On the basis of Taylor-Bishop-Hill's theory, many previous theoretical investigations have predicted that, at high rolling reductions, most of orientations should rotate along the 13 fiber from {110}< 112 > to {123}< 634 > and finally into the {112}< 111 > stable end orientations. Although some exceptions exist, experimental observations have shown, on the other hand, that the maximum on the S fiber is located still at about (123)< 634 > even after 97% cold rolling. In the present paper, high purity Al containing 50 ppm Cu was cold rolled up to 99.4% reduction in thickness and examined whether {112)< 111 > stable end orientation could be achieved experimentally. It was found that, with increasing rolling reduction above 98%, {110}< 112 > decreased, while orientations in the range between {123}< 634 > and {112}< 111 > increased, suggesting that crystal rotation along the beta fiber from {110}< 112 > toward {123}< 634 > and {112}< 111 > in fact took place. At higher rolling reductions, however, further rotation of this peak toward {112}< 111 > was extremely sluggish, and even at the highest rolling; reduction, it could not arrive at It {123}< 111 >. Such discrepancies between theoretical predictions and experimental observations should be ascribed to the development of dislocation substructures, which were formed by concurrent work hardening and dynamic recovery. Since such development of dislocation substructures are not taken into account in Taylor-Bishop-Hill's theory, it seems that they can not correctly predict the development of rolling textures at very high rolling reductions, i.e. stable end orientations. On annealing specimens rolled above 98% reduction in thickness, cube textures were very weak, suggesting that cube bands were almost completely rotated into other orientations during cold rolling. {325}< 496 >, which lay at an intermediate position between {123}< 634 > and {112}< 111 > along the beta fiber, developed strongly in the recrystallization textures.