236 BULLETIN 82, UNITED STATES NATIONAL MUSEUM. 



by a progressive decrease in the height, resulting from a planing off, by resorption, 

 of the dorsal pole; this results, owing to the hemispherical outline, in a progressive 

 broadening of the dorsal pole, which at the same time becomes flatter, and in the 

 elimination, one by one of the rows of cirrus sockets, so that the centrodorsal finally 

 becomes a broad flat disk with a single, often more or less deficient, irregular mar- 

 ginal row of cirrus sockets; the process continuing further, this disk becomes thinner, 

 the cirri, one by one drop off, the sockets close up, and the disk then begins to 

 decrease in diameter, finally retreating within the circlet of radials and sinking so 

 that the dorsal surface of the radials and of the centrodorsal both rest in a common 

 plane (figs. 152, 154-156, p. 221). In extreme cases the radial margin of the disk 

 is resorbed and becomes more and more concave, the interradial portion always 

 reaching to the ends of the basal rays, until a small thin sharply stellate plate 

 results (figs. 157-159, p. 221). 



The suppression of the cirri follows exactly the same lines as their development; 

 they first disappear one by one from the midradial- region of the centrodorsal (fig. 

 531, pi. 2); an incipient stage of this process is frequently noticed in certain of the 

 Thalassometridas (compare figs. 196 and 198, p. 237); then the whole of the radial 

 region becomes affected, so that the cirri are reduced to the interradial portions, 

 occurring, singly or in pairs, just beneath the interradial angles of the calyx; this 

 condition is permanently retained in the adult of Comatula purpurea (fig. 79, p. 

 132), and is often noticed, as an individual variation, in many of the species in 

 which the cirri are normally lost in the adult, as for instance, in Comanihina schlegelii 

 and in Comaster belli; at last these interradial cirri begin to drop away, so that only 

 one cirrus is left in each interradial angle, and finally all the cirri are discarded. 



P. H. Carpenter notes that the ventral surface of the centrodorsal of Comanthus 

 parvicirra is 10-sided or nearly so (figs. 243-245, and 247-249, p. 251), and is not 

 marked by shallow radial depressions like those seen on the ventral surface of the 

 centrodorsal of Antedon (figs. 280, 281, 283, p. 261, and 593, pi. 15). The radial 

 areas rise very slightly from their peripheral to their central margins, and are 

 marked by various indistinct ridges and furrows. Their sides rise towards the 

 five interradial elevations which, though not very much raised above the general 

 surface of the plate, are nevertheless very distinct; for they are wide and marked 

 by shallow grooves which occupy the greater part of their width, so that the sim- 

 ple ridge, as seen in Hatlirometra (fig. 290, p. 262) and Leptometra (fig. 287, p. 262), 

 is here represented by the two sides of the groove which is cut out along its median 

 line. In Antedon these sides meet at a very short distance from the central end of 

 the groove, so as to obliterate it (fig. 285, pi. 261). In Comanthus parvicirra, how- 

 ever, they approach one another very gradually, and only just meet within the 

 margin of the plate (figs. 243-245, and 247-249, p. 251); but the ridge formed 

 by their fusion does not end here as in Antedon, for it is continued a short dis- 

 tance beyond the general surface of the plate so as to appear as a short process 

 extending outwards from the angle between two sides of its external pentagonal 

 margin. Consequently these five short processes appear on the dorsal aspect of 

 the plate, prolonging its angles outward. The grooves which are thus cut out 

 along the median line of the interradial elevations on the ventral surface of the 



