研究采用眼動隨動顯示技術，通過操控提示正確/錯誤詞邊界線索的延遲時間，考察漢語閱讀中漢字、詞匯加工與詞切分的時間特性。實驗1發現，提示詞n+1邊界不影響總閱讀時間，但能積極影響詞興趣區內的眼動數據，這種影響隨著提示延遲時間的增加而逐步減弱;提示錯誤的詞n+1的邊界線索對總閱讀時間的影響則隨提示延遲時間的增加呈現倒"U"型的變化趨勢。實驗2發現，提示詞n邊界不影響總閱讀時間，但卻消極影響詞興趣區內的眼動數據，這種消極影響并沒有隨著提示延遲時間的增加而減弱;提示錯誤的詞n邊界對總閱讀時間和眼動數據均產生消極的影響，且這種影響都隨著提示延遲時間的增加而減弱。綜合兩項實驗的結果可以推測，單一的向上反饋假設和整體假設都不能全面解釋閱讀中的字詞加工過程，漢字加工與詞匯加工間存在交互激活過程。

Recently, word segmentation in Chinese reading has become an important research issue, especially concerning how readers segment sequences of characters into words. The feed forward hypothesis assumes that the visual information obtained from Chinese characters is initially fed into a character recognition system, and word segmentation follows after character recognition, leading to word recognition. The holistic hypothesis assumes word segmentation influences character recognition through feedback. According to the interactive hypothesis, word segmentation is an interactive process involving characters and words. In the present study, we investigated the timing of segmentation on word n+1 and word n to test these above mentioned hypotheses. The sentences used in the experiment consisted of 7 to 10 two-character words. These stimuli were balanced in four conditions following a Latin-square design. Experiment 1 contained two sub-experiments(Experiments 1A & 1B). We manipulated delay times in conditions facilitating the segmentation of word n+1 in Experiment 1A. There were four conditions in total: a control and three facilitating-segmentation conditions, in which the color of word n+1 changed from red to black, after word n was fixated for 40/120/160 ms. In the baseline, control condition normal sentences were presented in red. The results showed that none of these facilitating-segmentation conditions of word n+1 promoted sentence reading time, but they did positively influence eye movement data with reliable main delay time effects that did not fit the predictions of the holistic hypothesis. It was not possible to exclude the influence of exogenous attention from the results of Experiment 1A. Therefore, Experiment 1B, which adopted similar conditions as were used in Experiment 1A, was conducted. In this sub-experiment, however, the two adjacent characters not belonging to a word were grouped together, and their color was changed simultaneously. Thus, these manipulations interrupted the word n+1 segmentation. This sub-experiment was also able to test for the three hypotheses mentioned in the first paragraph. The results showed that all the interrupted manipulations negatively influenced eye movement data, and the influences on sentence reading time followed an inverted U shape in the function of delay times in the interrupted word n+1 segmentation conditions. The patterns of eye movements and sentence reading time in Experiment 1B differed from those in Experiment 1A, thus excluding the influence of exogenous attention from the results obtained in Experiment 1A. The results of Experiment 1B did not fit the prediction of the feed forward hypothesis. There were also two sub-experiments(A & B) in Experiment 2, in which the delay times of the conditions facilitating/interpreting the segmentation of word n were manipulated. Four conditions were used in Experiment 2A: a control condition provided as the baseline and 3 facilitating the segmentation of word n after a fixation time of 40/120/160 ms, changing the color from red to black. We found that these facilitating-segmentation of word n conditions do not promote sentence reading time, but that they negatively influence eye movement data without reliable main delay time effects. However, interpreting the segmentation of word n(Experiment 2B) negatively influenced the sentence reading time and eye movement data with reliable main delay time effects. The results of Experiment 2 did not fit the predictions of either the feed forward or the holistic hypothesis. In a word, the results from both experiments(Experiment 1 & 2) indicate an interactive process between character and word.