This study examined the impacts of online graduate students categorizing their online discussion contributions according to an instructor's criteria. Student messages before and after this categorizing activity were unitized by terminal punctuation; classified for function, skill, and level; and analyzed based on Henri and Rigault's (1996) content analysis framework. There was an unexpected decrease in the quantity of messages following the treatment. Even more surprising was the decrease in four quality measures of cognitive dialogue: percentage cognitive unit among four functions, percentage inference and percentage analysis or elaboration (or in-depth clarification) among five cognitive skills, and percentage high-level processing. This treatment seems to have inhibited cognitive dialogue, although qualitative data suggested that there may have been some advantages for some students.