356 HANS AND OLOF TKDIN 



discussion we will calculate with one factor for stem thickness, 

 the r of K. & P. 



CORRELATIONS BETWEEN STEM THICKNESS AND INTER- 

 NODE LENGTH AND THEIR INFLUENCE UPON 

 STEM HEIGHT. 



K. & P. are of the opinion that tallness of peas should he due 

 to a comhination of the factor Le for long internodes and a factor, 

 T, for thick stem. The following may be quoted from their paper 

 (pag. 5): »The other kind of semi-dwarf lacks the thick stem factor, 

 and in the absence of this factor the long internode factor cannot 

 build the stem segments of sufficient length to produce tallness in 

 plant.» So the opinion of K. & P. seems to be that the factor T, 

 introduced in the crosses by one parent (in our crosses by 0234), 

 should increase the length of internodes caused in the other parent 

 by Le . And further, when h is brought together with t, the inter- 

 nodes should be still shorter than in the Tie parent. This would 

 give an explanation of the transgressions in internode length, observed 

 in our crosses. If the view of K. & P. is correct, there ought to be 

 within the Le and the h part of the Fois, taken separately, a marked 

 correlation between stem thickness and internode length. The T 

 plants should not only have thick stem but also longer internodes 

 than the t plants. Now the coefficient of correlation between length 

 of internodes and thickness of stem has been determined — according 

 to the formula of Bravais and with the methods of Johaxnsen 

 (1913) — for the Lc and h parts of crosses I + II and III. In the 

 h part of crosses I + II the coefficient of correlation (r) is 0,22o + 0,082, 

 in the Le part of cross III r = 0,i2o + 0,05i, in the Le part of crosses 

 I + II and in the /,- part of cross III r is far less than its own standard 

 error. In cross IV, where both parents must have been Le, but Rapid 

 t and Stensärt T, Fo transgreedes the t parent in respect to internode 

 length, and segregates in stem thickness. No correlation is foimd, 

 however, between those two characters, the coefficient of correlation 

 being less than its own standard error. Thus in three cases out of 

 five there is no sign of correlation, in two cases there is a correlation. 

 The coefficient, however, is even there rather low and not sufficiently 

 marked (less than 3 times its standard error), the existence of a 

 correlation being therefore not quite certain. As the material is not 

 small, these results can hardlv be considered as occasional. At least 



