238 BULLETIN 82, UNITED STATES NATIONAL MUSEUM. 



this is first exposed by the removal of the centrodorsal from the dorsal surface of 

 tho radial p(>nta<;;on which rests upon it, five white rays are visible on a dark back- 

 ground. Uidoss the plate is iuunediatoly removed from the alkaline solution used 

 to effect its separation this distmction in color between the radial and the inter- 

 radial portions of its ventral surface rapidly disappears, owing to the destruction 

 of the pigments contained in the former. 



The development of these basal grooves is not only different in different speci- 

 mens of the various species of comatulids (figs. 229-234, p. 247, 235-242, p. 249, and 

 243-249, p. 251), especially among the Comasteridaj, but it varies to a certain extent 

 in the same individual (fig. 248, p. 251). Sometimes one or more of the basal grooves 

 may rapidly diminish in width and end well withm the periphery of the centro- 

 dorsal (figs. 243, 248, p. 251). Tliey may gradually dimmish (fig. 259, p. 255), or, 

 more rarely, gradually increase (fig. 229, p. 247), from the center to the periphery, 

 or the sides may bo quite parallel (figs. 266, p. 257, and 268, 270, p. 259) ; but usually 

 they increase slightly m diameter for a shorter or longer <listance, tapering off 

 gradually from tliLs pomt toward the periphery, thus having, as expressed by 

 Carpenter, a leaflikc appearance (figs. 244-249, p. 251). 



Except for very small fonns such as Comatilia iridometriformis, Comanthus 

 hennetti and C. pinguis (figs. 171-174, p. 231) are the only species in the Comasteridffi 

 in which the centrodorsal develops throughout life and shows but little trace of 

 progressive specialization in the adult stage; m most of the other species the centro- 

 dorsal is discoidal (figs. 160-162, p. 223, 163, p. 225, and 181, 182, p. 233), though it 

 may be rather thick, with a broad flat polar area and two or three marginal rows 

 of cirrus sockets bearmg functional cirri which in some cases, as in Comanthus 

 parvicirra, may be disproportionately small (figs. 160, p. 223, and 182, p. 233) or, as in 

 C. trichoptera, disproportionately slender and thm (fig. 330, p. 281). A number of 

 species commonly have the centrodorsal a very th'ui disk with a single row of cirrus 

 sockets which may be regularly (as in Comatula purpurea) or irregularly (as in 

 Comanthus parvicirra) mcomplete (figs. 79, p. 132, and 182, p. 233); others when 

 adult usually have the centrodorsal without cii-ri and pentagonal or stellate, but 

 frequently with one or two or even mor<> perfect cirri remaining, as Comanthus 

 annulata, Comanthina schlegelii or Comaster helli (fig. 182, p. 233); and a consider- 

 able numb<'r idways when adult have the centrodorsal smaU and stellate with never 

 a trace of cirri, as Comatula rotalaria, Comaster typica, Capillaster macrohrachius, 

 an<l Comantheria polycnemis ifig&. 153-159, p. 221, 162, p. 223, 164, p. 227, and 166, 

 168-170, p. 229). 



When very young, all the species of the C'omastcridse have centrodorsaJs exactly 

 like those of Antedon, and in all species alike they develop in exactly the same way. 

 The difference in the centrodorsals of the adults is therefore solely a <lifference in 

 comparative development, demonstrating a fundamental unity, and not a difference 

 in structure, implying a phylogenetic divergence. For mstance, the large hemi- 

 spherical centrodoi-suls of Comanthus hennetti or C. pinguis iiro merely controdoi-sals 

 of the most primitive comasterid type which, though greatly increased in size, 

 are not ontogenetically different from the centrodorsals of the early post-penta- 

 crinoitl stage; the centrodoi-sals of Comactinia or t)f Cojniss^ia, discoidal, with one 



