The effect of contrast on monocular versus binocular reading performance

Abstract

Abstract

Abstract:

Abstract  The binocular advantage in reading performance is typically small. On the other hand research shows binocular reading to be remarkably robust to degraded stimulus properties. We hypothesized that this robustness may stem from an increasing binocular contribution. The main objective was to compare monocular and binocular performance at different stimulus contrasts and assess the level of binocular superiority. A secondary objective was to assess any asymmetry in performance related to ocular dominance. In a balanced repeated measures experiment 18 subjects read texts at three levels of contrast monocularly and binocularly while their eye movements were recorded. The binocular advantage increased with reduced contrast producing a 7% slower monocular reading at 40% contrast, 9% slower at 20% contrast, and 21% slower at 10% contrast. A statistically significant interaction effect was found in fixation duration displaying a more adverse effect in the monocular condition at lowest contrast. No significant effects of ocular dominance were observed. The outcome suggests that binocularity contributes increasingly to reading performance as stimulus contrast decreases. The strongest difference between monocular and binocular performance was due to fixation duration. The findings may pose a clinical point that it may be necessary to consider tests at different contrast levels when estimating reading performance.

Introduction

It is well known that reading is impaired in the presence of binocular issues. Accommodative- or vergence-related deficits may cause blurred vision, diplopia, or asthenopic symptoms that affect reading speed and perseverance. There is however quite limited previous research describing the advantage of binocularity in reading performance. We have found a limited number of studies that have compared binocular and monocular reading performance using continuous text (Heller & Radach, 1999; Kanonidou, Proudlock, & Gottlob, 2010; Robinson, 1951), sentences (Jainta & Jaschinski, 2012), or word charts (Sheedy, Bailey, Buri, & Bass, 1986), that is reading both with and without context. Monocular reading resulted in a decrease of reading speed of up to 5% (Kanonidou et al., 2010; Robinson, 1951; Sheedy et al., 1986) along with increased fixation duration by up to 5% (Heller & Radach, 1999; Jainta & Jaschinski, 2012; Kanonidou et al., 2010). The length of progressive saccades during monocular reading has been found to both decrease (Heller & Radach, 1999; Kanonidou et al., 2010) and increase (Jainta & Jaschinski, 2012). 

Reading in daily life is done in a wide range of visual conditions where the quality of the text may be affected and suboptimal due to limited illumination or glare. Still, reading appears to be feasible in most cases. An additional interesting aspect is that experimental research on reading performance in visually challenging conditions has shown reading to be remarkably robust, for instance to induced blur (Jainta, Dehnert, Heinrich, & Jaschinski, 2011), filtering of spatial frequency (Jordan, McGowan, & Paterson, 2012; Kwon & Legge, 2012; Legge, Pelli, Rubin, & Schleske, 1985; Paterson, McGowan, & Jordan, 2013), and reduced stimulus contrast (Legge, Parish, Luebker, & Wurm, 1990; Legge, Rubin, & Luebker, 1987). This has led us to further investigate the contribution of binocular vision in the context of degraded stimulus properties and in this study specifically the effect of reduced stimulus contrast. 

During reading sensory input is acquired from an asymmetrical region around the fixation point called the perceptual span (Rayner, 1998, 2009). The perceptual span is strongly influenced by top-down processes hence its asymmetry weighted in the direction of reading (Rayner, 1998). Due to physiological retinal limitations only the characters within a limited central region, termed visual span (O'Regan, Lévy-Schoen, & Jacobs, 1983), can be resolved to the point of recognition. Outside this region the sensory input is coarser but provides preview information such as word length and spacing necessary for fluent reading. According to the visual span hypothesis, the size of the visual span has a direct effect on eye movements in reading. By varying the visibility properties of the stimuli, for example spacing, size, contrast, or spatial frequency, the visual span has been shown to shrink or extend with a codependency in binocular reading performance (Kwon & Legge, 2012; Legge et al., 2007; Legge, Mansfield, & Chung, 2001; O'Regan et al., 1983). 

Two important visual requirements have been identified to maintain habitual reading speed; namely to possess a visual acuity reserve, i.e., the print size relative to the visual acuity threshold (Whittaker & Lovie-Kitchin, 1993) and a contrast reserve, i.e., the print contrast relative to the contrast threshold (Legge et al., 1990; Legge et al., 1987). Through the interaction of the two eyes, the binocular visual performance is in certain respects, including acuity and contrast sensitivity, better than monocular, which may be of significance in particular when visual conditions are close to threshold. 

Research on the binocular summation effect in discrimination and recognition tasks has showed that the summation effect is strongly reduced as contrast is improved and the complexity of tasks are increased (Blake & Fox, 1973; Frisén & Lindblom, 1988). Bearse and Freeman (1994) investigated summation in orientation discrimination and found the greatest summation effect at 8% contrast but by 15% contrast the summation effect had diminished and monocular and binocular performance were approximately equal. Visual acuity has been shown to improve binocularly by 8%–11% at high contrast (Barany, 1946; Frisén & Lindblom, 1988; Heravian, Jenkins, & Douthwaite, 1990; Horowitz, 1949). At low contrast a further increase in binocular superiority has been observed; Home (1978) found an improvement in acuity by a factor of up to 1.4–1.5 when contrast was less than 10%. Banton and Levi (1991) measured Vernier acuity over a range of contrasts and found binocular thresholds to be approximately 50%–60% lower in the range from near threshold up to about 20 times the threshold. As contrast increases above these critical levels the summation effect rapidly diminishes. Jones and Lee (1981) compared monocular and binocular performances in a task involving identification of letters on cards at bright and dim light. Binocular performance meant fewer identification errors at both lighting conditions and a significant interaction effect showed binocularity to contribute more strongly at dim light where supposedly the contrast was lower. 

The binocular summation effect in discrimination outside the central foveal vision, i.e., theoretically a factor that could influence the width of the visual span, has been investigated in a limited number of studies. Zlatkova, Anderson, and Ennis (2001) used sinusoidal gratings to measure resolution acuity in foveal and peripheral (25° inferior) vision binocularly and monocularly. The binocular improvement in foveal vision was 5% while it was 16% in the periphery. Pardhan (2003) compared resolution acuity at fovea and at 8° eccentric positions (180/90/45/135) and with the stimuli presented at two levels of contrast, 95% and 7%. Foveal presentation in the horizontal meridian resulted in not significant summation effects of 5%–7% regardless of contrast. In the eccentric horizontal positions the summation effect reached 8%–16% with the low contrast stimuli consequently showing higher summation effects. 

Starting from the visual span hypothesis as a sensory limitation to higher processing during reading and applying the knowledge from the binocular summation literature, we hypothesized that the robustness in reading performance may stem from a gradually increasing binocular contribution as visual conditions get less favorable. The main objective of this study was to directly compare monocular and binocular reading performance at three levels of reduced stimulus contrast and to assess the level of binocular superiority as a measure of binocular contribution. A secondary objective was to assess any asymmetry in monocular performance and its possible correspondence to ocular dominance and also to compare dominant and nondominant eye to binocular performance. Ocular dominance is frequently referred to in experimental settings, for instance when one eye needs to be chosen, based on the assumption that performance may differ between dominant and nondominant eye. Research in support of this has found differences in search and recognition tasks (Money, 1972; Porac & Coren, 1979), saccade velocity (Oishi, Tobimatsu, Arakawa, Taniwaki, & Kira, 2005), and visually evoked potentials (Jagadamba & Karthiyanee, 2012). Other reports have found poor agreement between performance and dominance, for example in reading (Jainta & Jaschinski, 2012; Sheedy et al., 1986), and it has been suggested that ocular dominance only matters in monocular tasks (Mapp et al., 2003). There are different ways of classifying and testing for ocular dominance and the fact that dominance may change with different tests and test conditions (Evans, 2007) makes determination of dominance more complicated. Sighting tests (Ehrenstein, Arnold-Schulz-Gahmen, & Jaschinski, 2005; Mapp, Ono, & Barbeito, 2003; Rice, Leske, Smestad, & Holmes, 2008) are however a common way of deciding dominance and in this paper we refer to dominance as determined by sighting tests. 

Methods

The experimental design consisted of balanced repeated measurements where each subject silently read texts under all conditions while their eye movements were recorded. There were three levels of viewing conditions (monocular left, monocular right, and binocular) and three levels of stimulus contrast (Michelson 10%, 20%, and 40%). The order of the conditions and stimulus contrast were counterbalanced. Each text was read once by each subject. The dependent variables examined were reading speed, fixation duration, progressive saccade length, and proportion of regressive saccades. 

The inclusion criteria were age 18–40 years, nonsymptomatic at reading and near work, normal binocular vision, no diagnosis of reading difficulty or ocular disease, and an ability to read Swedish fluently. Subjects requiring eye glasses were not included because the frames of eye glasses frequently obstruct the eye tracker's view of the subject's eyes causing quality issues with the eye movement recordings. Subjects wearing contact lenses were allowed to have them. 

The study adhered to the tenets of the Declaration of Helsinki and was approved by the regional ethics review board in Stockholm, EPN (2012/1447-31/1). 

Subjects

A total of 18 subjects with a mean age of 24.0 (3.2) years were included in the study. Another six subjects participated but were excluded from the analysis; five due to not meeting the vision criteria and one due to quality issues in eye movement recordings. Eight subjects (44%) were female. All subjects were fluent in Swedish and had normal binocular vision status with stereovision of 60 s of arc or better. Details of visual status are available below. 

Apparatus

Eye movements were recorded with a Tobii T120 eye tracker (Tobii Technology, Stockholm, Sweden, http://www.tobii.com). The eye tracker is display mounted and records eye movements at 120 HZ using infrared video technology. The accuracy is 0.5° according to the manufacturer's specification. A nine-point sequence binocular calibration preceded each recording. During monocular reading one of the eyes was covered by an infrared transmissible occluder allowing binocular recording that is required by the eye tracker. 

Processing of the recorded eye movement data was done offline using Visiolyzer, a software developed at our lab. A dynamic dispersion model was applied to define fixations and saccades. A fixation was defined when the center of gravity of recorded fixation points stayed within a dispersion limit radius of 0.75° (2.7 character spaces) for a minimum of 50 ms. When the center of gravity moved out of the dispersion limit it was recorded as an initiated saccade. Progressive and regressive saccades were defined as a rightward and leftward movement, respectively, between two adjacent fixations. 

Stimuli

Nine texts from the International Reading Speed Texts (IReST) (Trauzettel-Klosinski & Dietz, 2012) were used in the experiment. The IReST texts are validated in 17 languages including Swedish. The Swedish texts have the exact same number of words (146), characters (684), and lines (16), and all texts have the same readability index (LIX 35) corresponding to the readability of fiction books. The mean word length is 4.61 (0.01) characters. The texts were presented as single paragraphs subtending 14.3° by 17.1° on the integrated screen of the eye tracker, which was placed 60 cm from the eyes of the subject (Figure 1). The font was Helvetica and the mean character width was 0.28°. Each IReST text was rendered to three levels of contrast (Michelson contrast 10%, 20%, and 40%) by keeping the background luminance constant at 210 candela/m² while the luminance of print was adjusted. The luminance levels were measured with a Hagner S4 lightmeter (Hagner AB, Solna, Sweden, http://www.hagner.se). The lowest contrast level (Michelson 10%) was decided based on screen resolution limitations and pilot testing. 

Figure 1

 

Stimuli used in the experiment with screen dimensions and visual angles. The contrast may differ from original when reproduced in paper.

Procedure

The subjects were first informed verbally and in writing about the experiment. All subjects gave informed written consent prior to the experiment. 

Second, the visual status of the subjects was assessed, including a brief history, visual acuity at far and near (Logarithmic Visual Acuity Chart 2000 “ETDRS” 4 m/40 cm, Precision Vision, http://www.precision-vision.com), stereo vision acuity (TNO random dot test, Laḿeris Ootech B.V., Nieuwegian, The Netherlands), cover test and fusional reserves (prism bar), eye motility, near point of accommodation, and near point of convergence (R.A.F. near point rule, Clement Clarke International, UK). A translated version of the Revised Convergence Insufficiency Symptom Survey (CISS) (Borsting et al., 2003; Rouse et al., 2004) was used to assess symptoms experienced during day-to-day near distance work. 

Next, ocular dominance was assessed at far (400 cm) and near (40–50 cm). At far the hole-in-card sighting test (Ehrenstein et al., 2005; Mapp et al., 2003; Rice et al., 2008) was used. In this test the subject held a card (20.0 × 12.8 cm) with both hands at arm's length looking at a black cross at 4-m distance through a 3.0-cm diameter hole in the center of the card. The subject was instructed to look with both eyes at all times. When the subject reported to have aligned the hole in the card and the black cross each eye was covered one at a time and the subject was asked to report which of the eyes that kept the alignment. This eye was recorded as the dominant eye. The test was repeated four times for all subjects. For near dominance assessment the near hole-in-card sighting test (Rice et al., 2008) was used. The same procedure as in the far assessment was used. 

The subject was seated comfortably and unrestrained in a steady chair in front of the eye tracker. The position of the eye tracker was adjusted horizontally and vertically to optimize the signal quality. A test run was made in order for the subject to get acquainted with the procedure. Initially there was the calibration procedure that was triggered by the test leader. Next, the subject read nine texts under the different reading conditions. The subject started each reading session by pressing space bar on a keyboard whereby a fixation cross was presented at the center of the screen. Two seconds later the full text was presented on the screen. The subject was instructed to fixate the cross carefully and then to start reading at his or hers own pace as soon as the text appeared and to finish by pressing space bar again as soon as the last word was read. 

The subjects were informed that the texts would be presented at different levels of contrast and they were instructed to read the texts for comprehension. They were also informed that they would need to answer questions about the content after finishing reading. There were four multiple choice questions with three alternatives for each text. After each text was finished the subject answered the multiple choice questions in writing. The experimenter was present at all times giving instructions before each reading session and monitoring the eye movements online. The total time for the subject's participation was 45 min to 1 hr and the average total time spent on reading nine texts was 7.1 min. 

Visual status

All included subjects reported to be symptom free when reading and all scored less than 21 (Rouse et al., 2004) on the CISS inventory. The mean binocular summation effect in visual acuity was 14%, i.e., the minimum angle of resolution was smaller (better) by this percentage during the binocular condition. In Subject 16 the cover test revealed a greater level of exophoria; however, because reporting and scoring showed the subject to be symptom free, displaying compensating fusional reserves and readily achieving 30 s of arc on the TNO test, this subject was included. (Table 1). 

Table 1

 

Details of visual status.

Subject	Age	Visual acuity far (LogMar)			Visual acuity near (LogMar)			Stereo vision test (Sec. Arc)	Covertest 400 cm (Prism Diopt.)	Covertest 40 cm (Prism Diopt.)	Breakpoint divergence 400 cm (Prism Diopt.)	Breakpoint convergence 400 cm (Prism Diopt.)	Breakpoint divergence 40 cm (Prism Diopt.)	Breakpoint convergence 40 cm (Prism Diopt.)	Nearpoint convergence (cm)	Nearpoint accommodation (cm)	Symptom score
		Right eye	Left eye	Binoc.	Right eye	Left eye	Binoc.
1	30	−0.08	−0.08	−0.14	−0.04	−0.04	−0.08	60	0	0	−12	12	−8	20	5	9	12
2	21	0.00	0.04	−0.08	0.00	0.08	−0.10	30	0	0	−10	20	−10	35	5	6	9
3	27	−0.04	0.00	−0.08	0.00	0.10	−0.06	30	0	−2	−12	20	−12	30	5	10	7
4	24	0.10	0.02	−0.06	0.04	0.02	0.02	60	0	0	−8	35	−8	25	8	10	18
5	20	−0.20	−0.20	−0.24	−0.04	0.00	−0.10	30	0	0	−6	20	−6	30	5	7	12
6	21	0.00	0.00	−0.08	0.00	0.06	−0.10	60	0	−6	−4	12	−8	25	7	8	11
7	21	−0.16	−0.10	−0.16	−0.04	0.00	−0.10	60	0	1	−16	20	−8	25	5	10	20
8	30	−0.06	0.00	−0.12	0.10	0.02	0.00	60	0	−6	−8	18	−8	12	10	10	16
9	28	−0.18	−0.18	−0.18	0.00	0.00	−0.06	30	0	0	−4	16	−6	16	10	10	15
10	24	−0.06	−0.08	−0.18	0.00	0.00	−0.10	60	0	0	−12	20	−12	25	5	8	4
11	21	−0.16	−0.18	−0.16	0.00	0.00	−0.08	30	−1	−4	−10	8	−10	16	5	6	7
12	21	−0.10	−0.12	−0.16	0.08	0.00	−0.06	30	0	−4	−10	16	−14	20	5	7	2
13	21	0.00	−0.10	−0.10	0.00	−0.10	−0.12	60	0	0	−10	12	−10	18	5	7	9
14	22	0.00	0.00	−0.14	0.00	0.00	−0.10	60	0	0	−12	10	−12	25	5	8	10
15	27	0.14	0.12	−0.06	0.10	0.10	−0.06	30	0	0	−4	25	−4	14	5	9	14
16	25	−0.18	−0.10	−0.20	−0.10	−0.08	−0.14	30	−12	−18	−25	35	−25	35	8	9	17
17	23	0.06	0.14	0.00	0.00	0.00	−0.06	30	0	2	−10	10	−6	25	5	9	11
18	26	0.00	−0.14	−0.10	0.00	0.00	−0.10	30	0	0	−4	8	−8	25	7	9	18

Ocular dominance

Test responses were enumerated according to a relative score (Zeri, De Luca, Spinelli, & Zoccolotti, 2011), i.e., the response at each trial was set as −1 in the case of left eye dominance and +1 in the case of right eye dominance. The relative score was calculated as the algebraic sum of responses divided by the number of trials. This resulted in a score ranging from −1 in the case of left eye dominance in all trials, 0 in the case of no dominance (50-50 mixed responses), and +1 in the case of right eye dominance in all trials. 

According to the hole-in-card sighting test performed at distance 10 subjects (56%) were right-eye dominant, seven subjects (39%) were left-eye dominant, and one subject (5%) displayed no dominance. In the near sighting test the distribution was 12 subjects (67%) right-eye dominant, five subjects (28%) left-eye dominant and one subject (5%) with no eye dominance. Four subjects (22%) switched dominance between far and near assessment. The subject displaying no dominance according to the tests was assigned to right-eye dominance in the analysis because this was the eye used first in both distance and near test. 

Statistics

IBM SPSS Statistics 22 (IBM Corp., U.S.A., http://www.ibm.com) was used for statistical analysis and the significance level was set to 5% (α = 0.05). A two-way repeated measures analysis of variance (ANOVA) (univariate) was conducted to examine the effect of contrast and viewing condition on reading speed and eye movements. The factor contrast had three levels in all analysis. Viewing condition had two levels in the analysis of monocular versus binocular performance and three levels in the analysis of binocular versus dominant and nondominant performance. A one-way repeated measures ANOVA or pairwise comparisons were used for follow-up tests in the event of a significant interaction effect. Any violation of the sphericity assumption was adjusted for with the Greenhouse-Geisser correction. A Bonferroni adjustment for multiple comparisons was applied in pairwise follow-up tests. 

Results

The subjects' participation in the experiment involved one visit where all data were collected. Each subject read three texts under binocular conditions and six texts under monocular conditions (three with either eye). Within each viewing condition the subject read one text at 10% contrast, one at 20% contrast, and one at 40% contrast. Reading speed was calculated as words read per minute (WPM) based on total reading duration and comprehension scores were enumerated as the percentage of correct answers. The proportion of regressive saccades was calculated as the number regressive saccades divided by the sum of number of progressive and regressive saccades. Below we first report the comparison of monocular and binocular reading; next we report on the effect of dominance on monocular reading; and finally we compare reading with the dominant or nondominant eye to binocular reading. 

Monocular versus binocular reading

Data from the left and right monocular tests were averaged to produce an average monocular value for comparison with the binocular value. Prior to averaging, the data from left and right eye were confirmed to be nonsignificantly different. Monocular reading speed decreased by 6.9% at 40% contrast, by 9.3% at 20% contrast, and by 21.1% at 10% contrast compared to binocular, but there was no significant interaction effect (Figure 2). There was a statistically significant difference in the main effect of viewing condition, F(1, 17) = 22.74, p < 0.01, η²_partial = 0.57. The mean monocular reading speed was slower than binocular by 22 WPM. A pairwise comparison showed binocular reading speed to be significantly higher at 10% contrast (p < 0.01) and 20% contrast (p = 0.02), while it was marginally significant at 40% contrast (p = 0.05). 

Figure 2

 

Binocular (squares) and monocular (circles) reading speed at the three levels of contrast. Error bars represent the standard error of the mean.

A significant difference was also found in the main effect of stimulus contrast, F(2, 34) = 32.44, p < 0.01, η²_partial = 0.66. The mean reading speed was slowed by 11.5 WPM between the 40% and 20% contrast and by 31.9 WPM between the 20% and 10% contrast level. A pairwise comparison showed significant differences between all three contrast levels, that is between 40% and 20% contrast (p = 0.03), between 40% and 10% (p < 0.01), and between 20% and 10% (p < 0.01). 

The mean fixation duration at monocular reading compared to binocular increased by 8.5% at 40% contrast, by 6.9% at 20% contrast, and by 24.6% at 10% contrast (Figure 3). There was a statistically significant interaction effect between effects of viewing condition and stimulus contrast, F(2, 34) = 14.89, p < 0.01, η²_partial = 0.47. 

Figure 3

 

Binocular (squares) and monocular (circles) fixation duration at the three levels of contrast. Error bars represent the standard error of the mean.

Follow-up tests were conducted to explore the combined effects of viewing condition and contrast level. Tests of viewing condition within stimulus contrast showed monocular fixation duration to be significantly longer at all three contrast levels (p < 0.01) and thus demonstrating a general effect of viewing condition on fixation duration. 

Follow-up tests of the effect of contrast level within viewing condition showed significant differences for both monocular, F(1.45, 24.61) = 51.95, p < 0.01, η²_partial = 0.75, and binocular, F(2, 34) = 7.27, p < 0.01, η²_partial = 0.30, reading. Pairwise comparisons of stimulus contrast within the monocular viewing condition showed fixation durations at 10% contrast to be significantly longer compared to both 20% and 40% contrast (p < 0.01). Within the binocular viewing condition the pairwise comparisons showed fixation durations at 40% contrast to be significantly shorter than both 20% and 10% contrast (p < 0.01). 

The interaction effect thus turned out as significantly increased fixation duration between 10% and 20% contrast during monocular reading, while the fixation duration was maintained during binocular reading. Secondly, the fixation duration during binocular reading was significantly reduced between 20% and 40% contrast, while it did not change significantly during monocular reading. 

For progressive saccade length there was no significant interaction effect (Figure 4). The two-way repeated measures ANOVA did however indicate a significant difference in the main effect of viewing condition, F(1, 17) = 4.95, p = 0.04, η²_partial = 0.23. The mean monocular saccade length was 0.2 characters longer than binocular. 

Figure 4

 

Binocular (squares) and monocular (circles) progressive saccade length (character spaces) at the three levels of contrast. Error bars represent the standard error of the mean.

In the main effect of contrast there was a significant difference, F(2, 34) = 19.72, p < 0.01, η²_partial = 0.54. The saccade length at 20% contrast was significantly prolonged compared to 10% contrast (p < 0.01) and also at 40% contrast compared to 10% (p < 0.01). 

The mean proportion of regressive saccades at monocular reading was 0.19 ± 0.06 at 10% and 20% contrast and 0.18 ± 0.05 at 40% contrast. At binocular reading the proportions were 0.19 ± 0.05, 0.18 ± 0.05, and 0.19 ± 0.06 at 10%, 20%, and 40% contrast, respectively. The statistical analysis revealed no significant interaction effects or differences in main effects. 

The comprehension scores in the monocular condition were 94.4 ± 0.1%, 95.8 ± 0.1%, and 93.7 ± 0.1% at 10%, 20%, and 40% contrast, respectively. In the binocular viewing condition the scores were 94.4 ± 0.1 %, 95.8 ± 0.1%, and 95.8 ± 0.1%, respectively. A Friedmann ANOVA showed no significant differences. 

Dominant versus nondominant eye reading

Irrespective of dominance determined at distance or near, there were no statistically significant interaction effects or differences in main effects in reading performance between dominant and nondominant eye. Nor did pairwise comparisons within condition and contrast show any significant differences. 

Dominant and nondominant versus binocular reading

Irrespective of dominance determined at distance or near the results of comparing dominant and nondominant eye versus binocular reading performance was similar with the significant differences falling out in the exact same performance measures. We have therefore limited this section to only include the effect of ocular dominance as determined at distance because we believe this is the more common testing distance. 

There were significant differences in reading speed in the main effects of viewing condition, F(2, 34) = 9.74, p < 0.01, η²_partial = 0.36, and stimulus contrast, F(2, 34) = 38.90, p < 0.01, η²_partial = 0.70 (Figure 5). The mean reading speed of dominant eye reading was slower than binocular by 23.6 WPM (p < 0.01), and the nondominant eye was slower than binocular by 20.4 WPM (p < 0.01). Reading speed differed significantly between all three contrast levels; at 10% contrast reading was slower by 34.4 WPM compared to 20% contrast (p < 0.01) and slower by 46.5 WPM compared to 40% contrast (p < 0.01); reading at 20% contrast was slower by 12.2 WPM compared to 40% contrast (p = 0.04). 

Figure 5

 

Binocular (squares), dominant eye (diamonds), and nondominant eye (triangles) reading speed (WPM) at the three levels of contrast. Error bars represent the standard error of the mean.

In fixation duration there was a significant interaction effect between the effects of viewing condition and stimulus contrast, F(1.91, 32.53) = 4.46, p = 0.02, η²_partial = 0.21 (Figure 6). At 10% contrast the fixation duration at dominant eye reading was increased by 55.5 ms compared to binocular and fixation duration at nondominant eye reading was increased by 66.7 ms compared to binocular. Follow-up tests comparing contrast within condition showed significant differences in fixation duration for dominant eye, F(1.50, 25.55) = 15.71, p < 0.01, η²_partial = 0.48, nondominant eye, F(1.24, 21.02) = 21.13, p < 0.01, η²_partial = 0.55, and binocular, F(2, 34) = 7.27, p < 0.01, η²_partial = 0.30. The fixation duration for dominant (p < 0.01) and nondominant eye (p < 0.01) both increased significantly between 10% and 20% contrast, while it did not change significantly for binocular reading. In binocular reading the fixation duration decreased significantly between 20% and 40% contrast (p < 0.01), while it did not change significantly for dominant and nondominant eye reading. Consequently, the pattern of the interaction effect when comparing dominant and nondominant eye separately to binocular looked similar to the comparison between averaged monocular performance versus binocular. 

Figure 6

 

Binocular (squares), dominant eye (diamonds), and nondominant eye (triangles) fixation duration (ms) at the three levels of contrast. Error bars represent the standard error of the mean.

For progressive saccade length there was no significant interaction effect. The main effect of viewing condition was almost marginally significant (p = 0.05) where the saccade length of the nondominant eye appeared to be more prolonged (Figure 7). 

Figure 7

 

Binocular (squares), dominant eye (diamonds), and nondominant eye (triangles) progressive saccade length (character spaces) at the three levels of contrast. Error bars represent the standard error of the mean.

For the main effect of stimulus contrast there was a significant difference, F(2, 34) = 25.21, p < 0.01, η²_partial = 0.60 (Figure 7). The mean saccade length at 10% contrast decreased by 0.6 characters compared to 20% contrast (p < 0.01) and by 0.7 characters compared to 40% contrast (p < 0.01). 

There were no significant interaction effects or differences in main effects in the proportion of regressive saccades between dominant, nondominant, and binocular reading. 

Discussion

In accordance with our expectations binocular reading proved to be more robust to the contrast reduction demonstrating a higher reading speed at all three contrast levels. The binocular advantage in the form of faster reading was highest (21%) at lowest contrast, with an apparent step reduction occurring at 20% contrast resulting in 9% faster binocular reading, followed by an apparent beginning level off to 7% faster binocular reading at 40% contrast. To our knowledge this has not been shown in previously published research and we can therefore not relate these findings directly to those of others. The step change in reduced reading speed between 10% and 20% contrast is however comparable to the findings by Legge et al. (1990) and Legge et al. (1987) who found a steeper decline in binocular reading speed to begin between 10%–30% contrast for text with similarly sized characters. In the present study we did however not find a statistically significant interaction effect. The comprehension score did not differ significantly between any of the reading conditions in this experiment. This is in part supported by previous research where no differences in comprehension score were found when altering the visual quality of the stimulus, for instance blurring of text (Jainta et al., 2011) and filtering of spatial frequency (Jordan et al., 2012). 

Eye movements

The strongest difference between monocular and binocular reading was prolonged fixation duration. As the contrast was reduced the fixation duration increased further, displaying a significant interaction effect. 

The increased fixation duration during monocular reading agrees with previous studies (Heller & Radach, 1999; Jainta & Jaschinski, 2012; Kanonidou et al., 2010); however, the durations in the present experiment were markedly longer (9%–24%), which is likely to be due to the lower stimulus contrast. Prolonged fixation durations has also been found at degraded stimulus properties such as reduced contrast (Legge, Ahn, Klitz, & Luebker, 1997), blurred text (Jainta et al., 2011), and reduced spatial frequency (Paterson et al., 2013). The significant interaction effect showed that contrast had an effect on fixation duration that was more adverse in the monocular condition at 10% contrast. This we believe to be an indication that the removal of binocular vision had an effect on at least some of the processes during the fixation as the stimulus properties got less favorable. The summation literature has shown that the binocular summation effect in discrimination tasks grows stronger with decreased contrast (Bearse & Freeman, 1994; Blake & Fox, 1973; Frisén & Lindblom, 1988; Jones & Lee, 1981) and with increased eccentricity (Pardhan, 2003). The visual information acquired during a reading fixation that has an effect on eye movements ranges from detailed information, letters, near the fixation point to more crude information (outline of text, word length, and spacing) towards the outer part of parafoveal region. Legge et al. (1997) found the reading time per word to be increasingly dependent on word length with reduced contrast. The removal of binocular vision combined with reduced contrast may be a contributing factor to the prolonged fixation durations possibly originating in a shrinking visual span associated with greater difficulty to discriminate letters or affecting the process of saccade planning, or both. 

The analysis of progressive saccade length indicated a prolongation during monocular reading compared to binocular. However, the difference was marginally significant and only when data from the left and right monocular tests were averaged to an average monocular value. When analyzing viewing conditions divided into binocular, dominant, and nondominant eye, the saccades of nondominant eye appeared to be more prolonged; however, the statistically significant difference then diminished due to the adjustments for multiple comparisons. We may therefore speculate that the observed prolongation of progressive saccade length in the present experiment is not a strong effect of viewing condition and that it therefore may be difficult to draw conclusions with the present sample size. Previous studies with similar sample sizes showed differences in the same order of magnitude, and the saccade length was in one case found to increase (Jainta & Jaschinski, 2012) and in another to decrease (Kanonidou et al., 2010). A power analysis (G*Power 3.1.7) made based on the outcomes of the present experiment indicates the requirement of 177 subjects to obtain a statistically significant effect (effect size = 0.21, α = 0.05, power 1-β = 0.80). 

The mean progressive saccade length did however decrease significantly as contrast declined. Previous research on reading of text of degraded visual quality found no difference (Jordan et al., 2012) or only a slight decrease (Jainta et al., 2011) in length of progressive saccades. Legge et al. (1997) found an approximately equal distribution between increased fixation duration and decreased progressive saccade length with decreased stimulus contrast, which was considered to be in support of a reduced visual span. In the present experiment viewing was reduced to monocular on top of reduced contrast, and still the progressive saccade length did not appear to differ significantly between viewing conditions. Because the proportion of regressive saccades or comprehension scores did not change significantly either, it appears that when forced to read at these challenging conditions, performance can be maintained at the cost of slowed reading where increased fixation duration appears to be the strongest difference. 

Our secondary objective was to assess the role of ocular dominance on performance. In accordance with previous research (Jainta & Jaschinski, 2012; Sheedy et al., 1986) we found no significant monocular asymmetry in performance related to dominance, nor did we identify any significant differences when comparing the dominant and nondominant eye separately to binocular performance. It suggests that ocular dominance has limited effects on performance when healthy subjects with normal binocular vision read paragraphs of text. However, in the present study we physically occluded the fellow eye during monocular condition. It can therefore not be excluded that monocular asymmetries, possibly dominance related, may occur during dichoptic presentation of stimuli, i.e., when both eyes simultaneously receive visual input but a specific object is presented to only one of the eyes. Previous research applying visual search tasks found monocular asymmetries in performance during dichoptic presentation of stimuli but not in purely monocular viewing conditions (Money, 1972; Porac & Coren, 1979). 

A limitation of this study is that contrast sensitivity was not measured. The included subjects were indeed fairly young and with a normal visual status, but it cannot be excluded that either eye may have had different contrast threshold. In the event that contrast threshold was substantially unequal and the stimuli contrast was much closer to threshold in one eye, it may have led to a very large binocular superiority effect. Upcoming experiments will incorporate contrast sensitivity to account for this effect and also to correlate contrast sensitivity and reading performance. Another aspect is the potential effect of binocular rivalry during monocular viewing. When patching one eye, rivalry may disrupt monocular viewing in the fellow eye; that is, the subject may experience transient symptoms of blur or blankouts. While the occluded eye was not completely patched in this experiment, and thereby possibly limiting factors such as inter ocular light adaption effects, we cannot exclude the potential contribution of binocular rivalry to the difference between monocular and binocular performance. 

More studies will be required to establish to what extent binocularity contributes to reading performance and the mechanisms behind it. Experiments including lower contrast levels may provide a more nuanced picture of binocular contribution and possibly reveal further interaction effects in reading performance. Involving subjects of different age groups may present other interesting findings given that eye movement pattern in older readers may differ from younger readers (Rayner, 2009), possibly related to changes in sensitivity to certain spatial frequencies (Paterson et al., 2013). With this study we have however found some quite striking differences between monocular and binocular reading performance related to stimulus contrast. Binocularity appears to provide an increasing advantage with declining contrast, mainly for the processes affecting fixation duration. The findings may in part explain the robustness of reading performance to degraded stimulus contrast. This may pose a further clinical point on the necessity to objectively evaluate reading performance in the presence of visual disturbances. It has previously been found that objectively measuring reading speed rather than relying on patient-reported outcome measures allows for an improved discrimination of the reading ability and also an earlier predictor of a decline in reading performance (Guralnik, Branch, Cummings, & Curb, 1989; Rubin, 2013). Given that reading in daily life needs to be feasible in a variety of visual conditions, sometimes suboptimal, it may be necessary to consider tests at different contrast levels when estimating reading performance. This could provide a sensitive functional measure of reading ability before and after treatment of visual disturbances affecting binocular vision. 

Acknowledgments

The work presented in this paper was supported by the Sigvard and Marianne Bernadotte Research Foundation for Children Eye Care. 

Commercial relationships: none. 

Corresponding author: Jan E. Johansson. 

Email: [email protected]. 

Address: Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden.