tained that this second group often stands in opposition to the first, we know that the reading must have existed in a common ancestor of the two special ancestors, and that therefore it can tell us nothing about the special characteristics of either.
260. The most delicate and difficult part of the use of groupings in criticism consists in judging how far a group loses its virtual identity by slight losses or slight accessions of constituent members. The least important losses and accessions from this point of view are evidently those which accompany fragmentariness of text, so that the change is not, for instance, from concurrence to opposition, but from concurrence to total absence, or vice versa: in such cases much depends on the number and variety of the remaining members. Others again, which look as if they ought to be important, are found in experience to be of little or no account: that is, if we treat separately the groupings with and without the varying member, the characteristics are found to be identical; so that the same results would have been reached by treating both forms of combination as a single group. An excellent example is supplied by many of the Alexandrian corrections in St Mark, where we have every binary and ternary combination of אCLΔ besides the full quaternion. But the accession or loss of any primary document should always be treated as constituting a new group until observation has shown that no real difference can be detected in the results. How easily readings having the same origin might come to have an attestation perpetually varying within certain limits may be readily understood, for instance in such an example as that just cited, as soon as we apprehend clearly the manner in which ordinary casual mixture came to pass. Whether
