kb58 wrote:
My analogy:
Take two pieces of PVC water pipe, one 1 foot long and the other two feet long. Put the end of one in your vice and grab the free end with ViceGrips - and twist. Repeat with the other one. Note that the longer one is easier to twist, and there you have it. The engine is putting equal twisting force through *different-length* axles. The longer one "winds up" more, so less torque ends up going through it. Because more power is going though one axle, the car pulls to one side.
I'm sure one of our engineers here could do the calcuations on how much a solid steel axle/CV joint combo is "twisting," but I'd bet a bottle of good hooch, it falls into the "Statistical noise" range.
One other problem with the axle twisting - some "equal length" setups were really three pieces, with the two angled halves equal length. And they had less torque steer than setups where the longer side was a single piece.
Bravenrace nailed it correctly. When the half-shaft isn't lined up straight into the spindle, you don't get all the torque driving the wheel; the amount of torque that isn't perpendicular to the axis of the wheel goes into trying to turn the wheel sideways. The reason equal length setups produce less torque steer is they also have equal angles - at least, with the suspension on a flat surface. Bumps and body roll can induce torque steer even then.
Some IRS setups can suffer from torque steer too, but the effect is less as not only are these nearly always equal length, but the shafts are often longer and thus the angles are shallower.
In reply to MadScientistMatt:
The RWD torque steer effect is also much lower because the wheels won't turn left and right. Well, they shouldn't anyway 
.
In any setup with a Torsen-style differential, you can see torque steer from unequal tire pressures left-right.
For an interesting read on FWD torque steer solutions, check out the Ford RevoKnuckle as well as the knuckle on the new Renault Megane RS.
