Linguistic arguments can be either explicitly realized (After she phoned him, …) or left implicit (After she phoned [∅], …). In production, the choice between these options is thought to depend on the contextual predictability of the implied referent. We investigated whether different contextual referents (single vs. multiple vs. underspecified) also affect the comprehension of implicit arguments, using the “maze” variant of self-paced reading. Our results suggest that, rather than predictability, other context-dependent pragmatic effects, such as the perceived genericness of actions, may influence how speakers comprehend implicitly encoded information.