The visual field can be divided into three regions: the foveal region (the central 2°in of vision), the parafoveal region (extending out to 5°on either side of fixation), and the peripheral region (beyond the parafoveal region) during reading. Although readers mainly acquire information from foveal region, effective extraction of information from parafoveally visible words is also essential to skilled reading. While numerous studies have examined the acquisition of visual information form during fixational pauses during reading, little is known about how rapidly useful visual information is acquired from different gaze locations when reading non-alphabetic languages such as Chinese. Therefore, the present study used disappearing text paradigm to investigate the extraction of useful information from papafoveal words to the right (word n+1) and left (word n-1) of fixation in Chinese reading. We manipulated simultaneously both the location (word n+1 and word n-1) and display time (0ms and 40ms) for parpfoveal words. In the disappearing text condition, the word n+1 or word n-1 disappeared a fixed period after the beginning of fixation (0ms or 40ms following the onset of fixation). Eyelink2000was used to record the eye movements of 40 college students. All participants read sentences formed from regular sequences of two-character words and each sentence was between 6 and 8 words long. In order to examine the lexical processing, each sentence containing either a high or low frequency target word. The results suggested that there were some differences in the time needed for information capture between word n+1 and word n-1.Readers could obtain most or all of necessary information of word n+1 for normal lexical and reading processing in the early period of fixation on word n(40ms). However, the information extraction from word n-1 occurred at relatively late period of fixation on word n (later than 40ms). In particularly,(1)serious disruption on overall reading times was caused by word n+1 for 0ms displays compare to control condition, and more important, the disruptions on overall reading times caused by 0ms and 40ms displays for word n+1 disappearing have significant differences. It indicated the readers obtain information of word n+1 during early 40ms of fixation on word n; whereas for word n-1 disappearance, the disruptions on overall reading times caused by 0ms and 40ms were equal. The same tendency showed on fixation count and regression; (2)there was a bigger difference of frequency effect between 0ms and 40ms display for word n+1 than word n-1, which also suggested readers obtain information of word n+1 during early 40ms of fixation on word n, thereby promoting the lexical processing of reading. To conclude, the information of word n+1 and word n-1 were both very important for Chinese reading, and the information was acquired at different period of fixation on word n. Readers could obtain information of word n+1 for normal lexical and reading processing in the early period of fixation on word n, but the information extraction from word n-1 occurred at relatively late period of fixation on word n. These results could be better accounted by processing gradient models in which there is some temporal overlap in the linguistic processing of a fixated and parafoveally visible word during reading.