Exploring a listener’s mental representation of tonal hierarchy typically uses several classes of tones or chords embedded in a musical context. Owing to this complexity, a gap exists between behavioral methods, such as the probe tone technique, and physiological studies using electroencephalography (EEG) that commonly require averaging over many stimulus presentations and thus are typically limited in the number of experimental conditions. Here, we propose a novel method for multivariate classification-based EEG feature extraction enhancing EEG multiclass differentiation in the domain of music perception. In a show-case application, we (a) investigate how 13 listeners rate the perceived harmonic distance of 11 classes of chord changes in a continuous stimulus train of major triad chords, (b) apply the proposed method to aggregate typical change-related components of event-related potentials into a compact 11-valued neural profile, and (c) compare both with various change representations derived from music theory. Although the behavioral profiles varied interindividually, showing influences of tonal categories and/or pitch distance, the event-related potential-based neural profiles revealed a dominant influence of pitch distance in 8/13 participants. Thus, a task-driven behavioral rating (reflecting tonal categories) indicates that pitch-based neural representations can be overridden in some participants, whereas in others they could dominate and lead to task-deviant (pitch-based) ratings. In summary, we demonstrate that multivariate analysis methods can extend the scope of music perception-related EEG studies with respect to the number of stimulus conditions at the single-participant level complementing established behavioral methods.