Background: Reading researchers have generally considered that reading is an interactive combination of top-down (higher-level language skill) and bottom-up (lower-level language skill) processes. Nevertheless, the mechanisms through which readers apply these skills for online text comprehension are unclear. Methods: The present study thus used eight classical Chinese (CC) texts and their corresponding vernacular translation (VT) texts for controlling the text structure and meaning to explore such mechanisms in high school students. Results: With partial-out to the influences of (a) students' language achievement scores, (b) word frequency and (c) word length, we observed no significant difference in comprehension accuracy between the CC and VT texts, and the CC texts involved a significantly lower reading speed than did the VT texts. Moreover, the first fixation duration, gaze duration, rereading time and total reading time for the CC texts were longer than those for the VT texts. For all events, CC text reading required a longer fixation duration and significantly longer rereading time and total reading time than did VT text reading. Further observations show that students comprehended CC texts by adjusting their lower- and higher-level language skills. Conclusion: These findings demonstrate that even if CC texts contain a relatively high number of low-frequency words, as readers get more and more contextual information from the text, they can gradually apply higher-level reading skills to understand the meaning of the text, which can ease the dependence on word decoding.