Simple and Complex Translation Schemes and Quine’s Argument for the Indeterminacy of Translation

Ken writes (a philosophy related post from my other blog):

Might a translator have good reason to favour a complex and context-dependent translation scheme over a simple and context-independent one?

For example, would a translator, or more particularly someone constructing a theory of meaning for gavagai-language, have a reason to prefer a scheme that gave the meaning of ‘etihw’ as follows:

x satisfies ‘etihw’ if and only if either ‘etihw’ occurs concatenated with ‘gavagai’ and x is a part of a white animal, or ‘etihw’ does not occur together with ‘gavagai’ and x is white

This is context-dependent because what ‘etihw’ means (its satisfaction conditions),and how it is to be translated into English, depend on what other words it occurs with in a sentence. A context-independent scheme might instead say:

x satisfies ‘etihw’ if and only if x is white.

The latter is prima facie much simpler than the former. But who cares? Is there any reason to prefer simplicity in the construction of theories of meaning? Is simplicity an indication of truth or just a pragmatic preference of the theorist who constructs the theory?

Chris Hookway, in his book on Quine (1988, Polity) and Crispin Wright, in his paper on Quine in the Blackwell Companion to the Philosophy of Language (1997), each give different reasons why a simpler theory is to be preferred.

Both think the grounds for rejection of complex satisfaction conditions lie in the connection of semantic theory with psychology. Semantic theory should not be unrelated, should possibly parallel, the psychological theory that explains the speaker’s linguistic competence. So the axioms should be things we can plausibly take to be realised in the speaker’s psychology. Both critics think the complex axioms fail this test. According to Hookway (1988:157) if we ascribe the complex axiom we must see the speaker as having two basic semantic dispositions, one to apply ‘etihw’ in one sort of context and one in another. But since it seems one could learn the word without two sorts of training the ascription doesn’t fit neatly with psychological theory. According to Wright (1997:412), the proposal breaks the methodological maxim not to ascribe to someone concepts that are inexpressible in their language to explain their competence in their language (except where necessary for other reasons). For example, no theory of English should require that English speakers possess the concept of Schadenfreude to be competent at English. And the meaning of ‘gavagai’ in gavagai language should not be undetached rabbit part since undetached rabbit part presupposes rabbit (in the sense that the concept is used not mentioned in the statement of satisfaction conditions of ‘gavagai’), but rabbit is inexpressible in gavagai-language if ‘gavagai’ means anything else).

These are both sensible points but they cannot possibly be right.

There is no legitimate inference from the complexity or otherwise of a statement of satisfaction conditions to the psychology of speakers of the language.

It is very simple to show that complexity or otherwise of statements of satisfaction conditions is an artefact of the language used to give the statements. Where the metalanguage is an extension of the object language, the satisfaction conditions will be homolinguistic and maximally simple and context-independent for that reason. Where the metalanguage is a genuine alternative to the object language, the satisfaction conditions may require all sorts of complication to remain adequate to behavioural data. But it cannot be insisted that satisfaction conditions descriptions be given in one particular metalanguage or other. If there is no ‘official’ metalanguage, there is no language-independent measure of complexity.

But, you say, the complex and the simple satisfaction conditions above are both given in the same metalanguage. It is a question of making complexity comparisons on the same scale (English in this case). This is true, but it doesn’t affect the point. The point is that the complexity or otherwise, relative to English, of a semantic theory for another language says nothing about the psychology of those speakers (who do not speak English and whose competence in a different language is under investigation, and where other languages besides English, including a metalinguistic extension of the object language itself, are available to state the satisfaction conditions).

The fact that linguistic facts about one language may require a complex description in one language and a simple description in another, and an even more complex description is yet another, (etc.) shows that the complexity or otherwise of linguistic facts, specifically satisfaction conditions, is a language relative matter.

The point is a general one, so it is permissible to vary the example a little to make it clear. If the minimalists are to be believed, satisfaction conditions for English include axioms like

x satisfies ‘red’ if and only if x is red

However, when Irish Gaelic is a metalanguage for English we should prefer a context-dependent statement (and here I make allowances for the general reader by using only two Irish words. Imagine the whole thing is in Irish)

x satisfies ‘red’ if and only if either ‘red’ occurs concatenated with ‘hair’ and x is rua or ‘red’ occurs without ‘hair’ and x is dearg

.
The complex axiom is preferable to a simple axiom such as (for example)

x satisfies ‘red’ if and only if x is dearg

Because the complex one captures the data of English linguistic behaviour better than the simple one, such as the fact that ‘Niall has red hair’ is more or less stimulus-synonymous with ‘Tá gruaig rua ag Niall’ (i.e. English speakers assent to former exactly when Irish speakers assent to the latter) and likewise for the pair comprising ‘Niall is drinking red wine’ and ‘Tá Niall ag ól fíon dearg’.

This case is sufficient to show, first, that complexity need not be ad hoc, for it is necessary here (though admittedly not in the original gavagai example) to capture the observed linguistic regularities, and second, that complexity relative to one medium (Irish) is compatible with simplicity in another (English).

But, you say, surely if two concepts were so related that one presupposed the other no matter what language they were expressed in, as might be the case with undetached rabbit part and rabbit, then it would after all be an objective matter whether a scheme that attributed one was more complex than a scheme attributing the other. I agree. But the point doesn’t bear on the gavagai example. Suppose for the sake of argument that ‘gavagai’ refers to undetached rabbit parts and that in any language that has it the concept undetached rabbit parts presupposes the concept rabbit, it still doesn’t follow that the scheme that says ‘gavagai’ is satisfied by all and only undetached rabbit parts is objectionably complex, because the ideology of gavagai-language cannot be read off extensional facts about it. Just because the English concept for undetached rabbit parts, namely undetached rabbt parts presupposes the concept rabbit doesn’t mean the gavagai-language concept for that same extension presupposes a more basic concept analogous to rabbit. Consider this formally similar example: The concept male parent presupposes the concepts male and parent, but the concept father, with the same extension, does not. (Children certainly seem to learn the latter concept before the other two). So extensions don’t give good direction when it comes to thinking about presupposition relations among concepts. So we can’t tell anything about gavagai from looking at undetached rabbit part and rabbit and so we cannot rule out undetached rabbit parts as a possible extension of ‘gavagai’ on that basis.

The upshot of the foregoing is that we are entitled to say this: It does not count against a translation scheme that it produce all sorts of complexity in the statement of satisfaction conditions of other words in the language, even if this is not necessary to capture linguistic behaviour, because descriptive complexity is not psychological deep.

Advertisements

One thought on “Simple and Complex Translation Schemes and Quine’s Argument for the Indeterminacy of Translation

  1. Jacob

    You say that Both Hookway and Wright give different reasons for why the simpler theory is to be preferred, and that both think that the reason why is to be spelled out or found in relationship to the connections between semantic theory and psychological theories.

    You then Quote Wright for suggesting that “the proposal breaks the methodological maxim not to ascribe to someone concepts that are inexpressible in their own language to explain their competence in their language”.

    You also say that this is a sensible point.

    Well, not knowing much about the discussion I must confess, (ad. Wright) I do not see why this is a sensible point relative to the general point you are inquiring about, namely whether there are any sound and good reasons for choosing simple over complex theories when it comes to semantic theories. I can see the point when the issue is over rabbit and Gavagai, but surely there must be examples of complex theories that do not introduce (necessarily) concepts inexpressible in the subjects own language), thus I do not see how Wright’s emphasis of the maxim, works as a criterion generally speaking for preferring simplicity to complexity.

    As for Hookway I am sure he has some theory as to why we ‘must see’ a speaker as having two simple dispositions rather than one complex – well whether that is so or not depends on whether speakers possess complex or simple dispositions – but surely – you cannot just take for granted that they do not, in fact possess complex ones.

    If the argument for why we must see the speakers as having two simple disp’ is that ‘that is the most simple’ it would be question begging in this context.

    Though I agree with you – in your conclusions – I am wondering whether there is not room to put in the knife at an earlier point and question a) what are the benefits of simplicity in general – surely this appeal to simplicity is derived from considerations about theory preference in general – where the concept of simplicity is not cashed out in a way that really mimics what is going on here. and b) if a good answer to a) can be given, then we need further argument for why it is that we should strive for simplicity in theories of meaning – in so far as these do not just follow on from theories in science – so even though a general argument for why we should prefer theories that are more simple in science could be given – we still need a further argument for how those arguments apply in the present case.

    Good reasons for being sceptical about simplicity as a theory virtue are found in : Philosophical Quarterly : WHEN IS PARSIMONY A VIRTUE? (p 216-236)
    Michael Huemer
    Published Online: Apr 16 2008

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s