On the determination of a function from spherical averages