An Eye Movement Study For The Guidance Mechanism Of Long-range Regressions In Chinese Reading

Regression is an important indicator which can reflect the reading situation.According to the regressing distance,the regression is split into two categories,long-regression and short-regression.And some researchers propose the spatial coding hypothesis and the verbal reconstruction hypothesis to explain the guiding mechanism of long-regression.Both the spatial coding hypothesis and the verbal reconstruction hypothesis have their rationality and are supported by some researches,but there are some disputes between the two theories.The spatial coding hypothesis suggests that the regressing eye movement is guided by spatial memory,but verbal reconstruction hypothesis suggests that the regressing eye movement is guided by verbal memory.In addition,the current studies have no definite conclusion on the question which is what guide the corrective saccade.Therefore,in order to solve these problems and improve relevant theories,the present study adopt some measures of regressions(initial regression size,initial regression error,cumulative regression size and regression reaction time)to explore the guidance mechanism of long-range regressions in Chinese reading through the following experiments.Experiment 1 used 2(Regression Condition: Regression with Reading,Regression without Reading)× 2(Word Frequency: High-Frequency Word,Low-Frequency Word)× 3(Word Location: First Third of the Sentence,Middle Third of the Sentence,Last Third of the Sentence)of the hybrid experimental design to explore the effect of memory on regressions.The regression condition is the variable between subjects,and the word frequency and word location are the variables within subjects.The results showed that initial regression error was lower in the reading group,and the reading group's initial regression size was significant in different word location conditions,but the non-reading group not.Word location was significant in all regressing measures.On the contrary,word frequency was not significant in all regressing measures.Experiment 2 used 2(Regression Visibility: Visible,Invisible)× 2(Word Frequency: High-Frequency Word,Low-Frequency Word)× 3(Word Location: First Third of the Sentence,Middle Third of the Sentence,Last Third of the Sentence)of the hybrid experimental design to explore the effect of text visibility on long-regressions.The regression visibility is the variable between subjects.The experimental procedure is similar to the reading group,except that the invisible group couldn't see the sentence which they have read before when they did the regressing task.Because the sentence was masked by ‘?'.The results showed that the initial regression error in the invisible group was smaller.Because the invisible group spent more time reading the sentences.Initial regression size was larger in the invisible group and response time also was slightly longer.But there was no significant difference in cumulative regression size between the two groups.The results of word frequency and word location were consistent with Experiment 1.Experiment 3 used 3(Target condition: Original Word,Synonym,Non-Word with Similar Form)× 3(Word Location: First Third of the Sentence,Middle Third of the Sentence,Last Third of the Sentence)of the design of the subject to explore the guidance mechanism of corrective saccades.The experimental procedure was similar to the reading group,except that target words in regressing sentences were no longer only use original words.The results showed that the cumulative regression size of non-word with similar form was the smallest in three target condition.There was no significant difference in cumulative regression size between original word and synonym.But response time was smaller in original word than synonym.Word location also was significant in all regressing measures.The experimental results proved that long-regressions were guided by memory in Chinese reading.Initial regression showed spatial selectivity,so it guided by spatial memory,which supported the spatial coding hypothesis.Corrective saccades need text visual information and verbal memory to guide.