Anisotropy in spatial summation properties of human Ocular-Following Response (OFR)

Abstract

Using sinusoidal gratings we show that an increase in stimulus size confined to the dimension orthogonal to the axis of motion leads to stronger Ocular Following Responses (OFRs) up to a certain optimal size. An increase beyond this optimum produces smaller responses, indicating suppressive interactions. In sharp contrast, when the stimulus growth occurs parallel to the axis of motion OFR magnitudes increase monotonically both for horizontal and vertical directions of motion. Similar results are obtained with 1D white noise patterns. However, the OFR spatial anisotropy is minimal with 2D white noise patterns, revealing a pivotal role of orientation-selective (i.e., cortical) mechanisms in mediating this phenomenon. The lack of anisotropy for 2D patterns suggests that directional signals alone are not sufficient to elicit this suppression. The OFR spatial anisotropy is potentiated if a stationary grating is presented for 600-1000 ms before its motion commences, further emphasizing the importance of static orientation signals. These results suggest that the strength of cortical spatial interactions is asymmetric—i.e., larger in the direction of the ends than the flanks of an orientation-selective receptive field—which corroborates the existing neurophysiological evidence.

1. Introduction

Center-surround interactions are the focus of many psychophysical as well as neurophysiological studies. Psychophysical work has shown that the detectability and discriminability of a stimulus (stationary or moving) can be strongly affected when other stimuli are presented at nearby and/or distant locations. Many factors influence the sign—suppression or facilitation—and strength of such interactions. These include the relative contrasts of the central and surround stimuli (Cannon & Fullenkamp, 1991, Cannon & Fullenkamp, 1996, Chen & Tyler, 2008, Ejima & Takahashi, 1985, Nurminen, Peromaa & Laurinen, 2010, Snowden & Hammett, 1998, Takeuchi & De Valois, 2000, Xing & Heeger, 2001), the center-surround spatial separation (Adini, Sagi & Tsodyks, 1997, Kapadia, Westheimer & Gilbert, 2000, Levi & Carney, 2011, Polat & Sagi, 1993, Polat & Sagi, 1994), the size of the surround (Cannon & Fullenkamp, 1996, Petrov & McKee, 2006, Saarela & Herzog, 2009, van der Smagt, Verstraten & Paffen, 2010), or its orientation and spatial frequency relative to the central stimulus (Chubb, Sperling & Solomon, 1989, Falkenberg & Bex, 2007, Solomon, Sperling & Chubb, 1993, Xing & Heeger, 2000).

The responses of neurons in visual cortex to their preferred stimulus are also modulated by stimuli located outside their classical receptive field. And many of the same factors outlined above can influence the sign and strength of such modulation (Cavanaugh, Bair & Movshon, 2002a, Henry, Joshi, Xing, Shapley & Hawken, 2013, Kapadia, Ito, Gilbert & Westheimer, 1995, Kapadia et al., 2000, Levitt & Lund, 1997, Mizobe, Polat, Pettet & Kasamatsu, 2001, Sceniak, Ringach, Hawken & Shapley, 1999, Shushruth, Nurminen, Bijanzadeh, Ichida, Vanni & Angelucci, 2013, Walker, Ohzawa & Freeman, 2000). In addition, single-unit recordings in V1 showed that with the same-orientation sinusoidal patches the suppressive interactions were stronger in the direction of the ends of an orientation-selective receptive field than in the direction of its flanks (Cavanaugh, Bair & Movshon, 2002b, Walker, Ohzawa & Freeman, 1999, but see also Webb, Tinsley, Barraclough, Parker & Derrington, 2003). At the neuronal population level this should result in anisotropy in the spatial summation of the oriented visual stimuli. However, a number of psychophysical studies failed to show such ends-flanks asymmetry (Cannon & Fullenkamp, 1991, Falkenberg & Bex, 2007, Foley, Varadharajan, Koh & Farias, 2007, Petrov & McKee, 2006, Xing & Heeger, 2001).

In this study we sought to identify a behavioral signature of this anisotropy, by measuring human ocular following responses (OFR). The OFR is a tracking eye movement elicited at ultra-short latency by the motion of a textured pattern (Gellman, Carl & Miles, 1990), and over the years it has emerged as a powerful behavioral tool for studying early stages of visual processing (Masson & Perrinet, 2012, Miles, 1998, Miles & Sheliga, 2010). Using sinusoidal gratings we revealed a robust anisotropy in the OFR spatial summation properties (Experiment 1). The anisotropy was greatly reduced for 2D noise patterns, but was still robust with1D noise (Experiment 2). This suggests that the anisotropy depends on the orientation content of the individual images, rather than resulting from the direction of pattern motion, that is it reflects orientation-selective mechanisms. These mechanisms seem to underlie the spatial anisotropy development in a stationary stimulus as well (Experiment 3). Some preliminary results of this study were presented in abstract form elsewhere (Sheliga, Quaia, FitzGibbon & Cumming, 2013).

2. Experiment 1: OFRs to sinusoidal gratings

2.1. Material and Methods

Most of the techniques will be described in brief, since they were similar to those used previously in this laboratory (Sheliga, Chen, FitzGibbon & Miles, 2005). Experimental protocols were approved by the Institutional Review Committee concerned with the use of human subjects. Our research was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki), and informed consent was obtained for experimentation with human subjects.

2.1.1. Subjects

Three subjects took part: two were authors (BMS and EJF) and the third was a paid volunteer who was not informed about the experiments’ purpose (AGB). All subjects had normal or corrected-to-normal vision. Viewing was binocular.

2.1.2. Eye-movement recording

An electromagnetic induction technique (Robinson, 1963) was used to record the horizontal and vertical positions of one eye (right eye in BMS and EJF; left eye in AGB). A scleral search coil was embedded in a silastin ring (Collewijn, Van Der Mark & Jansen, 1975), as described by Yang, FitzGibbon, & Miles (2003).

2.1.3. Visual display and the grating stimuli

Experiments were conducted in a dark room. Subjects’ heads were positioned using adjustable rests (for the forehead and chin) and a head band. A 21” CRT monitor was used for visual stimuli presentation; it was located straight ahead at 45.7 cm from the corneal vertex. The monitor screen had resolution of 1024 × 768 pixels (20.55 pixels/°, directly ahead of the eyes), a vertical refresh rate of 150 Hz, and a mean luminance of 20.8 cd/m². The video card signals were fed to the RGB inputs of the monitor via an attenuator (Pelli, 1997) and a video signal splitter (Black Box Corp., AC085A-R2). This arrangement permitted 11-bit grayscale resolution of black and white images.

The visual stimuli were vertical or horizontal gratings with sinusoidal luminance profiles (0.5 cpd; 32% Mickelson contrast) which underwent successive 1/8-wavelength shifts each video frame (18.75 cycles/sec). Gratings were confined to rectangular apertures centered directly ahead of the eyes, and whose horizontal and vertical dimensions were systematically varied. The aperture of the first subset of stimuli was alway ~6.2° wide (128 pixels ), whereas its height was assigned values from ~6.2° to ~25° (128 to 512 pixels) pixels in half-octave increments (i.e., 5 different heights). With horizontal motion (vertical gratings) the major axis of such stimuli was orthogonal to the axis of motion, whereas with vertical motion (horizontal gratings) the major axis of such stimuli was parallel to it. The aperture of the second subset of stimuli was always ~6.2° high, while its width varied from ~6.2° to ~25° in half-octave increments (i.e., 5 different widths). With horizontal motion (vertical gratings) the major axis of such stimuli was parallel to the axis of motion, whereas with vertical motion (horizontal gratings) the major axis of such stimuli was orthogonal to it. Refer to inserts in Figure 1 for examples: upper row – orthogonal configurations; lower row – parallel configurations. Each block of trials had 36 randomly interleaved stimuli: 9 aperture configurations¹, 2 axes of motion (horizontal vs. vertical), and 2 directions of motion: leftward vs. rightward or down vs. up.

Figure 1.

Experiment 1. OFRs to sinusoidal gratings of different size. Mean eye velocity profiles over time for subject BMS. Different sizes are coded by the darkness of individual traces (see the insert at the center). Each trace is the mean response to 62-66 repetitions of the stimulus. Left column: horizontal OFRs. Right column: vertical OFRs. Upper row: Orthogonal Configuration. Lower row: Parallel Configuration. Abscissa shows the time from the stimulus onset; horizontal dotted lines represent zero velocity; horizontal thick black line beneath the traces indicates the response measurement window. Grating examples are scaled versions of 0.5 cpd 32% contrast stimuli. Axis of stimulus motion is indicated by black arrows.

Experiment 1A

Experiment 1A compared responses to square patterns at different locations, with elongated stimuli that could be made from the summation of those square patches. Vertical sinusoidal gratings (0.5 cpd; 64% Mickelson contrast) undergoing horizontal shifts (18.75 cycles/sec) comprised the stimulus set. Gratings were confined to square or rectangular apertures. Square apertures measured ~6.2° by ~6.2° (128 by 128 pixels), were located at the horizontal or vertical meridian, and centered either at fixation, or ±6.2°, or ±12.5° away from it. Rectangular apertures were always centered at fixation. Their minor axes measured ~6.2°, while major axes were set to ~18.7° (384 pixels) or ~31.1° (640 pixels) along the horizontal or vertical meridian, which made such rectangular apertures spatially congruent to 3 (at fixation plus ±6.2°) or 5 (at fixation, plus ±6.2°, plus ±12.5°), respectively, square apertures stacked together. See upper panels of Figure 3 for the complete set of stimulus conditions of this experiment. 18 randomly interleaved stimuli comprised a single block of trials: 9 spatial configurations² and 2 directions of motion (leftward vs. rightward).

Figure 3.

Experiment 1A. (A-C) OFR spatial summation properties. The OFRs to square-aperture stimuli (sinusoidal gratings) required to spatially reconstruct a “bigger” rectangular-aperture stimulus are summed and divided by the OFR recorded to this “bigger” stimulus; the result is plotted as a function of the number of square-aperture stimuli used for such reconstruction. Circles: Orthogonal Configuration. Diamonds: Parallel Configuration. Thin dashed lines: horizontal – arithmetic sum prediction; diagonal – arithmetic average prediction. (D-F) Dependence of mean OFR amplitude on stimulus eccentricity. Square-apertures stimuli were used to obtain these dependences: stimuli could be located at the horizontal (filled circles; dotted lines) or vertical (open circles; solid lines) meridian, and centered either at fixation, or ±128 (±6.2°), or ±256 (±12.5°) pixels away from it. Grating examples on top of the figure are scaled versions of spatial configurations of 0.5 cpd 64% contrast stimuli. Subjects AGB (A, D), BMS (B, E), and EJF (C, F).

2.1.4. Procedures

Experimental paradigms were run by two PCs, communicating via Ethernet (the TCP/IP protocol). A Real-time EXperimentation software package (REX; see Hays, Richmond & Optican, 1982) was run on the first PC and provided the overall control of the experimental protocol and eye-movement data acquisition, display, and storage. The other PC utilized Matlab Psychophysics Toolbox extensions (Brainard, 1997, Pelli, 1997) to generate the visual stimuli.

A trial started with an appearance of a grating along with a fixation spot (dia. 0.25°) at the screen center. The fixation target disappeared and motion began if the subject’s eye remained within 2° of the fixation spot and no saccades had been detected (a velocity threshold of 18°/s) for a randomized period of 600 to 1000 ms. Following 200 ms of motion, the screen turned a uniform gray (luminance, 20.8 cd/m²) signaling the end of the trial. After 500 ms (an inter-trial interval) a new trial commenced. The subjects were given no instructions relating to the motion stimuli, but were asked to avoid blinking and/or shifting fixation except during the inter-trial intervals. The data were stored if no saccades were detected for the whole duration of the trial. Otherwise, the trial was aborted and repeated within the same block. Usually data collection continued for several sessions to permit good resolution of the responses (through averaging).

2.1.5. Data analysis

During the calibration procedure the horizontal and vertical eye position data were fitted with second-order polynomials to ensure position data linearization during the experiment. The signals were then smoothed with an acausal 6^th-order Butterworth filter (3 dB at 30 Hz). Trials with microsaccadic intrusions (<18°/s; i.e. those which avoided an online eye-velocity cut-off) were deleted. Finally, mean temporal profiles were computed for each experimental condition. To minimize the impact of directional asymmetries and improve the signal-to-noise ratio, the mean horizontal (vertical) eye position with each leftward (downward) motion stimulus was subtracted from the mean horizontal (vertical) eye position with the corresponding rightward (upward) motion stimulus (the “mean eye position”). Velocity (the “mean eye velocity”) was calculated as a difference between position samples 10ms apart (central difference method), and evaluated every millisecond. Response latency was defined as the moment in time after stimulus onset when the mean eye velocity first exceeded 0.1°/s. The OFRs were quantified by measuring the changes in the mean eye position signals—“OFR amplitude”—over the open-loop period, i.e., over the period up to twice the minimum response latency. For all the data of a given subject with a given stimulus set, this window always commenced at the same time after the stimulus onset (“stimulus-locked measures”); the actual time was determined by the shortest response latency, separately, for horizontal (66, 65, and 69 ms for subjects AGB, BMS, and EJF, respectively) and vertical (64, 65, and 72 ms for subjects AGB, BMS, and EJF, respectively) OFR datasets. However, to permit within-subject comparisons across the datasets, for a given subject the duration of this measurement window was the same for all conditions (64, 65, and 69 ms for subjects AGB, BMS, and EJF, respectively).

We used bootstrapping for all our statistical analyses. The p-value for statistical significance was set to 0.05. In case of multiple comparisons this value was divided by the number of comparisons (Bonferroni correction). All error bars in the Figures are one standard error of the mean (SEM; actually they were smaller than a symbol size in the vast majority of the OFR amplitude cases and, therefore, not visible on the graphs).

2.2. Results

Figure 1 shows mean horizontal eye velocity profiles for subject BMS in response to stimuli of different sizes (see grayscale coding of velocity traces). Conditions in which the axis of stimulus size growth was orthogonal to that of motion—which will be referred to as an Orthogonal Configuration (OC)—are plotted in the upper row, whereas conditions in which the axis of stimulus size growth and motion were the same—which will be referred to as a Parallel Configuration (PC)—are plotted in the lower row (note grating inserts in each panel). It can be clearly seen that in OC there is a range of stimulus sizes over which increases in size lead to smaller OFRs, with the largest stimuli—shown by the darkest traces—actually resulting in the weakest responses. In case of vertical motion the largest response is for the smallest stimulus, although had we explored sufficiently small stimuli they should inevitably have produced weaker responses. Thus these data indicate that there is an optimal stimulus size for OC. In sharp contrast, increasing size always produced larger OFRs for PC. Horizontal and vertical motion data occupy left and right columns of the Figure 1, respectively, and reveal the same pattern of results. Figure 2A-C quantifies these observations for three subjects. In each subject the semi-log plots of the OFR amplitude vs. stimulus size are close to linear for PC (filled symbols; dotted lines), whereas for OC they all show hyper-saturation, although the detailed shape is somewhat idiosyncratic (open symbols; solid lines). The differences in the OFR amplitudes to the same-size stimuli belonging to OC vs. PC emerge very soon—see the divergence of solid and dotted traces: as early as at 8.8° and no later than at 17.6°. With horizontal OFRs the OC vs. PC differences reached significance for 1, 3, and 2 largest stimulus sizes for subjects AGB, BMS, and EJF, respectively. With vertical OFRs the OC vs. PC differences reached significance for 4, 3, and 1 largest stimulus sizes for subjects AGB, BMS, and EJF, respectively. The OFR latency data are shown in Figure 2D-F, and reveal a clear dependence on size—larger the stimulus, sooner the OFRs are generated—with no apparent sensitivity to other stimulus attributes. The anisotropy we see in response magnitude is not present in response latency (p>0.05, ns).

Figure 2.

Experiment 1. (A-C) Dependence of mean OFR amplitude on stimulus (sinusoidal grating) size for subjects AGB (A; 82-90 trials per condition), BMS (B; 62-66 trials per condition), and EJF (C; 65-80 trials per condition). (D-F) OFR latency dependence on stimulus size for subjects AGB (D), BMS (E), and EJF (F). Circles: horizontal OFRs. Diamonds: vertical OFRs. Open symbols, solid lines: Orthogonal Configuration. Filled symbols, dotted lines: Parallel Configuration.

A useful way to characterize this anisotropy is by quantifying the extent to which the summation is or is not linear for the two configurations. We do this with an approach developed in our recent paper (Sheliga, Quaia, Cumming & Fitzgibbon, 2012): comparing responses to “elementary blocks” with responses to stimuli that are obtained by summing those blocks. We can then compare the observed response to that predicted by linear summation. We plot $\frac{{\sum }_{i=1}^N{R}_i}{R_{obs}}$ as a function of N, where R_i is the OFR amplitude to motion of the i-th “block”, N is the number of “blocks” required for spatial reconstruction of a bigger stimulus, and R_obs is the actual OFR amplitude recorded for this stimulus. In this plot, pure averaging will result in points sitting on the identity line, while linear summation will produce a horizontal line at unity³.

Experiment 1A compared responses to square elementary building “blocks” to those for larger stimuli occupying rectangular apertures (see Methods; note that only horizontal motion is used in this experiment). Figure 3A-C summarizes results for all subjects and shows a fit with the linear function

$$\frac{\sum _{i=1}^N{R}_i}{R_{obs}}=k^{\ast }N+(1-k)$$ (1)

Sheliga et al. (2012) called parameter k the “Averaging Coefficient” since k=1 corresponds to the arithmetic average ($R_{obs}=\frac{{\sum }_{i=1}^N{R}_i}{N}$), while k=0 corresponds to the arithmetic sum ($R_{obs}={\sum }_{i=1}^N{R}_i$). Figure 3A-C shows that the Averaging Coefficients were always greater than zero, i.e. the OFR spatial summation in both OC and PC is sub-linear. However, averaging is considerably stronger in OC than in PC: k_OC vs. k_PC compared as 0.28 vs. 0.08, 0.50 vs. 0.21, and 0.35 vs. 0.21 for subjects AGB, BMS, and EJF, respectively. The OC vs. PC differences were significant for both stimulus sizes (3-blocks and 5-blocks) in all three subjects. The OFR amplitude dependencies on stimulus horizontal vs. vertical eccentricity were very similar (Figure 3D-F)⁴, so this factor could not play any role in an emergence of the anisotropy in the OFR spatial summation properties.

Others have shown that rigid line-endings of spatially localized moving grating stimuli—“terminators” (e.g., Barthelemy, Fleuriet & Masson, 2010, Masson, Rybarczyk, Castet & Mestre, 2000)—contribute to OFRs of humans and nonhuman primates. This potential contribution is relevant for the interpretation of the PC vs. OC anisotropy results of this study. Indeed, increasing the stimulus size in the PC increases the number of terminators, whereas increasing the stimulus size in the OC keeps the number of terminators constant, but moves them peripherally. These changes in terminators might, in principle, account for PC vs. OC anisotropy effects. We, therefore, ran a control experiment in one subject (BMS), in which we utilized 3 of 5 stimulus sizes used in Experiment 1 (~6.2°, ~12.5°, ~25°) in two configurations, PC and OC. Horizontal and vertical edges of all stimuli were smoothed using a raised cosine profile (0.25 cpd). The OFRs still showed significant spatial anisotropy effects (Supplementary Figure S1), and thus the “terminator” mechanism contribution is not sufficient to explain the PC vs. OC anisotropy.

3. Experiment 2: OFRs to 1D and 2D white noise stimuli

The anisotropy identified in Experiment 1 indicates that spatial summation depends upon the stimulus. There are two possible sources for this: it may derive from orientation selective mechanisms, and hence depend on the orientation of the individual frames, regardless of any motion. Alternatively, it could be derived from direction selective mechanisms that could be activated by orientation-broadband images. To differentiate these possibilities, we measured spatial summation using 1D and 2D noise patterns. Both of these stimuli can be used to define motion direction, but the 2D noise is broadband in orientation.

3.1. Material and Methods

Only methods and procedures that were different from those used in Experiment 1 will be described here.

3.1.1. Visual stimuli

The visual stimuli were vertical 1D and 2D white noise patterns (48% Mickelson contrast; 9-pixel single check) undergoing successive 5-pixel horizontal shifts each video frame (~36.5°/s at 150 Hz monitor refresh rate). Noise patterns were confined to apertures identical to those (in size and number) used for gratings in Experiment 1. Examples of patterns are shown at the top of Figure 4. 36 randomly interleaved stimuli comprised a single block of trials: 9 aperture configurations, 2 types of stimuli (1D vs. 2D), and 2 directions of motion: leftward vs. rightward.

Figure 4.

Experiment 2. (A-C) Dependence of mean OFR amplitude on noise pattern size for subjects AGB (A; 120-130 trials per condition), BMS (B; 77-84 trials per condition), and EJF (C; 91-104 trials per condition). (D-F) OFR latency dependence on stimulus size for subjects AGB (D), BMS (E), and EJF (F). Circles: 1D noise. Diamonds: 2D noise. Open symbols, solid lines: Orthogonal Configuration. Filled symbols, dotted lines: Parallel Configuration. Noise patterns examples on top of the figure are scaled versions of 48% contrast stimuli. Axis of stimulus motion is indicated by black arrows.

3.2. Results

Figure 4A-C shows the OFR amplitude results for all three subjects, plotting them as a function of stimulus size. There is a striking difference between responses to 1D (circles) and 2D (diamonds) noise. There is minimal spatial anisotropy with 2D patterns as OC (open diamonds; solid lines) and PC (filled diamonds; dotted lines) responses show very similar dependence on size. On the other hand, the spatial anisotropy is very strong with 1D patterns, with very little change in response as a function of size in the OC condition (open circles; solid lines). With 1D noise the OC vs. PC differences reached significance for 2, 3, and 3 largest stimulus sizes for subjects AGB, BMS, and EJF, respectively. With 2D noise the OC vs. PC differences reached significance for the largest stimulus size in subject AGB only. However, in this particular subject the OC vs. PC differences for the largest stimulus size were significantly larger with 1D than 2D noise. The OFR latencies are plotted in Figure 4D-F, and, though rather noisy, are in general inversely proportional to stimulus size, regardless of stimulus configuration (OC vs. PC) and/or stimulus type (1D vs. 2D noise): with 1D noise the OC vs. PC differences reached significance for two intermediate stimulus sizes (12.5° and 17.6°) for subject AGB only, while no OC vs. PC latency differences were significant for the other two subjects.

4. Experiment 3: The impact of the pre-motion visual stimulus

In all of the data presented above, a stationary stimulus was present during the fixation period, and then this stimulus started to move. Experiment 2 suggested that it is the orientation content of individual images that suppresses responses to large stimuli. This raises the possibility that the stationary stimulus present during fixation contributes to the anisotropy. Experiment 3 addresses this issue, by comparing responses with and without the stationary stimulus during fixation.

4.1. Material and Methods

Only methods and procedures that were different from those used in Experiment 1 will be described here.

4.1.1. Visual stimuli and procedures

Vertical sinusoidal gratings (0.5 cpd; 32% Mickelson contrast) undergoing horizontal shifts (18.75 cycles/sec) were used. As in Experiment 1 they were confined to rectangular apertures. In the Orthogonal Configuration (OC) the aperture was ~6.2° (128 pixels) wide, whereas its height was set to ~6.2°, ~12.5°, or ~25° (128, 256, or 512 pixels). The aperture in the Parallel Configuration (PC) was ~6.2° high, and now its width was set to ~6.2°, ~12.5°, or ~25°. During the fixation period the display always contained a fixation target but could include or lack a grating pattern. In the former case, after a randomized period of 600 to 1000 ms the fixation target would disappear and the motion sequence would begin by horizontally shifting the grating already in place, while in the latter case, the disappearance of the fixation target would coincide with the appearance of the grating pattern which would make its first horizontal shift one video frame later (~7 ms). 20 randomly interleaved stimuli comprised a single block of trials: 5 aperture configurations⁵, 2 fixation conditions (grating present vs. grating absent), and 2 directions of motion: leftward vs. rightward.

4.2. Results

Figure 5 summarizes the OFR amplitude results. Data of each subject occupy a single row, while columns show fixation conditions: presence (left column) or absence (right column) of a grating pattern during the fixation period. Comparing data in two columns one can readily conclude that the difference between PC and OC responses was smaller in the grating-absent condition. This effect (a smaller difference in the grating-absent condition) reached statistical significance for 2, 1, and 1 largest stimulus sizes for subjects AGB, BMS, and EJF, respectively. Thus, the anisotropy in spatial summation (the difference between PC and OC) is greater when an oriented stimulus is present during the fixation period. Although the direction of changes was the same in all three subjects, their strength was not: in the grating-absent condition the OFR spatial anisotropy was minimal in subject AGB, while it was still substantial in subjects BMS and EJF.

Figure 5.

Experiment 3. Dependence of mean OFR amplitude on stimulus size. Data of each subject occupy a single row, while columns show fixation conditions: presence (left column) or absence (right column) of a sinusoidal grating during the fixation period. Open symbols, solid lines: Orthogonal Configuration. Filled symbols, dotted lines: Parallel Configuration. Subjects AGB: 209-228 trials per condition. Subject BMS: 193-199 trials per condition. Subject EJF: 108-128 trials per condition.

One possible complication with the interpretation is that the presence of the stationary grating may influence the pattern of fixational eye movements in the period before stimulus motion, and hence alter the effective retinal stimulus (Goffart, Quinet, Chavane & Masson, 2006, Martinez-Conde, Macknik, Troncoso & Hubel, 2009, Poletti & Rucci, 2010). This may explain some differences between conditions with and without a stationary grating, but it is hard to see how it could result in the anisotropy we observe in response to subsequent motion.

5. Discussion

We have previously shown that the amplitude of OFRs shows sub-linear spatial summation indicating the action of suppressive mechanisms for large sizes (Sheliga et al., 2012). Here we show that this suppression is very anisotropic for spatially oriented stimuli (sinusoidal gratings and 1D white noise)—the suppression is much greater when the stimulus is extended in a direction parallel to oriented contour than orthogonal to it. This anisotropy is not evident with 2D white noise stimuli, which suggests that spatial summation is not affected by directional motion signals alone. The anisotropy is potentiated if a stationary grating is present in the display for several hundred milliseconds before its motion commences.

The dependence on stimulus orientation suggests that the anisotropy in spatial summation is of cortical origin. Orientation selectivity first arises in the primary visual cortex, and therefore our result can only be attributed to the responses of neuronal populations of this and/or later stages of visual processing. The anisotropy we demonstrate also resembles one reported for cortical neurons stimulated with sinusoidal gratings. Cavanaugh et al. (2002b) showed that in primate V1 the suppressive interactions were stronger in the direction of the ends than the flanks of an orientation-selective receptive field. Because our 2D noise stimulus will activate a broad range of orientation selective neurons, the anisotropy of individual neurons would not give rise to anisotropy in the summed population response. Thus the anisotropy we observe might largely reflect the effect of orientation selective surround inhibition, and may therefore share the same mechanism as the surround suppression reported for OFR by Barthelemy (2006).

The suggestion that the anisotropy reflects orientation-selective mechanism rather than direction-selective mechanisms is further supported by the effects of a stationary oriented grating present in the display before being moved. This generates suppression of a population of orientation-selective neurons before the motion signal is added. This also implies that significant time is required for this suppression to develop – if it arose instantaneously, it should be equally effective in the condition with no stimulus present during fixation. Characterizing how this suppression evolves over time will be a subject of future research.

Interestingly the anisotropy we demonstrate here is more marked that has been reported in single neurons. This may reflect processes downstream of the striate cortex. However, it may also be reconciled with the responses of V1 neurons when the entire population response is considered. As our stimuli become larger, a larger population of V1 receptive fields will be engaged. How this population sum grows depends on more than the typical size tuning of individual neurons.

Similarly, it might seem at first sight that the lack of anisotropy for 2D noise stimuli makes it difficult to explain these results in terms of responses in MT. Many neurons there show reduction in response magnitude with increasing size, even for 2D noise patterns (Born, 2000, DeAngelis & Uka, 2003, Raiguel, Van Hulle, Xiao, Marcar & Orban, 1995). But although neurons with RFs at the center of our stimuli would be subject to this suppression, the increasing size adds new neurons to the active population. So the suppression of individual neurons means that the population response shows sub-linear summation, but it need not necessarily show a reduction in response for large sizes. Since we find sub-linear summation in all conditions, the observer OFR magnitudes may be compatible with a summed output of MT neurons.

Several psychophysical studies which used suprathreshold sinusoidal gratings did not show differences in perceptual judgments of various attributes of “central” stimuli when the location of the same-orientation “surround” gratings was manipulated (Cannon & Fullenkamp, 1991, Falkenberg & Bex, 2007, Petrov & McKee, 2006, Xing & Heeger, 2001). These results seem to be in consistent with observations made in this paper. However, in our OFR data there is also a parameter which does not show spatial anisotropy: the OFR latency (see Figures 2D-F and 4D-F). It may be that the signals driving the perceptual judgments are more reflected in the OFR latency than in the OFR amplitude. On the other hand, Foley et al. (2007) compared detection thresholds to sinusoidal gratings of different form and concluded that “tall thin patterns have essentially the same thresholds as short wide patterns of the same area.” Here the fact that the stimuli were presented at low contrast probably explains the difference. A number of neurophysiological studies have shown reductions in surround suppression at low contrast (e.g., Cavanaugh et al., 2002a, Kapadia, Westheimer & Gilbert, 1999, Sceniak et al., 1999).

In conclusion, we demonstrate anisotropy in spatial summation of visual inputs for the OFR. This is dependent on stimulus orientation rather than direction, which suggests that it probably reflects length suppression in the striate cortex.

• The Ocular-Following Response (OFR) spatial summation properties are anisotropic

• The OFR spatial anisotropy is strong for spatially oriented stimuli

• Unoriented moving stimuli (2D noise) produce minimal anisotropy.

• Static orientation signals play a role in the OFR anisotropy

• The OFR anisotropy resembles ends-flanks suppression asymmetries in V1neurons