Such tricotylous double races offer highly interesting material for inquiries into questions of heredity, as they have such a wide range of variability. There is little danger in asserting that they go upwards to nearly 100%, and downwards to 0%, diverging symmetrically on both sides of their average (50-55%). These limits they obviously cannot transgress, and are not even able to reach them. Samples of seed consisting only of tricotyls are very rare, and when they are met with the presumption is that they are too few to betray the rare aberrants they might otherwise contain. Experimental evidence can only be reached by the culture of a succeeding generation, and this always discloses the hidden qualities, showing that the double [422] type was only temporarily lost, but bound to return as soon as new trials are made. This wide range of variability between definite limits is coupled with a high degree of sensibility and adequateness to the most diverging experiments. Our tricotylous double races are perhaps more sensitive to selection than any other variety, and equally dependent on outer circumstances. Here, however, I will limit myself to a discussion of the former point. In the second generation after the isolation of stray tricotylous seedlings the average condition of the race is usually reached, but only by some of the strongest individuals, and if we continue the race, sowing or planting only from their offspring, the next generation will show the ordinary type of variability, going upwards in some and downwards in other instances. With the _Phacelia_ and the mercury and some others I had the good luck in this one generation to reach as high as nearly 90% of tricotylous seedlings, a figure indicating that the normal dicotylous type had already become rare in the race. In other cases 80% or nearly 80% was easily attained. Any further divergence from the average would have required very much larger sowings, the effect of selection between a limited number of parents being only to retain the high degree once [423] reached; so for instance with the mercury, I had three succeeding generations of selection after reaching the average of 55%, but their extremes gave no increasing advance, remaining at 86, 92 and 91%. If we compare these results with the effects of selection in twisted and fasciated races, we observe a marked contrast. Here they reached their height at 30-40%, and no number of generations had the power of making any further improvement. The tricotyls come up in two generations to a proportion of about 54%, which shows itself to correspond to the average type. And as soon as this is reached, only one generation is required to obtain a very considerable improvement, going up to 80 or even 90%.