A novel approach (‘voxel decomposition’): for inferring neural populations from fMRI responses to a large stimulus set. The results of our study suggest that the human auditory cortex contains distinct neural populations selective for music and speech, respectively. The sound set includes 165 natural sounds, each 2-seconds in duration. The sounds were intended to include many of the sounds people commonly hear in their daily life. Sounds were selected by first compiling a larger collection of 280 sounds that people commonly hear. We then discarded sounds that people had trouble recognizing (<80% performance on a 10-way forced choice recognition task, conducted online using Mechanical Turk). From those remaining, we selected the sounds that people rated as the most frequently heard in their daily lives.