In this study, we investigate trends in the relationship between what students know and the types of data that capture their attention over time in a science-based multi-user virtual environment. Longitudinal analyses of the patterns of data collected by 143 middle school students (nested within 5 teachers) showed that student prior knowledge was marginally (p < 0.10) related to variation in the attentional value (visibility and location) of data they collected over time, explaining about 2% of said variation. By contrast, accounting for the clustering of students by teacher was statistically significantly (p < 0.05) related to variation in trends in attentional values and explained 36% of said variation.