'The idea that language use is governed by clear cut conventions is a philosopher's fantasy.' Discuss.
In his Word and Object, Quine outlines an account for 'Radical Interpretation' - that is the learning of a language previously unknown, without the usage of interpreters or any form, simply through being immersed in the language. He begins by relating how, through a series of test-cases, one might come to align the utterance 'Gavagai!' in the new language with the utterance 'Rabbit!' in English. First, we might see, in the presence of what we would call a rabbit, the native exclaim 'Gavagai!', and form an initial hypothesis that he is alerting us (or some other Gavagai-speakers) to the presence of the rabbit. We might then attempt to place ourselves in a different situation with a rabbit and query 'Gavagai?', and see whether the Gavagai-speaker is willing to assent to our statement. To make sure that he is not simply noting the presence of a white thing, or an animal, we would also query 'Gavagai?' in the presence of a non-rabbit white thing, and a non-rabbit animal respectively, and see whether or not he was still willing to assent, even in the absence of the rabbit. If the Gavagai-speaker were willing to assent to 'Gavagai?' in the presence of a rabbit, and only in the presence of the rabbit (with some degree of wariness to prevent, for example, shocked silence convincing us that he is unwilling to assent, when if fact he would not assent to anything at that particular moment). It should be noted that Quine is using 'Gavagai!' and 'Rabbit!' as utterances, not simply words, and that, with his theory phrased in terms of assertability conditions, there will be no sense in which these utterances mean things simply as words, since we could never tell, for example, if the term 'rabbit' refers to a particular rabbit, to rabbitness, to a rabbit extended throughout time, to a time-slice of a rabbit, or whatever.
Quine suggests that a language could be learnt in this way and that, whilst occurrences of people finding themselves in situations where they simply do not understand a word of a language, without an interpreter are now very rare (given that there is likely to be a chain of bilingual people from a speaker of any language on Earth to any other), he suggests that this should be a plausible account of how such language learning would work, were someone to find themselves in a situation necessitating radical interpretation. I would suggest that he need not be so quick to deny that such practices actually occur - in fact I suspect that the comparison of this account with how a small child learns their first language leads to much of the attractiveness of the account. Furthermore, I would suspect that it is a frequent experience that linguists are only able to get so far in in learning a new language from books or tapes, and need to be truly immersed in a foreign language and culture before achieving true fluency. Regardless, however, Quine is only proposing a possible account, and need only show that his account could work, not that it is the actual process that we use to learn languages.
Two immediate problems present themselves, which I feel should be addressed before considering the meat of the argument, and the interesting results. The first of the problems for the account, as I have stated it, is the difficulty of performing a process of querying, if we really are totally new to a language. Whilst a proficient language user is able to understand the use of intonation to indicate an interrogative mood rather than an indicative one, it seems likely that this is a substantial part of the language, and would not be fully comprehended by someone about to learn the language - it isn't clear that an English speaker will be immediately able, for example, to identify what sentences of Chinese are questions, and which are simply statements. I think the response to this is simply to argue that we need not expect the Gavagai-speaker to understand the nuances of English's indications of questions, but simply hope for a moderately intelligent audience, and one willing to engage with us. If someone is able to see that an English-speaker cannot understand their language, it requires only that they wish to begin communication with us for them to respond in some way to the utterances that we might make - in the same way that given a baby pointing and saying 'That cat!' we are able to respond 'Yes, that is a cat', regardless of whether they have mastered the art of phrasing things as questions. So long as the Gavagai-speaker realises that our utterances are intentional (rather than random blurtings), he will be able to begin this process.
The second problem, similarly, is how we might recognise assent on the part of the Gavagai-speaker, if we really do not understand any part of their language. Again, I suspect that this should not be a substantial problem. Granted, a non-speaker of a language will not know the words for 'Yes' or 'No', but this is exactly why Quine suggests we simply need to grasp assent. It seems likely that it will be possible to understand assent quickly - in all likelihood it will be as easy as prompting 'Gavagai?' to the speaker's 'Gavagai!', and seeing what their reaction will be. Whilst we must still be wary, in just the way we must be about any as-yet unconfirmed word, we may still begin to establish a picture of what form the utterance of assent takes.
What Quine seeks to show with his account of language learning is that 'manuals for translating one language into another can be set up in divergent ways, all compatible with the totality of speech dispositions, yet incompatible with one another'. He argues that, since our learning experience is by necessity finite, there might be any number of ways in which we might extrapolate our experiential instances in order to create rules of translation, none of which would be the same, even though they all encompass the set of sentences we experience. The reason he creates his account in (detailed) terms of stimuli and (linguistic) responses, is that he takes an empirical line, whereby to say that someone means, for example, 'Rabbit!' by 'Gavagai!' is to say that they would be willing to assent to 'Gavagai?' if and only if they are in the presence of a rabbit, and nothing more. Since this is all he claims that a word can mean, there will be many ways to construe a particular utterance - we will not determine whether he will assent to 'Gavagai?' in the presence of a hare by any number of experiences involving rabbits, since we will not have determined whether he makes (as we do in English) a distinction between these two kinds.
Lewis rephrases Quine's discussion in quite an interesting way. He suggests that we imagine Karl, who does not speak our language, and that the debate is about how, if we were able to know everything about Karl's physical state (and history, and position in the universe), P, we might come to know AO and AK the set of beliefs that Karl holds in our language and his respectively. Lewis outlines a number of principles that he beliefs will allow us to begin to know parts of AK . The first such principle is the Principle of Charity, which basically asserts that we should attribute to Karl the same beliefs and desires as we have. He has no greater support for this claim than that it would be strange to want to attribute any other particular set of beliefs and desires that we might come up with. So, according to Lewis' account, we might be wise to assume that the utterance given in response to pointing a sharp stick at Karl would mean 'No, don't!', or some such. The next principle is that of Truthfulness - he argues that we must assume that someone largely wishes to express things they really do believe with their utterances, so that they only shout 'Lion!' or 'Ionlay!' if they actually believe there is a lion, and also only if they think others present believe the same. He outlines six constraining principles in total, and it is his thesis that these principles are able to undermine Quine's assertion of the indeterminacy of meaning. Lewis claims that, since there will never be two perfect 'solutions' - translation schema - which satisfy all his constraints, and if there were, it would simply show that were other constraining conditions that we had not yet discovered.
Chomsky claims that a language formed through the practices of Quine's account would have to be very small, since there would be very few sentences that have been subjected to adequate enough conditioning as to ingrain them firmly into our usage. I am not convinced by this objection, since it seems to rely on a far stronger level of conditioning than that for which Quine's account argues. Quine himself suggests that the necessary experience of a word is very limited before it is adopted:
'The linguist hears 'Gavagai' once, in a situation where a rabbit seems to be the object of concern. He will then try 'Gavagai' for assent or dissent in a couple of situations designed perhaps to eliminate 'White' and 'Animal' as alternative translations, and will forthwith settle upon 'Rabbit' as translation without further experiment - though always in readiness to discover through some unsought experience that a revision is in order.'
The inexperienced Gavagai-speaker will adopt words heard only very infrequently - perhaps only once - since this is enough (once we are set into a practice of adopting words associated with present stimuli), so long as we are willing not to be dogmatic, and to realise that there may well be instances where we later discover that a word is used differently from how we had previously thought.
Chomsky's suggestion rests on his claim that we cannot learn language in the way that Quine suggests since we obviously use sentence-schema for forming new (and unprecedented) sentences, rather than being able only to use sentences that we have heard expressed previously. I would suggest, however, that Quine's account fits quite happily with such pattern-based rather than sentence-based accounts of language, since in terms of dispositions to respond to stimuli we are quite reasonable in chopping up sentences and considering what effect varying different parts will have. We can quite happily conceive of a situation where we learn that the response to 'Rabbit!' will be different in the situation where there was a prior stimuli in the form of an utterance 'I shall kill that...', and we might even consider what difference it would make to the response if the word 'Baby!' were to appear after the same stimulus. It is only if we do take a holistic, complete-sentences-only view of language that we are unable to examine how parts if a sentence count as different stimuli, tempering or changing the meaning of what follows.
There seems to me, however, to be a fairly serious problem for Quine's account, which is hinted at by both Lewis' and Chomsky's objections. Quine's account of the indeterminacy of meaning seems to rely on a very strong prior notion of what meaning might be. A potential problem for Lewis' account of how might might establish a single translation scheme, is that, whilst his constraining principles do seem likely to provide only a single 'perfect' translation manual, it might seem that there is no reason to believe that this should be the translation manual, in the sense that it is correct. He has created a system by which to choose a single scheme if we are presented with many, but what reason do we have to assert that it actually does capture the meaning of the language being translated? The most sensible response for the defender of Lewis' account would be to argue that this is simply all the meaning consists in - that we can't form a 'correct' translation scheme according to his account, and then claim that it doesn't translate correctly. For Quine's account, likewise, it would seem sensible to assert that all that meaning is is the dispositions to respond to given stimuli. It doesn't make sense, therefore, to claim that meaning is indeterminate, since that would require a further, stronger, notion of what meaning should consist in. Clearly if Quine is going for an empirical account of meaning, he should deny this - and argue that there is nothing more to translation but that it should settle any dispute about what linguistic reaction might be made in a given situation. If he does not go down this route, it seems to me that whilst he is claiming that translation is simply to be viewed in terms of dispositions to respond to stimuli, he can't claim that translation has anything to do with the meaning of the utterances being translated.