Fourthly, you assess the referential and pragmatic accuracy of the translation by the tranlator's standards. If the translation is not a clear version of the original, you consider first whether the essential 'invariant' element of the text which consists usually (not always) of its facts or its ideas is adequately represented. However, if the purpose of the text is to sell something, to persuade, to prohibit, to express feeling through the facts and the ideas, to please or to instruct, then this purpose is the keystone of the in variance, which changes from text to text; and this is why any general theory of translation invariance is futile, and I am at least a little sceptical about making a rule of Tytler's 'the complete transcript of the ideas of the original work precedes style and manner of writing' or Nida's 'form is secondary to content' (though I accept that form in translation must be changed to accommodate meaning) given that the keystone of invariance may be expressed as much through words of quality (adjectives, concept-words, and degree) as through words of object and action.
After considering whether the translation is successful in its own terms, you evaluate it by your own standards of referential and pragmatic accuracy. You have to avoid criticising the translator for ignoring translation principles that were not established nor even imagined when he was translating. The main question here is the quality and extent of the semantic deficit in the translation, and whether it is inevitable or due to the translator's deficiencies. Further, you assess the translation also as a piece of writing, independently of its original: if this is an 'anonymous' non-individual text, informative or persuasive, you expect it to be written in a natural manner - neat, elegant and agreeable. If the text is personal and authori- tative, you have to assess how well the translator has captured the idiolect of the original, no matter whether it is cliched, natural or innovative.
