"Collaborative video annotation" (CVA) allows multiple users to annotate video and create a discussion asynchronously. This dissertation investigates 14 small-group CVA discussions held on YouTube in a pre-service teacher education course. Fourteen groups of 6-12 pre-service teachers (141 total) participated. Five of these groups (48 students) used a CVA platform where models of expert discourse had been planted into the video-the modeled condition. Five other groups (49 students) participated in two stages rather than one, without planted discourse models-the staggered condition. Four groups (44 students) annotated the videos without either design modification-the basic condition. A content analysis system, developed using published research in the analysis of asynchronous discussions, was used to measure substance frequencies, including three types of higher order thinking (HOT) present in annotations. Two other aspects of discourse were quantified: reference and judgment. Frequencies of HOT in the three conditions and joint occurrences of HOT + judgment and HOT + references were analyzed to provide insight about HOT in CVA as well as a comparison of the three conditions of instruction. In all three conditions, the most frequent substance type was the most concrete, observations. The modeled condition produced the most HOT, and the basic the least. In all three conditions, reference to the video was most frequent. However, reference percentages were more balanced in groups with design modifications, staggering participation and including discourse models. Analysis was more closely tied to other learners' annotations that the other types of HOT. HOT +Judgment measures did not differ across conditions, but different HOT types were associated with different kinds of judgment. The tendency for applications to include highly critical teaching suggestions exposed preconceived notions about teaching held by novices that differed from those of experts. A sequential analysis revealed that HOT appeared more often in chains than in isolation, and these chains appeared at higher frequencies in the conditions with design modifications-the most in the modeled discourse condition. The chains of HOT also uncovered that the act of scrolling back and re-watching video frequently correlated with HOT. This study provides evidence that HOT types and discourse tenor are associated. Analysis was found to be associated with positive judgments, applications were critical, and intellectual modesty was more often comprised of mixed judgments. HOT + Reference frequencies imply that designers should be aware of the relationship between HOT and learner annotations while understanding that the majority of annotations in CVA will likely be tied to video content. Furthermore, the design decisions that situated learners in discussions already in progress, made possible by the CVA mode, resulted in more HOT from learners. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]