What is true of Caxton’s early works is also true of a large number of other fifteenth century printed books. Mr. Hessels, after quoting the testimony of Mr. Bradshaw of Cambridge, and Mr. Winter Jones of the British Museum, refers to a large number of incunabula in which he has found evidence that this mode of printing was the common practice of the early typographers.[49] Assuming, then, that the first books were generally printed page by page, it will be seen that the stock of type necessary to enable the printer to proceed was but small. 2,700 letters would suffice for one page of the forty-two-line _Bible_; and for the _Rationale Durandi_, about 5,000 would be required. It is probable, however, that, as Bernard suggests, the printers would cast enough to enable one forme to be composed while the other was working, so that double these quantities would possibly be provided. Nor must it be forgotten that a “fount” of type in these days consisted not only of the ordinary letters of the alphabet, but of a very large number of double letters, abbreviations and contractions, which must have seriously complicated the labour of composition, as well as reduced the individual number of each type required to fill the typefounder’s “bill.” This feature, doubtless attributable to the attempt on the part of the early printers to imitate manuscript as closely as possible, as well as to the exigencies of justification in composition, which, in the absence of a variety of spaces, required various widths in the letters themselves, was common to both schools of early typography. M. Bernard states that, in the type of the forty-two-line _Bible_, each letter required at least three or four varieties; while with regard to Caxton’s type 1, which was designed and cast by Colard Mansion at Bruges, before 1472, Mr. Blades points out that the fount contained upwards of 163 sorts, and that there were only five letters of which there were not more than one matrix, either as single letters or in combination. Speaking of the _Speculum_, Mr. Skeen counts 1,430 types on one page, of which 22 are _a_, 61 _e_, 91 _i_, 73 _o_, 37 _u_, 22 _d_, 14 _h_, 30 _m_, 50 _n_, 42 _s_, and 41 _t_; besides which there are no less than ninety duplicate and triplicate characters, comprising one variation of _a_, 15 of _c_, 7 of _d_, 3 of _e_, 9 of _f_, 10 of _g_, 3 of _i_, 7 of _l_, 2 of _o_, 3 of _n_, 2 of _p_, 10 of _r_, 9 of _s_, 9 of _t_, varying in the frequency of their occurrence from once to eleven times, leaving but 541 other letters for the rest of the alphabet, including the capitals; {28} and of these last, from three to twenty would be the utmost of each required. Altogether, calculating 138 matrices (_i.e._, two alphabets of twenty-four letters each, and ninety double and treble letters) to be the least number of matrices required to make a complete fount,[50] the highest number of types of any one particular sort necessary to print a single page would be ninety-one.