Protein folding can begin co-translationally. Due to the difference in timescale between folding and synthesis, co-translational folding is thought to occur at equilibrium for fast folding domains. Thus, the folding kinetics of stalled ribosome-bound nascent chains should match the folding of nascent chains in real time. We test this assumption by comparing the folding of a ribosome-bound, multi-domain calcium-binding protein stalled at different points in translation with the nascent chain as is it being synthesized in real-time, via optical tweezers. In vitro, a misfolded state of the protein occurs readily, and on stalled ribosomes, the misfolded state still forms rapidly (1.5 s). Surprisingly, during active translation, this state is only attained after a long delay (~ 60 s), indicating that, unexpectedly, the growing polypeptide is not equilibrated with its ensemble of accessible conformations. Slow equilibration on the ribosome can delay premature folding until adequate sequence is available and/or allow time for chaperone binding, thus promoting productive folding. On the other hand, interactions between the nascent polypeptide and the ribosome exit tunnel represent one mode of regulating synthesis rates, which, in turn, are thought to affect protein folding. The SecM protein arrests its own translation as part of a feedback mechanism, and release of arrest has been proposed to occur by mechanical force at the translocon. This is the so-called translocon mechanical hypothesis. Using optical tweezers, I demonstrate that arrest of SecM-stalled ribosomes can indeed be rescued by force alone. Moreover, I will show that the force needed to release stalling can be generated in vivo by a nascent chain folding near the ribosome tunnel exit. I formulate a kinetic model describing how a protein can regulate its own synthesis by the force generated during folding, tuning ribosome activity to structure acquisition by a nascent polypeptide.