260 Shull. 



My observations fully confirm most of the statements made by 

 her, both as to the variability and the heredity of the heptandra-iotm. 

 I, too, have noted that "offspring of heptandra parents all show 

 dialysis and staminody of the corolla, but the extent to which these 

 malformations are exhibited varies not only in different individuals, 

 but in different flowers on the same individual". The least-developed 

 ktptandra~p\&nts in my cultures have had tubular corollas of the same 

 general form as those of the normal foxglove, but with the three 

 lobes of the lower lip inrolled at the margin and forming obvious 

 rudimentary anthers. (See Plate XK fig. 2.) The corollas modified in 

 this way are often more or less split between two or more of the 

 lobes on the ventral side of the corolla. At the opposite extreme of 

 the fluctuating series exhibited in my cultures was a form with nine 

 stamens, each of the five lobes of the corolla being represented by 

 a functional stamen of normal form. None was found with ten 

 stamens as reported by Miss SAUNDERS, nor was any observed in 

 which there was complete absence of petaloid structures, for even in 

 the most marked cases, containing nine stamens, there appeared two 

 linear-subulate petal-like phyllodes subtending the lower pair of normal 

 stamens, as shown (i, i) in most of the figures in PlateXK Not in- 

 frequently still larger lobes (f, f, in figs. 6, 8, and 9) of similar 

 character subtend the upper pair of normal stamens, and the dorsal 

 lip which is commonly undivided and truncate (figs. 3 and 4), becomes 

 in the plus-fluctuations laciniate, then divided into two filaments, 

 each with a half -anther (p, p, in the figures), and finally sometimes 

 with complete anthers, and between these two filaments a median 

 dorsal petaloid process (e). 



It is to be noted that these remnants of the corolla (i, i, f, f, e) 

 four of which subtend the four normal stamens and the fifth (e) 

 standing opposite the position which would be occupied by the fifth 

 stamen, were it present, constitute the five lobes of an "accessory 

 corolla" in which the symmetry is exactly the reverse of that seen 

 in the normal corolla, two lobes of the accessory corolla being ventral 

 and three dorsal and each occupying the position of one of the 

 sinuses of the normal corolla. The ventral components (i, i) of the 

 accessory corolla are much mor;e frequently in evidence than the three 

 dorsal lobes. They appear in a number of the published figures of 

 the heptandra-imm, while the dorsal components are much more 

 rarely seen, partly, no doubt, because the figures most commonly 

 represent flowers in which the dorsal lobe of the corolla is not suffi- 



