Abnormalities of coherent motion processing in strabismic amblyopia: Visual-evoked potential measurements

Abstract

Coherent motion responses of patients with mild to moderate strabismic amblyopia were compared to those of normals using visual-evoked potentials (VEPs). Responses were elicited by dynamic random-dot kinematograms that alternated at 0.83 Hz between globally coherent (left–right) and incoherent (random) motion states. Tuning curves were measured at the first harmonic of the global motion update rate (0.83 Hz) and at the first harmonic of the dot update rate (20 Hz) for spatial displacements 3.1 to 27.9 arcmin (1.6 to 9.3 deg/s). Responses locked to the changes in the global organization of the local direction vectors were an inverted U-shaped function of displacement/speed in the normal-vision observers and in the fellow eyes of the strabismus patients while the tuning function of the amblyopic eyes was shifted to larger displacements/higher speeds. Responses at the dot update rate were reduced in amplitude and altered in timing in both eyes of the patients. The results are consistent with both local and global deficits in motion processing in strabismic amblyopia.

Introduction

Amblyopia is traditionally considered to be a developmental disorder of spatial vision associated with the presence of strabismus, anisometropia, or form deprivation during early life (Ciuffreda, Levi, & Selenow, 1991). Recent studies however suggest that the abnormal processes that underlie amblyopia may not only cause deficits in spatial vision (e.g., poor visual acuity and reduced contrast sensitivity) but also deficits in oscillatory motion processing (Buckingham, Watkins, Bansal, & Bamford, 1991), motion aftereffect (Hess, Demanins, & Bex, 1997), global motion processing (Ellemberg, Lewis, Maurer, Brar, & Brent, 2002; Ho & Giaschi, 2006; Ho et al., 2005; Simmers & Bex, 2004; Simmers, Ledgeway, Hess, & McGraw, 2003; Simmers, Ledgeway, Mansouri, Hutchinson, & Hess, 2006), and contour integration (Chandna, Pennefather, Kovács, & Norcia, 2001; Hess & Demanins, 1998; Hess, McIlhagga, & Field, 1997; Kovács, Polat, Pennefather, Chandna, & Norcia, 2000; Kozma & Kiorpes, 2003). 

Global motion and contour integration deficits in amblyopia have been attributed primarily to “higher level” processes that operate after any initial losses in V1 (for recent reviews, see Kiorpes, 2006; Levi, 2006). Global stimuli, by design, require two levels of processing—a local level involving correspondence matches between local image elements and a global level involving the detection of larger-scale consistencies in local feature vectors such as direction of motion or orientation. In the case of global motion stimuli, a percept of motion is created by enforcing a consistent, element-by-element displacement rule across image updates. A large number of image elements such as dots or Gabor patches are used so that different global configurations (flow fields) can be portrayed. These stimuli thus consist of many local apparent motion sequences whose overall organization depends on both the patterning rule and the “coherence” or consistency of the local motions over large parts of the visual field. Importantly, the strength of the global motion percept can be controlled by varying the fraction of the elements obeying the global patterning rule either by replacing a fraction of the local motion pairs with random pairings (Newsome & Paré, 1988) or by adding variability to the local-motion trajectories (Williams & Sekuler, 1984). 

It has been widely suggested that sensitivity to coherent motion in dynamic random-dot patterns depends on the function of areas such as MT/V5 and MST that lie downstream of the primary visual cortex. Lesions of macaque MT result in elevated perceptual thresholds for coherent motion (Newsome & Paré, 1988). Neurons in these areas respond robustly to random-dot kinematograms portraying globally organized motion (Britten, Shadlen, Newsome, & Movshon, 1992; Celebrini & Newsome, 1994) and microstimulation of MT and MST can alter behavioral responses to motion (Celebrini & Newsome, 1995; Salzman, Britten, & Newsome, 1990). The human homologue of MT/MST is also robustly activated by global motion stimuli as measured both by fMRI (Braddick et al., 2001; Morrone et al., 2000) and MEG (Aspell, Tanskanen, & Hurlbert, 2005; Händel, Lutzenberger, Their, & Haarmeier, 2007; Lam et al., 2000; Nakamura et al., 2003; Prieto et al., 2007; Siegel, Donner, Oostenveld, Fries, & Engel, 2007). Differential responses to coherent motion in macaque V1 have not been observed (Snowden, Treue, & Andersen, 1992). FMRI activation in V1 is also generally not observed (but see Koyama et al., 2005) nor have MEG responses from the calcarine cortex been reported. Together these results have led to the view that motion signals are initially picked up in local regions of the image by cells in early visual areas (e.g., V1) that have small receptive fields, and these signals are subsequently pooled at a second stage, typically considered to be extra-striate area MT, where the true direction and speed is computed (Morrone, Burr, & Vaina, 1995; Simoncelli & Heeger, 1998). 

A number of studies have suggested that deficits in global motion processing in amblyopia are due primarily to abnormalities at the second or global stage of motion integration rather than at the first or local stage of motion detection. The strongest evidence on this point is the presence of abnormalities in eyes that have normal optotype acuity (Constantinescu, Schmidt, Watson, & Hess, 2005; Ho & Giaschi, 2006; Ho et al., 2005; Simmers et al., 2003, 2006). These studies have each found that global motion thresholds were elevated in the amblyopes' fellow eyes which posses high acuity and contrast sensitivity. A similar effect has also been seen in fellow eyes of amblyopic macaques (Kiorpes, Tang, & Movshon, 2006). Fellow eye abnormalities, if observed, indicate that amblyopia affects the development of motion mechanisms independent of effects that are due to a loss of acuity or contrast sensitivity at the local processing stage. 

In the present study, we have used dynamic random-dot kinematograms to record visual-evoked potentials (VEPs) in patients with strabismic amblyopia and in normal controls. Evoked responses have been recorded to these stimuli in previous studies of normal visual processing where the focus has been on motion onset responses that occur after stationary dot patterns begin to move (Bundo et al., 2000; Lam, Kaneoke, & Kakigi, 2003; Maruyama, Kaneoke, Watanabe, & Kakigi, 2002; Prieto et al., 2007; Ulbert, Karmos, Heit, & Halgren, 2001) or after transitions from random motion to coherent motion (Aspell et al., 2005; Händel et al., 2007; Lam et al., 2000; Nakamura et al., 2003; Niedeggen & Wist, 1999). The latter responses are more directly comparable with the psychophysical studies in that these studies are able to discount the effect of temporal transients that occur when static stimuli begin to move. 

We presented coherent and random motion in a periodic sequence that leads to motion onset and offset responses that are associated with the differences in the direction distribution of the coherent and random phases of the stimuli. Unlike previous studies, we have also measured responses that are time-locked to the dot update rate. These responses have not been previously recorded in motion VEP studies and provide an additional source of information about the putative first stage of processing than that of the local image transients. 

Amblyopia is a heterogeneous disorder, and previous studies have reported different patterns of loss in amblyopia that is associated with anisometropia or strabismus (Hou, Good, & Norcia, 2007; Levi & Klein, 1982; McKee, Levi, & Movshon, 2003). We chose to study patients with strabismic amblyopia because these patients have been reported to have abnormalities in both the amblyopic and the fellow eyes. We also recruited patients with mild to moderate amblyopia (optotype acuity from 20/40 to 20/80) in order to minimize the effects of unsteady or eccentric fixation that can occur in deep amblyopia. We compared evoked responses of amblyopic and fellow eyes of patients with strabismus to those of normal-vision eyes over a range of stimulus speeds and obtained responses that were time-locked to both the global and local motion updates of the display. 

Materials and methods

Observers

Fifteen normal-vision observers between 17 and 66 years (mean ± SD, 37.6 ± 13.7; 7 females) and ten patients with strabismic amblyopia between 24 and 60 years (mean age ± SD, 49 ± 12.9; 6 females) participated in the first experiment that compared normal and amblyopic-evoked responses. An additional 10 normal-vision observers between 22 and 49 years (mean ± SD, 29.6 ± 10.2; 6 females) participated in a control experiment that varied stimulus contrast. All participants were refracted under non-cycloplegic conditions by a pediatric ophthalmologist before the experiments. Visual acuity was evaluated with a constant crowding LogMAR chart (Bailey–Lovie) and was measured with best optical correction. Stereo acuity was measured with RANDOT stereotests (Stereo Optical CO., Inc.). All normal-vision observers had 20/20 or better optotype acuity in each eye and stereo acuity of at least 30 arcsec. They also had no prior history of strabismus, amblyopia, or any other eye disease. In the observers with strabismic amblyopia, horizontal and vertical angles of deviation were quantified with a prism-cover test at 0.3 and 6 m with and without optical correction. The presence of eccentric fixation in the amblyopic eye was determined with a Visuoscope while the fellow eye was occluded. Monocular fixation of more than 0.5 deg from the center of the fovea was classified as eccentric fixation. Stability of fixation was also judged during visuoscopy. Inclusion criteria for observers with amblyopia included (1) visual acuity between 20/40 and 20/80 in one eye, with the other eye being 20/20 or better; (2) constant ocular deviation or a history of prior strabismus surgery with or without anisometropia; (3) no history of visual deprivation (e.g., cataract, ptosis, etc.); (4) no other eye disease (e.g., cataract, glaucoma, lens implant, etc.); (5) no eccentric fixation and unsteady fixation; (6) no latent or manifest nystagmus or oculomotor dysfunction with the exception of strabismus. We recruited patients with a limited severity of amblyopia (optotype acuity from 20/40 to 20/80) to minimize the effects of unsteady or eccentric fixation that is more prevalent in deep amblyopia. The mean visual acuity in the amblyopic eyes was 20/55. Refractive errors were corrected for the testing distance (70 cm) in all observers during the experiments. The patient details are shown in Table 1. 

Table 1

 

Clinical characteristics of the patients with amblyopia.

Subject no.	Visual acuity		Refractive errors		Ocular alignment (Δ)	Strabismus onset	Surgery history	Stereopsis
	Fellow eye	Amblyopic eye	Fellow eye	Amblyopic eye
1	20/20	20/40	+1.75	+2.00	ET 14, R/L 2	1 year	1	>400″
2	20/16	20/80	+3.5+1.5 × 90	+2.75+3.25 × 120	XT 20. L/R 4	At birth	1	>400″
3	20/20	20/80	Plano	−7.50–1.00 × 50	XT 12, R/L 4 (V pattern)	3–4 years	No	>400″
4	20/20	20/40	−10.50–1.00 × 120	−7.50–2.00 × 40	ET 50, L/R 6	At birth	No	>400″
5	20/20	20/40	−1.75–0.75 × 65	+1.00	ET 16, R/L 12	Not clear	2	>400″
6	20/16	20/63	+2.00–0.50 × 94	+4.00–0.50 × 107	XT 2, L/R 17	Under 4 years	1	>400″
7	20/20	20/63	+1.50+0.75 × 115	+3.00+2.00 × 88	XT 8 (V pattern)	2 years	1	>400″
8	20/16	20/40	Plano	+2.75+3.00 × 105	ET 10	5 years	No	400″
9	20/16	20/63	−3.00	−8.00–0.50 × 90	Exophoria-XT 10	6 months	1	>400″
10	20/20	20/40	−4.75	−3.00–1.75 × 30	ET 20, R/L 4	Not clear	No	>400″

The research protocol was approved by the Institutional Review Board of the California Pacific Medical Center and conformed to the tenets of the Declaration of Helsinki. Written informed consent was obtained from the observers after the research procedure was explained. 

Stimuli and apparatus

Participants viewed random-dot kinematograms displayed on a color CRT monitor (Mitsubishi, Diamond Pro 2070 ^SB) running in monochrome mode (white dots on a black background). The display was set at a 640 × 480 pixel resolution and a 120 Hz refresh rate. The display area was 24 × 24 deg square at the 70 cm viewing distance. The random-dot kinematograms were composed of 12.4 arcmin white dots (105 cd/m ²) on a black background (5 cd/m ²). The dots had an average density of 2.5 dots per square degree and subtended slightly less than 10% of the total display area due to random overlaps between dots. A small fixation mark was presented in the center of the display. 

The display alternated between leftward/rightward coherent motion and incoherent motion with the direction of coherent motion alternating (e.g., leftward, random, rightward, random, etc.) in order to reduce the effects of motion adaptation. A full stimulus cycle thus lasted 2.4 s, but the data were collapsed across the two directions of motion to yield a 0.83 Hz (f1) cycle of the stimulus (0.6 s of lateral translation alternating with 0.6 s of random motion). All dots of both the random and coherent motion displays were updated at 20 Hz, and each dot was displaced by a fixed distance. In the coherent portions of the displays, all dots translated either leftward or rightward at 100% coherence. In the random motion portions, each dot was independently displaced in a random direction every 50 ms (Brownian motion). 

Each dot had a limited lifetime of 3 s (i.e., 60 updates at 20 Hz). During this lifetime, a dot's path of motion would alternate between random and lateral translation, as described above. At the end of its lifetime, the dot would be redrawn in a new, random location in the display area, before resuming its motion. To conceal visual artifacts caused by this re-positioning, the full set of dots were randomly assigned to 60 cohorts with staggered lifetime cycles, so that on any given update, less than 2% of the dots would be randomly re-positioned. 

VEP responses were measured over a range of dot displacements from 3.1 arcmin to 27.9 arcmin. The combination of spatial displacements and the 20 Hz dot update frequency resulted in a range of speeds from 1.6 to 9.3 deg/s. Large displacements per unit time produce fast speeds of apparent motion whereas small displacements produce slow speeds. We will refer to the coherent/incoherent motion alternation at 0.83 Hz as the global update rate and the apparent motion jumps at 20 Hz as the dot update rate. In first experiment, VEPs were recorded from each eye of 15 normal-vision observers and 10 patients with strabismic amblyopia at 90% contrast (Michelson definition). In second experiment, VEPs were recorded from non-dominant eye of 10 normal-vision observers at 90% and 20% contrast. A schematic depiction of high contrast (90%) stimuli is shown in Figure 1. 

Figure 1

 

Schematic illustration of the stimuli. A single, accurately scaled frame of the random-dot kinematogram is shown in the upper portion of the figure. The stimulus cycle is illustrated schematically in the carton. A full 0.83 Hz (f1) cycle of the stimulus consisted of 0.6 s of lateral translation and 0.6 s of incoherent motion at the same dot displacement/speed. The direction of coherent motion alternated.

VEP recording and procedure

VEPs were recorded with Grass gold-cup electrodes at five electrode sites (PO7, O1, Oz, O2, and PO8) of the international 10–10 system, each referenced to Cz (Nuwer et al., 1999). Impedance was measured and maintained between 3 and 10 kΩ. The EEG was amplified at a gain of 50,000 with amplitude band-pass filter settings of 0.1 to 100 Hz at −6dB (Model 12BIN/8CH-23C; Grass Instruments, Quincy, MA). The EEG was digitized to a nominal 16-bit accuracy at a sampling rate of 500 Hz via a National Instruments PCI-MIO16XE-50 card controlled by in-house software (PowerDiva) that also performed signal processing and user interface functions. The observers were instructed to fixate a small marker in the center of the monitor. A VEP recording session consisted of 12 trials per condition for all observers (12 conditions, 6 speeds, two eyes). The trials were randomly interleaved across conditions in blocks of 6 trials. Viewing was monocular, and the non-viewing eye was occluded with black eye patch during the experiment. 

VEP Analysis

We used a recursive least squares adaptive filter (Tang & Norcia, 1995) to determine VEP amplitude and phase for the first 5 harmonics (1f1, 2f1, 3f1, 4f1, and 5f1) of the 0.83 Hz global update rate and the first harmonic (1f2) of 20 Hz dot update rate. Group means were computed by coherently averaging the spectral coefficients for each trial for each observer, channel (electrode site), harmonic, and stimulus condition. 

Statistical analyses

Differences between conditions and patient groups were assessed using a multi-variate approach to repeated measures (multi-variate analysis of variance or MANOVA) that takes into account the correlated nature of repeated measures data (for a review, see Keselman, Algina, & Kowalchuk, 2001; Keselman et al., 2001). Because MANOVA is designed to work with scalar- rather than complex-valued data, we transformed each observer's accumulated trials average into a scalar value by projecting their data onto the mean vector average of the group for each condition. By doing this, we were able to preserve a degree phase sensitivity in the analysis at the within group level. Responses that are consistent in phase across observers within a given condition lead to larger average projected amplitudes, while those that are not consistent across observers do not. Additive EEG noise in the measurements is reduced by this measure relative to a fully scalar version of MANOVA. 

Results

The time-averaged-evoked response ( Figure 2A) and the temporal frequency spectrum ( Figure 2B) to coherent motion onset/offset contains activity that is locked to both the 0.83 Hz global update rate and the 20 Hz dot update rate obtained by averaging the records of the dominant eyes of 15 normal-vision observers at Oz derivation. Here it can be seen in Figure 2A that the response after coherent motion onset (at 0 ms) differs from that after motion offset (at 600 ms). The response after motion onset is positive and sustained. The response after motion offset is dominated by a negative peak. This asymmetry manifests itself in the spectrum as the presence of both odd and even harmonics of the 0.83 Hz global update rate (1f1, 2f1, 3f1, etc., see Figure 2B). Activity related to the dot update rate can be seen superimposed on the waveform in Figure 2A and as a spike at 20 Hz (1f2) in Figure 2B. The spectrum contains several harmonics of the global update rate, with the first harmonic (1f1, 0.83 Hz) being dominant, as well as the first harmonic (1f2, 20 Hz) being dominant in dot update rate. We therefore performed our analyses on these two components (1f1 and 1f2), as these were the dominant responses related to the global and the dot update rates. 

Figure 2

 

(A) Time-averaged evoked response for a 9.3-arcmin displacement (3.1 deg/s speed) coherent motion onset/offset stimulus obtained by averaging the records of the dominant eyes of 15 normal-vision observers at Oz. (B) Temporal frequency spectrum of the response in panel A. The response contained harmonics of the 0.83 Hz global update rate (1f1, 2f1, 3f1, etc.) as well as activity locked to 20 Hz dot update rate (1f2). The spectrum resolution is 0.4 Hz, and the first bin plots the DC component.

Experiment 1: Tuning functions for speed/displacement at the global update rate

We first compared the displacement/speed tuning of the fellow, the amblyopic, and the normal-vision eyes. This analysis used the first harmonic (1f1) as an indicator of mechanisms that differentially encode transitions between coherent and incoherent motion and vice versa. The first harmonic was also the dominant response component that was time-locked to the changes in global motion coherence. In normal-vision eyes, the response function has an inverted U-shape at Oz in each eye, with the maximum response occurring at 3.1 deg/s (9.3 arcmin displacement; see Figure 3B). 

Figure 3

 

Vector-averaged first harmonic (1f1) responses at Oz (A and B) and PO7 (C and D) derivations across 10 patients with strabismic amblyopia (left column, A and C) and 15 normal-vision observers (right column, B and D). Error bars plot SEMs. At Oz, the peak of the tuning function was at 3.1 deg/s in the fellow eyes (A, open circles) and at 6.2 deg/s in the amblyopic eyes (A, filled circles). The tuning function was shifted toward large displacements/fast speeds for amblyopic eyes compared to fellow eyes. This shift was also present at PO7 (C, open and filled circles). The peaks of both eyes in normals are around 3.1 deg/s at Oz (B) and are somewhat broader at PO7 (D).

There are several differences between the response functions of the patients and those of the normal observers. First, both eyes of the patients show reduced amplitudes overall ( Figure 3A). Secondly, the form of the tuning function is different in the fellow and the amblyopic eyes. The fellow eye shows a peak at 3.1 deg/s, as in the normal-vision observers, but the peak of the amblyopic eye response is shifted to higher speeds. The difference in tuning was reflected as a significant three-way interaction between the factors eye (dominant/non-dominant), diagnosis (patient/normal), and displacement ( F = 4.6, p = 0.0064) on a repeated measure MANOVA. There was also significant main effect of displacement of comparable magnitude. Follow-up analyses (MANOVA) indicated that the form of the global motion tuning function did not differ between fellow and normal-vision eyes, but that the overall amplitudes of the normal eyes were larger than those of the patients at Oz ( p = 0.035) but not at PO7 ( p = 0.151). 

The response functions for normal-vision observers were less peaked at PO7 than they were at Oz, with the response being relatively larger at the largest displacements. The tuning functions of the normal and patient eyes were thus more similar at PO7 than at Oz, and the three-way interaction between eye, diagnosis, and displacement was not significant at PO7 ( p = 0.34). A follow-up analysis indicated that the tuning functions differed between Oz and PO7 ( F(1,14) = 18.174; p = 0008). This distinction in tuning at the different electrode sites is lost in the patients. 

Dot update responses

Evoked responses time-locked to the 20 Hz dot update rate show a different dependence on displacement than the responses that are time-locked to the update of the global motion state. In both normal-vision observers and in patients with strabismic amblyopia, dot update responses increased in amplitude for larger displacements ( Figure 4). Maximal responses were recorded for 6.2 deg/s in the normal-vision eyes but at 9.3 deg/s—the highest speed—in each eye of the patients with strabismus. Figure 4 plots vector-averaged first harmonic of dot update responses at Oz across subjects, where the largest dot update response were recorded. 

Figure 4

 

(A) Vector-averaged first harmonic dot update responses at Oz across 10 patients with strabismic amblyopia. (B) Same as a for 15 normal-vision observers. (C) Plots the two groups together. Open symbols are for the dominant eye (circles for the patients and squares for the controls), and filled symbols are for the non-dominant eyes. Error bars plot SEMs. The dot update responses increased in amplitude with increasing speed/displacement.

The first harmonic of the dot update response (1f2) was of lower amplitude in both eyes of the patients, and the decrease was more severe in the amblyopic eyes. However, the three-way interaction (factors eye (dominant/non-dominant), diagnosis (patient/normal), and displacement did not achieve statistical significance; F = 0.793, p = 0.579). interaction between diagnosis and displacement when the amblyopic eyes were compared to non-dominant eye of normals was stronger but only marginally significant ( F = 2.238, p = 0.093). 

Because the MANOVA on projected amplitudes ignores differences in mean phase between groups, it is an incomplete test of differences between the patient and the normals. That is phase consistency is preserved within a condition by projection, but differences in the mean phase between groups are not accounted for. We therefore re-plotted the data in the complex plane and found that there is a clear difference in the response phases of normal eyes versus either eye of the patients. The data are shown in Figure 5, which plots the amplitude and the phase of 1F2 as a function of displacement. The time/phase origin of the plot is at 0 deg (3 o'clock), and increasing phase delay is in the counterclockwise direction. The arrows linking the data points indicate increasing values of stimulus displacement. In the normal-vision observers, the phase of the response shifts towards the phase origin as speed increases. This is consistent with the response speeding up as displacement increases. The phase of the peak amplitude response of the normal eyes is 45 deg. A similar trend is visible in both eyes of the patients, but in here the phases follow a different trajectory, ending at a value near 0 deg for the largest displacements rather than a value near 45 deg as in the normal-vision eyes. 

Figure 5

 

Polar plot of the amplitude and phase of the dot update response in normal-vision eyes (black; mean of two eyes), fellow eyes (green), and amblyopic eyes (red). Arrows indicate the direction of increasing speed/displacement conditions. The circles plot ±1 SEM of the vector mean for the highest speed (9.3 deg/s). In normal-vision observers, the phase of the response shows a consistent progression towards the origin, consistent with a speeding of the response with increasing stimulus/displacement. The phase trajectories of both fellow and amblyopic eyes are shifted relative to that of normal-vision eyes.

The phase of the steady-state response is related to latency, but it is not possible to assign an unambiguous latency for these responses because of the modulo 2pi (50 ms) uncertainty associated with the steady-state response. A 45-deg phase shift, for example, is consistent with 6, 56, 106, 156, etc., ms latencies. Consequently the approximately 45 deg absolute difference in phase angle between the normal and the patient eyes is consistent with a roughly 6-ms speeding of the response (50 ms/8) in the patients or a 44-ms slowing. 

Experiment 2: Contrast dependence of motion VEPs

Psychophysical studies have reported that global motion sensitivity does not depend strongly on stimulus contrast and is constant for contrasts above about 10% (Edwards, Badcock, & Nishida, 1996; van de Grind, Koenderink, & van Doorn, 1987). In amblyopia, the visibility of the dots does not appear to be the critical limiting factor in global motion sensitivity (Constantinescu et al., 2005; Ho & Giaschi, 2006; Ho et al., 2005; Simmers et al., 2003, 2006). Because these are the first evoked potential measures of global motion responses in amblyopia, we also wished to determine whether any of the effects of strabismic amblyopia we observed (tuning changes, amplitude reductions) could be mimicked by reducing the contrast of stimuli presented to normal-vision eyes. 

In this experiment, we compared evoked responses for high contrast dots (90%) and much lower contrast dots (20%) presented at the same mean luminance used in the main experiment. From the psychophysical literature, we expected that responses at these two contrasts should be similar, given that sensitivity to coherent motion saturates at around 10% contrast. VEPs were recorded from the non-dominant eyes of 10 normal-vision observers. Figure 6 plots the vector-averaged first harmonic (1f1) of the global motion response and the first harmonic of the dot update response at 90% and 20% contrast. In Figures 6A and 6B, there were no significant global motion displacement/speed tuning differences between low and high contrast conditions at either Oz ( F = 0.79, p = 0.40) or PO7 ( F = 0.38, p = 0.55). 

Figure 6

 

Control experiment for effects of dot visibility. (A) Vector-averaged responses of first harmonic (1f1) of the global update rate at Oz. (B) As in panel A, but for PO7. (C) Amplitude of the dot update first harmonic response (1f2) at the Oz derivation. Open symbols plot data taken at 90% contrast, and filled symbols plot data taken at 20% contrast from the non-dominant eyes of 10 normal-vision observers. Error bars plot SEMs. There were no significant effects of contrast.

A comparison of low and high contrast dot update responses (see Figure 6C) shows that the amplitudes at slower speeds were, if anything, higher at low contrast than at high contrast. At faster speeds, the amplitude of both high and low contrast responses was the same. A one-factor MANOVA however showed no significant difference between high and low contrast dot update responses ( F = 2.46, p = 0.15). This control experiment thus suggests that the difference in tuning and overall amplitude seen in Figure 3 is not due to limitations on the visibility of the dots but is rather a failure of motion integration. 

Discussion

These are the first measurements of coherent motion-evoked responses in amblyopia. We found abnormalities in two different response components associated with coherent motion stimuli: Responses that are time-locked to the global motion update rate are reduced in amplitude in both eyes of the patients, and the speed/displacement tuning function is shifted to higher speed/larger displacements in the amblyopic eyes. Secondly, the timing of the response that is time-locked to the dot update rate is abnormal in both eyes of the patients and the amplitude of this response is also reduced, especially in the amblyopic eye. 

A previous VEP study of motion processing in amblyopia that measured motion onset responses to shifting checkerboards found equivalent responses in both eyes of 6- to 14-year-old children with anisometropic, strabismic, or both types of amblyopia (Kubová, Kuba, Juran, & Blakemore, 1996). The apparent contradiction between their results and ours highlights several issues. First, the underlying generators for the shifting check response and the coherent motion response may differ. Secondly, Kubová et al. (1996) did not compare their patients to normal-vision participants, and we thus do not know if both eyes were actually normal, only that they were equal. Third, the motion onset VEP may have still been developing and thus not have attained a full level of function. It is possible that the fellow eye may achieve higher levels of performance than the amblyopic eye at the end of the developmental period. Our participants were adults and should have completed maturation. There is also the possibility that our patients who all had long-standing amblyopia could have shown regressive effects. Finally, the treatment history of the two groups could have differed. It will be important in the future to study motion processing in patients and age-matched normals prior to the initiation of treatment. In this way it will be possible to dissociate treatment effects from the natural history of amblyopia and its effects on motion processing. It would also be of interest to compare different types of motion stimuli to determine the generality of motion processing deficits. 

The present results confirm and extend previous psychophysical studies of coherent motion processing in amblyopia. These studies have found elevated thresholds for coherent motion in the amblyopic eye that are independent of contrast sensitivity losses (Kiorpes et al., 2006; Simmers, Ledgeway, & Hess, 2005; Simmers et al., 2006). Fellow eye abnormalities of coherent motion processing have also been described before in psychophysical studies of human (Giaschi, Regan, Kraft, & Hong, 1992; Ho et al., 2005; Simmers et al., 2003) and monkey amblyopia (Kiorpes et al., 2006). These studies have found elevated motion coherence thresholds or reduced maximum displacement thresholds. The presence of perceptual and evoked response abnormalities in the non-amblyopic fellow eye suggest a specific deficit in motion processing that is not linked to dot visibility but is rather associated with abnormal binocular interaction during development. 

Tuning shifts of the global update response in the amblyopic eye

The tuning shift in the amblyopic eyes observed at the first harmonic of the global update rate is consistent with the results of previous psychophysical studies in macaque and human. Kiorpes et al. (2006) have found that coherent motion sensitivity functions for the amblyopic eyes of macaques raised with experimental strabismus were shifted toward larger spatial scales and faster speeds compared to fellow eyes. Steinman, Levi, and McKee (1988) reported that amblyopic deficits in discrimination of temporal asynchrony were especially large at slow speeds. Using grating stimuli, Hess and Anderson (1993) measured contrast sensitivity for direction discrimination as compared to motion detection. They reported a selective elevation of direction discrimination thresholds at slow speeds. The VEP tuning shift was not seen in the high-acuity fellow eyes and is thus likely to be due to the loss of small receptive fields caused by amblyopia. 

Models of coherent motion deficits in amblyopia

Previous studies of coherent motion processing in amblyopia have argued that the amblyopic deficit involves a primary failure at a second stage of motion integration. In two-stage models, the first stage has small receptive fields that are sensitive to the direction and displacement of sequentially presented dots. Because the analysis at this stage is modeled as being confined to small regions of space, it has been referred to as being “local.” Integration over larger areas of the visual field is required to detect coherent motion, e.g., motion that involves systematic similarities of motion direction that occur over extents that are larger than the size of the first stage receptive fields. This type of analysis has been referred to as “global,” and it is believed that this type of analysis is only accomplished in extra-striate cortical areas such as MT. 

One of the arguments for a specific second stage abnormality has been the observation that deficits in sensitivity to coherent motion are observed even after an accounting for the visibility of the local elements have been made or, more compellingly, in non-amblyopic fellow eyes (Constantinescu et al., 2005; Ho & Giaschi, 2006; Ho et al., 2005; Simmers et al., 2003, 2006). Thus, the first stage of local motion processing has been assumed to provide an adequate input to the second stage, once dot or local element contrast was sufficiently high in the amblyopic eye and at all times in the high-acuity fellow eyes. Elevated motion coherence thresholds were thus attributed to an inability of the second-stage to adequately pool local motion vectors. 

More recently, a variation of the two-stage model has been proposed (Hess, Mansouri, Dakin, & Allen, 2006; Mansouri & Hess, 2006). In this view, elevated motion coherence thresholds are to due to an inability to handle noisy stimuli rather than to integrate local direction signals. The studies that have proposed this view have measured detection thresholds for a single element of the global display and then compared amblyopic and fellow eye direction discrimination thresholds at a contrast level that renders direction discrimination for single elements equal in the two eyes. When this is done, direction discrimination thresholds for multi-element global displays are normal. The normal direction thresholds in the absence of noise are taken as evidence for an intact integration mechanism. Elevated motion coherence thresholds are then attributed to a failure of the second stage mechanism to segregate signal and noise elements rather than to an inability of the second stage to integrate direction signals over space. This model however does not provide a means to explain elevated maximum displacement thresholds (Ho & Giaschi, 2006) that occur for fully coherent stimuli nor does it account for the abnormalities we observe with 100% coherent stimuli for responses tied to both the global and dot update rates. 

In the context of two-stage models, a critical observation is that the output of the “local” mechanisms will be time-locked to the dot update rate if this rate is not too high. We intentionally updated the dot positions at a rate that is slow enough to “tag” the responses that underlie the local apparent motion sequences but not too slow to destroy the appearance of global motion. We are aware of only two previous reports of dot update responses. These are from the single-unit literature (Cook & Maunsell, 2002, 2004) where these responses are noted in passing as being present due to software timing limitations. These responses have not been reported previously in the EEG/MEG literature. 

We are interested in measuring the dot update responses because they potentially provide access to the first stage direction-selective mechanisms. We find these responses to be reduced in amplitude in the fellow eye, as well as in the amblyopic eye for fully coherent stimuli. The timing of these signals is also altered. It is thus possible that abnormalities of coherent motion sensitivity, such as those seen in fellow eyes, may be due to timing or other abnormalities at earlier stages. In a two-stage feed-forward model of global motion integration, timing deficits in the first stage (V1) would be passed on to the second stage (e.g., MT) where they might interfere with global integration, independent of visibility. It is also possible that deficits we have observed at the occipital pole are due to abnormal feedback signals from higher-level areas. There is however a caveat: Activity that is time-locked to the dot update rate may not be exclusively due to the output of direction-selective cells. These responses could also arise from cells that are only sensitive to time-varying changes in luminance or contrast. 

The results of the second experiment suggest that the amblyopic deficit cannot be simply explained as reduction in visibility or contrast sensitivity. In our second experiment, neither the global motion tuning functions nor the dot update responses were affected by reducing the dot contrast by a factor of 4.5. 

If the global motion deficit in amblyopia were due to reduced visibility, then the global motion tuning function should have shifted toward larger displacements/faster speeds for low contrast stimuli that are reduced in visibility as are the high contrast stimuli we used to test the amblyopic eye. This was not observed. 

Comparison with previous evoked response studies

Responses to the onset of coherent motion from a static dot field (Nakamura & Ohtsuka, 1999; Prieto et al., 2007) represent an unknown mixture of direction-selective responses and flicker responses and are thus difficult to interpret. A more focused comparison can be made by comparing responses to the onset of coherent versus incoherent motion from a static dot baseline (Lam et al., 2003; Maruyama et al., 2002; Ulbert et al., 2001) or from a blank field (Schulte-Körne, Bartling, Deimel, & Remschmidt, 2004) or by recording coherent motion onset responses after periods of random motion with the VEP (Kavcic, Fernandez, Logan, & Duffy, 2006; Niedeggen & Wist, 1998, 1999; Schulte-Körne et al., 2004) or MEG (Aspell et al., 2005; Händel et al., 2007; Kaneoke, Bundou, Koyama, Suzuki, & Kakigi, 1997; Lam et al., 2000; Nakamura et al., 2003). In these latter cases, the potential contamination by local flicker modulation that occurs when transitions from static patterns or blank fields used is accounted for, and evoked activity specific to the differences in the direction distribution of the dot motion vectors is measured. Each of these studies has measured transient motion onset responses in the time domain, so it is difficult to make direct comparisons with our frequency-domain approach. 

How “global” are the responses?

The “global motion” literature has concentrated on mechanisms that integrate motion signals over regions of space larger than single dot-pairs because only these may respond differently to coherent motion versus random motion. With our unlimited lifetime stimuli, differential responses could arise from cells that integrate trajectories just over a small number of dot updates and thus a relatively small part of the visual field, or they could arise from cells that integrate over much larger areas. The literature generally considers any response to limited lifetime coherent motion displays to be “global.” For our stimuli, it is probably better to refer to our responses as being “coherent motion” responses rather than “global motion” responses to reflect the unknown size of the pooling region of the cells that generate the response we record. 

Acknowledgments

Supported by EY015790. The authors thank Margaret Q. McGovern for her assistance in recruiting the observers with amblyopia and to Suzanne P. McKee for comments on an earlier draft. 

Commercial relationships: none. 

Corresponding author: Chuan Hou. 

Email: [e-mail]. 

Address: Smith-Kettlewell Eye Research Institute, 2318 Fillmore Street, San Francisco, CA 94115, USA.