Simultaneous recordings of ocular microtremor and microsaccades with a piezoelectric sensor and a video-oculography system

Subjects

Eight subjects (5 males, 3 females) with normal or corrected-to-normal vision participated in the experiments. Experiments were carried out under the guidelines of the Barrow Neurological Institute’s Institutional Review Board (protocol number 04BN039). Written informed consent was obtained from each subject.

Eye movement recordings

Experimental setup and procedure

We mounted the piezoelectric sensor on the EyeLink II helmet to record eye position simultaneously with the two systems (Fig. 1A). Each subject laid supine on a hospital bed, looking up at a horizontally down-facing LCD monitor ∼ 43 cm from the subject. A licensed physician applied a drop of topical anesthetic (proparacaine) to the eye(s) that was/were to have a piezoelectric bimorph. The eyelid(s) of the eye(s) that was/were to have a piezoelectric bimorph was/were then retracted using polyethylene surgical tape. We then let the silicone tipped piezoelectric bimorph rest on the sclera (Fig. 1B) for a maximum of two sequential 40-s trials. Subjects performed one of two experimental tasks (Troxler fading or simple fixation; see Experiments section for details).

Experiments

Eye movement analyses

Microsaccade and OMT correlations with transitions to visible and invisible percepts

We correlated microsaccade production to the subjects’ perceptual reports, as in McCamy et al. (2012). Briefly, let X_M and X_R be the stochastic processes representing the onsets of microsaccades and intensification reports. For example, if s₁, s₂, …, s_k are the start times of all the microsaccades for a given subject, then X_M for that subject will be given by X_M(t) = 1 if t = s_i for some 1 ≤ i ≤ k, and X_M(t) = 0 otherwise; similarly for X_R. We obtained correlations of microsaccades with reports of intensification for each subject, using ${\xi }_{MR}\left(t\right)={\sum }_{n=-\infty }^{n=\infty }{X}_M(n+t)X_R\left(n\right)$ and then converting it to a rate (similarly for correlations of microsaccades with reports of fading). For each subject, correlations were smoothed using a Savitzky–Golay filter of order 1 and a window size of 151 ms (X_M and X_R were not smoothed). Average correlations are the average of the smoothed correlations (Fig. 4B). OMT correlations with reports of perceptual transitions were obtained in a similar fashion.

Statistics

To analyze the effect of the piezoelectric sensor on microsaccade magnitude and rate, we conducted separate single-factor repeated measures ANOVAs (one for each dependent variable) with the three measuring times (before sensor, during sensor, and after sensor) as the within-subjects factor. We conducted post hoc comparisons using Tukey HSD tests. To study the effect of the surgical tape and anesthetic on microsaccade magnitude, we used separate two-tailed paired t-tests (one for each dependent variable). To analyze the effect of the probe on the microsaccadic peak velocity–magnitude relationship, we used a two-tailed paired t-test on the slopes found from the robust linear regressions, for each subject. The significance level was set at α = 0.05. We analyzed the effect of the sensor, tape, and anesthetic on microsaccades using data from the fixation experiment only.

Simultaneous recordings and the effects of the piezoelectric sensor on microsaccades

Whereas the piezoelectric sensor can measure both microsaccades and OMT, video systems such as EyeLink II (SR Research) can measure microsaccades accurately, but do not have the resolution necessary to measure OMT (McCamy, Macknik & Martinez-Conde, in press). We extracted the microsaccadic component of the piezoelectric sensor data using a wavelet denoising technique (Al-Kalbani et al., 2007) and we detected the microsaccades from the EyeLink II data using a modified version of Engbert and Kliegl’s algorithm (Engbert & Kliegl, 2003) (Fig. 1C; see Materials and Methods for details). Because the piezoelectric sensor comes in contact with the eye, we wondered whether its presence might affect microsaccade dynamics, for instance by dampening eye movements during simultaneous piezoelectric and video recordings.

We determined the effects of the piezoelectric sensor on microsaccades using data from the fixation experiment; see Materials and methods for details. Subjects fixated a central spot and we recorded their eye position binocularly with EyeLink II while applying the piezoelectric sensor in one eye, both eyes, or neither eye.

Microsaccades occurred as binocular events in every condition, but when a single sensor was placed in one eye, microsaccade magnitudes in the eye with the sensor were significantly smaller than in the eye without the sensor (Figs. 2A–2B and 2C). Further, microsaccades in the eye without the sensor were significantly larger than microsaccades previous to sensor placement, and those in the eye with the sensor were significantly smaller than prior to sensor placement (Fig. 2C). Normal microsaccade magnitudes were restored upon sensor removal (Fig. 2C). Binocular application of the sensor did not alter microsaccade magnitude significantly, but we note that data from this condition were limited to few trials in only two subjects (not shown).

The eye with the sensor had a slightly lower (but not statistically significant, t(5) = 1.960, p = 0.107) peak velocity–magnitude slope (58 s⁻¹ ± 5 s.e.m.) than the eye without the sensor (62 s⁻¹ ± 4 s.e.m.) (Fig. 3), suggesting moderately decreased microsaccade velocities in the eye with the sensor.

The presence of the sensor did not affect microsaccade rates (with sensor: 1.47 microsaccades/s ±  0.36 s.e.m.; without sensor: 1.11 microsaccades/s ± 0.25 s.e.m.; t(5) = 2.075, p = 0.093).

In the Troxler fading experiment (see Materials and Methods for details), applying the sensor to one eye during monocular viewing (i.e. putting a patch on the eye without the sensor) also led to reduced microsaccade magnitudes, as compared to microsaccade magnitudes before sensor application (not reaching statistical significance, p = 0.081; n = 4 subjects) (Fig. 2D). Microsaccades after sensor removal were significantly larger than those occurring while the sensor was on. Thus, changes in microsaccade magnitude in the eye with the sensor were consistent with those observed in the fixation experiment.

Neither taping the eyelids open (with tape: 0.43° ± 0.07 s.e.m.; without tape: 0.46° ± 0.06 s.e.m.; t(2) = 1.764, p = 0.221) nor applying anesthetic drops (with anesthetic: 0.48° ± 0.06 s.e.m.; without anesthetic: 0.49° ± 0.06 s.e.m.; t(2) = 0.475, p = 0.682) affected microsaccade magnitudes. Thus, the changes in microsaccade magnitude described above were due to the application of the sensor.

Microsaccades but not OMT are correlated with perceptual restoration after Troxler fading

Here we set out to quantify the potential role of OMT in restoring faded vision during fixation. Because OMT is not necessarily conjugate (Martinez-Conde, Macknik & Hubel, 2004), we placed a single piezoelectric sensor in either the left or the right eye, and patched the eye without the sensor. Thus, this experiment consisted entirely of monocular recordings.

Four subjects fixated a central spot and continuously reported, via button press, whether an unchanging visual stimulus (a 2-lobe Gabor patch with 40% contrast), which was presented either foveally or peripherally (9°), was faded (or in the process of fading) versus intensified (or in the process of intensifying) (Martinez-Conde et al., 2006; McCamy et al., 2012) (Fig. 4A; see Materials and Methods for details). Microsaccade rates increased before perceptual transitions to intensifying targets and decreased before perceptual transitions to fading targets, in agreement with previous research (Martinez-Conde et al., 2006; McCamy et al., 2012) (Fig. 4B). Foveal and peripheral presentations of the Gabor patch resulted in equivalent modulations of microsaccadic rates before perceptual transitions, also consistent with previous results (McCamy et al., 2012) (data in Fig. 4B are collapsed across both eccentricities). The microsaccades detected with EyeLink II and the microsaccadic component from the piezoelectric recording produced comparable correlations with perceptual transitions (Fig. 4B and Inset from Fig. 4C). OMT frequency was not correlated with either type of perceptual transition (Fig. 4C).

Effects of the piezoelectric sensor on microsaccade dynamics

The oculomotor system adapts to adversities that would, unchecked, impair visual perception. For example, the oculomotor system adjusts its output to account for anatomical change due to growth, damage to the central nervous system or muscular control, and correction of visual refraction from glasses/contacts (Optican, Zee & Chu, 1985; van Donkelaar & Gauthier, 1996). Investigations of the oculomotor system’s adaptive ability have focused on saccadic adaptation, see Pélisson et al. (2010) for a review. In this paradigm, a subject makes saccades to successive cued locations. During the execution of each saccade, the cued location consistently changes (the subject is unaware of this change due to saccadic suppression) to induce a visual error signal at the termination of the saccade. The saccadic system senses this systematic error and slowly adjusts its output (i.e. it recalibrates itself) to compensate for and finally remove the error. The smooth pursuit system can adapt in similar ways as well (van Donkelaar & Gauthier, 1996). Here we encountered another situation where the oculomotor system adjusted its output (i.e. it changed its microsaccade magnitudes) to compensate for the adversity brought on by the piezoelectric sensor.

Lowering the sensor onto the sclera led to (at least) two changes to the state of the eye that could have affected microsaccade magnitude in our experiment: (1) translation and possible rotation of the eye, and (2) dampening of eye movements and deformation of the sclera (upon stabilization after the initial lowering of the sensor onto the eye).

The oculomotor system adjusts its output based on retinal (i.e. visual) and extraretinal (i.e. non-visual) signals (Collins & Wallman, 2012; Pélisson et al., 2010). In the absence of extraretinal signals, image displacements due to eye movements would be indistinguishable from those due to motion in the world; thus retinal signals, by themselves, are not sufficient for accurate visual perception. Extraretinal signals have two possible sources: (a) proprioceptive signals from the extraocular (EOM) muscles and (b) corollary discharge signals from the motor command center. The importance of each type of extraretinal signal is debated, but both are thought to contribute to normal perception (Balslev et al., 2012; Donaldson, 2000; Wang & Pan, 2012; Weir, 2006; Weir, Knox & Dutton, 2000); see Donaldson (2000) for a comprehensive review).

Placement of the piezoelectric sensor on the eye produces a similar scenario to that of the classic “eye press” experiments investigating the contributions of proprioceptive versus corollary discharge signals to perception (Gauthier, Nommay & Vercher, 1990; Rine & Skavenski, 1997; Stark & Bridgeman, 1983). In eye press experiments, one presses on the outer canthus (part of the sclera where the upper eyelid meets the lower one, towards the ear) of the eye while the subject indicates the direction of a target by pointing (Stark & Bridgeman, 1983). The procedure displaces the eye without a corresponding corollary discharge signal, and the experimenter measures the subsequent effects on spatial perception. Experimental conclusions are largely dependent on the assumption that the eye press does not affect the EOM proprioception signals. However, critics have argued that pressing on the outer canthus changes both corollary discharge and proprioceptive signals in a complex and, at the moment, undetermined manner (see Donaldson (2000) for a thorough discussion).Thus, it is unclear whether both translation and rotation occur due to the press and if so, to what extent. It is also unknown whether muscle spindles and palisade endings in the EOMs respond in the same way to passive (i.e. externally imposed) versus active (i.e. internally imposed) movements of the eye. Because of these unknowns, the eye press technique has fallen out of use and the conclusions from studies employing it are debatable (Donaldson, 2000; Rine & Skavenski, 1997). Thus, we can only speculate as to how the inferences from eye press studies relate to our results:

One plausible explanation is that the dampening force of the piezoelectric sensor may have caused microsaccades in the eye with the sensor to be smaller than the oculomotor system intended, thus causing a visual error signal (although subjects did not report any visual errors spontaneously, and received no queries about them) that led to the observed changes in microsaccade magnitude. The mechanical pressure may have also changed the proprioceptive signals of the eye with the sensor. Because classical saccadic adaptation is a gradual process that takes up to 100 trials to complete (i.e. several minutes) (Pélisson et al., 2010), and the “adaptation” we observed was very quick (i.e. the sensor was in the eye for only 80 s) it may be that the sensor affected proprioceptive signals mainly. Proprioceptive signal changes due to dampening – in addition to possible visual error signals in the eye with the sensor – may have resulted in more force and hence innervation in that eye to move as intended. To keep the saccades conjugate (i.e. to obey Hering’s law of equal innervation, which states that during saccades, both eyes receive equal innervation to corresponding muscles (Bahill et al., 1976)), microsaccades in the eye without the sensor may have increased in magnitude (i.e. due to increased innervation), without any damping. In agreement with this possibility, previous studies have found that oculomotor adaptation in one eye translates, at least partially, to the other eye (Albano & Arzola Marrero, 1995; Hopp & Fuchs, 2004; Optican & Robinson, 1980) (albeit oculomotor adaptation can also be disconjugate, see Hopp & Fuchs (2004) and Snow, Hore & Vilis (1985)).

Frens & Geest (2002) found that placement of scleral coil annuli on both eyes of human subjects led to reduced saccadic velocities (by about 5%) and increased saccadic durations (by about 8%). When they placed a coil on one eye only, the effect was still present in both eyes, although it was more variable; thus the authors hypothesized that the effects were non-mechanical, possibly due to an alteration in oculomotor signals. Whereas the OMT sensor has much greater mass than the scleral coil, a similar logic may apply; thus both mechanical and neural factors may have contributed to our observations.

Finally, placement of the sensor resulted in visible deformation of the sclera, which may have affected the shape of the cornea and caused optical blur. It is not clear whether scleral deformation could have affected microsaccade magnitude, but it may have introduced a larger visual error signal and thus contributed to our results.

A definite physiological explanation of the effect of the sensor on microsaccade magnitude remains elusive, due to all the unknown variables detailed above. Future research specifically designed to investigate the effects of changing proprioceptive and corollary discharge signals on microsaccades should provide further insight into how both types of signals combine to give the brain knowledge of gaze position during attempted fixation.

Video oculography systems such as the EyeLink II tracker used here are incapable of measuring OMT (McCamy, Macknik & Martinez-Conde, in press), and so we could not determine the potential effects of the sensor placement on the characteristics of OMT. However, a previous study limited to one subject (Al-Kalbani, 2010), using the same basic probe design as described here, found no significant difference between mean peak OMT frequency measured in both eyes, where the probe mass on one eye was increased by ∼30% (14 g to 18.5 g), compared to the probe used in the other eye. Neither was there any difference observed in the mean RMS OMT amplitude of the eyes with the heavier and lighter probes. Although no apparent relation was found between probe mass and mean RMS amplitude or frequency estimates, there was some greater variability seen in amplitude and frequency within individual records. Confirmation that the probe does not significantly suppress OMT activity would demand repeating the experiment with a “zero” mass probe (i.e. a non-contacting approach). Yet, physiological relevant frequency variations in mean peak OMT frequency clearly do transmit to piezoelectric probes, despite any potential damping, as established in many studies showing OMT frequency changes in different clinical conditions (Bojanic, Simpson & Bolger, 2001; Bolger et al., 2000; Coakley, 1983).

Bengi & Thomas (1968b) also investigated the effect of loading the eye on OMT spectra, and found a progressive reduction in the amplitude of velocity spectral density with increased moment of inertia (via weights on a contact lens) of between 5 and 15 g cm² .

Non-contact laser speckle interferometry offers the possibility of future OMT measurements without interfering with eye dynamics (Al-Kalbani et al., 2009).

Effects of microsaccades and OMT on perceptual restoration after Troxler fading

Microsaccade rates increased before perceptual transitions to intensification and decreased before perceptual transitions to fading, indicating that microsaccades counteract Troxler fading, in agreement with previous results (Martinez-Conde et al., 2006; McCamy et al., 2012) (Fig. 4B). However, OMT frequency was not correlated with perceptual restoration of faded targets in the present conditions (Fig. 4C).

It is important to note that the limited amount of data available from the piezoelectric recordings (Fig. 4C Inset) generated microsaccade correlations with perceptual transitions that were comparable to those obtained with the full video tracker recordings dataset (Fig. 4B); thus it seems unlikely that the lack of a correlation between OMT frequency and perception was due to insufficient data (because the same amount of data did render a correlation for microsaccades). Yet, our results do not completely rule out a contribution of OMT to combating visual fading. For instance, differences in OMT frequency or amplitude across subjects might cancel out in the average, thereby diminishing a potential correlation with perceptual transitions in our analyses.

In addition, OMT may improve or enhance other visual functions, such as signal detection (Funke, Kerscher & Wörgötter, 2007) or visual acuity (Hennig et al., 2002; Zozor, Amblard & Duchêne, 2009). Future research should investigate the potential effects of OMT in preventing and restoring faded targets of varied spatial frequencies, eccentricities and sizes, as well as its possible role in other perceptual phenomena.