V FLUCTUATION 99 



Total number of flowers per plant Mean, 115 P. E. ± 29 per cent, 

 bolls „ „ „ 50 P. £.+34 „ 



Correlation between total flowers and bolls : r=0"85. 



(Imperfection of correlation is mainly due to shedding.) 



Percentage of retained flowers : Mean, 40 '6 ; P. E. ±7 "5 per cent. 



These figures demonstrate the manner in which errors 

 accumulate during field experiments. Comparison of two 

 rows or plots of the same variety becomes more and more 

 erroneous as the plants grow older ; comparison by the 

 number of flowers has a lower P.E. than comparison by 

 the number of bolls ; this in its turn has a lower P.E. 

 than comparison by weight of seed-cotton picked ; this 

 again is more precise than comparison by ginned lint. 

 Fresh sources of error creep in at every stage, and the fact 

 that field plots have a P.E. oi ±6 per cent., at least, 

 becomes quite comprehensible. 



The boll. — The three chief characteristics of the boll 

 are its diameter, the number of loculi, and the shape. The 

 first, and last are determined from three full-grown 

 bolls per plant, measured with parallel-jaw callipers, the 

 form being expressed as a diameter in percentage of length. 

 The number of loculi is determined by examination of not 

 less than twenty bolls of one kind on each plant ; thus 

 forty bolls may have to be examined to obtain the critical 

 number, if half are trilocular and half quadrilocular ; 

 such a plant receives the designating formula of 3 "5. 

 Plants whose bolls are all trilocular are designated 3'0 ; if 

 all quadrilocular, 4'0, &c. For certain families the critical 

 number has been raised to forty bolls of a kind {e.g., 

 Fig. 63). 



Specimen data for these characters in Egyptian families 

 are : — - 



The fluctuation of the last character is low, presumably 

 because the number of loculi is diff'erentiated at an early 



H 2 



