Irrationality. In Dialectica 39, 345-354. DAVIDSON, D. (1985b). Deception and Division. In J. ELSTER (ed), The Multiple Self. Cambridge: Cambridge University Press. DENNETT, D. (1987). Reflections: Real patterns, Deeper Facts and Empty Questions. In D. DENNETT (1987), The Intentional Stance. Cambridge (Mass) London (England): MIT Press. DENNETT, D. (1979). True Believers. In D. DENNETT (1987), The Intentional Stance. Cambridge (MA): MIT Press. DENNETT, D. (1981). Making Sense of Ourselves. In D. DENNETT (1987), The Intentional Stance. Cambridge (MA): MIT Press. FESTINGER, L. (1957). A Theory of Cognitive Dissonance. Stanford: Stanford University Press. KENNY, A. (1976). Human Abilities and Dynamic Modalities. In J. MANNINEN and R. TUOMELA (eds), Essays on Explanation and Understanding. Dordrecht: D. Reidel. 21 RESCHER, N. and BRANDOM, R. (1979). The Logic of Inconsistency. Oxford: Basil Blackwell. STICH, S. (1990). The Fragmentation of Reason. Cambridge (MA): MIT Press. i Martin Davies drew my attention to the distinction between activation and co-attention and Greg Bognar suggested to me this way of phrasing it. ii Thanks to Martin Davies and Natalie Gold for extensive comments on the final draft of this paper. I am also indebted to Kim Sterelny, Peter Godfrey-Smith, Philippe Chuard, Philip Pettit, Phil Gerrans, Laura Schroeter, Christian List, Matteo Mameli and two anonymous referees for helpful suggestions. 22