This paper explores how a change in a default—specifically, an exogenously given reference point—affects individual preferences. While reference dependence is extensively studied, very little is known regarding the impact of reference points on individual choice behavior when the reference points themselves are not chosen (Reference Effect). We identify critical properties that differentiate between classes of reference-dependent models and test them. We find that the reference effect exists for asymmetrically dominated reference points, but we do not see any evidence of a reference effect for symmetrically dominated reference points. Some of the existing models are mostly consistent with our data but lack predictive power. None of the models offers the particular predictions that our experiment suggests. Finally, we also tease apart the differences between the reference effect and the asymmetric dominance effect (decoy effect), a well-known phenomenon observed in the literature on context-dependent choice.