Abstract: Scene was defined as a semantically coherent view. The human visual system could gather an incredible amount of information about scene in a glance. When a rapid sequence of scenes is presented, subject also could detect accurately a target (Potter, 1975).Why did human can distinguish the target from distractor? Friedeman (1979) thought that scene identity could be inferred from the identification of one or more key objects, so some researches pay attention to the characters of key objects. Some evidences supported that scene could be identified from scene semantic information; however others argue that low spatial frequency information could induce scene identification. Torralba & Oliva et al（2006）presented Contextual Guidance Model that were comprised two parallel pathways, one pathway computed local saliency and other computed global features. The saliency was bottom-up factor and processing of the global features affected by top-down knowledge. In the present study included tow experiments, some scene materials about human safety and eye movement paradigm were used. Experiment 1 examined whether scene recognition affected by target objects or not. A 2(safe scene/dangerous scene)×3（substitution/deletion/identical）design was used.12 participants were asked to learn 18 scenes, and after the learning to test the recognition of change scene or identical scene. Experiment 2 was designed to examine the influence of information attribute on eye movement during the scene recognition. Experiment 1 showed the recognition rate of scene and dwell time on interest area of target location were difference significantly under the three object change conditions. The results implied scene recognition depended on the target objects. Experiment 2 found semantics information could attract attention greatly, but the first fixation time more long if semantics information apart from perception information. So we concluded semantic information have attention prioritization, but it was affected by priming of perception information.

Key words: object changing, information attribute, scene recognition

摘要： 采用眼动研究范式，通过两个实验考察靶刺激变换与情景信息属性对情景再认的影响。实验一结果显示，靶刺激变换对情景再认、靶刺激所属兴趣区的凝视时间均有显著影响，这表明在情景再认过程中，观察者会有意识地搜索靶刺激，靶刺激具有诊断效应；实验二应用了知觉信息与语义信息重合和分离两种情景材料。结果显示，观察者对语义信息的首次注视次数显著多于知觉信息；观察者对知觉信息和语义信息分离条件下语义信息的首次注视时间显著长于重合条件下。这一结果提示，在情景识别过程中，语义信息具有注意优先性，但其优先性会受到知觉信息启动的干扰。

关键词: 靶刺激变换, 信息属性, 情景再认