No. 546] INFLUENCE OF STARVATION 



341 



Consider the standard deviations. 



As already noted, and as is clearly to be seen from the 

 graph, the standard deviations for the starved and fed 

 ancestral series show differences agreeing in general with 

 those seen in the means. This is to be expected, since 

 A and <r are generally closely correlated. For this rea- 

 son it is idle to discuss the influence of starvation or 

 feeding upon variability on the basis of the standard 

 deviation alone. 



Turning to the comparison series, we note that of the 

 28 differences, taken altogether, 17 are negative and 11 

 positive. The deviation from expectation is therefore 

 3 ± 1.79, and can not be asserted to be significant. 



Again taking Diff./jEaitf. > 3 as indicating differences 

 significant with regard to the errors of sampling, we note 

 that 17 cases out of 28 are statistically significant. Of 

 these 17 cases, 12 are negative and 5 are positive. Con- 

 sider averages as before : 



Navy, Within Strains, A = — .165 



Between Strains, A=— .053 



General Average, A — — .109 

 Ne Plus Ultra, A = — .595 



White Flageolet, A = + .145 



Again limiting comparisons to the strictly intra- 

 varietal, and segregating into direct and cross inter- 

 ramal comparisons, we find that of the ten direct com- 

 parisons possible in the four lots, six are negative and 

 four positive in the sign of the difference. Only four are 

 statistically significant, i. e., Diff./#di (f . > 3, and all are 

 negative. The average is — .221. Of the ten cross inter- 

 ramal comparisons, seven are negative and three are 

 positive in sign ; with regard to their probable error, eight 

 are significant ; of these five are negative and three are 

 positive. The mean for the series is — .181. 



Note the following points concerning the relative varia- 

 bilities as expressed by the coefficients of variation. 



Taken altogether, fifteen differences are negative and 

 thirteen are positive in sign. Accepting a difference of 



