The latent mechanism behind binocular advantage in reading

Abstract

Most individuals read binocularly, and previous studies have found a binocular advantage in reading speed. However, the underlying mechanism of the binocular advantage in reading remains unclear. In our study, we quantified contributions from basic visual functions, basic oculomotor functions, and reading-specific eye movements to the binocular advantage in Chinese reading speed, using six tasks and 32 metrics. Consistent with prior research, we confirmed a binocular advantage in Chinese text reading, with binocular reading being approximately 4% faster than monocular reading. Interestingly, although basic visual and oculomotor functions themselves exhibited binocular advantages, they did not account for the observed binocular advantage in reading among individuals with normal vision. This finding is particularly noteworthy because it provides an important normative reference for individuals with impaired vision, in whom basic visual and oculomotor functions may serve as critical explanatory factors for reading performance. In contrast, the concurrent reduction of three reading-specific eye movement metrics—fixation count, average fixation duration, and progressive saccade count—under binocular conditions well explained the binocular advantage in reading, despite these metrics not demonstrating a binocular advantage in isolation. Our results suggest that efficient parafoveal preprocessing and faster neural processing in binocular vision might play critical roles in binocular advantage in reading for individuals with normal vision.

Introduction

Reading is a demanding skill involving simultaneous operation and coordination of multiple distinct functions: vision, eye movement, attention, memory, and language comprehension. Most individuals read binocularly in the vast majority of situations. Previous studies have found a binocular advantage during reading for individuals with normal vision, expressed in more efficient word processing (Chen, Chen, & Liu, 2023; Jainta, Blythe, & Liversedge, 2014; Jainta & Jaschinski, 2012; Jainta & Joss, 2019; Jainta, Nikolova, & Liversedge, 2017; Nikolova, Jainta, Blythe, & Liversedge, 2018; Sheedy, Bailey, Buri, & Bass, 1986), faster sentence reading (Jainta et al., 2014; Jainta et al., 2017; Nikolova, Jainta, Blythe, & Liversedge., 2017; Nikolova et al., 2018) and shorter text reading time (Johansson, Pansell, Ygge, & Seimyr, 2014a; Johansson, Pansell, Ygge, & Seimyr, 2014b). However, the latent mechanism of binocular advantage in reading remains unclear. 

One potential explanation for the binocular advantage in reading is the enhanced visual function provided by binocularity. Basic visual functions, such as acuity and contrast sensitivity, are known to affect reading speed (Legge, Rubin, & Luebker, 1987; Whittaker & Loviekitchin, 1993) and studies have shown that binocular vision often outperforms monocular vision in visual acuity (Banton & Levi, 1991; Birch & Swanson, 1992; Cagenello, Arditi, & Halpern, 1993; Heravian, Jenkins, & Douthwaite, 1990; Horowitz, 1949) and contrast sensitivity (Blake & Levinson, 1977; Blake & Wilson, 2011; Campbell & Green, 1965; Home, 1978; Legge, 1984). Johansson et al. (2014a) found an increase in binocular advantage with decreasing text contrast, suggesting that the difference in visual contrast perception between binocular and monocular conditions may play an important role in reading. Furthermore, Jainta et al. (2017) investigated whether the binocular advantage in reading was due to the contrast reduction inherent in monocular reading. They simulated monocular reading conditions by reducing sentence contrast to 70% during binocular viewing but found no significant difference compared to full contrast conditions. However, at a further reduced contrast of 40%, reading speeds slowed globally, leading to the conclusion that contrast reduction alone does not account for the decreased efficiency in lexical processing observed under monocular conditions. 

The binocular advantage in reading may also stem from the different eye movement dynamics observed during binocular versus monocular viewing. Reading involves a sequence of saccades and fixations, where rapid saccades (approximately 30 ms) bring information to the foveal region of the retina (Liversedge & Findlay, 2000; Rayner, 1998). These saccades are followed by brief fixations (average 200-300ms), during which cognitive processing occurs (Raney, Campbell, & Bovee, 2014; Rayner, 2009). Eye movements during reading are influenced by both low-level oculomotor processes and high-level cognitive factors (Rayner, 1998; Rayner & Liversedge, 2011; Vitu, 2011). Therefore, differences in low-level oculomotor functions between monocular and binocular viewing may contribute to the binocular advantage in reading. For example, compared to monocular viewing, binocular viewing exhibits shorter saccade latencies and higher peak velocities in simple saccade tasks (Niechwiej-Szwedo, Chandrakumar, Goltz, & Wong, 2012; Niechwiej-Szwedo, Goltz, Chandrakumar, Hirji, & Wong, 2010), as well as better fixation stability during fixation tasks (González, Wong, Niechwiej-Szwedo, Tarita-Nistor, & Steinbach, 2012). 

Beyond basic oculomotor functions, eye movement dynamics also differ between monocular and binocular reading in normal vision individuals, as evidenced by metrics such as fixation count (Nikolova et al., 2017), fixation duration (Jainta et al., 2014; Jainta et al., 2017; Johansson et al., 2014a; Johansson et al., 2014b; Nikolova et al., 2017), saccade amplitude (Jainta & Jaschinski, 2012; Johansson et al., 2014a; Johansson et al., 2014b), and regressive saccade probability (Nikolova et al., 2017). Moreover, patients with abnormal reading-specific eye movements, including conditions like heterophoria (Jainta & Joss, 2019), strabismus (Perrin Fievez, Lions, & Bucci, 2018), and amblyopia (Kanonidou, Proudlock, & Gottlob, 2010; Stifter, Burggasser, Hirmann, Thaler, & Radner, 2005), exhibit a diminished binocular advantage in reading. This evidence suggests that distinct eye movement patterns during monocular and binocular reading may help explain the binocular advantage observed during reading. 

On the other hand, existing research presents diverse definitions of the binocular advantage in reading. Some studies conceptualize binocular advantage as the superiority of binocular viewing over an average of monocular viewing performances or a random selection from left and right eyes (Jainta et al., 2014; Jainta et al., 2017; Nikolova et al., 2017; Nikolova et al., 2018). Others, however, define it as the superiority of binocular viewing relative to dominant eye viewing (Jainta & Jaschinski, 2012; Jainta & Joss, 2019; Johansson et al., 2014a; Johansson et al., 2014b). This discrepancy prompts another interesting question: should there be no marked difference between the reading performance of the dominant and non-dominant eye, the two definitions might yield similar results. Conversely, if the dominant eye advantage exist in reading tasks, as observed in other visual tasks such as visual search (Liu et al., 2021; Shneor & Hochstein, 2006; Shneor & Hochstein, 2008), visually-guided action (Chaumillon et al., 2017; Chaumillon, Blouin, & Guillaume, 2014), and saccade (Chaumillon et al., 2022; Tagu, Doré-Mazars, Lemoine-Lardennois, & Vergilino-Perez, 2016; Tagu, Doré-Mazars, Vergne, Lemoine-Lardennois, & Vergilino-Perez, 2018), the two definitions will result in inconsistent outcomes, with the first definition overestimating the binocular advantage. Therefore it is crucial to assess the existence of a dominant eye advantage when exploring the binocular advantage in reading. 

In the current study, we quantified the contributions of basic visual functions, basic oculomotor functions, and reading-specific eye movements to the binocular advantage in Chinese text reading, with consideration for the dominant eye advantage to prevent an overestimation of binocular benefits, aiming to shed light on the mechanism behind the binocular advantage in reading. 

Method

Participants

A total of 28 native Chinese speakers with no reading difficulties took part in the experiment. All of them were right-handed undergraduate or graduate students (18 females and 10 males) aged from 19 to 31 years (24.04 ± 2.69). All participants had normal or corrected-to-normal monocular and binocular vision (with visual acuity better than 0.0 logMAR) and normal binocular status (with stereovision of 40 seconds of arc or better). They underwent testing with their best-corrected vision and wore their corrective glasses, if necessary. The study was approved by the Ethical Committee of the Institute of Psychology, Chinese Academy of Sciences (H22074, June 10, 2022). The study conforms to the standards of Declaration of Helsinki. Written informed consent was obtained from all participants before the commencement of the study. 

Experimental design

We quantified the binocular advantage in Chinese text reading (Task 6; see Figure 1C) and then accounted for this advantage by quantifying three possible factors: basic visual functions (Figure 1A), basic oculomotor functions (Figure 1B), and reading-specific eye movements (Figure 1C), using six tasks and 32 metrics (see Table 1). 

Figure 1.

 

Illustrations of the experimental procedure. (A) Three tasks to quantify basic visual functions. (B) Two tasks to quantify basic oculomotor functions. (C) Reading task to quantify the binocular advantage in reading and reading eye movements. (D) Each participant was tested under three different viewing conditions—binocular, dominant eye, and non-dominant eye—in random order for all tasks, except the stereoacuity test and binocular rivalry task.

Table 1.

 

Tasks and corresponding metrics. Notes: Avg, average; BCEA, bivariate contour ellipse area.

 

Before conducting any task, ocular dominance was assessed using the hole-in-card sighting test (Rice, Leske, Smestad, & Holmes, 2008). Each participant was then tested under three different viewing conditions—binocular, dominant eye, and non-dominant eye—in random order for all tasks, except the stereoacuity test and binocular rivalry task. During monocular viewing conditions, the untested eye was covered by an opaque eye patch (Figure 1D). 

Tasks

Task 1: Acuity assessment

Acuity measurements included far visual acuity (ETDRS chart, 4 m), near visual acuity (ETDRS chart, 40 cm), and stereoacuity (Titmus stereo test). 

Task 2: Quick contrast sensitivity function (qcsf) task

Spatial contrast sensitivity functions (CSFs) were assessed with the qCSF method (Lesmes, Lu, Baek, & Albright, 2010), which is a Bayesian adaptive procedure that utilizes prior knowledge about the general shape of the CSF to optimize stimulus selection, and an accompanying grating orientation identification task of 100 trials (Hou et al., 2010). The CSF was characterized by four parameters: peak gain, optimal spatial frequency, bandwidth, and low-frequency truncation level. In the grating orientation identification task, each trial started with a 294-ms cross-hair frame cueing the size and location of the upcoming grating stimulus, accompanied by a brief tone signaling the onset of each trial. This was followed by a 153-ms blank screen, and a 141-ms grating stimulus with a size of 3.0° × 3.0° and random orientation (45° or 135°). Participants were asked to report the grating orientation with keypress. No feedback was provided. A new trial started 588 ms after the participant's response. Based on the obtained contrast sensitivity functions, we derived two metrics: the area under the log CSF (AULCSF) and the spatial frequency corresponding to a contrast sensitivity of 1.0 (the high spatial-frequency cut-off, cutSF), which reflect general spatial visual representation (Applegate, Howland, Sharp, Cottingham, & Yee, 1998; Hou et al., 2010; Oshika, Okamoto, Samejima, Tokunaga, & Miyata, 2006) and high-frequency resolution (Huang, Tao, Zhou, & Lu, 2007; Zhang et al., 2015; Zhou et al., 2006) of the visual system, respectively. 

Task 3: Binocular rivalry task

Two static sine-wave gratings, with identical size (4.0° × 4.0°), contrast (0.75), and spatial frequency (1 cyc/°) but orthogonal orientations (45° vs. 135°), were presented dichoptically to the two eyes using a stereoscope. After vergence adjustment, two orthogonal gratings were displayed for 180 seconds, during which participants reported the dynamics of perceived orientation with keypress. The next trial started after a mandatory 30-second inter-trial interval. There were three trials in total. The dominance duration of each eye's perception was recorded separately, and the ratio of these durations was used to determine binocular balance extent. A dominant duration ratio of <1.0, 1.0, and >1.0 indicated left eye dominance, balanced sensory dominance, and right eye dominance, respectively. 

Task 4: Fixation stability task

The fixation stability task included three trials. Each trial started with a two-second red cross (RGB: 255, 0, 0), cuing the size and location of the upcoming target, presented at the center of the gray screen (RGB: 230, 230, 230), which was followed by a 0.5° red square (RGB: 255, 0, 0) that lasted for 15 seconds. Participants were asked to fixate on the center of the red square. The stability of fixation was assessed using a bivariate contour ellipse area (BCEA) that encompassed 68% of fixation points over the 15-second test interval, calculated by the following equation:  

\begin{eqnarray}BCEA = \pi {{X}^2}{{\sigma }_x}{{\sigma }_y}\sqrt {1 - {{p}^2}}, \quad \end{eqnarray}

(1)

where X² is the chi-squared value (2 df) corresponding to a probability of 0.68; σ_x and σ_y are the standard deviations in the horizontal and vertical directions, and p is the Pearson product moment correlation coefficient of the horizontal and vertical eye movements. The final fixation stability result for each participant was the average of three trials. Calibration of eye movements was performed prior to the task and repeated as needed during the experiment. 

Task 5: Simple saccade task

The saccade task used the same calibration procedure as the fixation stability task. In the saccade task, each trial started with a red cross (RGB: 255, 0, 0) presented at the center of a gray screen (RGB: 230, 230, 230) for a random duration between two and four seconds, which was followed by the display of a 0.5° red square (RGB: 255, 0, 0) at one of eight positions (±5° and ±10° from the cross horizontally and vertically) for three seconds. Participants were instructed to fixate on the cross first, and then perform saccades from cross to target as quickly and accurately as possible upon the appearance of the red square. The task comprised 64 trials, with each position being repeated eight times. For saccade performance in the simple saccade task, we identified primary saccade, which refers to the first saccade that landed near the target (within 3° of the target) after its appearance, and secondary saccade, which refers to the saccade that occurred within 250 milliseconds following the primary saccade (Deubel, Wolf, & Hauske, 1982; Prablanc, Masse, & Echallier, 1978). Specifically, we measured primary saccade latency, accuracy (the distance between the end point of the primary saccade and the target), average velocity, and peak velocity, as well as the count and accuracy of secondary saccades (the distance between the endpoint of the secondary saccade and the target). 

Task 6: Reading task

In the reading task, we measured participants’ reading and eye movement performance while they read texts silently for comprehension. There were ten different Chinese text materials, which were sourced from International Reading Speed Texts (IReST) that are designed for 17 languages including Chinese, and standardized in terms of content, length, difficulty, and linguistic complexity (Trauzettel-Klosinski, Dietz, & Grp, 2012). These Chinese IReST texts have also been validated as homogeneous and comparable for repeated measurements (Wang et al., 2018). Participants read two texts randomly selected from the Chinese IReST under each viewing condition. The texts were displayed as a single paragraph, subtending 24.3° by 9.8°, with black characters (RGB: 0, 0, 0) on a gray background (RGB: 230, 230, 230) in Song font, averaging 0.9° in size. Participants were instructed to read at their own pace and press the space bar on completing the text. Comprehension was assessed with a follow-up question about the text. Reading metrics consisted of two categories: those related to participants' reading speed, including the total reading time for the whole text (text reading time) and the reading time for each character (first fixation duration on character, character FFD), and those representing the participants' eye dynamics during reading text, including fixations (count and average duration), progressive saccades (count, average amplitude, average angle, average duration, average velocity, and peak velocity), regressive saccades (count, average amplitude, average angle, average duration, average velocity, and peak velocity), blinks (count and average duration), and average pupil size. 

Apparatus

Acuity (Task 1) was measured with ETDRS chart and Titmus test. For task 2 and task 3, which did not involve eye movement recording, stimuli were displayed on a CRT monitor with a resolution of 1600 × 1200 pixels and a refresh rate of 85 Hz. A special circuit was used to merge outputs from the red and blue channels of the graphics card to produce 14-bit gamma-corrected gray-level resolution (Li, Lu, Xu, Jin, & Zhou, 2003). Participants were positioned 114 cm away from the monitor. The experiment was programmed in Matlab with PsychToolbox extensions. 

For tasks 4, 5, and 6 that required the recording of eye movements, experimental stimuli were presented on an LCD monitor with a resolution of 1280 × 1024 pixels and a refresh rate of 60Hz. Participants sat 60 cm away from the monitor, with both chin and forehead rest to minimize head movement during the experiment. Eye movements were recorded by an Eyelink 1000 plus eye-tracking system (SR Research, Mississauga, ON, Canada) with a sampling rate of 1000 Hz. The eye movement experiment was programmed with EyeLink Experiment Builder software and the analysis of eye movement metrics was conducted with EyeLink Data Viewer software. During binocular viewing conditions, eye movement data were exclusively recorded from the right eye of each participant. 

Data analysis

To discern differences in metrics in the reading task (Task 6) as a function of viewing condition, we employed generalized linear mixed-effects models (GLMMs) which simultaneously controlled random effects introduced by participants and texts. GLMMs are advantageous as they allow interpretation on the original scale of response and adeptly handle non-normally distributed residuals, thereby circumventing the pitfalls of using raw or transformed data (Lo & Andrews, 2015). Each metric during reading was specified as a continuous dependent variable, and examined, using a gamma model in the family argument of the glmer function, as a function of viewing condition on a single-trial level. The viewing condition was entered as a fixed effect, specifying the participants and texts as crossed random effects, including intercepts and slopes (Baayen, Davidson, & Bates, 2008). We first constructed a model with a maximal random factor structure. When the maximal model failed to converge, we pruned the random effects structures to optimize the model (Barr, Levy, Scheepers, & Tily, 2013). R software v.4.2.1 with the “lmerTest” package was used to construct GLMMs. The main effect of the viewing condition was estimated using the Anova function from the “car” package, reporting chi-square values and corresponding p values. If the main effect was significant, a post hoc analysis with Bonferroni correction was carried out to compare differences in three viewing conditions using the “emmeans” package, reporting regression coefficients (bs, which estimate the effect size), standard errors (SEs), z-values, and corresponding p-values. 

To examine the relationship between the extent of binocular advantage in reading and the extent of binocular advantage in each metric from three possible factors (basic visual functions, basic oculomotor functions, and reading-specific eye movements), we conducted Pearson correlation analysis and carried out FDR correction for p values, using R software (v.4.2.1) with the “stats” package. The extent of binocular advantage was quantified by computing the difference between the mean values obtained under binocular and monocular conditions, followed by standardization using Z-score transformation. 

To explore the latent variables that contributed to the binocular advantage in reading time, we used the least absolute shrinkage and selection operator (LASSO) regression, which adeptly addresses both the multicollinearity issue and predictor selection by diminishing the coefficients of less important predictors to zero (Tibshirani, 1996). We applied LASSO regression on the extent of binocular advantage in three categories of metrics (basic visual functions, basic oculomotor functions, and reading-specific eye movements) to predict the extent of binocular advantage in reading time. The method used to calculate the extent of binocular advantage was the same as correlation analysis. The LASSO regression procedure with a ten-fold cross-validation was iterated 1000 times. In each iteration, we estimated the goodness of fit with all predictors in the LASSO regression model. The goodness of fit was calculated by R²,  

\begin{eqnarray}{{R}^2} = 1 - \frac{{\sum {{{\left( {{{y}_i} - {{{\hat{y}}}_i}} \right)}}^2}}}{{\sum {{{\left( {{{y}_i} - \bar{y}} \right)}}^2}}}{\rm{\ }}, \quad \end{eqnarray}

(2)

where y_i and \({{\hat{y}}_i}\) represent the observed and predicted binocular advantage extent in reading time for every participant respectively, and \(\bar{y}\) is the mean of observed binocular advantage extent in reading time across all participants. Then, we dropped one predictor each time and recalculated the goodness of fit for the reduced model with one predictor out. We repeated the procedure until each predictor was left out exactly once. After 1000 iterations, we used paired sample t-tests to ascertain the significance of the difference between the R² of the LASSO regression model with all predictors and the models with one predictor out. A significant reduction of R² on the removal of a specific predictor indicated its important role in explaining the binocular advantage in reading. The “glmnet” package in R software (v.4.2.1) was used to perform the LASSO regression analysis. 

To investigate differences in metrics from tasks 1, 2, 4, and 5 across different viewing conditions, we used one-way repeated measures analysis of variance (ANOVA) with post hoc Bonferroni pairwise comparisons. Greenhouse–Geisser corrections were used when sphericity was violated according to the Mauchly test, and in these cases, we report the unadjusted degrees of freedom for ease of interpretation and the corrected p-values. The “bruceR” package in R software (v.4.2.1) was used to perform the repeated-measures ANOVA analysis. 

To determine whether the sensory dominance of the two eyes (task 3) among participants in our study was balanced or not, a one-sample t-test was conducted with the “bruceR” package in R software (v.4.2.1). 

Results

Binocular advantage in reading time

In the reading task (Task 6), the average accuracy for comprehension questions was 92%, suggesting a high level of comprehension. We focused on two metrics directly related to reading time: the total reading time for the whole text (text reading time) and the first fixation duration on each character (character FFD). According to the hole-in-card sighting test, 21 participants (75%) were right eye dominant and 7 participants (25%) were left eye dominant. Based on the assignment of the dominant eye, we constructed generalized linear mixed-effects models (GLMMs) to investigate differences in reading time under three viewing conditions: binocular, dominant eye, and non-dominant eye. Prior to GLMMs analysis, four trials (2%) were excluded because of non-compliance with experimental instructions, and another four trials (2%) were excluded due to text reading time exceeding three standard deviations. Fixations shorter than 80 ms or longer than 1200 ms (8% of fixations) were also excluded. 

The results consistently demonstrated significant main effects of viewing condition on both text reading time and character FFD measures (χ² = 359.72, p < 0.001; χ² = 21.002, p < 0.001; see Figure 2). Subsequent post hoc analyses revealed that text reading time in binocular vision was shorter than that in the dominant eye (b = −861, SE = 68.8, z = −12.503, p < 0.001) and the non-dominant eye (b = −702, SE = 47.9, z = −14.639, p < 0.001). Additionally, the difference in text reading time between the dominant eye and the non-dominant eye was not significant (b = 159, SE = 82.7, z = 1.927, p = 0.162). Similarly, character FFD in binocular vision was shorter than that in the dominant eye (b = −9.48, SE = 2.11, z = −4.496, p < 0.001) and the non-dominant eye (b = −5.75, SE = 1.96, z = −2.939, p = 0.010). The difference in character FFD between the dominant eye and the non-dominant eye was also not significant (b = 3.73, SE = 2.08, z = 1.795, p = 0.218). These findings underscore the advantage of binocular vision in reading, without any observed advantage for the dominant eye in reading. Our study found that binocular reading was approximately 4% faster than monocular reading in terms of text reading time. In our study, the extent of binocular advantage in Chinese character recognition was approximately 6 ms, smaller than the approximately 25 ms reported in a prior study on two-character word recognition in Chinese reading (Chen et al., 2023). This discrepancy may stem from differences in experimental stimuli (e.g., complexity and word frequency) and the methodologies used to manipulate monocular and binocular viewing (i.e., eye patch vs. goggles). 

Figure 2.

 

Binocular advantage in reading. (A) Binocular advantage in text reading time (text-based global metric). (B) Binocular advantage in character FFD (character-based local metric). Data are means ± SE. **p < 0.01, ***p < 0.001.

Variables contributing to the binocular advantage in reading

The GLMMs analysis indicated a clear binocular advantage during silent reading of Chinese text, as evidenced by reduced text reading time and shorter character FFD. To further elucidate the mechanism underlying this binocular advantage, we tried to explore the significant contributing variables. The explanatory variables comprised three categories of metrics: basic visual functions, basic oculomotor functions, and reading-specific eye movements, totaling 30 metrics (see Table 1). When elucidating the binocular advantage in text reading time, character FFD was also considered as an explanatory variable. Initially, we calculated the correlation between the extent of binocular advantage in reading and the extent of binocular advantage in three categories of explanatory variables and then constructed LASSO regression models on the extent of binocular advantage in these explanatory variables to predict the extent of binocular advantage in reading. Given the absence of a dominant eye advantage in our findings, we defined the binocular advantage as binocular performance superior to the average monocular performance. Each index was thus derived by subtracting the mean value under monocular conditions from that under the binocular condition, followed by standardization using Z-score transformation. 

For the binocular advantage in text reading time, correlation analysis revealed that the extent of this binocular advantage was not linked to the binocular advantage in basic visual function metrics. However, this binocular advantage showed significant associations with the binocular advantage in two metrics of basic oculomotor functions (primary saccade latency and primary saccade peak velocity) as well as with most reading-specific eye movement metrics (character FFD, fixation count, average fixation duration, progressive saccade count, progressive saccade average amplitude, progressive saccade average duration, progressive saccade average velocity, regressive saccade count, regressive saccade average amplitude, and regressive saccade average velocity; see Figure 3). 

Figure 3.

 

Correlation between the binocular advantage in text reading time and the three categories of explanatory variables. *p < 0.05, **p < 0.01, ***p < 0.001. Notes: ave = average; sac = saccade; fix = fixation; pro = progressive; re = regressive.

To delve further, LASSO regression models were constructed on the extent of binocular advantage in three categories of explanatory variables to predict the extent of binocular advantage in text reading time. Variables meeting three criteria (average regression coefficient > 0.1 across 1000 iterations, selected in 100% of iterations, and significantly reducing model R² when removed) were identified as significant variables contributing to the binocular advantage in text reading time. Results pinpointed fixation count (β = 0.617), average fixation duration (β = 0.164), and progressive saccade count (β = 0.115) from reading-specific eye movements were significant predictors of the binocular advantage in text reading time (Figure 4). 

Figure 4.

 

LASSO regression on the three categories of explanatory variables to predict the extent of binocular advantage in text reading time. Red dotted box indicates the significant contributor to the binocular advantage in text reading time, which meets three criteria: average regression coefficient > 0.1 across 1000 iterations, presented in (A), selected in 100% of iterations, presented in (B), and significantly reducing model R² when removed, presented in (C). ***p < 0.001.

For the binocular advantage in character FFD, correlation analysis found that the extent of this binocular advantage was only correlated with the extent of binocular advantage in average fixation duration (Figure 5) and LASSO regression analysis also revealed that average fixation duration was the only significant contributor (β = 0.850) (Figure 6). 

Figure 5.

 

Correlation between the binocular advantage in character FFD and the three categories of explanatory variables. ***p < 0.001.

Figure 6.

 

LASSO regression on the three categories of explanatory variables to predict the extent of binocular advantage in character FFD. Red dotted box indicates the significant contributor to binocular advantage in character FFD, which meets three criteria: average regression coefficient > 0.1 across 1000 iterations, presented in (A), selected in 100% of iterations, presented in (B), and significantly reducing model R² when removed, presented in (C). ***p < 0.001.

Binocular and/or dominant eye advantage in the explanatory variables

We compared the performance of three viewing conditions in three possible factors to explore the binocular and/or dominant eye advantage in the explanatory variables (see Figure 7). 

Figure 7.

 

The performance of three viewing conditions in the explanatory variables. (A) Metrics in basic visual functions. Stereoacuity and rivalry are binocular functions. (B) Metrics in basic oculomotor functions. (C) Three metrics selected by LASSO regression in reading-specific eye movements. Three viewing conditions: B = binocular viewing condition, D = dominant eye condition, N = non-dominant eye condition. Data are means ± SE. *p < 0.05, **p < 0.01, ***p < 0.001.

For acuity (Task 1), we observed the advantage for binocular vision but not for the dominant eye. Specifically, for far visual acuity, a repeated-measures ANOVA revealed a significant main effect of viewing condition (F_(2,54) = 12.637, p < 0.001, \(\eta_{p}^{2}\) = 0.319). Post hoc analyses showed that binocular acuity surpassed both the dominant eye acuity (p < 0.001) and the non-dominant eye acuity (p = 0.002) and the difference between the acuity of the dominant and non-dominant eye was not significant (p = 0.335). A similar pattern was also observed for near visual acuity, where binocular acuity was superior (F(2,54) = 22.118, p < 0.001, \(\eta_{p}^{2}\) = 0.450; post hoc analysis: vs. dominant eye, p < 0.001; vs. non-dominant eye, p < 0.001). Additionally, there was no significant difference between the acuity of the dominant and non-dominant eye (p = 1.000). 

For contrast sensitivity (Task 2), we also observed a binocular advantage but not a dominant eye advantage. For AULCSF, a repeated-measures ANOVA revealed a significant main effect of viewing condition (F(2,54) = 99.918, p < 0.001, \(\eta_{p}^{2}\) = 0.787). Post hoc analyses found that the AULCSF of binocular viewing was greater than that of the dominant (p < 0.001) and non-dominant eye (p < 0.001) respectively. Additionally, there was no significant difference between the AULCSF of the dominant and non-dominant eye (p = 0.084). Regarding cutSF, a repeated-measures ANOVA revealed a significant main effect of viewing condition (F(2,54) = 7.727, p = 0.001, \(\eta_{p}^{2}\) = 0.223). Post hoc analyses showed that binocular cutSF was better than cutSF in the dominant (p = 0.012) and non-dominant eye (p = 0.009), respectively, and the difference between cutSF in the dominant and non-dominant eye was not significant (p = 1.000). 

For binocular balance measurement (Task 3), a one-sample t-test analysis revealed that the dominant duration ratio of the left eye to the right eye didn't significantly deviate from 1.0 (T₍₂₇₎ = −0.53, p = 0.601), indicating sensory balance between the two eyes for the participants. 

For fixation stability (Task 4), a repeated-measures ANOVA revealed a significant main effect of viewing condition (F(2,54) = 6.303, p = 0.009, \(\eta_{p}^{2}\) = 0.189), after removing outliers that fell beyond three standard deviations for each viewing condition. Post hoc analyses indicated that binocular fixation stability was better than both the dominant eye fixation stability (p = 0.002) and the non-dominant eye fixation stability (p = 0.010) respectively; no significant difference existed between the fixation stability of the dominant and non-dominant eye (p = 1.000). 

For simple saccade performance (Task 5), main effects of viewing condition were significant for primary saccade latency (F(2,54) = 12.032, p < 0.001, \(\eta_{p}^{2}\) = 0.308), primary saccade average velocity (F(2,54) = 9.803, p < 0.001, \(\eta_{p}^{2}\) = 0.266), and primary saccade peak velocity (F(2,54) = 13.428, p = 0.001, \(\eta_{p}^{2}\) = 0.332), but not primary saccade accuracy (F(2,54) = 1.894, p = 0.171, \(\eta_{p}^{2}\) = 0.066), after removing outliers that fell beyond three standard deviations. Further post hoc analyses consistently revealed a binocular advantage, with binocular performance superior to both the dominant eye performance (p < 0.001; p = 0.002; p = 0.005) and the non-dominant eye performance (p = 0.016; p = 0.005; p < 0.001). No significant dominant eye advantage was observed (ps > 0.05). For secondary saccade count, the main effect of viewing condition was not significant (F(2,54) = 0.400, p = 0.672, \(\eta_{p}^{2}\) = 0.015). For secondary saccade accuracy, the main effect of viewing condition was significant (F(2,54) = 3.457, p = 0.049, η_p² = 0.114) but no post hoc Bonferroni pairwise comparisons yielded statistically significant differences (ps > 0.05). 

The LASSO regression analysis identified three reading-specific eye movement metrics (Task 6) —fixation count, average fixation duration, and progressive saccade count—as pivotal in elucidating the binocular advantage during reading. To determine whether the binocular and/or dominant eye advantage exist for these metrics, we used Generalized Linear Mixed Models (GLMMs) with participants and texts as random effects and revealed that the main effects of viewing condition on these three metrics were not statistically significant (all p > 0.05). 

Discussion

In the current study, we examined the binocular advantage in Chinese text reading. To ensure a comprehensive understanding and to avoid overestimating this binocular advantage, we simultaneously assessed the potential advantage of the dominant eye. We replicated previous observations that the right eye is the dominant eye for the majority of participants when tested with the sighting test (Ehrenstein, Arnold-Schulz-Gahmen, & Jaschinski, 2005). Based on the assignment of the dominant eye, we investigated text-based global metric (text reading time) and character-based local metric (character FFD) across three viewing conditions (binocular, dominant eye, and non-dominant eye). Our findings were unequivocal: binocular vision surpassed monocular viewing in both metrics, underscoring the inherent advantage of using both eyes for reading tasks. This aligns with prior research that has documented the binocular efficiency in text reading (Johansson et al., 2014a; Johansson et al., 2014b) and lexical processing (Chen et al., 2023; Jainta et al., 2014; Jainta & Jaschinski, 2012; Jainta & Joss, 2019; Jainta et al., 2017; Nikolova et al., 2018). 

Notably, we didn't observe the dominant eye advantage in text reading time and character FFD. This could be interpreted in two ways. One possibility is that the dominant eye simply does not hold an advantage in reading, echoing several related studies that failed to discern the dominant eye advantage in reading speed (Jainta & Jaschinski, 2012; Jainta et al., 2017; Johansson et al., 2014a; Johansson et al., 2014b). Alternatively, it could imply that the dominant eye identified through the hole-in-card sighting test does not necessarily correspond to a “reading dominant eye,” suggesting a discrepancy between sighting and reading dominance (Mapp, Ono, & Barbeito, 2003). 

For the binocular advantage in text reading time, none of the metrics from basic visual functions were associated or selected by LASSO regression for prediction. This suggests that although basic visual functions act as foundational prerequisites for reading and inherently demonstrate a binocular advantage (as seen in visual acuity and contrast sensitivity), variations in these functions under different viewing conditions do not affect the text reading time. This could be attributed to the suprathreshold nature of the natural reading stimuli used in the experiment, which were well above the perceptual thresholds of individuals with normal vision. Even though binocular vision typically offers superior basic visual functions compared to monocular vision, the basic visual functions of a single eye are sufficient to clearly perceive the suprathreshold reading stimuli. Consequently, at the current reading stimulus level, the differences in basic visual functions between monocular and binocular viewing in individuals with normal vision may be too small to produce noticeable reading differences. However, when reading stimuli are close to or below the perceptual threshold, reading speed may depend on variations in basic visual functions under monocular versus binocular viewing (Legge, Ross, Isenberg, & Lamay, 1992; Legge, Rubin, Pelli, & Schleske, 1985; Whittaker & Loviekitchin, 1993). 

Two metrics from basic oculomotor functions, primary saccade latency and primary saccade peak velocity, were correlated with the extent of binocular advantage in text reading time. Additionally, the binocular advantage was observed in these two metrics, consistent with previous studies (González et al., 2012; Niechwiej-Szwedo et al., 2012; Niechwiej-Szwedo et al., 2010). In the simple saccade task, the primary saccade latency and peak velocity following the presentation of the target serve as indicators of saccade initiation and execution, respectively (Gilchrist, 2011). Correlation results suggest that in binocular conditions, the shortened time required for planning and executing saccades may be associated with the binocular advantage in reading. However, these two metrics were not selected by LASSO regression, indicating that basic oculomotor functions are not the primary factor contributing to the observed binocular advantage in text reading. This could be attributed to the fact that, in individuals with normal vision, the basic oculomotor functions of a single eye, while inferior to those of binocular vision in certain aspects (e.g., fixation stability, saccade latency, and saccade velocity), are not abnormally impaired. When oculomotor functions reach an abnormal level—as in individuals with amblyopia (Bhutada et al., 2022), strabismus (Lions, Bui-Quoc, Seassau, & Bucci, 2013), macular disease (Crossland, Culham, & Rubin, 2004), or dyslexia (Stein, Richardson, & Fowler, 2000)—oculomotor dysfunction may play a significant role in reading impairment. 

The extent of binocular advantage in three reading-specific eye movement metrics—fixation count, average fixation duration, and progressive saccade count—was highly positively correlated with and effectively predicted the binocular advantage observed in text reading time. These findings suggest that although the reduction in each individual metric under binocular vision compared to monocular vision did not reach statistical significance, the combined reduction across the three metrics may contribute to a decrease in text reading time during binocular reading. In other words, the binocular advantage in text reading time may arise from a synergistic reduction in fixation count and progressive saccade count, along with a decrease in average fixation duration, when both eyes are engaged in the reading process. The reasons for the reduction in these three metrics during binocular reading may provide insights into the underlying mechanism of the binocular advantage in reading. 

The reduction of fixation count and progressive saccade count may reflect efficient parafoveal preprocessing of information from the perceptual span during binocular reading. Both foveal vision (2° of visual angle around the foveal center) and parafoveal vision (2° to 5° from the fixation point) play critical roles in reading. The perceptual span refers to the amount of useful information that can be extracted during a single fixation. Studies employing the gaze-contingent moving window paradigm have shown that the perceptual span in Chinese reading is asymmetrical, extending approximately one character to the left of the fixated character and up to three characters to the right (Inhoff & Liu, 1998; Zang, Liversedge, Bai, & Yan, 2011). During monocular reading, the increased number of fixations and progressive saccades may be necessary to ensure extensive foveal processing of information. In contrast, during binocular reading, this information may be sufficiently preprocessed in the parafovea, reducing the need for additional fixations and progressive saccades. One possible explanation is that the perceptual span is narrower in monocular reading than in binocular reading. Alternatively, the perceptual spans of monocular and binocular reading may be comparable, but binocular parafoveal preprocessing may be of higher quality, obviating the need for subsequent foveal processing. Future research is needed to distinguish between these possibilities. Despite these alternative explanations, both converge on the same conclusion that parafoveal preprocessing is more efficient during binocular reading. This conclusion is corroborated by empirical studies that have systematically isolated foveal and parafoveal processing in monocular and binocular reading conditions, and demonstrated that binocular parafoveal preprocessing, compared to its monocular counterpart, facilitates faster recognition of target words (Chen et al., 2023; Jainta et al., 2014; Nikolova et al., 2017; Nikolova et al., 2018). 

The reduction in average fixation duration observed during binocular reading can be interpreted from both behavioral and physiological perspectives. Behaviorally, the parafoveal-on-foveal effect suggests that efficient parafoveal preprocessing in binocular viewing can shorten the time needed for foveal processing (Kennedy, Pynte, & Ducrot, 2002). Physiologically, this phenomenon may be attributed to the accelerated processing capabilities inherent to binocular vision. This is supported by neural studies that demonstrate more extensive and faster cortical activation in response to binocular stimuli compared to monocular stimuli. Neuroanatomical studies have shown that although layer 4 neurons in the primary visual cortex are driven by one eye, neurons above and below layer 4 are predominantly binocular, responding to input from both eyes. Additionally, the presence of eye dominance columns indicates that binocular neurons respond more robustly to monocular stimuli from a specific eye of origin (Hubel & Wiesel, 1962; Wiesel & Hubel, 1974). Recordings from animal studies have found that binocular stimulation leads to higher neuronal firing rates than monocular stimulation (Hubel & Wiesel, 1959; Hubel & Wiesel, 1962; Hubel & Wiesel, 1970). Similarly, human EEG investigations indicate that binocular stimulation, as compared to monocular stimulation, elicits visual evoked potentials (VEPs) with larger amplitudes (Di Summa et al., 1999; Heravian et al., 1990; Nicol, Hamilton, Shahani, & McCulloch, 2011; Suter, Bass, & Suter, 1993; Sutija et al., 1990) and shorter latencies (diSumma et al., 1997; Shimoyama et al., 1999). Human fMRI studies also provide evidence that binocular stimuli enhance BOLD signals compared to monocular stimuli (de Best, Raz, Dumoulin, & Levin, 2018; Moradi & Heeger, 2009). 

Both correlation and regression analyses suggest that the binocular advantage in character FFD can be primarily explained by average fixation duration. This implies that the behavioral advantage in parafoveal preprocessing and the rapid neural processing associated with binocular vision may facilitate faster lexical recognition during reading. 

In conclusion, our research echoes previous studies by confirming the binocular advantage in Chinese reading, with binocular reading being approximately 4% faster than monocular reading in terms of text reading time. This advantage was not attributable to basic visual or oculomotor functions but was instead explained by metrics related to reading-specific eye movements (fixation count, average fixation duration, and progressive saccade count). On the one hand, our results exclude the possibility that basic visual and oculomotor functions play important roles in the binocular advantage in reading for individuals with normal vision. This finding is particularly intriguing, as it provides an important normative reference for individuals with impaired vision, in whom basic visual and oculomotor functions may serve as critical explanatory factors for reading performance. On the other hand, the concurrent reduction of fixation count, average fixation duration, and progressive saccade count in binocular reading may shed light on the underlying mechanism of the binocular advantage in reading. Specifically, the reduction in fixation count and progressive saccade count under binocular conditions suggests that efficient binocular parafoveal preprocessing may play a pivotal role in the binocular advantage during reading (Chen et al., 2023; Jainta et al., 2014; Kennedy et al., 2002; Nikolova et al., 2017, Nikolova et al., 2018). Moreover, the shorter average fixation duration implies that accelerated neural processing of visual information in binocular vision might also contribute to the binocular advantage in reading (de Best et al., 2018; Di Summa et al., 1999; diSumma et al., 1997; Heravian et al., 1990; Hubel & Wiesel, 1959; Hubel & Wiesel, 1962; Hubel & Wiesel, 1970; Moradi & Heeger, 2009; Nicol et al., 2011; Shimoyama et al., 1999; Suter et al., 1993; Sutija et al., 1990; Wiesel & Hubel, 1974). 

Acknowledgments

Supported by Beijing Natural Science Foundation (5222028); National Key Research and Development Program of China (2023YFC3604100); National Science and Technology Innovation 2030 Major Projects (2022ZD0204800); and National Natural Science Foundation of China (32071056 and 32100864). 

Commercial relationships: none. 

Corresponding authors: Fang-Fang Yan; Chang-Bing Huang. 

Emails: [email protected]; [email protected]. 

Address: Institute of Psychology, Chinese Academy of Sciences, Beijing 100101, China.