PERSONAL EQUATION IN BREEDING EXPERIMENTS. 3^ r 



certainly negligible. Observer Xo. VII I. cla--ified eighteen 

 kerne]-, .ill told, differently in the recount than in the original. 

 Thi i i- only about one per cent, of the total kernels handled, 



and cannot be regarded as a significant error. In both of these 

 case- VI. and YIII.) the discrepancies had to do entirely with 

 the color classification. With observer IX. this was not the 

 case. < >n both ear 8 and ear 9 she classed two kernel- as -\veet 

 in tin recount which she had originally called starchy. Alto- 

 gether this observer < lassificd thirty-five kernels differently in the 

 r< (oiini from \\li.it -In- did in the original. This ho\\e\vr repre- 

 sents .1 P lati\e error of a little less than two per cent. V> very 

 i stress ' ould be laid upon such an error. 



I roin i In- i.iM. s it will be noted that there was a marked and 

 nearly uniform tendency on the part of all three observers to 

 underestimate tin \ellows (both starchy and sweet) and to over- 

 stimate the \\hin -. in the recounts as compared with the orig- 

 inal-. It seems probable that the cause of this lie-, in part at 

 lea-l. in a fading of the yellow color during the lime -ince the 

 inn- \\eiv made. Thus it may be that kernel- which were 

 plainly vellou \\lien lir-t c, united are now white or very nearly 

 so. A further fact which would indicate that fading had occurred 

 i- found in ihe menial impressions of the observer-. All three 

 found the material di-tinctly more diflicult to classil\- \\hen re- 

 counted than whi'ii ori^inalK counted. One feels certain that 

 a part, ai le.i-i. of ihi- i- due t<> a change in the man-rial it-elf. 



l\eco-ni/in- tully the nieauerness of the material, the fact- 

 so far as the) ;^o seem to indicate clearly that tin- -ame < >\ sen er 

 i- likcl\ to !a ity the same material in about th. -ame \\a\- 

 ever) lime. It a particular kind of bias is sho\\ n in one count 

 it \\ill appear i--(iiiially unaltered in succes>i\e tri.il-. Thi- i- 

 I u i '1 i.il >1\ UK ire 1 1 IK- < !' i 'I i-ervers especially experienced in dealing 

 \\ iih the data of \ a riat ion than in the case of tlm-e \\ it In mt -nch 

 experii nee, though figures are lacking to demon-irate thi^. 



V. 



\\ . c< urn- next to the consideration of the -ecotid ijm-lion pro- 

 at the beginning (p. 341). Thi- u a-: " I >o ( - -, unat it 'inter- 

 --' in mai/e im])l\ gametic ' intermedia teness '? In 



