The Role of Format Familiarity and Word Frequency in Chinese Reading

For Chinese readers, reading from left to right is the norm, while reading from right to left is unfamiliar. This study comprises two experiments investigating how format familiarity and word frequency affect reading by Chinese people. Experiment 1 examines the roles of format familiarity (reading from left to right is the familiar Chinese format, and reading from right to left is the unfamiliar Chinese format) and word frequency in vocabulary recognition. Forty students read the same Chinese sentences from left to right and from right to left. Target words were divided into high and low frequency words. In Experiment 2, participants engaged in right-to-left reading training for 10 days to test whether their right-to-left reading performance could be improved. The study yields several main findings. First, format familiarity affects vocabulary recognition. Participants reading from left to right had shorter fixation times, higher skipping rates, and viewing positions closer to word center.. Second, word frequency affects vocabulary recognition in Chinese reading. Third, right-to-left reading training could improve reading performance. In the early indexes, the interaction effect of format familiarity and word frequency was significant. There was also a significant word-frequency effect from left to right but not from right to left. Therefore, word segmentation and vocabulary recognition may be sequential in Chinese reading.


Introduction
Chinese is a typical ideographic language containing no spaces, thus differing from alphabetic languages such as English (Ma & Chuang, 2015).The basic unit of reading processing, whether in Chinese or English, is the word (Bai et al., 2008;Carrol & Conklin, 2014;Chen et al., 2021).For Chinese readers, the primary task while reading is word recognition (Chen et al., 2021;Fan & Reilly, 2020;Inhoff et al., 2000;Liang et al., 2017;Liu & Lu, 2018;Wang et al., 2018).The reader needs to segment the word from the text and recognize the whole word (Inhoff et al., 2000;Li et al., 2009; Participants Forty undergraduate students participated in Experiment 1, which contained 31 females and 9 males (mean age 20.50 ± 1.63 years).All of them are right-handed and native speakers of Chinese with normal or corrected-to-normal vision.Signed informed consent was obtained from each participant prior to the experiment.The Medical Ethical Committee of the author's university approved the experiment as compliant with the Declaration of Helsinki.

Design
The experiment had a 2 (format familiarity: reading from left to right, reading from right to left) × 2 (word frequency: high frequency, low frequency) within-subjects design.Four blocks were constructed, each containing 96 sentences.There were 24 sentences in each condition, and the conditions were rotated across files according to a Latin square.Sentences in each condition were presented in a blocked format, and the order of the sentences in each block was random.Twelve practice sentences, three for each condition, were included at the beginning of each experimental block.In addition, there were 24 filler sentences (six in each condition) that appeared randomly throughout the block.After each of the 22 sentences, a yes/no comprehension question was presented.In total, each participant read 132 sentences.
It generally took 15-20 minutes for each participant to complete this task.

Materials
Drawing on a published lexicon database developed by Cai and Brysbaert (2010), we selected 48 pairs of two-character words as target words.The unit of word frequency and character frequency was the occurrences per million words (OPM).Each word pair included an HF word and an LF word.The Journal of Eye Movement Research Chen, M. & Lu, J. ( 2023) 16(4):5 The role of format familiarity and word frequency 4 frequency difference between HF and LF words was reliably different.Numbers of strokes were matched for the HF and LF conditions.There was a marginally significant difference in first-character frequency between the HF and LF conditions.Table 1 shows the means and standard deviations.
We next constructed 96 Chinese sentences containing the target words but not including them at the beginning or end of the sentence.Each sentence was presented from left to right and from right to left.
Sentences were between 16 and 23 characters in length (M = 19.40characters, SD = 1.33).The experiment included four conditions: HF-left to right, HF-right to left, LF-left to right, LF-right to left (see Table 2).The experimental sentences were rated for naturalness, meaning the extent to which they are fluent, easy to understand, and compliant with norms (Bai et al., 2008).Ratings were given on a 7point scale by 40 individuals who did not participate in the eye-tracking experiment.The predictability test required another 28 participants to provide the following words given the sentence context before the target words.Naturalness and predictability were matched for HF and LF, ts < 1.   Note: Under the high frequency -familiar format, the high-frequency target words were presented from left to right.
Under the high frequency -unfamiliar format, the high-frequency target words were presented from right to left.In the low frequency -familiar format, the low-frequency target words were presented from left to right; and in the low frequency -unfamiliar format, the low-frequency target words were presented from right to left.The English translation for example sentence is "The medical insurance industry must follow the professional ethics specification" under four conditions.

Apparatus
The experiment recorded right-eye movements using Eyelink 1000 (SR Research, Canada).The sampling rate was 1000Hz, while the accuracy rate was a 0.5° visual angle.Stimuli were presented on a 19-inch Dell monitor at a resolution of 1024 × 768.Participants maintained a distance of 70 cm from the screen.Each character was 25 × 25 pixels, the visual angle was 0.80°, and the Song font was used.

Procedure
Before the experiment, the participants were told to read sentences from right to left in different conditions.Participants needed to understand the meaning of sentences as quickly as possible and then to press the space bar to read the next sentence.For some sentences, a comprehension question followed, which participants had to answer as correctly as possible.Chin rests were used to ensure that the participants' heads remained in a resting position with no head movement.Calibration was completed before the experiment to calculate the fixation point.Participants started the test after successful calibration.When necessary, eye location was recalibrated during the experiment.The experiment lasted about 20 minutes.Participants' correctly answered 91.0% of the comprehension questions, indicating that the sentences were predominantly read and understood.

Data preparation and analysis
In line with criteria from earlier research (Bai et al., 2008;Kumar et al., 2017;Li et al., 2016;Rayner, 2009a;Rayner et al., 2006;Sharmin et al., 2012;Strandberg et al., 2022;Wang et al., 2018), fixation durations shorter than 80 ms or longer than 800 ms were excluded.We also excluded data if: (1) a participant pressed the key incorrectly during the experiment, resulting in an interruption; (2) data were accidentally lost (e.g., due to head movement); (3) there were fewer than four gazes; or (4) data were outside three standard deviations.After excluding invalid data (2.8% of total data), we conducted data analysis.Collected data were analyzed using a linear mixed model (LMM) in the R environment (R Core Team, 2016), using the Ime4 package (Bates et al., 2012).We classify data as significant if the t-value exceeds 1.96 at the 5% level.Participants and items were specified as cross-random effects.We used posterior distributions for model parameters employing Markov-Chain Monte Carlo sampling to estimate p-values.Significance values reflected both subject and item variability (Baayen et al., 2008).Logtransformed analysis was performed on the analysis indexes, and logistic mixed model Ime4 was performed on the skipping probability and refixation rate.The LMM was used to analyze word frequency, format familiarity, and their interaction as fixed factors.If there was a significant interaction between word frequency and format familiarity, the HF condition was compared with the LF condition from left to right (Comparison 1), and the HF condition and the LF condition were compared from right to left (Comparison 2).

Experiment 1 Results
Table 3 shows the means and standard deviations of the eye movement measures for the target words.
Table 4 shows the fixed effect estimates for FFD, SFD, GD, RPD, TT, SP, RR, and ALP for the target words.The center of the word is the best fixation position for eye movements, and the reading efficiency is highest at the best fixation position.The farther the fixation position is from the word center, the lower the fixation efficiency (Bai et al., 2013;Li et al., 2011;Yan et al., 2010;Zang et al., 2013;Ma & Chuang, 2015).Compared with the familiar format, the average initial landing position was closer to the beginning of the word when reading efficiency was lower from left to right (b = −0.19,SE = 0.04, t = −4.69,p < 0.001).This interference may be caused by reading cost from right to left (Li et al., 2011;Ma, 2017).
Interestingly, the interaction between different indexes was not consistent.FFD and SFD generally are usually considered early indicators (Ma, 2017).In the early indexes, the interaction between format familiarity and word frequency was significant.The interaction was significant in the FFD (b = -0.070,SE = 0.030, t = -2.326,p = 0.021).From left to right, there was a significant difference between HF and LF (b = 0.048, SE = 0.015, t = 3.213, p = 0.002).From right to left, there was no significant difference between HF and LF (b = 0.014, SE = 0.022, t = 0.607, p = 0.545).There was a marginally significant interaction on the SFD (b = -0.071,SE = 0.036, t = -1.950,p = 0.053).The SFD was considered a good indicator of the semantic stage in vocabulary recognition and was greatly influenced by word frequency.

Experiment 1 Discussion
Experiment 1 investigated the individual and interaction effects of word frequency and format familiarity on Chinese vocabulary recognition.In all eye-movement indexes, there were word-frequency effects and format familiarity effects, with best reading performance found in the left to right, HF condition.FFD and SFD represented the early indexes which reflected the early processing in Chinese vocabulary recognition (e.g., Rayner, 2009b;Li et al., 2011;Ma, 2017;Huang & Li, 2020).There was significant interaction effect on FFD: the word-frequency effect appeared from left to right but not from right to left.We also found a marginally significant interaction effect on SFD: the word-frequency effect appeared from left to right but not from right to left.FFD and SFD represent early processing.By contrast, we found no significant interaction effects on the other time indexes, indicating that format familiarity affects only early processing in word recognition.Thus, for the early indicators, a lack of right-to-left reading experience had a more significant influence on HF words.This could explain some results in previous studies (e.g., Ma, 2017).There was also a significant interaction effect on RR, a sensitive indicator reflecting cognitive processing efficiency during fixation (e.g., Rayner, Pollatsek & Binder, 1998;Bai et al., 2008).Readers could fixate the optimal viewing position in fewer counts, thereby obtaining more information.This result suggests that format familiarity influences processing efficiency in word recognition.
The result in experment 1 may explain delay effect in previous study (Ma, 2017).The familiarity verification delays the appearing of word-frequency effect from right to left, the vocabulary processing was faster in the familiar format, the word-frequency effect only appeared on early indexes.According to the results of Experiment 1, the facilitation of format familiarity are due to features of the reading experience.The directional oculomotor activities in association with reading behind format familiarity may affect vocabulary recognition from top to bottom (Chen et al., 2021).There is rich experience from left to right, which caused the faster neural processing.While there is lack experience from right to left, which may have a delay in neural processing and reflect in the vocabulary recognition.Therefore, reading experience is considered as particularly important, reading training could improve reading experience quickly (e.g.Bai et al., 2008;Nedeljković & Pušnik, 2020;Chen et al., 2021).Based on the results of Experiment 1, we can infer that the impact of format familiarity on the word-frequency effect (the 7-point scale by 40 individuals who did not participated in the eye-tracking experiment.They were also rated for predictability by another 28 non-participating individuals.. Naturalness and predictability were matched for the HF and LF conditions. The materials used in the reading training were 60 Chinese essays (average number of words M = 936, SD = 45) chosen from Chinese high school textbooks, all were reversed from left-to-right format to right-to-left format using reversing software (see the Appendix for samples of the reading materials).
Note: The meaning of the Chinese sentence is that international schools need optimistic and responsible English teachers.

Apparatus
The apparatus was the same as in Experiment 1

Procedure
There were two stages: reading training (collective learning) and eye movement tracking (individual tests).For training, participants came to the laboratory every day.They sat in their allocated seats, each with a copy of the book containing the articles to be read was presented.Before reading began, the researcher gave the following instructions: Journal of Eye Movement Research Chen, M. & Lu, J. ( 2023) 16( 4):5 The role of format familiarity and word frequency 12 You will now read some articles.The sentences in the articles will be presented from right to left.
Please read carefully word by word and understand the article as much as possible.Seven reading comprehension questions will appear after each article.You are required to select the most appropriate answer based on the article and fill in the answer.
Participants then started to read after understanding the instruction.After reading each article, they completed the reading comprehension questions before moving on to the following article.The entire reading experiment lasted for 30 minutes.Each participant completed one reading exercise per day, comprising five articles.After 10 days of reading training, eye-movement testing began using the same procedure as in Experiment 1. Participants correctly answered 93.0% of the comprehension questions during eye-movement testing, indicating that the sentences were predominantly read and understood.

Data preparation and analysis
We applied the same data-selection criteria, analysis models, eye-movement measures, significance threshold, and p-value estimation approach as in Experiment 1 (e.g., Bai et al., 2008;Li et al., 2016;Liang et al., 2017;Rayner et al., 2006;Rayner, 2009b;Wang et al., 2018).After excluding invalid data (1.65% of the total data) according standards, which was mentioned in the experiment 1, data analysis was conducted.The LMM in this experiment analyzed the word frequency, the training effect, the format familiarity, the interaction of the word frequency and training, and the interaction of the word frequency and format familiarity as fixed factors.If there was a significant interaction between word frequency and format familiarity, the HF condition was compared with the LF condition from left to right before training
Table 8 combines the results of Experiment 2. There were significant word-frequency effects in FFD The role of format familiarity and word frequency Reading experience affects the speed and efficiency of Chinese vocabulary recognition.Ma (2017) dismissed the visual familiarity hypothesis, proposing that the inter-word spaces lower reading time by reducing lateral masking and facilitating word segmentation.However, word segmentation and word recognition are unified and indistinguishable.There seems to be a trade-off between familiarity and word segmentation (Chen et al., 2021), suggesting that the influence of familiarity cannot be excluded.Our study manipulated format familiarity using reading direction; the difference in format familiarity was attributable to the difference in reading experience.In the familiar condition of left to right reading, Chinese readers had a higher skipping rate and lower refixation probability.Reading experience is a high-level cognitive factor with a top-down effect on vocabulary recognition.The richer the reading experience, the higher the reading performance and efficiency, and the closer the fixation point to the word center (Bai et al., 2008;Chen et al., 2021;Li et al., 2009;Zang et al., 2013).The results of Experiment 2 support the view that reading experience influences vocabulary recognition in Chinese reading.The word-frequency effect could reflect the vocabulary-recognition stage sensitively (Liu et al., 2016;Liversedge et al., 2014;Ma, 2017;Ma & Zhuang, 2018;Rayner, Pollatsek & Binder, 1998;Yan et al., 2006).Notably, the interaction between format familiarity and word frequency was significant in the early indexes, which may indicate a delay in the word-frequency effect in right-to-left reading.This delay is similar to that found by Ma (2017).
Based on the SWIFT model, word segmentation and word recognition are top-down and bottom-up unified processes (Huang & Li, 2020;Li & Pollatsek, 2020).When word segmentation is completed, vocabulary recognition is too.However, the delayed word-frequency effect from right to left suggests that word segmentation processing may be independent in Chinese reading, consistent with the E-Z Reader model (Rayner & Pollatsek, 2007;Reichle et al., 2009;Ma, 2017;Chen et al., 2021), which proposes that readers conduct word segmentation and then vocabulary recognition when reading.From left to right, participants fully or partially completed word segmentation in the preview stage, then entered the vocabulary-recognition stage.Therefore, the word-frequency effect appeared in the early-fixation stage.From right to left, however, participants did not have sufficient reading experience and thus paid a higher cost, with word segmentation spilling over from the preview stage into the familiarity-verifying stage, thereby delaying the word-frequency effect on vocabulary recognition.If word segmentation and vocabulary recognition are the same process, there should always be an interaction between format familiarity and word frequency.The results showed such an interaction in the early indexes (FFD and SFD); however, in the late indexes, the word-frequency effect was significant from left to right and from right to left, meaning that format familiarity had no impact on the word-frequency effect, that the word segmentation process had been completed, and that the vocabulary-recognition stage had begun.
Therefore, the word-frequency effect always existed.In short, the processes of word segmentation and vocabulary recognition in late reading may be sequential, rather than concurrent.However, this processing mechanism is complicated and also involves character recognition.Whether there is an interaction between different levels should be explored in future research.
eye-movement measures for the target words: (1) first fixation duration (FFD)the duration of the first fixation on a word, irrespective of the number of fixations; (2) Single fixation duration (SFD)-the fixation duration when only one fixation was made on the word during first-pass reading; (3) gaze duration (GD)-the sum of all fixations on a word before moving to another word; (4) regression-path duration (RPD)-the sum of all gaze times looking back to the current word; (5) Total time (TT)-the sum of all fixations on the target word, including regressions; (6) Skipping probability (SP)-the probability of skipping the target region in the first reading; (7) Refixation rate (RR)-the probability of the target region being gazed at multiple times in the first reading; (8) Average Initial landing position (ALP)-the distance to the beginning of the target word for the first time.The time index units (FFD, SFD, GD, RPD, and TT) were measured in milliseconds.

(
Comparison 1), and the HF condition and the LF condition were compared from right to left before training (Comparison 2).If there was a significant interaction between word frequency and training, the HF condition was compared with the LF condition from right to left before training (Comparison 2), and the HF condition and the LF condition was compared from right to left after training (Comparison 3).

Note:
Standard deviations are provided in parentheses.FFD = First fixation duration; SFD = Single fixation duration; GD = Gaze duration; RPD = Regression-path duration; TT = Total time; SP = Skipping probability; RR = Re-fixation rate; AIP=average landing position.Table 8.Fixed effect estimates for FFD, SFD, GD, RPD, TT, SP, RR, AIP for the target words Note: *** p<0.001, ** p<0.01, * p<0.05, § p < 0.1.Interaction 1 represented the interaction between word frequency and format direction.Interaction 2 represented the interaction between word frequency and training.The highfrequency would be compared with the low-frequency from left to right before training (Comparison 1).The highfrequency and the low-frequency would be compared from right to left before training (Comparison 2).The highfrequency and the low-frequency would be compared from right to left after training (Comparison 3).

Table 1 .
The statistical characteristics of the experimental sentences and target words Note.Standard deviations are provided in parentheses.

Table 2 .
Example Chinese stimuli from the four experimental conditions.

Table 3 .
Eye-movement measures for the target words.

Table 4 .
Fixed effect estimates for FFD, SFD, GD, RPD, TT, SP, RR, and ALP for the target words.

Table 5 .
Statistical characteristics of the experimental sentences and target words Note: Standard deviations are provided in parentheses.

Table 6 .
Example Chinese stimuli from the two experimental conditions.

Table 7 .
Eye movement measures for the target words