It is often assumed that the delimitation and the subdivision of the various families which have been distinguished in the Angiosperms, do no longer offer serious difficulties. They would belong to those objects of study for which already long ago a fairly satisfactory solution was found. If we wish to be acquainted with this solution, the only thing we would have to do, would be to look up such works as Bentham and Hooker’s “Genera Plantarum” and Engler and Brand's “Nattirliche Pflanzenfamilien”. Some improvements might still be desirable, but these would be of minor importance only. These assumptions, however, are to be regarded as dangerous illusions. That the very serious nature of the shortcomings found in the delimitation and subdivision of these families, especially of the larger ones, is so often overlooked, is apparently due to an attitude of mind which is observed in a comparatively large part of the taxonomists, viz. a lack of interest in the development of a truly natural classification. This is not incomprehensible. Most of them spend the major part of their time in the elaboration of floras covering areas of more or less limited extent, and they are but rarely aware of the fact that the knowledge of the families which is obtained in this way, remains necessarily incomplete. Moreover, in the elaboration of a flora the most essential point is the construction of serviceable keys to the species as well as to the groups of higher rank, not the exact delimitation of these groups; to the latter end usually more material is required than the compiler of a flora has at his disposition. However, we must not overlook the fact that a key, in order to be serviceable, need not reflect the degree of affinity between the units with which it is dealing; in reality, such keys are often entirely or almost entirely artificial, and this applies therefore also to classifications which are based on such keys. To illustrate this, I will discuss here some of the characters by the aid of which in the Rubiaceae very suitable keys to the genera have been constructed, but which when they subsequently were used for the elaboration of a classification, led to entirely unsatisfactory results.