Recent studies have revealed that Chinese words are associated with psychological realities that are important for text comprehension. However, given the lack of spaces between words in Chinese text, readers of Chinese have to segment text into words during reading. Previous studies focused on word segmentation have revealed that this ability to preview and process text develops at a very early stage. However, there was little research on what information or cues were used by readers of Chinese to segment text into words. The assumption was that if readers of Chinese parsed a string of characters into words using a character-to-word processing, preview processing would result in word frequency effects on eye movement data, and that if the readers ascertained word boundary by top-down processing, such as anticipation, preview processing would result in word predictability effects. Four experiments were conducted to check these two hypotheses. In the first three experiments, we assumed that readers of Chinese parsed a string of characters into words using a character-to-word style in preview processing. There were two treatments of Chinese sentences in these experiments; all these experiments had a control condition in which as the nth word was fixated, and no words were masked for baseline comparison. An abnormal display condition was manipulated in Experiment 1, wherein the nth word was fixated, but the words located to the right were all masked by a series of “※,” which deprived the reader of previewing processing. The abnormal display condition was manipulated in Experiment 2 by keeping the nth word fixated, and masking the words located to the right of the n+1th word by a series of “※,” which provided a cue about the boundary of the n+1th word. Given that the results from Experiment 2 cannot exclude the influence of exogenous attention, Experiment 3 was conducted. Experiment 3 adopted a similar treatment as Experiment 2, but two adjacent characters that did not belong to a word were masked together. In first three experiments, the word frequencies (high and low) of target words that were embedded in the frame sentences were also manipulated. In the last experiment (Experiment 4), there were three display manipulations: control condition, a condition in which readers were deprived of previewing processing, and a condition involving the provision of boundary information about the n+1th word; the predictability of target words that were embedded in the frame sentences were also manipulated. The results of Experiment 1 showed that although eye movement data was negatively affected by being deprived of previewing processing, this manipulation did not have any influence on word frequency effects and, therefore, there was no interaction effect between display condition and word frequency manipulations. The results of Experiment 2 showed that providing boundary information of the n+1th word led to less gaze time on the target word, but here again, there was no interaction effect between the display condition and word frequency manipulations. The results of Experiment 3 showed that mask manipulation led to more gaze time on the target word than the control that excluded the influence of exogenous attention on the results of Experiment 2. It also showed that there was no interaction effect between display condition and word frequency manipulations. However, the results of Experiment 4 were interesting; there were some significant interaction effects between the display condition and word predictability manipulations. Specifically, we found that the manipulation in which readers are deprived of previewing processing eliminates predictability effects, and providing boundary information of the n+1th word decreases the discrepancy between words of high predictability and words of low predictability. The results of the first three experiments indicate that Chinese word processing in the preview cannot reach the desired vocabulary level, and that it must undergo character processing. Therefore, it was difficult for readers of Chinese to complete the segmentation of the n+1th word through a bottom-up characters-to-word processing. The results of Experiment 4 show that deprivation of preview processing can eliminate predictability and providing boundary information of the n+1th word can lessen predictability thereby indicating that top-bottom processing, such as anticipation, is an important reference cue to readers of Chinese to segment the n+1th word.