Elsewhere Virgil and also some lines of Columella and Varro go to prove in the same way that selection was applied by the Romans to their cereals, and that it was absolutely necessary to keep their races pure. There is little doubt, but that it was the same principle as that which has led, after many centuries, to the complete isolation and improvement of the very best races of the mixed varieties. It further proves that the mixed conditions of the cereals was known to man at that time, although distinct ideas of specific marks and differences were of course still wholly lacking. It is proof also that cultivated cereals from the earliest times must have been built up of numerous elementary forms. Moreover it is very probable, that in the lapse of centuries a goodly number of such types must have disappeared. [107] Among the vanished forms are the special barley and wheat of the lake-dwellings, the remains of which have been accidentally preserved, but most of the forms must have disappeared without leaving any trace. This inference is supported by the researches of Solms-Laubach, who found that in Abyssinia numerous primitive types of cereals are still in culture. They are not adequate to compete with our present varieties, and would no doubt also have disappeared, had they not been preserved by such quite accidental and almost primitive isolation. Closing this somewhat long digression into history we will now resume our discussion concerning the origin of the method of selecting cereals for isolation and segregate-cultivation. Some decades after Le Couteur, this method was taken up by the celebrated breeder Patrick Sheriff of Haddington in Scotland. His belief, which was general at that time, was "That cultivation has not been found to change well defined kinds, and that improvement can be best attained by selecting new and superior varieties, which nature occasionally produces, as if inviting the husbandman to stretch forth his hand and cultivate them." Before going into the details of Sheriff's work it is as well to say something concerning [108] the use of the word "selection." This word was used by Sheriff as seen in the quotation given, and it was obviously designed to convey the same idea as the word "lecta" in the quotation from Virgil. It was a choice of the best plants from among known mixed fields, but the chosen individuals were considered to be representatives of pure and constant races, which could only be isolated, but not ameliorated. Selection therefore, in the primitive sense of the word, is the choice of elementary species and varieties, with no other purpose than that of keeping them as pure as possible from the admixture of minor sorts. The Romans attained this end only imperfectly, simply because the laws governing the struggle for life and the competition of numerous sorts in the fields were unsuspected by them.