Two male Macaca mulatta monkeys (E and D) participated in this study. We implanted each monkey with fixtures to affix their head to the primate chair and a scleral search coil (20) to measure eye position in space. After the monkeys had recovered from the surgery, we trained them to track a small visual target in a dimly lit, sound-attenuating booth. The target was a 0.3° laser spot projected onto a cylindrical projection screen via two computer-controlled orthogonal mirror galvanometers. The primate chair was embedded in a servo-controlled multiaxis rotator (Actek, Seattle, WA). A custom program running in Spike2 (Cambridge Electronic Design, Cambridge, United Kingdom) controlled target movement, chair rotation, and the monkey’s reward via a CED Power1401 computer interface.

The monkey sat in a primate chair with its head restrained. We measured eye position with the electromagnetic search coil method (21). We rewarded the monkeys with applesauce for keeping their gaze within ±2° windows around the horizontal and vertical positions of the target spot for at least 0.5 s. Once they had learned to fixate the target spot, we trained them to make visually-guided saccades to a stepping spot that moved to random locations on the tangent screen within a ± 18° radius off straight-ahead. We delivered an applesauce reward (~0.16 ml per dollop, ~200 ml/h) by a pump (Masterflex tubing pump, Cole-Parmer, Vernon Hills, United States) every 2 s regardless of the amplitude, direction, or timing of the saccade, as long as it landed within the ±2° window surrounding the target. The targeting saccade was required to occur within 0.6 s of the target step and the subsequent fixation had to be maintained for at least 0.3 s for a reward to be given.

Under anesthesia, we implanted a vestibular prosthesis in the right ear of a rhesus monkey. This may have reduced vestibular function in the implanted ear, although this was not apparent in subsequent testing (17). Briefly, responses to the en-bloc velocity step testing, sinusoidal rotation testing (across a range of 0.05–2.0 Hz, ±10°), and constant peak velocity (across a range of 0.01–1.0 Hz, 80°/s) were normal. After full recovery from the implantation surgery, a series of transtympanic gentamicin injections were performed in monkey E, (18). Briefly, two holes were made in each tympanic membrane, and gentamicin (40 mg/ml, 0.2 ml/injection) was injected through one hole until the solution flowed from the other hole. This procedure was repeated 4 times (18). We suspect monkey D had an innate loss of the left horizontal semicircular canal because we could not find the left semicircular canal during the prosthesis surgery. Since the left horizontal semicircular canal had been dysfunctional enough by this loss and/or surgical damage, we did not inject gentamicin. Table 1 provides dates of these procedures and experiments.

We measured the VOR gain in complete darkness. The chair was moved sinusoidally at 0.5 Hz frequency and ± 10° amplitude. Once we confirmed the VOR gain had been reduced (see Result, Figure 1), we started the head-impulse test training (3, 22). The animals were required to fixate on a stationary target during chair rotation. We started with a slow horizontal rotation of the chair (25°/s). To make sure the animal stayed comfortable and safe, we did not exceed 150°/s head velocity and 2000°/s/s head acceleration and deceleration during the chair motion.

We conducted 43 experiments in 2 monkeys, 18 from monkey E and 25 from monkey D. We performed 5 tests, the VOR gain test, the head-impulse test, the target motion test, the visually guided saccade test, and spontaneous nystagmus test (Figure 1A).

We digitized eye, target, and chair position signals at 1 kHz using the Power1401 data acquisition/controller hardware (Cambridge Electronic Design, Cambridge, United Kingdom). Data were saved to a hard drive for later analysis.

A custom program running in Spike2 analyzed the saved data. It detected the occurrence of a saccade when eye velocity exceeded 150°/s within 50–800 ms after a program issued a target jump or initiation of chair movement and marked saccade onset and end when vector eye velocity exceeded or fell below 100°/s, respectively. The program measured saccade amplitude, peak velocity and duration, and the target distance before each saccade. The saccade attributes, target, and chair positions were exported to MATLAB (MathWorks, Natick, United States). Saccades whose initial eye positions differed from initial target positions by >5° were not analyzed.

To calculate the VOR gain, we first measured the slow phase velocity in the spontaneous nystagmus test session, estimating the background drift in the dark (monkey E: 4.6°/s leftward, 3.9°/s upward. Monkey D: 6.1°/s leftward, 5.4°/s upward; Figures 1B,C “Velocity” panel, blue line). Next, we measured the peak eye velocity during a slow phase in the VOR gain test session (Figure 1C “Velocity” panel, blue arrows). We also measured the corresponding peak head velocity (Figure 1C “Velocity” panel, black arrows). Finally, we calculated a gain value for leftward head rotation (=rightward eye movement) by dividing the peak rightward eye velocity (Figure 1C “Velocity” panel, blue arrow at “Ė_Rt”) by the peak head velocity to the left (Figure 1C “Velocity” panel, black arrow at “_Lt”). Similarly, a gain value for rightward head rotation (=leftward eye movement) was calculated by dividing the peak leftward eye velocity (Figure 1C “Velocity” panel, blue arrow at “Ė_Lt”) by the peak head velocity to the right (Figure 1C “Velocity” panel, black arrow at “_Rt”; (17)). The VOR gain was established by averaging 14 such measurements.

To demonstrate the eye movements during the head impulse test, we showed eye movement traces and saccade latencies (Figure 2). To evaluate the lack of effect of target movement speed on saccade latency in the data of the target motion test, we used one-way ANOVA (Figure 3).

To compare the characteristics of each saccade type, that is, putative non-visual and visual saccades during head impulse test, visually guided saccade, and quick phase nystagmus (Figure 4) and putative non-visual saccades during 100°/s and 150°/s head impulse tests (Figure 5), we plotted the main sequence, namely saccade peak velocity against saccade amplitude, and fitted them with a linear regression line. We then compared the slope of each saccade type across all 43 experiments (Wilcoxon signed rank tests). We considered a variable to be significant only when the value of p was less than 0.05.

Both monkeys had spontaneous nystagmus. Figure 1B shows a representative eye movement during the nystagmus test session in monkey D. The slow phase nystagmus velocity was 4.6°/s leftward, and 3.9°/s upward in monkey E and 6.1°/s leftward and 5.4°/s upward in monkey D (Figure 1B). Figure 1C shows a representative eye movement during the VOR gain test session in Monkey E. The VOR gain was 0.23 ± 0.039 to leftward head rotation (=rightward eye movement) and 0.45 ± 0.058 to rightward head rotation (=leftward eye movement). The other monkey’s VOR gain was 0.21 ± 0.038 to leftward head rotation and 0.76 ± 0.074 to rightward head rotation. Thus, we confirmed that both monkeys had vestibular impairment.

The eye movement response to a leftward head-impulse test (=rightward saccade) is shown in Figure 2A. In this representative example, the chair was rotated 15° to the left, while the target remained stationary, so the monkey needed to make an eye movement to the right in order to keep fixated on the stationary target. Residual VOR moved the eye initially, however, because the VOR gain was low, the distance between the eye and the target increased as the chair movement progressed. To catch the target, saccades were executed with variable latencies. Figure 2B shows the velocity traces of these trials. The latency distribution of all 26 trials in this session are in Figure 2C.

Since the duration of the chair movement was 300 ms in this experiment, saccades with less than 300 ms latency were considered covert saccades by definition (3, 4, 22). However, because 300 ms was longer than the duration of head movements in previous human studies (2, 3, 6, 9, 22, 23), we were concerned that these saccades with less than 300 ms latency might contain not only the so-called covert saccades but also overt saccades. Therefore, we differentiated the saccades in a different way: whether the saccade could be induced visually or not.

To induce saccades visually without vestibular stimulus, we moved the visual target with dynamics similar to those of the chair movement. The chair remained stationary in this condition. Figure 2D shows representative target movement and respective catch-up saccades. The relative target movement was identical to that during the head-impulse test in Figure 2A (grey line). The monkey made a catch-up saccade in order to fixate on the target; however, its latency was longer than those saccades with chair rotation. The latency distribution of all 31 trials in this session is shown in Figure 2F.

The shortest latency of the catch-up saccade in the target movement condition was 250 ms (dashed lines in Figures 2C,F). This suggested that the saccade whose latency was more than 250 ms during head impulse test could be induced visually, while the saccades with less than 250 ms latency during the head impulse test might be induced by a non-visual signal. Thus, we split the saccades during the head impulse test into 2 groups, the saccades with less than 250 ms latency as putative non-visually induced saccades and the saccades with more than 250 ms latency as putative visually induced saccades.

Figures 2G–I show the eye movement response to rightward head-impulse test (=leftward saccade). Similar to the leftward head-impulse test, the representative saccades were executed with variable latencies (Figures 2G,H). The latency distribution of all 34 trials in this session is in Figure 2I. Figures 2J–L show the target motion trials. Again, the latency was longer than those of saccades with chair rotation and the shortest latency of the catch-up saccade in the target motion trials was 250 ms (Figure 2L, n = 42).

Because the residual VOR moved the eye initially in the head impulse test, the visual input, i.e., retinal slip, was slower in the head impulse test than the target motion condition. Therefore, we compared the effect of retinal slip speed on catch-up saccade latency (Figure 3). We moved the target at 50, 60, and 70°/sec right- and leftwards (positive and negative values, respectively). The target speed did not have a significant effect on the saccade latency (One-way ANOVA, p = 0.59). Thus, the significant latency difference during the head impulse test and the target motion condition could not be explained by the difference in retinal slip velocity.

To verify that the characteristics of the saccades in monkeys were similar to that of previous human studies, we examined the correlation of saccade peak velocity as a function of saccade amplitude, namely the saccade main sequence. Figure 4A shows a representative experiment (monkey D) to compare the main sequence plots between these putative non-visual (HIT-NV) and putative visual (HIT-PV) rightward saccades (=leftward HIT). The correlation between saccades peak velocity and saccade amplitude for the non-visual saccades (regression line slope = 4.3, slope p = 4.03 × 10⁻⁵, r² = 0.05) was significantly weaker (p = 0.012) than for visual saccades (regression line slope = 29.0, slope p = 0.034, r² = 0.43; Figure 4A). Thus, the main sequences of non-visual and visual saccades were different, suggesting that the generation of these 2 types of saccades might be different.

Because the putative visual saccades might be induced visually, we compared its main sequence with that of visually guided saccades (collected during the visually guided saccade test, Figure 1A; VGS, Figure 4A) which are induced by a stepping target. The rightward visually guided saccade exhibited a stronger correlation than the visual saccades (regression line slope = 31.8, slope p = 1.00×10⁻¹⁵, r² = 0.87). We also plotted the main sequence of quick-phase saccades during nystagmus (collected by the spontaneous nystagmus test, Figure 1A). The quick phase saccades exhibited an intermediate slope between the visual and non-visual saccades (regression line slope = 10.8, slope p = 0.013, r² = 0.29).

Figures 4B–D summarizes the properties of the regression line among the 4 types of rightward saccades, that is, putative non-visual (HIT-NV) and putative visual saccade (HIT-PV) during head impulse test, visually guided saccade (VGS), and quick phase nystagmus, across all experiments from monkey E. We compared the 4 types of saccades to each other, so there were 6 comparisons (Wilcoxon signed rank tests with Bonferroni Correction for 6 times repetition, significant level: p < 0.0083). Figure 4B compares the slopes. HIT-NV exhibited a significantly lower slope than HIT-PV (p = 0.00064). The remaining 5 comparisons were not significantly different.

Figure 4C compares the 95% confidence interval of the slopes. HIT-NV exhibited a significantly smaller interval than HIT-PV (p = 0.0072). Figure 4D compares the r-square of the slopes. HIT-NV exhibited a significantly smaller correlation than HIT-PV (p = 0.00044).

Figures 4F–H summarizes 3 types of leftward saccades across all experiments from monkey E (no quick-phase nystagmus saccades to the left). Similar to the rightward saccades, HIT-NV exhibited a significantly lower slope and 95% confidence interval than HIT-PV (slope: p = 0.017, Figure 4F confidence interval: p = 0.00044, Figure 4G).

Figures 4I,J summarize all experiments from monkey D. This monkey showed significant differences in more comparisons. The consistent result from 2 monkeys and 2 directions were; 1, the slope of HIT-NV was lower than that of HIT-PV. 2, the slope of HIT-PV was not different from that of VGS. 3, the 95% confidence interval of HIT-NV was lower than that of HIT-PV.

Since the deficits of VOR were different between the monkeys and directions, we examined these factors on the slope of the main sequence. The median and IQR of the slope of rightward and leftward HIT-NV were 13.4 ± 7.7 and 11.6 ± 4.0, respectively, in monkey E (Figures 4B,F) (p = 0.085, Wilcoxon signed rank tests), and − 1.2 ± 9.1 and 4.4 ± 3.5, respectively, in monkey D (Figures 4I,J; p = 4.03 × 10⁻⁵, Wilcoxon signed rank tests). The p values of 2-way ANOVA for the slope of HIT-NV were 1.46×10⁻¹² (the factor monkeys) and 0.24 (the factor directions). Thus, there was a significant difference between the monkeys but no significant difference between the directions in HIT-NV saccades.

The median and IQR of the slope of rightward and leftward HIT-PV were 35.5 ± 25.4 and 18.8 ± 33.1, respectively, in monkey E (Figures 4B,F; p = 0.16, Wilcoxon signed rank tests), and 14.3 ± 28.4 and 33.1 ± 9.6, respectively, in monkey D (Figures 4I,J; p = 0.044, Wilcoxon signed rank tests). The p values of a 2-way ANOVA for the slope of HIT-PV were 0.19 (the factor monkeys) and 0.51 (the factor directions). Thus, in HIT-PV saccades, there were no significant differences between monkeys and directions. Finally, the median and IQR of the slope of rightward and leftward VGS were 39.1 ± 8.8 and 37.4 ± 4.7, respectively, in monkey E (Figures 4B,F; p = 0.031, Wilcoxon signed rank tests), and 30.4 ± 7.3 and 32.7 ± 5.5, respectively, in monkey D (Figures 4I,J; p = 0.25, Wilcoxon signed rank tests). The p values of a 2-way ANOVA for the slope of VGS were 2.14×10⁻⁷ (the factor monkeys) and 0.47 (the factor directions). Thus, for VGS, there was a significant difference between the monkeys but no significant difference between the directions.

Figures 5A,B show a representative experiment to compare HIT-NV that occurred during the 100°/s and 150°/s chair rotation. For both rightward (Figure 5A) and leftward saccades (Figure 5B), the slope of the regression line fitted to the main sequence for the 150°/s was lower than that in the 100°/s. Across 10 experiments from monkey E, the slope was significantly lower in the 150°/s than 100°/s [p = 0.03 for rightward (Figure 5C), 0.02 for leftward (Figure 5D)]. Thus, the saccade HIT-NV that occurred during the 150°/s had more extreme properties as regards to their main sequence relationships.