Ranunculus Bulbosus Scmiplcnus. 255 



flowers out of consideration. But even these did not 

 manifest any notable differences. I then compared these 

 ten curves with the part-curves determined from the 

 parents themselves (i e., with the number of flowers 

 counted on the seed-parents) and found no correlation. 

 As a matter of fact the seed-parent with the smallest 

 number of pleiopetalous flowers had the offspring with 

 the largest number. The following four curves of the 

 offspring of four seed-parents are extracted from my 

 records. Under M are given the numbers of petals in the 

 seed-parents of 1891. 



M NUMBER OF PETALS PER FLOWER 



— 5 6 7 8 9 10 11 12 13 14 15 16 17 18-23 Totals 



C 5—10 37 47 81 81 85 102 47 31 6 3 4 10 525 



C 6—10 25 67 80 75 117 77 75 45 30 10 6 1 2 3 613 



C 6—11 54 53 62 78 87 60 59 37 10 4 4 1 1 1 511 



C 7—11 52 57 76 77 95 64 26 13 460 



Another fact which points in the same direction is 

 that the plant which seemed to be far the best in the 

 summer of 1891, inasmuch as the apex of its curve was 

 over 11-12 petals, had offspring whose character cor- 

 responded exactly with that of the whole culture of 1892. 

 The improvement on the seed-parent therefore did not 

 justify the expectation of a real advance. 



For these reasons I then discontinued the experiment. 

 It seemed to me that the iinpossihUity of raising the 

 double race from my half race by simple selection z^'as 

 placed beyond all doubt. This result could only be ex- 

 pected from a further mutation. 



The extensi\-e material afforded by these cultures 

 has been utilized to find out how far the number of petals 

 per flower in tlie half race is determined, apart fn^m 

 selection, bv internal causes, and how far bv external. 



