240 BULLETIN 82, UNITED STATES NATIONAL MUSEUM. 



or at most two rows of cirri, represent an advanced stage of comasterid centrodorsal 

 development (though the arm structure in these two genera is much less spt>cializcd 

 than in Comanthus), while tlie stellate centrodorsals of Comaster typica or of Ca-pil- 

 laster macrohrachius are of the most specialized t3-pe. 



It is a curious fact, though one finding uinumerable parallels, that ui all of the 

 genera of the Comastcridas the centrodorsal starts, so far as wo know, from exactly 

 the same condition in the young, and develops along exactly the same Imes; in 

 Comanthus all the stages are found in the adults of the various species, but hi the 

 other genera the sum of the species taken together covers only a comparatively 

 small part of the entu-e developmental line. 



If we take the Ime of development of the comasterid centrodorsal and divide 

 it into four parts, markhig the division pouits A, B, C, and D, A being the Comanthus 

 bennetti type (figs. 171, 174, p. 231) (under which, in effect, all the very small species 

 such as Comatilia iridometriforrrds are include*!, as would be expected); D the small 

 stellate Comaster typica type (figs. 157-159, p. 221), B (figs. 146-148, p. 220) and C 

 (figs. 160, 161, p. 223, and 163, p. 225) intermediates, wo find that Comactinia, Lep- 

 tonemaster, Neocomaiella, Comissia and Nemaster all fall between B and C; Cominia 

 falls hi B; Comatella extends from A to C; Comanthus from A to D; Capillaster from 

 B to D; and Comatula an«l Comaster from C to D. Pal^ocomateUa is essentially 

 like Neocomatella, though it exhibits a tendency toward a columnar arrangement of 

 the cirrus sockets. 



It is interesting to note that, except for the very small species of Comatilm 

 and MicrocomatuJa, which are scarcely to be considered in this connection, the 

 West Indian comasterids and the comasterids occurring on the Atlantic coasts of 

 Africa are restricted in regard to the development of the centrodorsal to the interval 

 B-C, whereas those of tlie central East Indian region and of the more northern 

 l)ortions of Australia range from A to D with the emphasis, in Australia, on the D; 

 of other regions, the northwest and southeast African comasterids range only 

 between B and C like the West Indian, while the southern Japanese range from 

 A to C. 



It is evident from the tabulation given above that the comasterid genera which 

 show the most specialization in other characters have also the most specialization 

 in their centrodorsals, and also that extreme specialization, either in the direction 

 of a retention of a larval type of centrodorsal, or of very great reduction in the 

 size of that plate, is confined to the areas where extreme specialization in otlier 

 characters occurs. 



In the Innatantes the central jdate is not com])arable to the centrodorsal of 

 the other comatulids (figs. 505, 572, ])1. 7) ; I believe it to be tlie homologue of the 

 terminal stem plate plus all tlie columnals of the other comatulids. I am led to this 

 belief from the following circumstances: It lies in the body wall flush with the infra- 

 basals, and therefore can not be a columnal, for in all stalked crinoids the topmost 

 columnal supports more or less of the lower margin of the basals or of the under- 

 basals; this is a mechanical necessity, as otherwise the weight of all the calcareous 

 structures would have to be taken up by the soft interior structures immediately 

 above the stem, and by the sutures between the topmost columnal and the 



