of the variations prima facie belonging to it must be cleared away. The causes of the irrelevance fall under two principal heads, inability to express Greek distinctions, and freedom of rendering. Where the variation lies between two approximately synonymous words, it is often impossible to say which it was that the author of a given version had before him. Such version cannot therefore be cited for either variant, and the necessary absence of a version from the side of the primary Greek MSS in an apparatus criticus leaves it undecided whether the Greek original of the version had or had not their reading. A similar uncertainty attends grammatical forms partially identical in meaning, such as the aorist and perfect of verbs; and also, though not in all cases, the presence or absence of the article. The ambiguity caused by freedom of rendering is sometimes not essentially different from the preceding cases, namely, where the genius of the translator's language would have rendered literal translation of one of the Greek readings unendurably stiff, or even impossible, and the most obvious rendering of it coincides with what would be a literal representation of the other Greek reading.
272. But, apart from this involuntary licence, most translators are liable to deviate from their original by slight verbal paraphrase in just the same way as transcribers of the fundamental text: in other words, many associations of versions with Greek evidence in support of changes of diction are due to accidental coincidence. Every paraphrastic impulse which affects a transcriber is not less likely to affect a translator, who has a strong additional temptation to indulge the impulse in the fact that he is creating a new set of words, not copying words set one after another before him. One of the commonest
