A striking feature of contemporary analytical philosophy is its concern with exotic anthropological scenarios, stories in which we encounter an isolated and completely alien tribe and try to understand its language and activities. The most important source of this interest is Quine’s discussion of radical translation, which was continued by his most eminent follower Davidson under the label of radical interpretation. Radical translation/interpretation is interpretation from scratch, the attempt to understand the actions and utterances of a completely unknown community without the benefit of any previous acquaintance.1 Both Quine and Davidson use the idea of such anthropological encounters as a heuristic device. Its purpose is to ensure that we approach linguistic behaviour and the problem of meaning from a perspective which they deem proper. The expedition into the jungle is a campaign in support of a philosophical anthropology, a philosophical account of language and human behaviour in general. In Quine, and until recently in Davidson, this heuristic function has been linked to the idea that ‘radical translation starts at home’: all linguistic understanding is based on radical translation, and we have to interpret even our own utterances. Elsewhere I have argued that this is a mistake.2 In my view it is related to an error which Wittgenstein pinpointed in PI §§198–202, namely that of supposing that since it may always become necessary to interpret a rule, all rule following must involve interpretation.
To deny that we always engage in radical translation when we communicate is not to reject Quine’s and Davidson’s approach to genuine cases of radical translation. Indeed, the consideration of radical translation may serve a heuristic role in a philosophical anthropology precisely because it is such a special case. At any rate, this was the view of Wittgenstein, whose remarks on deviant practices and alternative forms of life are the second important source of the analytical debate. Before Quine, Wittgenstein discussed, albeit briefly, the ‘ethnological point of view’ or ‘anthropological method’ which we adopt when coming to understand such an (actual or invented) alien community.3 Like Quine and Davidson, Wittgenstein thought that we can learn something about the concept of understanding by investigating the question of whether and how radical translation is possible, and something about the concept of language by investigating the question of whether there are minimum requirements which a form of linguistic behaviour must meet in order to be intelligible to us.
In this chapter I argue that although Quine and Davidson provide important insights into radical translation, their overall conception of it is flawed, and should be corrected by reference to Wittgenstein’s contribution. I start by arguing that Quinean translation cannot even reach the meagre results it countenances without tacitly relying on hermeneutic assumptions and methods which he explicitly condemns in his arguments for indeterminacy. The second section of the chapter indicates how Davidson’s conception of radical interpretation departs from Quine, but it criticizes him for retaining the idea that radical translation is a matter of constructing a theory on the basis of non-semantic evidence. While the first two sections employ Wittgensteinian ideas to challenge Quine and Davidson, the final section develops Wittgenstein’s own alternative. It ends by sketching very briefly the consequences which conflicting conceptions of radical translation have for the topic of conceptual relativism.
Quine’s discussion of radical translation aims to provide a scientific theory which explains how a ‘meagre input’ of sensory stimulation gives rise to a ‘torrential output’ of structured verbal theorizing. According to Quine, human beings must be seen as black boxes whose ‘dispositions to verbal behaviour’ are triggered by external stimuli—’physical irritations of the subject’s surface’ (ORE 83; WO 207, 235). Verbal behaviour must not be described in terms of meanings, or as correct or incorrect, but only in terms of statistical regularities obtaining between movements, sounds and the environment. This reductionist behaviourism is sketched in chapter 1 of Word and Object, which provides the essential background for the celebrated discussion of radical translation in chapter 2. The discussion of radical translation serves as a thought experiment which describes language in purely extensional and behaviourist terms. The only evidence for radical translation which Quine allows is what sentences the natives assent to or dissent from in what circumstances. The master-problem he addresses is: ‘how much of language can be made sense of in terms of stimulus conditions?’ (WO 26; see also PTb 37, 48).
Quine’s answer is: very little. Beyond certain limits the translation of a completely alien language is ‘indeterminate’. There are mutually incompatible ‘translation manuals’ (WO 27–8), different ways of correlating native sentences with ours, all of which fit the facts about the natives’ linguistic behaviour equally well. What we can establish, according to Quine, is (WO 68):
In order to get even this far we need more than a description of the native tongue in terms of stimulus and response. We also need the famous ‘principle of charity’. According to this principle, our translation manuals should minimize the ascription of false beliefs, especially as regards observation sentences and logical connectives. For, Quine argues, it is ‘less likely’ that the interpretees hold obviously silly beliefs, such as contradictions, than that our translation is wrong (WO 59).
Even with the assistance of the principle of charity, however, translation remains indeterminate in several respects (ORE 67). The first results from semantic holism (WO §9). What has a specifiable empirical content, and hence a specifiable stimulus meaning, is not an individual sentence, but the ‘language’ or ‘theory’ as a whole. Semantic holism has striking consequences for radical translation. We can establish what sentences of the native language are stimulus-synonymous, but we cannot univocally translate these sentences into our language. For we may translate a given sentence differently by making compensating adjustments in the translation of other native sentences. Hence, there are mutually incompatible ways of pairing individual sentences which fit the totality of the natives’ behaviour equally well.
A second dimension of indeterminacy is the ‘inscrutability of reference’ (WO §12). Even if we could assign an objective meaning to the native sentences, we could not establish the referents of the terms occurring in these sentences, since that would depend on how we translate certain other native expressions. Assume that we have established that the stimulus meaning of the native sentence ‘gavagai’ is identical with that of our ‘There’s a rabbit.’ It nevertheless remains impossible to tell what the extension of ‘gavagai’ is, whether it refers to a rabbit, to an undetached rabbit part, or to something else. We cannot even tell whether it is a concrete general or an abstract singular term which refers for example, to a recurring universal, namely rabbithood. For the only way of removing these uncertainties is to ask in the native language questions like ‘is this the same gavagai as that?’ But that presupposes a prior translation of ‘the apparatus of individuation’, expressions like ‘the same’, articles, pronouns, etc. Once more there are different ways of construing the overall behavioural data.
The final dimension is ‘ontological relativity’. Understanding a language—determining its meanings and ontological imports—is doubly relative: not only to one of several possible translation manuals, but also to the choice of one of several possible languages to translate into. We are forced to project the ontology of some ‘background language’ or ‘theory’ onto the native language (ORE 49, 67–8; PL 81–2).
Quine uses the indeterminacy thesis to conclude that the notions of meaning and synonymy, and with them all other intensional notions, are illegitimate, since there are no criteria of identity for ‘meanings’ (PL 1–2, 67–8; ORE 23). There are various ways in which this conclusion might be resisted. One is to insist that talk about meaning does not require any criteria of synonymy. According to this line, even if there is no way in principle to establish whether two expressions mean the same, this epistemological result is irrelevant to the ontological question whether ‘meanings’ exist.4 Both Wittgenstein and Quine tried to avoid such a reification of meanings by replacing talk of meanings with talk of synonymy (M 258; A W L 30; FLPVa 11–12; Q 131–2). Quine would maintain, and Wittgenstein deny, that criteria of identity in general and of synonymy in particular must be context-independent and clear-cut (cf. WO 203 and PI §§ 214–16, 223–7). However, both would insist that talk of meaning presupposes that there are ways of telling whether two expressions mean the same. Although I cannot argue their case here, I think that they are right. Ascribing meaning to a word is not to relate it to an entity, let alone a verification-transcendent one. Such ascriptions would be senseless if there were no ways of explaining what a word meant, which in turn requires the possibility of providing synonyms.
Another reaction to the indeterminacy thesis is to question its specific components, notably the inscrutability of reference. For example, even by Quine’s austere standards it seems possible to determine whether ‘gavagai’ is a count-noun referring to a living animal, and hence is to be translated as ‘a rabbit’, or a mass-noun like ‘roast rabbit’. We watch a rabbit being turned into roast rabbit in a native’s company, and check whether the native still assents to the application of ‘gavagai’.
Yet another reaction is to increase the yield of radical translation through introducing mentalist elements into the behaviourist picture. Thus Dummett and Evans have argued that while a Quinean translation manual which merely pairs native and English sentences is indeterminate, this does not hold for a theory of meaning which would explain how the natives calculate the meaning of sentences from the semantic properties of their constituents.5 But it would seem that Quine can happily reply that the only legitimate evidence we have for such processes is the natives’ behaviour.
In order to resist the thesis of indeterminacy one must undermine the behaviourist methodology on which it rests. One way of doing this is to show that Quine’s method of translation cannot yield even the meagre results it is supposed to, without tacitly smuggling in either a prior understanding of the natives, or hermeneutical methods and intensional notions which he disowns. The answer to the question ‘How much of language can be made sense of by Quinean theory-construction?’ is not ‘very little’ but ‘none whatsoever!’ Either there is a better approach to radical translation than that of Quine, or such translation is impossible.
Given the paradoxical nature of the second possibility, it is tempting to regard this as a reductio ad absurdum of Quine’s behaviourist approach. Quine rejects that accusation (PTb, 37–8).6
His rationale is that ‘the behaviorist approach is mandatory’, because in learning a language ‘we depend strictly on overt behavior in observable situations’. But although there is no alternative to learning the natives’ language on the basis of what they say and do, there is an alternative to describing what they say and do in the behaviourist idiom of stimulus and response, an alternative to be found in Wittgenstein. If this is correct, the strategy of showing that Quinean translation is a non-starter holds out the promise of a reductio of behaviourism.7
Such a strategy must avoid armchair anthropology by keeping apart factual and conceptual issues. Both Wittgenstein and Quine rightly agree on the anti-genetic point that it does not matter how a language is acquired (see BB 12; PG 188 and RTC 138, 95, 206; WPEb 119–20). There is no contradiction in supposing that creatures might start to speak English without having learnt it at all. Equally, Quine and Davidson might return from the jungle with a perfect grasp of the native tongue, however austere their procedures. The question is whether they distort radical translation through a mistaken account of what it is to learn a completely alien language.
This anti-genetic lesson applies to a prima facie plausible misgiving about Quine’s method. Even sympathetic commentators sometimes complain that the Quinean discussion tends to ignore the fact that radical translation involves interaction between translator and native.8 However, as it stands, this objection is doubly inaccurate. For one thing, the Quinean translator is not in the hopeless position of someone who attempts to learn a language with the exclusive aid of tape-recorders and microphones dangling from trees.9 For he also observes the natives’ movements and their environment. Nor is he confined to observation. Rather, Quine tells us, he ‘takes the initiative’ (WO 29) by trying to elicit responses by uttering native observation sentences, for example in the identification of assent and dissent (see below). For another, contrary appearances notwithstanding, interaction is not an essential precondition for successful translation. Imagine an invisible translator along the lines of H.G. Wells’s Invisible Man, who can move freely among the natives without interacting with them or being noticed. There is no a priori reason why such an invisible translator should not be able to pick up the native language by observing the linguistic instructions given to native children. If they can do it, why shouldn’t he? Like them, he would (as a matter of fact) need to be privy not just to the initial stimulations and explanations given by the natives, but also to the corrections and clarifications with which they react to their children’s first attempts. Of course, his learning would be facilitated if he could ask questions and have his own mistakes corrected. In principle, however, the invisible translator could learn through mere observation, just as prodigies have learnt chess.
Consequently, Quine cannot be accused of ignoring the need for interaction. What he can be accused of is mischaracterizing this interaction. He is committed to describing the translator’s interference in native affairs as a matter of providing stimuli for a black box, a piece of verbal machinery with a certain input and output. It is partly because of this reductionist behaviourism that Quine cannot even reach the meagre result he promises. There is also another aspect of Quine’s method which has this effect. As we have seen, in his argument for the inscrutability of reference, Quine thinks that any procedure which is based on assumptions that are optional is unable to provide evidence for translation. This stands in stark contrast to what is known as the hermeneutical circle, the idea that in understanding a remote text or culture we have to start by making certain prima facie plausible assumptions about specific passages or actions, the validity of which is then checked against the plausibility of the overall interpretation to which it leads, which in turn is modified by reference to the rectified understanding of the specific passage, and so on. This approach would avoid inscrutability of reference in a realistic fashion. Unless we have reason to believe that the natives are more interested in rabbit parts or Platonic ideas than they are in rabbits, we shall start with the assumption that ‘gavagai’ refers to the whole animal. On the basis of that assumption, and others of a similar kind, we then provide a translation of the natives’ apparatus of individuation, which is tested by its plausibility in other cases, and so on. Quine is committed to ruling out this procedure as inadequate. But, as I shall argue, he himself tacitly relies on it in reaching his own results. He insists on a presuppositionless method in denying the scrutability of reference, but he is a closet hermeneuticist in the translations he licenses. Hence, we can either translate much more than Quine allows, or nothing at all. There are three prominent points at which this objection might apply, namely the move beyond observation sentences, the translation of the truth-functional connectives, and the identification of assent/dissent.
In order to get beyond the threshold of observation sentences like ‘it is raining’ to which all speakers assent in the same situations, independently of their background information, we need to ‘go bilingual’ first; that is, we must learn the native language as children do, and then translate ‘by introspected stimulus-synonymy’ (WO 46–7). Unfortunately, it is unclear what that phrase means, or even whether it is meant seriously. The most plausible gloss on it is that the bilingual translator observes things like ‘Whenever I would assent to “Ich bin ein Berliner” I would also assent to “I am a doughnut.”‘ But according to that gloss the translator does not introspect anything. Rather, he observes his own behaviour and makes inferences about his dispositions, just as he has previously done with the natives. Yet it is highly implausible to suggest that a bilingual person needs to establish his behavioural dispositions in order to explain the words of one language in terms of the other.
By Quine’s own acknowledgement, ‘going native’ transcends his method of translation; but it does not transcend his overall account, since he provides a behaviourist account of language acquisition (WO, ch.1). In any event, this first breakdown need not disturb him excessively, since the move beyond observation sentences is not listed among the results of his method. By contrast, univocal translation of the truth-functional connectives is. At first blush this seems justified. We can translate, for example, the native ‘blip’ as ‘and’ if for any two native sentences ‘p’ and ‘q’ the native assents to ‘p blip q’ if and only if he will assent to p and to q. However, this account already relies on assumptions which seem no less problematic than those which Quine deplores vis-à-vis the inscrutability of reference. Just as distinguishing between rabbits and rabbit stages presupposes prior translation of the apparatus of individuation, translating truth-functional connectives presupposes that
The final breakdown of Quine’s method, concerning the identification of assent/dissent, is the most fatal, since it would prevent Quinean translation from even getting off the ground. Radical translation cannot proceed by observing what situations prompt what utterances, since the motives for making or withholding an utterance vary widely. As the critics of earlier behaviourist theories of meaning pointed out, even the perspicuous presence of a rabbit may not lead the native to say ‘gavagai’: he may be too accustomed to or fed up with the sight of rabbits. To avoid these pitfalls, Quine insists that the translator himself must volunteer sentences in appropriate circumstances, ‘asking only for a verdict of true or false’ (TT 48). But at the same time it is essential to his behaviourist approach that the dispositions to assent or dissent which this procedure is thought to establish provide the only starting point of radical translation.
This restriction of the basis of radical translation is implausible. Putnam has urged that the evidential basis of radical translation should be extended from assent or dissent to declarative sentences, to include the natives’ questions or demands for explanation like ‘What does this word mean?’, which are no more difficult to identify than assent or dissent.10 This is plausible: even an invisible translator who cannot ask for explanations will have to recognize when the natives explain something to their offspring, and how they correct mistakes. But this is not just a low-key modification of Quine’s approach, as Putnam seems to suppose. For, as Wittgenstein showed, explanations of meaning are standards of correctness by reference to which subsequent applications of a word are assessed as correct or incorrect, meaningful or nonsensical (PI §54; PG 68, 143; M 276). Taking into account explanations and corrections introduces the idea that radical translation involves learning certain linguistic norms, a point of disagreement between Quine and Davidson on the one hand and Wittgenstein on the other (see the second section below).
If one accepts Quine’s behaviourist approach, and hence that the evidence of radical translation is confined to assent/dissent, the question is whether this evidence is available to a Quinean translator. Quine suggests that native assent/dissent can be identified as follows:
in asking ‘Gavagai’…in the conspicuous presence of rabbits …he has elicited the responses ‘Evet’ and ‘Yok’ often enough to surmise that they may correspond to ‘Yes’ and ‘No’, but has no notion which is which. Then he tries echoing the native’s own volunteered pronouncements. If thereby he pretty regularly elicits ‘Evet’ rather than ‘Yok’ he is encouraged to take ‘Evet’ as ‘Yes’…. However inconclusive these methods, they generate a working hypothesis.
(WO 29–30; see RIT 181n.)
Note in passing that it is unclear what sort of behavioural evidence could lead one to surmise that ‘Evet’ and ‘Yok’ are expressions of assent or dissent without indicating which one expresses assent and which one expresses dissent. I shall also leave aside the problem that our ‘Yes’ and ‘No’ have other functions besides the expression of assent and dissent. The real trouble is that this procedure is not presuppositionless. For a start, it presupposes that the translator has correctly translated the observation sentence—in our case ‘gavagai’—which he uses to elicit assent and dissent from the native. But that translation is subject to the vagaries which the improved behaviourist procedure of relying on assent and dissent was supposed to exclude, namely that the natives utter or fail to utter ‘gavagai’ for reasons that are extrinsic to the presence or absence of rabbits. Moreover, Quine’s procedure presupposes a mutual understanding between native and translator to which he is not entitled. It is assumed that the native understands that the translator’s ‘gavagai’ is meant as a question concerning the meaning of that expression, and not as a religious ritual involving rabbits, an attempt to query his hunting-rights, or simply a dumb repetition. In that last case, an ‘Evet’ in response to the translator’s echoing the native’s utterances would not be a sign of assent, but a rebuke for parroting.
These possibilities of misunderstanding are illustrated by the (apocryphal) etymology of the English term ‘kangaroo’. Supposedly one of the first Europeans pointed at a kangaroo and asked ‘What is this?’, to which the aborigines replied ‘kangooroo’, which in their language means ‘I haven’t a clue what you are talking about!’ There is also the all too real story of the head-hunters in New Guinea who gave their children the names of slain members of other tribes. Before killing their victims, they always asked them for their names. However, they often attacked distant communities with languages completely unknown to the head-hunters. As a result, the replies they got at knife-point were phrases like ‘Go to hell!’ or ‘Have mercy!’, which they nevertheless duly incorporated as proper names into their language.11 A more amusing case is that of the first French translation of Tongan. That language does not contain numerals above twenty. But when the French translator Labillardière persisted in asking for such numerals, he received expletives in reply, which he solemnly noted as Tongan numerals.12 Finally, a case I have personally witnessed. On the summit of a peak in the Alps a Prussian asked a local for the name of one of the many mountains to be seen. The local replied ‘Wehler?’, the Bavarian version of the High-German ‘Welcher?’ (Which one?). The Prussian was content with what he took to be an answer, and descended in the belief that he had seen the impressive Wehler-peak.
Such possible and actual misunderstandings demonstrate that Quine’s procedure presupposes that interpreter and native engage in a specific kind of dialogue, that is, perform certain types of speech-acts. Quine takes for granted that the native tries to teach his language to the translator, which (among other things) means that he will apply words in paradigmatic situations, and will try to correct the translator’s attempts to imitate his usage. Quine’s austere procedures do not, and cannot, account for this mutual understanding. But the only alternative to taking it for granted is to assume that the native knows that the bald white man from Harvard is trying to establish the stimulus meaning of his words. For Quinean translation to work, the natives had better read a translation of Word and Object!
To this one might object that Quine acknowledges that his method is ‘inconclusive’ and merely generates a ‘working hypothesis’ (WO 30). However, without assuming a framework of interaction, identifying assent/dissent would not just be hypothetical, a reasonable if inconclusive guess: it would be completely arbitrary. There would be no reason to suppose that the native’s reaction is at all relevant to assent or dissent. Moreover, any concession that the identification of assent/dissent is not presuppositionless means that Quine is here applying different standards from those at work when he propounds the inscrutability of reference. If there is no fact of the matter as to whether ‘gavagai’ refers to rabbits, then, by parity of reasoning, there is no fact of the matter as to whether the native assents to or dissents from the translator’s ‘gavagai’. Within Quine’s framework that would remove the possibility of translating anything, and hence lead to the nihilistic conclusion that understanding is impossible.
There is yet another problem with Quine’s procedure. In characterizing the native’s reaction as assent and dissent, he describes the output of the behaviourist experiment in richer terms than the input. The former is held to consist of surface irritations, more specifically, of patterns of stimulation at the surface of the perceptual organs. Quine prefers neural stimulations to sense data, because the empirical foundations of knowledge and language are intersubjectively accessible (a point urged by the physicalists of the Vienna Circle against the phenomenalists). But this has the disadvantage that the epistemic subjects are not aware of the alleged foundations of their beliefs. Quine realizes that most sentences are not about surface irritations, but insists that ‘some of them are elicited by surface irritations, and others are linked to surface irritations in less direct and more tenuous ways’ (TT 40). However, this amounts to a confusion of the causes of our beliefs—which include neural stimulations—with the evidence on which they are based, which the subject must be able to adduce, at least when prompted. Only the subject’s evidence is relevant to describing conditions of assent or dissent. For assent and dissent are not mechanical reactions, but forms of intentional (linguistic) behaviour. If the native screams ‘Yok’ because of being stung by a hornet, he has not dissented from the anthropologist’s ‘gavagai’. They are also intensional. One assents to or dissents from what is said, namely that things are thus-and-so. Against this last point Quine might insist that one assents to token sentences, since he regards ‘events of utterance’ as the bearers of truth and falsity (PL 13–14). But even if that view were tenable, it would not solve the problem at hand. For Quine himself states that to assent to a sentence is to pass a verdict on its truth which may be mistaken, and that the subject believes what is uttered (TT 48). This in turn implies that assent and dissent are not mechanical reactions, but responses to something that the native has understood, namely the anthropologist’s utterance.
This shows that the concept of assent that Quine actually deploys is intimately interwoven with epistemic and intensional notions. Quine might try to escape the objection by admitting (consistently with his treatment of semantic notions) that he should not really talk about assent or dissent but only about a behaviourist ersatz. It is unclear, however, what this ersatz could look like. To speak simply of positive and negative responses to verbal stimuli, for example, leaves open in what sense responses are so classified. In any event, Quine could not settle for such an ersatz. Unless assent expresses what the native believes to be true, it (and hence the notion of stimulus meaning which is defined by reference to it) becomes irrelevant not just to questions of meaning, which Quine might happily accept, but also to epistemology including his own ‘naturalized epistemology’. There would be no point in trying to minimize ascription of false beliefs, as the principle of charity bids us do. More generally, Quine’s whole discussion of radical translation would lose its point, which was to explain the link between our beliefs and theories and the data on which they rest. Another way of putting the point is this. Quine defines the (positive) stimulus meaning of a sentence as the class of all stimulations which would prompt assent. Stimulus meaning and stimulus-synonymy are not supposed to be more than a behaviourist ersatz of the discarded intensional notions (WO 66). However, they are not supposed to be less than Carnapian ‘explications’ of these notions. That is to say, they are alternatives which avoid the drawbacks of the originals (in our case, lack of criteria of identity), while serving their cognitive purposes (WO §§53–4). In our case this means that the notions of stimulus meaning and stimulus-synonymy should capture the ideas of cognitive significance and cognitive equivalence respectively (TT 47–51). This in turn means that stimulus meaning must include only stimuli which the speaker understands, and which are hence relevant to his beliefs. Unlike our intensional notions, a consistent behaviourist ersatz would lack the conceptual connections with epistemic concepts like belief, knowledge, etc. Quine repudiates our intensional concepts in the name of a reductionist behaviourism. But if my line of argument is correct, he must tacitly rely on these concepts, if his discussion is to have the implications he assigns to it.13
Davidsonian ‘radical interpretation’ differs from Quine’s radical translation in four important respects.14 Davidson rightly rejects Quine’s notion of stimulus meaning, on the grounds that it is based on the empiricist dogma that epistemic intermediaries, in our case neural stimulations, intervene between the world and our sentences. Consequently, he describes the conditions of utterance not in terms of surface irritations but in terms of macroscopic objects and events. Secondly, Davidson seeks to provide a ‘theory of meaning’ rather than a mere translation manual. Whereas the latter merely correlates the native sentences with ours, such a theory is supposed to specify what the sentences of both languages mean, namely by stating their truth-conditions. Although he is not consistent on the matter, Davidson often disowns the claim that competent speakers tacitly know his complex theory (which would introduce elements of which the subject is unaware, like Quine’s neural stimuli). Rather, he confines himself to the claim that someone who did know the theory would be able to speak the language. Thirdly, unlike Quine (at least in his early and middle writings), Davidson does not reject psychological terminology. Consequently he describes the task of interpretation as one of assigning meaning to the natives’ utterances, attributing ‘mental states’ to them (in particular propositional attitudes like beliefs and desires) and understanding their actions. These tasks hang together holistically (ITI 127). We could ascribe meaning to the natives’ utterances if we knew their beliefs and desires, and vice versa. Alas, at the start of radical translation we know neither what the natives mean, nor what they believe and desire. This holism of meanings, beliefs and desires leads to a final difference. Davidson treats charity not just as a pragmatic maxim of interpretative theorizing, but as a principle which is essential to the correctness of an interpretation. An interpretation which fails to make the natives’ beliefs come out as largely true, and their desires as largely intelligible, is not just less likely to be adequate: it must be inadequate. Davidson’s rationale for taking this line is that our only way of breaking into the holism of meanings, beliefs and desires is to maximize agreement with the interpretees, by assuming that most of their beliefs are true, and hence chime roughly with our own beliefs. If we find that translation is impossible because we cannot construe the natives’ beliefs and desires as by and large rational, we end up not with a less probable translation, but with the conclusion that they do not speak a language and do not engage in intentional action. Accordingly, we could never be in a position to judge that the natives had beliefs and desires radically different from ours (ITI 197). This introduces a normative element into linguistic understanding. We can make sense of others only in so far as we can treat them as agents who abide by certain standards of rationality.
To a considerable extent, therefore, Davidson’s philosophical anthropology moves away from Quine’s reductionist behaviourism. At the same time, however, this move is half-hearted and leads to tensions. To begin with, it is unclear whether Davidson can square his professed aim of ascribing meaning to utterances with his explicit acceptance of Quine’s thesis of the inscrutability of reference.15 Moreover, using charity as a normative principle that is constitutive of linguistic understanding seems incompatible with Davidson’s claim that Quine’s rejection of the analytic/synthetic distinction ‘has saved the philosophy of language as a serious discipline’.16 Obviously, there is no logical guarantee that native forms of behaviour which we might encounter will be rational in the sense required by the principle of charity. The normative force of the principle must be that unless we can treat the natives as rational, we cannot describe their behaviour as language. This suggests that this normative force derives from what we call ‘language’, ‘intentional action’, ‘rational behaviour’, etc. Hence it was only natural for David Lewis to put it to Davidson that on his account it must be analytic that anybody to whom we can ascribe prepositional attitudes must satisfy the conditions of the principle of charity. As a Quinean, Davidson rejected that slanderous imputation.17 More recently, however, he found himself compelled to admit that it ‘cannot be a factual question’ whether a creature with prepositional attitudes is approximately rational.18 Davidson needs to treat his own pronouncements on rationality as conceptual in precisely the sense he denounces elsewhere.
More serious than this straying from the Quinean line is that in other respects Davidson follows Quine too closely. For one thing, like Quine, Davidson identifies language and theory. But a language like English is not a theory. Even if Quine and Davidson were right in treating a language as a set of sentences, it is not a theory, since it must contain both sentences and their negations, which a coherent theory cannot. Moreover, the identity of a language is determined not by sentences, but by the principles for the formation of sentences, i.e. what Wittgenstein calls grammatical rules. The fact that Americans do not use the sentence ‘God save the Queen!’ does not show that their language differs from that of the British, since the rules of both idioms allow for the construction of that sentence. Finally, unlike a theory, a language does not predict anything, nor does it fit or face reality, and it cannot be true or false. Rather, it is statements in a language which do so, and which are potential constituents of theories.
In line with this assimilation, Davidson characterizes learning a language as constructing a theory. Equally, domestic understanding, for example my understanding of a particular utterance in English, is characterized as a matter of deriving hypotheses concerning that utterance from the general theory, which in turn is modified in the light of the success of that hypothesis. Unlike Quine, he distinguishes between theory construction in the natural and the psychological sciences, since the latter inevitably relies on canons of rationality, notably the principle of charity. Nevertheless, the requirement of rationality is superimposed by the interpreter on an array of brute data. The evidence of understanding, both domestic and anthropological, is ‘non-linguistic goings on’ like the movement of the lips and larynx.19 All we really perceive are sound patterns and bodily movements.20 Like Quine, Davidson extends the empiricist myth of the given from perception to speech and action. Linguistic understanding is a case of theory construction on the basis of observing conditions of utterance. This theorizing need not conform to the requirements of Quine’s behaviourism. The stimulus conditions may be described in terms of macroscopic objects and events rather than by reference to neural stimulations. But they must not be described in semantic or intensional terms like meaning, beliefs, desires, intentions. A ‘specifically semantical’ theory employing such concepts is to emerge only as the result of theorizing on the basis of ‘non-semantical evidence’.21
Moreover, Davidson joins hands with Quine in denying that language should be described in terms of linguistic rules, conventions or language-games.22 Even in our own language, they claim, what we encounter are not utterances which can be described as correct or incorrect, meaningful or nonsensical by reference to rules accepted by a linguistic community. Rather, we are given an array of sounds and movements non-normatively, and we confront the task of extrapolating their ‘meanings’ through explanatory hypotheses informed by, and in turn informing, a truth-conditional theory of meaning. A Davidsonian interpreter theorizes under the guidance of principles of rationality, but it remains a process of empiricist theory construction.
This contrasts sharply with Wittgenstein’s anti-reductionist and normative approach. Wittgenstein views language as an activity which is structured by rules, standards for the correct use of words which are evident in our explanations and corrections. Moreover, Wittgenstein insisted that we should not try to reduce concepts like rule following to anything more basic, since they are ‘FUNDAMENTAL’ to our linguistic practices (RFM 330). To naturalists, Davidson’s approach may appear to have an advantage over Wittgenstein’s, even if they reject Quine’s reductionism. Whereas Wittgenstein seems to take ‘higher phenomena’ like the normative nature of language for granted, Davidson holds out the promise of a naturalist transcendental argument: he does away with the idea of linguistic rules, and derives normative standards of rationality as preconditions of interpretative theory construction.
Against this I should like to argue that it is neither possible nor necessary to replace the normativist picture of language by that of interpretative theory construction. Both behaviourists like Quine and mentalists like Chomsky have characterized language acquisition as theory construction, respectively maintaining or denying that it can be done on an austere empiricist basis. Davidson seems committed to a similar picture. The ‘prior theories’ which, according to him, we adopt in everyday communication and adapt according to the principle of charity must presumably have been the result of theorizing during language acquisition. This underlying assumption, however, is absurd. The ability to engage in scientific theory construction clearly presupposes the ability to speak a language, and a very complex one at that.23 Some of Chomsky’s followers have been alive to this point and have been driven to the conclusion that in order to learn a language the child must already possess a ‘language of thought’. But that idea suffers from the same defect as Plato’s explanation of knowledge by reference to anamnesis: it simply pushes one step further back the problem of how we came to acquire any language.24
Even if we are entitled to presume language possession on the part of the hearer, Davidson’s picture is awry. For one thing, like Quine he insists that the only empirical evidence available to either enterprise concerns what people assent to under what circumstances.25 In Davidson’s case, this assumption seems mainly motivated by his desire to apply to radical translation a Tarskian theory of truth. But, as we have seen, the assumption is unwarranted. An anthropologist will often rely on identifying questions and explanations. And there are situations in which she would most conveniently start out from orders, exclamations and requests.
Moreover, understanding is not a matter of inferring the meaning of utterances or the mental states of speakers from a description of mere sounds and bodily movements. While we find it easy to describe human actions and utterances in the ‘rich’ semantic and intentional terms Davidson precludes, we are ignorant of the austere physical descriptions he condones. As Wittgenstein noted, we can describe a person’s features as ‘sad’, ‘radiant’ or ‘bored’, but do not know how to describe a person’s face in physical terms (Z §225). And as someone who has taken a course in phonetics I vouch for the claim that even a complex philosophical lecture is easier to understand than to describe in terms of its physical or phonetic features. In other words, we are able to state the conclusions of Davidsonian theorizing without necessarily being able even to understand the data from which they are allegedly derived. This suggests that the idea of theoretical inferences is misplaced here.
One might reply that the inference is subconscious. However, while human speech involves complex causal processes of which we are unaware, the physical and neurophysiological causes of our linguistic behaviour are not pieces of evidence from which we derive the meaning of what has been said. Davidson himself would not deny this. Instead, he might opt for a different line of defence.26 The non-semantic evidence which underlies the theory construction is not a phonetic description of the speaker’s utterance, but simply a reproduction of the utterance which the hearer is able to provide. In the light of her prior theory, she will then derive a T-sentence like ‘A’s utterance of “Snow is white” is true iff snow is white.’ In response I should like to ask what it is that the hearer has to be able to reproduce. Either it is the precise acoustic phenomenon—but that is even more difficult than giving a phonetic description, and certainly not required for understanding; or it is a token of the type-sentence like ‘Snow is white.’ But in that case the evidence is not pre-semantic, since to characterize something as a token of an English type-sentence is to characterize it as belonging to a particular linguistic system.
These considerations leave intact the austere version of Davidson’s project. That version is committed only to the view that the meaning of sentences and the significance of intentional actions could be derived from evidence which concerns only what people assent to under what circumstances. We can provide neutral phonetic descriptions of human language, and we could develop equipment to provide similarly neutral descriptions of facial expressions. However, this does not suffice to secure the evidence that Davidson requires. For, as argued above, identifying assent and dissent presupposes that anthropologist and native engage in a certain type of communication, and hence a certain kind of semantic knowledge.
Even if one grants the Davidsonian interpreter that evidence, there are reasons to suppose that his task is no less hopeless than that of the Quinean. For we lack the inferential procedures which would allow us to infer from such descriptions the meaning of utterances and actions. Davidson himself does not specify such procedures. This point is obscured by the fact that in his concrete examples he describes the macroscopic objects and events in the interpretees’ surroundings not in completely neutral—if macroscopic—terms, but in the kind of terms which occur in the native observation sentences themselves, for example as a ship passing by, or a tin of paint, or someone uttering ‘Snow is white.’ These terms, however, are not purely geometrical or physical. They are everyday terms which incorporate the (epistemic and conative) significance of those objects and events for creatures like us. And it is only because they do that we can apply rationality principles like the principle of charity to the observation sentences which contain them.
In my view Davidson’s failure to specify inferential procedures is no coincidence. The attempt to extrapolate the meaning of utterances from a physical description of sounds and movements is as absurd as trying to solve the problem ‘A ship is 20 feet long and 6 feet wide: so how old is its captain?’ (see BT 494). This is the point at which the Quinean discussion of radical translation converges with the Wittgensteinian discussion of rule following. Both show, in different ways, that such pre-semantic evidence leaves the meaning of our words and the sense of our utterances necessarily under-determined. Any finite sequence of numbers is compatible with indefinitely many functions. By the same token, any finite array of behaviour is compatible with ‘any number of rules’, if it is described in presemantic terms (BB 13). This means that any extrapolation of rules from behaviour neutrally described is, in principle, underdetermined. It does not mean, as rule-scepticism suggests, that the rule leaves its application under-determined. Whereas the relation between phenomena described in neutral (behaviourist, physicalist or naturalistic) terms is external, the relation between a rule and its application is, as Wittgenstein shows, internal (WWK 152–7): it is logically impossible that they should not stand in this relation, since the relation is constitutive of the relata. That a given behaviour is conducted according to such and such rules may indeed be an explanatory hypothesis or conjecture of an uninitiated observer. But that does not mean that those rules do not determine what counts as their correct application.
Such internal relations are de dicto, i.e. they depend on how we describe things (this is something Davidson accepts in his celebrated attack on the distinction between reasons and causes). The internal relation between a rule and its application is lost if the relata are described along the lines of Quine and Davidson, namely in presemantical, non-normative terms. If the rule formulation ‘Add 2’ and the utterance ‘1000, 1002, 1004’ are described phonetically, it is impossible to tell whether the latter is a correct application of the former. However, this is possible if both are described in terms of our normative practice in which the rule functions as a standard of correctness. In fact ‘1000, 1002, 1004’ is the correct continuation, since this is what we call ‘adding 2’ or ‘the series of even integers’. To insist on internal relations does not introduce any mysterious, supernatural phenomena. Internal relations are neither abstract nor mental. They are effected by our normative practice—the fact that we introduce, teach and explain standards of correctness, and criticize or justify performances by reference to them (see PI §201, LFM 83). But such relations emerge only if we describe human activities at the normative level at which the participants themselves do.
Both Quine and Davidson refuse to do this. At best they acknowledge assent/dissent under certain conditions. Normativity, by contrast, implies a distinction between two kinds of dissent: (a) rejecting an utterance as false, i.e. unfair to the facts; (b) rejecting it as incorrect, meaningless or nonsensical by reference to standards of correctness. Quine and Davidson repudiate this dichotomy. But it can be argued that without it linguistic meaning, the starting point of radical interpretation, vanishes. If an utterance like ‘The number 1 has an Italian hairdresser’ had the same logical status as an utterance like ‘Hanjo Glock has an Italian hairdresser’, namely that of being perfectly intelligible but false to the facts, the use of number-terms would have become completely arbitrary, and hence these terms would have lost all meaning. A practice without this distinction between the false and the senseless would at best be a communal phonetic babbling. In such a practice I could treat your utterance ‘I just met the number 1 with its new hair-cut’ as unusual, out of the statistical norm. But I could not reject it as unintelligible or demand an explanation. In such a Quinean scenario utterances and situations might still be linked by regularities. As a result hearers might still predict the behaviour of speakers on the basis of their utterances, and speakers might use words with the intention of causing a certain behaviour in hearers. But linguistic utterances would merely be empirical indicators of other phenomena, just as clouds indicate rain. They would have some indicative value (natural significance), but they could not be understood as having linguistic meaning. But without linguistic meaning there is no such thing as true or false statements, and hence no such thing as the assent or dissent which Quine and Davidson are preoccupied with.27
I have argued that neither domestic understanding nor radical translation is, or could be, a matter of constructing explanatory theories on the basis of pre-semantic evidence. The Quinean and Davidsonian methods of translation do not just differ from what we actually do in communication or radical translation; they are not up to the job. What makes ordinary understanding and interpretation possible is not evidence beyond human behaviour, as the mentalist opponents of Quine and Davidson have urged. Rather, it is the fact that for the participants such behaviour is ab initio infused with meaning and intentions. What we encounter are not mere sound patterns and bodily movements, but rule-guided behaviour.
Unfortunately, even if correct, these considerations do not solve the problem of how radical translation is possible. All they show is what it amounts to—not theory construction, but being introduced into a normative practice. We learn that certain utterances in certain situations count as saying such-and-such, that words can be combined in specific ways but not others, that it is a mistake to refer to certain objects by certain words, etc. We acquire a technique, and this will usually be a communicative and interactive process: we receive explanations and instructions, practise certain constructions, and are corrected or encouraged.
The problem is that in radical interpretation we ex hypothesi are not at first in a position to describe the native utterances and activities in the normative terms available to the participants. Here we are constructing what Quine calls ‘analytical hypotheses’. But what is their basis, given that it is a matter neither of straightforward application of familiar rules, nor of empiricist theory construction? It is clear that we must enter a hermeneutical circle here, and correct a provisional understanding of parts of the native language by reference to our understanding of the whole. The discredited demand for presuppositionless translation notwithstanding, the actual history of radical translation shows that this circle is not vicious, and does not lead to any indeterminacy. Definite mistakes have been made, and have definitely been rectified (as we have seen). But this leaves open the question of how precisely the circle operates in the communicative process of radical translation. How do we recognize, for example, that the native is explaining something to us, or correcting our first efforts? And how do we make sense of his explanations and corrections? Wittgenstein intimates an answer to these urgent questions:
The common behaviour of mankind is the system of reference by means of which we interpret an unknown language.
(PI §206)
We can solve the problem of radical translation, because we share with the natives certain basic forms of human behaviour.28 This idea is linked to the claim that ‘the speaking of language is part of a form of life’, i.e. of a communal practice in which our language-games are embedded (PI §23; RFM 335).
Some commentators have claimed that for Wittgenstein there is only one form of life for humans, and that different forms of life, notably those of non-human animals, are simply unintelligible to us.29 This would mean that what allows us to translate an alien language is the fact that we share with its speakers a common form of life, namely the human form of life. Wittgenstein often speaks of forms of life in the plural (e.g. PI II 226; RPPI §630; CE 404). But what he has in mind here are specific facts about human behaviour—he also calls them ‘facts of life’—which together characterize a form of life, a totality of communal activities.
Nevertheless, there are other reasons against ascribing to Wittgenstein the idea that there is a unique human form of life. One is his insistence that alternatives to our own conceptual schemes (what he calls ‘grammars’ or ‘forms of representation’) become intelligible if we assume that their protagonists lead a different kind of life, i.e. engage in communal practices which are based on different types of training and serve different purposes (PI II230; Z §§352, 387–8; RFM 38, 91; LFM 83, 201–2). Thus he imagines communities in which people measure with elastic rulers, or even sell piles of wood according to the area they cover, irrespective of their height. Another reason is that what Wittgenstein calls ‘the natural history of human beings’ (PI §415) includes not just basic activities which are shared by all human beings because of their inflexible biological make-up, but cultural activities which vary according to different times and places, such as measuring or doing mathematics and logic (RFM 352–3, 356, 399; RPPI §1109). In view of these facts it is reasonable to assume that ‘form of life’ does not refer to our common biological nature, but to a culture or social formation which is not shared by all human beings.
At the same time, like Quine and Davidson, Wittgenstein insists that there are minimum requirements which a form of linguistic behaviour must meet in order to be intelligible to us. Our form of life need not be identical with that of the natives; after all, even if we leave aside Wittgenstein’s fictional cases, we have managed to translate very remote languages such as Linear B, and to interpret very alien cultures, like that of the New Guinea head-hunters. But we could never start the hermeneutical process unless we shared with the interpretees certain forms or facts of life (RFM 414, 421).
This idea lies behind Wittgenstein’s puzzling remark ‘If a lion could talk, we could not understand him’ (PI II 223). On one reading this means that we could not understand a lion who utters English sentences like ‘I’m not interested in you: I just had an antelope’, which is obviously false (although one might, following Austin, question whether such a talkative creature could count as a lion). On a charitable reading, it means that if lions had a feline language of complex growls, roars, etc., we could never come to learn it. Why? Because their form of life, and their behavioural repertoire, are so alien to us. We could not make head or tail of their facial expressions, gestures and demeanour. Moreover, our ability to interact even with a tame lion is strictly limited. For related reasons we ‘could not find our feet’ with human beings who give no expression of feeling of any kind, and would be completely at a loss with spherical Martians (RPPII §568; Z §390; LC 2–3).
At this point it is imperative to be clear about what kind of things we need to share with the natives. What precisely does Wittgenstein’s ‘common behaviour of mankind’ include? What should it include? And how do his preconditions of radical translation differ from those of Quine and Davidson? One clear example is the kind of interaction tacitly presupposed by Quine. Unless the natives shared our desire to communicate with foreigners, as well as language-games of querying and correction, the mutual instruction between explorers and natives would not take place. This kind of interaction characterizes actual anthropological field-work. But if the above fiction of the invisible interpreter is coherent, it is not conceptually essential to radical translation. However, other features mentioned by Wittgenstein arguably are.
One such feature is behavioural universals. Thus he writes that the justification for translating words of an alien language as expressions of doubt or certainty ‘lies mainly, if not exclusively, in gestures, the facial expressions of the speakers, and their tone of voice’ (EPB 149; my translation). Unfortunately, Wittgenstein did not always stick to this insight, because he is sometimes too impressed by the cultural diversity of gestures. He suggests not only that we ‘wouldn’t know what genuine joy looks like with the Chinese’ (LWPPII 89; my translation), but even that ‘we understand Chinese gestures as little as Chinese sentences’ (Z §219). In the same vein, Quine mentions the idea that radical translation might be based on characteristic forms of behaviour such as gestures, only to object that gestures ‘are not to be taken at face value; the Turks’ are nearly the reverse of our own’ (WO 29).
This dismissal is precipitate. For example, although the Turkish gesture of dissent involves a vertical movement of the head, it is not nodding and can be recognized as a gesture of rejection since it also involves a sound which is clearly dismissive. Equally, Wittgenstein’s first claim is plausible, because the distinction between genuine expressions and pretence often relies not on straightforward criteria, but on very ‘fine shades of behaviour’, which are accessible only to observers familiar with the culture and personal character of the subject (see, for example LWPP II 61–8). But this does not license the implausible claim that we are as ignorant of the gestures and facial expressions of the Chinese as we are of their language. Without knowledge of Chinese culture we may have difficulties in distinguishing a genuine from an insincere smile, or an embarrassed smile from a relaxed smile, but we can distinguish either from scowling, for example. Moreover, even the distinction between genuine and insincere poses problems only in those cases in which the emotional import of the situation is unclear or ambivalent. Similarly, it is for the most part straightforward to distinguish threatening and submissive gestures, since these are tied up with characteristic forms of human action, and the gesture of pointing is shared by all known cultures.
As a matter of empirical fact some features of human behaviour— concerning gestures, facial expressions, demeanour and intonation—have transcultural significance.30 In another passage Wittgenstein himself suggests that one can recognize the behaviour characteristic of correcting the violation of a rule even in an unknown language (PI §54). If he is right, even an invisible radical translator could establish whether the native behaviour is indeed rule-guided, and profit from the natives’ specific corrections. Yet by itself, this point of contact is insufficient, however important it may be. For he could never make any reasonable guesses as to the impetus of a correction unless he shared other features with the natives. Some of these shared features are part of our animal nature, such as our needs for drink, food and shelter, our sexual drives and our reactions to physical danger. Others are the preserve of cultural and historical beings—such as our curiosity about what is alien, or our fascination with death. These shared features are not exclusively, or even primarily, cognitive in nature, but comprise conative and affective aspects of our lives. Thus we could not identify assent and dissent unless the natives shared certain fundamental preferences with us, such as the acceptance of food or drink, or the refusal of unpleasant things. This insight is prominent in Wittgenstein’s insistence that radical translation requires a substantial overlap in forms of life. It is absent from Quine’s version of the principle of charity, but present in Davidson’s, since the latter insists that it must be possible to treat both the beliefs and the desires of the aliens as largely rational.
At the same time Quine’s and Davidson’s principle of charity expresses important insights about the cognitive preconditions of radical translation. For one thing, unless we can treat the natives’ behaviour as abiding by certain fundamental laws of rationality, we cannot translate it. As Davidson has indicated in his celebrated attack on the idea of alternative conceptual schemes, we cannot even have any grounds for describing it as reasoning, and may even withhold the term ‘language’ from it.31 This is in line with Wittgenstein’s idea that a practice which does not conform to the so-called ‘laws of logic’ simply does not qualify as what we call ‘inferring’, ‘reasoning’ or ‘speaking a language’ (RFM 80, 89–95, 336; LFM 201–2, 214). There is another side to the principle of charity which has no echo in Wittgenstein’s discussion of radical translation, although it is in line with other observations of his. We cannot even start to translate the natives’ utterances unless we can take for granted that they share with us basic perceptual capacities. We take for granted that they can survey the scene around them and are aware of what goes on within their perceptual range. And this is a precondition for ascribing to them shared needs and desires. We cannot recognize them, for example, as refusing unpleasant things unless we can assume that they know that they are confronted with a knife rather than a piece of fruit.
However, Quine and Davidson distort these insights by approaching translation exclusively from the principle of charity. Unlike Quine, who in this respect is closer to Wittgenstein,
Davidson applies the principle not just to necessary truths or self-evident empirical truths, but ‘across the board’,32 i.e. indiscriminately to all types of beliefs. This suggests that a precondition of translating the natives is that we can count them right not just on fundamental issues, where disagreement would be unintelligible, but on most matters. Understanding would depend on maximizing agreement in quantitive terms. Contrast Wittgenstein’s remark:
If language is to be a means of communication there must be agreement not only in definitions but also…in judgements. (PI §242)
Davidson rightly stresses the second point, but in the process mistakenly discards the first. By insisting that we need to maximize agreement in order to understand other people, he puts the cart (truth) before the horse (meaning). For, by and large, we must understand what people say in order to judge whether they are speaking the truth. Sharing a language is ‘not agreement in opinions but in form of life’ (PI §241; see also RFM 353). By the same token, understanding an alien language presupposes convergence not of beliefs, but of patterns of behaviour, which in turn presupposes a framework of shared cognitive capacities, needs, emotions and attitudes.
On that basis, however, there is room for genuine disagreement, for example about beliefs concerning the causes of physical phenomena, or about the acceptability (moral or aesthetic) of certain desires. Such disagreement will often include those beliefs that play a fundamental role in the respective ‘world-pictures’, and which Wittgenstein discussed in On Certainty, notably propositions concerning fundamental scientific questions. Finally, once we exclude the need to maximize agreement, there are no prima facie reasons in favour of Davidson’s claim that there could not be genuinely different conceptual schemes of the kind envisaged by Wittgenstein. But that complex issue must be left for another occasion.33
My conclusion is that Wittgenstein presents a more accurate picture of radical translation, and thereby of human understanding, than either Quine or Davidson. On the other hand, it is only before the background of their elaborate and forceful discussion that we can appreciate the relevance and value of his cursory remarks. The most important thing, however, is to confront their contributions with each other. For, as all three have shown, the safari of radical translation may yield important insights not just for philosophical anthropology or philosophy of language, but also for epistemology, notably the problem of relativism.34
1 A terminological point: in the literature ‘radical translation’ and ‘radical interpretation’ are sometimes used to denote respectively the Quinean and Davidsonian methods of translation. I do not follow this usage, but rather employ ‘radical translation’ to refer to translation or interpretation from scratch, in order to bring out the different approaches to this task adopted by our three protagonists.
2 ‘The Indispensability of Translation in Quine and Davidson’, Philosophical Quarterly, 43 (1993); see also M.Alvarez, ‘Radical Interpretation and Semantic Nihilism: Reply to Glock’, Philosophical Quarterly, 44 (1994) and my ‘A Radical Interpretation of Davidson: Reply to Alvarez’, Philosophical Quarterly, 45 (1995).
3 CV 37; ‘Some Developments in Wittgenstein’s View of Ethics’, Philosophical Review, 74 (1965), 25.
4 J.Katz, ‘The Refutation of Indeterminacy’, in R.Barrett and R.Gibson (eds), Perspectives on Quine (Blackwell, Oxford, 1990), pp. 182–3.
5 Dummett, Frege: Philosophy of Language (Duckworth, London, 1981), pp. 374ff.; Evans, Collected Papers (Oxford University Press, Oxford, 1985), ch. 2.
6 It is not clear whom he has in mind. Of course, Searle (‘Indeterminacy, Empiricism and the First Person’, Journal of Philosophy, 84 (1987)) has claimed that Quine’s rejection of first-person authority concerning meaning is an absurd consequence of his behaviourism, but this attack would leave intact his approach to understanding the words of others, including radical translation.
7 In the same vein, Blackburn (Spreading the Word (Oxford University Press, Oxford, 1984), ch. 8) holds that the task of a ‘bleak’ Quinean interpreter is hopeless, but without showing why Quine cannot reach the little he claims or regarding this as a reductio.
8 See, for example, C.Hookway, Quine (Polity Press, Cambridge, 1988), pp. 172–3.
9 This scenario is suggested by Chomskian accounts, e.g. R.Bartsch and T. Vennemann, Semantic Structures (Athenäum Verlag, Frankfurt, 1972), p. 3.
10 Hilary Putnam, Philosophical Papers, vol. II (Cambridge University Press, Cambridge, 1975), pp. 257–8.
11 Paul Wirz, Die Marind-anim von Hollandisch Neu-Guinea, vol. I (L. Fehlgruber & Co., Hamburg, 1922), pp. 31–6.
12 See A.H.Wood, History and Geography of Tonga (Wilson & Horton, Auckland, 1938), p. 24. Quine himself (RR 44) mentions a mistranslation based not on a misunderstanding by the native, but by the radical translator.
13 According to Wittgenstein, this is a mistake which often underlies philosophical problems. See my ‘Philosophical Investigations §128: Theses in Philosophy and Undogmatic Procedure’, in Wittgenstein’s ‘Philosophical Investigations’: Text and Context, ed. R.L.Arrington and H.Glock (Roudedge, London, 1991).
14 See Donald Davidson, Inquiries into Truth and Interpretation (Oxford University Press, Oxford, 1984), chs 9–10, 16, 18 (subsequently abbreviated as ITI); Hookway, op. cit., pp. 167ff.; Evnine, Donald Davidson (Polity Press, Cambridge, 1991) ch. 6.2.
15 See my ‘A Radical Interpretation of Davidson’, op. cit., p. 211.
16 Donald Davidson, ‘A Coherence Theory of Truth and Knowledge’, in Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, ed. E.LePore (Blackwell, Oxford, 1986), p. 313.
17 Donald Davidson, Essays on Actions and Events (Oxford University Press, Oxford, 1982), pp. 272–3 (subsequently abbreviated as EAE).
18 ‘Reply to Essays’, in Essays on Davidson: Actions and Events, ed. B. Vermazen and M.Hintikka (Oxford University Press, Oxford, 1985), p. 245; cf. D.Lewis, Philosophical Papers (Oxford University Press, Oxford, 1983), vol. I, p. 112.
19 ITI 126.
20 Ibid., 161.
21 Ibid., 142–3.
22 See respectively WPEb 104–28; RTC 93–5, 138, 207–8 and ITI 171, 280; ‘A Nice Derangement of Epitaphs’, in Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, ed. E.LePore (Blackwell, Oxford, 1986), pp. 433–46.
23 See G.P.Baker and P.M.S.Hacker, Language, Sense and Nonsense (Blackwell, Oxford, 1984), p. 292.
24 J.A.Fodor, The Language of Thought (Crowell, New York, 1975). P.M.S.Hacker presents a powerful critique of this idea in Meaning and Mind (Blackwell, Oxford, 1990), ‘Thinking: The Soul of Language’.
25 ITI 230.
26 EAE 50–2.
27 I develop this argument against the Quinean assimilation of nonsense and falsehood at greater length in my ‘Wittgenstein vs. Quine on Logical Necessity’, in S.Teghrarian (ed.), Wittgenstein and Contemporary Philosophy (Thoemmes Press, Bristol, 1994), pp. 216–20.
28 Von Savigny, ‘Common Behaviour of Many a Kind’, in Arrington and Glock, op. cit., argues that this passage makes understanding dependent not on something we share with the interpretees, but only on their behaviour sharing certain regularities. Wittgenstein indeed insisted that such a consensus is a framework condition of rule following, and hence of speaking a language. But, as we shall see, he also insisted that we can learn an alien language only if we share a certain framework with its speakers; and the quoted passage is most naturally interpreted as expressing that thought.
29 For example N.Garver, ‘Naturalism and Transcendentality: The Case of Form of Life’, in Teghrarian (ed.), Wittgenstein and Contemporary Philosophy, op. cit.; A.Kenny, ‘Wittgenstein’s Meaning of Life’, Times Higher Education Supplement, 19 May 1989.
30 See, for example, C.E.Izard, The Face of Emotion (Appleton-Century-Crofts, New York, 1969); R.van Beldoyen, Characteristics and Recognizability of Vocal Expressions of Emotions (Foris, Dordrecht 1984).
31 ITI, ch.13.
32 Ibid., xvii.
33 See my ‘Radical Translation and Conceptual Relativism’, The European Legacy I (1996).
34 Thanks are due to Bob Arrington, Peter Hacker, John Hyman, James Young and to audiences at Graz, Oxford and Reading for comments and suggestions.