546 Journal of Agricultural Research voi. xix. No. u 



few degenerate plants are found. Irish Cobbler and Early Triumph 

 have behaved in just the opposite way, at least so far as degeneration is 

 concerned. Within the last two varieties, as grown under our condi- 

 tions, intelligent selection will give surprising results. With the Russet 

 Burbank, profitable increase in yield might result from selection, but 

 the improvement would be in no way spectacular. This type of selec- 

 tion work aims at increased production through the elimination of 

 plants yielding below the average of the population. 



The first work in line selection was undertaken in 191 3 with the selec- 

 tion from the crop of that year of a few high-yielding hills of Russet 

 Burbank (hills 201 to 206) and Rural New Yorker (hills 207 to 212). 

 These have now been tested for five years, and a summary of the results 

 is given in Table I. While these data do not represent a very large 

 amount of experimental evidence, I believe that the results indicate 

 what can be expected when effort is made to isolate high-yielding clonal 

 lines by hill selection and maintain them by mass selection based upon 

 tuber characteristics. Every effort was made to eliminate experimental 

 errors. Seed pieces were dropped 15 inches apart in rows 3 feet 9 inches 

 apart. Single-row plots were used with outside or guard rows in 

 every case. In 1914, these plots were necessarily small. In 1915, 

 i/40-acre plots were used, and since 191 5 plots of approximately i/20-acre 

 have been grown. No special effort has been made to control the size 

 of the seed piece; but in 191 4, 1915, and 1916 the plots were thinned to 

 single-stemmed hills. This method gives promise of being as reliable as 

 one based on the planting of uniform-sized seed pieces, for size of seed piece- 

 appears to influence yield only in so far as it is correlated with the number 

 of stems produced in the hill. In other years the size of seed pieces could 

 not have varied greatly, since the plan of cutting tubers of certain sizes 

 to so many seed pieces was followed, without reference to number of 

 eyes. The hills selected in 191 3 were single-stemmed hills and were 

 chosen on yield and tuber characteristics entirely. Of a large number 

 of hills dug with a fork, those were selected which gave the greatest 

 weight of tubers of good form. Tubers considered of good form were 

 those long for the variety, full and rounded at both ends, and oval 

 rather than round in cross section. In choosing seed by mass selection, 

 the same form characteristics were considered; and the seed was picked 

 after the digger. The seed for control plots was chosen by mass selection 

 from the population of the variety from which the hills originally came, 

 selection being based upon the same tuber characteristics. Single-plot 

 tests were employed. 



In Table I, I wish to call attention not so much to the 5-year per- 

 formance record of the hill lines as to the very marked variation between 

 their yields in 19 14 and 1915 when compared with the control plots. 

 If we take the 5 -year average as a reliable indicator, some of these lines 

 are apparently better yielders than others. I believe, however, that an- 



