258 D. F. JONES. 



pollen. In a total of 63,000 seeds produced in these experiments 

 only about 30 illegitimate seeds were observed. A larger number 

 of contaminations, however, would have the same appearance as 

 the legitimate seeds and so could not be detected but giving a 

 reasonable allowance to this source of unreliability the results 

 could not be appreciably altered. Moreover the error of this 

 kind is never all in one direction so that in the main it can be 

 disregarded. 



The accuracy with which the seeds have been classified de- 

 serves particular attention. Strains were selected to be used 

 which gave sharp differences between self-fertilized and cross- 

 fertilized seeds and in most cases separation was made very 

 satisfactorily. In a few mixtures there was some doubt and in 

 two experiments the seeds on the yellow endosperm plants could 

 not be distinguished. In these two cases the seeds were planted 

 and classification was made with the mature plants. Also in all 

 the other mixtures involving yellow and white endosperms a 

 sample was taken, after all the seeds of one class were mixed 

 together, and grown to determine the per cent, of error in sepa- 

 rating the seeds. Since the self-fertilized seeds give small inbred 

 plants, pure for yellow or white color, while the cross-fertilized 

 seeds produce large vigorous hybrids segregating into yellow 

 and white seeds, classification of the mature plants can be made 

 without the least doubt. However, it should be noted that the 

 better germination and greater vigor give the advantage to the 

 cross-fertilized classes in every case if there is any difference. 

 About 1 20 plants in each lot were grown and the per cent, of 

 error obtained was used to calculate the total amount of error 

 if all the seeds had been grown. Since the numbers of seed ran 

 up into the thousands it was impossible to grow all of them. The 

 figures showing the per cent, of errors found are arranged in 

 Table III. Only in one case is the number of wrongly classified 

 individuals above 3 per cent. In 25 out of 44 lots no faulty 

 separations were discovered. In the remaining cases the mis- 

 placed seeds tend to balance each other so that this source of 

 doubt can be largely removed. Calculating the figures without 

 regard to the error of classification gives practically the same 

 results as when this is taken into consideration. 



