262 Shull. 



ditions, but in each of these the ventral pair of normal stamens (k, k) 

 are united with the median ventral accessory stamen (g) to form a 

 triple structure, and in figs. 8 and 9 there is a partial or complete 

 union of the dorsal normal stamens (1, 1) with the lateral accessory 

 stamens (h, h). 



The periodicity in the degree to which the heptandra- character 

 is developed at different levels on the raceme, observed by Miss 

 SAUNDERS, was also noted in my cultures; but in no case was a 

 completely normal corolla found on any heptandra plant grown at the 

 Station for Experimental Evolution, such as Miss SAUNDERS states 

 appeared rarely in the higher parts of the racemes on certain plants 

 in her cultures which had the lowest grade of development of the 

 /te#tandra-cha.i:a.cter . My failure to find such normal flowers may 

 very well be due to the comparatively early death of the flower-stems 

 owing to injury from thrips. In my cultures as in those of Miss 

 SAUNDERS, the distinction between the heptandra plants and the 

 normal was absolute. 



Miss SAUNDERS has demonstrated that the heptandra-iorm is a 

 simple Mendelian recessive to the normal type, and the following 

 account of my experiences completely supports this conclusion. 



In January, 1906, the seeds received the previous autumn from 

 Mr. SMITH, proprietor of the Daisy Hill Nursery, Newry, England, 

 were sown in sterilized soil in the propagating-house, and there resulted 

 161 plants which, on May 21, were set into the garden at intervals 

 of one meter. Despite this early start, none bloomed in 1906, but 

 all formed large rosettes, a few of which were destroyed by rot in 

 the latter part of summer. Many more were lost during winter and 

 spring, so that not over one hundred plants bloomed in 1907. All 

 of these were of the keptandra-iQrm. The seeds had been collected, 

 according to Mr. SMITH, from a plant which had "lost its corolla 

 entirely, but not much its fecundity". It was therefore of the extreme 

 heptandra-tyipe. The offspring showed a great preponderance of the 

 more highly developed heptandra-form.) but varying to the least 

 developed extreme. 



A number of attempts were made to secure seeds by crossing 

 reciprocally both the least developed and the most highly developed 

 specimens in this family, with normal Digitalis purpurea, both white- 

 flowered and purple-flowered, which was growing in a neighboring 

 garden at a distance of about 100 meters from my cultures. The 

 attacks of thrips were so disastrous that almost no seeds were secured 



