Classification was always a major interest for Brian Vickery. In his last years he contributed to the theoretical debate on classification with original ideas that are not well known yet, as some of them were consigned to ephemeral Web pages or private discussion. This paper attempts to report them and to discuss their implications for the current theory of knowledge organization. Vickery’s proposals especially concern: application of the theory of integrative levels to classification as recommended by the Classification Research Group; the various dimensions of knowledge that are involved in the steps ‘from the world to the classifier’; progressive identification by science of phenomena within reality; interplay between the phenomena studied and the human activities providing the context and purpose for studying them; identification of facets of activities as well as facets of phenomena. It is concluded that these ideas offer a substantial contribution to the theory of knowledge organization, and thus should be known, discussed and further developed.