Prediction of item familiarity based on ERPs

A simple recognition task was used to investigate if the item familiarity of pictures can be predicted based on single trial ERPs during item presentation, to explore the possibility of using this property in a BCI application. Two experimental parts with equal learning phases but different ratios of old and new stimuli in a forced choice memory recognition test have been performed. We were able to predict item familiarity with accuracies above 70 % based on the ERPs elicited during item representation in both parts of the experiment. In some cases, the classification accuracy even exceeds the behavioral accuracy of the subjects. Usage of this property, for example in an education-oriented scenario, seems feasible in a BCI application.