Non-percussive audio sources (such as vocals) have much less prominent transients, and can wildly differ in dynamics from one another. For this reason, it is hard to make a consistent compression game around it.
Compressors affect the transients and overall dynamics, so percussive sounds with lots of transients are perfect for learning to hear compression nuances. After you get the idea it is easy to apply your learning to other non-percussive audio sources.