Recent studies have established that unlike Western thought, based on a two-valued logic (either True or False), Aymara thought is based on a three-valued logic, and is, therefore, capable of expressing modal subtleties which other languages can only capture through complex circumlocutions. Thus, to finish, there have been proposals to use Aymara to resolve all problems of computer translation. Unfortunately, owing to its algorithmic nature, the syntax of Aymara would greatly facilitate the translation of any other idiom into its own terms, but not the other way around. Thus, because of its perfection, Aymara can render every thought expressed in other mutually untranslatable languages, but the price of this is that once the perfect language has resolved these thoughts into its own terms, they cannot be translated back into our natural native idioms.4
One way out of this dilemma is to assume, as certain authors have recently done, that translation is a matter to be resolved entirely within the destination (or target) language, according to the context. This means that it is within the framework of the target language that all the semantic and syntactic problems posed by the source-text must be resolved. This is a solution that takes us outside the problem of perfect languages, or of a tertium comparationis, for it implies that we need to understand expressions formed according to the genius of the source language and to invent a ‘satisfying’ paraphrase according to the genius of the target language. Yet how are we to establish what the criteria of ‘satisfaction’ should be?
These were theoretical difficulties that Humboldt had already foreseen. If no word in a language exactly corresponds to a word in another one, translation would be impossible. At most, translation is an activity, in no way regulated, through which we are able to understand what our own language was unable to say.
Yet if translation implied no more than this it would be subject to a curious contradiction: the possibility of a relation between two languages, Alpha and Beta, only occurs when Alpha is closed in a full realisation of itself, assuming that it has understood Beta, of which nothing can be any longer said, for all that Beta had to say is by now said by Alpha.
Comparison
It has been repeatedly said that, in order to interpret our actual world (or the possible ones of which many of the texts we translate speak) we are moving within the framework of a semiotic system that society, history, education have organised for us. It has been said that we are defined by our language, and in this sense Barthes once said that language is radically fascist. If this is true, then any translation of a text coming from another language and another culture would be impossible. To repeat Quine, one cannot express in a jungle language a sentence such as neutrinos lack mass.
However the experiences we have analysed up to now have shown us that, to a certain extent, it is possible to translate. If different linguistic systems still look mutually incommensurable, they remain mutually comparable. Linguists and semioticians have been blackmailed for years by the idea that Eskimos have many different names to identify what we call snow, according to its physical state, but eventually it has become evident that they are not at all prisoners of their own language and understand perfectly well that when we say snow we speak of something similar to the different things they name individually. I pointed out in my first essay that Italians do not make any lexical distinction between grandson and nephew, and call both nipote (which at the beginning of an interaction may produce some misunderstanding), but they clearly have in their minds the difference between the son of their son or daughter, and the son of their brother or sister. A French person uses the same word, glace, to refer to both ice and ice-cream, but French people (even though they eat frogs and drive on the wrong side) do not put ice-cream cubes in their Scotch, nor even in their Pernod.
Thus, rather than a powerful metalanguage, we might elaborate a comparative tool, not itself a language, which might (if only approximately) be expressed in any language, and which might, furthermore, allow us to compare any two linguistic structures that seemed, in themselves, incommensurable. This instrument or procedure would be able to function in the same way and for the same reason that any natural language is able to translate its own terms into others by an interpretative principle : according to Peirce, any natural language can serve as a metalanguage for itself.
See for instance a table created by Nida5 that displays the semantic differences in a number of verbs of motion.
We can regard this table as an example of an attempt to illustrate, in English – as well as by other semiotic means – what a certain class of English terms mean. Naturally, the interpretative principle demands that the English speaker also interpret the meaning of limb, as indeed any other term appearing in the interpretation of the verbal expression. The eighteenth-century philosopher and linguist Degérando observed that in order to analyse even an apparently primitive term such as to walk there will be an infinite regression. It has been said that it is easier to define the word infarct than the verb to be. However, Nida’s table can be used by anybody who has assumed that terms such as limbs or contact are primitives.
Translation proceeds according to the same principle, and once again we are faced with negotiation processes. If one wished, for example, to translate Nida’s table from English into Italian, we would probably start by substituting for the English verbs, Italian terms that are practically synonymous: correre for to run, camminare for to walk, danzare for to dance, and strisciare for to crawl. As soon as we got the verb to hop, we would have to pause; there is no direct synonym in Italian for an activity that the Italian–English dictionary might define as ‘jumping on one leg only’. Nor is there an adequate Italian synonym for the verb to skip: Italian has various terms, such as saltellare, ballonzolare, and salterellare; these can approximately render to skip, but they can also translate as to frisk, to hop, or to trip, and thus do not uniquely specify the sort of alternate hop-shuffle-step movement specified by the English to skip.
Even though Italian lacks a term which adequately conveys the meaning of to skip, the rest of the terms in the table – limb, order of contact, number of limbs – are all definable, if not necessarily by Italian synonyms, at least by means of references to contexts and circumstances. Even in English, we have to conjecture that, in this table, the term contact must be understood as ‘contact with the surface the movement takes place upon’ rather than ‘contact with another limb’. To either define or translate, we thus do not need a fully fledged metalanguage at our disposition. We assume that all languages have some notion that corresponds to the term limb, because all humans have a similar anatomy. Furthermore, all cultures probably have ways of distinguishing hands from arms, palms from fingers, and, on fingers, the first joint from the second, and the second from the third; and this assumption would be no less true even in a culture such as Father Mersenne imagined, in which every individual pore, every convolute of a thumbprint had its own individual name. Thus, by starting from terms whose meanings are known and working to interpret by various means (perhaps even with gestures) terms whose meanings are not known, proceeding by successive adjustments, an English-speaker would be able to convey to an Italian-speaker what the phrase John hops is all about.
A parenthesis on translation and ontology
It is possible to compare the structure of two languages Alpha and Beta by referring two different terms A and B to the same human behaviour (i.e. jumping or crawling) or to a common parental relationship. Despite the fact that the French bois covers a semantic space different from the one covered by the Italian bosco, we are able to detect whether a French speaker is speaking of a collection of trees or of the material used to make a chair – and so on. Such remarks suggest two possible ontological solutions: (i) there are universal ways of segmenting the continuum of human experience even though our languages sometimes blur them; (ii) in the continuum of our experience there is (as I suggested in Kant and the Platypus) a ‘hard core of Being’, not something tangible and solid, as if it were a ‘kernel’ that, by biting into it, we might one day reveal, but rather some lines of resistance – by virtue of which, for instance, we can speak in every language of crawling, hopping or jumping human beings but not, and in no case, of flying humans.
If so, and irrespective of the solution, we should say that, even though no perfect language can univocally express these universal phenomena, it would be with them that languages should be confronted.
It is curious to remark that, while so many philosophical discussions have cast doubt on the very possibility of translation, since each language represents an incommensurable structure, it is precisely the empirical evidence of translation that challenges the philosophical assertions about the dependence of world views on language.
