Unintended hedging in ambiguity experiments

•We point out a potential hedging problem in standard Ellsberg-type experiments.•This problem may yield an incorrect classification of ambiguity averse subjects.•Subjects in experiments can eliminate all ambiguity in expected payoffs.•We propose a new classification for Ellsberg urns.

We describe an ambiguity hedging problem in Ellsberg experiments, where combinations of individually ambiguous bets eliminate aggregate ambiguity, and which may yield incorrect classifications of ambiguity averse subjects. We propose a new classification consistent with this hedging possibility.