One category of factors that can be interesting to predict and explain Brain-Computer Interface (BCI) users' performances are the one based on user stable characteristic such as their personality profile, their learning style, their spatial abilities or their age and gender. In our experiments, subjects were asked to complete psychometric and personality questionnaires, which aimed to assess different aspects of their personality and cognitive profile. The results of those different questionnaires provided us with the explanatory variables to build our different models using traits of the user. In addition, in this study, we build statistical models across experiments as factors that have already been identified as predictors of BCI performance in the literature were studied in single experiments and/or data sets.

So far, our results suggest that, anxious users might have difficulties controlling a BCI while users with high spatial abilities have a tendency to perform better. Furthermore, it seems possible to predict the performance of a session N+1, using the performance of a session N and the score of the vigilance factor. As the vigilance factor measures the tendency to trust others' motives and intentions, it might be interesting to provide deeper technical explanations to subjects with a high vigilance score, as they could be skeptical.

>However, our analyses suggest that the users' traits alone may not be enough to predict their MT-BCI performances across experiments, at least with data collected from a small number (here 3 or less) of sessions. This is a negative result, as it may suggest that users' traits may not have such a strong contribution to BCI performance variations.

Furthermore, BCIs are neurotechnologies using the user brain activity to control systems. In order to better understand why some users manage to produce better quality signals (stable, strong and more discriminant) than others, we used the neurophysiological characteristics of the users to build a model of performances.

As part of my research, I am interested in Motor-Imagery based BCIs. The user has to perform a motor-imagery (MI) task such as the imagination of a left- or right-hand movement. Such tasks share common neurophysiological mechanisms with real movement and therefore are interesting for the BCI technology. More specifically, motor imagery is related to the preparation of an actual movement output and represents meaningful neurophysiological dynamics of motor functions. Consequently, most subjects are able to control amplitude modulations of Sensorimotor Rythms (SMRs). However neurophysiological processes underlying Motor-Imagery BCIs vary over time and across subjects. Neurophysiological predictors of MI-BCI performances such as the amplitude or the stability of sensorimotor-rhythms (SMRs) at rest could help us build models of performance and progression for BCI users.

Our analyses suggest that users with highly variable EEG signals at rest find it more difficult to control BCIs, while those for whom the machine identifies an important mu rhythm (EEG signals oscillating around 10 Hz) as well as a stable beta pick (between 12-30 Hz), generally have good control performance.