As has been shown above, the history of the mammals is recorded, save in a very few instances, in terms of bones and teeth and, to the uninitiated, it might well seem that little could be accomplished with such materials. However, it is the task, and the perfectly feasible task, of palæontology to make these dry bones live. It is a current and exceedingly mischievous notion that the palæontologist can reconstruct a vanished animal from a single bone or tooth and, in spite of repeated slayings, this delusion still flourishes and meets one in modern literature at every turn. No doubt, much of the scepticism with which attempts to restore extinct animals are met by many intelligent people is traceable to the widespread belief that such off-hand and easy-going methods are used in the work. So far from being able to make a trustworthy reconstruction from a few scattered bones, competent palæontologists have been sometimes led completely astray in associating the separated parts of the same skeleton. More than once it has happened that the dissociated skull and feet of one and the same animal have been assigned to entirely different groups, just because no one could have ventured, in advance of experience, to suppose that such a skull and teeth could belong to a creature with such feet. In all these cases (and they are few) the error has been finally corrected by the discovery of the skeleton with all its essential parts in their natural connection. While the number of complete skeletons of Tertiary mammals as yet collected is comparatively small, it is often possible to construct a nearly complete specimen from several imperfect ones, all of which can be positively shown to belong to the same species. Such composite skeletons are almost as useful as those in which all the parts pertain to a single individual, though in making the drawings it is not easy to avoid slight errors of proportion. It must not be supposed that no successful restoration of missing bones is practicable; on the contrary, this can often be done very easily, but only when all the essential parts of the skeleton are known. Even if an unlimited number of perfect skeletons were available, of what use would they be? A skeleton is a very different looking object from a living animal, and how is it possible to infer the latter from the former? Do the many restorations of extinct mammals which this book owes to the skill of Mr. Horsfall and Mr. Knight deserve any other consideration than that due to pleasing, graceful or grotesque fancies, with no foundation of solid fact? To answer these questions, it is necessary first to consider the relations of the bony structure to the entire organism and then to discuss the principles in accordance with which the restorations have been made.