PURPOSE Rasch modelling was used to establish the validity and robustness of the Falls Behavioural (FaB) Scale for Older people. METHODS The sample comprised 678 community-residing elderly people aged 65 - 98 years. Data were analysed by calculating goodness of fit statistics, principal components analysis of residuals and by exploring the effectiveness of the category rating response-scale. RESULTS A partial-credit rating scale was the best fitting solution and the major change to the original version of the FaB. Analysis supported a 29-item intervention version and a briefer 24-item outcome evaluation (research) version. The latter produced item mean infit statistics of 1.00 (Z = 0.0, SD = 0.33) and mean outfit statistics of 1.03 (Z = 0.0, SD = 0.53), a person separation of 2.36 and internal reliability of 0.85. CONCLUSION The 29 item partial rating scale is valid, reliable and would be useful in clinical situations when used as a prompt for discussion and in raising clients' awareness of potential hazards; it also can be used as an outcome measure. The short form is a useful alternate for evaluating the effectiveness of fall reduction interventions that aim to encourage protective strategies when negotiating the environment, mobilizing and doing activities of daily living. Both scales would be improved by adding very difficult and very easy items to increase the range of ability levels of the people to whom it can be applied with precision.