Social annotation fosters collaborative learning by encouraging knowledge sharing and a community of inquiry. However, research has primarily focused on the cognitive aspect of social annotation. This study aims to contribute an emotional perspective to the existing literature on social annotation. Specifically, we used the valence-aware dictionary for sentiment reasoning algorithm to measure students' emotional tones in 1,954 comments posted during social annotation. We then utilised linear mixed-effect models to examine the effect of emotional tone on students' reading engagement and peer acknowledgement, respectively. Our findings indicate that students who posted more positive sentiment comments were more likely to spend more time engaging in social annotation and receive peer acknowledgement. These findings offer insights into the significance of emotional tone in social annotation and the design of scaffolding strategies to foster positive emotional tone.