A simple morphogen gradient based on the protein bicoid is insufficient to explain the precise (i.e., similar in all embryos) setting of anteroposterior gene expression domains in the early Drosophila embryo. We present here an alternative model, based on quantitative data, which accounts for all of our observations. The model also explains the robustness of hunchback boundary setting in unnatural environments such as published recently [Luccheta, Nature 434, 1134 (2005)]. The model is based on the existence of a secondary gradient correlated to bicoid through protein degradation by the same agent.