Abstract: A generalized weak dominance approach is used to test choice reversals across certainty, uncertainty and risk. In the case of pairwise choice where each alternative is generally better than the other on a single dimension, this approach models much human choice behavior as a process in which people seek to equate smaller difference between alternatives on one dimension, thus leaving the greater one-dimensional difference to be differentiated as the determinant of the final choice. The choice reversals are therefore seen as a consequence of the fact that what is seen as the greatest one-dimensional difference on one trial is no longer seen as the greatest on another trial. A “matching” task was designed to examine whether the knowledge of the value difference of the paired outcomes along each dimension will permit prediction of preferential choice. The overall test-retest results for various choosing tasks favor the equate-to-differentiate explanation. The finding supports the claim that the repeated choices can be consistent not because the chosen alternative is always of the greatest overall worth but because final choice is consistently based on a single fixed dimension on each trial.

Key words: repeated choices, choice reversals, weak dominance