Abstract: In the literature, there are two well-known positions concerning how new, incoming information is integrated with prior context during reading for comprehension. These include the immediacy hypothesis (Just & Carpenter, 1980), which proposes that all comprehension processes are immediately executed when a new piece of information in a text is encountered, and the delayed comprehension hypothesis (e.g., Bouma & de Voogd, 1974), which states that integrating the new information with context is delayed until reaching syntactic boundaries.
There is some evidence for the idea that both low- and high-level processes can occur immediately in reading (e.g. Cook & Myers, 2004; O’Brien, Shank, Myers, & Rayner, 1988； Poynor & Morris，2003). However, most previous research on the topic has been done in European languages. Since non-European languages such as Chinese and European languages differ in many important aspects, it is not clear whether the same processes are utilized to comprehend information expressed in the different languages. For example, unlike words in European languages, words in Chinese are not transparent units and do not have inflectional markings to indicate various forms of syntactic information (Chen, 1992). Indeed, given the special properties of Chinese words, some researchers (e.g., Aaronson & Ferres, 1986) have suggested that higher mental processes in reading Chinese would not be expected to function in an immediate manner.
Two experiments were designed to explore which of the two mentioned hypotheses can better account for the time course of integration in reading Chinese text.
Experiment 1
A self-paced, line-by-line, reading paradigm was used, in which participants read through a text by pressing a key each time they wanted a new line of text to appear. The participants were asked to read 12 short passages, each containing a target sentence that fitted well within the local context, but was either inconsistent with or irrelevant to background information from the early part of the text.
Results
A significant slowdown in reading time was observed on the target line in the inconsistent condition relative to the irrelevant, control condition. Also, the same pattern of results was found no matter whether the target line included a syntactic boundary or not, suggesting that integration starts before reaching the end of sentence boundaries.
Experiment 2
An eye-monitoring procedure was used in which Chinese readers’ eye movements were recorded during reading to determine whether the inconsistent information presented at the target sentence could be detected immediately. Two rating studies were conducted to locate critical information in the target sentence that was responsible for the conflict between the sentence and the background in the inconsistent condition. The stimuli and design were similar to those used in Experiment 1.
Results
The participants made longer first-pass fixations and less skipping at the critical regions in the inconsistent condition relative to the control. There were also reliable effects on some eye movement measures that reflected relatively late processes in reading (e.g., second-pass reading times and regression rates at the pre-critical and critical positions), indicating that the participants were trying to use contextual information to resolve the detected inconsistence.
Conclusions
The present findings indicate that higher integration processes can be conducted on-line, rather than delayed, in reading text written in a language with properties that encourage delayed processing, suggesting that immediate processing is likely a universal principle in reading comprehension

Key words: discourse comprehension, Chinese, self-paced reading procedure, eye-movement recording