After putting the copter together over a few exciting evenings, initial test flights showed some very strange control responses with pitch and roll being swapped. I could "correct" this by swapping my Aileron and Elevator receiver channels, but that made the autopilot's stabilization input completely unpredictable. I checked my wiring and documentation, and found the following:

The PDB that comes with the 3DR-B is oriented to use rotor 2 and 3 as the quad's front in x-mode. See page 7 of this doc:

http://stuff.storediydrones.com/ArduCopter3DRAssemblyInstructions

The APM2, however, thinks that rotors 1 and 3 are in the front:

http://code.google.com/p/arducopter/wiki/APM2RC

So, wiring the PDB correctly and hooking it up to the APM2 results in a copter that has its motors turned 90 degrees from its board. The solution is simple enough.

Maybe this is a well-known issue... It was news to me. Or maybe I've missed something obvious in the the build process. If your newly-built quad is behaving oddly, look into it.