This may be proved both directly and inferentially.[5] In some instances—as in that of an elegy upon Saint Columba in the Book of the Dun Cow—the dates of authorship are actually given. In others, we may depend upon evidence which, if not quite so absolute, is nearly as convincing. Even where the writer does not state that he is copying from older manuscripts, it is obvious that this must have been the case, from the glosses in his version. The scribes of the earlier Gaelic manuscripts very often found, in the documents from which they themselves were copying, words so archaic as to be unintelligible to the readers of their own period. To render them comprehensible, they were obliged to insert marginal notes which explained these obsolete words by reference to other manuscripts more ancient still. Often the mediæval copyists have ignorantly moved these notes from the margin into the text, where they remain, like philological fossils, to give evidence of previous forms of life. The documents from which they were taken have perished, leaving the mediæval copies as their sole record. In the Welsh Mabinogion the same process is apparent. Peculiarities in the existing manuscripts show plainly enough that they must have been copied from some more archaic text. Besides this, they are, as they at present stand, obviously made up of earlier tales pieced together. Almost as clearly as the Gaelic manuscripts, the Welsh point us back to older and more primitive forms. The ancient legends of the Gael and the Briton are thus shown to have been no mere inventions of scholarly monks in the Middle Ages. We have now to trace, if possible, the date, not necessarily of their first appearance on men’s lips, but of their first redaction into writing in approximately the form in which we have them now.