We extend Ellsberg’s two-urn paradox and propose three symmetric forms of partial ambiguity by limiting the possible compositions in a deck of 100 red and black cards in three ways: within an interval, within two disjoint intervals, and to two points. This paper experimentally investigates attitudes towards partial ambiguity and the corresponding compound lotteries in which the possible compositions are drawn with equal objective probabilities. We arrive at three key findings: (1) aversion to increasing the number of possible compositions in interval and disjoint lotteries, (2) significant association between attitudes toward partial ambiguity and compound risk, and (3) source preference over two-point ambiguity and two-point compound risk. The observed choice behaviour is compatible with multiple-prior models (e.g., maximin expected utility), but they cannot account for the observed link between attitudes toward partial ambiguity and compound lottery. By contrast, two-stage models (e.g., recursive rank-dependent utility) can account for much of the observed behaviour except for Key Finding 3. The overall findings help discriminate among ambiguity models in the literature.