If we adopt a Sign / Sign model, it is still necessary to
distinguish between the rewriting of signs within the same system and their transcoding. This
distinction appears significant to us: the elementary interpretive processes for which we previously
offered a typology (1987) were relevant to various sememic rewritings (in different texts in the same
language or in different languages). But those processes cannot be directly transposed to higher levels
of description, to units such as the rhetorical-prosodic period or the chapter, for example. And this
Text / Text model is quite different from a transcoding model.1
In addition, as far as the issue of the text is concerned, any rewriting of any sort is a
recontextualization, and cannot be described without describing also the parameters of situation and
context switching processes. Thus, sign A in context 1 is rewritten as sign B in context 2.
Nevertheless, the context is all of the text: in its turn, text A in situation 1 is rewritten as text B in
situation 2. The contextual and situational variations obviously require that interpretation be situated
in cultural history.
A related hermeneutic standard is attached to each sign system or language, and this
hermeneutic standard is a defining one. This is why a text cannot be interpreted through computation,
quite simply because a natural language is not defined by the same hermeneutic standard as a
language is. One can certainly, conventionally and at the cost of tremendous looseness, encode linguistic
signs with symbols and grammatical rules with syntactic rules (in the logical acceptation of the term).
But on the one hand such coding is not interpretation, and on the other hand it cannot be interpreted
(in the logical acceptation of the term). Worse still, it does not make computation (in the logical
acceptation of the term) possible. Throughout the whole history of the formal semantics of "natural" languages, despite exhilarating and worthy endeavours,
nobody has ever managed to bring forward any such computation, notably because compositionality
is foreign to the hermeneutic standards of natural languages.
It follows that the interpretation of a computational process does not have anything in
common with the interpretation of a text.2 And for
example, the fact that an interpreter of Prolog can be written in Prolog does not mean that this
language is equipped with hermeneutic circularity. On the other hand, in the case of a text,
intralinguistic commentary remains a possibility, just like translation, to some extent.
More generally, the various sign systems have their own interpretive modes, and as meaning
is the product of interpretation, they also have different semantics. This last remark follows the way
opened up by Lessing, whose Laocoon estimated that painting and poetry could not be
translated into one another. This position has hardly been followed, because of its dire consequences:
it questions the existence of an amodal conceptual-semantic level, such as traditional rationalism
conceived it, and upon which cognitivism still capitalizes.
In short, even if a general category of interpretation made possible that we set up a globalizing semiotics
which still would have some descriptive value, it is important to come up with a typology of
standards, modes and interpretive regimes pertaining to each semiotic system: indeed, those determine
the modalities of perception, recognition and comprehension for the signs in the system. This
enterprise implies that we enunciate the specific constraints of the expression-substance (rendered
discrete or not, linear or not, for example), and the standards of interpretation attached to each
generic semiotic performance.
1. In addition to the fact that natural languages are not codes,
transcoding is not sufficient to define interpretation, or more precisely, it circonscribes the question
of interpretation to syntactic interpretation: for example, compilation is a form of
transcoding, but not a semantic interpretation. [RETURN]
2. The various acceptations of to interpret are
often mixed together: depending on theories it is a question of highlighting, of expanding, of
reviewing, of specifying or affixing value, of rewriting, of transcoding, of building representations,
or even of compiling (as in the compiler-model theories of cognition). [RETURN]