which each of the rival readings is unobjectionable, so that either of them would be reasonably approved in the absence of the other. The latter kind, resting on 'transcriptional' probability, is not less valuable; but it is subject to analogous uncertainty, because in a vast number of cases each reading can be explained as a corruption of the other by reference to some tendency of scribes which is known to be often productive of textual change, and the tendency which actually operated in producing change in any particular case need not be the tendency which is most obvious to modern eyes. A few hours spent in studying a series of the countless corrections which no one would think of accepting will shew the variety of instinct to be found among scribes, the frequent disagreement between their instincts and our own, and, above all, the conflicting effects of different instincts in the same passage. Moreover, though normally a scribe's correction, or, more properly, corruption, should exhibit at once plausibility and latent inferiority, that is, should be condemned by transcriptional and by intrinsic evidence alike, the imperfection of our knowledge more commonly leaves unreconciled the apparent conflict of the two kinds of probability, arising out of the consideration that no scribe would consciously introduce a worse reading instead of a better. Lastly, all decisions made solely or chiefly on the ground of internal evidence are subject to the chances of mistake inseparable from single and isolated judgements: they lack the security given by comparison and mutual correction. Hence it is dangerous to fix the mind in the first instance on any kind of internal probability: the bias thus inevitably acquired can hardly fail to mislead where the authority of documents is not obviously clear and decisive at once. The uses of internal evidence are subordinate and accessory: if taken as the primary guide, it cannot but lead to extensive error.
Documentary evidence in its simplest form consists in the relative authority of individual documents; that is, in the relative antecedent probability that a reading attested by them is the true reading. This is what is meant when it is said in popular language that 'good MSS' should be trusted. A presumption of relatively high authority is conferred by priority of date; a presumption verified on the average by experience, but still no more than a presumption, because the exemplar from which a MS was copied may have been either only a little older than itself or of any earlier
