Is flicker-defined form (FDF) dependent on the contour?

Abstract

Flicker-defined form (FDF) is a temporally driven illusion within which randomly positioned background elements are flickered in counterphase to stimulus elements, creating the illusion of a contour in the region between the background and the stimulus dots. It has been proposed that FDF is dependent on the boundary region between the counterphase flickering dots. Is the stimulus area or the illusory contour itself (region between stimulus and background) paramount to the FDF percept? Circular stimuli were compared to ring stimuli to determine the relative importance of area and contour. The rings were tested in the following configurations: constant maximum diameter/variable area; constant area/variable contour; and constant contour/variable area. For rings with constant diameter, no effect of ring thickness was found. No effect of contour was found for rings of a constant area. For rings of constant contour, the smaller the area the greater the threshold. These results suggest a greater dependence on the area of a stimulus rather than its contour. Area dependence suggests that the theory of contour dependence by a fast extraction system is unlikely. This temporally defined magnocellular-dependent illusion is influenced by slow surface perception mechanisms of the parvocellular system.

Introduction

Do edges define figure versus ground when the edge itself is illusory? The phantom contour illusion, more recently referred to as flicker-defined form (FDF) (Quaid & Flanagan, 2005a), has been shown to rely heavily on the border between two regions of dots flickering in counterphase in order to create the percept of a figure (Rogers-Ramachandran & Ramachandran, 1998). Other studies have found that form-extraction information can be perceived with delays of 5 ms between figure and ground. This delay is consistent even when temporal frequencies are modified between 1.3 and 30 Hz (Fahle, 1993).Temporal synchrony, without spatial cues, is sufficient to elicit perception of a salient edge (Lee & Blake, 1999; Usher & Donnelly, 1998). 

Although temporal characteristics are enough to elicit perception, spatial characteristics can affect perception of temporally defined stimuli. The number of random dots per degree of visual space, which define the FDF illusion, affects the perception of the illusion (Quaid & Flanagan, 2005a). More specifically, the spatial content percentage (k), has been related to FDF perception. Spatial content percentage is the product of the area of individual dots, the number of dots within a given stimulus and the area of the stimulus. Thus, k, accounts for the area within the stimulus boundary which is flickering out of phase. This spatial content dependent effect has been found in other temporally defined stimuli (Lee & Blake, 1999). This may be due to a completion of the border creating a higher spatial frequency illusory edge, or it may simply be an area dependent effect, i.e., due to spatial summation. A greater number of random dots within the stimulus would give a higher spatial content for the stimulus, which gives an increased area of flicker, even though the area of the stimulus remains constant. This would still mean that more receptors would be activated. 

Although the spatial content percentage is important to FDF perception, a plateau for target size was previously reported (Quaid & Flanagan, 2005a). This is similar to the results found for temporally driven stimuli other than FDF, with a number of studies having shown that only small target sizes affect contrast thresholds (Mäkelä, Rovamo, & Whitaker, 1994; Tyler & Silverman, 1983). Medium and larger target sizes show no effect on contrast thresholds at eccentricities away from fixation. These results suggest that flickering stimuli are processed differently than static luminance-defined targets which are subject to principles of spatial summation at all eccentricities. 

We can compare FDF to other stimuli that share specific characteristics such as flicker (Mäkelä et al., 1994; Tyler & Silverman, 1983) for the similar temporal dynamics; and form-from-motion, for the ability to perceive a shape due to dynamic elements (Schoenfeld et al., 2003). Recent studies have shown that the visual system is sensitive to temporal synchrony (Lee & Blake, 1999; Usher & Donnelly, 1998) and temporal structure (Guttman, Gilroy, & Blake, 2007), particularly for contour binding (Bex, Simmers, & Dakin, 2001), but whether these systems use similar mechanisms is still unknown. Although some of these stimuli seem to rely on magnocellular and/or dorsal stream mechanisms, form-from-motion stimuli seem to be reliant on the interaction between the two streams and can be imperceivable even when motion and form perception are intact (Cowey & Vaina, 2000; Schenk & Zihl, 1997). In contrast, flicker perception is primarily dependent on the magnocellular system (Livingstone & Hubel, 1987). We believe that FDF is distinct from these stimuli because of the lack of perceivable temporal dynamics, as exemplified in the inability to perceive the surface phases when the border between the patches is covered (Rogers-Ramachandran & Ramachandran, 1998). 

The most similar stimulus to FDF is Lee and Blake's (1999) stimulus, which has no discernable temporal structure, but allows perception of shapes. According to Blake and Lee and others (Usher & Donnelly, 1998), stochastic (lacking structure) temporal structure is processed very efficiently by the human visual system. Thus, even this stimulus that provides no obvious cues to temporal structure is significantly different from FDF. 

Flicker-defined form is believed to be a predominantly magnocellular-based stimulus due to its dependence on high temporal frequencies, its perceived low spatial frequency (Goren, Quaid, & Flanagan, 2005), and its resistance to optical blur (Quaid & Flanagan, 2005b). The illusion can tolerate decreases in stimulus size and is enhanced by peripheral viewing (Quaid & Flanagan, 2005a; Rogers-Ramachandran & Ramachandran, 1998). Flicker-defined form thresholds have been shown to be determined by both the number of dots within the stimulus and the stimulus diameter (Quaid & Flanagan, 2005a). However, it is not understood whether the stimulus area or the border of the stimulus itself is the most important component of the illusion. Previous findings showed that when the region between the phase shifted random dots was covered, the two surfaces could not be distinguished, which is why the region in between the dots, referred to here as the contour, is believed to be the most important component. The importance of this contour was the basis for Rogers-Ramachandran's theory that this illusion is controlled by a fast-acting contour extraction system. This system was believed to be the magnocellular system. The current study aims to determine whether the contour is the most important component for perception of the illusion. If area is found to be equally or even more important, this would suggest that the fast-acting contour perception would not be the mechanism to explain FDF perception. 

Methods

The sample consisted of two sets of 3 subjects. The first set was aged 23, 23, and 31 years old, one female and two males. The second set consisted of two females (aged 24 and 27) and one male (aged 24). All subjects were naive to the purpose of the experiment. All subjects had normal, corrected vision (6/6 or better) and no known ocular abnormalities. Subjects viewed the stimulus with their right eye. 

Subjects were seated 32 cm from a Sony Trinitron monitor 20″ (Multiscan CPD-G500) using a resolution of 1024 × 768 pixels and a pixel pitch of 0.37 mm. At this distance, the monitor subtended 61.8 × 48.3 degrees. The refresh rate was 100 Hz. Luminance values ranged from 0 to 100 cd/m ². Twenty-three contrast values were presented, using this range of luminance values. Stimuli were generated on a PC running custom software in a LINUX-based environment. 

Stimuli's contrast values were measured in log Michelson units.  

$\frac{L_{\max }-L_{\min }}{L_{\max }+L_{\min }}.$

(1)

 

Stimulus

Dots subtending 0.34° were generated at random locations throughout the screen and were flickered at a temporal frequency of 16.67 Hz. Background elements were flickered in counterphase to stimulus elements, creating the illusion of a contour in the region between background and stimulus (Quaid & Flanagan, 2005a; Rogers-Ramachandran & Ramachandran, 1998). Stimulus dots were defined as those that fell within the boundary of the stimulus area. If a dot was overlapping this boundary, the percentage of area that fell within the boundary determined whether the dot was classified as stimulus or background. A dot that was 50% or more within the stimulus area was determined to be a stimulus dot. All dots were defined as either stimulus or background. This means that the boundary itself could potentially contain a portion of the background and/or stimulus dots. 

The current study was designed to determine the relative importance of the boundary between stimulus and background, as opposed to the amount of stimulus area. Circles and rings were used in all of the following experiments in order to ensure that all areas of the stimulus were equally salient. Using squares, gratings, or other shapes with obvious corners would have created particular regions that were more salient, particularly for the illusory contour. A second reason for choosing circles was a greater ease of manipulation of contour versus area. Using more complex stimuli such as gratings would make manipulation of contour, independent of area, significantly more difficult. 

Stimulus size, amount of contour and area were modified, along with ring thickness in order to determine FDF's dependence on spatial characteristics. Stimulus size is defined by the outer diameter of the stimulus. Stimulus size is not affected by increasing contour within its diameter. For example, Figure 1 shows the manipulations for Experiment 1 in which the stimulus size was not altered. Stimulus area was defined as the area between the inner and outer diameters of the rings. This area encompasses the locations where the stimulus random dots were found. During these experiments, the density of the dots was not manipulated, so the spatial content percentage did not change. The amount of contour in the stimulus was defined as the sum of the inner and outer contour that defined the ring. 

Figure 1

 

Stimulus structure. This figure shows the difference between stimulus and background elements. Note that the entire image would be covered with randomly located dots. Panels A and C are the background regions. The elements bounded by these regions are in phase with each other and are in counterphase to elements within the B region (the stimulus region). Stimulus area is the area of panel B. Contour is the length of the dashed line (i.e., the sum of the inner and outer contour).

Constant stimulus diameter-varying ring thickness

The first set of experiments used a set of ring stimuli within which stimulus size was kept at a constant 5° diameter, but the inner diameter was varied to give ring thicknesses of Ring I (2°), Ring II (1.5°), Ring III (1°), and Ring IV (0.5°) ( Figure 2). The stimulus size and spatial content percentage were kept constant while the area of the stimulus was modulated. Overall contour length of the stimulus increased as the diameter of the inner ring was increased, as even though the outer portion of the contour was unchanged, the inner portion was being modified. Thus, contour and area were negatively related in this experiment. 

Figure 2

 

Ring stimuli for experiment 1. Stimuli were of a 5° diameter, the inner circle was of diameter 0°, 1°, 2°, 3°, and 4° with corresponding ring thicknesses of 0°, Ring I (2°), Ring II (1.5°), Ring III (1°), and Ring IV (0.5°). Inner circle flickered in phase with background elements and in counterphase to elements within the ring.

Constant ring area-varying stimulus contour

The second set of experiments involved using a constant area equivalent to a 5° diameter circle (19.6 degrees ²), while modifying the amount of contour ( Figure 3). Five different stimuli were created with contours: Contour I (29.3°), Contour II (44.8°), Contour III (58.6°), Contour IV (72°), and Contour V (85.1°). These contours had the same areas and spatial content percentage but varied in their stimulus size. The outer diameters, defining stimulus size were: 6°, 8°, 10°, 12°, and 14°. 

Figure 3

 

Constant area stimuli, variable contour. Stimuli were of variable diameter of contour, but the amount of area (within the ring) was constant. Five different stimuli were created with contours: Contour I (29.3°), Contour II (44.8°), Contour III (58.6°), Contour IV (72°) and Contour V (85.1°). Inner circle flickers in phase with background elements and in counterphase to elements within the ring.

Constant ring contour-varying stimulus area

The third set of experiments involved keeping the amount of contour constant and varying the area ( Figure 4). For this condition, the amount of contour was matched to the contour of the 5° diameter circular stimulus (15.7°). Four different areas were tested: Area I (3.1 degrees ²), Area II (6.3 degrees ²), Area III (9.4 degrees ²), and Area IV (12.6 degrees ²). In order for the contour to remain constant and the area to change, the overall stimulus size, defined by the overall diameter of the stimulus, was modified. For this experiment, the outer stimulus diameters were 2.9°, 3.3°, 3.7°, and 4.1°. 

Figure 4

 

Constant contour stimuli, variable area. Stimuli were of variable total size and ring area, but the amount of contour (outer and inner circle combined) was constant. Four different areas were tested: Area I (3.1 degrees ²), Area II (6.3 degrees ²), Area III (9.4 degrees ²) and Area IV (12.6 degrees ²). Inner circle flickers in phase with background elements and in counterphase to elements within the ring.

Protocol

Subjects were instructed to fixate on a red dot at the center of the screen that was present at all times during experiments. Stimuli were presented for 720 ms (160 ms ramp up, 400 ms constant contrast, 160 ms ramp down). In a yes/no detection task, subjects indicated with a button press their ability to perceive the stimulus. In order to improve the variability of results, false-positives were recorded. False-positives were measured as responses that occurred within 180 ms of the stimulus ramp onset or for within the final 1 second of the interstimulus interval. Most subjects indicated a response while the stimulus was being presented, so a delayed response would typically indicate a response where there was no stimulus (i.e., before a presentation). For false-positives, any trials with more than 20% false-positives were excluded (Bayer & Erb, 2002). The exclusion of trials was very rare, as subjects were practiced observers. 

Contrast thresholds were estimated using a 4-2-1 staircase procedure with 2 reversals at the final crossing and recorded using log Michelson contrast units. Twenty-three log steps of contrast values ranged from −0.3 to 2.0. In most cases where thresholds of 1.70 or greater are reported, they reflect an inability to perceive the stimulus due to upper limit constraints. Specifically, the contrast difference between the stimulus and the background dots was unperceivable above a contrast of 1.70 log Michelson units. For all experiments, the effect of ring thickness was tested at each of four eccentricities (0°, 4.2°, 12.7°, 21.2°) within the inferior temporal quadrant along the 45° meridian, which correspond to stimulus locations of 3° × 3°, 9° × 9°, and 15° × 15°. All subjects were tested three times for each stimulus condition (ring thickness, location). 

Analysis

Mean and standard errors were calculated. Repeated measures ANOVA was used to determine the effect of different eccentricities, ring thickness (related to area of the stimulus), contour, and size and for any interaction between the variables. Tukey's honest significant difference (HSD) post hoc analysis was performed where significant effects of ANOVA were found. 

Results

Constant stimulus diameter-varying ring thickness

For this experiment, the stimulus size was kept constant (i.e., 5° diameter) while thickness of the ring was varied. Effects of ring thickness were not statistically significant ( F(4,8) = 1.16, p = 0.40, power = 0.22, Figure 5). 

Figure 5

 

Effect of ring thickness on detection thresholds. Stimulus size of 5° for all stimuli, with ring thicknesses of Ring I (2°), Ring II (1.5°), Ring III (1°), Ring IV (0.5°) and solid circle (thickness of 2.5°) at various eccentricities (indicated in upper left corner). Mean and standard error for each of 3 individual subjects is shown. Different symbols indicate the 3 subjects.

No significant eccentricity-dependent effects were found, although trends suggest that thresholds decreased with increasing eccentricity ( F(3,6) = 2.62, p = 0.15, power = 0.38). These trends could be seen in all three subjects with thresholds for 12.7° and 21.2° being lower than those found at 0° and 4.2°. The magnitude of this change varied by thickness of the ring and by subject. However, the interaction effects of ring thickness and eccentricity were not statistically significant ( F(12,24) = 0.95, p = 0.52, power = 0.39). The apparent trends revealed increasing thresholds with decreasing ring thickness, which were more pronounced at fixation. 

Constant ring area-varying ring contour

In this experiment, the amount of area (i.e., of the ring) was kept constant while the amount of contour (the sum of the outside and inside of the ring) was varied. The stimulus diameter changed. No significant variation in threshold was found for the effect of modified contour ( Figure 6, F(4,8) = 1.65, p = 0.25, power = 0.31) or the interaction between contour and eccentricity ( F(12,24) = 2.02, p = 0.069, power = 0.78). 

Figure 6

 

Effect of ring contour on detection thresholds. Stimulus size and area of the stimulus (area flickering in counterphase to background) were constant. Amount of contour created by the counterphase flickering dots was varied. Five different stimuli were created with contours: Contour I (29.3°), Contour II (44.8°), Contour III (58.6°), Contour IV (72°), and Contour V (85.1°). Mean and standard error for 3 subjects are shown. Different symbols indicate the 3 subjects.

Constant ring contour-varying ring area

In this experiment, the amount of contour (i.e., the sum of the outside and inside of the ring) was kept constant while the area of the stimulus was varied. Four areas were tested, which will further be referred to by their ranking: Area I (3.1°), Area II (6.3°), Area III (9.4°), and Area IV (12.6°). Due to a large amount of variability, between subjects at fixation, these data were excluded from the statistical analysis. When it is included, similar effects of area and area-eccentricity interactions are found. Changes in area were found to affect thresholds in an eccentricity-dependent manner ( Figure 7). Significant differences were found between stimuli of different areas ( F(3,6) = 211.97, p < 0.001). The interaction between area and eccentricity was also significant ( F(6,12) = 15.04, p < 0.001), meaning that the effect of area was significantly dependent upon the eccentricity in question. In particular, variation was most prominent for further eccentricities. 

Figure 7

 

Effect of ring area on detection thresholds. Stimulus size and amount of area created by the counterphase flickering dots were constant. Area of the stimulus (area flickering in counterphase to background) was varied. Four different areas were tested: Area I (3.1 degrees ²), Area II (6.3 degrees ²), Area III (9.4 degrees ²) and Area IV (12.6 degrees ²). Mean and standard error for each of 3 individual subjects were shown. Eccentricity is indicated in the upper left corner.

Larger stimuli had lower thresholds (as one might expect). All 3 subjects showed decreases in threshold with increasing area of the stimulus. Post hoc analysis showed that smaller stimuli have higher thresholds than larger stimuli. Area I had significantly higher thresholds (mean: 1.69 ± 0.01) than Area II (0.95 ± 0.06), Area III (1.05 ± 0.04) and Area IV (0.94 ± 0.03) at 4.2° ( p < 0.001). At 12.7°, Area I still had significantly higher thresholds than the two largest rings ( p < 0.05). This effect is also seen at 21.2° ( p < 0.01). 

Thresholds decreased with increasing eccentricity, which is in agreement with the observation that the contour's visibility improves with increasing eccentricity (Quaid & Flanagan, 2005a; Rogers-Ramachandran & Ramachandran, 1998). The Area I stimulus thresholds were significantly lower at 12.7° (mean:1.17 ± 0.05) and 21.2 (mean: 1.37 ± 0.07) versus 4.2° (mean: 1.69 ± 0.01, p < 0.01). Area I showed higher thresholds than Area III and IV at 4.2°, 12.7°, and 21.2° (p < 0.05). Area II, III, and IV showed no improvements with increasing eccentricity. In summary, smaller target sizes give improved sensitivity with increasing eccentricity, in some cases making centrally imperceptible stimuli, perceptible. 

Interaction between ring thickness and overall diameter

An analysis of the data was performed to determine the role of overall diameter in the perception of thin rings. Figure 8 shows a comparison of the thinnest ring in the first experiment to the thinnest ring in the second experiment. Both rings had thicknesses of approximately 0.5°. One of the circles (from Experiment 1) had an overall diameter of 5°, whereas the second had an overall diameter of 14°. Thresholds are lower for the larger stimulus, although this difference is small with large variability (average difference of <0.33 ± 0.08) and not statistically significant ( F(1,2) = 9.66, p = 0.09, power = 0.41).The majority of significant findings concerning FDF has pointed to the eccentricity dependence of different effects (e.g., area, contour). In this study, area manipulations have effects, which depend on the eccentricity of the target. Specifically, if effects of overall diameter were found that might influence the lack of contour manipulation effects, then we would expect them to also be affected by eccentricity, which they are not ( F(3,6) = 0.83, p = 0.52, power = 0.15). 

Figure 8

 

Effect of increasing ring diameter, while maintaining ring thickness. This graph shows averaged results across 3 subjects for rings of 0.5 degrees. The smaller ring has an overall diameter of 5 degrees, whereas the larger has a diameter of 14 degrees.

Discussion

In order to perceive the phantom contour illusion (i.e., flicker-defined form), it has been proposed that the dominant feature is the boundary between the counterphase flickering dot regions. Previous work suggested that the density of dots and the spatial content percentage affect perception of the illusion (Quaid & Flanagan, 2005a). This study aimed to determine whether the boundary is the most important component of the stimulus. The importance of the contour to FDF perception was used to support the fast contour extracting system theory, which is most likely the magnocellular system (Rogers-Ramachandran & Ramachandran, 1998). A dependence on the area of the target would suggest that although the surfaces cannot be distinguished, they play an important role in perception. Inability to distinguish phases means that if the surface plays a role in extracting the FDF contour, it alone, is insufficient. This means that the boundary relying on the contour extraction system must add additional cues. The first experiment intended to show how reducing the area of the stimulus while maintaining equal stimulus size affects perception. The second experiment maintained a constant area, while the amount of contour was modified. The third experiment maintained a constant amount of contour while the amount of area was modified. 

Importance of area

To determine whether the lower thresholds that were found for solid circles, as compared to a ring, were simply due to the presence of a greater out-of-phase signal, an experiment was conducted in which the amount of contour was kept constant. This experiment showed that changing the amount of area had a significant effect, even though the total amount of stimulus contour was kept constant. As area was decreased, performance degraded. 

The effects of area were dependent on eccentricity. At further eccentricities, the effect of area was more pronounced, i.e., improvements were greater farther from fixation. This was due to the illusion being more difficult to perceive at fixation (Quaid & Flanagan, 2005a), which meant that none of the sizes produced large improvements in thresholds. As mentioned in the Methods section, the current study limited the area and/or contour to that of the baseline circle. Further into the periphery, where the illusion is easier to perceive, the addition of area was initially (at small sizes) beneficial. Although, this improvement soon plateaus because the additional area was not as beneficial, similar to findings of other studies (Mäkelä et al., 1994; Tyler & Silverman, 1983). 

Importance of contour

The finding that the thinner the ring, the worse the perception was surprising if one espouses the view that FDF is dependent on the presence of a boundary between the counterphase flickering dots. If the boundary is important, then creating more “contour” should also improve performance. The data suggest that the presence of additional contour, while keeping the entire stimulus size constant, had no benefit to the perception of the illusion. 

A number of explanations can be postulated for this finding. First, it is possible that contour is less important than area in order to perceive FDF. Results show that when the area is decreased, the stimulus was harder to see, in spite of the presence of equivalent contour. 

A second, more likely explanation is that either sufficient contour or area, once suprathreshold, allows perception of the stimulus. The presence of either sufficient contour (e.g., a very thin ring which contains minimal area, but large amounts of contour) or area (e.g., a solid circle with high amounts of area but less contour than a ring which contains both internal and external contour), even if the other is below threshold, would still signal the FDF percept. The amount of contour may not be important, provided there was sufficient area flickering out of phase. Thus, even if the amount of contour was insufficient, the presence of a suprathreshold area may have allowed the stimulus to be seen. If the area was insufficient, adding additional contour maybe meaningless. This concept is in keeping with other studies which find that temporal cues can allow a stimulus to be perceived with greater ease than spatial summation alone can predict (Lee & Blake, 2001). 

The third possibility is that the stimuli used for contour manipulations in the current study were all suprathreshold. All of the experiments in which contour was manipulated gave very low contrast thresholds. This means that in the most difficult viewing conditions (small amounts of contour at low contrast) subjects were still able to detect the stimulus with ease. Thus, adding additional contour had no benefit as the smallest amount of contour was sufficient to allow perception. 

The fourth possibility is that aliasing between the inner and the outer contour may eliminate the benefit of additional contour. Previous unpublished data suggest that the illusory contour can be perceptually matched to a 2-cpd grating (Goren et al., 2005). Some of the contour-manipulated stimuli were less than half of a degree thick, which means that the inner and the outer contour of approximately 2 cpd perceptually overlap and combine into one contour, removing the benefit of inner and outer contour. 

The interaction between area, contour, and overall stimulus size

The current study aimed to determine which was more important to FDF perception: area or contour. It seemed that threshold changed with area, but not with contour. But could the changes in overall stimulus size (i.e., outer diameter) have affected results? 

A confounding factor to be considered in the second and the third set of experiments was the changing size of the stimulus. According to an earlier FDF study (Quaid & Flanagan, 2005a), circular stimuli of area 12.6°–28.3° (diameters 4°–6°) do not vary significantly in performance between fixation and 21.2°. Stimuli of diameter 4°–6° were easily detectible, when a standard density of background random dots was used. According to the Quaid and Flanagan (2005a) study, the most important determining factor for FDF perception was a constant “k,” which was dependent on the area of the stimulus and the density of the dots. 

Area-manipulated stimuli in the current study were smaller in stimulus diameter and area than the circular stimuli cited by Quaid and Flanagan (2005a). Some of the differences in performance may be due to problems of spatial summation (Barlow, 1958). Comparing stimulus sizes within the current study, the overall stimulus sizes were smaller for the third experiment (diameter: 2.9°–4.1°) than the other experiments (Experiment 1: 5°; Experiment 2: 6°–14°). In contrast, contour-manipulated stimuli were all larger in overall diameter than the baseline 5°. The increasing size of the contour-manipulated stimuli has a number of implications. 

Differences between contour-modifying and area-modifying experiments represent a trade off between the distribution of area (clustered or dispersed) and overall stimulus size. In the case of area-modifying experiments, size and area act in the same way. It is unclear whether the decrease in size or area of the stimulus was the reason for the degradation of performance. In contour-modifying experiments, the increase in size of the stimulus acts in opposition to the distribution of the area. Namely, in this set of experiments, although the overall stimulus size increased, the area was distributed as a thin ring, which is harder to perceive. It is possible that the overall increase in size compensates for the thinning of the ring. 

The benefit of increasing overall size is unclear, but there are a number of possibilities. Larger stimuli, although thinner, mean that the stimulus falls further into the periphery. Contour-modified experiments vary in size from 6° to 14°. A 14° stimulus will extend 7° into the periphery. This is important because, as the stimulus moves further into the periphery, the stimulus is much less dependent on area and size, as established in this and other studies (Quaid & Flanagan, 2005a). 

Variability effects are also in keeping with previous studies that suggested perception of the illusion was more robust at greater eccentricities (Quaid & Flanagan, 2005a, 2005b). The fragility of the illusion at fixation, found in both this and other studies of FDF, increases variability in thresholds both between and within subjects (Quaid & Flanagan, 2005a). 

A comparison of two rings of equal thinness (0.5°) shows that there was an advantage to larger overall size. This effect was most pronounced close to fixation. The improvements with increasing area in the second set of experiments were too large to be explained by the change in overall target size. An almost 3-fold increase in target size caused a maximum difference of less than 0.5 log Michelson percent contrast units ( Figure 8). A much larger effect can be seen from area manipulations, according to experiment two. It is possible that for the contour-manipulation experiments the addition of greater overall stimulus size counteracted the effects of contour manipulation. Although, the effects of contour manipulation were still significantly smaller than those for area, if the overall diameter size was able to compensate. This still suggests that although overall target size can compensate somewhat for decreases in area, or contour, the limitation imposed by area is much larger and cannot be compensated for by overall target sizes. Obviously, the area maintained in the contour experiments was sufficient for the illusion to be perceived. In the third experiment, in which area was manipulated, performance degraded significantly with decreasing area. Also to be considered is that increasing overall target size, with equal ring thinness, means an increase in area. Some of the effects of the increase in overall target size are potentially due to the increase in area. 

This study showed that there was an area-dependent component to perception of FDF, and that the boundary region alone could not model stimulus perception. These findings suggest that perception of FDF is based on a combination of stimulus size (i.e., the outer diameter) and stimulus area (i.e., the difference between the outer and the inner diameter), which compensate for the thinning of rings and decreasing area. These findings support previous data which place emphasis on the relationship between the size of the stimulus and the amount of area within that stimulus that flickers out of phase that has previously been labeled as the percentage spatial content (Quaid & Flanagan, 2005a). Rogers-Ramachandran's original theory that FDF-like stimuli use only the fast contour extraction system suggests that the characteristics of the surface, such as area, should not affect perception. The current study showed, along with the importance of spatial content percentage, that the fast contour pathway is insufficient to explain the percept. This indicates that FDF perception, a predominantly magnocellular phenomenon, is influenced by the interaction with the parvocellular mechanisms, which in turn is affected by stimulus eccentricity. 

At all eccentricities, greater stimulus area means that more receptive fields can be activated. As area is decreased, thresholds increase and performance degrades. Stimuli that were modulated in contour but maintained the baseline area had low thresholds. Thus, even though the rings were thinner, thresholds remained low. This is either due to the importance of area or to the compensatory benefit of increasing stimulus size. The first experiment in which stimulus size was maintained, but area was manipulated by thinning rings showed that overall stimulus size can compensate for decreasing area. 

The importance of area was eccentricity dependent. For smaller stimulus sizes (<5° in diameter), performance improved faster further from fixation but then stopped improving in contrast to the principles of spatial summation. Similar results were reported for FDF by Quaid and Flanagan (2005a), for other flickering stimuli (Mäkelä et al., 1994; Tyler & Silverman, 1983) and on other temporally defined stimuli by Lee and Blake (1999). 

In summary, FDF is a temporally driven illusion, which benefits from temporal synchrony but remains subject to the rules of other luminance-defined stimuli, close to fixation. Although FDF detection is subject to area and stimulus size thresholds, it is still easier to detect further from fixation, similar to other findings and in keeping with a typical magnocellular response (Quaid & Flanagan, 2005a). Thus, although FDF shares many properties with other first order stimuli, such as a dependence on area close to fixation, it's ability to compensate by a greater amount of contour, makes it distinct in its mechanisms and properties. According to a recent study, it is unlikely that this stimulus is entirely dependent on the magnocellular system (Skottun & Skoyles, 2006). The current study suggests that although there is a dependence on the fast-acting contour extraction system, which we believe to be the magnocellular system, FDF is still dependent on a slow surface system, which is most likely the parvocellular system. 

Acknowledgments

Custom software was created by James Cassidy. This research was funded by a CIHR training grant, OGSST and OGS to DG, and research support from Heidelberg Engineering to JGF. Thanks to Patrick Quaid for his many helpful suggestions, both in experimentation and in manuscript preparation. 

Commercial relationships: Heidelberg Engineering, research grant to JGF. 

Corresponding author: Deborah Goren. 

Email: [e-mail]. 

Address: 200 University Avenue West, Waterloo, ON, Canada N2L 3G1.