Variation and Correlation of Oats — Part II 211 



There is a decided increase, then, in the correlation of average weight 

 of kernels per plant with height of plant, length of head, and average 

 number of kernels per culm, as plants are crowded together more closely 

 in the row. The same condition exists also when average weight of 

 kernels is correlated with average yield of cuhn per plant, as noted above. 

 In line 62, however, in two cases (with yield and with height of plant) 

 the limit is reached at the medium condition of crowding, beyond which 

 there is a decrease in correlation. 



From the foregoing it is seen that different degrees of crowding in the 

 row may produce striking differences in the correlation between certain 

 characters of plants in the same pure line of oats. 



The differences between the correlation coefficients when the same 

 characters are correlated m line 50 range from a minimum of .024 ± .009 

 to a maximum of .514 ± .053. In line 62 the range of differences is from 

 .021 ± .010 to .561 zb .046. The smallest range was obtained when 

 average yield of culm per plant was correlated with average number of 

 kernels per culm of plant in one case and with average number of spikelets 

 per cuhn of plant in the other case, and the greatest range m both cases 

 was obtained when the fonxier character was correlated \vdth average 

 number of kernels per spikelet of plant. 



Whenever the average weight of kernels per plant is correlated with 

 another character there is always a considerable difference produced in 

 correlation by crowding. In every case when this factor has been used, 

 the lowest correlation has been found in conditions of least crowding. 

 The greatest correlation has usually been found in the condition of greatest 

 crowding. In two cases, however, both in line 62, the greatest correlation 

 is found in the intermediate condition, these cases being, respectively^, 

 with yield and with height of plant. It wiU be remembered that the 

 crowding in series 118 is the greatest that is dealt with, being considerably 

 more than that in series 122 (Table 11, page 154). There appears, then, 

 to be a point beyond which certain correlations decrease. 



The correlation between yield and other characters is changed by more 

 than .050 in six cases out of nme in both lines dealt wdth: m line 50 the 

 six relative characters are height, length of head, weight of kernels, diameter 

 of straw, number of kernels per spikelet, and breaking strength of straw; 

 in Ime 62 weight of straw is substituted for length of head, the other 

 characters being the same. 



