SUMMARYThe ideal classification of a group of organisms is one which takes account of all that is known about the group. It should, therefore, include information about its evolution. There are two aspects of evolutionary theory, (1) that which concerns the past history of the group and which can only be known with certainty from the fossil record, (2) that which concerns present‐day organisms and is expressed in the differences between advanced and primitive types. As regards angiosperms, there is only faint hope that their history will ever be known in sufficient detail for it to be made the basis for a phylogenetic classification. The extent to which the second aspect of evolutionary theory can be incorporated into classification depends on the ability of botanists to assess relative advancement.The first step in deciding which organisms are primitive is to discover which characters are primitive. For this purpose, a number of doctrines have been employed, often without due consideration of their philosophical justification. The doctrines of ‘conservative regions’, of ‘recapitulation’, of ‘teratology’ and of ‘sequences’ are unreliable, since their application is entirely subjective; in the absence of evidence from some other source, it is impossible to judge whether they are appropriate to the structures concerned. The doctrine of’ association’ is unsatisfactory because of the assumptions which have to be made before it can be applied. The doctrine of ‘the basic ground plan’, if applied in the strict sense, is relatively unprofitable; if applied loosely it is dependent on subjective judgement and if applied in the ‘Gestalt’ sense leads only to figments of the imagination. The doctrine of ‘correlation’ has been applied with success to the evolution of secondary wood in dicotyledons and to other characters, both floral and vegetative. Its results are of value since the subjective element is reduced to a minimum and, by its means, it is possible to assess the relative advancement of present‐day taxa.There have been many attempts to represent in two dimensions the phylogenetic classification of the angiosperms, but most of these are at fault because of a misconception of the tree of evolution. The ideal representation is visualized as resembling a target whose concentric rings correspond with relative advancement. On this are disposed the various families in such a way that closely similar ones are close together within circles of affinity, while, at the same time, their radial position corresponds to their relative advancement. In this way the second of the two aspects of evolutionary theory is catered for, and the classification may, therefore, be described as a phylogenetic one. The history of the group, if it becomes known from the fossil record, may then be inserted as a tree of evolution in a plane (corresponding to evolutionary time) which is perpendicular to tha