Filling-in along Horizontal and Vertical Meridians

Abstract

purpose. Filling-in is not homogeneous across the visual field; rather, fading time (FT) varies with eccentricity and polar angle. The reasons for this are unclear. The authors investigated FT along horizontal and vertical meridians in central and peripheral visual fields for luminance-defined targets, comparing Weber contrast sensitivity (CS_w). They also compared cone-mediated and rod-mediated filling-in with previously described photoreceptor densities. This represents the first investigation into rod-mediated filling-in.

methods. Fading times were recorded in nine volunteers using luminance-defined stimuli at eccentricities of 0°, 2.5°, 5°, 10°, and 20° for high-contrast stimuli and 0°, 1.25°, 2.5°, 3.75°, and 5° for low-contrast stimuli and were compared with CS_w at each location. Fading times were also recorded at 0°, 2.5°, 5°, 10°, and 20° using red-green stimuli and at 5°, 10°, and 20° under dark-adapted conditions and compared with cone and rod densities in the nasal, temporal, superior, and inferior hemimeridians.

results. Consistent anisotropy was evident in central and peripheral fields for luminance-defined targets and for chromatic targets, with horizontal meridians taking longer to fill in. The same pattern was observed in CS_w and the previously described cone density. Log FT and CS_w were correlated irrespective of visual field location. The ratio of rod-mediated FT to rod density decreased with eccentricity.

conclusions. Anisotropy between horizontal and vertical meridians for FT is consistent in central and peripheral fields, reflecting patterns in CS_w and cone density. This is discussed in the context of cortical processes underlying filling-in. For rod-mediated filling-in, more peripheral eccentricities are characterized by reduced FTs in relation to rod density.

Perceptual fading describes a filling-in phenomenon by which a peripheral target stabilized on the retina disappears after a certain fading time (FT). ^{1 2} Rather than the brain passively ignoring filled-in stimuli, evidence is accumulating that filling-in is an active cortical process through which a neural representation of the filled-in image is generated. ^{2 3 4 5} Filling-in is not homogeneous across the visual field. A number of studies show that FT decreases with increasing eccentricity of the target. ^{6 7 8} De Weerd et al. ⁷ suggest that this may be explained by cortical magnification. It has also recently been shown that FT varies with polar angle in the visual field, ⁹ with the horizontal meridian taking consistently longer to fill-in than the vertical meridian. The relationship between anisotropy and eccentricity has not been previously investigated, and the reason for anisotropy is unclear. We have investigated filling-in anisotropy along horizontal and vertical meridians of peripheral and central visual fields, comparing luminance-defined targets with detection thresholds. We also compared cone-mediated (red-green stimuli) and rod-mediated (dark-adapted conditions) filling-in with previously described photoreceptor distributions. ¹⁰ To our knowledge, this represents the first investigation into filling-in under dark-adapted conditions. 

The reasons for horizontal and vertical anisotropy of filling-in are unclear. However, Barrett et al. ¹¹ recently showed that the anisotropy is fixed to a viewer-centered (i.e., retinal-centered) rather than an environment-centered reference frame, suggesting an underlying hard-wired anatomic cause. One possibility is that because the vertical meridian uniquely corresponds to the intersection of left and right visual hemifields represented in opposite cortical hemispheres, filling-in of the vertical meridian requires interhemispheric transfer of visual information. This process has already been shown to make perceptual completion of illusory contours (another filling-in phenomenon) a more difficult task. ¹² Controversy exists concerning the degree of overlap of the hemifields along the vertical meridian, with some suggestion that greater overlap at the fovea could explain macular sparing in hemianopia. ¹³ If so, this could lead to differences in anisotropy between the central and peripheral visual fields. 

Another suggestion is that horizontal and vertical meridians have different cortical magnification factors, though there is yet no evidence for this anatomically. ¹⁴ Cortical magnification of the retinal surface is a multifactorial equation including photoreceptor density, ¹⁰ retinal ganglion cell/photoreceptor ratio, ¹⁵ and additional magnification in the retino-cortical pathway. ¹⁶ With respect to photoreceptor density, Curcio et al. ¹⁰ found that rod and cone densities in humans show anisotropy along meridians, with cone densities higher along the horizontal meridian but rod densities higher along the vertical meridian (within ±20° eccentricity). We investigated to what extent filling-in anisotropy is a reflection of photoreceptor distribution in the retina, given that photoreceptor densities contribute to cortical magnification. High-intensity isoluminant chromatic stimuli were used to investigate cone mediated filling-in, and low-luminance stimuli under dark-adapted conditions were used to investigate rod-mediated filling-in. 

Stürzel and Spillmann ¹⁷ have shown that stimulus salience is the key determinant of perceptual fading time. We test the hypothesis that differences in FT, with either eccentricity or polar angle, are caused by variations in sensitivity to the applied targets, such as Weber contrast sensitivity (CS_w). We have used a method to measure increment detection threshold (IDT) based on current automated perimetry techniques (where CS_w = 1/IDT). Since stability of fixation is known to influence FT, eye movements were also recorded in a subset of volunteers. 

Methods

Volunteers

Nine healthy volunteers, with no history or symptoms of ophthalmologic, otologic, or neurologic deficits, participated in the central and peripheral vision experiments (4 men, 5 women; mean age, 29.6 years; SD, 7.5 years). The same volunteers participated in the cone-mediated filling-in experiments, with four of the nine volunteers participating in the rod-mediated filling-in experiments with five other volunteers (4 men, 5 women; mean age, 33.2 years; SD, 6.8 years). All participants had corrected visual acuity of 20/20 or better for distance and N6 or better for near reading acuity, normal color vision using Ishihara plates, and binocularity of 120″ of arc or better using the TNO test. The experiments were performed under monocular best-corrected viewing of the dominant eye. The study received local ethics committee approval, was conducted with consent after the nature and possible consequences of the study were explained, and was performed in accordance with tenets of the Declaration of Helsinki. 

Visual Stimulus Generation

A calibrated visual stimulus generator video card (VSG 2/5; Cambridge Research Systems, Rochester, UK) with a DAC output resolution of 15 bits per gun was used to generate input to either a CRT monitor or an LCD projector in all trials. Each set-up was γ-corrected using a photometer and software (OptiCAL photometer; Cambridge Research Systems). Programs were written in Borland Delphi using the function library supporting the video card. Frame rates of 60 Hz were used to drive the video projector and 100 Hz to drive the CRT monitors. A resolution of 1024 × 768 pixels was used for all trials. Luminance readings were monitored using a radiometer (IL1700, R #106 radiance barrel, SEE038 detector; International Light, Newburyport, MA) with a photopic filter that matches the CIE V(λ) photopic curve to within 1% total area error. 

Central and Peripheral Vision Experiment

Measuring Fading Times.

A video projector (XGA Hitachi CP-X958; Hitachi, Maidenhead, UK) was used to project visual stimuli onto a rear projection screen, resulting in an image with dimensions 1.51 m × 1.12 m. The participant sat in a darkened room, 1.2 m in front of the screen, using a chin rest to stabilize the head position. The participant pressed a response box to record the FT. 

For measuring peripheral field FT, a 0.5°-diameter dark-gray spot (luminance, 9.14 cd·m⁻²) was projected on a mid-gray background (luminance, 12.10 cd·m⁻²; Weber contrast, −0.245) at eccentricities of 0°, 2.5°, 5°, 10°, and 20° along the temporal, nasal, superior, and inferior hemimeridians. For the central field, a lighter spot was used (luminance, 10.59 cd·m⁻²; Weber contrast, −0.125), projected at eccentricities of 0°, 1.25°, 2.5°, 3.75°, and 5° along the four hemimeridians. Eccentricities of 0°, 2.5°, and 5° in both experiments allowed comparison between experiments. 

A spot was displayed at each location, in random order, for up to 120 seconds. Thus, four spot locations were displayed along each hemimeridian, with an additional central spot trial, corresponding to 17 trials. The 17 trials were repeated 5 times on two separate days, resulting in 170 trials in all. The participant responded with a lever press upward when the dot was visible, followed by a return to the start position when the spot disappeared. At this point a new trial would begin, preceded by a 5-second rest period to allow time for the afterimage to fade. If the participant did not see the spot disappear before the end of the 120-second period, a “no filling-in” response was recorded. The participant fixated a red cross in all trials except for the central spot fixation trial. 

Measuring Weber Contrast Sensitivity.

A psychophysical test was devised to measure the participant’s increment detection threshold under conditions similar to the FT measurement. The test was based on the Humphrey Field Analyser (Carl Zeiss, Welvyn Garden City, UK) using a staircase method to measure static contrast perimetry. A CRT monitor (446XS; Nokia, Farnborough, UK) was used in favor of an LCD projection to provide greater resolution of contrast. Because of the curved surface of the monitor, screen locations of each spot were calculated experimentally with a flat plastic calibration film placed over the monitor. The participant was placed in a darkened enclosure, 33 cm from the screen, with the head stabilized on a chin rest. Screen background luminance was matched to that used in the filling-in trial (12.1 cd·m⁻²). A spot would randomly appear in 1 of 17 locations used in the FT delay measurement trial for a period of 200 ms. If the volunteer responded by pressing a mouse button (Bluetooth connection) within 1500 ms, the contrast of the spot would be lower when it next appeared. The contrast of each spot would decrease by a factor of \(\sqrt[3]{2}\) until it disappeared, after which point the contrast was increased by \(\sqrt[4]{2}\) until the spot appeared again. An error-checking algorithm was then performed, starting at the point at which the spot disappeared and increasing the contrast until the spot appeared again. This was repeated until two results were within \(\sqrt[4]{2}\) of each other. The average of the two measurements (expressed as Weber contrast = [luminance of figure − luminance of background]/luminance of background) was used to estimate the increment detection threshold. Weber contrast sensitivity is 1/increment detection threshold. All contrasts are expressed as absolute values rather than as negative contrasts, which result from using a dark spot on lighter background. 

Rod and Cone Vision Experiments

Measuring Fading Times.

For cone-mediated filling-in, a wide-screen CRT flat-screen monitor (GDM-FW900 Trinitron; Sony; Tokyo, Japan) was used to provide high-luminance chromatic stimuli. A red spot was placed on a luminance-matched green background (49.7 cd·m⁻²). Luminances were matched using the radiometer and photopic filter, with the red spot (diameter set to the width of radiance barrel) at maximum luminance and the green background adjusted until it matched the luminance of the red spot. For scotopic lighting conditions, volunteers were dark-adapted for 30 minutes in a completely blacked-out enclosure. The Sony monitor was covered with four high-strength neutral-density filters (299 1.2 ND filters; Lee Filters, Andover, UK). A blue filling-in spot of 2.62 cd·m⁻² was generated on the monitor (to maximize rod stimulation). Each filter allows 6% transmission of blue light (total transmission, 0.0013%), yielding a spot luminance of 3.4 × 10⁻⁵cd·m⁻². Consequently, the spot was only visible under dark-adapted conditions and was perceived as a white spot because only rods were stimulated. The background luminance of the monitor through the filters was 4 × 10⁻⁶ cd·m⁻², which was just visible under dark-adapted conditions. To allow stable fixation, a small incision was made into two of the filters to allow a low-intensity red-fixation cross to be visible under dark-adapted conditions. 

For cone-mediated filling-in experiments, a spot of 0.5° visual angle was projected randomly at eccentricities of 2.5°, 5°, 10°, and 20° along the four hemimeridians. Filling-in was also attempted at 0° under photopic conditions for up to 120 seconds. The 17 trials were repeated 5 times, resulting in 85 trials under each condition. The same protocol of timings was followed as for central and peripheral texture fading. For rod-mediated filling-in experiments, a spot of 2° visual angle was projected randomly at eccentricities of 5°, 10°, and 20° along the four hemimeridians. During dark adaptation, detection thresholds were recorded using a 2° spot randomly appearing at either 5° or 20° along the nasal hemimeridian for 200 ms. If the volunteer responded by pressing a mouse button within 1500 ms, the luminance of the spot would decrease by a factor of \(\sqrt[3]{2}\) . Otherwise the luminance increased by \(\sqrt[4]{2}\) . This allowed the luminance of the spot to be set below the level of the rod-cone break, ensuring rod stimulation only. 

Comparison with Photoreceptor Densities.

Human rod and cone densities described by Curcio et al. ¹⁰ were used to estimate the photoreceptor density at each FT spot location for cone-mediated and rod mediated filling-in. The equation given by Drasdo and Fowler ¹⁸ was used to convert internal retinal distance in millimeters to external visual angle in degrees. 

Eye Movement Recordings

To monitor differences in the stability of fixation when attempting to fixate the various target stimuli used in the filling-in trials, eye movements and blinks were recorded in a subset of volunteers (n = 4) using a high-resolution infrared pupil tracker at a sample rate of 250 Hz (EyeLink I; SensoMotoric Instruments GmbH, Berlin, Germany). Eye data were calibrated using a series of nine fixation points, projected individually in the shape of a 3 × 3 grid, ± 20° wide and ± 17.5° high. Otherwise, the experimental setup was similar to that used for the photopic and scotopic filling-in experiments. 

Three experiments were performed. In the first, given that different fixation patterns may result from viewing dissimilar targets during the central filling-in trial (when fixating a spot) and eccentric filling-in trials (when fixating a cross), fixating a spot (0.5° diameter, similar to the peripheral field filling-in experiment) for 30 seconds was compared with fixating a central cross for 30 seconds. This was repeated 5 times. In the second, to investigate the effect of having a distracting target near the fixation target or the filling-in target, eye movements were recorded during filling-in of a centrally fixated spot (0.5° diameter, similar to the central field filling-in experiment) and a peripheral spot (0.5° diameter, 10° eccentricity, similar to the peripheral field filling-in experiment), with and without the addition of a distracter (black cross 2.5° eccentric to the spot). Each condition was repeated 5 times, alternating conditions with and without the distracter. In the third experiment, eye movements were recorded from one dark-adapted volunteer when fixating targets at 5° and 10° (5 repeats). These were compared with filling-in during light conditions. 

Eye movement fixation stability was quantified by calculating the best-fit bivariate contour ellipse area (BCEA) ¹⁹ of the eye position data over time after data were removed during blinks and saccades. This represented the area of an ellipse that contained 68% of the data during fixation (1 SD). Data were corrected for any drift before the analysis. The bivariate contour ellipse was calculated using the equation:  

\[\mathrm{BCEA}\ {=}\ 2k{\pi}{\varsigma}_{\mathrm{H}}{\varsigma}_{\mathrm{V}}(1\ {-}\ {\rho}^{2})^{1/2}\]

where ς_H and ς_V are standard deviations of horizontal and vertical data, respectively, ρ is the product-moment correlation of horizontal and vertical data, and k is 1.14 (where P = 1 − e ^−k , with P the proportion of points within the ellipse, in this instance 0.68). The number of blinks per minute was also counted. 

Statistical Analysis

FT delay and CS_w were estimated for each hemimeridian by averaging measurements for 5°, 10°, and 20° eccentricities for peripheral-field testing and 1.25°, 2.5°, and 3.75° for central-field testing. For rod- and cone-mediated filling-in, FTs measured at 5°, 10°, and 20° eccentricities were added for each hemimeridian. Because of large intersubject variability of FTs, a within-subject design was used. Comparisons were made between horizontal and vertical meridians using a univariate general linear model, with participants introduced as random factors and mean FTs along the meridians as fixed factors (SPSS version 11; SPSS Inc., Chicago, IL). Nasal and temporal hemimeridians and superior and inferior hemimeridians were also compared using Bonferroni correction for multiple comparisons. 

Results

All FTs for one volunteer are shown as scatter plots in Figure 1for the conditions (A) full visual field, high-contrast stimuli, (B) central visual field, low-contrast stimuli, (C) chromatic red/green stimuli (cone-mediated filling-in), and (D) dark-adapted scotopic stimuli (rod-mediated filling-in). For conditions A to C, FT decreased with increasing eccentricity regardless of direction. In addition, reduced scatter of the points tended to be associated with increasing eccentricity. The volunteer could not fill-in the target at zero position (i.e., direct fixation) in the allotted time for chromatic stimuli (Fig. 1C)but could for conditions A and B. The presence of numerous outliers in the data, usually because of extended FTs, was assumed to result from blinks and unstable fixation, leading to resetting of the filling-in mechanism. For this reason median values were used as an average FT measure for each volunteer. Because median values of each volunteer approximated a normal distribution, parametric statistics were performed and means and standard errors (of median values) were used to describe the data. 

Mean FTs for all participants (± SEM) are compared with the mean CS_w measures (±SEM) for luminance-defined stimuli at various points across ±20° of the visual field in Figure 2and across the central ±5° of the visual field in Figure 3 . Fading time and CS_w were consistently higher along the horizontal meridian than the vertical meridian at equivalent eccentricities in the peripheral and the central fields (Figs. 2A 3A) . This was confirmed statistically, for FT (F = 31.8; P = 0.0005) and CS_w (F = 16.4; P = 0.004) of full-field, high-contrast stimuli (eccentricities 5°, 10°, and 20° summed) and for FT (F = 7.2; P = 0.03) and CS_w (F = 13.5; P = 0.01) of central field low-contrast stimuli (eccentricities 1.25°, 2.5°, and 3.75° summed). Three subjects could not fill-in the high-contrast target at 0°, and one subject could not fill-in any 2.5° target within the 120-second time limit. 

For central field, low-contrast stimuli, the anisotropy of FT was greater nearer the fovea (e.g., 1.25°). There was a dip at 0° eccentricity where FT was less than it was at adjacent 1.25° eccentricities. There were no significant differences between inferior and superior hemimeridians or between temporal and nasal hemimeridians for FT or CS_w (inferior versus superior, P = 1.0 and P = 1.0 for FT and P = 0.07 and P = 1.0 for CS_w; and temporal versus nasal, P = 1.0 and P = 1.0 for FT and P = 0.11 and P = 0.60 for CS_w, for full-field, high-contrast stimuli and central field, low-contrast stimuli, respectively). Log FT and CS_w were strongly correlated with full-field, high-contrast stimuli (Fig. 2B ; r ² = 0.93, P = 10⁻⁹) and central field low-contrast stimuli (Fig. 3B ; r ² = 0.84; P = 10⁻⁶). For full-field, high-contrast stimuli, the 0° central point was in accord with the trend for the rest of data, whereas for central field low-contrast stimuli, the point had a lower log FT than data with equivalent CS_w. This point was not included in the correlation. 

Points common to full-field and central-field experiments (2.5° and 5° eccentricities) show that anisotropy of horizontal and vertical fields for FT and CS_w were consistent for the two different levels of contrast tested. For example, mean FTs at 2.5° eccentricity, for horizontal and vertical meridians, respectively, were 35.5 seconds and 28.9 seconds for high-contrast stimuli and 13.7 seconds and 11.4 seconds for low-contrast stimuli. For 5° eccentricity, they were 18.4 seconds and 12.0 seconds for high-contrast stimuli and 10.5 and 8.7s for low-contrast stimuli. 

Comparison between cone-mediated FTs and cone density distributions described by Curcio et al. ¹⁰ are shown in Figure 4 . As for luminance-defined stimuli, consistent anisotropy between horizontal and vertical fields was evident, with the horizontal field taking longer to fill-in (F = 10.5; P = 0.01; eccentricities 5°, 10°, and 20° summed). Curcio et al. ¹⁰ also describe larger cone densities along the horizontal meridian compared with the vertical meridian at all eccentricities. Although mean FTs appeared elevated for the nasal field compared with the temporal field, there were no significant differences between temporal and nasal hemimeridians (P = 0.55) or between inferior and superior hemimeridians (P = 0.99). 

In a number of volunteers, filling-in of some central targets could not be accomplished under chromatic conditions in the allotted time. The number of participants in whom median FT values were greater than 120 seconds (where 120 seconds is the maximum duration of the target) was 8 subjects for 0° and 3 subjects for 2.5° temporal, 2 subjects for 2.5° nasal, and 2 subjects for 2.5° superior and 2 subjects for 2.5° inferior. Consequently, the FT at zero eccentricity is not plotted in Figure 4Aor correlated with cone density (Fig. 4B) . The relationship between FT and cone density was approximately linear, with a strong positive correlation (Fig. 4B ; r ² = 0.94; P = 10⁻⁸). 

Fading times were more even across the visual field under dark-adapted conditions compared with luminance-defined and chromatic stimuli (Fig. 5A) . There were no statistically significant differences between horizontal and vertical meridians (F = 1.03; P = 0.34) or between temporal and nasal hemimeridians (F = 4.2; P = 0.08); however, the inferior hemimeridian took significantly longer to fill-in than the superior hemimeridian (F = 11.4; P = 0.01). The relationship between FT and rod density was nonlinear (Fig. 5B) . The relationship between FT and rod density varied in a predictable way with eccentricity, as shown by Figure 5C , which demonstrates that the ratio of FT/rod density decreased with increasing eccentricity. Note that the x-axis of Figure 5Cis a logarithmic scale. The Weber contrast of the spot used during dark-adapted filling-in is shown in Figure 5Dfor one volunteer in relation to changes in the detection threshold over the course of the dark adaptation process. A rod-cone break is obvious for the 5° target and, to a lesser degree, for the 20° target, possibly because of the blue spot–stimulating S-cones in the peripheral retina. The contrast of the spot is below the rod-cone break and near the threshold of detection for the 5° target. However, all volunteers saw the appearance and slow fading of all targets presented at each retinal location. 

Stability of fixation, assessed using BCEA measurements of the eye movement recording data, was not considerably different when fixating either a spot or a cross (Fig. 6A) . Rather, intersubject variability was much higher than intrasubject variability between the two tasks. There were no obvious differences in BCEA or FT during central and peripheral filling-in with or without the addition of a distracter (Fig. 6B) . One subject showed large BCEAs when fixating a central target with or without the distracter (same subject) and also extended fading times (>50 s) compared with other volunteers fixating central targets. The same volunteer also showed long FTs compared with the other volunteers during the main experiments, when filling-in high- and low-contrast luminance-defined targets within the central 5°. However, FTs in the periphery during these trials were normal for this volunteer, as were cone-mediated and rod-mediated FTs. A separate statistical analysis was performed excluding this volunteer. Given that this had little effect on the outcome, data of this volunteer were left in the statistical analysis and graphical data. 

BCEAs recorded from one subject during filling-in of targets at 5° and 10° eccentricity (0.251 deg² and 0.226 deg ² , respectively) under dark-adapted conditions were similar to those recorded under light conditions (e.g., 0.258 deg ² when filling-in a target at 10°). 

Numbers of blinks recorded for the four volunteers were 7.6, 0.3, 1.6, and 2.2 blinks/min when directly viewing a spot and 7.7, 0.3, 3.3, and 0.9 blinks/min, respectively, when directly viewing a cross. During attempted filling-in of a central target, the effect of a distracter was to increase the number of blinks in three of the four subjects (1.2, 1.4, 4.9, 7.8 blinks/min increasing to 7.2, 0.3, 17.7, 14.0 blinks/min with the distracter). During attempted filling-in of a peripheral target, the effect of a distracter was less evident (4.1, 1.1, 0.0, 17.8 blinks/min increasing to 5.5, 0.0, 7.3, 9.9 blinks/min with the distracter). This shows that the effect of the distracter on blinks was related to being near the fixation target rather than the filling-in target. 

Discussion

We find anisotropy of filling-in to be consistent across central and peripheral visual fields for luminance-defined targets, with filling-in taking longer in the horizontal meridian than in the vertical meridian. This is related to differences in contrast sensitivity across horizontal and vertical meridians, where FT is exponentially related to CS_w for luminance-defined stimuli. The horizontal meridian also took longer to fade during cone-mediated filling-in, reflecting patterns in cone density previously described by Curcio et al. ¹⁰ We describe filling-in under dark-adapted conditions for the first time and find that rod-mediated FT is not directly related to rod density. 

Visual field anisotropy of filling-in has been previously described by Sakaguchi, ⁹ who found filling-in of stationary targets at 8° eccentricity took significantly longer along the horizontal meridian than along the vertical meridian. Barrett et al. ¹¹ also describe similar findings for a target at fixed eccentricity. We extend these findings by showing that this anisotropy is observed along the whole meridian and is present in central and peripheral visual fields. The reason for anisotropy of filling-in in the visual field is unclear. 

Perceptual fading was first described by Troxler (1804), ²⁰ who found that a small target presented in the peripheral field faded from view with steady fixation. Originally, this phenomenon was ascribed to local adaptation effects in the retina, ²¹ but perceptual fading has since been described for moving, ²² flickering, ²³ and textured backgrounds. ² Several lines of evidence support the view that filling-in is an active cortical process in which a neural representation of the completed visual stimulus is generated. Animal studies show responses in cortical neurons for visual stimuli causing filling-in related phenomena in humans, such as illusory contours induced through drifting bars and edges ²⁴ or brightness induction ²⁵ and bars extending over the blind spot. ²⁶ De Weerd et al. ³ describe neurons in monkey V2 and V3 cortex with gradually increasing activity that parallels the fading of scotomas in humans. Recently, using functional MRI in humans, Sasaki and Watanabe ⁴ have shown increased activity in area V1 during filling-in of a surface through neon color spreading. 

Sakaguchi ⁹ has suggested that horizontal/vertical anisotropy could be attributed to the uniqueness of the vertical meridian as intersecting right and left visual hemifields, thus requiring cortical interhemispheric communication. It is debated whether the degree of overlap of left and right visual hemifields increases in the region of the fovea and parafovea, ¹³ which could influence anisotropy. We find anisotropy of filling-in in the central field, which suggests either that overlapping of the hemifields does not influence horizontal/vertical anisotropy or that hemifield overlap is not significantly greater in the foveal region. 

An alternative suggestion to explain anisotropy, is that cortical magnification factors differ for horizontal and vertical meridians. ⁹ Most cortical magnification studies have not delineated between visual field quadrants, ^{27 28 29} though an early study by Daniel and Whitteridge ¹⁴ found no difference between horizontal and vertical fields. We show that anisotropy is evident for isoluminant chromatic red/green stimuli, resembling cone density patterns described by Curcio et al. ¹⁰ Photoreceptor density is related to cortical magnification through a complex relationship because cortical magnification also includes variation in retinal ganglion cell (RGC)/cone ratios across the retina. ¹⁵ There are no data, however, on the differences in RGC/cone ratio between horizontal and vertical meridians. Additional magnification of RGC inputs from the central visual field also occurs along the visual pathway after the retina. ¹⁶ Our data show that cone-mediated FT is surprisingly correlated to known distributions of cone density even with these additional factors. 

Numerous methods show that L/M cone ratios vary as a function of eccentricity, such as flicker photometry, ³⁰ high-resolution optical imaging of the retina, ³¹ and multifocal ERG. ³² The same methods show that L/M cone ratios vary greatly between persons. It is difficult to understand why we find strong linearity between FT and cone density given these variations. Interestingly, Hood et al. ³³ find that visual-evoked cortical response to L- and M-cone modulation show less variation than retinal L/M cone ratios, suggesting a gain modulation before the signal reaches the cortex. In addition, using ERG and psychophysics, Kremers et al. ³⁰ find that L/M ratios are strongly influenced by adaptation for the luminance channel but not for the chromatic channel, indicating the presence of a compensatory mechanism. These postreceptoral mechanisms may contribute to the surprising correlation between FT and cone density despite variations in L/M cone ratios across the retina and between subjects. 

One proposed scheme for perceptual fading is a two-stage process consisting of slow adaptation of the boundary separating the figure from the background and fast interpolation through which the target is invaded by the background. ³⁴ For example, De Weerd et al. ⁷ highlighted the importance of the figure boundary to FT, showing that FT is linearly related to the boundary length represented on lower-order visual cortex. In the present study we show that FT is exponentially related to CS_w for the target. Using the proposed scheme, the relation between FT and CS_w can be explained by the stronger edge representation brought by higher contrast sensitivity. There are some limitations to this scheme, however, such as experimental reports that the neural interpolation of filling-in begins substantially before the edge representation collapses. ³ Sakaguchi ³⁵ has shown that luminance, orientation, and color all have asymmetric effects on FT when features of the target and surround are switched, indicating that FT is not simply determined by the strength of the represented figure edge but by more complex visual features of target and surround. 

The current findings represent the first systematic investigation into filling-in under dark-adapted conditions. As they did in perceptual fading under photopic conditions, all volunteers described the slow fading of the target after a certain delay. However, unlike cone-mediated filling-in, rod-mediated FT is not directly related to photoreceptor density. Rather FT/rod density varies as a function of eccentricity for each hemifield with more peripheral eccentricities showing reduced FTs in relation to rod density. In addition, FT asymmetries along the meridians do not reflect those seen in rod density. 

Volunteers accomplished filling-in of targets using the central visual field when directly foveating on targets under photopic conditions. When directly viewing the higher contrast target, FT matched that predicted from CS_w measurements across the visual field. In contrast, when volunteers viewed a low-contrast target, FT for direct fixation was less than expected. Eye movement recordings suggest that this was not caused by stability of fixation but that it might have resulted from the greater blinking that occurred when targets were immediately adjacent to the point of fixation, perhaps because higher concentration levels were required. 

In summary, we show that anisotropy of filling-in for horizontal and vertical meridians is a consistent feature of the central and peripheral fields for luminance-defined stimuli, which is related to differences in contrast sensitivity. Anisotropy of filling-in was also evident for chromatic stimuli, reflecting cone distributions across the retina. The relationship between cone-mediated FT and cone density was surprising given the known variations in L/M ratios across the retina and the complex relationship between cone density and cortical magnification. We also describe filling-in under dark-adapted conditions for the first time and find FT is not directly related to rod density. 

 Supported by the Ulverscroft Foundation.

 Submitted for publication February 24, 2005; revised July 19, August 30, and September 14, 2005; accepted November 16, 2005.

 Disclosure: F.A. Proudlock, None; A. Khanna, None; I. Gottlob, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

FTs for one volunteer across horizontal and vertical meridians for each of the four conditions.

Figure 2.

 

(A) Mean FT (±SEM) plotted with CS_w for high-contrast stimuli applied at various horizontal and vertical eccentricities between ±20°. CS_w = 1/ICD (where ICD is increment detection threshold expressed as a Weber contrast ratio). (B) Strong linear relationship between log₁₀ fading time and CS_w independent of position in visual field (r ² = 0.93; P = 10⁻⁹; log FT = 0.019 CS_w + 0.66). Key indicates the eccentricity and hemimeridian of each point.

Figure 3.

 

(A) Mean FT (±SEM) plotted with CS_w for low-contrast stimuli applied at various horizontal and vertical eccentricities between ±5°. CS_w = 1/ICD (where ICD is increment detection threshold expressed as a Weber contrast ratio). (B) Relationship between log₁₀ FT and CS_w (r ² = 0.84; P = 10⁻⁶; log FT = 0.011 CS_w + 0.064). Key indicates the eccentricity and hemimeridian of each point.

Figure 4.

 

(A) Mean cone-mediated FT (±SEM) plotted with log₁₀ cone densities (from Curcio et al. ¹⁰ ). (B) Fading time and cone density are strongly correlated (r ² = 0.94; P = 10⁻⁸). Key indicates the eccentricity and hemimeridian of each point.

Figure 5.

 

(A) Mean rod-mediated FT (±SEM) plotted with log₁₀ rod densities (from Curcio et al. ¹⁰ ). (B) FT plotted against rod density. (C) Change in ratio of FT/rod density with eccentricity for each hemimeridian. Key indicates the eccentricity and hemimeridian of each point. (D) Change in detection thresholds of spots at 5° and 20° eccentricity for one subject over the course of dark adaptation. Lines are smoothed, averaging data for each point over the time course of a minute. (Arrow) Rod-cone break. (Dashed line) Contrast used for the filling-in spot.

Figure 6.

 

BCEA measurements of eye movement recording data used to assess fixation stability. (A) Comparison between fixating a 0.5°-diameter spot or a cross 0.5° × 0.5° (n = 4). (B) FTs and BCEAs for central and peripheral (10° eccentricity) filling-in with or without an additional distracter (2.5° from the spot) (n = 4).