BACKGROUND: Diarrhoeal diseases are an important cause of mortality in children younger than 5 years in sub-Saharan Africa. We aimed to evaluate the effect of three handwashing interventions on handwashing with soap (HWWS) after toilet use. METHODS: In this cluster randomised trial in Abidjan, Côte d'Ivoire, we randomly assigned communal housing compounds (1:1:1) to receive one of three interventions: a theory of normative social behaviour (TNSB) intervention, including provision of handwashing stations; handwashing stations only; and no intervention. The TNSB intervention was designed to shift the outcome expectation associated with HWWS from health to riddance of faeces-related disgust, and to increase the perceived descriptive norm and perceived handwashing publicness. Participants and fieldworkers were masked to the study objectives. The primary outcome was HWWS after toilet use, assessed at 1 month and 5 months follow-ups. Analysis was by intention to treat. This trial is registered at the Pan African Clinical Trial Registry, PACTR201501000892239. FINDINGS: Between April 10 and May 22, 2014, we identified 92 eligible compounds, of which 75 compounds were included. Follow-up data on HWWS were available for 23 compounds for the TNSB group, 25 compounds for the handwashing station-only group, and 25 compounds for the control group. The study ended in April, 2017. Compared with a frequency of 5% (29 of 604 occasions) in the control group, HWWS after toilet use increased to 9% (49 of 557 occasions; adjusted risk ratio 1·89, 95% CI 1·16-3·08) in the handwashing station-only group, and 24% (143 of 588 occasions; 4·82, 3·06-7·59) in the TNSB group, at the 1-month follow-up. The intervention effect was only sustained in the TNSB group (98 [22%] of 450 compounds; 2·68, 1·65-4·34). INTERPRETATION: A social norm-based handwashing intervention combined with disgust-inducing messages, with provision of handwashing stations, was effective at increasing HWWS after toilet use. The provision of handwashing stations alone had little effect. Future studies should investigate whether the same approach, when delivered via mass media, can have a similar effect to the face-to-face delivery used in this study. FUNDING: None.