4^o Bulletin 279. 



nearly every instance these changes were in a definite direction, either 

 positive or negative, from lower to higher values. 



Changes in the coefficients of correlation have been observed in other 

 plants. MacLeod (1899) and Shull (1902), working with Ficaria ranun- 

 culoides and Aster prenanthoides respectively, found that these constants 

 changed as the season advanced. In perennial plants we would expect 

 similar changes to occur from year to year. 



The fact that in timothy these constants are higher for 1907 than for 

 1905 and 1906 might at first thought lead to the conclusion that there 

 was an innate tendency toward a more perfect correlation, either positive 

 or negative, as the plants grew older, but this is not necessarily true, 

 for if, as has already been pointed out, changes in skewness may result 

 from changed environmental conditions it is not unreasonable to expect 

 changes in the coefficient of correlation to result from the same causes. 

 If the curves for height and weight be compared, it will be seen that 

 the changes in skewness are in opposite directions. These changes in 

 the distribution will of necessity affect the coefficient of correlation. 



Place variation as a factor in selection. — It has already been pointed 

 out that in the material under observation there has been a continual 

 shifting of the population from year to year as the result of place variation 

 and that these changes did not affect all the individuals in the same 

 manner. 



In view of these facts the question arises as to whether these seasonal 

 fluctuations are of sufficient importance to be taken into consideration 

 in the practice of selection. Will our knowledge of the condition of 

 an individual for a given season hold true for any other season or must 

 we keep the individual under observation for a long period of time 

 before we can determine its value? 



In order to answer this question, if possible, tables have been pre- 

 pared showing the correlations between weight for .T905 and 1906, 1906 

 and 1907, and 1905 and 1907 (Figs. 145, 146 and 147). Similar tables 

 have also been prepared for height (Figs. 148, 149 and 150). 



Since the differences exhibited by the population from year to year 

 are attributed largely to place variation, the numerical expression of 

 these differences may be designated as the coefficient of place variation, 

 a high coefficient denoting a low degree and a low coefficient a high 

 degree of place variation. 



