Musical harmony is essential in music perception. Nevertheless, harmony perception is severely limited in cochlear implant (CI) users. Most of the time, components of music chords occur simultaneously. However, the simultaneous presentation of components leads to channel interactions which in turn impair perception. A possible solution might be sequential instead of simultaneous presentation of chord components which we tested in a controlled training paradigm. Post-lingually and bilaterally implanted CI users as well as normal hearing participants were tested on short familiar and unfamiliar music sequences before and after feedback-based training (experimental group) or without training (control group). In 50 percent of trials, the final chord of a sequence was varied, e.g., changed from tonic to subdominant, and participants responded if the final chord was conclusive or not. The stimuli were presented as piano-like sounds and varied on three conditions: timing (simultaneous vs. sequential), tone length (staccato vs. legato), and spectral separation of chord components (wide vs. close). We hypothesize that the sequential presentation of harmony components elicits stronger training effects due to less channel interactions. Preliminary results of training as well as generalization effects on melody and speech perception will be discussed.