be further defined according to particular classes of readings.
258. Since in all cases tire inference depends on assumed homogeneousness of text, its basis may appear to be subject to uncertainty; for homogeneousness is interrupted by the intrusion of mixture, and it is theoretically possible that lost originals of groups might be mixed, as well as extant MSS. But the originals from which most groups which it is in practice worth while to keep in mind must have diverged can with difficulty be referred to so late a date as the times of general mixture, and no clear evidence of antecedent mixture has come to our own notice. The homogeneousness of the fundamental texts of all important groups may therefore, we believe, be safely trusted.
259. The limitation, more or less strict, to detached combination is necessary because otherwise the characteristics of the special common ancestor will be mixed up with the characteristics of a remoter and for present purposes less important ancestor. In all places where there is no variation D and the two associated versions are likewise found in combination, not the less truly because all other documents have the same reading; and this combination points with equal certainty to a single common ancestor: but here the single common ancestor was the apostolic autograph, followed perhaps by an indefinite number of immediate descendants; whereas what we want to know is the character of the special ancestor, as displayed either in departure from the original text or in fidelity shewn to it where others have departed from it. Similarly, where we find D and its associates agreeing with, for instance, אBCL and the Memphitic against all other documents, if we have ascer-
