From my post at AA a while back:
Bi-wiring does not bypass any crossover components.
Each driver has its own dividing network section. Each section has +/- inputs (at the binding posts), and +/- outputs (that go to the driver terminals). With single wire, all the + input leads are connected to the + post, and same for the - input leads.
With bi-wiring, a second set of posts is added, and then the network sections' input leads go to each dedicated pair of posts. Electrically, both configurations are basically the same. Bi-wiring requires two sets of speaker cables, of course. So what's the big deal?
The theory goes like this. The midbass woofer cone motion generates a back EMF, which can jump over into the tweeter network input and slightly modulate the tweeter. Of course, the tweeter network by definition filters out the lower frequencies, and rejects most of this corrupting signal. However, right around the crossover region, a small amount of intermod remains.
With bi-wiring, the back EMF would have to "travel backward" to the amp on the woofer speaker cable, and then jump over onto the dedicated tweeter cable before it could modulate the tweeter. So that's several feet more of wire. Won't the signal still affect the tweeter? Not really, because don't forget the very low output impedance of the amplifier is across the +/- legs of both speaker cable inputs. IOW, the amp effectively "shorts out" the back EMF signal, and prevents it from going down the tweeter cable.
Sonically, you hear a very small improvement in the crossover region. To me, it sounds like an increased "purity" and definition. The effect is subtle. I don't recommend it except for reference grade systems, where subtle improvements are more audible. However, I put it in the same class of tweaks such as cables, spikes, etc. Given enough tweaks (scientific ones that is), the tiny differences finally add up to become a *small* difference! And if not overly expensive, small differences are worth pursuing.
HTH.