Factors governing the speed of color adaptationin foveal versus peripheral vision

Romain Bachy1,* and Qasim Zaidi1,2

1Graduate Center for Vision Research, SUNY College of Optometry, New York 10036,USA2e-mail: qz@sunyopt.edu

*Corresponding author: rbachy@sunyopt.edu

Received October 7, 2013; revised December 18, 2013; accepted December 27, 2013;posted January 9, 2014 (Doc. ID 198967); published February 12, 2014

Troxler showed that fixated stimuli fade faster in peripheral than in foveal vision. We used a time-varying pro-cedure, to show that peripheral adaptation is faster and more pronounced than foveal adaptation for the threecardinal color modulations that isolate different classes of retinal ganglion cells. We then tested the hypothesisthat fixational eye movements control the magnitude and speed of adaptation, by simulating them with intermit-tent flashes, and attenuating their effects with blurred borders. Psychophysical and electrophysiological resultsconfirmed the eye movement-based hypothesis. By comparing effects across classes of ganglion cells, we foundthat the effects of eyemovements are mediated not only by the increase in size of receptive fields with eccentricity,but also by the sensitivity of different ganglion cells to sharp borders and transient changes in the stimulus. Finally,using the same paradigm with retinal ganglion cells, we show that adaptation parameters do not vary for the threeclasses of ganglion cells for eccentricities from 2° to 12°, in the absence of eye movement. © 2014 Optical Societyof America

OCIS codes: (330.1720) Color vision; (330.4060) Vision modeling; (330.5510) Psychophysics; (330.7320)Vision adaptation; (330.5380) Physiology.http://dx.doi.org/10.1364/JOSAA.31.00A220

1. INTRODUCTIONProlonged viewing of a stimulus causes neural adaptation atmany levels of the visual pathway [1–3]. Adaptation modifiesperceived attributes of the stimulus and, on cessation ofthe stimulus, can lead to a visual after-effect. For coloredstimuli, perceived colors desaturate with viewing and theafter-image on a neutral background is in the complementarycolors.

Troxler [4] showed that, when eye movements are mini-mized by fixation, stimuli fade faster in the periphery thanin the foveal vision. The slower fading, or even the nonfading,in foveal vision could be attributed to small fixational eyemovements that can selectively refresh responses of fovealneurons, which have narrow receptive fields (RFs), but can-not do the same for responses of the peripheral neurons,which have wider RFs [5–9]. Clarke and colleagues [10–13]examined the Troxler effect with a series of experimentson light adaptation in a dark environment. They characterizedthe time-course of peripheral brightness adaptation by aperipheral versus central equalization task [11], as well asthe recovery time following exposure with different visual pat-terns, including gratings [12]. They hypothesized that slowercentral adaptation was based on the effects of eye movementsat the borders, and had a postreceptoral, but precortical, locus[13]. Although the Troxler effect has been exploited for somestriking color illusions, e.g., Hinton’s Lilac Chaser [14], andhas been the subject of many investigations [15,16], the effectseems not to have been experimentally investigated with colorstimuli or by simulating the effects of fixational eyemovements.

Zaidi et al. [17] devised a new color after-image paradigm tostudy color adaptation. Two halves of a bipartite disk werefirst increased, then decreased in contrast as a temporalhalf-sinusoid [Fig. 1(a)] along one of the three cardinal direc-tions of color space [18,19]. Observers saw the contrastincrease, then decrease to zero, then reverse as an after-image[Fig. 1(b)]. They were asked to report the point of perceivedidentity (identity-point) of the two halves using a clock face[17]. Since this point preceded the point of physical equality,the contrast at that point provided an estimate of the magni-tude of adaptation (nulling-contrast). Modulation along thecardinal axes isolates the responses of the three types of gan-glion cells [20]: Δ�L −M� isolates cells that project to theParvo layers in the lateral geniculate nucleus (LGN), Δ�L�M � S� cells that project to Magno layers, and Δ�S� isolatescells that project to Konio cells; thus, we will call them PC,MC, and KC ganglion cells. The same paradigm, used within vivo electrophysiological recordings, revealed that all threeclasses of primate retinal ganglion cells adapt in a subtractivefashion (equivalent to a high-pass filter [21]); however, with atime-course much slower than photoreceptor adaptation. As aresult of the adaptation, the ganglion cell responses return tobaseline before the sinusoidal modulation has reached itszero-crossing, and then exhibit a rebound response that formsthe after-image signal for later neural centers.

In this study, we compared the speed and magnitude ofmonocular color adaptation in foveal and peripheral visionusing the color after-image procedure. We then tested thehypothesis that the difference between foveal and peripheraladaptation is due to eye movements by deriving predictions

A220 J. Opt. Soc. Am. A / Vol. 31, No. 4 / April 2014 R. Bachy and Q. Zaidi

1084-7529/14/04A220-06$15.00/0 © 2014 Optical Society of America

from the Zaidi et al. [17] ganglion cell adaptation model(Fig. 2). In this model, a postreceptoral adaptation signalaccumulates continuously, but also decays as a negativeexponential function of time, and is subtracted from the in-stantaneous postreceptoral input [Fig. 3(a)]. We used thehalf-sinusoid modulation as the adapting stimulus becausethe model predicts that a multiplicative scaling on the postre-ceptoral signal, as could happen by a decrease in sensitivity,would not alter the timing of the identity-point [Fig. 3(b)

versus Fig. 3(a)], but if eye movements repeatedly movethe RF of a cell across the border of the stimulus, alternatingstimulus and background as input, the identity-point will bedelayed [Fig. 3(c) versus Fig. 3(a)].

2. EXPERIMENT 1: FOVEAL ANDPERIPHERAL ADAPTATION WITH ANDWITHOUT JITTER SIMULATIONIn the first experiment, we used the after-image method tomeasure the identity-point of adaptation to slow sinusoidalmodulations along all three cardinal directions for the centralfovea and a peripheral location. Then, to simulate the effectsof eye movements moving RFs across the stimulus border, werapidly turned the stimulus on and off during the modulation,and made the same measurements at the same locations. Wechose to simulate the effects of eye-jitter rather than control it,because the required stabilization could only be achieved witha tracking adaptive optics scanning laser ophthalmoscope,and these are not yet capable of producing color modulationsthat can isolate classes of ganglion cells [22].

A. MethodsA3.2° radius bipartite diskwas displayed in the center of a CRTin a dark room and viewed from a chin-rest at 1.0 m. The colorsof the two halves were slowly modulated on the monitor to op-posite ends of each of the cardinal color axes [23]: equilumi-nant red/green Δ�L −M� (R/G), equiluminant yellow/violetΔ�S� (Y/V), and light/dark Δ�S �M � L� (L/D), as 16 s sinusoi-dal half-cycles (1/32 Hz). The temporal cycle began and endedat a 62.85 cd∕m2 mid-gray level, whichwas also the color of thebackground. The time of the identity-point, where the per-ceived contrast was null (R�t� � 0), was reported by using afoveal clock (Fig. 4). The trial ended when the observer re-ported that the after-image had faded. On the jitter simulation

(a)

(b)

Fig. 1. Stimulus time-course: (a) physical stimulus contrast varies as a half-sinusoidal cycle so the contrast Q�t� lies between 0.0 and 1.0.(b) Perceived stimulus time-course reaches identity, or R�t� � 0, before Q�t� � 0, and then a negative color after-image occurs [note the reversecolors on the last three panels of Fig. 1(b)]. The value of Q�t� at R�t� � 0 provides a measure of adaptation.

Fig. 2. In the computational model, a decay functionA�t� � e−Δ�t�∕τ isconvolved with the stimulus signal Q�t�. The result is subtracted fromthe stimulus signal so the response is R�t� � Q�t� − Q�t� � A�t�.

(a) (b) (c)

Fig. 3. (a) Response R�t� to a full contrast stimulus Q�t�. Identity-point (black line) precedes the point of physical equality (gray line).(b) Q�t� reduced by ω to simulate weaker sensitivity. The identity-point is the same as with the full contrast (black line). (c) StimulusQ�t� with pulses to background to simulate eye jitter. The null re-sponse (black line) occurs later compared to (a) and (b) (dashed line).

(a) (b)

Fig. 4. Stimuli sample: (a) L/D 3.2° foveal stimulus (C). (b) L/D 3.2°peripheral stimulus (P) at 8° eccentricity.

R. Bachy and Q. Zaidi Vol. 31, No. 4 / April 2014 / J. Opt. Soc. Am. A A221

trials, the stimulus was returned to the background level for95ms for every 800ms. Four conditionswere tested: 3.2° fovealstimulus without pulses or with pulses, and 3.2° peripheralstimulus without pulses or with pulses, centered at an eccen-tricity of 8° (Fig. 4). The large stimulus size was used becauseperipheral adaptation was too rapid to make useful compari-sons for small 2° stimuli. Measurements on retinal ganglioncells up to 8° have shown that at all eccentricities used in thisstudy, RF centers are fed by single cones, but the size of theRFsincreases with eccentricity [24,25].

Seven observers with normal color vision participated inthe experiment. Each of them completed 20 trials per condi-tion after a training session of three trials of each condition.

Two time values were recorded for each trial: the identity-pointR�t� � 0, and the endof the color after-image. Fromeveryidentity-point, we computed the corresponding physical con-trast value Q�t� of the stimulus in the range of 0.0–1.0, with1.0 being the maximum excursion possible on our monitor.The color after-image duration reports guaranteed that observ-ers did not perceive any residual after-image for the next trial.The after-image duration measurements were much noisierthantheidentity-pointmeasures,andwerenotusedforanalysis.

B. ResultsFigure 5 shows the mean contrast values Q�t� at the identity-point R�t� � 0 for each of the four conditions for each of thethree color axes, averaged across all seven observers, accom-panied by the 95% confidence intervals (see Appendix A Fig. 9for individual results). The data on the four conditions wereused to compare three pairs of stimulus conditions: theclassical case of center versus periphery without jitter, andpresentations with eye movement jitter simulation versuswithout, for each of the two locations. Statistical tests for eachcolor axis were based on paired comparison t-tests for all 140paired trials, i.e., each trial was its own control across observ-ers and sessions (see Table 1).

In the conditions without the jitter simulation, the adapta-tion time-course was significantly slower at the fovea thanthe periphery for all three cardinal axes, confirming theTroxlereffect for our experimental conditions. It is unlikely that fasteradaptation in the periphery could be causedby aweaker signal,whether receptoral or postreceptor, because the ganglion cellsubtractive adaptationmodel [17] shows that theweaker signalwould not change the identity-point setting [Fig. 3(b)].

With jitter simulation for the foveal stimuli, the adaptationtime-course was similar to the condition without jitter for all

three cardinal axes, suggesting that, in the absence of our si-mulated jitter, fixation eye movements were already reducingadaptation (the effect was statistically significant but quitesmall for L/D). Peripheral jitter simulation slowed down adap-tation for the R∕G and Y∕V axes, but did not significantlychange the adaptation time course for the L∕D axis. Thatthe effect of the jitter simulation for the chromatic modula-tions was to decrease adaptation to peripheral stimuli, with-out affecting the foveal adaptation, indicates that the weakerfoveal adaptation under usual conditions is likely to be due tothe effects of eye-jitter. That jitter simulation did not substan-tially affect foveal or peripheral brightness adaptation, sug-gests that the jitter simulation we used was not sufficientto change the effects of eye movements on the transientMC ganglion cells that convey brightness contrast to higherareas [26–28].

3. EXPERIMENT 2: BLURRED VERSUSSHARP STIMULUS BORDERSThe efficacy of the remarkable Lilac Chaser illusion [14] ispartly due to blurred stimulus borders. A blurred border

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Fig. 5. Stimulus contrast values at identity-points: symbols show the mean results across seven observers times 20 trials. Bars showing the 95%confidence interval of the mean were the same size as the triangles (right-pointing for center C, upward-pointing for periphery P, left-pointing forcenter with jitter simulation C/J, and downward-pointing for periphery with jitter simulation P/J).

Table 1. Experiment 1: Paired Comparison t-Tests(Stim1–Stim2) for the Three Cardinal Axesa

R∕G Y∕V L∕D

Stim1 Stim2 ΔQ�tR�0� p < ΔQ�tR�0� p < ΔQ�tR�0� p <

C P −0.18 0.0001 −0.10 0.0001 −0.08 0.0001C C/J 0.01 0.1976 0.02 0.0505 0.03 0.0017P P/J 0.10 0.0001 0.12 0.0001 0.01 0.5475

aΔQ is the mean difference tested for being significantly different from zero(N � 140).

(a) (b)

Fig. 6. Stimuli sample: (a) L∕D 3.2° foveal stimulus with blurred edge(C/B). (b) L∕D 3.2° peripheral stimulus with blurred edge (P/B) at 8°eccentricity.

A222 J. Opt. Soc. Am. A / Vol. 31, No. 4 / April 2014 R. Bachy and Q. Zaidi

attenuates the effects of eye-jitter as compared with a sharpborder. Therefore, as a further test of the eye-jitter hypothesis,especially for the achromatic brightness condition, we re-peated the nonjitter conditions of Exp. 1, but where the outeredges of the bi-partite disk were either sharp, as before, orwith a spatial blur decreasing linearly in contrast for the outerthird [Fig. 6(a)]. We tested four conditions for each of thethree color axes: 3.2° peripheral stimulus with sharp orblurred edges, and a 3.2° foveal stimulus with sharp or blurrededges (Figs. 4 and 6). A subset of the six observers from Exp. 1participated in Exp. 2. Each of them completed 20 trials after atraining session of three trials. Identity-point and after-imagecessation times were recorded for each trial, similar to Exp. 1.

A. ResultsFigure 7 shows contrast values Q�t� corresponding to the oc-currence of the identity-point (R�t� � 0), averaged over all sixobservers (see Appendix B Fig. 10 for individual results),along with bars showing the 95% confidence intervals. Theresults are plotted to facilitate comparison of adaptationmagnitude in blurred versus sharp stimuli for each of the threecolor axes, separately for the foveal and peripheral presenta-tions. Statistical tests for each color axis were based on pairedcomparison t-tests for all 120 paired trials (see Table 2). Forfoveal and peripheral presentations, adaptation was signifi-cantly faster when the edges were blurred for the L∕D axis,but there was no significant adaptation difference for thechromatic Y∕V and R∕G axes. MC ganglion cells are tempo-rally transient and have a spatially bandpass response, soattenuating the effects of eye-jitter with stimulus blur shouldhave the effect on adaptation that is shown in the results. PCand KC ganglion cells have a more sustained and low-passresponse, so the gradient at the stimulus border should haveless effect.

4. ELECTROPHYSIOLOGICALMEASUREMENTS OF ADAPTATIONVERSUS ECCENTRICITYZaidi et al. [17] measured in vivo neuronal responses for KC,PC, and MC ganglion cells in stationary primate eyes. Eachcell was exposed to sinusoidal modulation from gray to eachpole of its preferred cardinal axis. The stimuli for KC and PCwere 5° uniform circular patches covering the RFs. Thestimuli for MC were stationary sinusoidal gratings of spatialfrequency 0.5 cycles per degree, with the cell centered atthe peak or trough. Response histograms for all classes ofcells resembled the left-bottom panel in Fig. 3(a), i.e., the cells’responses returned to baseline (zero-crossing) before thestimulus. By fitting the model in Fig. 2, we estimated zero-crossings for each ganglion cell and the correspondingcontrast at that level Q�t� (see [17] for details). Figure 8 plotsthese adaptation magnitude estimates against eccentricity,and shows that adaptation magnitude is essentially constantfrom 2° to 12° in the absence of eye movements.

Table 2. Experiment 2: Paired Comparison t-Tests(Stim1–Stim2) for the Three Cardinal Axesa

R∕G Y∕V L∕D

Stim1 Stim2 ΔQ�tR�0� p < ΔQ�tR�0� p < ΔQ�tR�0� p <

C P −0.13 0.0001 −0.06 0.0005 −0.04 0.0021C C/B −0.01 0.2916 −0.02 0.0554 −0.11 0.0001P P/B −0.01 0.6373 0.00 0.9128 −0.15 0.0001

aΔQ is the mean difference tested for being significantly different from zero(N � 120).

S B S B S B0

0.2

0.4

0.6

0.8Foveal

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Peripheral

R/G Y/V L/D R/G Y/V L/D

Fig. 7. Stimulus contrast values at identity-point. Symbols show the mean results across six observers times 20 trials (upward/downward-pointingfor R/G axis, left/right-pointing for S axis, and square/diamond for L/D axis). Adaptation to sharp edge stimuli are plotted on S-labeled columnswhileadaptation to blurred edge stimuli are plotted on B labeled columns. Bars show the 95% confidence interval of the mean.

0 2 4 6 8 10 12 14

0

0.2

0.4

0.6

0.8

1

Q(t

) fo

r nu

ll re

spon

se

Eccentricity (deg)

Fig. 8. Stimulus contrast values for RGCs’ null responses relative toeccentricity. Circles denote KC ganglion cells responses to Y∕V axismodulation. Triangles denote PC ganglion cells responses to R∕G axismodulation (upward-pointing triangle for green-center cells anddownward-pointing triangle for red-center cells). Squares denoteMC ganglion cell responses to L∕D axis modulation. Edged symbolsare for ON-type cells and filled symbols are for OFF-type cells. A first-order linear regression shows no tendency (gray line).

R. Bachy and Q. Zaidi Vol. 31, No. 4 / April 2014 / J. Opt. Soc. Am. A A223

5. DISCUSSIONWe first showed that the Troxler effect can be measured re-liably for all three classes of retinal ganglion cells with ourafter-image generation method. Next, we tested the hypothe-sis that the difference between foveal and peripheral adapta-tion is due to miniature eye movements, which are inevitablewhen observers are trying to fixate. In the absence of beingable to perfectly stabilize images on the retina, we tried to sim-ulate the effects of eye-jitter by turning the adapting stimulusto the background level for short pulses. This manipulationequated peripheral adaptation to foveal adaptation for thetwo chromatic cardinal axes, possibly because PC and KCganglion cells have long integration times. Since this manipu-lation did not work for the achromatic axis, we tried theopposite task of attenuating the effect of eye-jitter by blurringthe edge of the stimulus. This manipulation increasedadaptation for both foveal and peripheral stimuli, indicatingthat eye-jitter may play a role at both locations because ofthe transient nature of MC ganglion cells. That blur had similar

effects on both locations for achromatic stimuli and alsoexplains why increasing the effect of eye-jitter by addedpulses did not remove the difference between foveal andperipheral brightness adaptation. Finally, we showed thatelectrophysiological recordings from retinal ganglion cellsconfirm our inferences from the psychophysical data.

In conclusion, our results support the classical hypothesisthat the effects of fixational eye movements account for thedifference between foveal and peripheral color and brightnessadaptation. However, MC cells have larger RFs than PC cells,so comparisons of eye movement effects on different classesof ganglion cells reveal that the effects are also mediated bythe spatial and temporal response properties of ganglion cells,not just by the size of the RFs.

APPENDIX A: INDIVIDUAL RESULTSEXPERIMENT 1

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8R/G

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8Y/V

Q(t

) fo

r R

(t)

= 0

C P C/J P/J0

0.2

0.4

0.6

0.8L/D

Q(t

) fo

r R

(t)

= 0

Fig. 9. Experiment 1: individual results for the seven observers, O1–O7, from top to bottom.

A224 J. Opt. Soc. Am. A / Vol. 31, No. 4 / April 2014 R. Bachy and Q. Zaidi

APPENDIX B: INDIVIDUAL RESULTSEXPERIMENT 2

ACKNOWLEDGMENTSWe are grateful to Rob Ennis and Barry Lee for extractingthe eccentricities of the retinal ganglion cells and allowingus to present the adaptation data, to our observers for theircareful and patient observations, and to NEI for supportingthis work through grants EY07556 and EY13312 to QasimZaidi.

REFERENCES1. H. E. Smithson, “Sensory, computational and cognitive compo-

nents of human colour constancy,” Phil. Trans. R. Soc. B 360,1329–1346 (2005).

2. Q. Zaidi, “The role of adaptation in color constancy,” in Fittingthe Mind to the World: Adaptation and After-Effects in High-Level Vision, C. W. G. Clifford and G. Rhodes, eds. (OxfordUniversity, 2005), pp. 103–131.

3. M. A. Webster, “Evolving concepts of sensory adaptation,”F1000 Biol. Rep. 4, 21–28 (2012).

4. D. Troxler, “Über das verschwinden gegebener gegenstände in-nerhalb unseres gesichtskreises [On the disappearance of givenobjects from our visual field],” Ophthalmologische bibliothek 2,1–53 (1804).

5. S. Martinez-Conde, S. L. Macknik, and D. H. Hubel, “The role offixational eye movements in visual perception,” Nat. Rev.Neurosci. 5, 229–240 (2004).

6. S. Martinez-Conde, S. L. Macknik, X. G. Troncoso, and T. A.Dyar, “Microsaccades counteract visual fading during fixation,”Neuron 49, 297–305 (2006).

7. J. Krauskopf, “Effect of retinal image stabilization on the appear-ance of heterochromatic targets,” J. Opt. Soc. Am. 53, 741(1963).

8. R. W. Ditchburn, D. H. Fender, and S. Mayne, “Vision with con-trolled movements of the retinal image,” J. Physiol. 145, 98–107(1959).

9. L. A. Riggs, F. Ratliff, J. C. Cornsweet, and T. N. Cornsweet, “Thedisappearance of steadily fixated visual test objects,” J. Opt.Soc. Am. 43, 495–500 (1953).

10. F. J. J. Clarke, “Rapid light adaptation of localised areas of theextra-foveal retina,” Opt. Acta 4, 69–77 (1957).

11. F. J. J. Clarke, “A study of Troxler’s effect,” Opt. Acta 7, 219–236(1960).

12. F. J. J. Clarke, “Visual recovery following local adaptation of theperipheral retina (Troxler’s effect),”Opt. Acta 8, 121–135 (1961).

13. F. J. J. Clarke and S. J. Belcher, “On the localization of Troxler’seffect in the visual pathway,” Vis. Res. 2, 53–68 (1962).

14. M. Bach, “Hinton’s “Lilac chaser”,” http://www.michaelbach.de/ot/col_lilacChaser/index.html (2005).

15. A. E. Welchman and J. M. Harris, “Filling-in the details onperceptual fading,” Vis. Res. 41, 2107–2117 (2001).

16. M. Millodot, “Variations extra-fovéales du phénomène detroxler,” Psychologie Française 12, 190–196 (1967).

17. Q. Zaidi, R. Ennis, D. Cao, and B. B. Lee, “Neural locus of colorafterimages,” Curr. Biol. 22, 220–224 (2012).

18. A. M. Derrington, J. Krauskopf, and P. Lennie, “Chromatic mech-anisms in lateral geniculate nucleus of macaque,” J. Physiol.357, 241–265 (1984).

19. J. Krauskopf, D. Williams, and D. Heeley, “Cardinal directions ofcolor space,” Vis. Res. 22, 1123–1131 (1982).

20. H. Sun, H. E. Smithson, Q. Zaidi, and B. B. Lee, “Specificity ofcone inputs to macaque retinal ganglion cells,” J. Neurophysiol.95, 837–849 (2006).

21. T. von Wiegand, D. Hood, and N. Graham, “Testing a compu-tational model of light-adaptation dynamics,” Vis. Res. 35,3037–3051 (1995).

22. D. W. Arathorn, S. B. Stevenson, Q. Yang, P. Tiruveedhula, andA. Roorda, “How the unstable eye sees a stable and movingworld,” J. Vis. 13(10), 22–41 (2013).

23. Q. Zaidi and D. Halevy, “Visual mechanisms that signal the di-rection of color changes,” Vis. Res. 33, 1037–1051 (1993).

24. D. M. Dacey, “The mosaic of midget ganglion cells in the humanretina,” J. Neurosci. 13, 5334–5355 (1993).

25. B. B. Lee, P. R. Martin, and U. Grünert, “Retinal connectivity andprimate vision,” Prog. Retinal Eye Res. 29, 622–639 (2010).

26. T. Yeh, B. B. Lee, and J. Kremers, “The time course of adaptationin macaque retinal ganglion cells,” Vis. Res. 36, 913–931 (1996).

27. D. H. C. Nothdurft and B. B. Lee, “Responses to colouredpatterns in the macaque lateral geniculate nucleus: patternprocessing in single neurones,” Exp. Brain Res. 48, 43–54(1982).

28. A. Valberg, B. B. Lee, P. K. Kaiser, and J. Kremers, “Responses ofmacaque ganglion cells to movement of chromatic borders,”J. Physiol. 458, 579–602 (1992).

S B S B S B0

0.2

0.4

0.6

0.8Foveal

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Peripheral

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Foveal

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Peripheral

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Foveal

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Peripheral

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Foveal

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Peripheral

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Foveal

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Peripheral

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Foveal

Q(t

) fo

r R

(t)

= 0

S B S B S B0

0.2

0.4

0.6

0.8Peripheral

Q(t

) fo

r R

(t)

= 0

Fig. 10. Experiment 2: individual results for the six observers,O1–O6, from top to bottom, are the same as O1–O6 in Fig. 9.

R. Bachy and Q. Zaidi Vol. 31, No. 4 / April 2014 / J. Opt. Soc. Am. A A225