DERIVATION OF THE FLORA AND FAUNA 361 



florescence; in some cases cauliflory is observed. A much more uncommon type 

 is the unbranched stem terminated by a tuft of large leaves (if entire = Grise- 

 BACH's Clavija form, j^j. I. ii), either pollacanthic with lateral inflorescences 

 or hapaxanthic when, after a number of years, a terminal inflorescence ends the 

 life of the individual; this type is the most unusual of all. All kinds have in com- 

 mon the very short internodes and the short-lived leaves, a self-evident condition 

 for the formation of a compact tuft. A reduction of the rosette tree to herbaceous 

 state would result in a compact caudex multiceps or a single basal rosette; an 

 hapaxanthic tree would become a therophyte. 



These growth forms are by no means restricted to island habitats but are 

 also found in all continents, they belong to many different genera and families 

 and are characteristic of such dicotylous families as Araliaceae, Caricaceae, Theo- 

 phrastaceae etc.; typical examples are found in Compositae (e.g. Espeletia, species 

 of Senecio), Epacridaceae [Dracophyllum], Lecythidaceae (Grias), Meliaceae (Ca- 

 rapa), Rutaceae [Spatkelia], Sapindaceae, and so forth. Among the monocotyle- 

 dons the vast majority of palms belong here, the Pandanaceae should of course 

 be mentioned, and well-known examples are scattered through the Liliiflorae (spe- 

 cies of Aloe, Cordyline, species of Yucca, Dracaena, Fourcroya etc.), Ravenala in 

 the Musaceae, and Piiya Rainiondii in Bromeliaceae. Hapaxanthic trees are few; 

 Krause [162) called attention to the interesting rutaceous Sohnreya excelsa Krause 

 from Amazonas. 



Thus, even if rosette-trees are more or less widely distributed over the globe 

 in warmer regions, it is a fact that they are a particularly conspicuous feature in 

 island floras. In the Pacific they are very plentiful in the Hawaiian Islands, espe- 

 cially among the Compositae and Lobeliaceae, also the hapaxanthic type repre- 

 sented; they occur in the Galapagos Islands (Grisebach II. 512), we find many 

 in New Caledonia, they constitute a large proportion of the poor Juan Fernandez 

 flora (see Chapter IX), and they are scattered over Oceania. In the Atlantic they are 

 numerous in Macaronesia (e.g. Campa7iula Vidalii on the Azores, MusscJiia on Ma- 

 deira, species of Aeoniuni, Dracaena, EcJiiiwi, Melanoselinum, Sonchus, Sinapide7i- 

 dru7n, etc.). I suppose that some are found also on the islands of the Indian Ocean, 

 but I have no reliable information. I may be entirely wrong, but I have the idea 

 that this is an old-fashioned growth form, a relict element in island floras, and that 

 the high volcanic islands of the Pacific and the Atlantic, which are so like each 

 other in geology and topography, are, in their present shape, of approximately the 

 same age and date back to a period of land submergence and great volcanic 

 activity. Perhaps also the "inland islands", the high volcanic mountains of Africa, 

 deserve to be mentioned in this connection, famous as they are for their magni- 

 ficent tree Lobelias and Senecios. 



Rosette trees seem to favour open, sunny situations; this is certainly the case 

 in Juan Fernandez, where they are definitely adapted to such habitats. Single 

 specimens of a few species are sometimes found growing in the shade, with the 

 result that the internodes become much longer, the dense tuft dissolves and the 

 formation of flowers is suppressed. ScillMPER believed that they are adapted to 

 a very windy climate and fit to withstand the pressure of strong winds better 



