According to the results of previous studies, the gist of a scene can be identified unless than 100msec, and violation of gist can influence the allocation of attention. In other words, people tend to allocate attention to the object that is inconsistent with the gist of a scene. To investigate whether contextual consistency can affect scene perception of the scene perception, two experiments were conducted. In both experiments, a 3x2 factorial design with scene presentation times(2s, 5s, 10s) as between-subject factors, and object-background contextual consistency(consistent, inconsistent) as within-subject factors was used. In experiment 1, participants viewed line-drawing pictures of natural scene in preparation for a memory test. Eye movements were recorded during scene perception. The results of Experiment 1 showed that people make different eye movement depend on contextual consistency. The initial fixation index showed fast to the inconsistent object than consistent one. The eyes more often return to inconsistent objects, and remain fixated longer on that object. This results accords with the results of previous studies. If an object is identified that is inconsistent with the gist of a scene, it is
attract attention. Moreover, the correct recollection rates about inconsistent object were better than consistent object. In particular, the correct recollection rates about inconsistent object’s position information was significantly higher. Experiment 2 was equal to Experiment 1 except that a dual-task paradigm. To allocate divided attention on the inconsistent object, participants must react to the position of probe embedded line-drawing picture every second. However, the similar patterns in Experiment 1 were observed in Experiment 2. Even when the attention was divided on probes deliberately in the divided attention task, contextual consistency still affected scene perception. Implications and the limitations of these results and the future directions of research were discussed in the final discussion section.