In preparation for a large phase-3 trial of pneumococcal polysaccharide/conjugate vaccine among infants in The Gambia, the relative merits of community and individual randomisation were considered. The impact of vaccination might be enhanced in a community randomised trial if there was a substantial 'herd' effect. This might occur if those vaccinated comprise a substantial proportion of potential transmitters of infection. However, there are few data on the sources of pneumococcal infections in The Gambia and with the high degree of mobility of people in The Gambia, it seems unlikely that any herd effect would be strong. In the absence of a herd effect, a community-randomised trial would have lower power for the mortality end-points compared to an individually randomised trial of the same size. In addition, a community-randomised trial might not provide sufficient control against potential confounders and blinding might be difficult to sustain if the vaccine has a strong effect. An individually randomised trial seems a better strategy under the conditions prevailing in The Gambia.