Up/Down Anisotropies of Vertical Saccades in Healthy Children From 6 to 10 Years of Age

Abstract

Purpose.: Although the overall development of saccades in children has recently gained increasing interest, the precise characteristics of vertical saccades remain understudied. This study focused on the development of vertical saccades and their interaction with vergence movements.

Methods.: Thirty-one children (mean age: 7.33 ± 0.21 years) performed vertical saccades with an eccentricity of 7.5° at 40 and 150 cm viewing distance, in a simultaneous paradigm (simultaneous offset of central dot and onset of peripheral target).

Results.: The results revealed shorter latency, more vertical conjugate postsaccadic drift after upward saccades, and less horizontal vergence during and after upward saccades in children. Furthermore, the intrasaccadic convergence decreased progressively with children's age for upward saccades. Relative to adult behavior, children present more hypometric saccades, longer latencies, higher vertical disconjugacy, and a different pattern of horizontal vergence during vertical saccades.

Conclusions.: This result suggests that up/down asymmetries are built progressively in mutual interaction with a perceptive peripheral bias such as up being perceived as far and down as near.

In everyday life, we make two to three saccades per second. These rapid stereotyped eye movements direct the image of the object of interest to the fovea in order to increase its visibility. Horizontal saccades have been repeatedly characterized. By contrast and despite their implications in reading, exploration of visual three-dimensional space, locomotion, and navigation, vertical saccades remain understudied. Unlike horizontal saccades, vertical saccades present several up/down anisotropies. For instance, Collewijn et al.¹ used the scleral sensor coil technique to investigate four adult subjects performing vertical saccades with a large range of amplitude (1.25°–70°). Their results revealed that upward saccades tend to undershoot the target (by approximately 10%), whereas downward saccades tend to overshoot it. More recently, Yang and Kapoula² obtained similar results in a large population of young adults and a group of elderly subjects. 

Other studies focusing on the latency of vertical saccades during a gap and overlap paradigm^3–5 revealed that most subjects presented reaction time anisotropy during both tasks. In particular, the study demonstrated shorter latency for upward than for downward saccades. A similar finding was also observed in monkeys.⁶ This effect was attributed to a different processing of the visual information for the upper and lower visual fields. In other words, the visual system would respond differently to stimuli received by different quadrants of the field. This hypothesis fit with the anatomical differences observed in the retina,⁷ as well as in the ventral and dorsal extrastriate cortex, dedicated to the superior and inferior contralateral quadrant respectively,⁸ yet, and in line with the study by Miller,⁹ the study by Yang and Kapoula¹⁰ showed that young adults and elderly persons present no significant up/down latency differences. It should be noted that even in the studies by Goldring and Fischer⁴ and Honda and Findlay,³ not all subjects presented this anisotropy (9 of 12⁴ and 9 of 13³). Overall, the up/down latency differences seem to present considerable intersubject variability. 

Vertical saccades also present upward and downward differences in terms of binocular coordination. Collewijn et al.¹ and Enright¹¹ found that the vertical binocular coordination of subjects' saccades is almost perfect for both upward and downward directions. However, their results revealed a difference in terms of horizontal vergence during vertical saccades: a divergence movement occurs during upward saccades, as it does for horizontal saccades, whereas a convergence occurs during downward saccades. 

How do all these aspects, accuracy, latency, binocular coordination, and horizontal vergence, relate to vertical saccade behavior in children? To our knowledge, only one study examined the characteristics of vertical saccades in children and teenagers (8–19 years of age).¹² The investigators recorded subjects' eye movement during horizontal saccades from 10° to 15° target eccentricity and vertical saccades from 5° to 10° target eccentricity and examined their latency, accuracy, and peak velocity. The authors found that although the latencies decreased with age, the accuracy and mean velocity were constant over time. In particular, latency to 5° downward saccades was significantly longer than that of all other saccades, and there was a negative correlation between the latency of vertical saccades and age. Finally, there were no differences in the peak velocity for upward and downward saccades. 

Here, we proposed an integrative study examining all aspects of vertical saccades: latency, accuracy, duration, velocity, vertical binocular coordination, horizontal vergence, and postsaccadic drift after vertical saccades. The study was conducted with a group of children 6 to 10 years of age and focused primarily on the examination of upward and downward saccade anisotropies during the children's development. 

Materials and Methods

Subjects

Thirty-one children participated in this study, ranging from 6 to 10 years of age (mean age: 7.33 ± 0.21 years old); nine children were between 6 and 7 years old, nine were between 7 and 8 years old, six were between 8 and 9 years old, and seven were between 9 and 10 years old. They were recruited through contacts with friends and families. All subjects were healthy, did not present neurologic, vestibular, or ophthalmologic symptoms that could affect their performance, and had normal or corrected-to-normal vision (the range of visual acuities was 8/10 to 12/10, and range of stereoacuities were 30 to 60 arc seconds. The dominant eye was measured with the unilateral cover test: the subject fixated on a target at 5 m, and the orthoptist covered alternately each eye, observing the viewing eye. If the viewing eye did not move, then it was considered the dominant eye. This test is one of the most commonly used techniques in clinical evaluation. The stereoacuity was measured with the TNO test (test of stereopsis). The investigation adhered to the principles of the Declaration of Helsinki and was approved by our institutional human experimentation committee, the Comité de Protection des Personnes Ile de France VI (no. 07035), Necker Hospital in Paris, France. Informed parental consent was obtained for each subject. 

Visual Display

The visual display consisted of light-emitting diodes (LEDs) mounted on a vertical table, adjusted to subjects' eyes, and placed at 40 or 150 cm away from subjects' faces. The subject's chair was moved farther from or closer to the table to obtain different distances. Subjects were seated in an adapted chair with a chin rest. Five LEDs were used. The LED at the center of the visual field served as the initial fixation light. Two LEDs were placed at the top and at the bottom of the visual field (7.5°), and two others were placed to the left and to the right of the visual field (10°) (Fig. 1A). 

Figure 1

 

Experimental setup. The table was adjusted vertically. The subject was seated in a chair facing the vertical table, the central fixation LED at eye level. The other LEDs were at 7.5° from the center, up or down; one eccentric LED at a time was on (A) and the experimental paradigm used for the simultaneous condition: when the central fixation point was turned off, the eccentric target appeared immediately (B).

Table

 

Summary of Results and Comparison With Adults

 

Oculomotor Task and Procedure for Calibration

A calibration test was run first, consisting of an LED target stepping by 10° as follows: from center to left, center, right, center, up, center, down, center; this sequence was repeated twice. All LEDs (center and eccentric) stayed illuminated for 1500 ms. Subjects were asked to make a vertical saccade to the target LED as rapidly and accurately as possible. Two calibration blocks were performed at 40 and 150 cm, respectively. In each block, vertical saccades were randomly interleaved between the upward and downward directions. 

Oculomotor Task and Procedure for Vertical Saccades

A simultaneous paradigm was used. At every trial, the central LED at eye level was presented first for 1500 ms. It was then switched off, and simultaneously, the target LED appeared (either up or down). Each block contained 20 trials (10 up and 10 down) that were randomly interleaved. This test was repeated twice at 40 cm from the screen (vergence angle of 9° required) and at 150 cm (vergence angle of 2.5° required) (Fig. 1B). 

Eye Movement Recording

Vertical eye movements were recorded binocularly with the Chronos (Chronos Vision, Berlin, Germany; available in the public domain at www.chronos-vision.de) rapid video eye tracker, using a high-frame rate complementary metal–oxide–semiconductor (CMOS) sensor.¹³ This system has an optimal resolution less than 0.1° for vertical eye movements. The sample rate was set at 200 Hz, and the relevant image data (i.e., during each trial) was stored on the hard drive for offline analysis. Eye movements were recorded with the subject's head in a head rest to avoid head movements that could interfere with the calculation of the accuracy of saccades and of vergence. No visible head movement was noticed by the investigator as the duration of the test was relatively short. Ideally, simultaneous head movement recording would be of interest, but this is a common limitation of studies in children. 

Data Analysis

The system gives two individual calibrated eye position signals, from which we calculated the conjugate signal (i.e., the mean of both eyes signals [(left eye + right eye)/2]) and the disconjugate signal (i.e., the difference between the two eyes [left eye − right eye]). Saccade onset was defined as the time when conjugate eyes velocity exceeded 30°/s and the offset when the eyes velocity signal dropped below 10°/s. The vertical postsaccadic drift consisted of the 160 ms following the end of the saccades. The process was performed automatically by the computer, and the verification was made by visual inspection of the individual eye position and velocity traces. Vertical and horizontal component amplitudes (conjugate and disconjugate) were defined with the following markers (Fig. 2): the conjugate vertical saccades (between “i” and “e”), and the conjugate horizontal components during the vertical saccade (between “i” and “e”); both vertical and horizontal disconjugacy during the vertical saccade (between “i” and “e”) and during the early postsaccadic fixation period (between “e” and “d”). 

Figure 2

 

Typical recording of vertical saccades obtained by averaging the position signal of the two eyes (LE + RE)/2 (a). Letters “i” and “e” indicate the onset and end of the saccades, respectively; “d” indicates the end of the drift of the eyes at 160 ms after the end of the vertical saccade. Horizontal movements during the vertical saccades (b). Disconjugacy signal of the two eyes (LE − RE) (c).

We measured the following parameters: the latency (i.e., the time between stimulus onset and initiation of saccades); the gain or accuracy of the saccades (ratio movement amplitude/target amplitude); the mean velocity (ratio amplitude/duration), and the maximum velocity of the saccades and their duration. We also examined the amplitudes of horizontal and vertical conjugate components during the vertical saccade and the disconjugate components (vertical and horizontal) during and after the vertical saccades. Eye movements in the wrong direction, with latency shorter than 80 ms (anticipation) or longer than 1000 ms or contaminated by blinks were rejected. Overall, 10% of trials were removed mostly because of blinks, and anticipation was scarce (<3%). 

Age effect was evaluated with a linear regression model applied to the different saccadic parameters, with the children's age (months) as the independent variable and the saccadic parameter for upward and downward saccades as dependant variable. The analysis was done in two distance conditions: at 40 and 150 cm. For up/down anisotropies, we used the Shapiro-Wilk test to test for normality of data distribution. As normality failed, a nonparametric Friedman ANOVA was performed with individual mean parameters with the viewing distance (40 cm or 150 cm) and the direction (upward versus downward) as main factors. Subsequently, the Wilcoxon test was used for pair-wise comparisons. A Bonferroni correction was used to correct for multiple comparisons and lowered the significance level form 0.05 to 0.0125. Finally, we used a Mann-Whitney U test to compare children's and adults' vertical gains and disconjugacy for each distance and direction. 

Results

Effect of Distance and Direction

Temporal Parameters.

Latency of Vertical Saccades.

The Friedman ANOVA revealed a significant effect of condition (χ²_(3,26) = 3.43, P < 0.032). Pair-wise comparison revealed a significant difference in latency between upward and downward saccades at far distances (Wilcoxon, Z = 2.01, P < 0.04). The vertical saccade latency at far distance appeared shorter for upward than for downward movements (upward: 298 ± 86; and downward: 339 ± 103). A significant difference was also found for downward saccades between near and far distances (Z = 2.87, P < 0.004). Group means are shown in Figure 3. 

Figure 3

 

The latency of the vertical saccades for both directions (down and up) at near (black) and at far (white) distances. Vertical bars represent standard error. Asterisks show the significant differences.

Mean Velocity of Vertical Saccades.

Conditions (up/down, near/far) showed no significant effects (χ²_(3,31) = 2.65, P = 0.44) on the mean velocity of vertical saccades. Thus, the speed of vertical saccades was stable regardless of direction and distance (upward: 122 ± 32; and downward: 131 ± 36 at near; and upward: 127 ± 31; and downward: 132 ± 31 at far). 

Peak Velocity of Vertical Saccades.

Results showed no significant effect, either in direction or in distance (χ²_(3,31) = 5.67, P = 0.13) on the peak velocities of the vertical saccades (upward: 240 ± 66; and downward: 257 ± 65 at near; and upward: 243 ± 54 and downward: 251 ± 56 at far). 

Duration of Vertical Saccades.

The Friedman ANOVA test showed no significant condition effect (χ²_(3,31) = 3.65, P = 0.30) on the duration of the vertical saccades (at near upward: 59 ± 11 and downward: 56 ± 12; at far and upward: 57 ± 12 and downward: 56 ± 10). 

Spatial Parameters: Conjugate Component.

Accuracy of Vertical Saccades.

The accuracy of saccades was measured as the gain (i.e., ratio of the amplitude of the vertical saccades made to that of the target eccentricity). The results showed no significant main effect of condition (χ²_(3,31) = 1.26, P = 0.74). As shown in Figure 4, the gain of the vertical saccades does not change significantly with direction or distance. In fact, both upward and downward saccades show hypometria in children, regardless of the viewing distance. 

Figure 4

 

The accuracy of the vertical saccades for both directions (down and up) at near (black) and at far (white) distances. Vertical bars represent standard error.

Vertical Conjugate Drift Following Vertical Saccades.

For the vertical conjugate postsaccadic drift, results showed a significant condition effect (χ²_(3,26) = 12.6, P < 0.006). The Wilcoxon test showed a significant difference in the vertical conjugate postsaccadic drift at near distance after upward versus downward saccades (Z = 3.96, P < 0.001). Figure 5 shows group means. The drift following downward saccades is relatively small (mean: 0.086 ± 0.31 at near and −0.07 ± 0.5 at far) and significantly smaller than drifts following upward saccades (mean: −0.35 ± 0.45 at near distance and −0.25 ± 0.48 at far distance). 

Figure 5

 

The vertical conjugacy after the vertical saccades for both directions (down and up) at near (black) and at far (white) distances. Vertical bars represent the standard error; *significant differences.

Disconjugate Components.

Vertical Disconjugacy During Vertical Saccades.

The Friedman ANOVA test showed a significant effect of condition for the vertical disconjugacy (χ²_(3,26) = 9.62, P < 0.022). However, the Wilcoxon test corrected with the Bonferroni method revealed no significant differences between the two distances (40 cm and 150 cm) and between the two directions (up and down). The mean values were really small (close to 0° for saccades at near and less than 0.5° for far distance). This is similar to what is known in adults.² 

Vertical Disconjugacy Following Vertical Saccades.

The Friedman ANOVA showed no effect for the vertical disconjugate postsaccadic drift (χ²_(3,26) = 1.95, P = 0.58). The mean postsaccadic vertical disconjugacy were near zero (upward: 0.06 ± 0.78; and downward: −0.08 ± 0.6 at near; upward: 0.00 ± 0.49; and downward: −0.11 ± 0.49 at far). 

Horizontal Vergence During Vertical Saccades.

For the horizontal vergence, the Friedman ANOVA showed a significant effect of conditions (χ²_(3,26) = 22.47, P < 0.001). Pair-wise comparisons showed a significant difference between upward and downward saccades at near distance (Z = 3.02, P < 0.002) and at far (Z = 3.41, P < 0.001). Figure 6 shows that the horizontal vergence is always convergent (positive value) and that the convergence is higher for downward saccades (mean: 1.06 ± 0.83 at near and 1.01 ± 0.75 at far distance) than for upward saccades (mean: 0.40 ± 0.67 at near and 0.35 ± 0.98 at far distance). As a consequence, and similar to adults' behavior,^1–3 children's downward saccades are associated with higher convergence than upward saccades. 

Figure 6

 

Horizontal vergences are shown during the vertical saccades for both directions (down and up) at near (black) and at far (white) distances. Vertical bars represent the standard error; *significant differences.

Horizontal Vergence Following Vertical Saccades.

The Friedman ANOVA test showed a significant conditional effect on the horizontal vergence following the saccade (χ²_(3,26) = 49.65, P < 0.001). The pair-wise comparison showed a significant difference at near distance between upward and downward saccades (Z = 4.7, P < 0.001). Figure 7 shows a slightly larger horizontal vergence drift after downward saccades (mean: −0.86 ± 0.43 at near and −0.74 ± 0.49 at far) than after upward saccades (mean: −0.14 ± 0.60 at near and −0.41 ± 0.33 at far); such drift level counter the horizontal vergence occurring during the vertical saccade. 

Figure 7

 

Horizontal vergences are shown after the vertical saccades for both directions (down and up) at near (black) and at far (white) distance. Vertical bars represent the standard error; *significant differences.

A summary of the results and their significance are shown in the Table. Direction-significant effects were found only for vertical conjugate postsaccadic drift and for horizontal vergence during and after the vertical saccades. 

Age Effect

Correlation Between Children's Age and Saccade Parameter.

The linear regression model showed no significant effect of age for all the following saccade parameters: gain, latency, duration, mean velocity, peak velocity, and vertical disconjugacy of the saccades (vertical binocular coordination). This was the case for both saccade directions (up and down) and for both distances (near and far). All R² values were weak and varied from 0.000 to 0.11, and all P values were superior to 0.05, yet for upward saccades at 150 cm, there was a significant correlation between the horizontal vergence taking place during the vertical saccades and the age (R² = 0.29, P < 0.002). Figure 8 shows that the horizontal convergence at 150 cm for upward saccades decreased with age. 

Figure 8

 

Linear regression plot of the horizontal vergence during the upward saccade at 150 cm as a function of age (in months) of the subjects.

Comparison Between Children and Adults.

Adults were not included in this study, but a comparison was done here with the data acquired in studies from Yang and Kapoula,^2,10 using a setup similar to that used here. Data for children and adults are summarized in the Table. 

Accuracy of Vertical Saccades: Adults Versus Children.

Saccade accuracy is expressed as gain (i.e., saccade amplitude over target eccentricity amplitude). The Mann-Whitney U test showed significant differences between the values for children and those for adult for all directions and distances (U test results = 44, 46, 41, and 86; P < 0.01 for near distances up and down and far distances up and down, respectively). As shown in the Table, the children presented smaller gain values than adults; children primarily undershoot their targets, whereas adults overshoot targets. 

Vertical Disconjugacy of Vertical Saccades: Adults Versus Children.

The Mann-Whitney U test showed that vertical disconjugacy was significantly different between children and adults but only for upward saccades at far distance (U test results = 94; P < 0.029). The vertical disconjugacy of vertical saccades appeared larger in children than in adults (Table, means). 

Qualitative Comparison: Adults Versus Children.

As we had no access to the whole data set but only to the group means, we compared the latency and the horizontal vergence values of children with these from young adult's values qualitatively only (Table). Children presented longer latencies than adults under all conditions. For children, latencies were shorter for upward than for downward saccades, whereas for adults¹⁰ this seemed to be the case only for target eccentricities of 15°. Children did not present the same behavior as adults for the horizontal vergence during vertical saccades; in particular, children always showed positive values (convergence), whereas adults always showed negative values (divergence) for upward saccades, yet the convergence from children is stronger for downward saccades than upward saccades. Thus, inspection of children's versus adult's behavior suggests a progressive development toward adult behavior. 

To summarize, most saccade parameters remained stable over the 6- to 10-year age span. The only aspect that changed was the decrease of convergence during upward saccades toward divergence (i.e., a pattern similar to that shown by adults). Furthermore, some aspects of the properties of vertical saccades appeared less mature in children than in adults—namely in terms of accuracy, of latency, of binocular coordination along the vertical axis and the pattern of horizontal vergence occurring during vertical saccades. 

Discussion

Latency

In line with published reports,^1,3,4 we report here that children display a latency asymmetry similar to what has previously been observed in adults (i.e., a shorter latency for upward than downward saccades that appeared significant at far distance). In adults, such asymmetric latency was not reported in all studies,^9,10 attributable to the different distances used in these studies or in the mode of initiation of saccades. The hypothetical perceptual superiority of the inferior visual field^7,8 could explain why eye movements are more quickly triggered in the superior visual field. Because the target presented in the superior visual field cannot be accurately processed in the periphery, it would be necessary to trigger the saccade as fast as possible. 

In the present study, the latency did not change significantly between children from 6 to 10 years old. Note however that the latency values reported here are still higher than the mean latency observed in adults (Table),¹⁴ which may be indicative of the ongoing maturation of the cortical oculomotor areas (e.g., parietal and frontal).^15,16 

Dynamics

The dynamics of vertical saccades (duration and mean and peak velocity), which were similar for up and down directions and for the execution distance (near or far) fit with those of studies in children by Salman et al.¹² and in adults by Collewijn et al.¹ The present results extend the results of Salman et al.¹² in a younger population, showing an absence of up/down differences for children from 6 years old. Thus, the dynamics of vertical eye movements appear to already be mature at this age. Furthermore, the saccade network, responsible for making fast saccades, depends essentially on the saccade generator at the brainstem level and on the cerebellum.¹⁴ These structures are already developed and myelinated early on.¹⁷ 

Accuracy

Our results showed that all saccades are hypometric in children (Fig. 4), as was found in the study by Salman et al.¹² In adults, upward saccades are hypometric, whereas downward saccades are hypermetric. These anisotropies could be explained by the hypothetical perceptive bias of a tilted vertical visual axis¹ (Table): If stimuli in the upper visual hemifield are misperceived as being farther away than they really are, then the saccade command would be programmed to a smaller size. In other words, this hypothetical tilt may be an adaptation to the environment and could be learned during childhood. 

Vertical Conjugate Postsaccadic Drift

We have shown here the presence of a downward drift after upward saccades, similar to that shown in adults.² After downward saccades, although our children also presented a downward drift at far fixation, as do adults, we noted here an upward drift at near distance (Fig. 5). However, the values of drift after downward saccades were near zero, similar to that found in adults.² If conjugate postsaccadic drift results from hypothetical pulse-slide-step mismatch,¹⁴ then our data suggest that quasi-optimal control of such command is obtained more readily for downward than for upward saccades. 

Vertical Binocular Coordination During and After Vertical Saccades

It is important that saccades are the same for the two eyes in order to minimize the vertical disparity, to avoid double vision and obtain single fused vision.¹⁸ In adults, studies of vertical saccades found that the vertical disconjugacy is well controlled during and after the movement, even for larger saccades, being less than 1°.^1,2 We also found here vertical disconjugacy smaller than 1° (Table). 

Horizontal Vergence During and After the Saccade

The results here show that the horizontal vergence is always convergent in children but tends to become divergent for upward saccades as age increases. Such progressive decrease of convergence tends toward the adult behavior.^1,2 If one assumes that there is little developmental change in the muscles and brainstem oculomotor neurons, then the different behavior in children between 6 and 10 years of age is more in line with the hypothesis of Collewijn et al.¹ of a perceptive bias in the vertical visual axis, slowly implemented by learning. In such context, our data indicate that this bias progressively maturates with age in children and gives rise to the motor correlate, that is, the adjustment of the horizontal intrasaccadic vergence. We propose that control of horizontal vergence is a continuous integrative process that optimizes oculomotor behavior in direction and in depth. 

In terms of values, we can see that children present higher amplitudes of horizontal vergence for downward than for upward saccades, in line with the study by Yang and Kapoula.² It is also proof that children tend toward the adult behavior of horizontal vergence during vertical saccades. 

As for horizontal vergence after the vertical saccades in children, the results here show more divergence after downward than after upward saccades, which is counter to the vergence observed during vertical saccades. This observation favors the theory by Collewijn et al.¹ 

Acknowledgments

The authors thank Sylvette Wiener-Vacher and Layla Ajrezo for the clinical examination of the children and Jean-Rémi King for the English proofing of the manuscript. 

Supported by a grant from the Fondation de France. 

Disclosure: C. Gaertner, None; Z. Kapoula, None