It is here that the manuscripts come to our aid. To take, indeed, any single manuscript, as Dr. Grosart did, and select this or that reading from it as seems to you good, is not a justifiable procedure. This is simply to add to the editions one more possible source of error. There is no single manuscript which could with any security be substituted for _1633_. Our analysis of that edition has made it appear probable that a manuscript resembling _D_, _H49_, _Lec_ was the source of a large part of its text. But it would be very rash to prefer _D_, _H49_, _Lec_ as a whole to _1633_.[41] It corrects some errors in that edition; it has others of its own. Even _W_, which has a completer version of some poems than _1633_, in these poems makes some mistakes which _1633_ avoids. If the manuscripts are to help us it must be by collating them, and establishing what one might call the agreement of the manuscripts whether universal or partial, noting in the latter case the comparative value of the different groups. When we do this we get at once an interesting result. We find that in about nine cases out of ten the agreement of the manuscripts is on the side of those readings of _1633_ which are supported by the tests of intrinsic probability referred to above,[42] and on the other hand we find that sometimes the agreement of the manuscripts is on the side of the later editions, and that in such cases there is a good deal to be said for the later reading.[43] The first result of a collation of the manuscripts is thus to vindicate _1633_, and to provide us with a means of distinguishing among later variants those which have, from those which have not, authority. But in vindicating _1633_ the agreement of the manuscripts vindicates itself. If _B_'s evidence is found always or most often to support _A_, a good witness, on those points on which _A_'s evidence is in itself most probably correct, not only is _A_'s evidence strengthened but _B_'s own character as a witness is established, and he may be called in when _A_, followed by _C_, an inferior witness, has gone astray. In some cases the manuscripts _alone_ give us what is obviously the correct reading, e.g. p. 25, l. 22, 'But wee no more' for 'But now no more'; p. 72, l. 26, 'his first minute' for 'his short minute'. These are exceptionally clear cases. There are some where, I have no doubt, my preference of the reading of the manuscripts to that of the editions will not be approved by every reader. I have adopted no rigid rule, but considered each case on its merits. All the circumstances already referred to have to be weighed--which reading is most likely to have arisen from the other, what is Donne's usage elsewhere, what Scholastic or other 'metaphysical' dogma underlies the conceit, and what is the source of the text of a particular poem in _1633_.