DEFINITE DIFFERENCES 85 



characteristics as those which were exhibited by the 

 two original parental strains. The only difficulty in the 

 way of invariably distinguishing the two original kinds, 

 after their segregation in the offspring of the cross, lies 

 in the fact that the smallest seeds of the large type 

 may be smaller than the largest seeds of the smaller 

 strain, and this is a difficulty which applies equally 

 to the original strains before crossing, as well as to the 

 case of the evening primrose fruits just mentioned. 



Now, it is clear that if we mixed together the seeds 

 of several different races of peas in the proper pro- 

 portions, the result might lead to a normal distribution 

 of the kind presently to be described. The several 

 races, however, would none the less be perfectly distinct, 

 even though we could not separate the individual 

 seeds belonging to each by any direct method.* Such a 

 mixture of races would constitute a decided pitfall for 

 the unwary statistician, and it is well to remember 

 that, after even the most elaborate mathematical 

 analysis* the final result cannot be clothed with any 

 greater amount of certainty than the facts from which 

 the calculation set out. Those who have made a large 

 expenditure of intellectual effort in such processes have* 

 unfortunately* a natural tendency to overlook this 

 elementary fact. 



Prior to the application of statistical methods to a 

 particular case of normal variation a number of pre- 

 liminary processes have to be gone through. 



* It would generally be possible to decide which strain a 

 particular seed belonged to by sowing it and observing the 

 variation of its offspring. 



