SL texts and items are more or less translatable rather than absolutely translatable or untranslatable.
--J. C. Catford (A Linguistic Theory of Translation, 1965; p.93)
J. C. Catford (1917-2009) is a British linguist and translation theorist of worldwide renown, reputed to be the founder of translation linguistics on a par with Eugene Nida of the United States and Georges Mounin of France. He was a professor of linguistics at the University of Edinburgh in the United Kingdom and the University of Michigan in the United States. J. C. Catford was hailed as “the world’s most influential translation linguist” by Nida, and he was an expert in systematically presenting translation theory in London School of Linguistics
The last chapter of Catford’s A Linguistic Theory of Translation, chapter fourteen, focuses on the differences between different languages and the difficulties that cultural differences bring to translation, namely the so-called “the limits of translatability”.
Firstly, Catford pointed out that the limits of translatability in total translation are, however, much more difficult to state. Indeed, translatability here appears, intuitively, to be a cline rather than a clear-cut dichotomy. SL texts and items are more or less translatable rather than absolutely translatable or untranslatable. In total translation, translation equivalence depends on the interchangeability of the SL and TL text in the same situation--ultimately, that is, on relationship of SL and TL texts to (at least some of) the same relevant features of situation-substance.
Secondly, translation fails--or untranslatability occurs--when it is impossible to build functionally relevant features of the situation into the contextual meaning of the TL text. Broadly speaking, the cases where this happens fall into two categories. Those where the difficulty is linguistic, and those where it is cultural.
In linguistic untranslatability the functionally relevant features include some which are in fact formal features of the language of the SL text. If the TL has no formally corresponding feature, the text, or the item, is (relatively) untranslatable. Linguistic untranslatability occurs typically in cases where an ambiguity peculiar to the SL text is a functionally relevant feature--e.g. in SL puns. Ambiguities arise from two main sources, (i) shared exponence of two or more SL grammatical or lexical items; (ii) polysemy of an SL item with no corresponding TL polysemy.
Cultural untranslatability means that, what appears to be a quite different problem arises, when a situational feature, functionally relevant for the SL text, is completely absent from the culture of which the TL is a part. This may lead to what we have called cultural untranslatability. This type of untranslatability is usually less “absolute” than linguistic untranslatability. In many cases, what renders “culturally untranslatable” items “untranslatable” is the fact that the use in the TL text of any approximate translation equivalent produces an unusual collocation in the TL. To talk of “cultural untranslatability” may be just another way of talking about collocational untranslatability: the impossibility of finding an equivalent collocation in the TL. And this would be a type of linguistic untranslatability.
Reference：
Catford, J. C., A Linguistic Theory of Translation, Oxford: Oxford University Press, 1965.