认知与发展心理学研究室知识库

很多读者在默读文本的过程中，仿佛都能听到来自大脑中的一个声音，也在读着相同的内容，这个现象被称作内部言语。该现象己经得到了大量的研究。在早期，内部言语存在的证据大多来源于自我报告法、内省法，随后实证的研究虽然兴起，但很少是在自然阅读中进行的，尤其是在中文这样一种非字母语言系统中。本研究中，我们设计了两个眼动实验来寻找中文阅读中内部言语存在的证据，除此之外，更要探究内部言语和读者眼动之间的关系，进而说明阅读中内部言语的特征及其作用。

实验一中，将两个两字关键词嵌入到相同的句子框架中组成两个句子，其中一个关键词的发音时间较长，另外一个关键词的发音时间较短。首先要求被试默读这些句子并记录他们的眼动，然后要求被试出声朗读相同的句子并记录他们的发音情况。结果发现，被试在关键词上的发音时长与被试在关键词上的总注视时间、接收回视概率之间存在显著的相关，被试在关键词上的发音时长与被试在关键词后区域的总注视时间存在显著的相关。

实验二中，操纵两字关键词前区域的文本成为饶舌条件(关键词前区域的4-8个汉字中，超过70%具有和关键词的首字相同的声母)，结果发现，这样的操纵会使得关键词的发音变得困难，从而造成发音时间的延长。而在控制条件中，关键词之前区域文本的声母和关键词的首字声母之间没有关联。要求被试默读这些句子并记录他们的眼动。该实验的结果与实验一的结果模式相似，被试在饶舌条件中对关键词的总注视时间显著高于控制条件，在饶舌条件中对关键词后区域的接收回视概率、向外回视概率显著高于控制条件。

实验一表明一个词的发音时长影响了被试注视该词时的眼动轨迹，实验二表明阻碍一个词的发音影响了被试注视该词时的眼动轨迹。两个实验的结果为中文阅读中内部言语的存在提供了有力证据，并且为工作记忆假设提供了支持。两个实验中我们都发现了总注视时间和回视概率上的显著效应，我们推测这是由于读者在阅读的过程中，有一种保持眼动和内部言语同步性的趋势所致，由于内部言语时间过长，读者就需要不断的回视，而读者保持眼动和内部言语同步性可以使他们更好地将信息储存在工作记忆中。此外，读者阅读时内部言语与其发音之间的关联恰恰说明了内部言语与外部言语之间的相似性，在一定程度上对阅读中内部言语的特征提供了解释说明。

Many readers can hear themselves speaking when they read silently. This phenomenon is usually called inner speech and has been widely documented in literature. The existence of inner speech is mainly supported by evidence of self-report and introspection, but only a few experimental studies. This is especially true for Chinese, which is not an alphabetic writing system. In the present study, we reported two eye movement experiments which was designed to seek experimental evidence of inner speech during Chinese reading, and to investigate the relation between eye movements and inner speech, thus, to provide supports for the important function and nature of inner speech during reading.

In Experiment 1，two two-character target words were embedded in a same sentence frame. One of the target words had longer oral reading time and the other had shorter oral reading time. Participants firstly read the sentences silently while at the same time their eye movements were monitored. After this, participants read the same sentences aloud and the time they spoke each word were recorded. We did not find reliable correlation between oral reading time and first-pass reading times on the target word (including first fixation duration and gaze duration). However, we found a reliable correlation between participants' total reading time, regression possibility and their oral reading time on the target region. Words have longer oral reading time also have longer total reading time and higher regression in possibility during silent reading.

In Experiment 2, the oral reading time of the target words was manipulated by varying whether words before the target region could introduce a tongue-twister or not. In the tongue-twister condition, the initial consonants of several characters before the target word were the same as the first character of the target words. It has been reported that this kind of manipulation could introduce tongue-twister, and oral reading time of the target word is prolonged. In the control condition the characters preceding the target word did not share initial consonant with the target words. Participants' eye movements were monitored while they read sentences silently in these two conditions. The results showed the similar pattern as the results of Experiment 1:first-pass reading time on the target word did not differ between the two conditions, while total reading time on the target words were significantly longer in the tongue-twister condition than in the control condition, and regression out possibility on the post-target region in the tongue-twister condition were significantly higher in the tongue-twister condition than in the control condition.

Results of Experiment 1 indicated that oral reading time of a word influences eye movements on it, and results of Experiment 2 indicated that preventing the articulation of one word also influences eye movements on it. Taken together, the results of two experiments showed that oral reading time of a word has strong correlation with the time readers spend on them during silent reading. These results provide clear evidence for the existence of inner speech during silent Chinese reading. Results also showed that oral reading time did not influence first-pass reading time on the target word but affected total reading time. These results seem to suggest that when people are reading, they are eager to pursue the synchronicity of their eye movements and inner speech, but it is hard to do so during first-pass reading. Thus, readers need a regression if inner speech needs too long time. It is likely that the synchronicity of inner speech and eye movements can help people to maintain information in working memory. Besides, our results can provide some interpretation of the nature of inner speech during reading to some extent.