4'58 Bulletin 279. 



among the lower values where there is a high degree of divergence from 

 year to year. 



The marked changes shown by certain individuals clearly indicate 

 the advisability of a study of the population for several years, or at 

 least for as long a period as the improved type is to be employed in 

 a system of rotation, before selections are made. 



While these conclusions apply directly to timothy it is not unreason- 

 able to expect them to hold true for the majority of perennial grasses 

 and possibly for herbaceous perennials in general, though experimental 

 evidence is necessary to establish this assumption. 



It is hoped that the data presented are sufificient to emphasize the 

 importance of a long-continued study of a given population before 

 arriving at any definite conclusions regarding its condition, otherwise 

 the conclusions may be erroneous. For example, if the present study 

 of timothy had been confined to 1905 only, we would have concluded 

 that the distribution for height in this species was normal, while if 

 the data for 1907 had been taken the deductions would have been 

 entirely different; hence conclusions based on a single collection of 

 material do not give a true indication of the state of the population 

 except to show its condition for a given time and place. 



SUMMARY 



This investigation comprises a study of 3,505 plants for a period 

 of three years, the same individuals being under observation for the 

 entire period. 



A wide range of variation was found in all the characters discussed 

 in this paper. As indicated by the coefficient of variability, the greatest 

 variation in the characters studied statistically was in weight and the 

 lowest in height. 



With the exception of earliness, the mean values were highest in 1906. 

 This was due, to a large extent at least, to the favorable conditions 

 during that season for the growth of timothy. 



There was found to be practically no correlation of the characters 

 studied statistically except in the case of weight and height. 



Multimodal curves were found to be the rule rather than the excep- 

 tion. The changes in multimodality are attributed to fluctuations in 

 environmental conditions. 



Changes in the distribution resulting in an increase or decrease in 

 skewness are due largely to environmental influences, or to heterogeneity 

 arising from the different periods of maturity within the population. 



