So it is in many cases. If a variety is observed in a number of specimens at the time of its first discovery, and at a locality not studied previously, it takes the aspect of an old form of limited distribution, and little can be learned as to the circumstances under which it arose. If on the contrary it occurs in very small numbers or perhaps even in a single individual, and if the spot where it is found is located so that it could hardly have escaped previous observation, then the presumption of a recent origin seems justified. What has to be ascertained on such occasions to give them scientific value? Three points strike me as being of the highest importance. First, the constancy of the new type; secondly, the occurrence or lack of intermediates, and last, but not least, the direct observation of a repeated production. The first two points are easily ascertained. Whether the new type is linked with its more common supposed ancestor by intermediate steps is a query which at once strikes the botanist. It is usually recorded in such cases, and we may state at once that the general result is, that such intermediates do not occur. This is [580] of the highest importance and admits of only two explanations. One is that intermediates may be assumed to have preceded the existent developed form, and to have died out afterwards. But why should they have done so, especially in cases of recent changes? On the other hand the intermediates may be lacking because they have never existed, the change having taken place by a sudden leap, such as the mutations described in our former lectures. It is manifest that the assumption of hypothetical intermediates could only gain some probability if they had been found in some instance. Since they do not occur, the hypothesis seems wholly unsupported. The second point is the constancy of the new type. Seeds should be saved and sown. If the plant fertilizes itself without the aid of insects, as do some evening-primroses, the seed saved from the native locality may prove wholly pure, and if it does give rise to a uniform progeny the constancy of the race may be assumed to be proved, provided that repeated trials do not bring to light any exceptions. If the offspring shows more than one type, cross-fertilization is always to be looked to as the most probable cause, and should be excluded, in order to sow pure seeds. Garden-experiments of this kind, and repeated trials, should always be combined [581] with the discovery of a presumed mutation. In many instances the authors have realized the importance of this point, and new types have been found constant from the very beginning. Many cases are known which show no reversions and even no partial reversions. This fact throws a distinct light on our first point, as it makes the hypothesis of a slow and gradual development still more improbable.