Brain–computer interface and semantic classical conditioning of communication in paralysis

► We propose a classical semantic conditioning procedure for communication in the CLIS. ► It is shown that semantic conditioning of cortical responses can be used to enable basic BCI communication. ► SVM-rbf classifier provided the highest accuracy. We propose a classical semantic conditioning procedure to allow basic yes–no communication in the completely locked-in state as an alternative to instrumental-operant learning of brain responses, which is the common approach in brain–computer interface research. More precisely, it was intended to establish cortical responses to the trueness of a statement irrespective of the particular constituent words and letters or sounds of the words. As unconditioned stimulus short aversive stimuli consisting of 1-ms electrical pulses were used. True and false statements were presented acoustically and only the true statements were immediately followed by electrical stimuli. 15 healthy participants and one locked-in ALS patient underwent the experiment. Three different classifiers were employed in order to differentiate between the two cortical responses by means of electroencephalographic recordings. The offline analysis revealed that semantic classical conditioning can be applied successfully to enable basic communication using a non-muscular channel.