The development of live viral vaccines relies on empirically derived phenotypic criteria, especially small plaque sizes, to indicate attenuation. However, while some candidate vaccines successfully translated into licensed applications, others have failed safety trials, placing vaccine development on a hit-or-miss trajectory. We examined the determinants of small plaque phenotype in two dengue virus (DENV) vaccine candidates, DENV-3 PGMK30FRhL3, which produced acute febrile illness in vaccine recipients, and DENV-2 PDK53, which has a good clinical safety profile. The reasons behind the failure of PGMK30FRhL3 during phase 1 clinical trial, despite meeting the empirically derived criteria of attenuation, have never been systematically investigated. Using in vitro, in vivo and functional genomics approaches, we examined infections by the vaccine and wild-type DENVs, in order to ascertain the different determinants of plaque size. We show that PGMK30FRhL3 produces small plaques on BHK-21 cells due to its slow in vitro growth rate. In contrast, PDK53 replicates rapidly, but is unable to evade antiviral responses that constrain its spread hence also giving rise to small plaques. Therefore, at least two different molecular mechanisms govern the plaque phenotype; determining which mechanism operates to constrain plaque size may be more informative on the safety of live-attenuated vaccines.