(2) But when the question is as to what these bounds may be, we can only answer in a general way that, according to the general theory of evolution, they must be such as are imposed by heredity, coupled with the degree to which external conditions of life (and possibly also use-inheritance) are capable, in given cases, of modifying congenital characters. These are the only causes which the theory of descent can consistently recognise as producing variations in determinate directions. (3) Inasmuch as variation presupposes the existence of parts that vary, and inasmuch as the variation of parts can only be in the alternative directions of increase or decrease around an average, it follows that, in the first instance at all events, every variation, if determinate, must be so only in one or other of these two opposite directions. (4) In as far as variations are summated in successive generations, so as eventually to give rise to new structures, organs, mechanisms, &c., natural selection is theoretically competent to explain the facts, without our having to postulate the operation of unknown causes producing variations in determinate lines,--or not further than is stated in paragraphs 1 and 2. (5) Nevertheless, it does not follow that there are not such other unknown causes; and, if there are, of course the importance of natural selection as a cause of adaptive modification would be limited in proportion to their number and the extent of their operation. But it is for those who, like the late Professors Asa Gray and Nägeli, maintain the existence of such causes, to substantiate their belief by indicating them.