Well, in theory with *any* speaker in any configuration, the narrower
the range of frequencies that you demand it play, the better it is
going to sound--that just means that the waveforms that is has to
reproduce are that much less...."cluttered" and less complex. It also
will mean that the speaker will have to dissipate less power, thus
resulting in less distortion caused by non-linear speaker behavior.
So, ideally, the best situation would be to have one speaker
reproducing exactly one frequency using a "brick wall" (infinite dB/
octave roll-off) filter...and yet somehow have all of these 1Hz-
bandwidth speakers acting as a single point source.

However, in practice, we don't deal with "brick wall" filters (infite
dB/octave roll-offs after the crossover point), so there's going to be
overlap. And in this overlapping range between drivers, you're going
to have phase shifts introduced by the crossovers themselves. Also, we
can't stack a bunch of drivers atop one another and achieve the
equivalent of a single point source--you have to either put them side-
by-side or on to of one another on the same axis, in which case the
high frequency drivers would be much closer to you than would the low-
frequency drivers.

So, in the real world of imperfection and compromise, you do the best
you can and divide up the audio spectrum into smaller "bit-sized"
bands that are handled by specialized drivers.

Every speaker is going to have a natural frequency response (absent
any enclosure) that is going to "roll-off" at both the high and low
ends...loudspeakers are essentially bandpass devices--there is no such
thing (to the best of my knowledge) as a true "full-range" (20Hz -
20kHz, +/- say 3dB). This is pretty obvious if you apply unfiltered
music through a subwoofer--you won't hear the "tssk tssk tssk" of a
high hat, nor will you get very good output in the 20Hz range (again,
absent an enclosure of some type). Likewise, unfiltered music applied
to a midrange will result in most of the guitars and vocals getting
through, but not much in the way of cymbals and kickdrums. The same
goes for tweeters.

But we rarely let speakers play with unfiltered program source--we
apply a lowpass filter to the subwoofers, a bandpass (low + high pass)
to the mids, and a high pass to the tweets. The LP filter on the subs
keeps Barry White from sounding like he's in your trunk, and allows
you to crank up your bass amp and get more "oomph" out of it since the
human ear is not as sensitive to distortion at the low end of the
ear's response curve. The high-pass filters on the mids and tweets
protect these smaller, delicate speakers from being literally roasted
by high-energy, low-frequency signals (a typical tweeter playing full
range will probably only handle about 0.5 to 1 Watt before it goes
*poof*); and the low-pass filter on the midrange speakers keep them
from trying to reproduce frequencies for which they were not designed--
which is right around where those speakers have their resonant peak
and start sounding screechy.

I know I've taken the long way around answering this question, but I
figured that the basics of why we use crossovers needed to be laid
out. Does all of this help, or did I just muddy the waters for you?

-dan