AbstractPhantom decoys are alternatives that asymmetrically dominate a targeted alternative and yet lead to increased selection of the target when the decoy is declared to be unavailable. This effect is difficult to explain within most standard theoretical accounts of decoy effects. The current experiments tested between three explanations of this effect: (1) the relative advantage model based on loss aversion, (2) similarity substitution, and (3) range weighting. In Experiment 1, participants were presented trinary choice sets, with half of the sets containing a phantom decoy in one of five possible locations within the attribute space. Phantom decoy effects were robust across all decoy locations but one, and the pattern of effects most closely corresponded to predictions of the relative advantage model. Experiment 2 used a within‐subjects manipulation of the five phantom decoy locations. The overall pattern of effects most closely corresponded to predictions from the relative advantage model, as did the pattern for the group of participants who exhibited the strongest phantom decoy effects. Copyright © 2007 John Wiley & Sons, Ltd.