Growth and variation in maize. 



119 



five equal parts there would be lU'S plants in each part. Yet for 

 present purposes it is obviously impossible to consider that part of a 

 plant falls in one quintile, and the remainder in another, at a given 

 measurement. In such cases the best division that can be effected is 

 to have 11 plants in each of four quintiles and 10 plants in the fifth. 

 In other cases two or more i)lants having the same measurement may 

 fall on the divitling line l)etween two quintiles. In such cases it is 

 necessary to include all of these plants in one or the other quintile. 

 In such cases as this the so-called quintiles do not have equal areas. 

 However, with the ungrouped data, there are very few cases where 



Fig. 5. Diagram to illustrate the quintile division of a frequency curve. 



more than one plant falls on a quintile boundary line. In series A and B, 

 each of wliich contain ,54 plants, the convention of allowing 11 plants 

 in each of the first four quintiles and 10 plants in the fifth has been 

 adopted. Thus the fifth quintile is , not equal to the others in area but 

 in this way the successive distributions are made comparable one with 

 another. In series C there are fifty -five plants so that each quintile 

 contains 11 plants in nearly every distribution. Wlierever it has been 

 possible allowance has been made for these defects in the quintile 

 distributions (cf. p. 123). 



In order to get the data before us for discussion it will be con- 

 venient to give tables showing for the three series the quintile distri- 



