Despite its potential, online dialogue (online dialogue) can be superficial. Following Vygotskian (1978) and design experiment approaches (Brown, 1992), this study explores a labelling feature that allows students to tag parts of their messages. Data comes from 4 sessions of a graduate education course. Students engaged in 2-3 graded online activities in groups of 3-4. Students contributed labels for subsequent sessions. Field-notes and descriptive statistics suggested there were 7 labelling groups, 7 non-labelling groups, and 3 difficult-to-categorize groups. Types of labelling use emerged: interactive labelling, elaboration labelling, and interactive elaboration labelling. Labelling correlated to the quality of online dialogue (r=0.410). The Nelson-Denny text-comprehension measure, task-specific motivation, and labelling predicted approximately 25% of the variance in quality online dialogue, F(3,38)=5.149, p less than 0.05; adding labelling was significant. Narrative analyses suggested that some types of labelling more effectively supported online dialogue than did others. Content analyses (n=696 coded "paragraphs") found the interactive elaboration labelling group contributed proportionately more segments coded as critical thinking than did the elaboration labelling group ((M=0.96 vs. M=0.50), especially more analysis and inference. Labelling correlated to performance on the final examination r=0.283. A model including the final, the Nelson-Denny, task-specific motivation, online dialogue marks, and labelling was significant, (F(4, 37)=8.257, R=0.672, R[superscript 2]adj. = 0.415), but adding labelling was not. (Contains 5 figures and 5 tables.)