Respecting this mould we have scarcely more material for conjecture than with regard to the first punches and matrices. The principle of the bipartite mould was, of course, well known already. The importance of absolute squareness in the body and height of the type would demand an appliance of greater precision than the uncertain hollowed cube of sand or clay; the heat of the molten lead would point to the use of a hard metal like iron or steel; and the varying widths of the sunk letters in the matrices would suggest the adoption of some system of slides whereby the mould could be expanded or contracted laterally, without prejudice to the invariable regularity of its body and height. By what crude methods the first typefounder contrived to combine these essential qualities, we have no means of judging[38]; but were they ever so crude, to him is due the honour of the culminating achievement of the invention of typography. “His type mould,” Mr. De Vinne remarks, “was not merely the first; it is the only practical mechanism for making types. For more than four hundred years this mould has been under critical examination, and many {18} attempts have been made to supplant it. . . . But in principle, and in all the more important features, the modern mould may be regarded as the mould of Gutenberg.” [Illustration: 2. Specimens illustrating the variations in the face of type produced by bad casting.] It may be asked, if the matrices were so truly justified, and the mould so accurately adjusted, how comes it that in the first books of these Mentz printers we still discover irregularites among the letters—fewer, indeed, but of the same kind as are to be found in books printed by the artists of the ruder school? To this we reply, that these irregularities are for the most part attributable neither to varieties in the original models, nor to defects in the matrix or the mould, but to the worn or unworn condition of the type, and to the skill or want of skill of the caster. Anyone versed in the practice of type-casting in hand-moulds, is aware that the manual exercise of casting a type is peculiar and difficult. With the same mould and the same matrix, one clever workman may turn out nineteen perfect types out of twenty; while a clumsy caster will scarcely succeed in producing a single perfect type out of the number. Different letters require different contortions to “coax” the metal into all the interstices of the matrix; and it is quite possible for the same workman to vary so in his work as to be as “lucky” one day as he is unprofitable the next. In modern times, of course, none but the perfect types ever find their way into the printer’s hands, but in the early days, when, with a perishable matrix, every type cast was of consequence, the censorship would be less severe,[39] and types would be allowed to {19} pass into use which differed as much from their original model as they did from one another. Let any inexperienced reader attempt to