A sharper contrast between good representatives of a race and their atavists is perhaps to be seen in no other instance. All the details contribute to the differentiation in appearance. The whole stature of the plants is affected by the varietal mark. The atavists are not, as in the case of the poppies, obviously allied with the type by a full range of intermediate steps, but quite distant from it by their rarity. There seems to be a gap in the same way as between the striped flowers of the snapdragon and their uniform red atavists, while with the poppies the atavists may be viewed as being only the extremes of a series of variations fluctuating around some average type. From this reason it is as interesting to appreciate the hereditary position of the atavists of twisted varieties as it was for the red-flowered descendants of the striped flowers. In order to ascertain this relation it is only necessary to isolate some of them during the blooming-period. I made this experiment in the summer of 1900 with the eighth generation of my race, and contrived [407] to isolate three groups of plants by the use of parchment bags, covering them alternately, so the flowers of only one group were accessible to insects, at a time. I made three groups, because the atavists show two different types. Some specimens have decussate stems, others bear all their leaves in whorls of three, but in respect to the hereditary tendency of the twisting character this difference does not seem to be of any importance. In this way I got three lots of seeds and sowed enough of them to have three groups of plants each containing about 150-200 well developed stems. Among these I counted the twisted individuals, and found nearly the same numbers for all three. The twisted parents gave as many as 41% twisted children, but the decussate atavists gave even somewhat more, viz., 44%, while the ternate specimens gave 37%. Obviously the divergences between these figures are too slight to be dwelt upon, but the fact that the atavists are as true or nearly as true inheritors of the twisted race as the best selected individuals is clearly proved by this experience. It is evident that here we have a double race, including two types, which may be combined in different degrees. These combinations determine a wide range of changes in the stature of the plants, and it seems hardly right to use the [408] same term for such changes as for common variations. It is more a contention of opposite characters than a true phenomenon of simple variability. Or perhaps we might say that it is the effect of the cooperation of a very variable mark, the twisting, with a scarcely varying attribute of the normal structure of the stem. Between the two types an endless diversity prevails, but outwardly there are limits which are never transgressed. The double race is as permanent, and in this sense as constant, as any ordinary simple variety, both in external form, and in its intimate hereditary qualities.