We see in these examples, selected as fairly typical, that some of the variants have evidently the character of errors, while in other cases the difference of reading is due to an alternative version. The circumstances, however, of these two cases are not distinguishable, the errors are supported by as much authority as the rest, and it must be supposed that both have the same origin. If then we assume that such variations as we find (for example) in i. 3396, 3416, v. 30, 47, 82, 368, 2694, &c., are due to the author, as is almost certain, there can be no doubt that the form of text which is given by the group AJM in combination with the second and third recensions is the later of the two: and if the group H₁ ... B₂ represents an earlier type as regards this class of variation, it must surely do so also as regards the errors, which, as we have seen, stand upon the same ground in respect of manuscript authority. As we cannot help believing that the author wrote originally ‘To holde hir whil my lif may laste,’ v. 82, and ‘The more he hath the more he greedeth,’ v. 394, so we may reasonably suppose that errors such as ‘it’ for ‘hid,’ i. 1755, ‘that diere’ for ‘that other,’ ii. 833, ‘what’ for ‘war,’ iii. 1065, existed in the copy which first served as an exemplar. It may be observed here that in cases where revision seems to have taken place, we can frequently see a definite reason for the change; either the metre is made more smooth, as i. 1770, 2622, 3374, ii. 671, 751, 1763, iii. 765, 2042, 2556, iv. 234, v. 368, 1678, &c., or some name is altered into a more correct form, as where ‘Element’ is changed to ‘Clemenee,’ iv. 985, with a corresponding alteration of the rhyme, or the expression and run of the sentence is improved, as i. 368, 3416, v. 30, 1906, 6756, &c. In particular we note the tendency towards increased smoothness of metre which is shown in dealing with weak _e_ terminations. It is to be assumed on the principles which have been stated that the group ERCLB₂ and the other manuscripts which agree with them represent with more or less accuracy the first form of the author’s text, that H₁YXG and a few more form a class in which correction and revision has taken place to some extent, but partially and unsystematically, and that AJM &c. give us the first recension text in a much more fully revised and corrected form.