Hostname: page-component-74656f9884-j7cmm Total loading time: 0 Render date: 2026-11-07T09:19:29.360Z Has data issue: false hasContentIssue false
  • English
  • Français

Can Theoretical Underdetermination support the Indeterminacy of Translation? Revisiting Quine's ‘Real Ground’

Published online by Cambridge University Press:  05 February 2010

Sophie R. Allen
Sophie R. Allen
Affiliation:
St Peter's College, Oxford
Rights & Permissions [Opens in a new window]

Abstract

It is commonly believed that Quine's principal argument for the Indeterminacy of Translation requires an untenably strong account of the underdetermination of theories by evidence, namely that that two theories may be compatible with all possible evidence for them and yet incompatible with each other. In this article, I argue that Quine's conclusion that translation is indeterminate can be based upon the weaker, uncontroversial conception of theoretical underdetermination, in conjunction with a weak reading of the ‘Gavagai’ argument which establishes the underdetermination of the sense and reference of subsentential terms. If underdetermination is considered to be a widespread phenomenon in science, or in inductive reasoning more generally, then the Indeterminacy of Translation will be widespread too. Finally, I briefly consider two issues concerning the scope of this conclusion about the Indeterminacy of Translation: first, whether the argument presupposes behaviourism; and second, whether indeterminacy is restricted to the case of radical translation. I argue that the answer to both these questions is negative, and thus that the thesis of semantic indeterminacy remains relevant to those who disagree with Quine about some issues concerning the nature of mind and language.

Information

Type
Research Article
Information
Philosophy , Volume 85 , Issue 1 , January 2010 , pp. 67 - 90
Copyright
Copyright © The Royal Institute of Philosophy 2010

W. V. Quine used two strategies to support the indeterminacy of translation: the ‘argument from below’, a reflection on the methodology employed in radical translation, including the gavagai example, which doubles as an argument for the inscrutability of terms; and the ‘argument from above’, based upon Pierre Duhem's thesis that theories are underdetermined by the evidence for them. Although it was predominantly with the former strategy that Quine introduced the indeterminacy thesis,Footnote 1 in later work he maintained that ‘the real ground of the doctrine is very different, broader and deeper’,Footnote 2 and from then on he concentrated on bolstering the argument from above in reaction to continued criticism. In the course of this paper, I will investigate whether there is a plausible formulation of theoretical underdetermination that can be used to support translational indeterminacy, since it is commonly claimed that this cannot be done.Footnote 3

Much has been written on the indeterminacy of translation – some would say too much – and it is with some hesitation that I enter the fray.Footnote 4 However, I think that there is one way in which the argument from above can be invoked in support of translational indeterminacy which appears to have been neglected by commentators, although it is sometimes alluded to by Quine himself. In what follows, I will not attempt to give an exegesis of Quine, nor to defend the consistency of his philosophical system as a whole – a full consideration of his views goes well beyond the scope of this paper – and, although I will mark points of apparent agreement between myself and Quine along the way, some of the avenues which I explore will be distinctly unQuinean in their concerns.

In the first section, I will recount the way in which the phenomenon of theoretical underdetermination was initially taken (by Quine and others) to support the indeterminacy of radical translation, and how Quine strengthened the underdetermination thesis in order to avoid objections to his initial formulation. In the second section I will briefly consider why strengthening the underdetermination thesis at best reduces the plausibility of the argument for indeterminacy, and at worst creates insurmountable problems for it. Third, I will argue that there is a way in which the argument for the indeterminacy of translation can be understood which does not require the stronger version of the underdetermination thesis. In the final section, I will preempt two objections and explore the extent to which this version of the argument should be acceptable to those who do not subscribe to Quine's other views.

1. Radical translation

The first exposition

Briefly (and uncritically) sketched, Quine maintains that two (or more) manuals for translating from one language into another may be compatible with all dispositions to verbal behaviour and yet incompatible with each other.Footnote 5 This entails, as Quine explains it, that ‘in countless places, they [the translation manuals] will diverge in giving, as their respective translations of a sentence of one language, sentences of the other language which stand to each other in no plausible form of equivalence however loose’.Footnote 6 Thus, given two divergent translation manuals T1 and T2 to translate between languages LN and LH, a sentence S of LN will be translated such that:

T1: ‘S’ in LN means ‘P’ in LH

T2: ‘S’ in LN means ‘Q’ in LH

where P and Q are sentences of LH and ‘P’ does not mean ‘Q’.Footnote 7

The failure of sentence S to translate determinately is not a unique and isolated feature of S itself, so T1 and T2 cannot be used interchangeably.

To explicate how this situation comes about, Quine invokes the now well-known project of radical translation: it is the task of the linguist to construct a translation manual from the Native language to the Home language with neither the aid of interpreters, bilinguals, nor prior information relating the structure or syntax of Native gleaned from the study of related languages. The linguist begins by matching observation sentences of Native to those of her own language by equating stimulus situations; that is, by matching a sentence which the Native speaker is disposed to utter (or assent to) with one in the Home language to which the linguist is disposed to assent in the same circumstances.Footnote 8 ‘Gavagai!’, construed as an observation sentence, is identical in stimulus situation with the (in this case) English ‘Rabbit!’, ‘Lo, an undetached rabbit part!’, ‘Lo, a rabbit stage!’, ‘Lo, an instance of rabbithood!’ and so on, and hence the phrases are stimulus synonymous. Since all provoke assent and dissent in identical circumstances, the linguist cannot discern which of the English alternatives is the correct one – that is, what, intuitively speaking, the native speaker means – and so it does not matter which of the alternative translations the linguist chooses, they are holophrastically interchangeable.

From these translations of observation sentences, the linguist projects into the native's morphological associations and grammatical trends. She isolates and attributes functions to subsentential terms in order to construct tentative analytical hypotheses, incorporating Native analogues of identity and individuation, quantification, pronouns, plurals and the like, by which further non-observational sentences of Native can be dissected and translated. It is in this endeavour that arguments for the inscrutability of subsentential terms become important, including the Gavagai example,Footnote 9 and the examples of proxy functionsFootnote 10 and Japanese classifiersFootnote 11: there is a choice of one-one mappings of subsentential terms from Native to Home. The reference of ‘gavagai’, qua subsentential term, is inscrutable: on the basis of the available data, there is no fact of the matter about whether ‘gavagai’ denotes rabbits, undetached rabbit parts, rabbit stages, instances of rabbithood and so on. Given suitable adjustments to analytical hypotheses elsewhere in the manual, each mapping will make for an empirically adequate translation manual.

This construction of the translation manual is subject to a linguistic analogue of the Duhemian underdetermination of physical theories according to which two theories may be compatible with the same body of empirical observation whilst being incompatible with each other. This underdetermination arises because a hypothesis does not face empirical test standing alone but against a background of theoretical assumptions; the confirmation of theories (construed as sets of sentences) is holistic. Thus, when an observation conflicts with a particular hypothesis in a theory, there is a choice between resolving the conflict by revising the hypothesis itself, or by making adjustments elsewhere within the theory (which include treating the recalcitrant observation as error). Despite this, theoretical underdetermination is not generally regarded as a troublesome aspect of scientific methodology: there will usually be a possible, future observation by which to distinguish which of the two (or more) competing, underdetermined theories is correct, even though there will then be another rival theory compatible with the new set of observations and incompatible with the theory just adopted. In the case of language, the linguist must choose which part of the translation manual to revise in the face of a recalcitrant observation, and so there is no reason to expect that two linguists working independently on the same language would construct manuals acceptable to both.

In practice, of course, the construction of a translation manual would be subject to the application of the usual theoretical virtues, such as simplicity, elegance and parsimony, according to which scientific theories are adjusted and revised.Footnote 12 But the application of such constraints could only dispel the problem of theoretical underdetermination if they could be epistemically grounded – that is, if they could be shown to have some role in determining the truth of one theory rather than another – and thereby be accorded some factual basis, rather than being purely aesthetic or pragmatic virtues. But in the case of natural languages, perhaps even more so than in the case of theories in the natural sciences, it is not obvious that the simpler formulation of a language is the correct one, nor even that there is an inter-linguistic notion of simplicity to be applied.Footnote 13 Thus, more than one translation manual will fit the behaviour (verbal and physical) of the Native speakers and so there will be a choice about which one to adopt. In the absence of any semantic facts by which to choose between competing translation manuals and from which it might be possible to extrapolate constraints on theory choice, such as what counts as linguistic simplicity, radical translation is indeterminate.

The second exposition

The above brief account of how the indeterminacy of translation arises, based in part upon the underdetermination of theories by data, accords with Quine's earlier explication of his thesis. However, in reply to an objection raised by Noam Chomsky, Quine revised his view. Chomsky's objection is that Quine has merely shown the ‘indeterminacy’ involved in translation to be a special case of the more general underdetermination of scientific theories, and as such it is neither surprising, nor cause for concern,Footnote 14 and he accuses Quine of drawing an unwarranted distinction between the normal inductive uncertainty of scientific enquiry and the indeterminacy involved in the construction of a translation manual. If there are ‘objective’ methods based upon physical facts by which we might make a rational decision about which physical theory to accept, then why should these not also be called upon to distinguish between linguistic theories developed on the basis of observable, behavioural evidence? Even if there are no such facts, further observation will permit choice between translation manuals as it does in physics: it seems odd, Chomsky notes, that in physics we are entitled to ‘work within the framework of a tentative theory’, while ‘it is impermissible to make general statements about language’.Footnote 15 If the epistemological analogy between physics and language does not break down, then the choice between competing theories can always be resolved by further observation, and so translation at its worst is no worse off than physics.

Quine's response is to appeal to a stronger version of theoretical underdetermination: theories can be compatible with all possible observation and yet incompatible with each other, such that they are empirically equivalent and yet logically distinct.Footnote 16 There is no fact to decide between rival theories; they will remain underdetermined even when all possible observations are taken into account. One example he provides of such empirically indistinguishable theories are the competing geometries of physical space suggested by Henri Poincaré: on one conception, rigid bodies move around infinite space without change in size; while, on the other, bodies in finite spherical space shrink uniformly as they move away from the centre of the sphere.Footnote 17 The latter theory admits of an entity – the centre of space – which has no counterpart in the former. In such cases, given two incompatible translations, A and B, of the Native physical theory, we may attribute either A or B to the speakers of Native, although we ourselves may have accepted theory A, even in cases where the choice of translation is guided by simplicity.Footnote 18 Moreover, interrogation of the Native scientists about their theory would occur in Native and so could be interpreted according to either translation manual, so this will not help to determine which of the theories the Native speakers hold. The significance of asking whether the Native speakers really believe theory A or theory B is thrown into doubt.

In addition to strengthening the underdetermination thesis, Quine makes the further claim that the degree to which translation is afflicted by indeterminacy depends upon the amount of ‘empirical slack’ acknowledged in physics, the degree to which we take our physical theories to be underdetermined by observation. If this phenomenon is only recognised in the higher reaches of theoretical physics, then indeterminacy of translation will only occur in this area of discourse. On the other hand, if one maintains with Quine that underdetermination also affects discourse about the ‘ordinary traits of ordinary bodies’, then the scope of the indeterminacy of translation is broader, affecting talk about ordinary, observable, middle-sized objects too.

2. Problems with strong theoretical underdetermination

While Quine's original exposition of the indeterminacy of translation left it unclear why translation encounters any more difficulties due to theoretical underdetermination than theories in the physical sciences do, reliance upon the stronger version of theoretical underdetermination proves to be a very problematic route for the defence of the indeterminacy of translation to take. First, strong theoretical underdetermination creates difficulties for Quine's criteria of theoretical identity, thereby endangering his own maxim of ‘no entity without identity’, since it is not clear upon what basis he is warranted in asserting that two theories which are compatible with all possible observations are really two. This, in turn, threatens the consistency of two of Quine's principle theses: the indeterminacy of translation and his commitment to the non-relativistic truth of our best scientific theory. Many authors have taken this as an indication that his system is on the point of collapse and, finding the indeterminacy of translation the least congenial of Quine's theses – and perhaps also considering it to be the most poorly supported – this is usually the first in line for rejection.Footnote 19

In the absence of any convincing empirical examples of strong theoretical underdetermination,Footnote 20 Quine made recurring attempts in his later work to overcome the difficulty of providing an account of theoretical identity which is compatible with the thesis of strong theoretical underdetermination: When are two empirically equivalent theoriesFootnote 21 which are alike in considerations of ‘simplicity and naturalness’, significantly distinct, rather than merely lexically or phonetically different? Quine vacillates in his answers to this question, but by the time of writing Pursuit of Truth he claimed to have reduced the options to two alternatives: ‘the sectarian line' which is to reject ‘all the contexts of the rival theory's alien terms … and plead that we have no access to truth than our evolving theory, however fallible’; or, ‘the ecumenical line’, suggested by Davidson, where ‘both theories count as separately true, the truth predicate now being understood as disquotation in an inclusive and theory-neutral language in which both theories are couched’.Footnote 22 Elsewhere he comments ‘I no longer consider the question of identity of theories to be worth pausing over. We can just speak of drastically unlike theory formulations, irreducible one to the other by any known system of sentence-to-sentence reinterpretation, and let the notion of theory go.’Footnote 23 The problem with this latter formulation is that is not clear whether it would be possible for the theories in question to be strictly logically incompatible, or whether we would be able to gerrymander some translation manual between the two theories by introducing terms from one into the other – effectively combining the two theories into one overarching one – to remove the seeming incompatibility.Footnote 24 Eventually, of course, such a strategy would yield an unwieldy theory which, although empirically acceptable, would probably fail to meet our usual constraints of ‘simplicity and naturalness’, but it seems unlikely that this failure would aid the case for translational indeterminacy.

Quine's difficulty with individuating strongly underdetermined theories threatens his ‘argument from above’ for the indeterminacy of translation. For example, Robert Kirk plays upon this absence of a criterion of significant theoretical difference to challenge Quine over the consistency of strong theoretical underdetermination with his toleration of differing levels of indeterminacy according to the amount of empirical slack acknowledged between observation and theory.Footnote 25 Kirk imagines that a theory M, held universally among Martians, is to be translated. If one allows, with Quine, that at least some areas of Martian discourse are not susceptible to theoretical underdetermination, then all non-theoretical sentences are translatable with no more than the weak underdetermination of the kind acceptable in science. Then, given a textbook expounding M from scratch, we could translate all those sentences of M in which theoretical terms do not appear or are merely mentioned, as in verbal definitions. We can also translate the non-theoretical elements of sentences in which theoretical expressions are used. This partial translation does not give us a grasp of M, since we still do not understand the sentences in which theoretical terms are used, but our predicament is no worse than that of the Martian students of M, for they too must grasp the theoretical sentences by studying how theoretical expressions are embedded in less theoretical ones, noting the nomological interconnections within the theory and how theoretical expressions figure in these laws. Moreover, it is conceivable that the Martian theory M is isomorphic with an English theory A, such that the untranslated Martian theoretical expressions could be mapped onto each other one-to-one in all contexts. Thus, word-to-word substitution into the partial translation would provide a theory acceptable to English speaking physicists as a textbook of A. The only difference between A and M would be a trivial difference in vocabulary.

Although this is not a general counterexample to Quine's indeterminacy thesis, as it relies upon the isomorphism of the physical theories M and A, Kirk emphasises that it cannot be treated as a special exception since the isomorphism of the theories is logically independent of their being strongly underdetermined. If Quine's argument from strong theoretical under determination held, it should also apply in cases of isomorphism. Thus, Kirk's counterexample shows that, if any translation is determinate, this provides an Archimedian point from which to translate into the realm of theoretical discourse without indeterminacy setting in, so the inference from strong theoretical underdetermination to the indeterminacy of translation is invalid. Conversely, if translation is nowhere determinate, then Quine's concession that translational indeterminacy might not apply to all areas of language is clearly inadmissable, leaving his thesis of translational indeterminacy absurdly strong for all but the most dedicated semantic irrealists.

3. Radical translation revisited

The difficulties which strong underdetermination creates for individuating strongly underdetermined theories – and thereby for the consistency of the indeterminacy of translation with Quine's concession that the extent of indeterminacy varies according to the extent of theoretical underdetermination – appear to be insurmountable.Footnote 26 One of these three theses will have to go and, on the presumption that Quine requires strong underdetermination to support the indeterminacy of translation, it is frequently presumed that the indeterminacy of translation is the thesis that must be dropped.Footnote 27 I think, however, that the introduction of strong theoretical underdetermination as Quine's ‘real ground’ is something of a red herring: it is not required to support the indeterminacy of translation and therefore the attacks which focus upon it in order to refute the indeterminacy of translation are misdirected. In some later work, Quine also maintained this position:

‘In treating of the underdetermination of theories by evidence it is a poor idea to assume compatibility with all possible data, though in an earlier paper I put it that way, for it is both unrealistic and irrelevant.’Footnote 28

I shall now attempt to explicate the indeterminacy of translation once again in order to show why the weak theoretical underdetermination with which Quine began is all that is required to support the indeterminacy of translation.

To illustrate this point, it is best to return to the case of the radical translator. She equates stimulus situations, thereby matching observation sentences of Native with those of her own language, and begins to construct analytical hypotheses and a tentative account of the Native grammar which together constitute a translation manual. This tentative account of the syntax and semantics of Native must begin at a very fundamental stage. The translator must form hypotheses about whether the sentences of Native can be broken down into individual words at all, or whether the language is polysynthetic where sentences are ‘wordlike’, constructed from more fundamental morphological units which cannot stand alone. She must attempt to determine the relevance (if any) to semantics or syntax of the ordering of subsentential units, and of the tonality with which Native speech is pronounced; what types of constructions in Native function as verbs, adjectives, nouns, pronouns and so on; whether the language divides any of these according to gender, or some other ‘metaphysical’ hierarchy; and which cases, moods, tenses and aspects (if any) the Native language displays.

Both Quine's argument from below, and his arguments for the inscrutability of subsentential terms have a bearing on this project, but they do not need to be construed as independently adequate arguments for the indeterminacy of translation and inscrutability of reference respectively in order to have sufficient force to bolster Quine's case. For each observation sentence, and each subsentential term, there will be a choice of one-one mappings from Native to Home, so the translator provisionally accepts one of the options, revising her choice of mappings – or another part of the translation manual – if she encounters evidence which conflicts with her decision. Similarly, the translator's hypotheses about the grammatical and phonetic structure of Native will be open to revision.Footnote 29 Furthermore, the translator need not be a passive observer of the Native's speech behaviour, since she could set up situations about which to question the Native speaker in order to try to determine which of two (or more) alternative hypotheses is the more suitable one.Footnote 30 However, although this technique might rule out potential translations, it will not determine which translation is the correct one, since the translator will always be faced with different ways in which her theory might be revised. The evolving translation manual itself is a weakly underdetermined theory.Footnote 31

At each stage of evidence gathering, the translator can use her evolving translation manual to determinately translate Native sentences into those of the Home language with nothing more than ordinary inductive uncertainty. Even though an independent translator may construct an alternative manual – i.e. one that maps Native sentences onto different sentences of Home – the differences between the two manuals will not be inexplicable, and some possible observation may permit a principled decision between the two. However, this will not settle on a ‘correct’ translation manual, since there will always be another possible manual with which all the observations of the Native's speech behaviour made so far are compatible. This weak strain of underdetermination of the translation manual is not problematic, nor is its output indeterminate.

However, in areas of the Natives' discourse which are subject to theoretical underdetermination – physical theory (say), and perhaps some other areas of theoretical discourse as well – the status of translation alters. Now, the translator is dealing with not one, but two underdetermined theories – the translation manual and the physical theory – which are being simultaneously confirmed by the same set of possible observations. But the translator cannot juggle the simultaneous confirmation of both theories by the same evidence – this would be akin to solving a single equation containing two variables – so she must fix one theory as a parameter in order to proceed with the construction of the other. But, whichever theory is fixed (physics or the translation manual), it is not open to revision by observation while it is fixed, its development is artificially stalled at an arbitrary stage, and so even in the light of all possible observations, it will remain underdetermined. Thus, strong theoretical underdetermination arises as a consequence of the argument for the indeterminacy of translation which requires as a premise only the uncontentious claim that theories are weakly underdetermined; Quine was wrong to call strong theoretical underdetermination his real ground for the indeterminacy of translation.

One can, of course, presume that the translator will be able to hold one theory fixed (the translation, say) while amending the other (the physics); then alter the translation manual while assuming that the physics can be trusted. But this alternating procedure could not be expected to yield similar translation manuals, were it to be carried out by different linguists (or even by the same linguist at different times).Footnote 32 So, although one might not have satisfied the conditions for producing strong underdetermination, it would not be required to give rise to translational indeterminacy: there would be competing translation manuals which mapped identical Native utterances onto distinct utterances in the Home language; translation would be indeterminate without strong underdetermination.

The impossibility encapsulated in the equation-solving metaphor above is quite deliberate: one cannot solve a single equation containing two unknown variables without arbitrarily fixing one of the variables. Similarly, for the linguist, it is impossible to separate what the Native speaker means (the semantics of his language) from what he believes (his physical theory,Footnote 33 expressed in his language) on the basis of any possible observations, since the confirmation of one of these precludes the possibility of confirming the other.Footnote 34 Here we really are, as Quine puts it, ‘weighing up incommensurable values. How much grotesqueness may we allow to the Native's beliefs, for instance, in order to avoid how much grotesqueness in his grammar or semantics?’.Footnote 35 This impossibility is, I think, what Quine is alluding to when he says that ‘radical translation is a near miracle’.Footnote 36 The same problem does not occur when translating areas of discourse which are not underdetermined, however, since in this case, the linguist can in principle rectify her mistakes in constructing the translation manual on the basis of observation.

It is important to note here, since this point has often been confused, that the indeterminacy afflicts the conflicting output of the translation manuals, the mapping rules of which are themselves only underdetermined. As Quine notes, ‘a manual of [Native]-to-English translation constitutes a recursive, or inductive, definition of a translation relation’ but the ‘indeterminacy of translation consists rather in conflict in the outputs [of the manuals] themselves’,Footnote 37 once the manuals can no longer be compared on the grounds of observation. Translation is indeterminate where physics is underdetermined because it is translation of a physical theory which is already underdetermined from the Native scientist's perspective, into the Home physical theory which is also underdetermined, via a translation manual which is itself underdetermined, and there is no reason to expect that either theory, Native or Home, is underdetermined in the same manner. If we knew which theoretical virtues, or ‘super-empirical’ constraints were utilized by the Native physicists – that is, we knew what respective weights they placed upon simplicity, explanatory power, parsimony, naturalness and the like – and we knew how these are extrapolated from observation (assuming here that such considerations are not merely pragmatic), then we would have much more of a guide as to how their theory was constructed and how it was confirmed by the observation sentences which we can translate. That is, we could roughly work out the Native theory for ourselves. But asking the Native physicist which constraints he and his colleagues use would just give us more theory to translate, leaving the translator no better off, and in the absence of such knowledge, we cannot impose our own theoretical virtues in order to determine what we think the Native theory might be.Footnote 38 This is not to say that in practice we never apply our own standards – in the absence of anything better, there is little other choice – but this is an imposition which serves to cement the indeterminacy, rather than a technique which can help us decide which theory the Native physicist holds.

The scope of this indeterminacy is up for grabs in two dimensions. First, the extent to which the indeterminacy of translation extends from the lesson of radical translation towards translation of idiolects of the Home language depends upon features of language acquisition, in particular, how much similarity there is between an infant learning language and the radical translator's case.Footnote 39 Second, it depends upon the amount of empirical slack acknowledged in physics. It may occur only in the translation of theories concerning the farthest theoretical reaches of subatomic physics, cosmology, metaphysics and the like. But if, with Quine, one believes that underdetermination extends to talk of ‘ordinary traits of ordinary bodies’, then the translation of talk about these is indeterminate too. It is worth pointing out here though that when Quine made his comments about the variable amount of ‘empirical slack’ or theoretical underdetermination which might be acknowledged in physics, he was thinking in terms of the troublesome, strong form of theoretical underdetermination, such that theories may be underdetermined by all possible evidence for them and yet incompatible with each other. The scope of such strong theoretical underdetermination – if it occurs at all – is more restricted than the weak underdetermination upon which I have based my argument for the indeterminacy of translation, and one might think that in the latter case that we are stuck with underdetermination, and hence indeterminacy, in all, or virtually all, areas of discourse. This would be so, if the much-maligned distinction between theory and observation cannot be defended, since there would be no area of language free of theoretical, or theory-laden discourse. On the other hand, if such a distinction (however fuzzy) can be defended, then there will be room for disagreement about where exactly it lies; and underdetermination in physics, and with it indeterminacy of translation, will only afflict discourse on the theoretical side.Footnote 40

I have attempted to show that the argument for the indeterminacy of translation does not require the questionable thesis that some theories are subject to strong theoretical underdetermination. Indeterminacy arises because the mappings of meaning relations in the translation manual and the physical theory are orthogonal to each other, but both are confirmed by a common evidential source. The omnipresent underdetermination which afflicts them both creates troublesome indeterminacy only from the point that they must be considered jointly. After this stage, I conclude that Quine is right: insofar as correctness of translation goes, there is no fact of the matter about which translation is correct, nor can we overlook this phenomenon as we would weak theoretical underdetermination.Footnote 41

4. Some objections to this account

In the final section of this paper, I will briefly attempt to preempt two of the objections which might be made to the account I have given, in addition to attempting to assess its limitations.

Behaviourism

The first complaint is the oft-repeated reprise that Quine is a behaviourist and so his conclusions about language are irrelevant to anybody who does not share this account of the mind (or the lack of it). However, there are two points to note in response to this complaint. The first, as Alexander George has argued,Footnote 42 is that although Quine eschews talk of mental entities, his behaviourism is not the Skinnerian caricature which is usually ascribed to him, rather it is a product of his naturalism and also, one might add, his physicalism. It does not preclude the existence of ‘innate’ neurological ‘mechanisms of learning-readiness’,Footnote 43 including those which could account for language learning, and his postulation of such mechanisms minimizes the differences between his own view and that of Chomsky who considers that the innate universal grammar to be a sub-system of the mind/brain which is present prior to the acquisition of a particular language. Where Chomsky and Quine differ here is that Chomsky is happy to talk in terms of entities in linguistics which lack immediate, obvious physiological grounding in the hope that, like ‘valence’ in chemistry, more fundamental, broadly physical bases will be found for these entities on further empirical investigation, while Quine's physicalism leads him to reject such non-physical vocabulary whenever he can.

But the existence of innate mechanisms – a universal generative grammar, perhaps – does not remove the difficulties encountered in radical translation: innate generative grammar might be a necessary condition for language acquisition, but it is not a sufficient one. In constructing the translation manual, the radical translator must provide an account of enough of the grammatical structure of Native to produce lexical strings on appropriate occasions exhibiting what the Native speaker would count as correct surface grammar, the deep syntactic structure is no business of hers (unless, of course, she also happens to be a Chomskian linguist). Even if the radical translator had explicit knowledge of a formulation of generative grammar, were it possible to provide one, this would only aid her to the extent that it might place constraints on which rules and transformations might occur in Native.Footnote 44 The formulation of these constraints would be underdetermined, just like the rest of the translation manual, but even if they were not, they would not be sufficient to determine or predict the surface grammar of sentences of Native which the innate mechanism merely facilitates – the data and explananda of the radical translator – since that will also have been shaped by contingent features of the Native's environment and history.

Secondly, Quine's behaviourism, such as it can be called that, is not required by the argument for the indeterminacy of translation: if one postulates meanings as mental or neurophysiological entities, it is easy enough to arrive at the same result.Footnote 45 These internal meanings will have been determined by the initial state of the brain (or organism) and by its causal-environmental history, and so will be peculiar to each individual. If this is the case, Karl Schick argues:

‘Because there is no public way of telling what other people's meanings are, there is no public way of telling how other people put their meanings into words. Because of this, there can be no guarantee that one learns to put meanings into words as other people do. And there can be no guarantee that one's translations preserve meanings.’Footnote 46

Translation would remain indeterminate even when meanings are considered to be entities which are internal to the speaker, rather than sticking with Quine's dispositional account of language.

One can, of course, make ontological and epistemological presuppositions which will deal with the threat of translational indeterminacy. One could, for example, follow David Lewis and introduce epistemic constraints on the amount of underdetermination, or inductive uncertainty, there is, and thus minimize the extent to which translation manuals, or other scientific theories, could diverge.Footnote 47 Or, one could make an ontological claim about the existence of determinate meanings, existing either as abstract objects, mental entitiesFootnote 48 or in virtue of worldly (and also, perhaps, other-worldly) entities such as properties, classes or kinds.Footnote 49 Or, one could do both. But the resolution of scepticism about meaning by making presuppositions, whether these are ontological or epistemological, in effect solves the problem by begging the question against the Quinean view. So I will not pause to assess the justification and merits of such accounts of language here.Footnote 50

Underdetermination does not happen at home

The second objection I will consider to the account that I have given is that although this account may work in the case of radical translation, it cannot be extended to the case of homophonic translation; that is, it will not work in cases in which both speaker and translator speak the Home language. The reason for this is that my account relies on the translator's hypotheses about the semantic and syntactic structure of the speaker's language – her translation manual – being weakly underdetermined, but in the case of two English speakers (say), it might be objected that this could not occur. For instance, Kirk suggests of Quine's thesis that

‘There is no interesting indeterminacy or inscrutability at all unless it holds also for the domestic case, which is free from syntactic and semantic differences.’Footnote 51

Whether one thinks Kirk's statement is true or not depends upon what one thinks a language is, and what one thinks speakers can and need to know (or, be disposed to do) in order to communicate. The former point is an ontological one concerning what counts as a language and how languages are to be individuated. Roughly speaking, there are two strategies to do this: one can take the folk ontology at face value which suggests that languages, such as English, Gaelic, or Lakota, can each be treated as a unified entity of such a sort that it is possible to be right or wrong about its structure and semantics; or, one can individuate languages with reference to an individual speaker, as the idiolects of speakers, or as I-languages, as they are internally represented in the speaker's mind or brain. If one takes the first option – call it, following Chomsky, the E-language option – then there is a public body of syntactic and semantic rulesFootnote 52 for a speaker to acquire. Although different speakers may do this more or less adequately, it seems likely that fluent speakers will acquire the same rules when they learn the language,Footnote 53 in which case there are no semantic or syntactic differences in the domestic case, except harmless differences where one speaker possesses vocabulary that another lacks (but which he could, in principle, learn). One might, on this option, even define the language in terms of its grammatical rules, in which case, Kirk might be correct by definition: two English speakers cannot, in principle, differ in the grammatical rules they use if this shared grammar also precludes a version of the gavagai argument; that is, if it also ensures that the words of each speaker share publically communicable sense. In such cases, the argument I have given for the indeterminacy of translation will not extend to the domestic case, since homophonic ‘translation’ is free from underdetermination.

On the other hand, although the individuation of languages as idiolects or I-Languages does not entirely preclude the sharing of linguistic knowledge between speakers, it certainly makes it less likely. On acquiring language, a language learner's theory (her ‘translation manual’, to put it in Quinean terms) will be underdetermined by the evidence which she has for constructing it, which consists in other speakers using their idiolects in what they consider to be appropriate conditions. The mere sharing of an innate language faculty, such as a generative grammar, is not sufficient to rule out underdetermination of the speaker's syntax and semantics – as was discussed above – it is merely an aid to her acquiring whichever fuzzily-defined, public language(s) she is exposed to but is not, I presume, infallible in this respectFootnote 54 and not all of the semantic and syntactic differences between herself and other speakers will be noticed or corrected.Footnote 55 Thus, there is room for underdetermination and inductive uncertainty in the case of homophonic translation; pace Kirk, the domestic case is not free from semantic and syntactic differences and so the argument that I have given for the indeterminacy of translation applies in this case as well.

5. Conclusion

I have argued in this paper that Quine did not need to rely on strong theoretical underdetermination in order to defend the indeterminacy of translation, nor does his argument rely upon the acceptance of behaviourism. The indeterminacy of translation occurs when two weakly underdetermined theories – the translation manual and physical theory – are being confirmed simultaneously from a common evidential source. In order to deal with two underdetermined theories, a translator must hold one of them fixed, treating it as if it were not underdetermined in order to utilise the available evidence to develop and confirm the other theory. Most probably, she will have to assume that the translation manual which she has developed on the basis of the Native's verbal behaviour and observable reactions to the world is correct, thereby ceasing to treat her account of the Native's syntax and semantics as underdetermined and open to revision on the basis of further observations except insofar as she will attempt to map the novel theoretical terms of Native onto those in the Home language. Only if this technique results in the Native's beliefs seeming too ‘grotesque’ by her own lights might she alter the linguistic theory further, but here she is imposing her own theoretical beliefs and constraints upon the Native in order to revise her account of the syntactic or semantic structure of Native. The output of the translation manual will be indeterminate in the sense that a competing translation manual could always be constructed which gave an alternative account of what the Native's utterances mean.

References

1 Quine, W. V., Word and Object (New York: John Wiley and Sons, 1960), Chapter 2Google Scholar. All citations will refer to the 18th edition (1992).

2 ‘On the Reasons for Indeterminacy of Translation’, Journal of Philosophy 67 (1970), 178.

3 Most critics focus upon Quine's later, stronger version of the underdetermination thesis which he advanced to counter criticism of the original version. See Section 2, especially footnotes 18 and 24.

4 Given its subject matter, it is perhaps not surprising that Quine's account of translational indeterminacy, and his arguments for it, have been interpreted by different philosophers in a wide variety of different ways: on one hand, if Quine is correct about the indeterminacy of translation, this confusion is only to be expected; and, on the other, if he is not, a certain amount of confusion is inevitable anyway, given the vast literature his views generated and the ways in which Quine himself revised his arguments, both for the sake of defending his conclusion that translation is indeterminate and in order to defend the consistency of the indeterminacy of translation within the broader framework of the rest of his philosophical system.

5 Word and Object (Op. Cit., note 1), Chapter 2.

6 Ibid., 27.

7 Here, the notion of meaning is used fin an intuitive sense in order to argue against the commonly held view that sentences can have determinate meanings and the ‘ill-conceived notion within traditional semantics, namely, sameness of meaning’ (Quine, W. V., ‘Indeterminacy of Translation Again’, Journal of Philosophy 84 (1987), 10CrossRefGoogle Scholar).

8 There are two prima facie difficulties here. The first, raised by Jaakko Hintikka, is that assent and dissent cannot be determinately recognised in order to begin the radical translation procedure (‘Behavioural Criteria of Radical Translation’ in Words and Objections: essays on the works of W. V. Quine, Donald Davidson and Jaakko Hintikka (eds.) (Dordrecht: D. Reidel, 1969)). Since this worsens the plight of the radical translator, I will not consider this point further here and will accept Quine's claim that the translator must make conjectures about the Native's signs or words of assent and dissent, and that, ‘if he is wrong in guessing these signs, his further research will languish and he will try again’ (‘Indeterminacy of Translation Again’ (Op. Cit., note 7), 6). The second difficulty pertains to what is involved in equating stimulus situations; since these are defined relative to the stimulation of an individual's nervous system, a Quinean cannot mean this to be a case of literal identification and so some other account must be sought. This matter troubled Quine, but it would be something of a digression to investigate his attempts to solve the problem, since it arises out of his commitment to some form of behaviourism. (See Pursuit of Truth, (Cambridge, MA: Harvard University Press, 1990; revised edition 1992), 40–45.) In section 4, I will consider whether commitment to Quinean behaviourism is mandatory for those who support the indeterminacy of translation.

9 Word and Object (Op. Cit., note 1), 51–57.

10 Ontological Relativity (New York: Columbia University Press, 1969), 55–62.

11 Ibid., 34–38.

12 In practice, it would also be extremely unusual for interpreters to work independently, rather than their conferring, standardizing and thus, in effect, unifying their disparate manuals by fiat. This factor explains, in part, why translators and interpreters rarely encounter the phenomenon whch Quine highlights, even if it does exist. Moreover, conferences and research exchanges serve a similar standardizing purpose among scientists who work in areas in which new vocabulary is frequently required, so their experience is not generally one of semantic disagreement either. These points seem to be overlooked by Dorit Bar-On who presents the experiences of actual translators and interpreters as a criticism of Quine. (See ‘Indeterminacy of Translation – Theory and Practice’ Philosophy and Phenomenological Research 53 (1993), 781–810.)

13 See section 3 for further discussion of this point.

14 Chomsky, Noam, ‘Quine's Empirical Assumptions’, in Words and Objections, Davidson, Donald and Hintikka, Jaakko (eds.) (Dordrecht: D. Reidel, 1969)Google Scholar. I will discuss Chomsky's other main objection concerning Quine's behaviourism in section 4.

15 Ibid., 66.

16 ‘Reply to Chomsky’, in Words and Objections, Donald Davidson and Jaakko Hintikka (eds.) (Dordrecht: D. Reidel, 1969). This theme continues in ‘On the Reasons for Indeterminacy of Translation’ (Op. Cit., note 2).

17 Poincaré, Henri, Science and Hypothesis (London: Walter Scott Publishing, 1902), Chapter 4Google Scholar.

18 ‘On the Reasons for Indeterminacy of Translation’ (Op. Cit., note 2), 180.

19 Quine's strengthening the underdetermination thesis provoked a torrent of philosophical criticism, including Bar-On, Dorit, ‘Semantic Indeterminacy and Scientific Underdetermination’, Pacific Philosophical Quarterly 67 (1986), 245–263CrossRefGoogle Scholar; Boorse, Christopher, ‘The Origins of the Indeterminacy Thesis’, Journal of Philosophy 72 (1975), 362–387CrossRefGoogle Scholar; Hockney, Donald, ‘The Bifurcation of Scientific Theories and the Indeterminacy of Translation’, Philosophy of Science 42 (1974), 411–427CrossRefGoogle Scholar; Jaggar, Alison, ‘On One of the Reasons for the Indeterminacy of Translation’, Philosophy and Phenomenological Research 34 (1973), 257–265CrossRefGoogle Scholar. I remain unconvinced by some of these criticisms but, since I disagree with the need for Quine's stronger version of the underdetermination thesis, I will not pursue this issue here.

20 See, for example, Kirk, Robert, ‘Indeterminacy of Translation’, The Cambridge Companion to Quine, Gibson, Roger F. (ed.) (Cambridge: Cambridge University Press, 2004), 170–172Google Scholar.

21 Here the theories are to be construed as ‘global systems of the world’ accounting for all possible observations.

22 Pursuit of Truth (Cambridge, MA: Harvard University Press, 1990), 99–101.

23 ‘Three Indeterminacies’, in Perspectives on Quine, R. Barrett and Roger F. Gibson (eds.) (Oxford: Blackwell, 1990), 13.

24 See, for example, Lars Bergström, ‘Underdetermination of Physical Theory’, The Cambridge Companion to Quine (Op. Cit., note 20), 101–104.

25 Kirk, Robert, ‘Underdetermination of Theories and Indeterminacy of Translation’, Analysis 33 (1973), 195–201CrossRefGoogle Scholar; ‘More on Quine's Reasons for the Indeterminacy of Translation’, Analysis 37 (1977), 136–141.

26 Even Quine thinks that Kirk's counterexample works. See Kirk, ‘More on Quine's Reasons for the Indeterminacy of Translation’ (Op. Cit., note 25), 141, footnote 1.

27 Commentators on Kirk's counterexample usually take this approach. See, for example, Peter Smith, ‘Kirk on Quine's Real Ground’, Philosophical Studies 27 (1975), 427–431; Bradley, M. C., ‘Kirk on Indeterminacy of Translation’, Analysis 36 (1975), 18–22CrossRefGoogle Scholar. Kirk also does not appear to think that there are any options apart from denying the indeterminacy of translation.

28 ‘Comment on Bergström’, in Perspectives on Quine, Op. Cit., note 23, 53.

29 For example, the translator may decide that her account of which phonemes Native contains (that is, which sounds make a difference to meaning) is not broad enough, if a large number of Native words turn out to be homonyms. For instance, early attempts (by English speakers) to translate and write Inuktitut failed to distinguish between the ‘k’ sound and the glottal ‘q’, and the sounds ‘n’ and ‘ng’, resulting in an unusually large number of homonyms.

30 See, for example, Evans, Gareth, ‘Identity and Predication’, Journal of Philosophy 72 (1975), 343–363CrossRefGoogle Scholar.

31 If the ‘argument from below’ is construed in this way as resulting from weak theoretical underdetermination, then Quine was right to call the ‘argument from above’ as being his real ground for the indeterminacy of translation, since the former is just the occurrence of theoretical underdetermination in linguistics. However, as will become clear, I do not think that he required strong theoretical underdetermination to state his case as he could have relied on the overlap of weakly underdetermined theories in physics and linguistics to do so.

32 This difficulty mirrors one in the philosophy of science concerning whether enquirers holding two (or more) radically distinct scientific theories will ultimately ‘converge’ towards a single true theory. I have argued previously that I do not see a non-question-begging, naturalistic reason for this convergence. See Allen, Sophie R., ‘Deepening the Controversy over Metaphysical Realism’, Philosophy 77 (2002), especially 534–5CrossRefGoogle Scholar.

33 In practice, of course, one cannot assume that one's interlocutor is a physicalist, but this simplifying assumption is permissible since it will serve to make radical translation somewhat easier, since it will limit the range of Home terms onto which the Native theoretical terms can be mapped.

34 This interdependence of belief and meaning had enormous influence on the work of Donald Davidson. An alternative analogy to the equation-solving one might be the epistemic situation described by Heisenberg's uncertainty principle (or, ‘indeterminacy principle’) according to which the greater the precision with which the position of a fundamental particle is determined, the lesser the precision with which its momentum can be measured (and vice versa).

35 ‘Indeterminacy of Translation Again’ (Op. Cit., note 7), 8.

36 ‘Indeterminacy of Translation Again’ (Op. Cit., note 7), 9.

37 Op. Cit., note 22, 48–9.

38 Kirk argues that the indeterminacy thesis is trivial if it is merely based on the Native's dispositions to verbal behaviour, as Quine originally said. In order that it is not, he continues, we must also take the Native's higher-order dispositions to revise his theory in the light of conflicting evidence into account. But it is difficult to see how we are to find out about these higher-order dispositions, except with the help of the underdetermined translation manual that the translator accepts and so mistakes in the latter will also be carried over into our attempts to determine what these higher-order dispositions are. See Kirk, ‘Indeterminacy of Translation’ (Op. Cit., note 20), 156–159 and 170.

39 I will address a potential difficulty concerning indeterminacy of translation in the domestic case in the next section.

40 I have argued elsewhere that theoretical underdetermination is a widespread phenomenon and extends to talk of ordinary traits of ordinary bodies which would leave much of everyday language affected by indeterminacy. See Op. Cit., note 32 for arguments which imply this.

41 Even assuming that I am right in my interpretation of Quine, I can only guess at Quine's reasons for introducing strong theoretical underdetermination as the real ground for the indeterminacy of translation. It may arise from an initial failure to recognise the conflation of the indeterminacy thesis with the thesis of general theoretical underdetermination. This was perhaps motivated by the worry that there may be some substance to the charge of Chomsky and others that the indeterminacy of translation is merely a special case of the weak underdetermination of any theory by observation (see Op. Cit., note 14). ‘Translation’ is a crucially ambiguous term: it is both an activity (the concern of linguists) and the product or output of that activity (as in ‘That book is a translation of Carnap's Der Logische Syntax der Sprache’). Chomsky appears to conflate Quine's theses, concerning a thesis about the underdetermination of theories or translation manuals (weak theoretical underdetermination) with the resulting indeterminacy of the products of those translation manuals (the indeterminacy of translation). He misses the point that the former is the method by which we produce the latter and that the two theses are applicable over distinct domains—weak underdetermination applies to the mapping rules and analytical hypotheses of the translation manual while the indeterminacy of translation concerns its output—and thus takes himself to be vehemently defending the legitimacy of his own discipline as suitable subject of empirical enquiry. Quine's overzealous reply, with the introduction of strong theoretical underdetermination, unwittingly gives weight to the objection which it does not deserve. In later work, Quine is more careful to distinguish the ways in which ‘translation’ can be understood. See Pursuit of Truth (Op. Cit., note 22), 48–9, quoted in the previous paragraph but one. There are other diagnoses of what exactly the misunderstanding is between Quine and Chomsky. See, for example, George, Alexander (‘Whence and Whither the Debate Between Quine and Chomsky?’, Journal of Philosophy 83 (1986), 489–499)CrossRefGoogle Scholar who concentrates about where their disagreement would lie were Quine talking about the translator/linguist constructing a Chomskian universal generative grammar, rather than the more empiricist linguistic paradigm of Bloomfield and Boas which I take Quine to be interested in. I have no idea whether I am right, George is right, or neither of us are, but I do not take his account to be mutually incompatible with mine except regarding some trivial issues.

42 ‘Whence and Whither the Debate Between Quine and Chomsky?’, Journal of Philosophy 83 (1986), 489–499.

43 ‘Linguistics and Philosophy’ in The Ways of Paradox (Cambridge, MA: Harvard University Press, revised edition 1976), 57.

44 Presuming here that Native is a human language. Even if Chomsky is correct about innate grammar, I think it is plausible that there may be languages whose speakers do not possess this feature (although they may have another innate mechanism). Chomsky would deny these were ‘languages’ by definition, although I think we could be presented with persuasive evidence against his view; evidence of a complex and creative communication system, for instance, whose speakers clearly did not share any innate mechanisms with human beings.

45 See for example, Schick, Karl, ‘Indeterminacy of Translation’, Journal of Philosophy 69 (1972), 818–832CrossRefGoogle Scholar. Dagfinn Føllesdal has also sketched some ideas on this subject (in Perspectives on Quine, Op. Cit., note 23, 107–109) although I do not know whether he has pursued them elsewhere.

46 ‘Indeterminacy of Translation’, Op. Cit., note 45, 828.

47 ‘Radical Interpretation’, Synthese 27 (1974), 331–344; ‘New Work for a Theory of Universals’, Australasian Journal of Philosophy (1983), reprinted in Properties, D. H. Mellor and A. Oliver (eds.) (Oxford: Oxford University Press, 1997).

48 The former approach is typified by Fregean semantics, the latter by the work of Katz, Jerrold, including Semantic Theory, (New York: Harper and Row, 1972)Google Scholar; ‘Analyticity, necessity, and the epistemology of semantics’, Philosophy and Phenomenological Research 57 (1997), 1–28.

49 See, for instance, the semantic externalists such as Kripke, Saul (‘Naming and Necessity’, in Semantics of Natural Language, Davidson, D. and Harman, G. (eds.) (Dordrecht: D. Reidel, 1972), 253–355)CrossRefGoogle Scholar, Donnellan, Keith (‘Proper Names and Identifying Descriptions’ in Semantics of Natural Language, Davidson, D. and Harman, G. (eds.) (Dordrecht: Reidel, 1972), 356–79)CrossRefGoogle Scholar, and Putnam, Hilary (‘Is Semantics Possible?’ reprinted in his Mind, Language and Reality: Philosophical Papers, vol. 2. (Cambridge: Cambridge University Press, 1975), 139–52CrossRefGoogle Scholar; ‘The Meaning of “Meaning”’in Language, Mind, and Knowledge, Minnesota Studies in the Philosophy of Science VII, K. Gunderson (ed.) (Minneapolis: University of Minnesota Press, 1975), 131–193). Raatikainen, Panu has recently argued that the indeterminacy of translation can be avoided by taking the externalist route ‘On how to avoid the indeterminacy of translation’, The Southern Journal of Philosophy 43 (2005), 395–414CrossRefGoogle Scholar. However, Putnam (in conversation) has said that he meant his externalism to be independent of whether entities which determine meaning such as ‘natural’ kinds objectively exist; this might significantly limit the utility of his account of meaning in avoiding the indeterminacy of translation.

50 I have discussed the justification of some of Lewis's claims elsewhere. See Op. Cit., note 32.

51 Op. Cit., note 20, 168.

52 One might not require semantic rules in addition to syntactic ones, if one maintains that semantics is supervenient on syntax.

53 If a speaker's acquisition of the language falls short of his gaining knowledge of the same semantic and syntactic rules as other speakers, then indeterminacy of translation will affect this case too on an epistemological level: the speaker's translations, and hence his knowledge, of the meanings of another's language will be indeterminate, although ontologically speaking determinate meanings exist.

54 One might be able to get rid of this difficulty by presupposing that a very large amount of the conceptual structure of language is innate, thus that the meanings that each speaker learns are so constrained as to rule out underdetermination of their language acquisition. I am rather doubtful that this strong nativist assumption could be made plausible, especially since it would have to be strong enough to rule out underdetermination (thus precluding indeterminacy) while being weak enough to permit the explanation of the wide range of semantic and syntactic differences between languages and the numerous cases of misunderstanding and linguistic error which occur even in homophonic translation.

55 Even when a child's language use is corrected by another more fluent speaker, there is no guarantee that it will be corrected in favour of standard usage.