242 C. W. Andrews — ySilpyornis from Madagascar. 



Of the skull no complete specimen is known, but a nearly perfect 

 cranial portion, the anterior part of the beak, and a mandible were 

 described and figured in the paper above referred to (pi. viii). The 

 foramen magnum is large and oval, the occipital condyle pedunculate, 

 and there is a precondylar fossa. The paroccipital processes are 

 large, but only extend downwards a little below the condyle. The 

 lambdoidal crest is weak, and the occipital surface passes into the 

 roof of the skull by a gentle curve. The surface of the frontals 

 is marked by several rows of deep pits which indicate the presence 

 of a crest of large feathers. The tympanic cavity is shallow, but 

 the temporal fossa is very deep aad narrow ; there is a large 

 zygomatic process. The basi-temporal platform is very prominent : 

 anteriorly it bears a pair of large basi-pterygoid processes, and 

 laterally it is grooved by the eustachian canals, which remain open 

 as in Dinornis. There is only a single facet for articulation with 

 the quadrate. This bone is chiefly remarkable for the great ex- 

 pansion of its orbital process, which forms a broad plate with 

 a rough inner surface in which some pneumatic foramina occur. 

 The single convex facet for articulation with the cranium is some- 

 what triangular in outline, and the articular surface for the mandible 

 is completely divided into two facets, the inner convex in all 

 directions, the outer concave from side to side. On the inner side 

 of the bone there is a broad surface for union with the hinder 

 extremity of the pterygoid (71), on the outer a deep cup for the 

 posterior end of the jugal arch. 



The rami of the mandible are vei'y broad (from side to side) and 

 massive. The articular surface for the quadrate is deep, and there 

 are well-developed posterior and internal angular processes. 



The vertebral column as reconstructed consists of twenty true 

 cervicals and eight vertebrae bearing free ribs, of which probably 

 the three anterior are cervico-dorsals. The fused pelvic vertebree 

 are about twenty in number. It is, of course, possible that 

 a mistake may have been made in the number of vertebrae, but as 

 far as it is possible to judge by the gradual increase and decrease in 

 size of the various processes from one region to another, the column 

 as now mounted must be very nearly correct. 



In the atlas vertebra, of which a very perfect specimen is 

 preserved (Fig. 1, A), the cup for the occipital condyle is deep and 

 circular in outline : it is nearly complete, the cleft in its dorsal 

 margin being comparatively narrow and only extending downwards 

 about one-third of the diameter. The posterior articular surface is 

 shaped somewhat like the letter B, the straight border being the 

 dorsal margin, while the two lobes are two nearly completely 

 separated, slightly concave, articular surfaces for the axis. The 

 inferior angles of these lobes are produced into short processes 

 forming lateral hypapophyses : there ' is no trace of a median 

 hypapophysis. The neural arch is very large, and its two halves 

 are completely fused above. The pedicels are narrow, but superiorly 

 the arch is greatly expanded, forming a backwardly projecting wing 

 which forks posteriorly into au upper and lower lobe. The upper 



