Still more significant is the illusion in [Fig. 3], in which the convexity is very clear. The length, etc., of the lines makes no difference in the illusion.
On the other hand, in [Fig. 4] the diagonals must be definitely thicker than the parallel horizontal lines, if those are to appear not parallel. That the inclination is what destroys the appearance of parallels is shown by the simple case given in [Fig. 5], where the distance from A to B is as great as from B to C, and yet where the first seems definitely smaller than the second.
Still more deceptive is [Fig. 6], where the first line with the angle inclined inwards seems incomparably smaller than the second with the angle inclined outwards.
Fig. 4.
Fig. 5.