Successful Single-Session Neural Self-Regulation Through Neurofeedback Varies Between Features

Neurofeedback (NFB) and Brain-Computer Interface (BCI) research seldom present within-session individual learning dynamics. This is even though a large proportion of NFB and BCI users cannot learn the neural self-regulation required to control the feedback. Understanding the time course and learning dynamics between subjects will enable us to design more effective NFB and BCI protocols that promote the learning of neural self-regulation. In this study, we aimed to analyze individual learning trajectories of self-regulation of four different cortical rhythms, in terms of both frequency and spatial selectivity. Twenty healthy subjects performed four sessions of NFB training, each session with feedback reflecting a different cortical rhythm as measured with an electroencephalogram. We specifically tested frontal midline (fm) Theta, occipital Alpha, unilateral centrotemporal sensorimotor rhythms (SMR), and central Beta. We show that all subjects were able to self-regulate at least two of these features; however, with varied specificity in the spatial and frequency domains. Unexpectedly, we show that none of the subjects succeeded in regulating fm Theta. Using a clustering approach, we identified two different learning dynamics among the learners across features: a linear increase/decrease and a non-linear plateau-like trajectory. This is the first NFB study employing an intra-subject cross-over experimental design, enabling the direct comparison of neural self-regulation between multiple features. Our results provide important insights into the "non-learner" problem, showing that it is not a feature-universal personal trait. We further show feature-specific spatial and frequency selectivity of neural self-regulation, providing important considerations for future NFB protocols.