Main topics:Movement analysis in clinical practice, Functional outcome measures in mobility, Analysis of clinical movement data Introduction and aim: The identification of subjects at risk of falling has become a strategic goal in health policy programs and relies on clinical screening tools. Instrumental balance assessments can improve the predictive performance of these clinical scales [1]. Frequent fallers (FF) are known as subject at very high risk of fall and do not need any further screening to be categorized. Conversely, the early identification of future new-fallers among older people would permit to focus the delivery of fall prevention programs on selected individuals only (intervention appropriateness). Single posturographic parameters (PP) have been proven to differentiate between non-fallers (NF) and FF but not between NF and first or, more in general, rare fallers (RF) [2]. In this study, we applied the statistical technique of Principal Component Analysis (PCA) on a set of posturographic data obtained from a large sample of elderly to investigate thediscriminant capacityof tunedmixtures of posturographic variables. Patients, materials and methods: The study population consisted of 130 cognitively able individuals (clinical dementia rating, CDR≤0.5), 58 M and 84 F, mean age 77±7 years and age range 62–91 years, seen consecutively at the Memory Clinic of the Regional Hospitals of Mendrisio and Lugano, Switzerland. Subjects were categorized as NF (N=67), RF (one or two falls, N=45) and FF (more than two falls,N=18) according to their last year fall history. Postural stability was assessed in five different conditions: eyes open (EO) and closed (EC) on both a firm and a soft surface (EOFS, ECFS, EOSS, ECSS), and in dual task, that is with eyes open on a firm surface while performing a cognitive task (backward counting). Protocol details are described in [3]. Principal Component Analysis (PCA, Kaiser criterion, Varimax rotation) [4]was used to select themost significant features among the set of 17 parameters that characterize the posture maintenance task. PCA-derived parameters, rather than the individual PP, were used to test, in each task, statistically significant differences between the NF, RF and FF groups (Wilcoxon test, alpha=0.05). Results: In the EOFS and EOCS tasks, the PCA analysis showed that 4 PCs accounted for 88% and 89% of the variance in the whole set of parameters. The Kaiser-Meyer-Olkin measure of sampling adequacy had a value of 0.78 and of 0.77, respectively, thus confirming a good factor analysis. PCA allowed to reduce the number of PP introducing new ones (i.e. the PCA derived parameters) characterizedbybeing, for each experimental trial, a linear combination of themost significant PP (loading value higher than 0.4 in absolute value). The PCA-derived parameter based on a combination of a set of CoP medio-lateral variables in the EOFS condition was different between NF and FF (P