SELECTION AND V A K I A HI I IT Y 



with flics, selection for change in chaeta number not merely robbed, 

 the stock of its fertility: it also changed the pattern of pigmentation, 

 the mating discrimination and the number of spermathecac. This 

 last change was especially marked : instead of the normal two, as 

 in the parents, selected females had from none to five spermathecac. 



Correlated responses to selection such as these may be due, in part, 

 to one gene affecting several characters, the pleiotropy which we 

 discussed earlier. But there is also another agency at work. Chaeta 

 numbers showing the effects of selection and at first associated with 

 a lower fertility have, later in the same experiment, become asso- 

 ciated with a higher fertility. The two effects must therefore be 

 due to different genes, and the cause of the correlated response must 

 be linkage. Crossing-over within a chromosome will bring about 

 recombination, not merely within the polygenic combination 

 affecting the character in which we are interested, but also within 

 the other polygenic combinations which are intermingled with it 

 along the chromosome. So in selecting for changed combinations 

 of one set of genes, change in the combinations of sets affecting 

 other characters, for which no selection is practised, will also be 

 brought about (Fig. 75). And if these changes cause a sufficiently 

 large unbalance, their consequences will appear as an alteration in 

 the expression of the character they control; an alteration for 

 which there is no direct selective cause. When this has happened, 

 however, continued recombination of the gene sets can eventually 

 lead to a reassociation of the characters, of the kind we see when 

 the flies with changed chaeta number regain their fertility. Even 

 pleiotropy may, as we saw, merely express complete linkage. But, 

 with pleiotropy, the reassociation of characters would have been out 

 of the question. 



Although the response must go in a fixed direction for the 

 character on which selection is primarily acting, the correlated 

 responses in other characters need not do so. The correlated changes 

 will therefore often be deleterious. Correlated response should in 

 consequence act as a brake on the primary response to selection, 

 both in nature and in experiment. Decline and death is characteristic 

 of selection lines in experiment and bears witness to the truth of 

 this principle. The better adjustment in one character must be paid 

 for by the worse adjustment, at least for a time, of others. The 



298 



