When fleas are hatched in a nest they have no choice but to attach themselves to the young mammals or birds. But even in that case they frequently leave their hosts and do not for very long remain stationary. Moreover, when a host dies and becomes cold the fleas invariably leave their quarters, which explains how it may happen that Carnivora get infested with the fleas of their prey. This change of hosts which is always occurring makes it impossible to draw conclusions from material collected in zoological gardens where many animals are herded together. In menageries, too, the normal conditions of breeding are absent. A German naturalist collected 2036 fleas from theatres, concert-halls, ball-rooms, schools and barracks in the grand-duchy of Baden and found that more than fifty per cent. were dog-fleas (_Ctenocephalus canis_). What the proportion may be in other parts of Europe we have no materials from which to form a judgment. In zoological gardens cat-fleas (_Ct. felis_) are generally numerous in most of the cages. It is, of course, well known to every zoologist that species are not fixed or constant and that various forms of mammal or of bird tend to show geographical variations. When a long series of skins are laid out on a table and carefully examined it is seldom that those from the west of any great region cannot be picked out and distinguished from those obtained in the east. So we also get northern and southern forms of the same species varying slightly. These variations are perceptible in many forms of insects, and zoologists now describe these local races as subspecies and designate them with trinomials. No one, however, knows enough as yet about all the various forms which are assumed by fleas to attempt, except in a few instances, to do so in the case of these animals. The study of Siphonaptera is still quite in its infancy. We know little or nothing of the minute geographical variation of fleas. That there is such a thing can already be seen in a few species. In the meantime the study of variation must be postponed until collectors have amassed a more plentiful amount of material; and it is best to treat all forms which are to all appearance constantly different as being specifically distinct until more is known about variation. Any classification of fleas that may now be attempted can only be tentative. It will be enough for present purposes if the reader will remember that the Order Siphonaptera can be divided into three groups or families: (1) the chigoes and their allies, which are the most parasitic fleas (_Sarcopsyllidæ_); (2) the typical fleas to which the majority of species belong (_Pulicidæ_); and (3) the bat-fleas (_Ceratopsyllidæ_), which have certain peculiarities that will be described in a later chapter.