Page:The Library, volume 5, series 3.djvu/217

 MIRACLE CYCLES. 205 For a critical text, therefore, an editor, while adopting the recension of H, should take D as his basis, this being the oldest manuscript and that in which the linguistic forms are in general least corrupt. But it will serve as a basis only. He will need to collate it throughout with H. Where D and H differ he will consult B, or should B fail, then K or W. If B, or K W, supports H, he should replace the reading of D by that of H. If B, or K W, supports D, he will not be able to argue that D is correft, but he will know that there is a divergence between H and /3. In such a case, and in the absence of any independent authority such as P, he will have to rely upon internal evidence to determine which is the more likely to be the original reading. He will remember that the scribe of H has been convifted of editing his text, and that the mathematical probability is about two to one in favour of /3. W. W. GREG.