Ribosomes can stall during translation elongation in bacteria and eukaryotes. To identify mechanisms by which ribosome stalling affects expression of the encoded protein, we develop an inverse approach that combines computational modeling with systematic perturbations of translation initiation rate, the number of stall sites, and the distance between stall sites on a reporter mRNA. By applying this approach to ribosome stalls caused by amino acid starvation in the bacterium Escherichia coli, we find that our measurements are quantitatively inconsistent with two widely used kinetic models for stalled ribosomes: ribosome traffic jams that block initiation, and abortive (premature) termination of stalled ribosomes. To account for this discrepancy, we consider a model in which collision from a trailing ribosome causes abortive termination of the stalled ribosome. This collision-stimulated abortive termination model provides a better fit to measured protein synthesis rates from our reporter library, and is consistent with observed ribosome densities near stall sites. Analysis of this model further predicts that ribosome collisions can selectively stimulate abortive termination of stalled ribosomes without fine-tuning of kinetic rate parameters. Thus ribosome collisions may serve as a robust timer for translational quality control pathways to recognize stalled ribosomes.