MONOGRAPH OF THE EXISTING CRINOIDS. 379 



In the majority of the comatulids the radials are just visible beyond the edge 

 of the centrodorsal, or terminate just at the edge (figs. 96-98, p. 159, and 228, p.245). 

 The portion concealed by the centrodorsal is horizontal, but the portion extending 

 beyond it, while often horizontal, is usually more or less turned upward toward the 

 dorsoventral axis, and may even be parallel to that axis (figs. 94, p. 155, 110, p. 176, 

 and 112, p. 179). This slanting of the exposed portion of the radials indicates 

 that the transformation from a primitive vertical to a secondaiy horizontal position 

 has not quite been completed, but that the radial has reclined to an angle equal 

 to that proportion of the angle included by lines drawn from the center of the 

 dorsal surface of the radial pentagon, one to the middle of the distal outer edge of 

 the radial and the other to the middle of the proximal outer edge, which is equal 

 to the proportionate length of the free outer edge (measured perpendicularly) as 

 compared with the dorsal length beneath the centrodorsal. It is thus evident 

 that in no case does the comatulid radial depart greatly from a horizontal position. 



There is but slight correlation between the comparative condition of the radials 

 and the various systematic groups, though in general the most primitive families, 

 such as the Pentametrocrinidse (figs. 113, 114, p. 181, 119, p. 185, 120, p. 187, and 

 121, p. 189) and the Atelecrinidae (figs. 123, p. 192, 124, 125, p. 193, 227, 228, p. 245, 

 414, p. 319, and 573, 574, pi. S), show the least approach toward a horizontal position, 

 this tendency increasing with specialization until in certain of the Comasteridse (figs. 

 164, p. 227, 165-170, p. 229, and 181, 182, p. 233) we find the condition perfected. 



It is curious that the angles of the articular faces of the radials do not show a 

 definite correspondence to the recumbency of the radials as a whole. While as 

 a general rule there is a close relationship between the angles at which the articular 

 face is inclined to the dorsoventral axis and the angle at which the radial as a whole 

 is inclined to the horizontal, yet the former is far more constant in any given genus 

 or family, and is therefore a far more reliable systematic character. While the 

 latter is greatly affected by ontogenetical changes, the former is fairly constant 

 throughout life, and thus it comes about that in certain forms, as in very large 

 specimens of certain species of Pentamdrocrinus, the radials may be quite concealed 

 exteriorly by the centrodorsal and perfectly horizontal, while the articular faces 

 are still inclined toward each other at an angle of 90 (or toward the dorsoventral 

 axis at an angle of 45) as in the young. 



The Macrophreata, in all of which the angle made by the radial articular faces 

 to the dorsoventral axis is relatively large, tend to maintain a moderate angle of 

 basal inclination, though in the more specialized subfamilies of the Antcdonidae, 

 particularly those including phylogenctically overgrown species inclining (when 

 proportionately very large) toward the development of oligophreate characters, 

 the angle of basal inclination frequently becomes 180; in the Oligophreata the 

 angle between the direction of the articular faces and the dorsoventral axis is much 

 less than in the Macrophreata, and in the most highly specialized forms these faces 

 may even be parallel to the dorsoventral axis, as for instance in many of the Coma- 

 steridro, and here we find that the radials are always very nearly, often quite 

 recumbent, even if, as in many of the comasterids, they are not at all concealed by 

 the centrodorsal. 



