In some instances the degree of their inheritance is only a very small one. The isolated tricotyls yield 1 or 2% of inheritors, in some cases even less, or upwards up to 3 or 4%. If the experiment is repeated, no amelioration is observed, and this result remains the same during a series of successive generations. In the case of _Polygonum convolvulus_, the Black bindweed, I have tried as many as six generations without ever obtaining more than 3%. With other species I have limited myself to four successive years with the same negative result, as with spinage, the Moldavian dragon-head, (_Dracocephalum moldavicum_), and two species of corn catch-fly (_Silene conica_ and _S. conoidea_). Such poor races hardly afford a desirable material for further inquiries. Happily the rich races, though rare, may be discovered also from time to time. They seem to be more common among cultivated plants and horticultural as well as agricultural species may be used. Hemp [420] and mercury (_Mercurialis annua_) among the first, snapdragon, poppies, _Phacelia_, _Helichrysum_, and _Clarkia_ among garden-flowers may be given as instances of species containing the rich tricotylous double races. It is very interesting to note how strong the difference is between such cases and those which only yield poor races. The rich type at once betrays itself. No repeated selection is required. The stray tricotyls themselves, that are sought out from among the original samples, give hereditary percentages of a much higher type after isolation than those quoted above. They come up to 10-20% and in some cases even to 40%. As may be expected, individual differences occur, and it must even be supposed that some of the original tricotyls may not be pure, but hybrids between tricotylous and dicotylous parents. These are at once eliminated by selection, and if only the tricotyls which have the highest percentages are chosen for the continuance of the new race, the second generation comes up with equal numbers of dicotyls and tricotyls among the seedlings. The figures have been observed to range from 51-58% in the majority of the cases, and average 55%, rarely diverging somewhat more from this average. Here we have the true type of an ever-sporting variety. Every year it produces in the [421] same way heirs and atavists. Every plant, if fertilized with its own pollen, gives rise to both types. The parent itself may be tricotylous or dicotylous, or show any amount of multiplication and cleavage in its seed-leaves, but it always gives the entire range among its progeny of the variation. One may even select the atavists, pollinate them purely and repeat this in a succeeding generation without any chance of changing the result. On an average the atavists may give lower hereditary figures, but the difference will be only slight.