Verbs in the passive voice have been described as linguistic features which can exert ideological effects because they typically omit the agent, or ‘doer’ of an action. These missing agents are said to be potentially ‘mystified’ to readers. But to what extent are these agents actually mystified to readers, and can we predict when mystification is likely to occur? A text-based analytical framework for inference prediction focusing on agency mystification was applied to two corpora of US newspaper editorials. Eighty-four percent (...) of passive verbs were predicted not to mystify agency for casual readers. To test this method of inference prediction, a reader response study was carried out in which readers were asked to identify agents that had been deleted from texts via the addition of agentless passives. By a standard of 50% accuracy, 74% of passives were found to be non-mystifying. These results suggest that agents missing from texts are likely to be inferred correctly by readers in a majority of cases and that mystification may be predictable from text analysis. (shrink)