This single exceptional plant has been the starting point of my double variety. It was not remarkable for its terminal head, which exhibited the average number of rays of the 21-rayed race. Nor was it distinguished by the average figure for all its heads. It was only selected because it was the one plant which had some secondary heads with one ray more than all the others. This indication was very slight, and could not have been detected save by the counting of the rays of thousands of heads. [499] But the rarity of the anomaly was exactly the indication wanted, and the same deviation would have had no signification whatever, had it occurred in a group fluctuating symmetrically around the average figure. On the other hand, the observed anomaly was only an indication, and no guarantee of future developments. Here it should be remarked that the indication alluded to was not the appearance of the expected character of doubling in ever so slight a measure. It was only a guide to be followed in further work. The real character of double flower-heads among composites lies in the production of rays on the disk. No increase of the number of the outer rays can have the same significance. A hasty inspection of double flower heads may convey the idea that all rays are arranged around a little central cluster of disk-florets, the remainder of the original disk-florets but a closer investigation will always reveal the fallacy of this conclusion. Hidden between the inner rays, and covered by them, lie the little tubular and fertile florets everywhere on the disk. They may not be easily seen, but if the supernumerary rays are pulled out, the disk may be seen to bear numerous small florets at intervals. But these intervals are not at all numerous, showing thereby that only a relatively small number of tubes has been [500] converted into rays. This conversion is obviously the true mark of the doubling, and before traces of it are found, no assertion whatever can be given as to the issue of the pedigree experiment. Three more years were required before this first, but decisive trace was discovered. During these years I subjected my strain to the same sharp selection as has already been described. The chosen ancestor of the race had flowered in 1896, and the next year I sowed its seeds only. From this generation I chose the one plant with the largest number of rays in its terminal head, and repeated this in the following year.