When More Is Not Better: Effects of Multimodal Motor Imagery Guidance on EEG Discriminability in MI-BCI

Multisensory guidance is assumed to enhance motor imagery training by increasing engagement and immersion. However, how adding sensory modalities shapes the specificity of neural representations during motor imagery remains unclear. We reveal a paradoxical tradeoff between engagement and neural discriminability during motor imagery tasks. Using motor imagery as a controlled testbed, we compared static visual, synchronized visual, visual-auditory, and visual-vibrotactile guidance within a unified experimental framework. Self-reports tentatively suggested greater imagery vividness, controllability, and adaptability under multimodal guidance, but these subjective benefits did not translate into stronger task-specific EEG discriminability. Instead, synchronized visual guidance alone produced the most robust sensorimotor event-related desynchronization and yielded the highest decoding accuracy across multiple classifiers. By contrast, adding auditory or vibrotactile cues increased perceived engagement while weakening the specificity of sensorimotor oscillatory features. These findings uncover a multimodal paradox in motor imagery, demonstrating that increasing sensory richness does not necessarily strengthen task-specific neural representations.