This article addresses the conceptualization and definition of message variables in persuasion effects research. Two central claims are advanced. First, effect-based message variable definitions (in which a message variation is defined in terms of effects on psychological states, as when fear appeal variations are defined on the basis of differences in aroused fear) impede progress in understanding persuasion processes and effects and hence should be avoided in favor of definitions expressed in terms of intrinsic message features. Second, when message variations are defined in terms of intrinsic features, message manipulation checks, under that description, are unnecessary but similar measures may usefully be understood and analyzed as assessments of potential mediating states.