BOTANY: A. B. STOUT 
133 
giving concentration in the epidermis of the upper surface of the leaves. Some 
of the variations involve very striking changes in epidermal pigmentation, 
superficially resembhng chimeral relationships. During the first seven gen- 
erations, sixteen distinct color patterns were obtained by bud variation and 
isolated by selection; fifteen of these were produced as marked sudden varia- 
tions, and six of the types also occurred as fluctuating variations. One pat- 
tern has appeared only as a fluctuation. Between any of these types a wide 
range of intermediates was found from which many additional patterns might 
probably have been isolated. There has been reversion to parent patterns; 
colors that have been lost or thrown out have reappeared; yellow-green pat- 
terns have given pure green sports and later the yellow has reappeared. 
A calculation of the frequency of bud variations on the basis of the esti- 
mated number of buds that developed into branches shows clearly that de- 
crease of red occurred with about twice the frequency as did increase of red; 
likewise decrease of yellow occurred about twice as often as increase of yellow. 
Among sister lines of clonal descent there were marked differences in the ratios 
of frequency for any one change in coloration. The most frequent change was 
loss of yellow with increase of green, for which the frequency ratio was 
1 : 2960. The change occurring with least frequency was that of increase of 
epidermal red pigmentation to solid red for which the ratio was 1 : 19,250. 
Variations in leaf form were fully as striking as those of pigmentation. 
Deeply laciniate-leaved forms arose in thirteen instances as fluctuations af- 
fecting an entire plant, and in one case as a decided bud variation. A strik- 
ing feature of the laciniate character was the marked periodicity in its de- 
velopment; plants of the variety having strongly laciniate leaves in winter 
produced entire leaves in summer, while the great bulk of the sister plants of 
other varieties produced only entire leaves. In general these variations in 
form are continuous and the extremes are in decided contrast with each other. 
Selection for extremes and for intermediates has in every case given a prog- 
eny of marked constancy but with further fluctuations and sporadic varia- 
tions about a new mode. The types thus arising are for purposes of prop- 
agation the equivalents of the 'Kleinarten' or 'biot3^es' that commonly 
occur in species propagated by seed. In their bearing on the theories of con- 
tinuous variation and the effects of selection, my results are quite identical 
with those obtained by Castle and Phiflips (1914) in their study of color pat- 
terns in biparental reproduction in rats and interpreted as indicative of actual 
variation in the hereditary units. The results are also quite identical with 
those Jennings (1916) has obtained with Difflugia, by asexual propagation 
analogous to that I have used in Coleus. 
It is quite clear that the changes seen in Coleus do not involve a permanent 
loss of definite hereditary units by vegetative segregation. There is also no 
opportunity for any such recombination of multiple modifying factors as is 
assumed to give similar variations in sexually reproduced progeny. As far 
as is known vegetative propagation gives the greatest possible degree of purity 
