Abstract A generalized taxon concept (GTC) is proposed with a method for revealing and ranking difficult taxa at any level in the taxonomic hierarchy. The method is based on cluster quality, defined jointly by the compactness of a cluster's contents and its isolation from its informational neighbours. The cluster contents are individuals in the case of species and at higher levels, taxa from the rank below. A standard, quality threshold value is obtained from clustering accepted taxa in the informational region. If the quality value of a problem cluster lies at or above the threshold it is accepted as a taxon and ranked with others at the current level. If it lies below, and is likely to be informationally useful, it may be accepted as a sub-taxon such as a subgenus or subfamily. Provision is made for coarsely scored data. The clustering is mainly based on homogeneity, where possible with a rapid, fuzzily cladistic de-weighting of symplesiomorphies by self-graded factors. The strengths of inter-item reactions such as breeding and DNA-DNA hybridization may also be used. The method is agglomerative so that it can rapidly reveal polythetic groups which may be riddled with exceptional property states caused by long exposure to natural selective forces. All this fits the evolutionary outlook of the GTC, which sees taxa as fuzzy clusters of populations and lineages sharing much of a genetic memory, moulded by a unique history of evolution and extinction. Practical problems of methods based on this and other taxon concepts are briefly compared. The GTC's approach offers important refinements that could be valuable in helping to speed up urgent surveys of biodiversity, especially in the moist tropics.