which makes it difficult to read; and not compressed laterally, as all later type has grown to be owing to commercial exigencies. There was only one source from which to take examples of this perfected Roman type, to wit, the works of the great Venetian printers of the fifteenth century, of whom Nicholas Jenson produced the completest and most Roman characters from 1470 to 1476. This type I studied with much care, getting it photographed to a big scale, and drawing it over many times before I began designing my own letter; so that though I think I mastered the essence of it, I did not copy it servilely; in fact, my Roman type, especially in the lower case, tends rather more to the Gothic than does Jenson's."
By the middle of August eleven punches had been cut for the new fount, and Morris had determined that Caxton's "Golden Legend" should be the first large work produced by his press. He himself had recently acquired a copy of the edition of 1527 printed by Wynkyn de Worde. The Kelmscott "Golden Legend" was, however, set up, not from this, but from Caxton's own first edition of 1483. The almost priceless original was borrowed, under a heavy bond in case of loss or injury, from the Cambridge University Library for this purpose. It was transcribed for the press—a work of such laborious magnitude, and one executed so patiently and carefully as to deserve commemoration—by Miss Phillis M. Ellis, the daughter of his old friend and collaborator. Ellis himself took the chief part in superintending the accuracy of the text, a work of no small difficulty.
On the 29th of August, 1890, Morris writes to him from Kelmscott:
"Please pardon me for not answering your letter sooner; you know my little ways. Also I did want
