No. 565] CHANGES PRODUCED BY SELECTION 13 



however, show that such judgments would have been 

 superficial. The general type of the plant did appear to 

 be fixed, but the frequency distribution for number of 

 leaves of the F 5 and F 6 populations were not the same. 

 Strictly speaking, they were not fixed. What would be 

 the result of selecting (and selfing) extremes from these 

 different families for a number of years? A tentative 

 answer to this question is to be obtained by examining 

 the remainder of our tables. 



The tables are arranged roughly in the order of the 

 effect that selection has had in changing the mean of the 

 various families that were the starting points of this part 

 of the experiment. The selections were grown near Bloom- 

 field, Connecticut, on the light sandy loam of that region, 

 soil typical of that which produces the famous Connecti- 

 cut Eiver Valley wrapper tobacco. Duplicate experi- 

 ments with several of the original families were made at 

 New Haven, Connecticut, however, on an impoverished 

 soil not fitted to grow a good quality of tobacco even after 

 supplying large quantities of tobacco fertilizer, and in 

 the condition used not fitted to grow good crops of any 

 kind. Two families were also grown in triplicate, the 

 third selections being planted at Forest Hills, Massachu- 

 setts, on a very fine type of rich garden land which brought 

 out maximum luxuriance of growth, but which did not 

 produce good tobacco quality. These experiments were 

 not true repetitions of the experiments at Bloomfield, 

 Connecticut, since aliquot portions of the seed from the 

 selfed plant grown there were not sent to the other places 

 to be grown. But they were duplicates in that each 

 family came from the same F 4 or F 5 mother plant, 

 although, beginning with the F 5 or F fi population, differ- 

 ent selfed seed plants furnished the starting point of selec- 

 tions carried on independently. In this way there were 

 afforded a greater number of chances to see what selec- 

 tion could do. 



Table III shows the results obtained from family No. 

 77. This family arose from an F 5 plant having 23 leaves, 

 one below the modal leaf number if we may judge from 



