1 92 STUDIES IN SEEDS AND FRUITS 



seeds of other orders. The ratio may have a generic value in 

 some cases, as in Erythrina, Guilandina, Phaseolus, and Hibiscus ; 

 and in others it may vary considerably within the limits of a 

 genus, as in Anona and Ipomcea^ so that as possible disturbing 

 influences the one may be set against the other. Whilst 

 ranging widely in some orders, as in Leguminosae, it may 

 be comparatively uniform in others, as in Malvaceae and 

 Sapotaceae. Size has little or nothing to do with the ratio, 

 excluding very small seeds less than one-tenth of a grain, which 

 are not here discussed. Looking down the lists, we find large 

 and small seeds frequently associated on account of the 

 similarity in their ratios. Thus the seeds of Canna indica and 

 of Mammea americana have similar ratios, although about 220 

 seeds of the first will be required to make the weight of a 

 single seed of the second. So also, if we compare the imper- 

 meable seeds of the two species of Entada^ we find that their 

 ratios are not far apart, notwithstanding that at least 80 seeds 

 of E. polystachya are required to weigh down a single seed of 

 E. scandens. Then, again, with the two sapotaceous plants, 

 Lucuma mammosa and Achras Sapota, we notice that the propor- 

 tional weights of the seed-coats are nearly the same, although 

 the difference between the weights of a single seed are as 220 

 to 9*5. Size, of course, goes with weight in all these cases. 

 The con- In resting seeds the seed-coat ratio is sufficiently constant 



seed-coat* C to f rm a character for the species, though, as we have seen 

 ratio .within above, it may Vary considerably within the limits of a genus. 



the limits of ' t ' ' j i i r r 



a species. Its constancy tor a species is illustrated in the results tor a tew 

 plants given below ; but my data do not often lend themselves 

 for such a comparison, since in determining this relation it was 

 my usual practice to employ a number of seeds at the same time. 



Guilandina bonducella, number of seeds tested 22, range of seed-coat 

 ratios 53 to 64. 



Guilandina bonduc^ number of seeds tested 8, range of seed-coat 

 ratios 44 to 52. 



Entada scandens^ number of seeds tested 5, range of seed-coat ratios 

 37'7 to 40'5. 



