268 



J. ARTHUR HARRIS 



able, however, to emphasize the fact that the problem of deter- 

 mining these relationships is not a simple one and that in con- 

 sequence the measures here gi\'en are very rough prehminary 

 approximations merely. The results are, indeed, qualitative 

 rather than quantitative, although they are given numerical 

 expression. 



The difficulties to which this condition is due are essentially 

 two-fold: inabilitj' to control closely enough the physical con- 

 ditions to which the seeds were subjected in germination and 

 inability to measure with accuracy the time required for develop- 

 ment up to any selected stage. 



These obstacles are of the kind which can to a. considerable 

 extent be overcome in more refined work. If one inquires why 

 these more precise methods were not applied, the answer is sim- 

 ple. Practicability is a primary requisite in any investigation. 

 By working with large numbers, a general solution of the prob- 

 lem could be obtained incidentally in the carrying out of other 

 studies b}^ an additional amount of labor which although exces- 

 sive was not prohibitive. Now that the major lines are blocked 

 •out greater refinement in any detail may be profitably sought. 



The chief experimental difficulties lie in the control of tem- 

 perature and substratum moisture. These, especially tempera- 

 ture, differ considerabh' from experiment to experiment, but I 

 believe such fluctuations have no great effect upon the correla- 

 tion coefficient (except possibly as they may be influenced by 

 differences in mortality) for, in no case were two or more lots 

 of material combined. 



Probably a greater source of error in the correlation coefficients 

 is the variation in the soil moisture of the pots of an individual 

 experiment. Notwithstanding great care, I fear that this fac- 

 tor was far from satisfactorily controlled. It cannot disturb 

 the general trend of the results for the seeds were individually 

 labeled, thoroughly shuffled and scattered quite at random 

 through the house. Thus no particular series or type of seed 

 was subjected to better or worse conditions than others. It 

 must, however (besides its possible influence through variations 

 in the percentage of mortalit}') , influence the variability in time 



