The current model of measles virus (MV) pathogenesis implies that apical infection of airway epithelial cells precedes systemic spread. An alternative model suggests that primarily infected lymphatic cells carry MV to the basolateral surface of epithelial cells, supporting MV shedding into the airway lumen and contagion. This model predicts that a mutant MV, unable to enter cells through the unidentified epithelial cell receptor (EpR), would remain virulent but not be shed. To test this model, we identified residues of the MV attachment protein sustaining EpR-mediated cell fusion. These nonpolar or uncharged polar residues defined an area located near the binding site of the signaling lymphocytic activation molecule (SLAM), the receptor for MV on lymphatic cells. We then generated an EpR-blind virus maintaining SLAM-dependent cell entry and inoculated rhesus monkeys intranasally. Hosts infected with the selectively EpR-blind MV developed rash and anorexia while averaging slightly lower viremia than hosts infected with wild-type MV but did not shed virus in the airways. The mechanism restricting shedding was characterized using primary well-differentiated human airway epithelial cells. Wild-type MV infected columnar epithelial cells bearing tight junctions only when applied basolaterally, while the EpR-blind virus did not infect these cells. Thus, EpR is probably a basolateral protein, and infection of the airway epithelium is not essential for systemic spread and virulence of MV.