200 Metrical Anribates of ^VlKat PlauLs 



§ 111. The Glume Length : Rachis-Length Ratio. 



As a pifliininary to further work, the correlation between glume- 

 length and rachis-h'ngth was evahiated from data obtained from the 

 iyi4 crop of the two pure lines. There were 15U plants of each pure hne 

 and the coefhcients of correlation (r) were: 



Polish = + 0-295 ± 0-050, 

 Kubanka = + 0-469 ± 0-043. 



It seemed not improbable that the lowness of the correlations was due 

 to the fact that the experimental plants were not alike in respect of 

 the total number of ears (or tillers) produced per plant.~ Consequently 

 tiie 1920 plants — the ones observed in the main investigation — were 

 classified as "one-ear." "two-ear," etc. plants and the correlation between 

 glume-length and rachis-length was evaluated for the separate classes. 

 In the case of every class, the main ear only of every plant was dealt 

 with in the correlations. Table I contains the correlation coefficients (r): 



Table I*. Coefficients of Correlation between Glume-Length at\d Rachis- 



Lenglh for one-car, ttco-ear, and three-ear Polish and Kubanka jdant^. 



(Only the main ear of the plant was observed.) 



* All tho cori'olatioiis in llic labk- are jjusitive. The unit of nieasuroMiont throughout' 

 is 1 mm. 



The values of the correlations are high, botli absolutely and in 

 relation to their own probable errors but even a correlation of unity 

 lietween two variables does not imply a constant ratio between them 

 unless tlie regression lines pass througii the origin (0 . 0). This fact 

 was well illustrated by the results obtained when the ratio glume- 

 length/rachis-length was evaluated for individual ])lants. The ratio 

 iluctuated as wildly as did the glume-length and to show this it is 

 necessary to give no more than the coefficients of variation (viz. 

 V = 100a/ 3/, where a = standard deviation and M = mean). These are 

 given in Table II. 



