ADDITIONAL EVIDENCE 129 



to explain this low water-percentage ; but at all events this 

 fact shows how necessary it was to avoid comparing seeds 

 of different families when connecting impermeability with 

 diminished water-contents. 



I come now to the additional evidence in support of the Further 

 principle typified by Guilandina honducella in Chapter IV, that show that in 

 the seed-coverings of impermeable seeds possess the same s'JJ'ds'Sfe^'^'^ 

 quality of ultra-dryness as the kernel, though often in a ^^^gg^J^^^.^ 

 somewhat diminished degree, and the same quality of supply- same quaHty 

 ing the deficiency by absorbing water from the air, the nessas'thT 

 larger water-percentage of the coats being usually associated kernel, 

 with a diminished absorptive capacity of the freshly exposed 

 material. 



Most of my results are given in the table a few pages 

 later ; but I will confine the discussion as before to legu- 

 minous impermeable seeds. All the kinds of seeds there 

 tested possess this quality of ultra-dryness for the coats as 

 well as the kernel, though the presence of oil in the kernel 

 of Adenanthera pavonina somewhat alters the regime. Most of 

 the results represent the average of three or four or more 

 experiments, the absorptive capacity in air and the water- 

 percentage being determined independently. In spite of 

 possible disturbing effects, due to variation in the seeds 

 and in the atmospheric conditions, the data thus obtained 

 go fairly well together. But in two cases, those of Guilandina 

 honducella and Entada scandens^ this disturbing influence was 

 removed by a simple expedient ; and these experiments have 

 been specially added to the others, since they are not only 

 the most critical but the most decisive. With Guilandina 

 honducella the coats and kernel of each seed were divided 

 between two samples, so that the water-percentage and the 

 absorptive capacity in air were simultaneously determined 

 from similar examples. With Entada scandens the two 

 samples of the seed-coverings and the two samples of the 

 kernel were obtained from one large seed weighing about 

 500 grains. They gave the following results for the water- 



9 



