How to run speaker wires...

Twisted pair cabling for digital signals, such as CAT5e and other ethernet cabling, is twisted for three reasons. First, it results in reduced cross-talk between the pair of wires carrying the "transmit" signal (going from, say, device 1 to device 2), and the pair of wires carrying the "receive" signal (going from device 2 to device 1). Second, twisting provides better impedance control, to maintain waveform integrity. Third, it helps assure that interference pickup (rfi/emi) will be essentially equal between the + and - polarities of each signal, which will result in their rejection by the receiving device, which responds to the difference between + and -.

None of that has any relevance to speaker cables, though, due to the vastly lower frequencies that are involved, and also because the impedance levels are extremely low (which will tend to "short out" rfi/emi pickup, not to mention that a speaker will be unable to respond to anything at rfi/emi frequencies).

So I think that twisting would be pointless. As far as bundling + with + or + with -, I doubt that it will make any difference, and if it does I would imagine it would be cable dependent (specifically on dielectric thickness and type). So trying it both ways, as was suggested, is probably the only way to tell.

Regards,
-- Al

Parallel conductors create capacitance -- albeit, tiny if reasonably far from one another (say ¬2"). Twisting solves that, creating slight inductance... You can even combine: cross pairs of equidistant conductors.
I have preferred the parallel method on low power signals, and crossed on speakers. Try, carefully.

Gregm - wire has inductance even without twisting (approx. 0.5uH/ft). Twisting doesn't remove or lower capacitance between wires.

Making twisted pair of speaker wires does nothing, as Almarg stated, but makes you feel good since you did something.

Do you really think manufacturer wouldn't twist wires if there would be any - even smallest improvement (advantage over competition)?