Here are two recordings of the spectral averages of a girl speaking deeply (resonantly) and nasally. The original recordings were analyzed in MaxMSP, deconstructed into the frequency domain, and averaged. Essentially, what you are hearing is the resonant frequency of the sentence over time. The two attached pictures represent the sinusoidal resynthesis of the averaged recording.
I am doing all of this to try and better understand the spectral distribution of different speech patterns.