Mental-imagery-based brain-computer interfaces (MI-BCI) control external devices using specific thoughts or mental imagery. MI-BCIs are promising for many applications, but because they lack reliability, they are rarely used outside laboratories. Therefore, user training is critical for controlling an MI-BCI, and users must stably be able to generate specific EEG patterns. However, optimal training methods for acquiring this ability are still being investigated. Onomatopoeias are sensory expressions that symbolically describe sounds, scenes, and feelings. A multimodal approach for training with visual and auditory imagery using onomatopoeias was used to investigate its effects on how user perform.