During initiation of eukaryotic translation, the eIF3 complex acts as a scaffold that mediates recruitment of the 40S ribosomal subunit to the 5′ UTRs of cellular mRNAs. However, eIF3's functions seem to go beyond this general scaffolding role, and eIF3 promotes translation of hepatitis C viral RNA by directly binding a structured RNA element. Furthermore, eIF3 affects cellular differentiation and growth and is linked to various cancers, thus suggesting that it may direct the translation of specific cellular mRNAs. Cate and colleagues now provide PAR-CLIP analyses demonstrating that 4 of the 13 eIF3 subunits indeed interact directly with RNA. Genome-wide sequencing revealed that eIF3 was predominantly associated with the 5′ UTRs of mRNAs involved in cell growth processes. JUN and BTG1 5′ UTRs were among the top eIF3-binding targets, and reporter assays showed that, as expected from eIF3's canonical function, the eIF3-binding site of JUN was required for translation. Surprisingly, BTG1's eIF3-binding site conferred translational repression. To understand how eIF3 could act as a positive or negative regulator of translation, the group used SHAPE analyses and showed that the eIF3-binding sites occurred in conserved stem-loop structures. Although eIF3 could bind the JUN 5′-UTR stem-loop directly, it appeared to require additional factors to bind the BTG1 5′ UTR, thus suggesting that eIF3's modes of binding to the RNA stem-loops may underlie its ability to activate or repress translation. Thus, this study uncovers a new form of translational control for cellular mRNAs and describes an unexpected dual role for eIF3. (Nature doi:10.1038/nature14267, 6 April 2015).