In a decision problem comprised of multiple intermediate choices, subjects may fail to take into account the interdependencies between their choices. We design and deploy a novel experiment to understand how people make decisions in such problems. We provide revealed preference tests of three models of choice bracketing: broad, narrow, and partialnarrow. We apply these tests in three experiments to determine how subjects bracket in portfolio allocation under risk, social allocation, and induced-utility shopping experiments. 40-44% of our subjects are consistent with narrow bracketing, while only 0-15% are consistent with broad bracketing. Classifying subjects while adjusting for models' predictive precision, 75% of subjects are best described by narrow bracketing, 14% by broad bracketing, and 3% by intermediate cases.Neoclassical economic theory assumes that agents broadly bracket their choices, "assessing the consequences of all of them taken together" [Read et al., 1999]. Behavioral economics usually takes the other extreme view, and interprets experimental subjects' choices as though they narrowly bracket their choices by ignoring all the other choices they face inside or outside of the lab. For example, experiments measuring risk aversion assume that subjects ignore all wealth accrued from decisions outside of the experiment (e.g. Tversky and Kahneman 1992;Holt and Laury 2002). Similarly, those that measure social preferences rely on subjects bracketing equity considerations only in terms of lab payments, ignoring all other people and other decisions outside of that interaction (e.g. Fehr and Schmidt 1999). 1 Since multiple interrelated decisions are the rule rather than the exception, choice bracketing plays an important role in predicting and interpreting behavior. For instance, leading behavioral models, like loss aversion, make predictions that are sensitive to how a person brackets