I think I know the answers, but then again

I have a "hobby" project to evaluate intelligibility of speech with
varying parameters ( primarily sample rate and bandwidth ).

[ for perspective, in a previous incantation I "discovered" formants ;]

My sample universe is a reading of the Bible recorded under studio
conditions. As it is a commercial recording, I'm assuming Nyquist is the
least restrictive criterion it meets.

Am I correct thinking that if I digitally filter my source to less than
3 kHz i can safely decimate to 8 kHz sample rate?