244 BULLETIN 82, UNITED STATES NATIONAL MUSEUM. 



(liscoidal, always cirrifcrous, van'injjr from thin to thick, the dorsal ]iole always 

 somewhat smaller than the base, the slopmg sides slightly convex; the dorsal pole 

 is usually smooth, sometimes faintly pitted, and is most commordy flat or more 

 or less concave, less frequently, and usually only in small species, more or less 

 convex; the cutus sockets are large and crowded, and are arranged in from one 

 to three of four (but mostly in one or two) alternating rows; the central cavity 

 of the centrodorsal is comparatively small. In these families the centrodorsal 

 has reached the same stage of development, and is practically the same throughout 

 all the species; it furnishes (except in regard to the excavation for the basal rays, 

 wliich will be explained later) no valid systematic characters; though the number 

 of rows of cutI, the comparative concavity or convexity of the dorsal pole, and the 

 occasional markings on its surface are in some cases good specific indices, none of 

 them can be relied upon. Like the size of the central cavity, the size of the cirrus 

 sockets, and the proportionate size of the dorsal pole and consequent angle which 

 the sides make ^vith the base, they are sometimes useful as a supplement to char- 

 acters exhibited by other structures; but at the best they are uncertain, in respect 

 to both generic and specific differentiation. 



In the Thalassometridffi (figs. 93, p. 153, 94, p. 155, 95, p. 157, 96, 97, p. 

 159, 187-190, p. 235, 191-198, p. 237, and 199-205, p. 239) and in the Charito- 

 metridffi (figs. 99, p. 160, and 206, 207, p. 239) the case is quite different; here 

 the centrodorsal takes on a considerable variety of form and becomes of great 

 importance, both generically and specifically. In the Thalassometridse the centro- 

 dorsal is usually rounded-conical, but less than t\vice as liigh as broad at the base, 

 and the lateral surface is usually separated by more or less pronounced mterradial 

 ridges into five radial areas, each of which contains usually two, more rarely three, 

 definite columns of cirrus sockets. The dorsal pole is usually small, and, though 

 sometimes flat, is usually ornamented in some way, either pitted or thicldy covered 

 with small tubercles or spines, and the interradial ridges and the inferior margin 

 are often similarly ornamented. 



In Ptilometra (figs. 93, p. 153, and 187, 188, p. 235) the centrodorsal is very 

 large, thick discoidal, the sides only slightly oblique, the dorsal pole broad and 

 flat; the cirrus sockets are arranged in two or three crowded alternatuig rows in 

 one species, while in the other the rows tend to lie directly imtler each other, so 

 that the cirri are nearly or quite in 15 columns, three to each radial area, though 

 the radial areas are not in any way marked off, and the columns are closely 

 crowded against each other. 



In Asterometra (figs. 43, p. 77, 94, p. 155, and 189, 190, p. 235) and in Pkro- 

 metra the centrodorsal varies from long corneal to columnar, being usually cohnnuar 

 basally with the portion beyond the cirrus sockets conical, the very small i)olar 

 area vdih five rounded tubercles which are radial in position; the sides are more 

 or less flattened, and are divided into five radial areas by broad and more or less 

 deep grooves or furrows, each radial area containing two columns of cirrus sockets 

 of from two to (rarely) four each, which are separated from eacli other bj' narrower 

 and less prominent (midradial) grooves than those delimiting the radial areas. 

 This tjT)e of centrodorsal is essentially like that found in the larger and more spe- 



