ipoS.] IMPROVEMENT OF THE POTATO. 397 



wirth (41), in three years' observations with both early and late 

 varieties found a relation between yield and the following charac- 

 ers: Flat shape, number of stalks per plant, length of growing 

 period, height of plants and number of tubers per plant. He also 

 states that the greater the number of stalks from a plant, the 

 thinner are the stalks. 



Some of these correlations may help materially in eliminating 

 some of the undesirable plants from the progeny of crosses from 

 which we are trying to build up new varieties. There may also be 

 many pairs of characters with high percentage correlation which 

 have not yet been noticed but which will be brought out with 

 further statistical studies. It is doubtful, however, if many of 

 these characters which appear to be related in certain varieties, 

 are to be regarded as real correlations characteristic of the species. 

 Reasoning from statistical studies of the writer on maize and sugar 

 beets, which may or may not be analogous, it would seem that cor- 

 relations which are likely to be of most practical value in making 

 selections will probably be found in very narrow blood lines (ele- 

 mentary species?). For example, in ear-to-the-row tests of dent 

 maize there was planted an ear with a peculiarly shaped tip. This 

 ear yielded very highly, and of its progeny when grown and their 

 yield compared, in nearly every case, those were found to average 

 highest whose mothers had this peculiar -tip. 



In hybridization, pairs of characters may be found which are 

 inherited as a single character. When both characters are desir- 

 able, this would be a decided help, but when one character is unde- 

 sirable, there is only the consolation of knowing the difficulty of 

 finding exceptions to the rule. Johannsen (58) states however 

 that, "Crossing is the means of breaking the correlation/' 



In selection, those correlations are of greatest value which al- 

 low us to eliminate plants through correlations of characters in the 

 young vines, with characters in the tubers. It is doubtful, however, 

 whether weak correlations should be used in originating varieties. 

 They are probably of practical value only when the life history of 

 the variety is known. 



