18 EXPERIMENTS WITH DROSOPHILA AMPELOPHILA. 



there been any apparent tendency to settle down to any definite type.* 

 The "center of disturbance " has remained in the distal portion of the 

 marginal cell closely related to the second longitudinal vein. Next to 

 the second, the third vein has been the most affected, then the fifth; 

 but I have failed to fix the abnormalities in these veins. They are, 

 apparently, all the effects of the disturbing factor or factors, centered 

 on the second longitudinal vein in the marginal cell. 



The hundreds of families studied showed that it is impossible to pre- 

 dict, from the character of the ancestors, what the form of the abnor- 

 mality will be in the offspring. The most that one can do is to give an 

 approximate estimate of the percentage of abnormal individuals and a 

 still less exact prediction of the average intensity of the abnormalities. 



THE BEARING OF THESE DATA UPON PROPOSED LAWS OF HEREDITY. 



In my former paper (1907) I considered that normal venation is more 

 or less dominant over abnormal in the Mendelian sense. Such was the 

 case in the early part of the work, although, as was pointed out, it was 

 the spirit only and not the letter of the law which was followed. When 

 a normal fly, having normal ancestors, was crossed with an abnormal 

 one, practically all the offspring were normal. The abnormalities which 

 did appear were slight, but there was no doubt about their presence. 

 Matings 318 to 322 (see table 36) illustrate such cases. The offspring 

 of matings 347 to 353 are second-generation hybrids from such a cross. 

 They show a condition not very divergent from the Mendelian expecta- 

 tion. 



Since the number of offspring in most of the families considered here 

 is large, the Galtonian formula can be tested in single families, and it is 

 evidently not at all in accord with the data. Neither is Pearson's modi- 

 fication of it. The fact that normal X abnormal gave, in large families, 

 practically all normal completely negatives for these data all theories 

 which are founded on the hypothesis of equipotency of the two parental 

 characters. 



On the other hand, while the results of certain matings accord with 

 Mendelian expectation, the fit is far from good in the majority even in 

 the early generations. For instance, we have seen that neither normal 

 nor abnormal breeds true. A Mendelian recessive would be expected to 

 do so ; therefore we can not consider either normality or abnormality 

 to be Mendelian unit-characters in that sense. 



*Here again (see p. 10) the similarity to the experience of breeders of spotted ani- 

 mals is interesting. Castle (1905), for example, found that "one can by selection 

 progress in either direction through this series of changes, either increasing or de- 

 creasing the number and extent of the pigment patches, but it is impossible without 

 long-continued selection to fix the color-pattern at any particular stage in the series; 

 perhaps it is wholly impossible to do so, as Cue'not (1904) asserts on the basis of his 

 studies on mice, but this I very much doubt." 



