Examination of a simple evolutionary model shows that the deficiencies of a phenogram, as an estimate of an historical cladogram, are not as great as sometimes imagined. In reality, there are grounds for a fair degree of confidence, particularly in relationships between taxa of higher rank. It can be shown, for instance, that uniformity in rates of evolution is not critical. What is required is, roughly, that later stocks should not evolve for substantial periods at rates much greater than in earlier stocks and/or that sister-species should diverge appreciably before undergoing further splitting. Likewise, the convergent elements of resemblance need not be small, or even uniform: simply, and roughly, later lines should not converge upon each other less than upon earlier lines. The latter holds also for “phylogenetic” reconstructions, based on what are believed to be resemblances in derived attributes only. There seem to be reasonable grounds for believing that such conditions are often met in Nature, and that a phenogram therefore supplies a reasonable phylogenetic hypothesis, for testing against further characters and taxa (but not methods) as they become available. This analysis does not conflict with the pragmatic aspects of classification, and can supply some criteria for choice of a numerical method; for instance, there are grounds for favouring “average linkage” as a clustering strategy.