this exploration of the ancestry of documents has been performed already at an earlier stage of the investigation, for we could have made little progress if we had not been able to recognise certain more or less defined groups of documents as habitually attesting analogous ancient readings, and thus as being comparatively faithful representatives of particular ancient texts. But we are now enabled both to verify with increased exactness the earlier classifications, and to extend them to other documents the texts of which were too ambiguous at first sight to allow them to be classified without the aid of standards external to themselves.
200. The evidence is supplied by the numerous variations in which each variant can at once be assigned with moral certainty to some one of the ancient texts, to the exclusion of those variations in which the grouping of documents is at this stage obscure. At each variation we observe which ancient text is attested by the document under examination. The sum of these observations contains the required result. Neglecting petty exceptions as probably due to some unnoticed ambiguity, unless they happen to be of special clearness, we find that the document habitually follows some one ancient text; or that it sometimes follows one, sometimes another, but has no characteristic readings of the rest; or again that it follows all in turn. Thus we learn that it has transmitted one ancient type of text in approximate purity; or that it is directly or indirectly derived by mixture from two originals of different defined types; or that it has arisen from a more comprehensive mixture. The mixture may of course have taken place in any proportions, and the same observations which bring to light the various elements will supply also a fair estimate of the
