no A NATURALIST IN THE PACIFIC CHAP. 



in species of Cuscuta. When the seed floats, as with Ipomea 

 pes caprae, I. glaberrima, &c., the unoccupied space is relatively 

 large ; and when, as with I. bona nox and I. turpethum, the 

 behaviour of the seeds is irregular, some floating, and others 

 sinking, a corresponding variation exists in the extent to which 

 the seed-cavity is filled. This applies also to the irregular 

 behaviour of the seeds of Ipomea peltata and of Argyreia 

 tilisefolia. A singular instance is afforded by the seeds of 

 Ipomea insularis, collected by me in Fiji and Hawaii. Those 

 from Fiji were incompletely filled, and consequently buoyant. 

 Those from Hawaii were more densely packed and sank. 

 . . . The three British species of Convolvulus illustrate the 

 same principle, namely, C. arvensis, with non-buoyant seeds ; 

 C. soldanella, with buoyant seeds ; and C. sepium, with seeds 

 irregular in behaviour. 



In the case of plants of the Corivolvulaceae, possessing buoyant 

 seeds, there is always evidence of marked shrinking of the seed- 

 contents before the final setting and hardening of the seed-coats. 

 The embryo often appears shrivelled and dried up, and is almost 

 brittle, so that large spaces are produced in the seed-cavity. If 

 we partly divide such a seed and place it in water, the embryo 

 absorbs water rapidly, and within an hour is soft, healthy-looking, 

 and much swollen, the interspaces being filled with a jelly-like 

 mucilage. It is therefore evident that absolute impermeability 

 of the seed-coats is essential for the successful transport by sea- 

 currents of the floating seed ; and we can only suppose that the 

 shrinking of the seed-contents takes place before the final setting 

 of the tests. That with the buoyant seeds the coats are quite 

 waterproof was illustrated in many of my experiments where, after 

 a period of flotation covering several months, and sometimes a 

 year or more, the seed-contents were still quite dry and shrunken. 

 The limit of buoyancy, as I have shown in Chapter IX., depends 

 on an attempt at germination on the part of the floating seed, 

 which then absorbs water, softens, swells, and sinks. 



It is, therefore, not a matter of surprise that non-buoyant seeds 

 of the Convolvulacese do not gain floating power after prolonged 

 drying of many months. It is also to be expected that, as we find 

 in Fiji, when a characteristic shore-species with buoyant seeds like 

 Ipomea pes caprae extends far inland, the seeds retain their floating 

 powers. Seed-buoyancy of this description is, on the face of it, 

 purely mechanical. 



Another type of the buoyant seeds of the first group is 



