It's a good question. The simple answer is, the entire audible frequency range matters.
Above about 2 KHz, the ears use volume (amplitude) differences between the left and right channels to figure out imaging. Below about 700 Hz, they use phase differences. Between 700 Hz and 2 KHz, it's complicated.
It's actually a bit more complicated than that. The amount of reverberant (reflected) signal vs. direct signal determines image depth behind the plane of the speakers. The reflected signal is that which bounces off from the recording venue (or is added by the sound engineer). So when you hear a singer dead center at the front plane of the speakers, that's all direct sound which has been recorded equally in the left and right channels. If you hear drums behind the singer, that's due to reflections that are coherent with the direct sound coming from the singer. Your ears-brain system is built to figure spatial relationships like this out using amplitude and phase differences in the left and right channels.
When I say the entire audible frequency range matters, what I mean is that amplitudes (volumes) must be reproduced at the level which they were recorded for each frequency. This is called the frequency response and some call it the amplitude response. These are the curves that you see on the Axiom product pages. It's also important for this response to not change during quiet passages (microdynamics) and peaks (macrodynamics). If the system loses accuracy during whispers and peaks, there goes the image and soundstage.
You'll run across the words accuracy and linearity. Accuracy measures the degree to which the output from the system reflects the input. By output, I mean level and frequency. Linearity measures the degree to which accuracy is maintained across all frequencies and amplitudes (from very quiet to very loud).
The more accurate and linear a system is, the better the imaging and soundstage. System = the entire audio chain.
