Drosophila Armadillo and its vertebrate homolog beta-catenin are key effectors of Wingless/Wnt signaling. In the current model, Wingless/Wnt signal stabilizes Armadillo/beta-catenin, which then accumulates in nuclei and binds TCF/LEF family proteins, forming bipartite transcription factors which activate transcription of Wingless/Wnt responsive genes. This model was recently challenged. Overexpression in Xenopus of membrane-tethered beta-catenin or its paralog plakoglobin activates Wnt signaling, suggesting that nuclear localization of Armadillo/beta-catenin is not essential for signaling. Tethered plakoglobin or beta-catenin might signal on their own or might act indirectly by elevating levels of endogenous beta-catenin. We tested these hypotheses in Drosophila by removing endogenous Armadillo. We generated a series of mutant Armadillo proteins with altered intracellular localizations, and expressed these in wild-type and armadillo mutant backgrounds. We found that membrane-tethered Armadillo cannot signal on its own; however it can function in adherens junctions. We also created mutant forms of Armadillo carrying heterologous nuclear localization or nuclear export signals. Although these signals alter the subcellular localization of Arm when overexpressed in Xenopus, in Drosophila they have little effect on localization and only subtle effects on signaling. This supports a model in which Armadillo's nuclear localization is key for signaling, but in which Armadillo intracellular localization is controlled by the availability and affinity of its binding partners.