A Real Reason to Value Linguistic Diversity

May 7, 2017 – 10:17 pm

There’s been some discussion on other sites about whether we should regret the loss of small languages, and if so, why. Many of the usual points have been made about how change is not something that should be seen as bad just in itself, and we can’t expect to preserve language speakers in a cultural ghetto so that we can enjoy just knowing that their tiny language is being spoken while we engage the rest of the world in conversation in a more successful tongue. On the other hand people will claim that the lessening of diversity in the world is something to be regretted.

I tend to agree with all of the above, but I want to give a reason for preserving linguistic diversity that doesn’t get mentioned often enough. Languages are a very convenient way of looking at the way the mind works. And the larger the number of languages in the sample the less likely it is that characteristics of the mind will be extrapolated from merely contingent features of that sample. I am put in mind of an article I once read concerning the language of the Maricopa Indians of Arizona.

David Gil analysed the language [’Aristotle Goes to Arizona, and finds a language without “and”‘ in Zaefferer (ed.) (1991) Semantic Universals and Universal Semantics] and concluded that it lacks any coordinate construction even at its deepest level, and has, indeed, no syntactic category of coordinator. (I won’t bother describing this category here, it should be intuitively obvious what he’s talking about.) Approximate translation of English coordinated phrases is achieved in Maricopa by various means but these always involve a degree of subordination of one (English equivalent) phrase. In fact the basic difference is that Maricopa, like many other languages but to an extreme degree, tends to prefer subordination in its structures, whereas the Indo-European languages have a much flatter structure.

Some examples will show you what this means in practice. Here are just two ways the sentence ‘I saw John and Bill’ could be translated by a Maricopan.

[S: [NP: ][VP: [S: [NP: Johns][VP: [NP: Bill][V: udaavm (accompany)][V: ni’yuuk (see)]]]]

in which the closest thing to an ‘and’ is a verb ‘to accompany’, but the structure of the sentence is clearly not as we defined it to be for a coordinating construction because the elements are not directly dominated by the root node.

[S: [NP: ][VP: [S: [NP: Johns][NP: Bills][VP: mat teevm (be together)]][V: ni’yuuk (see)]]]

in which the closest thing to an ‘and’ is now a regular intransitive verb, but again the structure fails to be that of a coordinating construction because the conjunct elements are of category NP while the root node is of category S. In a coordinating construction recall that the category of the construction is the same as the category of its coordinated parts.

So what’s the importance of all this? Well if one has a theory – and it was popular at one stage – that LF is to be derived from S-Structure, then it strongly suggests that there is no coordinating structure in the LF of a Maricopa sentence; and similar considerations show that no analogue to LF with coordinators is derivable under the same constraints from any other accepted level of linguistic representation in the current model. Of course there are possible functions which could create coordinating structures from Maricopa S-structures, the fact that they would need to be used by any translator between Maricopa and English assures us of this, but that is not the same as claiming that those functions can be justified as transformations in the linguistic system itself. Moreover, if such transformations did apply the problem would then be to explain how monolingual Maricopa speakers came to acquire their rules of application, since there are no linguistic data available to that speaker which even hint at their existence.

And the importance of this is that if a language like Maricopa did not exist then it would have been unlikely that anyone would have proposed it as a possible language. In short, the store of languages is a set of experiments and data for our investigations into our language faculty, which is about the most essentially human of all our faculties. Fewer extant or recorded languages means less data about ourselves, means less access to the truth about ourselves.


Post a Comment