This general position is confirmed by a minute comparison of 1590 and 1596. To take the more massive changes first: in 1 596 Spenser completely remodelled the conclusion of Book III. Instead of bringing _Scudamour_ and _Amoret_ together, as in 1590, he left them still parted, hoping thus to prolong the interest of their story into Book IV, and so to form a link between the two volumes, which he desired to be read as one continuous poem. For this he sacrificed five glorious stanzas, one of them the most rapturous that he ever wrote. The three stanzas which he substituted are far inferior, as he must have known; but they served his purpose. He also added a new stanza at I. xi. 3. He rewrote single lines, in the interests of sound or sense; he altered single words or phrases; and he made--what is even more significant--several minute changes of order designed to improve the rhythm. Let me add that most of these changes are more happily inspired than the second-thoughts of poets have sometimes been. I hasten to make two admissions. The first volume of 1596 was not reset afresh from Spenser’s manuscript. It was printed from a copy of 1590. In the nature of the case, while it escapes some of the blunders of its original, it reproduces others, and perpetrates some new. Nor did Spenser do more than glance at the proof. The 1596 volumes, as we have seen, were printed rapidly; the poet was busy,[11] and such time as he had for proof-reading was given to the new books. I infer that the alterations which he made in 1590 were made not on the proof, but on the copy. In no other way can we account for that combination of author’s corrections with printer’s errors which marks the first volume of 1596. And this conclusion is strengthened by another consideration. It is one of the worst faults of 1596 that it so often ignores _F. E._ But the significance of the fault has been overlooked. Making corrections on the copy, Spenser did not trouble himself about errors that he had already noted in _F. E._; had he made his corrections on the proof, they could not have escaped him. I believe this to be a true account of the relations of these two texts. But when all is said there remain many places where we cannot pronounce on mere inspection whether an alteration is the author’s or the printer’s, but must be guided by a calculation of probabilities, inclining (e. g.) to the author where there is clear evidence of his hand in the neighbourhood of the vexed passage, to the printer where the _ductus litterarum_ in both readings is suspiciously alike. The most important of these places are discussed in the Critical Appendix to Vol. III.