What would be an effective method for empirically testing the impact of timbre on music recognition?