Content and Consciousness

by Daniel C. Dennett

  The point of thinking in terms of propositional attitudes even where no neat sentences of propositional attitude can be produced is that Intentional objects, even under the linguistic interpretation given them here, lead almost inexorably to metaphysical excesses,17 and the characteristic of these objects that accounts for this is one that it can be argued serves precisely to show that Intentional objects are not any kind of objects at all. This characteristic is the dependence of Intentional objects on particular descriptions. As criterion (3) indicates, to change the description is to change the object. What sort of thing is a different thing under different descriptions? Not any object. Can we not do without the objects altogether and talk just of descriptions? When John wants the stick on the ground, and the stick turns out to be a snake, it would be wrong or meaningless to say that John wanted the description ‘the stick on the ground’, for we do not want descriptions and so in Brentano’s terms his mental phenomenon is not directed upon the description, but is it not directed by the description? It is suggested that Brentano’s thesis might be altered to read ‘All mental phenomena are directed by (or simply: related to) unique descriptions or whole propositions which usually, but not always, have reference to real objects in the world.’ Thus Brentano’s thesis becomes, what it is often supposed without argument to be by writers in the field, simply that mental phenomena differ from physical phenomena in having a content, or relating to meaning, in the sense that their identity as individual phenomena is a matter of the unique descriptions or propositions to which they are related.

  Raising the subject level of discussion back up from phenomena to talk about phenomena, from things to sentences, the point is this: Intentional sentences are intensional (non-extensional) sentences.18 Briefly and roughly, the extension of a term is the class of all things of which the term is true, or to which the term refers. Thus the extension of ‘Presidents between 1961 and 1968’ is the class containing the two members, Kennedy and Johnson. The intension of a term is, roughly, its meaning. The terms ‘Democratic Presidents between 1956 and 1968’ and ‘Twentieth-century Presidents whose names begin with adjacent letters of the alphabet’ have the same extension as ‘Presidents between 1961 and 1968’, but clearly all three have very different meanings or intensions. ‘Goblin’ and ‘sphinx’ have the same extension (the null class), but different intensions. The going scheme of logic, the logic that both works and is generally supposed to suffice for all scientific discourse (and, some hold, all significant discourse), is extensional. That is, the logic is blind to intensional distinctions; the intersubstitution of coextensive terms, regardless of their intensions, does not affect the truth value (truth or falsity) of the enclosing sentence. Moreover, the truth value of a complex sentence is always a function of the truth values of its component sentences. Criteria (2) and (3) indicate that Intentional sentences do not follow the rules of extensional, truth-functional logic, and hence they are intensional. This expression of the position leads us to the central claim of the Intentionalists, that Intentional phenomena are absolutely irreducible to physical phenomena. Put in terms of sentences, the claim is that Intentional sentences cannot be reduced to or paraphrased into extensional sentences about the physical world. The claim goes beyond the obvious fact that Intentional sentences are intensional, and hence cannot be, as they stand, extensional – to the more remarkable claim that no sentence or sentences can be found which adequately reproduce the information of an Intentional sentence and still conform to extensional logic. This is to be contrasted with the situation with respect to, for example, Quine’s sentence:

  (13) Giorgione was so called because of his size.

  This sentence, intensional as it stands since it does not admit the substitution salva veritate of the coextensive term ‘Barbarelli’, can be replaced by the sentence

  (14) Barbarelli was given the name ‘Giorgione’ because of his size

  which is extensional, since the occurrence of ‘Barbarelli’ in the sentence is replaceable salva veritate by ‘Giorgione’, and the occurrence of ‘Giorgione’ is within inverted commas, and hence, by standard convention, refers to the name ‘Giorgione’ and not the man. Since it refers to the name, ‘Barbarelli’ is not coextensive with it; its coextensive brethren include ‘the maximizing form of “Giorgio” ’ and ‘the name which appears on page seven of the art gallery catalogue’, and so forth, and these are all substitutable salva veritate. That is, making two legitimate substitutions in (14) we get

  (15) Giorgione was given the name which is the maximizing form of ‘Giorgio’ because of his size

  which has the same truth value as (14). The Intentionalist claim is that no extensional sentence – or longer paraphrase – could reproduce the sense of an Intentional sentence in the manner in which (14) reproduces the sense of (13).

  This claim has been argued for in a number of different ways. Since the conclusion involves a negative existential claim, viz., that there are no such paraphrases, it can never be absolutely established, but only made extremely compelling.19 Quine attempts to generate this conviction by taking the case of indirect quotation or oratio obliqua (‘x says that p’), and arguing that ‘for all its tameness in comparison with other idioms of propositional attitude, and for all its concern with overt speech behaviour, [it] seems insusceptible to general reduction to behavioural terms; the best we can do is switch to direct quotation, and this adds information.’20 That is,

  (16) He says, ‘it is raining’

  is not a satisfactory paraphrase of

  (17) He says that it is raining

  for (16) adds information and hence could in many circumstances be false while (17) was true, and vice versa.21 One is invited to reflect on the impossibility of there being any physical state of affairs that would be in force always and only when someone was saying that it is raining – an act that might be accomplished on one occasion by merely nodding, on another by shouting ‘es regnet’, on another by saying ‘You bet it is’. If so overt an activity as saying that something is the case is not subject to behavioural, extensional paraphrase, what hope is there for such hidden, private phenomena as believing and imagining? Quine and Chisholm also present arguments about believing and intending, of which the central point is that efforts to provide behavioural analyses of these two phenomena are doomed by a vicious circle of implications.22 Take, for example, the belief that it is raining. What behaviour would clinch it that A believes it is raining? No matter what is suggested, it will turn out that this is a clincher demonstrating that A believes it is raining only if we assume that A has some particular purposes or intentions. A’s saying ‘It is raining’ or answering ‘Yes’ to the question ‘Is it raining?’ only counts as evidence on the assumption, inter alia, that A intends not to deceive us and ‘intends’ is an Intentional idiom. A’s finding a tree or roof to stand under is no more evidence, for it depends on A’s intending to stay dry. If ascription of belief always depends on an assumed ascription of intention, the converse holds as well. A’s intention to stay dry is not behaviourally demonstrated by his cowering under the tree except on the assumption that he believes it is raining, that he believes that he would get wet if he did not stay under cover, and so forth. A survey of the other Intentional and mongrel Intentional idioms shows that the use of any one of them has implications about beliefs and intentions, so the circle that prevents a behavioural paraphrase of belief and intention sentences infects the whole realm of the Intentional. It is, of course, no argument against this that behavioural data are ‘for all practical purposes’ completely reliable as clues to Intentional ascriptions, for we are not concerned here with practical purposes, but with theoretical foundations.

  IV TWO BLIND ALLEYS

  The Intentionalist thesis of irreducibility is widely accepted, in one form or another, and there are two main reactions to the impasse: behaviourism and Phenomenology. The behaviourist argues that since the Intentional idioms cannot be made to fit into the going framework of science, t
hey must be abandoned, and the phenomena they are purported to describe are claimed to be chimerical.23 Thus Quine, in one of his most pragmatic and behaviouristic moments, is ready to turn his back on Intentional idioms entirely, allowing them in his casual speech, but banishing them from the language of theory. ‘One may accept the Brentano thesis as showing the indispensability of Intentional idioms and the importance of an autonomous science of Intention, or as showing the baselessness of Intentional idioms and the emptiness of a science of Intention. My attitude, unlike Brentano’s, is the second.’24 This position, which is shared by behaviourist psychologists, is not merely the position that what you refuse to listen to cannot bother you. Our evidence that ‘there really are’ Intentional phenomena coincides with our evidence that in our ordinary language we speak as if there were, and if a science of behaviour could be successfully adumbrated without speaking as if there were these ‘things’, the insistence that there really are Intentional phenomena would take on a hollow ring. Behaviourism would attempt to discover extensional laws governing the occurrence of events (animal – including human – motions) that are initially given extensional, non-Intentional characterizations. If a truly predictive, extensional science of animal and human behaviour (specified in pure ‘motion’ terms and including all human verbal behaviour) could be produced, then the existence of Intentional idioms could be safely explained away as a peculiarity of natural languages, perhaps on a par with noun genders and onomatopoeia. Allowing working science to serve as ontological arbiter, one could claim that there really aren’t any Intentional phenomena, and hence no science of Intention is needed.

  Unfortunately for the alternative of behaviourism, however, so far the attempts to produce such an austere Stimulus-Response science have been notably unsuccessful. While behaviouristic research on animals and men over the last several decades has been undeniably fruitful from the point of view of crucial data obtained, these gains have been achieved independently of – and, in many instances, in spite of – the theories the experiments were intended to confirm or disconfirm. One could even make a case for the claim that the value of experimental results has been in inverse proportion to the extent to which the shibboleths of orthodox behaviourism have been honoured. The difficulty the behaviourist has encountered is basically this: while it is clear that an experimenter can predict rate of learning, for example, from the initial conditions of his mazes and the experience history of his animals, how does he specify just what is learned? It is certainly not the case that what the rat in the maze learns is a sequence of skeletal motions (for as Lashley’s famous experiments show, impediments in normal skeletal motion do not prevent the rat from getting itself to the food).25 Nor does it learn to move through a series of spatial juxtapositions no matter what. What it learns, of course, is where the food is, but how is this to be characterized non-Intentionally? There is no room for ‘know’ or ‘believe’ or ‘hunt for’ in the officially circumscribed language of behaviourism; so the behaviourist cannot say the rat knows or believes his food is at x, or that the rat is hunting for a route to x. The generalization of learning and the goal-directedness of the resultant behaviour have withstood all efforts to date to account for them as pure constructs out of the stimulus and response biography of the animal, and the nature of the theoretical failures points to the possibility of a fundamental error in the approach.26 The effect of these frustrations has been a relaxation of scruples, a tacit acceptance of Intentional characterizations, so that interesting research can continue.27 This strongly suggests, but does not prove, of course, that psychological phenomena must be characterized Intentionally if they are to be explained and predicted, that no science of behaviour can get along without the Intentional idioms.

  What then of the alternative of Phenomenology, the establishing of ‘an autonomous science of Intention’?28 An Intentional science of behaviour would characterize the events of its domain in fully Intentional terms. Its programme would be to relate actions, beliefs, desires, intentions, rather than the supposedly ‘pure’ events of the behaviourists (stimuli and responses characterized in extensional, ‘physical motion’ terms). Explanations in such a science would characteristically take a form like ‘His desire to find shelter prompted him to try to find a way into the box’, and what is immediately apparent, but not as important as has sometimes been claimed, is that we ordinarily explain behaviour in the Intentional mode.

  What is peculiar about such explanations is that they are not causal explanations in the more or less Humean sense of the term. The key word in the example above is ‘prompted’; it is not to be replaced by ‘caused’. The Humean doctrine is that causes must be identifiable independently from their effects, for otherwise the statement of cause and effect will not be contingently, empirically true, as it must be, but analytic, i.e., true only in virtue of the meanings of the words. This independent identification and concomitant contingency is missing, however, when the antecedent is an intention, the consequent an action, and the same can be seen to be true for all Intentionally characterized antecedents and consequents.29 To see what this difference amounts to, consider first a faulty argument about a case of causal explanation, e.g., the claim that conception is the cause of pregnancy. Hume requires causes and effects to be independently identified, but part of what we mean by conception is that in the absence of interfering factors pregnancy results, so there is a conceptual (not merely contingent) connection between conception and pregnancy. Then conception cannot be the cause of pregnancy. What is wrong with this argument is that although conception can be characterized as what pregnancy follows from, it can also be given independent characterizations, in physiological terms, which make no mention of pregnancy. If conception is defined as the cause of pregnancy, then it follows that ‘Conception is the cause of pregnancy’ is analytic, but it does not follow from this that conception is not the cause of pregnancy! Provided there is a way of alternatively characterizing the event which is conception, it can be a perfectly good Humean cause.

  The situation with Intentional explanations is different, however. It follows directly from the Intentionalist’s irreducibility hypothesis that no independent characterization of an Intentionally characterized antecedent is ever possible. To say that a particular Intentionally characterized antecedent could be characterized in another way is to say that either the Intentional sentence announcing the occurrence of this antecedent has an extensional paraphrase (and this is ruled out ex hypothesi), or the Intentionally characterized antecedent can be given a different Intentional characterization, but this is contrary to the fundamental principle of Intentionality, that Intentional phenomena are individuated by their characterizations – a different characterization means a different phenomenon. To take a particular example, consider a case where I intentionally open a door and walk out of a room. Does my intention to open the door cause me to open the door? If so, then we must be able to find another characterization for this intention, for as it stands it is certainly not conceptually independent of the consequent action. This independent characterization would have to be either extensional – but we are supposing for the moment that this is impossible – or Intentional. What other Intentional characterization of this intention could there be, though? The Intentionality of intentions is just that they have unique characterizations; the intention to leave the room, for example, is not the intention to open the door, and so could not serve as an alternate characterization of our initial antecedent. Nor can we avoid this dilemma by declaring that the cause of my opening the door was not my intention to open the door, but my intention to leave the room (which is an antecedent characterized independently of the consequent). For if I do leave the room, and do this intentionally, we shall have to find a cause for this as well, and if it is to be independently characterized it will have to be a yet more ulterior intention, say, the intention to see my brother, and so forth. It will not do for an Intentional science to try to get along in this way with an indefinitely large nest
ing of narrowly characterized actions caused by more widely characterized intentions (opening the door being caused by the intention to leave the room, leaving the room being caused by the intention to see my brother, etc.) for if opening the door is an intentional action of mine, it follows that I must have had the intention to open the door and I must have opened the door because I had the intention to open the door. This must be the case since it is possible to intend to do X, and do X, and yet not do X intentionally, because one did not do X because of the intention. For example, an actress can intend to scream, actually scream, and yet not intentionally scream, for though she did intend to scream at the time, she actually screamed because she was genuinely frightened. So the ‘because’ of Intentional explanations steadfastly resists treatment as a causal ‘because’; we must explain A’s intentional action X by saying A did X because he intended to do X, and this intention cannot be given the independent characterization it needs to be a proper cause.

  This impasse represents in itself an uncomfortable peculiarity for the ‘autonomous science’, but there is worse to come. The hegemony of Hume-style causal explanation is not so secure that non-causal explanation counts as an overwhelming objection to Intentional science, but the closed nature of Intentional explanation has a further consequence that amounts to a radical asymmetry of our scientific world view – in many ways like the asymmetry feared by the identity theorists (see Chapter I). In the domain of causal explanation, that a particular event a is followed by another event b is explained by the invocation of some more general causal law, to the effect that all events of type A (which includes a) are followed by events of type B (which includes b), and this law may be explained in turn by being subsumed under or deduced from still wider laws. In Intentional explanation, on the other hand, the sequences of events are so characterized that the occurrence of a particular consequent action is explained by the occurrence of a particular antecedent, say a perception or belief or intention, and there is no room for the question of why this consequent should follow this antecedent, and hence no room for any general law ‘explaining’ the sequence. For example, having said that my intention to leave was followed by my walking to the door, there is no room for the question: why should that result (as opposed to, say, opening my mouth or raising my arm) follow the intention to leave. The ‘covering law’ to the effect that all intentions to leave are followed by walking to the door is silly and unnecessary; the occurrence of my walking to the door has already been explained by citing my antecedent intention.