86 THE AMERICAN NATURALIST [Vol. XL VIII 



seven positive differences are at least 2.5 times their prob- 

 able error; four of them are over five times their prob- 

 able error. The mortality is therefore almost certainly 

 selective, with a tendency to leave the surviving popula- 

 tion with seeds distinctly heavier on the average than 

 those which were planted. On the other hand, there are 

 the three cases in which the seeds which produced plant- 

 lets were on the average lighter than those which failed 

 to germinate. One of these differences is only 2.2 times 

 its probable error, and so perhaps not statistically trust- 

 worthy. Of the other two, one is over 5.5 times and the 

 other nearly 4 times its probable error. There can be 

 little doubt that in at least one of these cases there is a 

 tendency for the lighter seeds to show a viability greater 

 than that of the heavier. In garden beans, too, strong 

 evidences of differences between strains in this regard 

 have been pointed out. 



The interpretation of the variabilities offers greater 

 difficulties than does that of the means. More data and 

 more refined methods of analysis are necessary for a final 

 solution of the problem. It appears, however, that in 

 seven of the ten series the variability of the seeds which 

 survived is less than that of those which failed. This is 

 true whether absolute variability as measured by the 

 standard deviation or relative variability as expressed 

 by the coefficient of variation be used in the comparison. 



As far as these data go, therefore, they are in general 

 accord with those for Phaseolus. In both of these Legu- 

 minosse the mortality which occurs before germination is 

 not random but differential. But in both cases, and espe- 

 cially in Pisum where the seeds used are of commercial, 

 not pedigreed, stock and number as yet only about 10,000, 

 far wider series of experiments and much refinement of 

 methods of analysis are necessary to establish fully the 

 nature and the immediate (physical or chemical) cause of 

 this selective death rate. 



Cold Spring Harbor, N, T., 



