Abstract: In the present study, visual search in the real world images task was used to explore how visual system integrated scene context and target template to guide eye movement and improve search efficiency. Whether the visual system used scene context and target template simultaneously or not. If scene context and target template were used simultaneously, how did they influence the scanning epoch and the verification epoch; if not, which one was the first to be used by visual system, scene context or target template. In the task, two fixations were displayed in sequence. Between the fixations, a cue (picture or words) identifying the search target was presented. After that, a scene picture was displayed, and the participants’ task was to find the target, look at it and press the space bar on the keyboard. In experiment, participants’ eye movements were monitored by the SMI eye tracking when they were performing the task. For the experiment, the trial event was as follows. Each trial began with a central fixation cross for 200ms, followed by a cue identifying the search target for 800 ms. The cue was either a word identifying the target or an exactly matching picture of the target. The cue was followed by a central fixation point lasting 200ms. Once the fixation was over, the real-world scene appeared. There were three types of test scene: the target in a high probability region (consistency), the target in a low probability region (inconsistency), and the background of scene was segmented into 48 pieces and arranged randomly (random). The experiment was thus a 2 × 3 design with cue type (word vs. picture) as the within-subject variable and scene context (consistency, inconsistency and random) as the between-subject variable. Participants were asked to locate the target as quickly and as accurately as possible, to look at it, and to press Space key as soon as the target was found. Participants were given twelve practice trials prior to the experiment. Results of accuracy and reaction time showed that no speed accuracy trade-off existed, the objects were detected more rapidly and accurately in the consistent scenes context than in the inconsistent and random scenes, the objects were detected more rapidly and accurately when the cue was picture than those when the cue was word, and the interaction between scene context and target template reached the significant level. Eye movement analysis found that the main reasons for improving search performance were the reduction of the duration of scanning phase and verification phase. Futhermore, in the initiation epoch, only scene context influence the initiation time, in the scanning epoch, scene context and target template could simultaneously influence the most eye movement indices, such as, scanning time, average fixation duration, fixation number, and in the verification epoch, scene context and target template also could simultaneously influence the most eye movement indices, such as, verification time and saccade number. These finding suggested that the visual system could use scene context and target template to guide eye movement and improve search performance simultaneously and the time loci of scene context and target template influencing visual search process were different. Specifically, in the initial epoch, only scene context played a guiding role and next these two factors played a role in the scanning epoch and verification epoch. Based on these findings, the authors proposed the interactive model between scene context and target template in the visual search.

Key words: scene context, target template, visual search, eye movements

摘要： 以真实场景图像中的物体搜索为实验任务, 操纵场景情境和目标模板, 采用眼动技术将搜索过程分为起始阶段、扫描阶段和确认阶段, 考察场景情境和目标模板对视觉搜索过程的影响机制。结果发现, 场景情境和目标模板的作用方式及时间点不同, 二者交互影响搜索的正确率和反应时, 仅场景情境影响起始阶段的时间, 随后二者交互影响扫描阶段和确认阶段的时间及主要眼动指标。作者由此提出了场景情境和目标模板在视觉搜索中的交互作用模型。

关键词: 场景情境, 目标模板, 视觉搜索, 眼动