THE ENDOSPERM 171 



even of fusion. Ordinarily it must be dependent upon polar 

 fusion, and in some cases upon triple fusion, as indicated by 

 the behavior in Naias cited above ; but in the failure of these, 

 other conditions may cause nuclear division and the formation 

 of endosperm. 



While in the majority of plants the endosperm may be re- 

 garded as fully developed, either to remain as a permanent 

 tissue of the seed or to be more or less resorbed by the growing 

 embryo, there are certain plants in which it is abortive or even 

 suppressed. It consists of only a few scattered nuclei, or at 

 most of a parietal layer of free nuclei, in Naiadaceae, most Alis- 

 maceae, Juncagineae, and Hydrocharitaceae, all of which belong 

 to the Helobiales among Monocotyledons. The tendency of the 

 endosperm to become abortive in this particular alliance is evi- 

 dently very strong, although, as Hall 50 has shown in Limno- 

 charis, the endosperm may finally develop and become packed 

 about the embryo. With the exception of the Helobiales, disap- 

 pearance of the endosperm seems to be very rare, having been 

 reported in Tropaeolum and Trapa; and among the Orchida- 

 ceae the endosperm seems to be entirely suppressed, the polar 

 nuclei, as a rule, neither fusing nor dividing. 



Humphrey 17 has called attention to what he calls a pro- 

 gressive series in the development of the endosperm among the 

 Scitamineae, but which seems to be best interpreted as a retro- 

 gressive series. In the Musaceae an abundant starch-bearing 

 endosperm either fills the sac (Ileliconia) or nearly so (Stre- 

 litzia), the peripheral cells often forming an aleurone layer; 

 in Zingiberaceae (Costus) the endosperm is several layers thick 

 in the lower part of the sac and only aleurone-bearing ; in Can- 

 naceae (C. indica) the endosperm is a single aleurone-bearing 

 layer lining the sac; while in Marantaceae (Thalia dealbata) 

 the endosperm is probably not represented at all in the mature 

 seed. 



Strasburger 4 has called attention to the two general meth- 

 ods of endosperm formation among Angiosperms. In the ma- 

 jority of plants observed it begins with free nuclear division ; 

 but in many cases, chiefly among Dicotyledons, the first division 

 of the primary endosperm nucleus is accompanied by a wall 

 dividing the sac into two chambers (Fig. 74). While these 

 two methods of initiating endosperm formation are quite dif- 



