x /Contributors tive and positive approach characterises his work in philosophy generally. In September 1996, an international group of about twenty philosophers gathered in the Bohemian spa, Karlovy Vary, to participate in the immensely exciting and enjoyable entreprise of interpreting and criticising Davidson. A selection of their papers, together with Davidson's opening paper 'Externalisms', are presented in this book. The participants greatly appreciated Davidson's careful replies to all the papers, and the editors are happy that Davidson's reactions, included in the last chapter of this volume, will now also be available to the public. The succinctness, comprehensiveness and systematic structure of Davidson's concluding essay make it unnecessary to try to characterise in this Foreword Davidson's philosophical views or the topics discussed in this book. Although Davidson's article is structured by topics rather than persons, one name does appear in a subtitle: not surprisingly, it is Quine's. The relation between Davidsonian and Quinean philosophy is a fascinating issue and one much discussed; as such it appears in several articles collected in this volume. Willard Van Orman Quine was among the participants of the Karlovy Vary symposium on Davidson's philosophy and contributed substantially to the discussion. He would undoubtedly have had a lot to say about the views expressed in this book. Sadly enough, he will not be among its readers. There are four persons, apart from the contributors themselves, that the editors wish to thank especially for help with bringing about this volume: Dagfinn Fellesdal, Oslo and Stanford, John Perry, Stanford, and Gunnar Svensson and Henrik Ahlenius of the department of philosophy at Stockholm University. Petr Kotatko Peter Pagin Gabriel Segal 1 Externalisms Donald Davidson Our beliefs are objective, not, of course, in being unprejudiced and formed in the light of all the evidence, but in the sense that they are true or false, and that, with few exceptions, their truth depends on matters independent of us. Our other thoughts and attitudes, in so far as they have a propositional content, are also objective: it is an objective question whether or not our intentions or desires will be fulfilled, whether our hopes or fears are realized, whether our suspicions are justified, whether our orders will be obeyed, whether we will find what we seek. Thought is objective, and we know that it is. But it is not obvious what it is about thought that makes this possible. This question is not, in itself, epistemological. It is not the question how knowledge of the world is possible, or what justifies the beliefs we have. Skeptics think beliefs are objective, that is, that they are objectively true or false, for unless at least some of our beliefs were objectively true or false, there would be nothing to doubt. What skeptics doubt is whether we have good reason to hold those of our beliefs that happen to be true, that is, whether we have knowledge. If we cannot justify our beliefs, the skeptic maintains, then we must question whether the world is at all like what we believe it to be. If we are not to be skeptics about the possibility of knowledge of the external world or other minds we must reject the view that all knowledge of the world depends on objects or phenomena that are directly present in in- 1 2 / Donald Davidson dividual minds, objects such as sense data, impressions, ideas, raw feels, or propositions, objects that might be just as they present themselves to us even if the world were very different than it is. I do not deny that such objects and phenomena exist; it hardly matters. What matters is whether such objects are taken as epistemically basic, and therefore as epistemic intermediaries between our minds and the rest of the world, the messengers on which we must depend for all news of what goes on outside. Such theories are not only individualistic, but also subjectivist; according to such a theory, each person's world is constructed from material available in consciousness, material connected only indirectly, if at all, to the world. Empiricism-is a form of subjectivism insofar as it claims that the ultimate evidence for beliefs about the external world is something nonconceptual that is directly given in experience. This empiricist dogma should not be confused with the harmless doctrine that the source of all empirical knowledge is the senses. The only alternative to subjectivism is externalism, a view that makes the connection between thought and the world intrinsic rather than extrinsic—a connection not inferred, constructed, or discovered, but there from the start. There are two major species of externalism: social externalism, which maintains that the contents of our thoughts depend, in one way or another, on interaction with other thinkers; and perceptual externalism, which holds that there is a necessary connection between the contents of certain thoughts and the features of the world that make them true. The form of social externalism I find the most interesting is one Saul Kripke tentatively promotes in the name of Wittgenstein: someone is said to have grasped the truth conditions of a sentence if one uses the sentence in situations relevantly similar to the situations in which others would, or do, use that sentence (Kripke 1982). The form of perceptual externalism I find most defensible is one to which I have myself long subscribed, and which has more recently been well stated and defended by Tyler Bürge (Bürge 1988). Both of these forms of externalism suffer, however, from one or another defect or omission, at least as marshaled in response to the demand for an understanding of objectivity. Kripke's suggestion—or Wittgenstein's, if it is his—valuable as it is, is not satisfactory as a solution to my problem. One reason is that Kripke does not even attempt to give an account of the contents of particular thoughts; at best, he explains when two thoughts are the same (or when two utterances have the same meaning). Part of the reason for the failure to give an account of the contents of thoughts is due to the fact that Kripke dwells almost exclusively on mathematical examples, and so cannot connect the contents of thoughts with the empirical circumstances of speech behavior. EXTERNALISMS / 3 Furthermore there is the question how the content, however it is determined, is recognized as the same from one occasion of utterance to another, for Kripke's answer leaves us with another question. Kripke's answer is that an agent goes on in the same way as before (that is, means what he did before, or is following the same 'rule') if he goes on as others do or would. The further question is Warren Goldfarb's: he asks, if we can't tell when a single speaker's replies constitute his going on in same way, how can we tell when the replies of one speaker are relevantly like those of another speaker (Goldfarb 1985). Perhaps it will be said in Kripke's defense that we must take the notion of similarity for granted. The question was not, the defense continues, when one response of a speaker is like his previous responses, but rather what makes some responses correct and others not. A series of responses by a single speaker just shows that that person has one disposition or another; the notion of correctness or incorrectness has no application in such a situation. But when we add other speakers we have added a radically new element and with that addition we can say that correctness is defined as going with the crowd. But this, too, seems unsatisfactory. For how can the simple fact that two or more people have gone on in the same way introduce the distinction between following a rule and just going on in one way or another? All the sunflowers in the field turn together to face the sun: are they following a rule? Surely not in the required sense; they have no concepts and so cannot misapply concepts. Simply adding further creatures with identical dispositions cannot turn dispositions into rule-following. Doesn't Kripke himself show he is aware of the point when he quotes Wittgenstein as saying '...no course of action could be determined by a rule, because every course of action can be made to accord with the rule'(Wittgenstein 1953:20)? It seems clear that this difficulty remains no matter how many creatures follow the same course of action. What is missing is something to satisfy what Kripke calls the 'basic requirement' for a rule, namely that the rule should tell the speaker what he ought to do in each new instance (the word 'tell', italicized, is Kripke's). Or, more strongly still: 'The relation of meaning and intention to future action is normative, not descriptive' (Kripke 1982:37). But then, is it clear that Kripke has solved his own problem? For why ought I to go on as others do? What makes their behavior my norm? It does not answer this question to say that the teacher, or society, tells me. They can punish me if I fail to toe the line, but fear of punishment can't, in itself, give me the idea that there is anything more wrong with my action than that others don't like it. What is missing, one might say, is the idea of understanding. Introducing more than one creature does add something basic to the situation with one creature, for with the possibility that their ac- 4 / Donald Davidson tions may diverge we have introduced the gap needed to make sense of the notion of error. But the mere possibility of divergence, even when combined with sanctions to encourage conformity, does not introduce the sort of norm needed to explain meaning or conceptualization. However, despite these unanswered problems, I am persuaded that the basic idea is right: only social interaction brings with it the space in which the concepts of error, and so of meaning and of thought, can be given application. A social milieu is necessary, but not sufficient, for objective thought. Perceptual externalism, as described by Bürge and others, supplies one of the elements missing in Kripke's Wittgensteinian account by suggesting how particular contents can be assigned to our perceptual beliefs, and so explains in part how thought and language are anchored to the world. But there are serious lacunae in Burge's account of perceptual externalism. There is, first, the problem of the content of perceptual beliefs. According to Bürge, it is the 'normal' cause of a perceptual belief that determines the content of the belief. But what is the normal cause? It could be anything from the stimulation of nerve endings to the original big bang. Clearly, we need a way to isolate the right cause J Second, Bürge gives no serious account of error, and without an account of error, there is no way to distinguish between having a concept and simply having a disposition. How do we identify a mistake as a mistake? In other words, Burge's account fails to solve, or even address, the Wittgensteinian problem. These are both problems about relevant similarity. Fake cows are in one way relevantly similar to real cows—that's why we make mistakes. But in another way, fake cows aren't at all like cows—they don't fall under the concept cow. It is because we occasionally mistake fakes for the real thing that we can be said to have concepts, to classify things, and so sometimes classify them wrong. There's nothing amiss with a concept that includes the fakes with the reals—that's why it's hard to say what makes it possible to go wrong. It takes only a thoughtless disposition to distinguish cows from other things; it takes only another mindless disposition to treat cows and fake cows alike. What is difficult is to explain what is going on when someone thinks a fake cow is ä cow, for that requires having the concept of a cow. The problem of relevant similarity comes up also in connection with the idea that the content of a perceptual belief depends on its common or usual cause. Even to approach the question, we must have some idea which causes are being gathered together. Since any set of causes whatsoever will Quine quite properly raised this question (in conversation) in connection with my original suggestion that the content or an observation sentence was its 'common cause'. By this I then meant, not its cause for variotis speakers, but for one. I had not yet appreciated the importance of the social element. EXTERNALISMS / 5 have endless properties in common, we must look to some recurrent feature of the gatherer, some mark that he or she has classified cases as similar. This can only be some feature or aspect of the gatherer's reactions (perhaps these are verbal reactions), in which case we must once again ask: what makes these reactions relevantly similar to each other? Wittgenstein's problem once again. It would not surprise me if, at this point, someone were to say in exasperation, but of course we just do react in similar ways to certain stimuli: we shy away from extreme heat, we shrink from loud sounds, we smile when smiled at. It is a brute fact that we class certain things together in these ways. Of course it is easy for us to describe the stimuli and the responses, since we are among the creatures that naturally make these classifications. So of course we have predicates—concepts—that reflect these characteristics of ours. If we are going to describe anything, we have to use the equipment we have. None of this shows that it isn't true that creatures like us naturally classify things as we do. This is, of course, right, and it points us in the direction of Kripke's Wittgensteinian answer to the problem of rule following. It is a brute fact—a fact about each brute, including you and me—that each reacts to stimuli as it does; but it is a suggestive fact that many of these brutes react in more or less the same ways. It is a suggestive fact, but what exactly does it suggest? As we just noticed, mere crowd behavior in itself can't explain conceptualization; it can't even explain error. What it can do, however, is make space for something that can be called error: room for error is created by cases in which one individual deviates from a course of action when the crowd does not. For example, when a fake cow is introduced among the real cows, the crowd ceases to mutter 'Cow' while the deviant individual persists in saying 'Cow'. This is not enough, as we remarked just now, to justify an attribution of error; we can still say there simply are two different dispositions at work. Room has been made for error, but error itself has not been explained. What must be added in order to give an account of error is something that can count as recognition or awareness, on the part of the those who share reactions, of each other's reactions. The basic situation, as now set forth, contains a minimum of three elements: two creatures and a world of objects, properties and events the creatures can discriminate in perception. Of course the social environment would typically be more numerous, but it will simplify the story if we can for the moment stick to this irreducible triad. Each of the creatures is carrying out its habitual inductions, whether learned or inherited. This means that for each creature there are stimuli which can be classed together by virtue of the similarity of the responses. There will be cases that fit this de- 6 / Donald Davidson scription: creature A displays responses similar to its past responses at the same times and in the same places that creature B displays responses similar to its past responses. (The responses of A may or may not be similar to B's responses; A may run, while B cowers.) We can then ask what stimuli plausibly cause these repeatedly correlated responses. This common cause will typically be some objective feature of the environment. Nothing in this story suggests the necessity of thought, conceptualization, or error, even if the creatures occasionally fail to act in concert. No doubt we can say, looking on with our conceptual advantage, that one of the creatures has gone wrong when it reacts, perhaps verbally, to the lake cow as it does to real cows. The problem is to put the creatures in a position to think this. Something must be added which will make the difference between the cases where the creatures act in unison, and the cases where they do not, available to the creatures themselves. This can be done by adding a further element of association or natural induction: each creature associates the other creature's responses with stimuli from the shared world. (Perhaps this does not have to be learned in some cases.) Let us assume that our two creatures, A and B, can often observe one another when both can observe cows. Now suppose that creature A responds to cows by uttering the sound 'Cow'; creature B then associates cows with the utterances of 'Cow' by A. Creature B similarly associates cows with A's utterances of'Vache'. A and B are now each in a position to notice occasions on which the responses of the other differs from the other's usual responses (perhaps one is badly placed for seeing a cow; or one has better eyesight or a keener sense of smell). When I speak of responses differing, I mean: A's responses are not taken by B to be the same as A's earlier responses, while B's responses are the same as B's earlier responses. The basic triangle is now complete: there are causal connections between cows and our two reacting creatures, by way of perception, and there are causal connections between the creatures, also by way of perception. Finally, these two sets of connections are correlated, for each creature associates cows with the cow-reactions of the other creature, except, of course in the relatively rare case of error. ,, Take an example. I observe a lioness stalking a gazelle. I describe this situation in this way by correlating two aspects of the situation: gazelle behavior and lion behavior (the lioness changes course to intercept the gazelle; the gazelle changes course to evade the lioness). I can confirm my theory that the gazelle is the object salient for the lioness by (among other possibilities) noting the responses of other lionesses to gazelles on other occasions. But even this further evidence leaves it up to me to decide what the object of the lioness's attention is; I decide on the basis of what catches my attention. If a second lioness joins the first in pursuit of the gazelle, I Extern Ali sms / 7 can eliminate such complete dependence on my own choice of salient object in this way: I class together the responses of lioness A with the responses of lioness B in the same places and at the same times. The focus of the shared causes is now what I take to be the salient object for both lionesses. I no longer have to depend on my own choice of the relevant stimulus of the lionesses' behavior. In the exceptional case where responses differ / can say: one of the lionesses has erred. The challenge is to put the lions in a position to distinguish these cases. To do this we have to eliminate the dependence on my arbitrary (or interested) choice of relevantly similar responses on the parts of the lionesses. A further element enters when the lions cooperate to corner their prey. Each watches the other while both watch the gazelle, noting the other's reactions to the changes of direction. Here is my thesis: an interconnected triangle such as this (two lionesses, one gazelle) constitutes a necessary condition for the existence of conceptualization, thought, and language. It make possible objective belief and the other propositional attitudes. It fills in the gaps in social and perceptual externalism, at least the versions, due to Kripke and Bürge, that we have been discussing. It clearly incorporates the essential element in Kripke's 'solution' to the problem, which requires that we distinguish between a mere disposition and rule following. That element is provided by the second creature, who stands for society. The simple model I have described does not, to be sure, suggest where the norm lies when the creatures differ in their responses; but neither does Kripke's account of the matter. The point is not to identify the norm, but to make sense of there being a norm, and this has been done if we can point to the difference between the preponderance of cases where the creatures respond alike and the deviant cases where they diverge. I insisted that it is not enough that a third party be able to observe or describe these two cases; I claimed it was necessary that the existence of the contrast be available to the creatures themselves. This essential element enters the triangle when the creatures observe each other's reactions to the very phenomena they are both observing. The triangle also makes plain, what Kripke's mathematical examples could not, the role perception plays in the social establishment of a ground for objective thought. It does this in two ways: first, it allows the creatures involved to observe each other's reactions. But it also makes (partial) sense of the idea that there is a single focus for the joint reactions. Mathematical examples simply ask us to assume there is a common focus; in the case of perceptual objects, while shared reactions help locate the focus. Our triangular model thus makes a step toward dealing with another troublesome feature of Bürge's perceptual externalism, the indeterminate nature of the contents of perceptual beliefs. That difficulty arose because 8 / Donald Davidson there seemed to be no way to decide the location of the objects and features of the world that constitute the subject matter of perceptual beliefs; Bürge told us only that the content was given by the 'usual' or 'normal' cause. But this did not help choose between proximal and distal stimuli, or anything in between, in the causal chain. By introducing another perceiver, it is possible to locate the relevant cause: it is the cause common to both creatures, the cause that prompts their distinctive responses. Both creatures observe a cow. They do not share the neural turbulence that stirs in their brains, nor their retinal stimulations. They do not share photons streaming in, but the cow is mutually sighted and perhaps otherwise sensed. We can call it a form of triangulation. In the case of sight, it is quite literally a matter of triangulation: the lines of sight converge on the shared stimulus. It would not count as triangulation if there were no signal passed between the observing stations, but we have provided for this by requiring the observation of each other's responses to the shared stimulus by the creatures involved. I have emphasized at a number of points the way the notion of relevant similarity can play an essential role without being noticed: thus Kripke, in the attemptMo say when a creature's subsequent response was relevantly similar to an earlier response, helped himself to the notion of the relevantly similar responses of another creature. Bürge helped himself to the idea of relevantly similar stimuli. The reason it is so natural to avail ourselves of such concepts is, of course, that they are ours. The trouble is, we move in a circle when we appeal to what is natural for us in order to explain conceptualization itself—to explain, in other words, what is natural for us. Triangulation faces the problem in the only way it can be faced, by introducing the interaction of classifying agents. The difference is the difference between an external commentator slipping in his categories to make sense of an isolated creature, and a participant observing another participant doing his or it's thing. The social scene provides a setting in which we can make sense of the transition from mere dispositions to conceptualization. I mentioned earlier the fact that by using mathematical examples, Kripke deprived himself of one apex of the triangle, the apex occupied by the shared perceptual object. This left no way of determining the potential content of the purported rule, no observable object external to the learner and the teacher to be objective about. But here I was not entirely fair to Kripke: he does mention two examples which bring in external objects or features of the world. After saying that it is a 'brute fact' that people go on in the way they do rather than in some other way, he asks, Can we grasp how someone could be presented with a number of green objects, and be told to apply the term 'green' just to things 'just like these', and yet apply the term learnt as if it meant 'grue'? It would EXTERNALISMS / 9 seem that if we find our own continuation to be inevitable, in some sense we cannot (Kripke 1982:98). This is the case of our acting as detached observer: no interaction is involved. The following is closer to the situation I have been describing: A child who says 'table' or 'That's a table' when adults see a table in the area (and does not do so otherwise) is said to have mastered the term 'table': he says 'That's a table', based on his observation, in agreement with the usage of adults, based on their observation. That is, they say, 'That's a table' under like circumstances, and confirm the correctness of the child's utterances (Kripke 1982:99). Perhaps this is a correct interpretation of Wittgenstein; I do not know. But it still seems to me an incomplete statement of the nature of mastery of a word or concept. All that this analysis demands of the child is that he have acquired a certain habit, a disposition, a disposition that matches that of the adults. This is a disposition which (aside from the inessential fact of saying 'table' instead displaying some other distinctive form of behavior) the child might have had independent of any training or even learning. Mastery of the word or concept requires in addition, I believe, that the child understand that error is possible, that 'That's a table' expresses a judgment that has a truth value independent of its merely being uttered. It may be that to add this condition is to fail to grasp Kripke's or Wittgenstein's point that nothing purely internal, no feeling, attitude, or intention should feature in the explanation. But to rule out awareness of truth conditions, or of the possibility of error, is to reduce mastery of a word or concept to aping the behavior of others. To suggest that this is exactly what accepting a 'way of life', using words as others do, comes to is to miss the difference between two senses of use. In the sense in which use might be claimed to be meaning, two people may use a word in the same way when one is prepared to apply it in a certain situation, and the other is not. This is not mere aping. The fact that someone has grasped the distinction between following the crowd and being right is not without observable behavioral fallout; one such fallout is willingness to change one's mind (and therefore what one says), not because one differs from others, but because the change fits better with further observations. I will return to this general theme presently, but first I want to describe another precursor of my triangulation thesis, a precursor which has had a large influence on me. The precursor is Quine's account of radical translation. Quine has always been a sort of externalist, though this is not easy to see. Part of the reason it is—or was—not easy to see is that Quine long claimed that all empirical content rested, in the end, on something entirely private, what he called 'stimulus meaning'. It is true that stimulus meaning was defined in terms of physical events at our nerve endings; but as far as 10/ Donald Davidson epistemology is concerned, stimulus meanings could as well have been classes of sensations or sense data. This was in fact what Quine wanted: a scientifically respectable substitute for sense data, or whatever else provided the basis for the various forms of empiricism, such as Carnap's, Russell's, or C. I. Lewis's. The result was that the ultimate 'input', or 'evidence', or 'data' (these are Quine's words), all that (according to Quine) we have to go on in building up a picture of the world, was private. The subject matter of our utterances, and the contents of our thoughts, was determined by the world beyond our skins only as mediated by some sort of unconceptualized 'evidence'. If the impacts of the world on our skins were unchanged, but the world very different, we would never know it. This is a form of skepticism that did not trouble Quine, in part, perhaps, because he seemed, at least at times, to view material objects as mere 'nodes' or 'posits' in a scheme of which all that could be demanded was that it successfully predict sensory input. In this respect Quine was no worse off than other empiricists, at least empiricists who, like Quine, hold that all our 'information' or 'data' about the world can be characterized in terms of patterns of sensory stimuli or sensations, percepts, some uninterpreted 'given'. Quine lias modified his views on meaning in recent years, and stimulus meaning in particular no longer plays the role it did (Quine 1995; Quine 1996). But his epistemology remains resolutely individualistic; though he allows that society provides obvious gains in precision and the amassing of knowledge, there is no reason in principle why we could not win an understanding of the world on our own. This is not a position which can get a grip on the objective character of thought. Nevertheless, there has always been a strong externalist element in Quine's theory of meaning, and therefore also in his overall view of the contents of the pro-positional attitudes. Although the meaning of each person's utterances of observation sentences was at one time determined by the proximal stimulus (i.e., patterns of sensory stimulations), the only clue Quine provided to this content was by way of a comparison with other people's patterns of stimulation. This comparison was put into practical effect by the translator noting when, and under what circumstances, the speaker to be translated was prompted to assent to or dissent from a sentence; the translator was then to translate the speaker's sentence by the sentence the translator would assent to under the same circumstances. The Quine of Word and Object and many later works says the criterion of correct translation is a matching of stimulus meanings—of patterns of stimulation. But in fact, of course, no ordinary, or even extraordinary, translator has any idea what the patterns of stimulation of someone else, or of himself, are; he will in fact translate on the basis of what he perceives in the world and that he assumes the speaker also perceives. There is then, a certain ambivalence in Quine's position. On the EXTERNALISMS / 11 one hand, the epistemology is empiricist, based on private 'evidence'; on the other hand, the theory of meaning, as implemented by the process of radical translation, is basically social. Disposed from early days to distrust empiricist foundationalisms, I chose the social, and therefore externalist, position from the start/ If we make this choice, Quine's picture of radical translation satisfies the conditions for triangulation I have been describing. We have two interacting people, their interaction mediated by the external objects, situations, and events they mutually observe. This is how Quine has recently described radical translation (Quine 1996). It is not clear how this form of externalism fits with Quine's epistemology, which maintains that our perceptual beliefs are based on something private and given to the senses. Nevertheless, Quine teaches us that what a speaker means by what he says, and hence the thoughts that can be expressed in language, are not accidentally connected with what a competent interpreter can make of them, and this is a powerfully externalist thesis. Let me now go back over some of the ground we have covered, taking things in the natural order of increasing complexity, following what may be the history of the development of thought in each individual. I mentioned first a single animal, isolated in the sense that its relations to other animals do not supply any of the necessary conditions of propositional thought. Such an animal could, at least in theory, come equipped from the start to cope with endless contingencies, and could learn more. We are familiar with machines which give us the idea, without having to suppose they are capable of thought or conceptualization. We, observing and describing such a creature (or machine), say that it discriminates certain shapes, objects, colors, and so forth, by which we mean that it reacts in ways we find similar to shapes, objects, and colors which we find similar. But we would be making a mistake if we were to assume that because the creature discriminates and reacts in much the way we do, that it has the corresponding concepts. The difference, as I keep emphasizing, lies in the fact that we, unlike the creature I am describing, can, from our point of view; make mistakes in classification. We judge, come to believe, on looking, that something is a table, and that the table is brown and rectangular. The creature is reacting appropriately; it is not judging. 2 In adopting Quine's radical translation to my radical interpretation, I opted from the start for the distal (external) stimulus as the basis for interpreting what Quine called 'observation sentences', rather than Quine's proximal stimulus meaning. See Truth and Meaning (1967) and Semantics for Natural Languages (1970), both reprinted in (Davidson 1984). I emphasized this difference only later (Davidson 1983). 12 / Donald Davidson It does not add anything important to suppose that our creature responds to the behavior of other creatures, the way fish or birds swim or fly in formation. But it does make a significant difference if the creature associates the responses of other creatures with aspects of the world to which it also responds, for it then has a double take on the world. This double take benefits the creature in that it is now able to respond to threats or opportunities it does not independently sense; and it puts us in a position to note that the basic triangle is in place which is necessary if anything is to count as error, and to locate the features of the world to which the creatures are responding. Certain primates, for example, make easily distinguished sounds when different predators appear: one sound for eagles, one for snakes, one for lions. Other members of the group, though they may not have seen the danger, react to the appropriate threat by hiding, running, or climbing a tree. A more elaborate case is this: there is a species of bird in Africa which craves honey, but cannot break into a hive. When such a bird discovers a hive, it will fly over human honey-hunters indicating, by the direction and length of the segments of its flight, the location of the hive, correcting the course of the human hunters as necessary. When the human hunters have broken open the hive, they share its contents with the bird. For all the amazing sophistication of the bird's behavior, there is nothing in this story to suggest that we ought to attribute propositional attitudes like intention, belief, or desire to the bird; the bird follows a set pattern, but there is no reason to suppose it has the idea of a failure to carry out an intention, or for a belief to be false, or a desire misguided. But although we are not in a position to speak of conceptualization, many of the conditions for conceptualization and thought are present. To recapitulate once more, these are: 1) With two creatures in general responding in characteristic ways to distal stimuli, we can speak of the focus of their responses, the common cause of the most frequent cases. 2) If thought were present, this would provide an obvious indication of the content of the thought. 3) There is a space for the concept of error, which appears when there is a divergence in normally similar reactions. 4) Finally, and perhaps most important, since this is what we are most apt to miss: we have moved the reliance on our notion of relevant similarity into the realm of animal interests, where it surely belongs. We have accomplished this, not by projecting thought-like distinctions onto inanimate nature, as Plato and Aristotle both did in their ways ('dividing at the joints'), but rather by seeing the relevant similarity of one animal's responses through the eyes of another animal. We have embraced the Witt- EXTERNALISMS / 13 gensteinian intuition that the only legitimate source of objectivity is in-tersubjectivity. Mere similarity of response is obviously not enough for thought, however, not even when one animal's responses to events and features of the world serve to touch off responses appropriate to those same events and features in other creatures. This triangular arrangement is a necessary, but not sufficient condition, of thought. What must be added to make the conditions sufficient? If we could answer this question we would have an analysis of thought. It is hard to think what would satisfy us which did not amount to a reduction of the intentional to the extensional, and this, in my opinion, is not to be expected. What further progress we can make will be in the direction of theory building within the realm of the rational, not reduction of that realm to something else. What must be added to the basic triangle of two or more creatures interacting with each other through the mediation of the world if that interaction is to support thought? The unhelpful answer is that the relation between the creatures must include linguistic communication. For unless the creatures can communicate, unless they can engage in the exchange of propositional contents, there is no way they can take cognitive advantage of their ability to triangulate their shared world. They must, in other words, recognize each other as embodied minds with a location in a common space and time. The reason this answer is not very helpful is that it assumes what was to explained. Of course if there is language there is thought, so it cannot be easier to explain the former than the latter. Nevertheless, it is useful to recognize the somewhat surprising fact that the social element that is essential to language is also essential to thought itself. What makes this relation between thought and language useful in the present context is that it makes it possible to study thought by studying language. We exploit this possibility when we think of the case where the basic triangle is complete except that our two creatures do not understand one another. Each, we may suppose, has learned a language in a social setting, and has the general knowledge of the world and of people that this presupposes. We simply drop the assumption that the languages are the same. Each creature then has the task of coming to understand the other's language without the help of a bilingual informant or a dictionary. What makes understanding in such a situation radical is that since thought depends on language, our would-be interpreters cannot assume without circularity that they can fathom the thoughts of the other independently of understanding their speech. This is the problem of how radical interpretation is possible, and I believe it can, in outline, be solved. 14 / Donald Davidson There is an intermediate stage which deserves independent discussion, the stage at which a person is first introduced to the world of thought and reason, the stage of language acquisition. In this situation society is the teacher, whether deliberately or not, for even if a child has only one teacher, that teacher was taught by another, and so on back until we reach a stage at which no one is more a teacher than another. Foregoing speculation on how it might all have begun, let us consider how it begins for someone joining the society of those who are already seasoned speakers. For an individual, the process must at some early point involve the learning, through ostension, of words and concepts for the most easily observed and discriminated objects, events, and features of the learner's environment. We come into the world equipped to recognize faces, colors (even when reflected wavelengths and saturation vary according to amount and character of light), shapes (even when distance and perspective greatly alter what reaches the retina), and objects (though again, under many aspects, from many angles and under varied illumination). We are probably better, at least at first, in lumping together cow or rabbit appearances than in lumping together appearances of the same cow or rabbit after an interrupted viewing. What matters, in any case, is that for starters we learn words for what we preverbally discriminated; we learn from creatures very like ourselves, and could not learn from creatures very unlike ourselves. I spoke just now of learning words and concepts. This is, of course, the goal of the process. But something else must come first. At the start, the teacher has language and the concepts that come with it, the beginner does not. What comes first for the beginner is simple conditioning of sounds to sights (or whatever is otherwise perceived). 'The origin and the primitive form of the language game is a reaction; only from this can more complicated forms grow. Language—I want to say—is a refinement. "In the beginning was the deed'" (Wittgenstein 1993:395). This remark of Wittgenstein's is in the same vein as his remark that 'The basic form of our game must be one in which there is no such thing as doubt* (Wittgenstein 1993:377). What we are describing is on the road to speech and conceptualization, but it is not yet there. Once one has mastered a concept, or words that express it, one can always doubt whether it applies in a given case. But one cannot doubt the application from the start. Here is a way of putting this central point. Someone who is consciously teaching a beginner the use of a word may think of herself as simply passing on a meaning that already attaches to the word. But from the learner's point of view, the word—the sound—is being endowed with a meaning. This is why doubt makes no sense at the start. The first examples, the first things ostended, must, from the learner's point of view, belong to the application of the expression. This is so even if the teacher is at EXTERNALISMS / 15 fault from society's point of view, whether because the teacher does not know how others use the word, or because the teacher has mistaken fake cows for real ones. From the learner's point of view, any mistakes or doubts must come later. If we know what the objects are to which the learner associates the learned sound, then to the extent that we are like the learner, we know (i.e., can reasonably induce) what the learner, when he is capable of meaning anything, will mean by the learned sound. The question what others besides the learner, even his teacher, means by the sound is irrelevant; the sound has been given a meaning by the learning process itself, and we will misinterpret the learner if we assume that for him it has any meaning not connected with that process. This is a large part of the lesson of externalism, and we must never lose sight of it, for it is the reason it is impossible to be generally mistaken about the aspects of the world we have learned to conceive through ostension. I have been talking all along as if learning a first language depended on natural induction. I am aware that Chomsky and those in his thrall resist the application of this notion on the ground that basic grammar is wired in, and that early language acquisition is too rapid to count as learning or to require anything like induction. Let it be so. But whether or not we call acquiring them learning, we are not born knowing the words of Bantu or Armenian. It may well be that one hearing is enough in many cases to give us all we need in order to 'go on' in the right way. This cannot change the point I am making, provided the one hearing is in the presence of an osten-der who is ostending an appropriate object, event, property, or state of affairs. We must remember, of course, that if going on in the right way means only going on as the teacher does, this is not enough in itself to constitute mastery of word or concept; there remains the grasp of truth conditions, the awareness of the possibility of error. Nor will punishing or correcting what the teacher sees as error suffice, for the same reason. We can say that in the potential gap between how the learner goes on and how the teacher goes on there is room to introduce the idea of error, not only in the mind of the teacher, for we are assuming it is there, but also in the mind of the learner. But how that room gets filled it is probably beyond our powers to tell. Our failure to provide an analysis of the concept of error or, as we could as well say, of the concept of objectivity, or of thought itself, does not mean there is something hopelessly mysterious about these concepts; it only reflects the fact that intentional phenomena cannot be reduced to something simpler or different. Nor does our failure mean that true grasp of concepts as opposed to mere following of routines does not reveal itself in behavior. What 16 / Donald Davidson we cannot do is say, in non-intentional terms, in what the difference consists. References Bürge, T. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85:649-63. Davidson, D. 1983. A Coherence Theory of Truth and Knowledge. Kant oder Hegel, ed. D. Henrich., 423-38. Stuttgart: Klett-Cotta. Davidson, D. 1984, Inquiries into Truth and Interpretation. Oxford: Oxford University Press. Goldfarb, W. 1985. Kripke on Wittgenstein on Rules. Journal of Philosophy 82:471-88. Kripke, S. 1982. Wittgenstein on Rules and Private Language. Oxford: Oxford University Press. Quine, W. V. 1995. From Stimulus to Science. Cambridge, Mass.: Harvard University Press. Quine, W. V. 1996. Progress on Two Fronts. Journal of Philosophy 93:159-63. Wittgenstein, L. (1953). Philosophical Investigations. Mew York, MacMillan. Wittgenstein, L. (1993). Philosophical Occasions 1912-1951. Cambridge, U. K.: Hackett Publishing Co. 2 Davidson's Objections to Quine's Empiricism Lars Bergström 1 Introduction There are many similarities between Donald Davidson's philosophy and W. V. Quine's, but there are also differences. One difference has to do with Quine's empiricism. As everyone knows, Quine has argued that the notorious 'two dogmas of empiricism' should be abandoned, but he himself is definitely still an empiricist (see Quine 1953). Davidson, on the other hand, has drawn attention to a 'third dogma of empiricism', which he attributes to Quine among others. Davidson has argued that this third dogma should also be given up, and he suggests that when it is gone there is nothing left to call empiricism (Davidson 1973-74). In this paper I shall describe this apparent disagreement between Davidson and Quine in more detail, and I shall argue that Quine's empiricism is not really undermined by Davidson's arguments.1 1 I am grateful lo Donald Davidson, Roger Gibson, Sören Hüggqvist, Peter Pagin, W. V. Quine and Folke Tersman for helpful comments on earlier versions of this paper. 17