Abstract: We consider a risk averse decision maker who dislikes ambiguity as in the Ellsberg urns. We analyze attitudes to ambiguity when the decision maker is exposed to unrelated sequences of ambiguous situations. We discuss the Choquet expected utility, the smooth, and the maxmin models. Our main results offer conditions under which ambiguity aversion disappears even without learning and conditions under which it does not. An appendix analyzes compound gambles within the expected utility model and demonstrates how to rank them.