It is evident that the cause of this difference does not lie in the nature of the monstrosity, but is due to the criterion upon which the selection is made. Selection of the apparently best individuals is one method, and it gives admirable results. Selection on the ground of the hereditary percentages is another method and gives results which are far more advantageous than the former. In the lecture on the pistillody of the poppies we limited ourselves to the selection of the finest individuals and showed that there is always a manifest correlation between the individual [424] strength of the plant and the degree of development of its anomaly. The same holds good with other monstrosities, and badly nourished specimens of rich races with twisted or fasciated stems always tend to reversion. This reversion, however, is not necessarily correlated with the hereditary percentage and therefore does not always indicate a lessening of the degree of inheritance. This shows that even in those cases an improvement may be expected, if only the means can be found to subject the twisted and the fasciated races to the same sharp test as the tricotylous varieties. Much remains to be done, and the principle of the selection of parents according to the average constitution of their progeny seems to be one of the most promising in the whole realm of variability. Besides tricotylous, the syncotylous seedlings may be used in the same way. They are more rarely met with, and in most instances seem to belong only to the unpromising half-races. The black bindweed (_Polygonum Convolvulus_), the jointed charlock (Raphanus Raphanistrum), the glaucous evening-primrose (_Oenothera glauca_) and many other plants seem to contain such half-races. On the other hand I found a plant of _Centranthus macrosiphon_ yielding as much as 55% of syncotylous children [425] and thereby evidently betraying the nature of a rich or double race. Likewise the mercury was rich in such deviations. But the best of all was the Russian sunflower, and this was chosen for closer experiments. In the year of 1888 I had the good luck to isolate some syncotylous seedlings and of finding among them one with 19% of inheritors among its seeds. The following generation at once surpassed the ordinary average and came up in three individuals to 76, 81 and even 89%. My race was at once isolated and ameliorated by selection. I have tried to improve it further and selected the parents with the highest percentages during seven more generations; but without any remarkable result. I got figures of 90% and above, coming even in one instance up to the apparent purity of 100%. These, however, always remained extremes, the averages fluctuating yearly between 80-90% or thereabouts, and the other extremes going nearly every year downwards to 50%, the value which would be attained, if no selection were made.