An inactivated vaccine against respiratory syncytial virus (RSV) was compared with two live vaccines. The inactivated (GC) vaccine consisted of glutaraldehyde-fixed bovine nasal mucosa cells persistently infected with RSV and emulsified with oil adjuvant. The live vaccines were a modified virus (MV) derived from a bovine strain of RSV and a temperature-sensitive mutant (ts-1) derived from a human strain. The GC vaccine was inoculated subcutaneously into 12 calves and the live vaccines intramuscularly into eight calves each. Nine unvaccinated calves acted as controls. The vaccines were administered in two doses 3 weeks apart and all calves were challenged intranasally with 2 X 10(7) p.f.u. of bovine RSV 3 weeks after the second dose. At the time of challenge calves given GC, MV and ts-1 vaccines had mean serum neutralizing antibody titres of 25, 19 and 2 respectively; mean titres of IgG1 antibody by radioimmunoassay were log10 4.5, 1.3 and 2.6 respectively and mean zone areas by single radial haemolysis (SRH) were 107, 27 and 36 mm2 respectively. Eleven of 12 calves given GC vaccine were completely protected against challenge but all control animals and those given the two live vaccines were infected. The mean peak titre of virus in nasal swabs of control calves was 3.0 log10 p.f.u./ml and the mean duration of virus shedding was 6.8 days. Both these parameters were significantly reduced in animals given MV and ts-1 vaccines: mean peak titres were 2.1 and 2.4 log10 p.f.u./ml and mean duration of shedding was 3.4 and 3.3 days respectively. Thus, protection correlated better with RSV antibody detected by radioimmunoassay and SRH than with neutralizing antibody. These results are discussed in relation to the possible mechanism by which protection was mediated.