and 4a (pl. XXI.), the former showing the outline of the sternum of Notornis from the ventral face with that of Ocydromus superposed in dotted outline on the left side, and that of Porphyrio in broken outline on the right; while fig. 4a shows the right half of the sternum of Notornis with that of Tribonyx superposed in dotted outline.

The sternum of Notornis is broad and flat; its anterior edge is somewhat emarginate, as in Ocydromus, and is devoid of all trace of the manubrium or rostrum (fig. 4, 4a, r) found in Porphyrio, and to a less degree in Tribonyx. The coracoid grooves are even more widely separated than in Ocydromus, instead of having merely the width of the rostrum between them, as in the other genera. The diminution in width of the sternum from its anterior to its posterior end is very gradual; in this respect Notornis most nearly approaches Tribonyx. The external xiphoid processes (e. x. p.) are divergent, not expanded at their distal ends, and are proportionally shorter than in either of the allied genera; the middle xiphoid process is blunt and unossified, the bone terminating in a straight transverse edge, about six mm. from the actual extremity of the process. In this again the resemblance between Notornis and Tribonyx is of the closest kind: the middle xiphoid both in Porphyrio and Ocydromus is completely ossified, terminating in the former by a truncated edge, while it is deeply emarginate in the latter.

The keel of the sternum is feebly developed, being hardly deeper, proportionally to length of trunk, than that of Ocydromus. Its anterior edge has nothing of the strong forward convexity found in Porphyrio, but passes almost insensibly into the ventral edge; in this respect the resemblance to Ocydromus would be great, but for the fact that in the latter a strong bifid thickening (fig. 4, k) is formed at the junction of the anterior and ventral borders, whereas the corresponding thickening in Notornis is less marked and shows no tendency to division; the resemblance to Tribonyx is here very marked.

Another point connected with the fightlessness of Notornis is the very slight lateral curvature of the sternum; its two sides enclose a dihedral angle (fig. 5 B.) which is nearly as open as that of Ocydromus (A) and considerably greater than in Tribonyx or Porphyrio (C and D). This transverse sternal angle as it may be called, seems to be pretty constantly more open in flightless birds than in the normal members of the same group; its increase, and the correlated diminution of the keel, cause the sternum to approach to the ratite type, as is especially well seen in Didus, Cnemiornis, Stringops, and Aptornis, and to a less extent in Nesonetta, Ocydromus and Notornis.

On the whole the sternum of Notornis differs from that of Tribonyx in much the same way as the latter from that of Porphyrio. Tribonyx is, in all important respects a mean between the two extremes furnished by Porphyrio and Notornis. Ocydromus, on the other hand is, in some respects,

intermediate between Tribonyx and Notornis, while in others it is in advance of Notornis, in just the same way as the latter is in advance of Tribonyx, i.e., deviates more from normal carinate characters.

In the shoulder-girdle the four genera form a very interesting series: this is shown in fig. 6 (pl. XXI.), in which all four shoulder-girdles are reduced to a common length of trunk, and the coracoids are superposed upon one another, so as to coincide in direction. As regards relative length of coracoid and scapula, Notoris is seen to intermediate between Ocydromus and Porphyrio, Tribonyx having the coracoid a little longer than that of Notornis, the scapula somewhat shorter. The same series is observable in the curvature of the scapula; this is greatest in Porphyrio, next comes Tribonyx, then Notornis, and finally Ocydromus with a scapula nearly straight save at its distal end.

A similar gradation is seen in a more important point, namely, in the angle enclosed between the adjacent portions of the coracoid and scapula. As was first pointed out by Professors Huxley and Newton^*, one of the most marked features of the Carinatce is the fact that the coraco-scapular angle never approaches 180° as in Ratitee, and is usually less than 90°; the only exceptions mentioned by either author being Didus and Ocydromus, in which the angle is slightly over 90°. This, then, is another morphological character which has a definite relation to the power of flight, the coraco-scapular angle, like the transverse sternal angle, being found, speaking generally, to increase pari passu with diminution of that power. It would, however, be a mistake to suppose that there is anything like a constant relation between flightlessness and increase of the coracoscapular angle. I find, for instance, that it is less in Tetrao than in Vultur, and that of the two skeletons of Stringops in the University Museum, one has the angle less than 90° on both sides, while in the other the angle on the left side is just over a right angle, that on the right being the same as in the other skeleton: so that the angle has undergone little or no increase in a bird in which the carina sterni is practically obsolete, and the furcula rudimentary. But the strangest exception to the rule that the coracoscapular angle in the Carinata is less than 90°, is furnished by that paragon of flying birds, the albatross, in which the angle is fully 100° the same, though to a less degree, is the case in the Nelly (Ossifraga).^† Thus Diomedea and Ossifraga must be added to the above list of exceptions, as well as Stringops (?), Cnemiornis, Aptornis, Tribonyx, and Notornis. As a very general