Eccentricity dependence of orientation anisotropy of surround suppression of contrast-detection threshold

Abstract

Both neurophysiological and psychophysical data provide evidence for orientation biases in nonfoveal vision—specifically, a tendency for a Cartesian horizontal and vertical bias close to fixation, changing to a radial bias with increasing retinal eccentricity. We explore whether the strength of surround suppression of contrast detection also depends on retinotopic location and relative surround configuration (horizontal, vertical, radial, tangential) in parafoveal vision. Three visual-field locations were tested (0°, 225°, and 270°, angle increasing anticlockwise from 0° horizontal axis) at viewing eccentricities of 6° and 15°. Contrast-detection threshold was estimated with and without a surrounding annulus. At 6° eccentricity, horizontally oriented parallel center–surround (C-S) configurations resulted in greater surround suppression compared to vertically oriented parallel center–surround configurations (p = 0.001). At 15° eccentricity, radially oriented parallel center–surround stimuli conferred greater suppression than tangentially oriented stimuli (p = 0.027). Parallel surrounds resulted in greater suppression than orthogonal surrounds at both eccentricities (p < 0.05). At 6° the horizontal center was more susceptible to suppression than a vertical center (p < 0.001) for both parallel and orthogonal surrounds, while at 15° a radial center was more susceptible to suppression (relative to a tangential center), but only if the surround was parallel (p = 0.005). Our data show that orientation anisotropy of surround suppression alters with eccentricity, reflecting a link between suppression strength and visual-field retinotopy.

Introduction

It has long been observed that our sensitivity to visual contours is not the same for all stimulus orientations. Thus, the oblique effect refers to superior performance in processing stimuli that are constructed of vertical or horizontal orientations compared to oblique orientations. This effect has been reported for a number of psychophysical tasks, including orientation discrimination (Bouma & Andriessen, 1968; Heeley & Buchanan-Smith, 1996; Westheimer, 2001; Westheimer, 2003a; Westheimer & Lavian, 2013), contrast detection (Campbell & Kulikowski, 1966), resolution of grating orientation (Rovamo, Virsu, Laurinen, & Hyvarinen, 1982), scene perception (Nasr & Tootell, 2012), and vernier acuity (Westheimer & Beard, 1998). The presence and strength of the oblique effect varies with visual eccentricity (Mansfield, 1974; Rovamo et al., 1982; Westheimer, 2003a). For example, Rovamo et al. (1982) reported that the oblique effect observed in foveal human vision becomes progressively weaker moving away from the fovea through to 8° of visual eccentricity, and disappears between 8° and 18° for a task involving resolution of the orientation of a grating. For eccentricities greater than 18° (up to 30°), the oblique effect was replaced by a radial effect in which a meridionally oriented grating patch was resolved more easily than one that was perpendicular to the visual-field meridian. Similarly, Westheimer (2003a) studied orientation discrimination of lines at the fovea, at eccentricities of 9° and 18°–20°, and observed decreased thresholds for horizontal and vertical lines relative to oblique ones at all eccentricities. Further, outside foveal vision, orientation discrimination indicated better sensitivity for radial than tangential lines, especially along oblique meridians. 

Alternate methods that have been used to study orientation anisotropy in human and nonhuman vision include neurophysiological recordings (De Valois, Yund, & Hepler, 1982; Mansfield, 1974; Payne & Berman, 1983; Vidyasagar & Henry, 1990; Vidyasagar & Urbas, 1982), optical imaging (Wang, Ding, & Yunokuchi, 2003; Xu, Collins, Khaytin, Kaas, & Casagrande, 2006), visual evoked potentials (Bonds, 1982; Mansfield & Ronner, 1978), functional magnetic resonance imaging (fMRI; Furmanski & Engel, 2000), and, more recently, magnetoencephalography (MEG; Cichy, Ramirez, & Pantazis, 2015) techniques. For instance, measures of differences in blood-oxygen-level-dependent (BOLD) responses using fMRI techniques has demonstrated a shift from an oblique effect to a radial effect with increasing visual eccentricity (Furmanski & Engel, 2000; Mannion, McDonald, & Clifford, 2010). Furmanski and Engel (2000) reported stronger responses in V1 for cardinal orientations (i.e., for horizontal and vertical) than oblique orientations when stimuli were presented at 4.5° visual eccentricity in humans. However, for stimuli presented at 7.2° eccentricity, Mannion et al. (2010) reported that fMRI BOLD responses in visual areas V1, V2, V3, and V3A/B were lowest for horizontal orientations, intermediate for vertical orientations, and highest for oblique orientations when grating orientation was considered independent of the visual field. When the corresponding angular orientation was considered, the response was higher for radial orientations compared to tangential orientations in visual areas V1, V2, V3, V3A/B, and hV4. 

Moving further away from the fovea, Sasaki et al. (2006) reported increased fMRI BOLD responses for radial over tangential orientations in V1 of both human subjects and nonhuman primates when stimuli were presented at 15.5° eccentricity. They also found greater sensitivity for contrast detection of gratings with radial orientations than tangential orientations in human subjects based on psychophysical observations at 15.5° eccentricity. 

A combined inference that arises from these reports is that orientation anisotropy depends on stimulus eccentricity. In fact, orientation selectivity itself and anisotropies based upon visual-field location may have their origins early in the visual pathways, possibly even in the retina (Levick & Thibos, 1982; Passaglia, Troy, Ruttiger, & Lee, 2002) and lateral geniculate nucleus (Shou & Leventhal, 1989; Vidyasagar & Urbas, 1982). Despite repeated evidence in the literature that orientation anisotropy of detection and discrimination thresholds are dependent on eccentricity, to our knowledge no study has systematically investigated the effect of a surround stimulus on the perception of a central target in relation to eccentricity. Analogous to surround suppression observed in neurons in the visual pathway, numerous psychophysical studies have demonstrated that detecting or discriminating a grating pattern can be facilitative or suppressive depending on several factors related to surrounds (Cannon & Fullenkamp, 1991; Levitt & Lund, 1997; Nguyen & McKendrick, 2016b; Yu, Klein, & Levi, 2002). These factors include contrast, orientation, spatial frequency, and eccentricity in the visual field. Of particular relevance to our study is the observation by Petrov and McKee (2006) that at 6° parafoveally, a radially placed surround layout produced stronger surround suppression of contrast-detection thresholds than a tangential layout by a factor of 1.7. Note that the aim of their study was not to investigate the effect of center or surround grating orientation on suppression strength, nor did they investigate whether the effect held or was altered for retinotopic locations farther from fixation. 

Here we studied whether surround suppression of contrast detection depends on retinotopic location in parafoveal vision using stimuli that are designed to test V1 orientation properties in human observers (Nguyen & McKendrick, 2016a; Shushruth et al., 2013). Parallel center and surround orientations were first chosen to test the effect of retinotopic location on surround suppression, since such collinear patterns have maximum suppression strength for a particular spatial frequency and surround contrast. Another aim of our study was to understand if there is an anisotropy of suppression strength that depends upon the retinal meridian at two chosen parafoveal eccentricities (6° and 15°). Finally, we also studied the effect of surround suppression by parallel and orthogonal orientations on processing of center orientation at 6° and 15° eccentricities. 

Methods

Participants

Four experienced psychophysical observers and one inexperienced (four unaware of the experimental purpose; three women, two men; ages = 23–32 years, M = 28) participated in the experiment. Participants provided written informed consent, and the study was approved by the University of Melbourne Human Research Ethics Committee according to a protocol consistent with the Declaration of Helsinki. Subjective refraction was conducted to determine refractive error, and all subjects were emmetropic with monocular visual acuity of 6/6 or better without optical correction. Ocular health was assessed using a slit-lamp biomicroscope and ophthalmoscope to ensure absence of ocular pathology or media opacities. 

Stimuli and procedures

Procedures to generate stimuli were written in MATLAB v7.6 (MathWorks, Natick, MA) and displayed on a gamma-corrected monitor (G520 Trinitron; Sony, Tokyo, Japan; frame rate = 100 Hz; resolution: 1,024 × 768 pixels; maximum luminance = 100 cd/m²) interfaced with a ViSaGe graphics system (Cambridge Research Systems, Ltd., Kent, UK). The screen was viewed monocularly with the left eye in a dimly lit room from 573 mm (stimuli at 6°) or 400 mm (stimuli at 15°). Participants sat comfortably using a chin and forehead rest to secure steady positioning throughout the experiments. For each participant, overall testing duration was about 4 hr and was completed in three or four sittings. 

The target stimulus was a circular sinusoidal grating similar to the near surround stimuli of the perceptual contrast-matching study by Shushruth et al. (2013) and was presented at 6° or 15° in order to allow comparison between eccentricities. Stimulus parameters at 6° were as follows: center radius = 1°, inner surround radius = 1.5°, outer surround radius = 2.5°, spatial frequency = 1 c/°. Stimuli were always presented at the center of the screen to ensure consistent luminance, and the fixation spot (0.1° radius) was positioned according to the visual-field location being tested and remained on the screen all the time. Participants were instructed to fixate on the fixation spot. Experiments testing performance for the central target alone and the center surrounded by a 40% contrast annular grating of the same phase and spatial frequency were tested in separate blocks. Test locations were horizontal nasal (0°), oblique inferior (225°), and vertical inferior (270°) from the foveal fixation (see Figures 1 and 2). At each test location, contrast-detection threshold was estimated for horizontal, vertical, radial, and tangential grating orientations with and without the surround annulus. The surround condition (with or without), stimulus location (0°, 225°, or 270°), and orientation (horizontal, vertical, radial, or tangential) were presented for each participant in a random sequence in separate experimental runs. 

Figure 1

 

Stimulus illustration for parallel center–surround gratings at horizontal, oblique, and vertical retinotopic locations, with mean suppression ratios for horizontal and vertical orientations at eccentricities of 6° (left panel) and at 15° (right panel); error bars indicate 95% confidence interval of the mean.

Figure 2

 

Stimulus illustration along horizontal, oblique, and vertical retinotopic locations, with mean suppression ratios for radial and tangential orientations at eccentricities of 6° (left panel) and 15° (right panel); error bars indicate 95% confidence interval of the mean.

A two-interval forced-choice method was used to determine the contrast-detection threshold. The target was presented randomly in the first or second interval with a stimulus duration of 150 ms and an interstimulus interval of 500 ms. The participant indicated the interval which carried the target by pressing a button (CB6, Cambridge Research Systems). The target contrast was varied using a three-down, one-up staircase with a step size of 20% and converging on the approximate 79% correct level (Wetherill & Levitt, 1965). The staircase was terminated after six reversals and the last four reversals were averaged. This leads to an average of around 35 presentations per run. Each threshold estimate was repeated twice and the average of the two taken for analysis. A ratio of the contrast-detection thresholds with and without the surround was taken as the suppression ratio. A ratio of 1 denotes no suppression. 

The same procedure was repeated for threshold estimation at 15° eccentricity, by projecting the fixation point at 15° visual angle while the stimulus remained on the center of the screen. All participants performed the 15° task after the 6° task. Stimulus size and spatial frequency were scaled for cortical magnification at nasal (0°), average of temporal and inferior (225°), and inferior (270°) visual fields (Rovamo & Virsu, 1979). Respectively for 0°, 225°, and 270° visual-field locations, stimulus parameters were as follows: center radii (°): 2.07, 2.12, 2.04; inner surround radii (°): 3.1, 2.95, 3.06; outer surround radii (°): 5.16, 5.3, 5.1; spatial frequency (c/°): 0.48, 0.47, 0.49. Surround contrast remained at 40%. 

Statistical analysis

The main data analysis was conducted using SPSS 20.0. Normality and sphericity of data were tested using the Kolmogorov–Smirnov normality test and Mauchly's sphericity test. Repeated-measures analysis of variance (ANOVA) was conducted to compare various within-subject factors detailed in the Results section. A p value of 0.05 was considered as the criterion for statistical significance. Permutation tests were conducted using RStudio (RStudio, 2016). 

Results

Parallel center–surround suppression: Horizontal and vertical orientations

Figure 1 shows the stimulus orientation at retinotopic locations horizontal, oblique, and vertical. All individual data are shown in Supplementary Figure S1. We determined separately whether there was a bias for horizontal or vertical orientation for parallel (maximum suppression) center–surround (C-S) configurations at 6° or 15° eccentricity. A two-way repeated-measures ANOVA—3 retinotopic locations (0°, 225°, 270°) × 2 orientations (horizontal, vertical)—indicated a significant main effect of orientation at 6°, F(1, 4) = 94.8, p = 0.001, but no such effect at 15°, F(1, 4) = 0.12, p = 0.75. In other words, at 6° the horizontal configuration exhibited stronger suppression than the vertical (left panel). Such an effect was not present at an eccentricity of 15° (right panel). We also used permutation tests to shuffle data labels (i.e., horizontal vs. vertical, 1,000 samples), and significance was reassured (p < 0.01). The bar graphs of Figure 1 show the mean suppression ratio of horizontal and vertical patterns at 6° (left panel) and 15° (right panel) eccentricities. There was no significant main effect of retinotopic location, nor a Retinotopic location × Orientation interaction at either 6° [F(2, 6) = 2, p = 0.22; F(2, 6) = 2, p = 0.78] or 15° [F(1, 5) = 3, p = 0.16; F(2, 7) = 3, p = 0.1]. 

Parallel center–surround suppression: Radial and tangential orientations

Figure 2 plots the mean suppression ratio, with the stimuli classified as radial or tangential. All individual data are shown in Supplementary Figure S1. The mean suppression ratio of parallel C-S orientation for the radial configuration was significantly higher than for the tangential configuration at 15° (right panel), F(1, 4) = 11.5, p = 0.027, but not at 6° (left panel), F(1, 4) = 0.012, p = 0.919. In summary, the suppression ratio of a parallel C-S configuration of radial orientation is greater than that of tangential orientation at 15° eccentricity but not at 6° eccentricity. Further, we used permutation tests to shuffle data labels (i.e., radial vs. tangential, 1,000 samples), and significance was reassured (p < 0.05). There was no main effect of retinotopic location at 6°, F(1, 6) = 3, p = 0.12, or at 15°, F(1, 6) = 2, p = 0.25. The Retinotopic location × Orientation interaction was significant at 6°, F(2, 7) = 21, p = 0.001—indicating a greater suppression for tangential than radial orientation at the 270° location—but not at 15°, F(2, 6) = 0.22, p = 0.75. Note that horizontal and vertical data for the 0° and 270° locations are reordered for radial and tangential data, whereas at 225° it is a different set of data. 

Suppression-ratio difference between parallel and orthogonal center–surround orientations

Since there is a horizontal orientation bias at 6° for parallel C-S configurations, we further explored if the orientation of the center grating is key for this observation. Figure 3 shows parallel and orthogonal suppression ratios when the center is horizontal (left panel) or vertical (right panel). A three-way ANOVA (3 retinotopic locations × 2 center orientations × 2 surround orientations) indicated main effects of center orientation, F(1, 4) = 109.7, p < 0.001, and surround orientation, F(1, 4) = 27.4, p = 0.006. As has been repeatedly shown in the literature, a parallel surround resulted in significantly more suppression than an orthogonal surround (Xing & Heeger, 2000). A new observation is that the horizontal center was susceptible to greater suppression (overall suppression in the left panel) than the vertical center regardless of the surround orientation (parallel or orthogonal). 

Figure 3

 

Mean suppression ratios for parallel and orthogonal center–surround gratings along horizontal, oblique, and vertical retinotopic locations for horizontal (left panel) and vertical (right panel) center orientations at 6°; error bars indicate 95% confidence interval of the mean.

Similarly, given that there is a radial orientation bias at 15°, we further explored if the orientation of the center grating is key to this observation. Figure 4 shows parallel and orthogonal suppression ratios when the center is radial (left panel) or tangential (right panel) to the visual-field meridian. A three-way ANOVA (3 retinotopic locations × 2 center orientations × 2 surround orientations) indicated, as expected, that parallel surrounds confer significantly greater suppression than orthogonal surrounds, F(1, 4) = 17, p = 0.015. Furthermore, Center × Surround interaction, F(1, 4) = 30.8, p = 0.005, indicated that the radial center was susceptible to greater suppression only if the surround was parallel (gray bars in left vs. right panels). As can be seen in the figure, yellow bars remain nearly unchanged in both panels. 

Figure 4

 

Mean suppression ratios for parallel and orthogonal center–surround gratings along horizontal, oblique, and vertical retinotopic locations for radial (left panel) and tangential (right panel) center orientations at 15°; error bars indicate 95% confidence interval of the mean.

In order to confirm that the distinct orientation biases we observed at 6° and 15° cannot be ascribed to the size or the spatial frequency scaling change according to the cortical magnification factor, we conducted an additional control experiment. The stimulus parameters used for 15° in the main experiment were presented at 6°, and the contrast-detection threshold with and without surround were estimated for two participants. The aim was to determine whether the bias at 6° switched from horizontal to radial when larger stimuli with lower spatial frequency were presented. Neither a radial nor a horizontal bias was observed under this condition (data not shown). 

Discussion

The main results of our study were twofold. First, horizontally oriented parallel C-S configurations resulted in greater suppression of the center than did vertical configurations at 6° eccentricity, regardless of visual-field location—these locations being nasal horizontal (0°), inferior oblique (225°), and inferior vertical (270°). Second, at 15° eccentricity radially oriented and parallel C-S configurations resulted in the greatest suppression. Although demonstration of general eccentricity dependence of orientation anisotropy is not novel, our report is the first of such effects for stimuli involving both center and surround mechanisms. 

Furthermore, we studied the role of center and surround orientations with relation to visual-field meridians. We found that a horizontal center grating is more susceptible to suppression than a vertical center at 6°. However, at 15° a radial center showed greater susceptibility to suppression than a tangential center, but only for the parallel surround condition. In addition, at 6° eccentricity if the surround orientation changed from parallel to orthogonal, suppression was released completely (Figure 3); but at 15°, suppression was retained even for an orthogonal surround (Figure 4). This is partly consistent with previous literature showing that facilitation (a suppression ratio below 1 in our study) disappears with increasing eccentricity (Xing & Heeger, 2000). Another novel observation here is that the orientation tuning effects of C-S interactions on contrast-detection ability depend not only on eccentricity but also on the relation of the center orientation to the specific visual-field meridian. 

Regarding the anisotropic effect of a surround on detection, how do our results fit with known properties of surround suppression? Surround suppression of contrast detection is usually weak in foveal vision but increases with eccentricity (Petrov, Carandini, & McKee, 2005; Snowden & Hammett, 1998). Here we show that surround suppression of contrast detection does not increase in a globally homogenous fashion in the visual field, but that instead there are variations that depend on the orientation of center and surround patches, retinal eccentricity, and the retinal meridian. Stimulus orientation has been studied extensively in order to understand excitation and inhibition of receptive fields, both classical (Ferster & Miller, 2000; Hubel & Wiesel, 1962) and extraclassical (Angelucci & Bressloff, 2006), of neurons in the visual system. It has been shown in cat (Gilbert & Wiesel, 1989) and monkey (Malach, Amir, Harel, & Grinvald, 1993) visual cortices that long-range connections selectively connect cells that have similar orientation preferences. In particular, long-range connections link neurons with co-oriented and coaxially aligned receptive fields. Bosking, Zhang, Schofield, and Fitzpatrick (1997) showed that neuronal axons of tree-shrew V1 extend for a greater distance along the axis of the receptive field than orthogonal to it. Although our study does not directly address mechanisms underlying surround suppression of contrast detection, it sheds some light on a mechanistic difference of suppression with eccentricity variation in human vision. Combining the anatomical evidence of a greater extent of horizontal connections along the radial axis of the receptive field in nonhuman primates with our observation that a radial orientation bias of surround suppression is seen only at 15°, we speculate that a greater involvement of long-range connections at 15° eccentricity relative to 6° may play a complex role. It is possible that at greater eccentricities, local inhibitory interneurons may be brought into play to a more significant extent. 

Though the aims and design of our study are different from those of Petrov and McKee (2006), who explored how the surround spatial layout with respect to visual field influences suppression strength at 6° parafoveal vision, it is nevertheless interesting to compare our findings to theirs. Petrov and McKee reported an increase of surround suppression of contrast detection for a radially oriented spatial layout compared to a tangential one, but for stimuli presented at 6°. They tested just two spatial layouts (collinear and flanking) in two visual-field meridians (vertical and horizontal) and found greater suppression strength for vertical-collinear and horizontal-flanking layouts compared to the other two. One should not be misled that the radial orientation bias we report here is completely different from Petrov and McKee's results, since they deliberately canceled the influence of grating orientation by presenting vertical gratings in all four conditions, whereas we tested the effect of different grating orientations in different visual-field locations. 

Despite wide differences in methodology, an interesting finding that aligns well with our study is the dependence of the orientation anisotropy on eccentricity that has been reported in some earlier studies in the absence of surround configurations. Psychophysical observations report a radial orientation bias outside the fovea both in human subjects (Rovamo et al., 1982; Sasaki et al., 2006; Westheimer, 2003a) and in animals (Sasaki et al., 2006). The bias for radial orientation has been observed in different ways. During development, retinal ganglion cells in the cat are added radially as the eye grows, and thus the dendritic fields tend to get stretched out radially (Leventhal & Schall, 1983). Further, there is a preponderance of cells tuned to radial orientations in the visual cortex of cats (Schall, Vitek, & Leventhal, 1986). A recent study also suggests that the thalamic afferents to the input layers in V1 in the macaque carry signals tuned largely to the radial orientation (Jayakumar, Mohan, Levichkina, & Vidyasagar, 2014). Though the radial bias may simply be an anatomical consequence of retinal development, it may nevertheless favor efficient processing of the optic flow of information, where the visual scene expands radially with self-motion in the environment (Lee, 1980). In fact, in one of the extrastriate areas of the cat (lateral suprasylvian area), more than half the cells respond vigorously to optic flow (Sherk, Mulligan, & Kim, 1997). 

In addition to fMRI studies, as mentioned in the Introduction, it has recently been shown that orientation information is decodable from MEG signals in humans (Cichy et al., 2015). That study found that signals generated from cardinal and oblique orientations were equally decodable from MEG responses centered over V1. The data suggested evidence for a radial bias, with orientation information from grating stimuli (radially unbalanced) being more readily classified than signals generated from spiral stimuli (radially balanced). Note that the results from fMRI and MEG are complimentary but not identical. MEG has much better temporal precision (but poorer spatial precision) than fMRI and avoids some of the problems of indirect extraction of neural activity from the BOLD signal. 

One striking result we obtained was the clear susceptibility of horizontal centers to suppression. This is consistent with a number of earlier studies using different methodologies. For example, Mannion et al. (2010) found horizontal stimuli to give the weakest BOLD response in fMRI. Our stronger suppression for horizontal centers surrounded by horizontal gratings than in the case of other orientations is also consistent with the finding by Essock, DeFord, Hansen, and Sinai (2003) that with natural scenes, sensitivity was poorest for horizontal orientations. It also follows their argument that with the divisive normalization supposed to be happening in more natural scenes, the pooled inhibition from the larger proportion of cells tuned to horizontal and vertical orientations (Coppola, Purves, McCoy, & Purves, 1998; Mansfield, 1974; Mansfield & Ronner, 1978; Orban & Kennedy, 1980) will also be greater. 

Another important point to mention here is that although our stimuli were briefly presented (150 ms), and the size of the stimulus patch was matched to that of Shushruth et al. (2013)—who studied orientation effects in surround suppression in V1—the precise contributions of retina, lateral geniculate nucleus, V1, and beyond to these psychophysical observations are not fully clear. One possibility is that the radial effect may be inherited from the retina (Leventhal & Schall, 1983), and surround suppression then added in V1 (Gilbert & Wiesel, 1990). In such a case, the mechanism for the orientation anisotropy for the suppressive component may be derived from feed-forward connections from lateral geniculate nucleus, long-range connections within V1, or feedback connections from extrastriate cortex (Angelucci & Bressloff, 2006). However, the relatively local (near surround) changes that we investigated here are less likely to be predominantly extrastriate (Shushruth et al., 2013). 

Another visual function that shows a radial/tangential anisotropy is visual crowding, which refers to the decreased recognition of a target due to the presence of surrounding flankers (Pelli, Palomares, & Majaj, 2004). Although visual crowding and surround suppression are thought to be distinct phenomena (Malavita, Vidyasagar, & McKendrick, 2017; Petrov, Popple, & McKee, 2007), they share some similarities. Of relevance here are observations from studies that show that the crowding zone is elongated along the radial axis rather than tangentially, since if flankers are placed along the radial axis, crowding increases more than if it is placed along the tangential axis at a given eccentricity, as shown in psychophysical (Greenwood, Szinte, Sayim, & Cavanagh, 2017; Nandy & Tjan, 2012; Toet & Levi, 1992) and fMRI (Kwon, Bao, Millin, & Tjan, 2014) studies. Similarly, our data show that surround suppression of contrast detection also shows a radial bias, although with some eccentricity limitations. The commonality of radial–tangential anisotropy between crowding and surround suppressions implies a common link between their neural mechanisms, possibly early in the visual pathway where orientation input is encoded. 

With regard to the horizontal orientation bias observed in our study and others, one interesting question is whether such cardinal orientation bias is a consequence of visual experience. Kittens with restricted visual experience for a particular orientation have been shown to have difficulties in perceiving lines along other orientations, and an increased number of neurons with optimal orientation the same as the orientation that they were exposed to (Blakemore & Cooper, 1970; Hirsch & Spinelli, 1970). Apart from complete visual deprivation, experimentally induced astigmatism in kittens has also been shown to lead to changes in cortical receptive fields (Freeman & Pettigrew, 1973), suggesting that overrepresentation of cardinal orientations in the environment may be the key factor in producing the preference for cardinal orientations in the cortex. Studies in human subjects have also supported the idea that the pervasiveness of vertical and horizontal contours in the visual environment may induce the development of the oblique effect (Baddeley & Hancock, 1991; Coppola et al., 1998; Hirsch, 1985; Switkes, Mayer, & Sloan, 1978). The occurrence of such biases in orientation discriminations of virtual lines (such as three circles arranged in a line) indicate that the effect is very likely not restricted to V1 (Li & Westheimer, 1997; Westheimer, 2003b). Furthermore, Bayesian modeling applied to data collected from Gabor-patch stimuli and photographs of the environment (Girshick, Landy, & Simoncelli, 2011; Tomassini, Morgan, & Solomon, 2010) has reported that human perception is significantly biased toward cardinal orientations when there is an increased uncertainty of orientation judgements. This suggests that a process of comparing an observer's internal model to the local distribution of stimulus orientations is used to fine-tune the orientation selectivity of the human visual system. 

Finally, we stress that caution needs to be exercised in interpreting our and other psychophysical results in relation to orientation anisotropies emerging from single-neuron and imaging studies. First, a psychophysical data point is the response of the human visual system as a whole and depends upon processing at multiple levels, especially on the final processing that occurs in extrastriate areas. Second, single-neuron studies and imaging studies themselves are not directly comparable. Not only do the imaging studies lack the spatial and often the temporal resolution of single-unit studies, they generally reflect more the presynaptic, synaptic, and dendritic activity rather than the action potential response that single-unit studies usually report (Logothetis, Pauls, Augath, Trinath, & Oeltermann, 2001). Thus, when there is good correspondence between psychophysical results and a particular physiological parameter, it may be reasonable to consider the neural responses as a possible correlate of the psychophysical phenomenon—an approach we have taken here. However, the lack of such correspondence does not necessarily invalidate one or the other of the studies, but points to the need to be aware of the additional processing occurring at different levels of the visual pathway. 

Conclusion

In conclusion, we found stronger surround suppression for horizontally oriented parallel center–surround gratings at 6° eccentricity, while there was a stronger suppression for radially oriented parallel center–surround gratings at 15° eccentricity. A change in the form of orientation anisotropy with eccentricity reflects a link between surround suppression and visual-field retinotopy. Further, a horizontal center was more susceptible to suppression than a vertical center at 6° regardless of the form of its surround (parallel or orthogonal), whereas at 15° a radial center was more susceptible to suppression than a tangential center, but only when the surround was parallel to the center. Our results demonstrate that the orientation-tuning properties of surround suppression of contrast detection depend on center orientation, retinal meridian, and eccentricity. 

Acknowledgments

This project was supported by ARC Discovery Project 140100157 (AMM), a Melbourne International Research Scholarship (MSM), and a Melbourne International Fee Remission Scholarship (MSM). 

Commercial relationships: none. 

Corresponding author: Allison M. McKendrick. 

Email: . 

Address: Department of Optometry and Vision Sciences, The University of Melbourne, Parkville, Victoria, Australia.