168 



Pearl and Surface. 



to get less than four series out of one lumdred which would agree as 

 closely or more closely with the theoretical distribution than does the 

 observed distribution. Thus the agreement is exceedingly good. 



Table 24. 



Showing the frequency distribution of the theoretical mean 



quintile positions obtained from the values given in the last 



column of table 23. Also the observed percentages taken 



from table 16. 



This close agreement does not necessarily mean that there are 

 only two pairs of factors concerned in the growth of these plants. It 

 might be possible to get a series of weighings for three, four or more 

 factors wliich would give as good or better fit than that obtained above. 

 The breeding data necessary to determine the probable number of fac- 

 tors in tiiese plants is not available. For tliis reason it is not desir- 

 able to make any more complicated assumptions than are necessary to 

 give a reasonal)le explanation of the facts. 



Two alternative coucei)tions of the action of these factors may 

 be postulated. In the first place it may be conceived that all the pairs 

 of factors affect the growth of the plant throughout the season. On 

 the other hand it may be assumed that certain factors affect the growth 

 only during a certain stage of its growth. It is somewliat suggestive 

 to imagine that each of the four cycles into which the growth 

 curve can be divided is controlled by separate factors. The present 

 data are not sufficient to give evidence for or against such a sup- 

 position. 



