Long-range twinkle induction: an achromatic rebound effect in the magnocellular processing system?

Christopher W. Tyler and Lani Hardage



After viewing a unstimulated region surrounded by a dynamic noise stimulus, viewers report the perception of prolonged dynamic twinkle in the unstimulated region. This twinkle aftereffect may be induced over long ranges of the visual field, up to 10º from the edge of the noise in central vision. Our previous studies of the properties of this aftereffect suggested mediation by the magnocellular processing system. We therefore evaluated the properties predicted by the magnocellular hypothesis by varying the coloring, temporal and spatial frequency of the stimulus. No aftereffect could be induced by an equiluminant color stimulus or by luminance noise below the temporal frequency of 5 Hz. The aftereffect obtained by luminance noise above 5 Hz was stronger for larger inducing elements. These results are consistent with known properties of the magnocellular processing system.



Dynamic perceptual aftereffects have presented a puzzle of particular interest to vision research because they imply that a steady neural signal is producing the appearance of a fluctuating visual percept. The well-known waterfall effect, and its cousin, the moving train illusion, are familiar examples of negative motion aftereffects induced in a patterned field by adaptation to motion in the opposing direction. Dynamic aftereffects on a blank field are a special class of effects produced by stimuli that may be either static or moving. Purkinje (1823) mentioned an orthogonal streaming aftereffect in a blank field after adapation to a set of static lines, as studied by Dvorak (1870) and MacKay (1957) among others (see Wade, 1977).

A new kind of dynamic aftereffect was recently reported by Ramachandran and Gregory (1991) to occur in unstimulated regions of the visual field in response to the cessation of a dynamic noise adapting surround. The unstimulated test region was filled with the percept of dynamic twinkling noise of fine grain, which could last for many seconds. This aftereffect differs from the motion aftereffect in having no coherent direction of motion and in that no aftereffect is seen in the stimulated surround (except perhaps an unusual degree of quiescence).

Recently, we discovered that such dynamic noise (twinkle) aftereffects were not just local but were induced at suprisingly long distances from the edge of the inducing noise, as much as 10º away (Hardage et al 1995). A static depiction of the inducing stimulus and the percept resulting under one of the experimental conditions is provided in Fig. 1. (Spillmann et al 1992 also reported long-range induction of a noise percept as an aftereffect of a dynamic noise surround, but they characterized it as static noise.) The long range of induction, which implicates long-range rebound effects propagating across the visual field, distinguishes the induced twinkle aftereffect from the motion and similar aftereffects, which are restricted to the region of previous stimulation.

(The term ‘induced’ is included in its name to connote this spatial opponency of its occurrence.) We also found that the spatial properties of the inducing noise had little effect on the perceived grain or the duration of the induced twinkle aftereffect over a range of element sizes from 9 - 60 arc min (Fig. 2), unlike the motion aftereffect, which is tuned to spatial frequency content.


Fig. 1A. One frame of a dynamic inducing stimulus consisting of large checks surrounding a 6º uniform test area. B. Depiction of the dynamic twinkle aftereffect as it appeared about 3 sec after surround offset. Fine-grain twinkle filled the entire test area immediately after surround offset, but retreated toward the center of the test area as time progressed. This observation excludes the possibility that the aftereffect in the center of the test area was merely due to perceptual filling-in from an aftereffect localized close to the edge of the inducing surround. C. One frame of the chromatic noise surround. D. Depiction of the absence of a twinkle aftereffect at equiluminance as predicted by the magnocellular hypothesis.


We now address the question of whether such long-range, dynamic interactions occur in the chrominance system. Purely chromatic dynamic noise is readily visible; its appearance gives no reason to suppose that the chrominance system would not support the same aftereffect as the luminance system. On the other hand, we found (Hardage and Tyler 1995) that noise frame rate had to be faster than 10 frames/sec to generate a significant achromatic aftereffect. Thus the implied spatial interactions of induced twinkle are restricted to dynamic conditions of both stimulation and perceptual response, suggesting an origin at some level of the magnocellular pathway. We therefore evaluated the properties predicted by the magnocellular hypothesis (Merigan et al 1990; Schiller et al 1990; Kremers et al 1992; Merigan et al 1992) by varying the color, temporal frequency and spatial frequency of the surround stimulus. The magnocellular pathway is reputed to be essentially color-blind, so that no aftereffect would be predicted for purely chromatic noise. We also measured the twinkle aftereffect as a function of achromatic contrast, to quantify the aftereffect duration predicted by the achromatic contrast equivalent to the maximum cone contrast available for chromatic noise. Additionally, temporal frequency was varied for comparison with stimuli designed to probe the magnocellular processing system in primates.




Random-element stimuli were generated with a Macintosh IIfx at a non-interlaced frame rate of 67 Hz and displayed on an Apple color monitor subtending 23.5 deg horizontally by 17.5 deg vertically at a viewing distance of 57 cm. In the center of the display was a circular test patch of fixed diameter of approximately 6º for all element sizes tested. The random-element surround consisted of black and white, or red and green, square pixels, of either small (4.4 arc min) or large (44 arc min) extent. The stimulus elements had a dot density of 50% black or red and 50% white or green so as to fill the surround area, and were refreshed with a new set of random dots on each frame. The central patch was either gray, in the case of the achromatic surround stimulus, or yellow, in the case of the red/green surround stimulus.

Luminance of the inducing noise surrounds was measured in the testing room with an ambient light level of 2.1 candela/m2 +0.1 sem. Luminance levels for the achromatic surround were reduced by viewing through a neutral density filter in order to equate mean luminance with that used in the color test, 9.1 cd/m2. The Michelson contrast was varied from 65%, with light elements at 15.0 cd/m2 and dark elements at 3.2 cd/m2, to 1%, with light elements at 9.19 cd/m2 and dark elements at 9.01 cd/m2. The mean luminance of the uniform central test patch was set at 9.1 cd/m2.

Color CIE values and luminance output of the monitor were measured by means of a Minolta Chromameter. Maximum output was 18.2 cd/m2 for the red gun, 59.8 cd/m2 for the green gun. In order to achieve luminance contrast values of + 80% in the color stimulus, the maximum green output that could be used was the equiluminant value of 18.2 cd/m2, resulting in a mean luminance of 9.1 cd/m2 for the red/green surround stimulus. Color contrast levels calculated in terms of the equivalent cone contrast energy elicited by the red and green surround stimuli. At a luminance contrast of 80%, color contrast could be varied from maximum of 89% down to 0%, or isochromaticity. The color of the central patch was set at this isochromaticity point.

Cone contrasts for the heterochromatic surrounds were calculated in the following manner: energy elicited by the red and green elements in long-wavelength and middle-wavelength cones was calculated according to the standard formula for spectral sensitivity of the R and G cone types for each phosphor output (Wyszecki et al 1982):

  and '

where are the spectral distributions of the phosphors and are the spectral efficiencies of the cone types. (Short-wavelength cones are not considered because energy elicited in them by these red and green stimuli was negligible.) Contrast seen by each type of cone was calculated as the Michelson contrast elicited by red (r) and green elements (g) according to the formula and. The overall cone contrast of the surrounds was then calculated as the square root of the summed squared contrasts in each cone, or (CL2 + CM2 )/. Cone contrasts ranged from 76% for a maximum luminance contrast of 80% down to 23% for the luminance contrast of 0% (equiluminance).



Two observers with normal visual acuity viewed the screen binocularly; a third observer, author LH, was functionally monocular (i.e., aphakic with counting fingers vision in the right eye since age 5). Not all observers were tested on every experimental condition. Each trial consisted of a 20-sec period of adaptation to the inducing noise surrounding the static test area and a subsequent test period during which any dynamic twinkle aftereffect was timed. Observers fixated a dark spot subtending 3.3 arc min in the center of the test area throughout. It should be emphasized that there was no stimulation in the test area at any time during the trial.

After observers adapted to the dynamic noise stimulus surrounding the uniform central patch, they were presented with a uniform test screen with a luminance of 9.1 cd/m2, either grey for the black/white or yellow for the red/green inducing surrounds, respectively. They responded with a computer-controlled timing key when a criterion was met for disappearance of the dynamic scintillation perceived in the test patch. This criterion was the disappearance of the fine scintillating spots of a range of sizes and scintillation rates, seen in the test patch after cessation of the adapting stimulus.

All trials were performed in block of 1.5 hr, with the trial conditions for a particular session interleaved. The trial conditions included achromatic and chromatic contrast levels for two element sizes and several adapting rates; these conditions are defined more fully in the sections describing the experiments in which they were used. Eye blinks during adaptation or during the aftereffect did not appear to affect the twinkle aftereffect.



We tested the magnocellular origin hypothesis by evaluating the twinkle aftereffect induced by chromatic and achromatic surrounds under similar conditions of luminance contrast, element size and noise frame rate. Previous studies of the twinkle aftereffect (Hardage and Tyler 1995) showed optimum conditions for induction of the effect to be central viewing (with surrounding noise) of a uniform test area at least 3º in diameter, adaptation duration of at least 15 sec, noise elements at least 9 arc min on a side, and binocular viewing by observers with binocular vision. (The monocular subject’s performance was equivalent to that of binocular observers in binocular viewing).

For this study, three experiments were designed to evaluate the magnocellular hypothesis. The first experiment calibrated the effect of achromatic contrast on aftereffect duration for adaptation to a noise surround with either large or small elements. The second experiment searched for an aftereffect to chromatic noise surrounds for both large and small sizes, as an added achromatic component in the noise was reduced to zero. The third experiment was designed to ensure that we had not missed a chromatic aftereffect with a different (lower) temporal frequency range than the achromatic aftereffect by determining the effect of varying frame rate on aftereffect duration for both achromatic and combined chromatic/achromatic surrounds.



Experiment 1: Characterization of the dynamic aftereffect by varying luminance contrast of achromatic surrounds

Studies of the response of primate magnocellular and parvocellular systems in response to contrast-modulated flicker rates (Merigan and Maunsell 1990; Schiller, Logothetis et al. 1990; Kremers, Lee et al. 1992; Merigan, Byrne et al. 1992) have shown that magnocellular lesions degraded flicker detection at high temporal and low spatial frequencies. We therefore designed achromatic and chromatic surrounds, with their luminance levels equated, to preferentially stimulate the magno and parvo systems, respectively, and compared twinkle aftereffect durations induced in response to changes in luminance contrast.

Fig. 2. Lack of effect of noise element size on twinkle aftereffect duration for three observers (replotted from Hardage and Tyler 1995). Perceived duration of twinkle was very stable down to test area diameters of about 0.2º (12’), but showed idiosyncratic fluctuations below this level. Note that the perceived twinkle grain was also invariant with noise element size (cf., Fig. 1).

 A previous study with achromatic surrounds only had shown that the dynamic twinkle aftereffect was relatively independent of element size down to 9 arc min (Fig. 2). Two sizes of stimulus element were tested in the present series of experiments; one size within the range previously tested (44 arc min), and one much smaller (4.4 arc min), to identify any preference of the parvo system for high spatial frequencies in the new chromatic stimulus.

The achromatic stimulus in the first experiment consisted of black and white square elements of either 4.4 or 44 arc min that completely filled the inducing surround. The noise surround covered the entire monitor screen except for the uniform test area, which was a circular 6º-region centered on fixation. We measured the duration of the aftereffect, seen on a neutral uniform field of 9.1 cd/m2 luminance immediately after cessation of the adapting stimulus, as a function of luminance contrast in the adapting stimulus.

Results for three observers are shown in Fig. 3. Achromatic noise surrounds of the larger element size were themselves detectable at 1-2% luminance contrast (open arrows at the x-axis) and produced detectable twinkle aftereffects above about 10% contrast. The duration of the dynamic twinkle aftereffect shown in Fig. 3 for two observers after 20-sec adaptation to noise elements of the larger size rose with luminance contrast from no aftereffect at 5-10% luminance contrast to a maximum of 5-8 sec duration at 70% contrast. Aftereffect duration after adaptation to noise surrounds of the smaller size was significantly shorter for one observer. Threshold contrast for detection of the adapting surround was much lower (1-2%) for the larger elements than for the smaller elements (5-7%).


 Fig. 3. Duration of twinkle aftereffect as a function of contrast of achromatic noise surrounds, for two sizes of noise pixel. Arrows show detection thresholds of surrounds per se for the two pixel sizes.

Experiment 2: Characterization of the dynamic aftereffect from chromatic surrounds

The observers were tested with the chromatic noise surround at two element sizes. Beginning with the red and green surround elements set at the same luminance values, yielding zero luminance contrast, equal amounts of luminance were added to green elements and subtracted from red elements (open symbols) or vice versa (filled symbols) to measure the effect of luminance contrast on the twinkle aftereffect duration. Looking first at absolute detectability, chromatic noise surrounds that were equiluminant were detectable at 0.7% chromatic contrast, but dynamic modulation was not visible until 1.6% chromatic contrast. Added luminance modulation in the chromatic noise surrounds were detectable at 3% luminance contrast, but again, flicker was not visble in the stimulus until a luminance contrast of 6%. In comparison to detectability of the achromatic stimulus at 1-2% luminance contrast, the detectability of luminance contrast in the chromatic stimulus was thus impaired by the the presence of the chromatic noise modulation, implying some interaction between the two modalities.

Results for two observers are shown in Fig. 4. No twinkle aftereffect was detectable for the chromatic inducing surrounds at the equiluminance point (0 added luminance contrast in Fig. 4). At this equiluminance point, the equivalent cone contrast of the chromatic noise was 23%. The dotted lines in Fig. 4 shows the aftereffect duration for a luminance contrast of 23%, at 2.5-4.0 sec for the two observers. Thus, if a chromatic contribution to the aftereffect were based on the equivalent cone contrast, the equiluminant stimulus would have induced an aftereffect of about 3 sec duration. That no aftereffect at all was detected at equiluminance clearly demonstrates that there was a negligible chrominance contribution to the aftereffect.


Fig. 4. Duration of twinkle aftereffect as a function of added luminance contrast in mixed luminance/chromatic noise surrounds. At equiluminance (zero added luminance contrast), virtually no aftereffect was induced into the test area. Aftereffect duration was an approximately symmetric function of positive and negative luminance contrast and was generally less than that induced by pure luminance contrast (dashed lines, replotted from Fig. 3).


The V-shaped functions show the strength of the aftereffect when luminance contrast was added to the chromatic noise. Maximum added luminance contrast of 65% produced aftereffect durations of 5-10 sec. The aftereffect duration elicited by the purely achromatic surrounds (dashed lines, replotted from Fig. 3) shows that adding the chromatic component to the noise surround never enhances the aftereffect, and in fact significantly reduces its duration under most conditions tested (difference greater than 2 times the standard error bars). This comparison implies that adding chromatic contrast to a given luminance contrast not only does not enhance the aftereffect but has the inhibitory effect of reducing aftereffect duration.

Only in the fine-check condition for very low contrasts were hints obtained of enhancement of twinkle induction in the presence of added chromatic noise relative to the luminance noise case.


Experiment 3: Effect of adding chromatic noiseon frequency range of twinkle aftereffect

To ensure that the lack of a induced twinkle aftereffect for chromatic stimuli was not a function of the rate of the adapting stimulus, two observers were tested with both purely achromatic and joint achromatic/chromatic surrounds at a range of stimulation rates. The object of this experiment was to determine whether the presence of chromatic modulation in the stimulus showed any tendency to enhance the twinkle aftereffect duration at lower frame rates than the optimal frame rate for an achromatic stimulus determined from previous experiments. Naturally, the effect of frame rate could not be measured for equiluminant chromatic surrounds because no aftereffect was obtained at the equiluminance point. We therefore compared the temporal range of the luminance aftereffect to that for noise surrounds that also contained a high chromatic contrast (red/green noise with 80% luminance contrast, red bright and green dark). If there was any tendency for the presence of chromatic modulation to generate afterefeffects at lower temporal frequencies, it would be expected to lower the temporal range of the aftereffect in this mixed stimulus.

Because the element color from one frame to the next was randomly assigned, we define the frame rate as the cut-off frequency for the surround stimuli, implying that slower frame rates were also present in the stimulus. For example, at a cutoff-frequency of 33 Hz, some stimulus elements changed at 16 Hz, some at 11 Hz, some at 8.3 Hz, etc.

Results are shown in Fig. 5 for a luminance contrast 80% in the red/green surrounds (filled symbols) and a 65% luminance contrast in purely achromatic surronds (open symbols). At frame rates above 5 Hz, all observers showed substantial aftereffect durations for both types of adaptation. At any given cut-off frequency, aftereffect durations were generally longer for achromatic surrounds, despite the fact that the maximum cone contrast of achromatic surrounds (65%) was lower than the achromatic contrast of the joint chromatic/achromatic surrounds (80%). At 5 Hz and below, no aftereffect could be measured in 2 of the 3 observers, while the other observer showed only minimal aftereffect relative to his results at high frame rates. But the presence of chromatic modulation did not expand the range of the aftereffect to lower temporal frequencies for any of the observers.


Fig. 5. Duration of twinkle aftereffect as a function of temporal cutoff frequency of the noise frame rate in luminance only (open symbols) and mixed luminance/chromatic (filled symbols) noise surrounds. Aftereffect duration was negligible below 5 Hz and was generally lower when induced by chromatic noise than when induced by pure luminance contrast.



Contrast effects

The contrast manipulation of Fig. 3 shows that twinkle aftereffects were induced beginning at about 0.6 log units above the detection threshold for the surround noise. Aftereffect duration then increased in rough proportion to log contrast, although with a different curvature for the two observers. The onset of the twinkle aftereffect at about 10% inducing contrast might seem to be at variance with the magnocellular hypothesis, which would predict that the aftereffect should be seen at low contrast. It should be noted, however, that the threshold for detecting the high frequency peripheral noise is about a log unit higher than for static gratings, so the range of magnocellular phenomena derived from such noise should be expected to be similarly elevated. In other words, the threshold for twinkle induction occurs at an equivalent contrast of about 2%, when translated into low-frequency sensitivity terms. Since there is no clear model for twinkle induction, the specific level at which it should first occur is, in any case, hard to predict.

Curiously, there is almost no effect of size on aftereffect duration (<0.2 log units), despite a reduction of 0.4-0.8 log units in the detectability of the fine checks relative to the coarse ones. This dissociation may be related to the appearance of the induced twinkle, which was always fine-grained even when the surround checks were very coarse. This property may imply that the mechanism of twinkle induction is itself tuned to high spatial frequencies and hence gives a relatively weaker aftereffect to the coarse checks (even though some other mechanism may exist to detect those coarse checks with greater sensitivity).


Absence of chromatic induction

The data of Fig. 4 at zero added luminance-contrast provide clear evidence of an absence of (or negligible contribution of) chrominance input, for either noise-element size, into the mechanism of surround induction of the twinkle aftereffect. Whatever that mechanism may be, this result bears out the prediction from the dynamic properties of the aftereffect (Hardage and Tyler 1995) of its identification with the magnocellular pathway, since that pathway is known to be essentially colour-blind. Thus even the experimental strategy of favoring the parvocellular mechanism by the use of small element sizes failed to elicit significant twinkle induction at equiluminance (Fig. 4B, values at zero added luminance-contrast).

The same result was obtained when reducing the frame rate of the inducing surrounds. The reduction of twinkle aftereffect duration to zero for slow frame rates in Hardage and Tyler (1995) was replicated for luminance noise, and addition of high-contrast chromatic modulation to the noise surround failed to extend the perceived aftereffect to slower frame rates. Thus, even stimulation of the parvocellular system in its optimal temporal frequency range was insufficient to elicit any report of twinkle induction, again as predicted by the magnocellular hypothesis.

It is interesting to note that, contrary to the simple magnocellular hypothesis, substantial motion aftereffects have been reported for equiluminant chromatic stimuli (Cavanagh et al 1985; Mullen et al 1992). In this respect, the properties of the motion aftereffect thus differ from those of induced twinkle, just as they did for its spatial properties (see Introduction). Note that Cavanagh and Favreau (1985) have argued that, as with chromatic motion detection, the chromatic motion aftereffect is more likely a chromatic signal into the same motion channel than into a separate (parvocellular) motion system. These properties of the chromatic motion aftereffect, and their absence from the induced twinkle mechanism, clearly make the case that the two aftereffects have different neural substrates, at least in some respects.


Inhibitory effect of chromatic component

Although chromatic noise could neither induce a twinkle aftereffect or extend the range of that induced by luminance noise, there was an effect of the presence of a chromatic component in the noise. Fig. 4 shows that the aftereffect duration as a function of the luminance contrast in the surround may be significantly reduced by the presence of chromatic noise, relative to the durations reported for purely achromatic surrounds. In this respect, induced twinkle does share the same property as the sensitivity to chromatic motion and its aftereffect (Cavanagh et al 1984; Cavanagh and Favreau 1985; Mullen and Boulton 1992).

This result suggests two possibilities:

i) that there is an inhibitory interaction between the chrominance signal and the luminance signal in the visual processing stream.

ii) that the red and green elements are processed by separate neural mechanisms, so that the effective contrast in each colour mechanism is half that of the stimulus as a whole. If the responses were then recombined in some nonlinear or probabilistic fashion, the net result would be a reduction in aftereffect duration. In this model, it is not appropriate to regard the mixed luminance/chrominance signal as a linear combination of the two properties.


In conclusion, this new aftereffect seems to be a robust indicator of a profound spatial interaction operating over long ranges of the visual field and a phenomenon that may be firmly identified with the magnocellular pathway (to the extent that we know the properties of this pathway in humans). With the recent spate of studies on long range interactions in even primary cortical regions (Fries et al 1977; Nelson et al 1978; Fiorani et al 1992; Field et al 1993; Polat et al 1993; Kitano et al 1994; Kovacs et al 1994; Polat et al 1994; Kapadia et al 1995; Kitano et al 1995), the properties of a perceptually explicit manifestation of such interactions becomes of particular interest. The identification specifically with the magnocellular processing stream conforms with the typical properties of that stream as rapid, long-range, processing of achromatic stimulus events. It is nevertheless surprising to observe such a vivid hallucination of dynamic noise in a non-stimulated area of the field.



Cavanagh P and Favreau O E, 1985 "Color and luminance share a common motion pathway" Vision Research 11: 1595-1601

Cavanagh P, Tyler C W and Favreau O E, 1984 "Perceived velocity of moving chromatic gratings." Journal Optical Society of America A 1: 893-899

Dvorak V, 1870 "Versuche uber die nachbilder von reizveranderungen." Sitzungsberichte der Wiener Akademie der Wissenschaften 61: 257-262

Field D J, Hayes A and Hess R F, 1993 "Contour integration by the human visual system: Evidence for a local "association field" Vision Research 33: 173-193

Fiorani M J, Rosa M G, Gattass R and Rocha-Miranda C E, 1992 "Dynamic surrounds of receptive fields in primate striate cortex: a physiological basis for perceptual completion?" Proceedings National Academy Sciences USA 89: 8547-8551

Fries W, Albus K and Creutzfeldt O D, 1977 "Effects of interacting visual patterns on single cell responses in cat's striate cortex" Vision Research 17: 1001-1008

Hardage L and Tyler C W, 1995 "Induced twinkle aftereffect as a probe of dynamic visual processing mechanisms." Vision Research 35: 757-766

Kapadia M, Ito M, Gilbert C and Westheimer G, 1995 "Improvement in visual sensitivity by changes in local context: parallel studies in human observers and in V1 of alert monkeys" Neuron 15: 843-856

Kitano M, Kasamatsu T, Norcia A M and Sutter E E, 1995 "Spatially distributed responses induced by contrast reversal in cat visual cortex" Experimental Brain Research 104: 297-309

Kitano M, Niiyama K, Kasamatsu T, Sutter E E and Norcia A M, 1994 "Retinotopic and nonretinotopic field potentials in cat visual cortex." Visual Neuroscience 11: 953-77

Kovacs I and Julesz B, 1994 "Perceptual sensitivity maps within globally defined visual shapes" Nature 370: 644-646

Kremers J, Lee B B and Kaiser P K, 1992 "Sensitivity of macaque retinal ganglion cells and human observers to combined luminance and chromatic temporal modulation" Journal Optical Society America A 9: 1477-1485

MacKay D M, 1957 "Moving visual images produced by regular stationary patterns" Nature 180: 849-850

Merigan W H, Byrne C E and Maunsell J H R, 1992 "Does primate motion perception depend on the magnocellular pathway?" Journal of Neuroscience 11: 3422-3429

Merigan W H and Maunsell J H R, 1990 "Macaque vision after magnocellular lateral geniculate lesions" Visual Neuroscience 5: 347-352

Mullen K T and Boulton J C, 1992 "Interactions between colour and luminance contrast in the perception of motion" Ophth & Physiol Optics 12: 201-205

Nelson J I and Frost B J, 1978 "Orientation-selective inhibition from beyond the classic visual receptive field" Brain Research 139: 359-365

Polat U and Sagi D, 1993 "Lateral interactions between spatial channels: Suppression and facilitation revealed by lateral masking experiments" Vision Research 33: 993-999

Polat U and Sagi D, 1994 "The architecture of perceptual spatial interactions" Vision Research 34: 73-78

Purkinje J (1823). Beitrage zur Kenntniss des Sehens in subjecktiver Hinsicht. Berlin, G. Reimer.

Ramachandran V S and Gregory R L, 1991 " Perceptual filling in of artificially induced scotomas in human vision." Nature 350: 699-702

Schiller P H, Logothetis N K and Charles E R, 1990 "Role of the color-opponent and broad-band channels in vision" Visual Neuroscience 5: 321-346

Spillmann L and Kurtenbach A, 1992 "Dynamic noise backgrounds facilitate target fading" Vision Research 32: 1941-1946

Wade N J, 1977 "Distortions and disappearances of geometrical patterns" Perception 6: 407-433

Wyszecki G and Stiles W S (1982). Color Science: Concepts and Methods, Quantitative Data and Formulae. New York, John Wiley & Sons.