认知与发展心理学研究室知识库

眼动是人类主动获取信息的重要手段，阅读是人类获取知识和技能的重要途径之一。理解阅读中的眼动控制机制，是现代认知心理学研究的一项重要的任务。对拼音文本的研究表明，词的加工对眼动控制，尤其是眼动目标的选择起着非常重要的作用，读者在选择眼跳的目标时，倾向于将眼跳落在词的中央位置附近。但是中文文本中没有词之间的空格，在中文阅读中词的加工对眼动目标选择的影响可能并不如拼音文本显著， 我们认为中文阅读中的眼动目标选择更有可能是基于加工效率的，具体来说，读者在每一个注视点上会尽可能多地加工信息，当一个注视位置的加工效率降低到一定程度后， 他们会将眼跳跳向未处理的信息位置。本研究系统地验证和探讨了基于加工效率的眼跳目标选择策略。
研究一通过两个实验对基于加工效率的眼跳目标选择策略进行了验证。 根据这个策略的预测，如果当前注视区域内的信息加工容易的话，向右的眼跳长度会更长。实验1 中，我们分析了 4字的兴趣区，在单词条件下，兴趣区中的 4 个字是一个成语，在双词条件下，兴趣区中的 4 个字是两个双字词组成的词组。结果发现，离开兴趣区的眼跳长度在单词条件下要比双词条件长。实验2中，兴趣区是一个 2字的高频词或者低频词，同样发现了离开高频词的眼跳长度要比低频词更长。这些结果支持了我们提出的基于加工效率的眼跳目标选择策略。
研究二进一步验证该策略， 同时探讨了读者对右侧信息加工到什么程度时便会产生眼跳。实验3 采用了边界范式进行研究，当读者的眼睛跨过一个不可见的边界时，整个句子消失，同时在跨过边界后的注视点的正下方呈现一个汉字。该汉字是边界后的第一个字或第二个字，或者是与他们字形相似或语音相似的字，或者是非字。读者的任务是判断这个字是真字还是非字。根据基于加工效率的眼跳目标策略的预测，如果读者在眼跳前对右侧信息加工的更深，则眼跳长度更长，落点位置更远；同时，如果对右侧信息加工程度更深，则反应时判断的时间会更短。因此，眼跳长度会对字的判断的反应时产生影响。实验结果发现，当给读者呈现边界后第一个字时，仅在语音相似条件中发现了落点位置的影响，眼跳长度的增加促进了读者对语音相似的字的判断。当给读者呈现边界后第二个字时，在相同字和字形相似条件中发现了落点位置的影响，随着眼跳长度的增加，读者的判断时间是减少的。该实验结果首先证明了基于加工效率的眼跳目标选择策略是可靠的，其次说明读者加工注视区域内信息时，并不是对区域内信息进行完全的加工，而是对这些信息进行部分加工，并估计他们能加工多少信息，从而将眼睛跳向无法加工到的位置。对这些信息的部分加工包括了正字法信息和语音信息，且语音信息的加工是滞后于正字法信息加工的。
研究三则探讨了眼跳计划的完成时刻。在拼音文本的阅读中的眼动模型均假设眼跳是通过两个阶段完成的。在不稳定阶段，眼跳的计划是可以取消的；而在稳定阶段，眼跳的计划不能被取消，而一定会被执行。实验 4 通过停止信号任务与阅读任务的结合探讨了中文阅读中眼跳的计划在什么时刻就不能被取消了。 当读者的眼睛跨过一个不可见的边界后的一段时间内，给读者呈现了听觉信号，要求读者在听到这个信号时，立刻停止眼跳，将眼睛停留在当前的注视位置。结果发现，中文阅读中读者的眼跳计划不能被取消的阈限值在105毫秒左右，且不受到词频的影响。
总之，在中文的阅读中，读者会采用一种基于加工效率的策略来选择眼跳的目标。具体来讲，读者在每一个注视点上会尽可能多地加工信息，当一个注视位置的加工效率降低到一定程度时，眼睛将移动到未处理的信息位置。本研究的结果有助于建立适合中文阅读的眼动控制模型， 进而揭示跨语言的阅读中可能存在的统一的眼动控制机制，加深我们对阅读的认知机制的理解。

In alphabetic scripts,  the interword spaces can guide eye movement target selection, so readers adopt a word-based saccade target selection strategy. They usually select word center as their saccade target. However, in Chinese reading system, there are no interword spaces. This  dissertation  proposed  a processing-based strategy for saccade target selection in Chinese reading in which  readers estimate how many characters they can process on each fixation, and then program their next saccade so that the eyes fixate somewhere beyond.  In this  dissertation,  the processing-based saccade target selection strategy was examined in the following three studies.
In Study 1, whether properties of words in a fixated region affect the length of the outgoing saccade was investigated. According to the processing-based saccade target selection,  the easier the processing  of the fixated region is, the longer the outgoing saccade will be. In Experiment 1, the critical region was a 4-character string which was either a word (one-word condition) or  two 2-character word phrases (two-word condition). In Experiment 2, the critical region was either a high frequency word or a low frequency word. The outgoing saccade length from the last fixation on the critical region was longer in the one-word condition than the two-word condition in Experiment 1 and was longer in the high frequency condition than in the low frequency condition in Experiment 2. The results were consistent with the prediction of processing-based saccade target selection strategy.
In Study 2, the depth of reader processing the parafoveal information on a fixation was examined. A variant of the boundary paradigm was introduced in Experiment 3. In this paradigm, after the participant’s eyes crossed an invisible boundary, only a target character remained on the screen, displayed at the saccade landing site. The target character was  either identical, orthographical or phonological similar to the first or second character after the boundary, or was a non-character. Readers were required to judge whether the target characters were real characters. The results showed that, when the first characters after the boundary were shown to readers, their reactions times (RTs) to the target characters were affected by landing position only in phonological similar condition. RTs were shorter as the landing position went further. When the second characters after the boundary were shown to readers, their RTs to the target characters were affected by landing position in the identical and orthographical similar condition.
RTs were shorter as landing position went further. These results are consistent with the predictions of a processing-based strategy. When a saccade is executed, the processing depth of variant properties of parafoveal information is different, which indicated that readers might program a saccade based on partial computation of information in fixated material.  
In Study 3, the time at which the saccade program cannot be inhibited and whether this time would be affected by word frequency was explored.  The paradigm in Experiment 4 combined  stop signal  paradigm with  reading  task. When readers’  eye moved across an invisible boundary, a stop sound was played after a delay. Readers were required to stop their eye movement when the sound was played. The results showd that the time for saccade programming  till labile stage was about 105ms, and was not influenced by the word frequency of their fixated word. Readers cannot cancel or change the saccade program after this time.  
In  conclusion,  this  dissertation  confirmed that Chinese readers do use a processing-based saccade target selection strategy. Readers  estimate how many characters they can process on each fixation, and then program their next saccade so that the eyes fixate somewhere beyond them. This dissertation will help us develop a model that is best suitable for Chinese reading, and also improve our understanding on the mechanisms of eye movement control in Chinese reading and further the general mechanisms of reading among all the languages.