Experimental research in behavioural nutrition is often limited by practical applicability. In the present study, we assess the reproducibility and validity of a new experimental method using food replicas. A total of fifty-seven people were invited on two separate occasions with an interval of 2 weeks to serve themselves a meal from a fake food buffet (FFB) containing replica carrots, beans, pasta and chicken. The external validity of the FFB was assessed in a second study by comparing meals served from replica foods (beans, pasta, chicken) with meals served from a corresponding real food buffet (RFB). For the second study, forty-eight participants were invited on two separate occasions; first to serve themselves a meal from the FFB or an RFB and 2 weeks later from the other buffet. The amounts of food items served and (theoretical) energy content were compared. Correlation coefficients between the amounts of fake foods served were 0·77 (95 % CI 0·68, 0·86) for chicken, 0·79 (95 % CI 0·68, 0·87) for carrots, 0·81 (95 % CI 0·69, 0·89) for beans and 0·89 (95 % CI 0·82, 0·93) for pasta. For the FFB meal and the RFB meal, the correlations ranged between 0·76 (95 % CI 0·73, 0·91) for chicken and 0·87 (95 % CI 0·77, 0·92) for beans. The theoretical energy of the fake meal was 132 kJ (32 kcal) lower compared to the energy of the real meal. Results suggest that the FFB can be a valuable tool for the experimental assessment of relative effects of environmental influences on portion sizes and food choice under well-controlled conditions.