One of the important issues in silent reading comprehension is whether comprehension is a bottom-up or top-down process. Another issue is what the role of orthography and phonology is on semantic retrieval. Previous researches investigated these issues by using eye movements to measure the reading behaviors. A number of studies attempted to address the issues by controlling the frequency of the words in a sentence, but obtained contradictory results. To reexamine these issues, we manipulated the orthography and phonology of a target word in a sentence and observed readers’ behavior in their silent reading comprehension. During the silent reading, subjects looked through the words surround the target word a number of rounds. The fixation time of the first round implied that the manipulation did not affect the reading speed of either the preceding word or the following word. In the following rounds, more fixation time took place in both the preceding and following words. Accordingly, the silent comprehension process in the early stage seems to be a bottom-up process, while it more likes a top-down process in the latter stage. It is also found that the comprehension process took less look-back counts and shorter reading time in the case of phonology manipulation than that in the orthography manipulation. This phenomenon indicates that phonological information plays a more important role on semantic retrieval for the silent reading comprehension.