We implement an experiment to elicit subjects’ ambiguity attitudes in the spirit of Ellsberg's three-color urn. The procedure includes three design elements that (together) have not been featured in similar experiments: Strict ambiguity preferences, a single decision, and a mechanical randomization device with an unknown distribution (to both subjects and experimenters). We use this device in order to eliminate possible ‘strategic’ ambiguity related to subjects’ beliefs about the experimenters’ motivations. In addition, we survey 40 experimental studies on Ellsberg's two- and three-color problems, and find that, on average, slightly more than half of subjects are classified as ambiguity averse. Our results, with our new design, fall on the low end of the range of results in the surveyed studies, and are comparable to a control test where “strategic” ambiguity was induced.