There are two aspects to this question. You need to first consider the effect of roll on the wing, and then the subsequent effect of sideslip.
A dihedral wing in S&L flight will produce equal lift on both wings as the aoa and effective relative airflow (ERA) is the same on both sides. Introduce a balanced roll, and the ERA on the downgoing wing appears to come more from below, increasing the angle of attack. The upgoing wing has a correspondingly reduced angle of attack. If the roll is not enforced with more or continued aileron input, this effect (true of any wing) will tend to resist the roll. A dihedral wing will, if the roll is stopped at a small bank angle, still have a greater amount of effective lift from the into-roll wing than the outer wing due to its lift vector being closer to the vertical, and will tend to return to S&L flight.
Going back to the rolling scenario, as the roll is introduced, a component of the weight acts sideways and introduces a sideslip element to the equation. This means that in a RH roll, the RH wing leads the LH wing. Therefore the ERA meets the wing at a point closer to the RH wing tip and leaves the wing at a point closer to the LH wing tip. If you look at a diagram of this dihedral wing from behind, you can see that the effect is that the RH wing has a greater aoa than the left.
This effect is only relevent following a disturbance in roll, and is referred to as lateral ststic stability. Once in a balanced (ie no sideslip) turn, with a fixed angle of bank, it is irrelevent whether or not the wing has dihedral.
Edited to remove bullet from foot and take the JCB away!
[This message has been edited by scroggs (edited 16 November 2000).]