This control affects the animation, sometimes called brightness or stridency, in the voice.
The value 50 is used as the default and with this setting the voice will sound as if it's normal.
Reducing this value makes the voice more mellow and soft. Extremely low values might almost be whispered.
Increasing value to 100 makes the voice very forceful, almost shouting.
The value can be specified as a floating point value to control the richness very finely.
The behavior of the spoken voice when controlled by the aural style properties does very much depend on the in-built speech capabilities of the platform. The Macintosh already supports a very versatile and realistic voice synthesizer as an integral component of its operating system. It would be interesting to experiment with this if it were integrated with a usable browser.