The early stages are in the main similar to those of the Anura. The body of the embryo is, as pointed out by Scott and Osborn, ventrally instead of dorsally flexed. The metamorphosis is much less complete than in the Anura. The larva of Triton may be taken as typical. At hatching, it is provided with a powerful swimming tail bearing a well-developed fin: there are three pairs of gills placed on the three anterior of the true branchial arches. Between the hyoid and first branchial arch, and between the other branchial arches, slits are developed, there being four slits in all. At the period just before hatching, only three of these have made their appearance. The hyomandibular cleft is not perforated. Stalked suckers, of the same nature as the suckers of the Anura, are formed on the ventral surface behind the mouth. A small opercular fold, developed from the lower part of the hyoid arch, covers over the bases of the gills. The suctorial mouth and the provisional horny beak of the Anura have no counterpart in these larvæ. The skin is ciliated, and the cilia cause a rotation in the egg. Even before hatching, a small rudiment of the anterior pair of limbs is formed, but the hind-limbs are not developed till a later stage, and the limbs do not attain to any size till the larva is well advanced. In the course of the subsequent metamorphosis lungs become developed, and a pulmonary respiration takes the place of the branchial one. The branchial slits at the same time close and the branchiæ atrophy. The other types of Myctodera, so far investigated, agree fairly with the Newt. The larva of Amblystoma punctatum (fig. 84) is provided with two very long processes (_s_), like the suctorial processes in Triton, placed on the throat in front of the external gills. They are used to support the larva when it sinks to the bottom, and have been called by Clarke (No. 98) balancers. On the development of the limbs, these processes drop off. The external gills atrophy about one hundred days after hatching. It might have been anticipated that the Axolotl, being a larval form of Amblystoma, would agree in development with Amblystoma punctatum. The conspicuous suctorial processes of the latter form are however represented by the merest rudiments in the Axolotl. [FIG. 84. LARVÆ OF AMBLYSTOMA PUNCTATUM. (After Clarke.) _n._ nasal pit; _f._ oral invagination; _op._ eye; _s._ balancers; _f.l._ front limb; _br._ branchiæ.] The young of Salamandra maculata leave the uterus with external gills, but those of the Alpine Salamander (Salamandra atra) are born in the fully developed condition without gills. In the uterus they pass through a metamorphosis, and are provided (in accordance with the principle already laid down) with very long gill-filaments[57]. [57] Allen Thomson informs me that the crested Newt, Triton cristatus, is in rare instances viviparous.