认知与发展心理学研究室知识库

最佳注视位置(optimal viewing position, OVP)效应是眼动领域中一种十分有趣的现象，是指首次注视在信息单元（如词语或物体）的中央偏左位置时，信息识别的效率最高，具体表现在反应时较短、发生再次注视的概率较低。目前，OVP效应背后的机制仍存在争议。眼球运动类模型认为它由读者眼球运动神经的限制和知觉经验决定。而认知控制类模型则认为它主要受到实时进行的高水平语言信息加工的影响。以往研究大多通过比较真词和不包含语言信息的假词的OVP效应来试图分离高水平语言因素和低水平视觉因素对OVP效应的贡献，然而至今未得出一致结论。本研究通过设计不同的加工需求水平，巧妙地分离了高低水平因素对OVP效应的影响，得到了稳定的结果。

研究一包含三个实验，采用仅涉及低水平视觉加工的任务（实验1和2要求被试判断词长，实验3要求被试判断词语属于汉语词还是英语词）来考察OVP效应的机制。实验1选择了三组汉语熟练度不同的被试探究汉语熟练度对OVP效应的影响，实验2和3在实验1的基础上加入词语熟悉性变量，结果发现OVP效应完全由低水平眼球运动或视觉因素主导，汉语熟练度和词语熟悉性这些高水平语言变量不影响OVP效应，此结果支持了眼球运动类模型对OVP效应的解释假设。研究二包含两个实验，采用涉及语义加工的高水平任务（实验4要求被试判断刺激材料是真词还是假词，实验5要求被试判断是否认识刺激材料的含义）考察OVP效应的机制，变量的选择与研究一基本相同。结果发现，研究一中OVP效应不受汉语熟练度和词语熟悉性的影响，研究二中不论眼睛首次注视在什么位置，高熟悉性词均表现出较短的反应时和较低的再次注视概率，即高熟悉性词语的OVP效应几乎不存在，但低熟悉性词的OVP效应则非常明显。而且高低熟悉性词的OVP差异仅体现在高汉语熟练度的情况下。总之，高水平语言因素显著影响了OVP效应。这些结果既验证了认知控制类模型解释假设的合理性，同时也不排斥眼球运动类模型对OVP的解释，即高水平语言因素和低水平视觉因素均显著交互影响信息加工效率。

结合两个研究，本论文发现，读者眼球运动的制约/视觉因素和语言因素均对OVP效应有贡献，但贡献量同认知加工需求相关。当读者只进行低水平视觉加工任务时，眼球运动的制约/视觉因素对OVP效应的影响居于主导地位；当读者进行高水平加工任务时，语言因素对OVP效应的影响更大，特别体现在熟悉性高的刺激。这些结果表明眼球运动类模型和认知控制模型在解释OVP效应中均有可取之处，只是适用情境不同。

The optimal viewing position (OVP) effect refers to the longstanding finding that where the eyes initially fixate within an information unit (word or object) influences how easily that word can be recognized. For words presented in isolation, the OVP is a little to the left of the center of words at which both reaction time and refixation probability are minimized. However, there is an ongoing controversy about the contribution to the OVP phenomenon of low level oculomotor strategies and high level linguistic processes. According to oculomotor models, low level oculomotor strategies or visual factors largely determine the OVP effect. On the contrary, cognitive control model holds that high level linguistic variables exert a robust influence on the OVP effect. For this controversy, we conducted five experiments from the perspective of processing depth, and aimed to dissociate the impact of low level oculomotor strategies and high level linguistic processes on the OVP effect.

StudyⅠconsists of three experiments which only involved low-level visual processing. In Experiments 1 and 2, participants were asked to judge whether the stimulus was short or long (Length categorization task). In Experiment 3, participants were asked to judge whether the presented word was English or Chinese (language categorization task). In Experiment 1, three groups of participants with different Chinese proficiency levels were recruited to investigate the influence of language proficiency on the OVP effect. Experiment 2 and 3 explored the effects of both language proficiency and word familiarity on the OVP effect that based on Experiment 1. The first study showed that the OVP effect was stable, and it did not be affected by high-level linguistic factors like word familiarity and language proficiency. Thus, the results supported oculomotor models. In Study Ⅱ, two experiments were conducted to investigate the mechanism of OVP effect by using high-level linguistic tasks. We asked participants to perform lexical decision task in Experiment 4, and asked participants to judge whether they knew certain words (familiarity categorization task) in Experiment 5. This study showed that high level linguistic variables largely affected the OVP effect. Initial viewing positions strongly influenced the processing for unfamiliar words, but not for familiar words. Furthermore, this difference was only true for high language proficiency people. Thus, the results of study both supported oculomotor models and cognitive models.

In summary, both low-level oculomotor strategies and ongoing cognitive factors influence information processing efficiency. When readers performed a ‘shallow’ task which only involved perceptual processes, low-level oculomotor strategies contributed more significantly than high-level linguistic processing for the OVP effect. While readers performed a ‘deep’ task which needed to process the meaning of the stimuli, the results supported both oculomotor models and cognitive models. But the weight of high-level linguistic processing was a little heavier than low-level oculomotor strategies, especially for familiar stimuli. Thus, these results more likely reflect the fact that the OVP effect primarily depends on low-level oculomotor strategies, and high-level linguistic processes may be as modulators of default oculomotor strategies for information processing.