270 



J. ARTHUR HARRIS 



necessitated the examination of the experiments eight times a 

 d&y. Since as many as 8000 pots were sometimes involyed in 

 a single planting the labor of recording was exhausting in the 

 extreme, especially when the plants were germinating most rap- 

 idly. I have most heartily to thank Mr. Leo Macy for taking 

 two of the eight daily periods. 



With regard to the statistical technique, little need be said. 

 Ordinary correlation tables were formed and the coefficients cal- 

 culated by a convenient modification of the well known product 

 moment method.* 



II. PRESENTATION OF DATA 



The correlations, with the number of seeds upon which they 

 are based and their probable errors, are given in table 1.^ 



The 50 determinations are divided into 36 positive and 14 

 negative. Now, if there were no biological relationship what- 

 ever between the weight of the seed and the time required for 

 its germination, the result of an experiment would not be a cor- 

 relation of zero, but a low (positive or negative) value due to 

 the unavoidable errors of sampling or of experiment. There 

 being — as far as known — no complex of errors tending regularly 

 to throw the results in the positive or in the negative direction, 

 the outcome of a series of experiments should (within the limits 

 of fluctuation determined by the laws of chance) be equally 

 divided between positive and negative. Since on the assump- 

 tion of no physiological relation between the weight of the seed 

 and the time required for gennination, the chances of positive 



* Sheppard's correction was applied to the second moment for both time and 

 weight. 



The publication of the fifty large tables of data cannot reasonably be expected. 

 They have been carefully verified, and all constants calculated at least twice. 

 The correlations were then again checked within certain limits by the calculation 

 of the correlation ratios. Only the end results — the correlations — are published. 

 The physical constants for seed weight may be found in the second paper on 

 selective mortality. Those for time required for germination are omitted, since 

 I hope later to obtain constants for much more closely controlled conditions. 



' The key letters are those used in other papers on Phaseolus. They open 

 the way to considerable information concerning the seed. The final letter, 

 separated from the others by a dash, is the number of the germination test. 



