Stimulus Size Affects RodInfluence on Tritan ChromaticDiscrimination

Roger Knight,* Steven L. Buck,Maria PereverzevaDepartment of Psychology, University of Washington, Seattle, WA 98195–1525

Received 16 August 1999; accepted 2 September 1999

Abstract: A signal-detection procedure was used to measurethe influence of rods on S-cone-mediated (“tritan”) chro-matic discrimination in normal trichromatic human observ-ers. Rod influence on discriminability of both incrementsand decrements in S-cone excitation was assessed by com-paring cone-plateau (minimizes rod influence) and darkadapted (maximizes rod influence) conditions. For 1° stim-uli, rods impaired tritan discrimination, especially for S-cone decrements. For 8° stimuli, rods had no systematiceffects. This confirms that rods have the same direction ofeffect (impairment) on tritan discrimination as has beenshown for Rayleigh (L-M) discrimination. However, the sizedependence of the rod effect on tritan discrimination isopposite to that previously shown for either Rayleigh dis-crimination or hue appearance. This provides further evi-dence for separable rod influences on S-cone and L-M conepathways.© 2000 John Wiley & Sons, Inc. Col Res Appl, 26, S65–S68,

2001

Key words: chromatic discrimination; hue; rod vision; S-cones; tritan; psychophysics; human color vision

INTRODUCTION

After a period of dark adaptation, signals from rod photo-receptors can influence observers’ reports of subjective huesof spectral lights1–4 and the spectral wavelengths (loci)associated with unique (single hue percept) and binary (two

equally balanced hue percepts) hues.5–7 Such rod influencesaffect the distribution across the 400–700 nm spectrum ofall four basic hue percepts (red, green, yellow, blue) and allunique- and binary-hue loci. Rod influences on hue appear-ance are made most salient by using large, eccentricallyviewed stimuli with light levels reduced to about 1.5 logscot td.1,6,7 Studies of the spectral distribution and timecourse of these effects suggest separable rod influences onhue are mediated by S-cone (tritan) and L-M (Rayleigh)pathways.8

Rods can also influence chromatic discrimination at me-sopic light levels. In trichromatic observers, rod activity hasbeen shown toimpair both constant-luminance Rayleighdiscrimination thresholds9 and S-cone mediated discrimina-tion on the Farnsworth–Munsell 100-Hue Test.10 However,rod activity has been shown toimprovewavelength discrim-ination in S-cone monochromats.11 The basis for this dif-ference between normal and color-deficient observers is notknown.

The present study seeks to determine whether the appar-ent rod impairment of tritan discrimination found with theFM100 is replicated with more-rigorous psychophysicalmethods and control of stimuli and fixation. The presentstudy also seeks to determine whether the rod influence ontritan discrimination grows larger with stimulus size, as hasbeen shown for rod influence on hue5,6 and on Rayleighdiscrimination.9 The finding of similar size dependence ofall three effects would permit the most parsimonious mod-eling of rod influences on human color vision.

METHODS

Three color-normal adults (ages 24–42) served as observ-ers. Females MP and BM, who were naive as to priorhypotheses, and male RK were kept blind to their experi-mental results until the completion of all conditions.

* Correspondence to: Roger Knight, Department of Psychology, Univer-sity of Washington, Seattle, WA 98195-1525 (e-mail: rknight@u.washington.edu)

Portions of these data were presented at the 1999 Annual Meeting of theAssociation for Research in Vision and Ophthalmology, Ft. Lauderdale,FL, May 1999

Contract grant sponsor: NIH; Contract grant number: EY03221© 2000 John Wiley & Sons, Inc.

Supplement Volume 26, 2001 S65

A signal-detection psychophysical procedure was used toestimate the detectability of fixed-sized chromaticity shiftsof a homogeneous stimulus field that was either 1° or 8°diameter and centered 7° eccentric in the temporal visualfield. Two channels of a previously described1 computer-controlled Maxwellian-view apparatus were used to presentthe test stimulus and a dim red 1°-square fixation cross. Thechromaticity shifts produced increments or decrements ofS-cone stimulation, with little change of stimulation of L orM cones or rods. A pair of interference filters, nominally430 and 520 nm, was chosen that produced chromaticitiesinsitu that fell close to a tritan confusion line. Flicker pho-tometry (15 Hz sinusoidal exchanges) was used to adjust thetwo lights to equal luminance individually for each ob-server. The overall light level was adjusted to 1.6 log S-conetd (about 1.5 log scot td) at the starting chromaticity (0.131,0.378 in 1931 CIE space), which was formed by nearlyequal contributions of the two lights. S-cone increments anddecrements were produced by exchanging the 430 and 520nm lights at equal luminance.

Table I shows the % changes of excitation of all fourreceptor types that accompanied the chromaticity shifts forboth 1° and 8° fields. The sizes of the chromaticity shiftswere selected, on the basis of pilot testing, to be slightlyabove threshold for each stimulus size and were, therefore,smaller for the 8° field and larger for the 1° field. For aspecific field size, increments and decrements were of equalapparatus stepper-motor change and nearly equal chroma-ticity shifts. Calculations are based on the DeMarcoet al.cone fundamentals12 with foveal macular pigment densitiesremoved and the CIE scotopic luminosity function. Thespectral irradiance of the two mixture lights and of eachcombination was measured at 3-nm intervals by means of aGamma Scientific spectroradiometer, calibrated against arecently certified standard source.

The observer initiated each trial onset, typically withdelays of 10–20 s between trials. The test field came on atthe beginning of the trial and remained on throughout the4-s duration. On “signal” trials, the chromaticity shift (eitherincrement or decrement of S-cone stimulation) began 2 safter trial onset and 250 ms after a warning tone. Thesmaller chromaticity shifts used for the 8° field took 200 msfor completion. The larger chromaticity shifts used for the1° field took 1000 ms (and entailed exactly 5 times theexcursion of the apparatus stepper motors). The chromatic-ity remained at the target setting for the remainder of the

trail. On “blank” trials, the warning tone sounded but thechromaticity did not change.

The observer’s task was to say at the end of a trialwhether it had been an increment signal trial, a decrementsignal trial, or a blank trial. Noise delivered through head-phones and movement of identical but functionally irrele-vant stepper motors provided auditory and vibrationalmasking of extraneous stimuli. Increment trials and decre-ment trials were randomly mixed with blank trials in blocksof 20 trials. Half the trials in each block were actuallyblanks, the other half were evenly split between incrementsand decrements. Observers received no feedback after theirresponse. Detectability (d9) was calculated on the basis of50 signal and 100 blank trials spread over 10 blocks.

The effect of rods on these tritan discriminations wasassessed by comparing performance with the same physicalstimulus, but under two different adaptation conditions. Inthe “bleached” condition, all measurements were made onthe cone plateau between 3–8 min after a xenon flashbleach, when rods could contribute minimally, if at all. Inthe “dark-adapted” condition, all measurements were madeafter the observer had been dark adapted for at least 30minutes, when rods could contribute maximally. Both dark-adapted and bleached conditions were run in the same dailysession, with initial adaptation condition counterbalancedacross days.

RESULTS

The preselected chromaticity shifts yielded fairly compara-ble detectabilities for the 1° and 8° fields, as shown in Fig.2. Values ford9 ranged mostly between 1.0–2.0 for the 8°

FIG. 1. Coordinates of start (triangle) and ends of chroma-ticity shifts used for the 8° field (circles) and 1° field(squares). S-cone decrements are toward 520 nm; S-coneincrements are toward 430 nm.

TABLE I. Receptor stimulation changes during chro-maticity shift.

% Change in stimulation

S M L Rods

1° FieldIncrement 141.2 10.56 20.72 12.7Decrement 238.7 20.54 10.75 22.7

8° FieldIncrement 17.5 10.11 20.15 10.5Decrement 27.7 20.11 10.15 20.5

S66 COLOR research and application

field and between 1.5–2.5 for the 1° field. The overall hitand false-alarm rates for the tritan chromaticity shifts weregenerally 60–70% and 10–20%, respectively.

Figure 2 (top panel) shows thed9 values for the 8° fieldfor both dark-adapted (DA) and bleached (BL) conditionsfor both increments (1 or INC) and decrements (2 or DEC)of S-cone stimulation for all three observers and for themeans across observers. Figure 3 shows the influence ofrods on tritan discrimination directly as the arithmetic dif-ference in d9 (Dd9) between dark-adapted and bleachedconditions for each observer and shift direction. Positivevalues indicate that rods enhanced detectability; negativevalues indicate rods impaired detectability. The figuresmake clear that, for the 8° field (top panel), no two observ-ers show the same pattern of direction of rod influence onS-cone increments and decrements, and there is no net rodeffect on either increments or decrements across observers.

Figures 2 and 3 also show that, for 1° fields (bottompanels), the pattern of results is quite different. All threeobservers show small but consistent rod impairments of

detectability of tritan increments and much larger rod im-pairments of detectability of tritan decrements.

DISCUSSION

The results show, by means of a rigorous psychophysicalprocedure and careful control of stimuli and fixation, thatrods can impair tritan chromatic discrimination. This isconsistent with our previous report10 of selective rod-medi-ated impairment of tritan discrimination on the FM100. Thefailure to find a conspicuous or consistent rod influenceusing the 8° field is also consistent with our prior investi-gations13 that have used other psychophysical methods(method-of-adjustment, Boynton and Kambe14) and ex-plored a range of light levels and starting chromaticities.Both studies that found rod effects on tritan discriminationhave used small stimuli (1–2° extent). All our work with 8°stimuli show no pronounced or consistent rod effect. Asurprising aspect of this size dependence of rod influence ontritan discrimination is that it is opposite in direction to thesize dependence found for rod influence on either Rayleighdiscrimination9 or hue appearance.5,6 The basis for thisdifference is unclear, but ultimately has to be accounted foras part of a comprehensive account of rod influence on colorvision.

The finding of much greater rod influence on S-conedecrements than on S-cone increments was unexpected.

FIG. 3. Rod influence on tritan discrimination as shown bydifference in d9 values (Dd9) for dark-adapted and bleachedconditions. Positive values indicate rod enhancement. Neg-ative values indicate rod impairment.

FIG. 2. Detectabilities (d9 values) calculated from signal-detection procedure for (top) 8° field and (bottom) 1° field forboth S-cone increment (1 or Inc) and S-cone decrement(2 or Dec) conditions. Data for individual observers areshown at left and center and for means (6 1 s.e.) acrossobservers at right. Rod influence on tritan discrimination isshown by differences between dark-adapted (DA, filled) andbleached (BL, open) bars.

Supplement Volume 26, 2001 S67

Because rod stimulation levels changed so little with thechromaticity shifts used in the present study, and the incre-ment/decrement rod effect asymmetry was so pronounced,it seems unlikely to be related to differences or artifactsbased on rod difference thresholds. Further work is neededto uncover the basis of the asymmetry.

The present results also sharpen the apparent discrepancybetween the impairment of chromatic discrimination thatrods can produce in normals and reports of enhancement ofchromatic or wavelength discrimination in S-cone mono-chromats.11 However, a reconciliation of these apparentdiscrepancies may still be possible. For S-cone monochro-mats, rod and cone excitations can simultaneously changesubstantially, sometimes even in counterphase, as wave-length changes. However, for normal observers, the con-straint of keeping LM-cone stimulation constant results instimulus situations in which rods happen not to be sostrongly modulated and may instead provide an effectivelysteady signal. As is embodied in the Boynton and Kambe14

model of chromatic discrimination, visual signals may in-fluence discriminability in either an opponent or a nonop-ponent fashion. Increasing a nonopponent influence alwaysreduces discriminability, but increasing an opponent influ-ence can either increase or decrease discriminability, de-pending on whether that influence increases or decreases thepolarization of the opponent site. Possibly the studies to datewith normal observers have revealed only a nonopponentimpairment of chromatic discrimination due to rod signals.Perhaps rod signals might also enhance discriminability innormals by reducing opponent polarization, if they could bemore strongly modulated.

Finally, although it is unconventional to comment on thepercepts underlying chromatic discrimination judgments,we want to point out that our observers phenomenal reportsraise the possibility that the 8°-field discriminations weremore clearly based on hue changes than were the 1°-fielddiscriminations (even though the cone-plateaud9 valuesindicate that the chromaticity changes were more detectableto cone-mediated vision for the small field than for the largefield). By their nature, chromatic discriminations made un-der different conditions may be based on changes alongdifferent perceptual dimensions, which may in turn reflecteither quantitative or qualitative differences in the contribu-

tions of underlying substrate pathways. Perhaps rods haveless influence on the discriminability of S-cone-mediatedhue changes, which may have dominated on 8° fields, thanon the discriminability of saturation or brightness changes,which may have dominated on 1° fields. Hopefully, futurestudies can assess the perceptual specificity of rod influenceon chromatic discrimination and help us understand itsrelationship to the interaction of rod and cone signals inspecific neural pathways.

ACKNOWLEDGMENTS

Supported by NIH grant EY03221.

1. Buck SL, Knight R, Fowler G, Hunt B. Rod influence on hue-scalingfunctions. Vis Res 1998;38:3259–3263.

2. Knight R, Buck SL. Time courses of multiple rod influences on hue.Vis Res, to appear.

3. Buck SL, Knight R. Time-dependent changes of rod influence on hueperception. Investigat Ophthalmol Vis Sci 1998;39:S160.

4. Knight R, Buck SL. Rod influences on hue perception: Effect ofbackground light level. Col Res Appl Suppl 2001;S60–S64.

5. Buck SL, Knight R. Modeling rod contributions to extrafoveal hueperception. . In: IS&T/OSA Optics in the Information Age. Spring-field, VA: IS&T; 1997. p 16–18.

6. Buck SL, Knight R, Bechtold J. Effect of rod stimulation on uniqueand binary hue judgments. In: IS&T/OSA Optics in the InformationAge. Springfield, VA: IS&T; 1997. p 11–15.

7. Buck SL, Knight R. Opponent-color models and the influence of rodsignals on the loci of unique hues. Vision Res, to appear.

8. Buck SL. What is the hue of rod vision? Col Res Appl; 1999 ICVSProc, 26:S57–S59.

9. Nagy AL, Doyal JA. Red-green color discrimination as a function ofstimulus field size in peripheral vision. J Opt Soc Am 1993;10:1147&1156.

10. Knight R, Buck SL, Fowler G, Nguyen A. Rods affect S-cone dis-crimination on the Farnsworth–Munsell 100-Hue Test. Vision Res1998;38:3477–3481.

11. Reitner A, Sharpe LT, Zrenner E. Is colour vision possible with onlyrods and blue-sensitive cones? Nature 1991;352:798&800.

12. DeMarco P, Pokorny J, Smith V. Full-spectrum cone sensitivity func-tions for X-chromosome-linked anomalous trichromats. J Opt Soc Am1992;9:1465–1476.

13. Knight R, Feyma J, Fowler G, Shepard D, Buck SL. How do rods andS-cones interact in hue discrimination? Opt Photonics News 1998;9:74.

14. Boynton RM, Kambe N. Chromatic difference steps of moderate sizemeasured along theoretically critical axes. Col Res Appl 1980;5:13–23.

S68 COLOR research and application