212 



PHYSIOLOGIC GENETICS 



cannot be predicted a priori. However, experiments suggest, although they are really 

 too few to justify a generalization, that the response may be expected to continue for 

 about 20 generations and that the total improvement one might reasonably hope to 

 achieve by selection is probably of the order of 4 to 10 standard deviations. These con- 

 clusions, however, rest on a very tenuous factual basis. 



The complicating factor which lays any prediction of the response from a know- 

 ledge of the heritability open to doubt is this: several experiments in which selection 



Fig. 36. Two-way selection for six-week weight of mice 



The generation means are plotted against the cumulated selection differentials, and 

 linear regression lines are fitted to the points. The slopes of these lines, which estimate the 

 realized heritabilities, are: upwards, 0.175 + 0.016; downwards, 0.518 + 0.023. 341 



was made in both directions, that is, for an increase and also for a decrease of the 

 character, have shown that the rate of response in the two directions was not equal. 

 One such experiment is illustrated in figure 36. The heritability estimated from the 

 response, which may be called the realized heritability to distinguish it from estimates 

 based on the resemblance between relatives, was 17 per cent in the upward direction 

 and 52 per cent in the downward direction. The heritability, if it had been estimated 

 from the resemblance between relatives, would presumably have been about 35 per 

 cent, and, if a prediction had been made for upward selection, the response achieved 

 would have fallen short of expectation. Since the reasons for these asymmetrical 

 responses are not yet fully known, there is no means of predicting when they are likely 

 to occur. Consequently it would be unwise to put much faith in any prediction of a 



