SUMMARY: When extracts from plants infected with various strains of tobacco mosaic virus were ultracentrifuged, the non-infective supernatant fluids still contained 0.5-5 yo of the protein serologically related to the viruses. The small, mostly spherical, particles aggregated to form short rods as the antigen was progressively purified by precipitation with acid or salts. It formed long rods when heated in pH 5-5 buffer or when incubated with trypsin. *4s the particles increased in length, their serological behaviour in precipitation tests changed from ' somatic' to 'flagellar' type.Purified preparations of the unsedimented antigen from plants infected with either of two virus strains contained 0.1-0.2~0 phosphorus, seemingly in the form of a ribose nucleic acid. No evidence was obtained that the preparations were mixtures containing some particles with the 0-5 yo phosphorus characteristic of infective virus and some particles of protein free from nucleic acid.One virus strain produced a higher ratio than the others of unsedimented to sedirnented antigen. The amount of unsedimented antigen was correlated with the total content of anomalous protein when the protein was increasing rapidly, but later it fluctuated unpredictably. No conditions were found that consistently favoured its accumulation, but when plants systemically infected with the type strain were kept a t 36", the total amount of antigen decreased, while the amount unsedimented sometimes increased.The proportion of the total antigen now obtained as poorly infective nucleoprotein is much less than 10 years ago, when a third of it sedimented in the ultracentrifuge but failed to compact into a pellet. Now the uncompacted sediment, with all the host plants and virus strains used, contains only a trivial part of the total antigen.The virus released into sap when leaves are minced is, weight for weight, more infective than the virus that remains in the leaf residues until it is released by fine grinding.The early work on the anomalous protein contained in virus-infected plants was done primarily to identify material that could be related to the viruses themselves, and its main aim was to make preparations with the maximum infectivity and the minimum of different components. Little attention was given to non-infective material, and the emphasis laid on the homogeneity of purified preparations served only to strengthen the general assumption that viruses multiply by the replication of an infecting particle to give a result that is uniform except for occasional variants regarded as equivalent to mutations. When we (Bawden & Pirie, 1945a, b ) found that plants infected with the Rothamsted tobacco necrosis virus or with tobacco mosaic virus (TMV) contained more than one type of anomalous protein, we questioned the validity of this assumption, and have since argued (Bawden & Pirie, 1950a, b, 1952, 1953 that infection is more usefully regarded as a change in the protein metabolism