Abstract: Inference provided necessary insurance for correct comprehension and fluent reading. A prevailing way to clarify the meaning of inference was through its function for comprehension (i.e., inference for local coherence, inference for global inference, and predictive inference). Based on the evidence from narrative text experiments, Minimalist Hypothesis, Memory-based text processing view and Constructionist theory had different predictions on the cognitive process and properties of the generation of inference for global inference and predictive inference. However, they had a clear consensus that inference for local coherence could be coded on-line during reading. In view of both agreement between opposing theoretical positions and the evidence from narrative studies, it was somewhat surprising that evidence for the generation of local causal inference is not clear when comprehension of expository text was considered. For expository text on highly technical topics, participants weren’t able to produce online scientific causal inferences for local coherence expect when inferential strategies were explicitly encouraged; whereas for expository text on more familiar topics, participants generated causal inferences to maintain local coherence. Thus it raised a question that whether the construction of causal inference for local coherence could be realized voluntarily in expository comprehension like that in the narrative reading. This research was designed to throw some light on the question.
Three experiments were conducted. Experiment 1 was to explore whether participants could compute immediately causal inference for local coherence while all necessary premises were presented explicitly and compactly. In experiment 2, necessary premises were expressed implicitly with two examples. Experiment 3 varied the availability of the premises by manipulating the distance between critical pieces of information. Participants read the short expository on familiar topic and finished a probe task after reading each passage. All materials were presented on a monitor controlled by computer. Participants read the passages in a self-paced manner, advancing the text one line at a time by pressing the space bar. Participants were instructed to read carefully so that they would be able to judge whether the probe word appeared in the text. The reading times of the line containing the conclusion were recorded and analyzed.
The results showed that in the reading of expository on familiar topics, the causal inference for maintaining local coherence was computed spontaneously when the conclusion was followed by all necessary premises, being presented explicitly or implicitly; the causal bridging inference was impaired by separating the premises and the conclusion with two sentences.
The present findings indicated that the construction of causal inference for local coherence in expository was differ from that in the narrative comprehension. Also, it implied that the activation of causal inference to maintain local coherence might be a strategic process followed by the principle that costing little recourse to means much. We argued that with such principle, it might provide a relatively general interpretation for different conclusions from the researches of causal inference in narrative and expository text comprehension.

Key words: expository, causal inference, local coherence