We have previously shown that stable base-pairing at a translational initiation site in Escherichia coli can inhibit translation by competing with the binding of ribosomes. When the base-pairing is not too strong, this competition is won by the ribosomes, resulting in efficient translation from a structured ribosome binding site (RBS). We now re-examine these results in the light of RNA folding kinetics and find that the window during which a folded RBS is open is generally much too short to recruit a 30S ribosomal subunit from the cytoplasm. We argue that to achieve efficient expression, a 30S subunit must already be in contact with the mRNA while this is still folded, to shift into place as soon as the structure opens. Single-stranded regions flanking the structure may constitute a standby site, to which the 30S subunit can attach non-specifically. We propose a steady-state kinetic model for the early steps of translational initiation and use this to examine various quantitative aspects of standby binding. The kinetic model provides an explanation of why the earlier equilibrium competition model predicted implausibly high 30S-mRNA affinities. Because all RNA is structured to some degree, standby binding is probably a general feature of translational initiation.