[rspencer]

It depends on the type of normalization.

Peak normalization just raises the volume so that the peaks reach the set level (e.g., if you set to -.01dB the highest peaks will now be at -.01dB). The dynamic range is preserved, as everything is raised equally. No different than turning up the volume on your stereo.

RMS normalization normalizes to an average, may utilize compression, and can crush the dynamics if heavy-handed. Any use will tend to reduce the dynamics, but it can be done well.

On the whole, obviously, peak normalization is the way to go. It's just adding gain to a source to get it where you want it.