Background: Collaborative reading can facilitate students' understanding of complex learning materials. High-quality annotations provided by peer learners are essential for successful collaborative reading. However, it remains to be understood how annotation quality affects reading comprehension. Objectives: A simulated collaborative reading environment was constructed for interactive highlighting and commenting. The environment was used to conduct a laboratory experiment and examine the benefits of reading annotations by others. Methods: Forty-seven college students participated in the study and were randomly assigned to two conditions. The constructive annotation group was provided with annotations containing information that pertained to the constructive level of cognitive processing, while the active annotation group was supplied with annotations containing information that pertained to the active level of cognitive processing. Pre- and post-tests were used to measure learning gains. System log files and eye-tracking data were analysed to compare and contrast the differences in underlying processing. Results and Conclusions: The results showed that constructive and active annotations incurred different reading patterns, with the constructive annotations attracting more attention than the active annotations. The results implied that instructors could use seed annotations of a constructive level in reading assignments to facilitate reading comprehension.