194 A NATURALIST IN THE PACIFIC chap. 



that, as before described, marks the final setting of the seed-coats 

 and the ultimate maturation, as it is termed, of the seed. During 

 this shrinking process the kernel also shrinks within the seed-tests, 

 and cavities are thus produced within the seed-shell, on the 

 relative size of which depends the buoyancy of the seed, neither 

 the seed-shell nor the kernel possessing independent floating- 

 power. These cavities, as illustrated in the figures given in 

 Chapter XII, are of two kinds. That usually produced, being the 

 one that mainly determines the buoyancy, is a large central hollow 

 caused by the arching outwards of the cotyledons during the 

 shrinking process, such as is found also in the seeds of Entada 

 scandens, Mucuna urens, and some other Leguminous littoral 

 plants. With such seeds the kernel never rattles when the 

 seed is shaken, since the cotyledons lie in close contact with the 

 seed-shell. The other kind of cavity is produced between the seed- 

 shell and the kernel by the general or partial shrinking of the 

 kernel away from the shell, the cotyledons remaining in apposition, 

 as shown in the figures. When the shrinking away from the shell 

 is general, the kernel lies loose within the shell, and the seed 

 rattles when shaken. When the shrinking is partial the cavity is 

 on one side of the seed and the kernel is fixed. 



Professor Schimper (p. 164) remarks that the buoyant seeds of 

 Cjesalpinia bonducella all rattle when shaken, and that it is to the 

 incomplete filling of the seed-cavity, thus indicated by the loose 

 kernels, that the buoyancy of the seed is due. The rattling of the 

 kernel was, however, quite exceptional in the seeds handled by me, 

 even in the case of originally buoyant seeds kept for five 

 years. Seeds with loose kernels were, in fact, more frequent 

 with non-buoyant seeds than with those that floated. Thus in 

 Fiji I found that whilst with the buoyant seeds 17 to 20 

 per cent, had loose kernels, with non-buoyant seeds the pro- 

 portion was as much as 60 per cent. 



The normal cause of buoyancy is, therefore, a large inter- 

 cotyledonary cavity with the cotyledons lying in close contact with 

 the seed-shell ; but the two kinds of cavity may sometimes be 

 combined. Out of a number of buoyant seeds of Ca^salpinia 

 bonducella examined by mc, 80 per cent, owed their buoyancy 

 solely to a large central cavity (4 to 5 mm. across). In 6 per 

 cent, it was due solely to the shrinking of the kernel away from 1 

 the seed-shell ; whilst in 14 per cent, it was to be attributed partly 

 to a reduced central cavity (2 to 3 mm. wide), and partly to a space 

 outside the kernel. The only difference noted in the structure of the 



