Tricot y Is, Ucmi-tricotyls, and Tcfrocotyls. 363 



laria nodosa produced, in the harvest of 1894, a mean 

 of 2% and a maximum of 5.5% tricotyls per seed-parent. 

 I harvested the seeds from two tetracotyls and ol^tained 

 0.5% and 3% tricotyls. Amongst the 2000 seedhngs 

 wliich these cultures contained, there were 30 tricotyls, 

 3 hemi-tricotyls and only 2 tetracotyls. 



On the other hand a tetracotylous plant of Aspcrula 

 a-ziirca gave 7% and the corres]:)onding tricotylous seed- 

 ]:)a rents only 2%, in 1892. Of this 7%, there were 5% 

 tricotylous, 1% tetracotylous, and 1% hemi-tricotylous. 

 I bred tlie tetracotyls of Aniarantiis speciosus for two 

 s^enerations, in 1893 and 1894. In the summer of 1893, 

 9 tetracotylous plants were left to flower; 3 proved to be 

 fasciated, but the rest gave values varying from 1 to 

 7.5% with a mean of 5%. I counted for each seed- 

 parent 500-1000 seedlings. The corresponding tricotyl- 

 ous culture gave values from 2.5% to 7.5%, that is, a 

 mean of 4.5%; from each of the 15 seed-parents from 

 700 to 1000 seedlings being examined. We see there is 

 practically no difference between the two cases. Together 

 the tetracotylous parents produced only 6 tetracotylous 

 offspring among 4000 seedlings, and the tricotyls 13 

 among 10,000; that is to say, they behaved in regard 

 to this character as variants of the same race. I then 

 selected the tetracotylous offspring of the tetracotvlous 

 seed-parents for a continuation of the race in 1894, but 

 observed no further progress, the percentage in tetra- 

 cotyls being only 0.2%. 



The question suggests itself whether the proportion 

 of tetracotylous seedlings, perhaps, simph- obevs the laws 

 of proba1)ility. The splitting of a cotyledon may be im- 

 agined to be distributed at random over a group of sav 

 100 individuals, and we mav ask. how manv times a 



