486 THE VOYAGE OF H.M.S. CHALLENGER. 



' This becomes at once very evident when we consider the two great subdivisions of 

 Lyssaciua and Dictyonina, which, according to most mo'dern investigators of sponges, 

 and in my opinion also, may be recognised in the Hexactinellida. The question at 

 once arises whether these two divisions, regarded in classification as of approximately 

 equal importance, are to be expressed by the forking of a common stem, or are not 

 rather to be considered as the continuous but successive divisions of one ramified tree. 

 In the first case we must suppose that the two divisions with divergent forms originated 

 almost or quite contemporaneously from a common ancestral form, and that each 

 developing by itself in a special direction gradually exhibited the distinctive characters 

 of the modern forms. In the second case we have to suppose that the ancestors of the 

 hio-her division must at first have had the characters of the lower, and have been 

 systematically included within it, — we must suppose, that is to say, that the one 

 division sprang from the other. 



But before I proceed to the discussion of this and similar questions, I shall briefly 

 review the relative opinions of previous investigators. 



In his researches on Hexactinellida,^ published in 1875, Marshall says: — " The state 

 of skeletal coalescence " (in which the axial canals of the framework beams are said to 

 form one connected anastomosing system) " I regard as phylogenetically oldest, as 

 that from which have developed the Hexactinellida with free siliceous elements, and 

 especially those with predominantly hexradiate spicules, which may be regarded as 

 simply inherited. By adaptation we account for the large series of frequently very 

 beautiful forms, for which Bowerbank has invented such an elaborate nomenclature. 

 The third state, that of fusion, appears to arise in diS'erent ways : ( 1 ) by the simple 

 union of the ensheathing substance of two adjacent spicules ; (2) by lamella-like 

 structures which extend like bridges between two adjacent but not directly apposed 

 spicules ; (3) by the development of lamellar layers of flinty material round two 

 parallel and adjacent spicules." 



In the memoir entitled Ideen fiber die Verwandtschaftsverhaltnisse der Hexact- 

 inelliden,- Marshall has more definitely expressed his conclusions as to the phylogeny 

 of the Hexactinellida. Starting from a Chalynthus-like ancestral form without skeleton, 

 he regards it as probable that in the w^all of this simple sack somewhat firmer longi- 

 tudinal, circular, and radial strands of " hardened protoplasmic material " were developed, 

 intersecting at right angles, and forming a connected fibrous framework with square or 

 cubical meshes. "The next form," Marshall continues {loc. cit, p. 119), "is thus a 

 sponge with simple connected siliceous lattice work, in which the central canals are 

 also connected, and which has not yet acquired any functionally important free spicules. 

 From such a simple Protohexactinellid there have developed, on the one hand, forms 

 like Sclerothamniis, with single free spicules, and, on the other hand, forms in which 



1 Zeitschr. f. iviss. Zool., Bd. xxv., Supplement. - O/j. cit., Bd. xxvii. p. Ill 



