Brightening prospects for understanding the neural coding of perceived luminance(Invited Lectures at the 27th Annual Meeting)

(1)

The Japanese Psychonomic Society

NII-Electronic Library Service The JapanesePsychonomic Society

TheJopanese

.Jburnai

of

.Rs),chenomic

Science

2eQ9,Vol.2g,No,1.,95-!06

Lecture

Brighteningprospects

for

understanding

perceived

luminance

MarkE.

McCouRT*

and

Barbara

Nbrth Dakota

State

VdeiversiCy

the

neural

coding

of

BLAKESLEE*

*

Along

with color, depth and motion,

brightness

isa fundamental quality of vision, and understanding the neural mechanisms of

brightness

perception

is

a topicof

intense

interest

and

controversy, both histortcallyand incontemporary vision research. With few exceptions, modern

textbooks still_promote the

fiction

that

brightness

induction

(e.g.,

simultancous

brightness

con-trast)results

from

lateral

inhibition

in

isotropic

filters,

such as the cireularly-concentric

fields

found inthe retina. However, because brightness induction occurs over visual angles farinexcess

of the

dimenstons

of

individual

retinal receptive

fields

(up

to

10 degrees

visual angle), "fill-in"

accounts of brightness were _proposed

based

on cortlcal mechanisms.

A

second

historical

challenge

to

retina],accounts of

brightness

induction was

White's

effect,inwhich

the

brightness

of mid-gray

patches situated on the dark and

bright

bars

of a square-wave _grating was opposite to that predicted by the output of circularly-concentric receptive fields.White's effect was a watershed

event which caused spatia]

filtering

accounts of

brightness

_to

be

abandoned, and encouragedi the

develepment

of

high-level

theoriesof

brightness

perception

based

largely

on

Helmholtzian

idea

of

"unconscious _inference". _Through ₂₅

year of systematic analysis of the _grating induction effect

Barbara

Blakeslee

and

I

have

developed

a "second-generatien7'

theory of

brightness

_perception

which

is

based on oriented multi-scale spatial

filtering

which

incorporates

well-known properties of early cortica! processing such as contrast normalization.

We

have

recently applied

the

ODOG

model toevaluate "anchoring"

as an explanation of

lightness

_perception,

with the

_goal

ef clarifying

and synthesizing theunderstanding of brightness and lightness.

Key

words: brightness, grating

induction,

computational modeling

What

is

Brightness?

Brightness,

a]ong with color and motion,

is

a

fun-damental _quality of hurnan vision. Bn-ghtness is

defined

as the attribute according towhich a visual

stirnulus appears to be more or lessintense,or to

emit more or

Iess

light.

Brightness

ranges

from

bright to dim, Unrelated achromatic colors

Cpre-sented alone

in

a

dark

field)can vary only

in

bright-ness

(CIE,

l970).

Brightness

is

correlated with

lumi-nance, especially forunrelated stjmuli, and another

common

definition

of

brightness

is

_perceived

lumi-nance

{Arend,

1993).

The

CIE

adds the _property of

Jightness

toretated achrornatic stimuli

{presented

in

a

display

containingmultiple stimulD.

Lightness

is

the

attribute･according to whieh a visua] stimulus

ap-* _Department _of _Psychology, _Center _for _Visua]

Neuroscience,

North

Dakota

State

University,

Fargo,

ND

58108-6050.

U.S.A.

pears toemit more or

less

light

in

_proportion tothat

emitted

by'

a similar(li

itguminated

area _perceived as

"white". _Thus, _the _CIE _definition

of ]ightnessis

rela-tive

brightness,

Lightness

ranges

from

very

light

or

white, to very

dark

or

b]ack.

Although

unrelated

colors can appear white, only related colors have a

gray or black component and _possess a _perceptual

dimension

(blackness)

that

does

not exist

for

unre-lated

celors.

This

added

dimension

arises through

spatial interactions,revealed in some instances by

induction

effects thatcan occur only

between

{Wyszecki

&

Stiles,

1982;

WyszeckL

1986;

Lennie

&

D'Zrnura,

1988;

Pokorny,

Shevell,

&

Smith,

1991),

71heUtilingofBrightness "tusions

Perceptual illusionsprovide

information

about the

rnechanisms underlying nerma! visual _perception,

including

that of

brightness

and

Iightness.

A

large

(2)

NII-Electronic Library Service

96 The

Japanese

Journal

of Psychonomic

Science

Vol.28,

No.

1

and growing number of

intriguing

brightness

illu-sions have been introduced over the past several

decades; however the number and diversity of

pro-posed explanations

for

these

illusions

is

cumbersome

(Kingdom

& Moulden, 1988; FiorentinL

Baumgart-ner,

Magnussen,

Schil]er,

&

Thomas,

1990;

Gilchrist,

Kossyfidis,

Bonato,

Agostini,

Cataliotti,

LL

Spehar,

Annan,

& Economou,

].999;

Adelson, 2000),

Al-though phenomenal

illusions

themselves are oftcn

argued to support a _particulartheory of brightness

coding, _quantitative

data

based

on experiments

which critically test these claims

is

often lacking.

The

goal of our reeent research

(Blakeslee

&

McCourt,

1997,

1999,

2001,

2e03,

2004,

2005,

2008;

Blakeslee,Pasieka & McCourt, 2005;Blakeslee,Reetz,

&

McCourt,

2008,

2009)

has

been

to remedy these

deficienciesby measuring and modeling the spatial

interactions

between

different

areas of the visual

fieldthrough the _quantitative study of brightness

illusions.

Collecting

quantitative psychophysical

data

on

brightness

Musions

enlarges the

quantita-tivedatabase tocTitically testtheoriesof brlghtness

perception. These

data

inform the continued

devel-opment of amechanistic model of

brightness

percep-tion,the

ODOG

model of Blakeslee and McCourt

(1999).

The

ODOG

model

has

simplified our

under-standing of the rnechanisms underlying

brightness

perception by simultaneously encompassing a large

number ofseemingly

diverse

brightness

_phenomena

with a

history

of

different

explanations. These

ex-planations

include

low-level

spatial

filtering

mecha-nisms-the modern equivalent of lateral-inhibitjon originally _proposed by Mach

{1838-1916)

and Iater

elaborated

by

Hering

(]834-1918);

explanations

in

terms of

T-

and

X-junctions

(Todorovic,

1997;

Zaidi,

Spehar,

&

Shy,

1997};

higher-level

mechanisms

in-volving _perceptual

inferences

about

depth

and!or

transparencyLsuch as these firstproposedi

for

illu-mination by He]mholtz

(1821-1894);

and

explana-tions

in

which

the

key

factor

is

perceptual grouping

(GIIchrist

et aL,

]999;

Ross

&

Pessoa.

2000)

derived

from

the

Gestalt

principle ot "belongingness"

(Gil-christ et aL,

1999;

Kingdom,

1997),

The de'finjngfeatures of the ODOG model, which

include

multiscale spatial

filtering,

orientation

seLec-tivity

and responsc normaljzation, are characteristics

of cortical visual processing

(Rossi

&

Paradiso,

1999;

Rossi, Rittenhouse, & Paradise, 1996;

Gilbert,

Das,

Ito,

Kapadia,

&

Westheimcr,

1996;

Geisler

&

Al-brecht, 1995).

We

contend that explanations couched

in

terrns

of "higher-level"

mechanisms are

not required to explain the rnajority of the wide

variety of brightness

illusions

we have examined

(Blakeslee

&

McCourt,

1997,

1999,

2001),

and that these illusionsare more

_{parsimoniously}

accounted

for

by

the

ODOG

model,

The

explanatory _power of

the

ODOG

model

does

not necessarily confiict with

junction

or grouping analyses, and may actually

rep-resent amechanistic

basis

for

both.

Final]y,

thereare

a nurnbeT of effects thatremain unexplained

by

the

ODOG

modeL

A

careful analysis of these Musions

will help to refine the modeL and todetermine the

circumstances under which higher-lcvel factors do,

in

fact,

exert unique

infiuences

on

brightness

percep-tion,

A

_Spatiat

Nttering

Approach

Simultaneous Bn'ghtness Contrast and

Grating

induc-tion

The brightness of a region of visual space depends

upon the luminance of adjacent regions, Brightness

induction,

in

¢ludes

both

assimilation and contrast

effects. Assimilation occurs when thebrightness of a

test region

becomes

more similar

in

brightness

to

adjacent regions. In_general,assimilation effects

oc-cur

in

complex

displays

with small

(high-frequency)

patterns

(He]son,

1963;

Smith,

Jin,

&

Pokorny,

2001).

Contrast effects occur when the brightness of a test region appears more different in brightness than

adjacent regions.

A

wel]-known example

is

simulta-neous

brightness

contrast

<SBC).

SBC

_produces a

(nearly)

homogeneous

brightness

change within an enelosed test

field

such thata gray patch on a white

background looksdarker than an equiluminant gray

patch on a black background

[Fig.

1(a)].This effect

has been well _quantified with respect to inducing background and testfieldluminance

_(Heinemann,

1955).

Although

SBC

decreases

with

increasing

te$t

fieldsize,brightness

inductien

occurs

for

testfields

as

large

as 100

(Yund

&

Arrnington,

1975).

Since

this

(3)

NII-Electronic Library Service The JapanesePsychonomicSociety

M.

E. McCouRT

and

B. BLAKEsLEE:

Neural

coding of _{perceived1umlnance}

97

dilstancefar exceeds the

dimensions

of retinal or

LGN

receptive fie]dsinmonkey

(DeValois

& Pease,

1971;

Yund,

Snodderly,

Hep]er,

&

DeValois, 1977;

'li'

"',i.i M

l/-･-1,･

-･

'

.sE-E'RruE[et-=en9eoo=cu.=-E="o==o-evE=cueE

e.Ts e.se O.25

DeValois

&

DeValois,

1988),

a common explanation

for

SBC

has

been

thatthe

brightness

of the test field

isdetermined

by

the

information

at the edges oi the

75 50 25 o

･25

-50

-75

75 50 25 owa

-25so

60LE

-T5LL

75el-sosdi 2Sat o

-25

-sg

-TS

O.75 O,50 O,25

,t,/.!gtt/

igk

rmk."

(f)

}･･i_"1,, t/.. rdiEE

'X',,x,,

l"''''"''''"''"'11

.,,1'

1,,,,ll

l.,/

{t

l

4/l･K x'L,kI,,x/" e,75 O,50 O,25 D.75

(j)

H _32o

'fi'

k1

(k)

t }7 O,50 O,25

o256 5ri2 76e tD24

755D25

a-25.50-75o256

512 76S t024

Spatial

Position

(pixels}

Fjgure 1

{a-d)

Four

of the stimuli used to measure the effect of test

field

width on induction

magnitude.

Display

width

is

320;

test

field

widths of

la,

60,

!20

and

320

are

illustrated.

Test

field

height is10. Sinewave inducing contrast was constant at

O.75.

Test

field

luminance

was set to the

rnean of the display

(50

cdlmZ), Note that _panel

(a)

isa

"c]assical"

simu]taneous brightness contrast

(SBC)

stimulus

(i.e.,

_two IDX10 test fields),and that _panel

(d)

isa standard _grating induction

{GI)

stimu]us

(i.e.

a continuous test

field

spanning

the

display).

(eAh)

Point-by-point

brightness

matches

(at

O.250

intervals)

across the test

fields

of

displays

il]ustrated

in panels

{a-d).

Open symbols _plot mean

brightness matches made tothe testfields

(proportion

mean Iuminance ± ₁s.e.m.);

filled

symbols

in

(e)

are brightness matches to the

inducing

grating.

The

light

gray

line

depicts

the veridical

luminance

profileof the stimulus dispLay along a horizontal linethrough the vertical center oi the test

fie]d

and

display.

<i-1)

Solid linesrepresent slices taken through the

ODOG

mode[ filteroutput ioreach o"he

stimulus disp]ays in_panels

(a-d),

The lightgray ljneagain depictsthe veridical luminance profile of

the

$timulus display

taken

at the vertical center of the test

field.

Notc the excellent qualitative and

quantitative agreement between the ODOG model output and the corresponding poinVby-point

brightness

rnatching

data,

The model captures the magnitude and structure of brightness induction

within the homogeneeus test fields

Ci-1),

as well as the brightness of the inducing _grating itself

(D,

SBC and GI are thus demonstrated to be congruent phenomena, which are both accounted forby the

(4)

98 The

Japanese

Journal

of

PsychonormcScience

VoL 28,No. 1 bounded region and

is

subsequently

fi11ed-in

or

as-signed to theentire enclosed area

(Shapley

&

Enroth-CugelL

1984:

Cornsweet

&

Teller,

1965;

Grossberg

&

Todorovic, 1988; Paradiso & Nakayama, 1991;Rossi & Paradiso, 1996;Paradiso

&

Hahn,

1996;

for

review

see

Kingdom

&

Moulden, 1988). Evidence suggests

that this explanation istoo sirnple, and that

_distal

factors

must _play a role

(Arend,

Buehler.

&

Lock-head, l971;Land & McCann, 1971; Heinemann, 1972;

Shapley

&

Reid,

1985;

Reid

&

Shapley,

1988).

Grating induction

(GI),

unlike SBC, isan induction effect that_produces a spatial brightness variation

(a

grating)

in

an extendecl test

field

[Fig.

I(d}].

The

perceived contrast of the induced _grating decreases

with

increasing

inducing

_grating

frequency

and with

increasing test fieLdheight

(McCourt,

1982},such

that

canceling contrast

is

constant

for

a constant product of inducing frequency and testfieldheight

(McCourt,

1982; Folcy & McCourt, 1985).

GI,

like

SBC,

extends over

large

distances,

since

it

is

still

observed intestfieldsat leastas largeas 6n

(Blakeslee

&

McCourt,

I997).

Unlike

SBC,

however,

homogene-ous

brightness

fi1]-in

cannot account for

GL

Several

brightness models have been

proposed

that

incorpo-rate non-hornogeneous

fill-in

mechanisms

(Grossberg

& Mingolla, 1987; Pessoa, Mingolla,

&

Neumann,

1995),

however,

these

mode]s

huve

not been

success-fully applied to _grating induction. Blakeslee and

McCourt

(1997)

suggested

that

GI

might

be

under-stood

in

terms of the output of paralle]spatial

filter-ing across mu'!tiple spatial scales

_(Moulden

&

King-dom,

1991>.

An

attractive featureof thisapproach is

that

both

the

low-pass

spatial

frequency

response of

GL and the invariance of

induction

magnitude with viewing distance

(i.e.

the tradeoff between the

ef-fect$of inducing _grating spatial frequency and test

field

heighO,

are

both

_{parsimoniously} explained by

multiscale spatial

filtering.

Despite the

fact

that

SBC

is

typicallyconsidered a

homogeneous

brightness

effect

dependent

on a

fi11-in

mechanism, whereas _grating induction _possesses spatial structure wbich cannot bc produced by a

homogeneous filt-inmechan,ism, ithas bcen

sug-gest.edthat eithcr SBC isa special law frequency

instance ef _grating

induction

(McCourt,

1982), or

that

_GI

isa _particularcase of

SBC

(Zaidi,

1989;

Moul-den

&

Kingdom,

1991).

Blakeslee

and McCourt

{1997)

asked whether the mechanism(s) underlying

GI could also account

for

SBC,

or if

fundamentally

different

mechanisms were required toexplain these

two effccts. The structure and magnitude of

induc-tion

in

both

GI

and

SBC

stimuli were measured

wherc the

inducing

¢onditions for the two effects

were rendered as similar as _possible

by

employing

one cycle of a

low

frequency

sinewave grating as the

inducer, Test fielddimensions spanned a range that

incorporated

both

c]assic

SBC

and

GI

configurations

[see

Figs.

ICa-d)],

At

each of threetest

field

heights

(1=,

30 and 6C),_{point-by-point} brightness matches

{Heinemann,

1972;

McCourt,

1994}

were obtained at

intervalsof O.250,fortestfieldwidths of 320

{the

GI

condition),

140,

120,

8e,

60,

30

and

IC

[Fig.

1(e-h)].

Potnt-by-point brightness matches were analyzed to assess systematic changes ininduction structure

(i.e.,

departures

from

the sinusoidal

brightness

variation

seen in320 wide testfieldGI condition} and in the

average magnitude of

brightness

and

darkness

in-duction within the test

fields,

as a

function

ot test fieldheight andi width. In the widest test fields

induction

structure was well accounted

tor

by

the

sinewavc _pattern observed inthe

GI

condition. As

test

field

width

decreased

thesinewave amplitude of

the induced structure

in

thetest

field

decreased

_Ci.e.

the

pattern

flattened),

and eventually

bccame

nega-tive

(i.e.,

showed a reverse cusping) at the narrower

testfieldwidths. Both the structure and magnitude

of brightness induction as a function of changing

test

field

1iejghtand width were parsimoniously

accounted

for

by the output of a

differentially

w･eighted, octave-interval array of seven

difference-of-Gaussian

CDOG)

filters

[Fig.

1(ILI)],Thls array oi spatial filtersdiffered from those _previous]y

em-ployed to rnode] spatial vision in that itinc]uded

mechanisms tuned to much lower spatial

frequen-cie$.

We

postulate

thal

such

filters

exist at _those

levelsof thevisual system where

brightness

_percepts

are determined.

Recent

evidence shows _that a sig-nificant number of cells

in

cat and monkey _primary' visual cortex respond ina manner correlated with

brightness over distances farlarger than thesize of

(5)

M

E. McCouRT

and

B. BLAKEsLEE:

Neura1

coding of perceived luminance 99

their "classica]"

receptive

fields

(Rossi

et a].,

1996;

Gi]bertet al. 1996).Thus lateralinteractionsat early

levelsof visual _precessing, or feedback from hierar-chically

higher

processing areas

(Larnme

&

Roelf-sema, 2000),may allow the integration of brightness

information

over

]arge

regions of visual space,

This

simplc

filtering

explanation can also be _generalized

to account forother brightness phenomena

includ-ing

Zaidi's

₍₁₉₈₉₎

GI

demonstrations

showing

both

localand distal effects;

Shap]ey

and Reid's

(1985)

contrast and assimilation

demonstration

modeled as

due totheintegration of localcontrasts across space;

and theinduced spots seen at the street intersections

of theHermann

Grid

classically explained

in

terms of on- and off-center receptive fields

(Fiorentini

et a].,

1990).

Thus,

the model of

Blakes]ee

and

McCourt

{1997)

brings

together with a common explanation a

variety of seemingly

diverse

brightness

phenomena

with a history of differentexplanations that include

local

spatial

filtering,

filling-in

and edge

integra'tion.

IVhitels

EtTectand fk)dorovids SBC

Denvonstration

B]akeslee

and

McCourt

{1999)

subsequently

ad-dressed a _group of effects, including the White effect

(White,

1979)

[Fig,

2(a)],

and a

SBC

demonstration

of

Toderovic

(1997}

[Fig.2(b)],

which cannot be

ac-counted forby isotropiccontrast models such as the DOG moclel or

fi11-in

models. IntheWhite effect. _gray

tesL _patches of identicaL

]uminance

_placed on _the

black

and white

bars

of a square-wave _grating

ap-pear differentin bTightness,

What

makes the effect

so

interesting,

however,

is

that the

direction

of the

brightness change isindependent of the aspect ratio of the test_patch. Unlike

SBC

and GI,theWhite effect

does

not correLate with theamoun't of

black

or white

border in contact with the test_patch, For exarnple,

when

the

_gray _patch

is

a vertically oriented

rectan-glesittjng atop a white stripe of a vertical grating,

it

has

two short sides that are

in

contact

(above

and

be]ow)

with the coaxia! white bar upon which itsits,

and two leng sides

(left

and right) thatare incontact with theflanking black bars

[see

Fjg.2(a)].Thus the

testpatch has greater contact with the dark fianking

bars _yet itappears da'rkerthan a simiLar _gray _patch

fianked

by

white

bars.

This

is

not an assimilation

:II'

i

e

sEIiGii

tEI:fi

Figure 2,

_(a)

An

example ot the

White

lus.

(b)

Todorovic's

variation of a

ous brightness contrast stimulus, In both

stimult the _gray test _patches are

minant,

but

appear

different

in

brightness.

These

effects are not explicab]e

by

contrast

models utilizing isotropicspatial filters.After

Blakeslee and McCourt

{1999).

effect since even ifthe height of the test _patch is

reduced so

it

has

more extensive

border

contact with

the bar on which itjssitting

(i.e.

the coaxial white

bar),

the

direction

of the effect

is

unchanged

(White,

1979, 1981), White

{1979)

cencluded that

explana-tions

{whether

contrast or assimilation) that

depend

simply on the relative amounts oi

black

and white surrounding the_gray elements could not explain the

effect,and that

directional

(orientation)

properties of

the inducing _grating must be importanL A number of _qualitative filteringexplanations have been

offered

for

the

White

effect.

White

himself

proposed

a mechanism called "pattern-specific _inhibition"

{White,

1981L

based

on the notion that elongated cortical fiLtershaving similar preferred orientation and spatial

frequency

selectivity, and which received

their input from adjacent retinaL locations,might

tend to inhibit each other and thus produce the

effect Foley and McCourt

(1985)

suggested that

hypercornplex-like cortical filterswith small centers and elo'ngated surrounds might be responsible for

(6)

100 The

_Japanese

_Journal

of Psychonomic

Science

VoL28,

No.

1

dual mechanism model toexp]ain the results of an

investigation

in

which they varied the

height

of

both

the

fianking

and coaxial

inducing

bars. They

con-cluded that a localmechanism, mediated

by

circu-larly

symmetric center-surround receptive

fields,

op-erated a]ong the

borders

of the test _patch and

pro-duced

a _particularlystrong signal at thecorner

inter-sections of the testpatch with the coaxial bar.

Ac-cording totheir model itis_thiscorner signal that

_in

some

<unspecified}

manner

disproportionately

weights the coaxial bar relative to the fiank and

induces brightness

into

the testpatch,

Additionally,

they proposed that a more spatially extensive mechanism was required

to

allow the coaxial

bar

te

exert an

infiucnce

on the

brightness

of thetest

_patch

throughout

its

length. This mechanism issimilar to thatproposed

by

Feley

and

McCourt

(1985).

Other

exp]anations of the White effect are based

on higher-order _perceptual inferences involving

depth and!or transparency, and the

Gestalt

notion of

"belongingness"

(Agostini

&

Profitt,1993; Taya,

Ehrenstein & Cavonius, 1995; Spehar, Gilchrist

&

Arend,

1995;

Anderson,

1997;

Ross

&

Pessoa,

2000).

According to the

Gestalt

approach, _perceptual

or-ganization

{such

as relative

depth

relations

in

the

Whjtc stimulus) influence brightness contrast such

that surfaces _{predorninantly} interact

(contrast)

with other surfaces wjth which they are _grouped.

Agostini and Profitt

(1993)

and Gilchristet al,

(1999')

argued that

in

the

White

effect thetestpatch appears

]ighter

(or

darker) when itison the black

(or

white}

bar

because

oi the phenomenal

impression

that

it

Ltbelongs

_to"

or has been

`Lgrouped

with" that bar.

According to Gilchrist et aL

{1999)

the _grouping

factor at work

here

i$

the

T-junction.

which

is

thought tosignal

depth

through occlusion.

Of

note,

however, isthat both Zaidiet aL

(1997)

and

Todoro-vic

(1997)

argue that while an explanation

based

on an analysis of local

junctions

inthe stimulus, spe-cifically T-junctions,can account for"rhite'seffect, it

does net require that T-junctions contribute to per-ceptual organization.

Although both theT-junction and _grouping

ana]y-ses offer useful rules for_{qualitatlvely}_predicting _the appearance of various

brightness

effects, they

fall

(a)

(d)

(b)1234ss7

Ce)(D

(g)

2 s

*as-me

s

*ma-ma

tw*me-me

amil

*

ma=wa

(tztzb*ew-me

%*pt-ew

ipg･;tt'i'E /･ti(c) /.n:l:IobOGCerrte;Fr;nyuenc.y'telgeg/

(h)

z

Figure

3.

A

diagrammatic

representation of

the

Oriented

Difference-of-Gaussians

(ODOG)

model of

brightness

erception,

(a)Illustra-tion of a two-dimensional riented

difference-of-Gaussian

(ODOG)

filter.

(b)

Seven

filtcrs,

with center

frequencies

spaced at octave

intervals,

are summed within orjentation

after being weighted across frequcncy

(c}

using a _power

functjon

wtth a slope of

O.1.

(d)

The

resulting six multiscale spatial

filters,

one for each orientation, are convolved with stimuli of interest

(e>,

in_thisinstance a

White

stimulus.

The

convolution outputs

(/t)

are

pooied

across orientation according to their

space-averaged root-mean-square

{RMS)

tivity level

_(g)

to _produce a resultant output

(h),

After

Blakeslee and

McCourt

(1999).

Table

1

Differenccof Gaussian space constants

Filter

Space

constant

Center

Surround

1234567

.0470.094D.18sC.37sO

･7so

1.so 3o

.Og3n.18so.37so

-7sz

1.sc 3e

6c

NII-Electronic

(7)

M.

_E.

McCouRT andB. BLAKEsLEE: Neural coding of _perceived luminance 101

short of

identifying

an underlying mechanism.

B]akeslee and McCourt

(1999>

were able teprovide

such a mechanistic exp]anation in the form of an

oriented-difference-of

Gaussians

{ODOG)

modet.

The

oriented filtersof theODOG rnodel were _produced

by

setting the rati,oof

DOG

center/surround space

con-stants to 1:2 inone orientation and to

1:1

in

the

orthogonal orientation

[Table

1].

A

_gray leve],

repre-sentation ofsuch an

ODOG

filter

appears

in

Fig.

3(a}.

The center iscircular and thesurround extends be-yond theeenter

for

a

distance

of twtcethecenter size

inone or,ientation but isthesame sizeas thecenter

in

the orthogonal orientation, These filterscan be

de-scribed as

Gaussian

blobs

with

inhibitory

fianksor as

simp]e-like cells

{such

as those

found

in

the cortex of

monkey or cat) that are orientation and spatial

fre-quency

selective. The ODOG mode]

is

irnp]emented

jn

6

orientations

(O,

30,

60,

90,

-30

and

-60

degrees).

Each orientation

is

representecl

by

seven

volume-balanced filtersthat _possess center frequencies

ar-ranged at octave intervals

{from

O.1-6.5c/d). The seven spatial

frequency

filters

[Fig.

3Cb)]

within each

orientation are summed after weighting across

fre-quency using a _power

function

with a slope of

O,1

[Fig.

3(c)],This slope

is

consistent with the shallow

lew-frequency

fall-off

of thesuprathreshQld contrast

sensitivity

function

that

is

expected to

be

associated

with the high-contrast stimu]i under investigatien

(Georgeson

&

Sullivan,

1975).

The

resulting six rnultiscale spatial

filters,

one _per ortentation, are

con-volved with the stimu]us of tnterest

[Fig.

3(d-e)j,

'The

filter

outputs

[Fig,

3{f)]

are pooled across

orienta-tion according to their space-averaged root-mean-square

(RMS)

activity

leveL

as computed across the

entire irnage,The pooling isinaccord with a simple

response norrnalization in which the filteroutputs

are weighted such that the

RMS

contrast

in

the

"neural _images"

across orientation channels are equated

[Fig.

3(g)].Response nonlinearities found

in

neurons

in

cat and rnonkey visual cortex, such as contrast _gain control and the rapidly accelerating

increase

inresponse at low contrast and saturation at

high

contrast, may represent the physiological

sub-strate forthis type of response normalization

(Geis]er

&

Albrecht,

1995).

Note

thatwhen the

filters

of the

ODOG

model are linearlysummed across the

full

range of orientations within each spatial

frequency

these filterscombine to produce a

DOG

filter.

Thus

the DOG model of Blakeslee and McCourt

(i997)

is

simply a subset of the

ODOG

mode]

in

which the

filter

outputs are

linearly

pooled,

As

mentioned

_previously,

the

defining

featuresof

the

ODOG

model e.g., rnultiscale spatial

frequency

filtering,

orientation selectivity and response

nor-malization, are response characteristics thatare rou-tinely observed at early cortical stages of visual proc-essing

in

both

cat and rnonkey

(Rossi

&

Paradiso,

1999;Rossi et al. 1996;

Gilbert

et al. 1996;

Geisler

&

Albrecht, 1995). ILis specifically the addition of orientation selectivity and response norma]ization,

however, that allows the model to accoun,t

for

an-isotrepic

effects such as the

IVhite

effect.

An

intui-tive sense

for

the model can be obtained

from

exam-ining

Fig.

3(d-f),

When

the

long

axis of the

multis-cale

ODOG

fi]ter

!svertical, as

it

is

in

theorientation represented by the top row of Fig.3(d-f),the

convo-lution

output of this

filter

with the

White

stimulus

shows the _greatest activity

in

the region of the test

patches and _produces the

White

effect.

Although

the

top and

bottom

edges of the

inducing

grating are

also a good stimulus for _this filter,the inducing

grating

itself

is

not.

This

situation

is

largely

re-versed in the convolutjen output of the multiscale

filterwith ahorizonta]orientation, represented

in

_the

fourth

row of

Fig.

3(d-f).

Here

the activity _generated

by theinducing _grating is

high

compared tothatfor

the testpatches.

Added

together,

however,

thesetwo

filterorientations represent both the test_patches and

the

inducing

grating,

Response

normalization _prior

to summation simply weights the

foatures

extracted

by these two filtersequally. This _prevents high

contrast

features

Csuch

as the

inducing

grating)

cap-tured at one orientation from swamping ]ower

con-trastfeatures

{such

as the test_patches) captured at another orientation.

Blakeslee and McCourt

(1999)

showed that the

ODOG

model _{qualitatively} _predicts the relative

brightness

of

the

testpatches

in

the

White

effect, the

Todorovic SBC demonstration. GI and SBC, and

quant,itativelypredicts the relative rnagnitudes of

(8)

102

The

Japanese

Journal

of Psychonomic

Science

VoL

28,

No.

1

A=caeE[.g=oa9eeocre.aE=Jmc=ovreEcreeE O,3O,2o,ae,e-O,1-e,2-O.3 o .. MM

ma

$

ts

60 30 O.3e,2o]o,e-e,t-O,2-O.3 o .. BB

ee

_as

ma

A c co ¢ DE E

2

--3o

g

:-; cr ) ¢

-so

3!.

-60 = m

1

o

L3e

.o-.

r. L m

.zo

rs

-di

or-30

-6e

SBC3 SBCI Gi3 Gerl W4 W2 T

Stimulus

CendMen

Figure 4. The bar _graph

_Cleft

ordinate) _plots

the

deviatien

of mean rnatching

luminance

from

the mcan

luminance

(as

a _proportion of mean luminance)

for

various brightness stimulus conditions, The error

bars

are

95%

confidence limits.Data from two subjects appear in the upper and lower _panels.

Condition

SBC3

refers to

brightness

matches obtained

in

simultaneous

brightness

contrast conditions

[Figure

la] where test_patch height and width werc 30;condition

SBCI

refers to

le

test_patches, The

bars extending above the mean represent brightness matches fer test

_patches

on the dark background

(whtch

appear

brighter

than the rnean), while the bars extending below the mean represent the test

patch rnatehes on the bright background

(which

appear darker than

the

mean).

for

two GI displays:a O.03125 cfd sine wave

inducing

_grating with a test

field

height of

3"

(GI3L

and a

O.l25 c/d sinewave

inducing

_grating wlth a test

field

height of 10

{GII).

The conditions labeled W4

and

W8

_plot the magnitude of the

White

effect

for

a O,25 cfd and a O.5 cfd square wave

inducing

grating,respective]y. For the O.25 cfd inducing _grating,test_patch width was

2a

and testpatch

height

was 40, For the O,5 c7d inducing _grating,test _patch width was

10

and testpatch

height

was

20. Note

that for these two conditions the

bars

extending above the linerepresent matches to test _patches

located

on the

dark

bars

of the

inducing

_grating while those below the lineare matches

to

the

test

patches

located on the bright bars of the inducing _grating. The finalcondition,

{T),

_plots the

magnitude of brightness induction inthe

Todorovic

stimulus

[Figure

2(b)].

The

bar

extending above

the

mean

lurninance

represents the match tothe testpatch on thedark

inducing

background with the

overlapping white squares. The bar extending be]ow the mean isthe match

to

the

test

_patch on

the

white background with the overlapping black squares. Inducing _patterns of 100% contrast were used

in

all

brightness

displays.

The

symbols are read against the Tight ordinate andi represent the ODOG

rnodel outputs tothe test

fielcls

in

each stimulus condition, The

fi11ed

symbols are the _predictions for

the matches that appear as dark bars and the open symbols are the _predictions forthe matches that

appear as white

bars.

The

ODOG

model output and the ernpirical

brightness

matching data are clearly

similar across a wide variety of brightness phenomena. After Blakes}ee and McCourt

(1999}.

these brightness effects as measured psychophysi-cal]y using brightness matching

[Fig.41.

This mechanistic explanation

does

not necessarily confiict

with T-junction or _grouping ana]y･ses,

but

may, at

]eastto some extent, serve as a mcchanism for

both.

Indeed, tQ the extent that

junctions

infiuence

"higher-]evel"

grouping, and to theextent that filters

of the

ODOG

rnode] capture the eperations of

iunc-tions and grouping, one might expect all these ap-proaches to yteld similar results

(Todorovic,

1997;

(9)

M. E.

McCouRT

and

B. BLAKEsLEE:

Neural

coding of perceived luminance 103

A

stdi

:s2.2

gee-,g.=:v==m--)

ep =S 32302S2624222018 il･l-I . t'TT

TT"'."

,

. ri44140 1136-a. oa32 tst28

:-ord24

.)m

G120 i or116aa2 o 3e 6e ge no ase lse

Difference

in

Phase

BetweenTestPatch

anci BlackBar

(degrees)

Figure 5,

Judged

lightness

(fi11ed

symbols,

left

ordinate), replotted from

White

and

White

(1985},

as a function of test _patch spatial

phase.

Open

symbols _ploL _predicted test patch brightncss from

ODOG

model output

averaged across

the

width of the

test

patch

(right

ordinate).

ODOG

model output

ly predicts the linearphase-brightness

tionship reported by White and White

(1985>.

After Blakeslee and McCourt

(1999).

Blakeslee

&

McCourt,

1999).

The

ODOG

rnodel has

the advantage,

however,

in

that

it

makes

quantita-tive predictions about the relative size of various

brightness effects and _provides an explanation

for

a

larger variety of brightness effects. For example,

SBC

and

GI

do

not contain

T-junctions

or

X-junctions

and cannot be addressed

by

a

junction

analysis

(B]akeslee

& McCourt, 1999). There isalso

no explanation for

GI

based

on either

Gestalt

group-ing or

GilchrisVs

anchoring

hypothesis

(Gilchrist

et

al.,

1999).

In

addition,

Blakeslee

and

McCourt

(1999)

showed that the

ODOG

model accounts for the

smooth transition in mean brightness seen in the

White

effect

[Fig.

5]

when the relative phase of the

test _patch isvaried relative tothe inducing _grating

(White

&

White, 1985}

_[Fig.

6(a-D].

This smooth

transitLon

is

not readily exp]ai,ned

by

a

T-junction

or

grouping analysis, Fi,nally,point-by-point

bright-ness matching revealed brightness variations across

the test _patches of White stimuli

[Fig.6(g-1)]

(Blakeslee

& McCourt, 1999),as we]1 as GI and SBC

stirnuli

(Blakeslee

&

McCourt,

1997>

that

accord with

ODOG

model predictions.

Only

spatial

filtering

can easily account

for

these types of

brjghtness

gradi-ents.

Conclusions

and

lileiture

Directions

It

is

clear that _the

_ODOG

model can account fora

large

number of diverse

brightness

effects that

have

previously been explained by a Nny'idevariety of

djffe-rent proposed

brightness

mechanisms,

These

expla-nations include low-level iiltering,

fi11-in,

edge

inte-gration and

junction

ana]ysis, as well as higher-level mechanisms

involving

perceptual

inferences

about

depth and/or transparency, and explanations

in

which the

key

factor

is

visual _grouping

based

on

such concepts as the Gestalt _principleof "belonging-ness.'' The fact that all of the induced brightness

effects reviewed here can

be

parsimoniously ac-counted

for

by

the

ODOG

model suggests

that

these

particulareffects

_prtmarily

reflect the operations of

early-stage cortical filtering,and that explanations

in

terms of '`higher-level"

grouping mechanisrns are not

required.

There may, however, be other situations where

higher-order effects on

brightncss

do

occur.

For

example, several clajms have

been

made

(including

our own>

for

an effect oftransparency on perceived brightness

_(Adelson,

1993; Anderson, 1997;

King-dom,

Blakeslee,

&

McCourt,

1997).

In

the

interests

of

parsirnony,however, careful stud}, isrequired to

de-termine thecircumstanees under which higher-order

factors,

such as transparency, exert a uniqzae

influ-ence on brightness,and todetermine the magnjtudes

of these

_higher-order

effect$.

For

example,

in

a

care-fully

controlled study,

Kingdom

et al,

(1997)

demon-strated a small effect of _perceived transparency on

the brightness of the testpatch in a

SBC

stimulus.

Multiplicative transparency affected brightness in

such a way that subjects perceived the testpatch te be brighter than inother configuratton conditions.

Somewhat

surprisingly, this

_is

consistent with an

explanation whereby the transparency was _partially

discounted Erom the brightness of the test patch.

Carefu]ly

sorting out those brightness effects that

are and are not accounted forby low-level mecha-nisms, as well as rneasuring their relative

magni-tudes, will _provide needed

direction,

_precision and

insight into the

investjgation

of brightness

(10)

104 The

Japanese

Journal

of Psychonomic Science Vol.28,No. 1

256 a92 128 64 o256 t92 12B 64 e256 t92 t28 64 o

rl

{d)

aili･mu

O.7 O.5-E=E'Rtu e.3E=.9= O,70ocrEee2 O,5ow.EE="a) e,3eec=stu e,7o==cu ¢ =

MM

i i

(g).t.7rpm.tt..t...t...t.

BB

c a

-.-

..--J

-=CL-JoL

¢ sttam>=cuInt iiilI{l

ligi

fiIiilll

IIIll

i

{e)Ii';

.l'

g}

{k)

ri60 "28 ge

{

(h)

Figure 6.

_(a-c)

The

to the inducing relationship with

testpatches

have

been

s

by half

[Fig.

5], eliminating the effect,

displays

taken

linesrepresent views of the ODOG

matches

{with

symbols, as read

output closely

After Blakeslee and McCeurt

(1999).

e,50

_l

'

(l]

O,30 o 2s6 512 768 1024 2s6 St2 76B Spatia[PesitEen{pixel$}

White stimulus Mustrating

the

effect of shifting

the

ph

grating.

(a)

In the standard configuration the gray Lhe

black

bar,and a 180fiphase re]ationship with the

hifted

tothe right

by

450 phase angle; this

The test patches in panel

(c>

have been shitted by 900

(d-f)

The lightgray lines

depict

the veri

'

along a

horizontal

line

through the vertical center of the

corresponding slices taken through the

ODOG

model

rnodel output

(solid

lines)illustratedin _panels

<d-f),

an

95%

confidence

intervals)

obtained at seven locatiens across

against

left

ordinate).

parallelsthe observed brightness variations across the test

(l}

2565t2T6B

s16o

-a

8 :.t28

.l

-e96

-e

at ri60 n2s 96

ase of

'the

testpatch relative

test patch

is

in

a OO phase

white bar, In panel

(b)

both

reduces the magnitude oi the effect

phase angle, completoly

dical

luminance

profiles

of the stimulus

test

field

and

disp]ay.

Solid

filteroutput,

<g-l)

Magnified

d point-by-point

brightness

each 20 testpatch

(open

Data

from

two subjects

(MM

and

BB)

are shown,

ODOG

model

patches in

these

stimuli.

Acknowledgement

This work was supported by grants frornthe

Na-tienaLScience Foundation

_{NSF:

IBN-0212789), the

National

Eye InstLtute

_(NEI:

ROI EYO14015) and the

NationaL

Center for

Research

Resources

_(NCRR:

P20

RR020151>.

NEI

and

NCRR

are components of the

National

Institutes

of

Health

(NIH).

The

contents of

thisreport are solely the responsibllity of thc authors and do not necessarily reflect the oMcial views of the

NIH or NSF.

References

Adelson,

E.H,

(1993).

Perceptual

organization and

the

judgrnent

of

brightness

Science,

262,

2044.

Adelson,

E,

H. (2000).

Lightness

perception and

ness

illusions.

In

Gazzaniga

M. (Ed.).

The

IVbw

Cog-nitive Neurosciences,

2"d

ed,

Cambridge,

MA: MIT Press.

Agostini,

T. &

Profftt,

D.

R. (1993).

Perceptual

ganization evokes simultaneous

lightness

contrast.

Perception,

22,

263L272.

Anderson,

B. C1997).

A

theory of

jl]usory

lightness

and transparency in monocular and binocular

(11)

M.

E. McCouRT

and

B.BLAKEsLEE:

Neural

codlng of

_perceived

luminance

105

ages:

The

role of contour

junctions.

Pigrception,

26,

419-453.

Arend, L.E.

(].993).

Mesopic lightness, brightness,

and brightness contrast,

Perception

and

Psycho-physics,

54(4),

469-476.

Arend, L,E,,Buehler,

J.

N.

&

Lockhead, G,R,

(1971)

Difference information in brightness _perception.

Percoption

and 1?sychophysics.

9,

367-37e,

Blakeslee,B.,&

McCourt,

M.

E. _(1997).

Simi]ar

mecha-nisms underlie sirnultaneous brtghtness contrast

and _grating

induction.

Vision

Research,

37,

2849-'

2869.

BlakesLee,

B. &

]vlcCourt,

M.

E. {1999).

A

multiscale spatial

filtering

account of the

White

effect,

simul-taneous brightness contrast and grating induction.

VisionResearch, 39,4361-4377.

'

Blakeslee,

B. &

McCourt,

M,

E,

(2001).

A

multiscale

spatial

filtering

account of theWertheimer-Benary

effect and the corrugated Mondrtan. Vision

Re-search,

41,

2487-2502,

B]akeslee,B. & McCourt, M, E.

(2003).

A multiscale

spatial filteringaccount of brightness _phenomena.

In:

Harris,

L.,

&

Jenkin,

M,

(Eds.),

Levels

of

tion,

Springer,

NY,

New

York.

Blakeslee,

B,,

&

McCourt,

M.

E. {2004).

A

unified

ory of

brightness

contrast and assimilation

porating oriented multiscale spatial fi]teringand

contrast normalization. VisionResearch, 44, 2503.

Blakeslee, B.

&

McCourt,

M.E,

(2005>.

A multisca]e

spatial

filterjng

account ef _grating

induction

and

remote brightness induction effects: Reply to

vinenko.

fercoption,

34, 793-802.

Blakeslee,B.

&

McCourt,

M.

E. (2008),

Nearly taneous brightness

induction.

foumal

of

Wsion,

8

(2):

15,

1-8,

http://journatofvision.org18/2115L

doi:10.1167f8.2.15.

Blakeslee,B. Pasieka, SNF.,& ],CcCourt,rv[.E.

_{2005).

Oriented

multiscale spatiaL

filterlng

and contrast

normalization:

A

parsimonious model of

ness induction ina continuum of stimuli including

White, Howe and simultaneous brightness

trast.

Vision

Research,

45,

607-615.

Btakeslee,

B. Reetz,

D.,

&

McCourt,

M.

E. (2008).

ingtoterrnswith ]ightnessand brightness:Effects of stimu]us configuration and instructions on

ghtness and lightness

judgments.

lbormal

of

Vision,

8(11):3,1-14, http://journa]ofvision.orgf8/1113f,

dei:10.ll6718.11.3.

Blakes]ee,B.,Reetz,D.,

&

McCourt,

M.

E. (2009},

tialfilteringversus anchoring accounts of

ness/Iightness _perception instaircase and

neous

brightnessllightness

contrast stimuli.

Ibur

nat

of

Esion, 9(3):22,1-17.http:/fjournalofvision.

org!9/3122Ldoi:10.1167/9.3.22.

CIE

(1970).

international

Lighting

Vocabulary.

3i`Ed,

Publ.

No. 17,

_(El,O.

Paris:

Bureau

Central

de

la

CIE,

Cornsweet,

T.N.

&

Teller.

D. (1965).

Relation

of

in-crement threshelds to brightness and luminance.

foumaat

of

the

QPtical

Society

of

America,

55(10),

1303-1308.

DeValois, RL.

&

DeValois,

K.K.

(1988}.

Spat'ial

Iiision.New York:

Oxford

University Press.

DeValois,

R,

L. &

Pease,

P.

L. (1971).

Contours

and

contrast:

Responses

of monkey

lateral

_geniculate cells toluminance and color figures,Science,171,

694-696.

Fiorentini,

A.

Baumgartner,

G. Magnussen,

S.

Schil-ler,P.H.

&

Thornas,

J.

P,

_(1990).

The _perception of

brightness

and

darkness:

Re}ations teneuronal

ceptive

fields.

InSpillman,

L,,

&

Werner,

J,

S,

(Eds.),

Visual PleTception:The IVeuroPhysiolQgical

tions,

San

Diego:

Acadernic

Press,

Foley,

JM.

&

McCourt,

M. E.

_(1985}.

Visual

grating

induction.

Iburnal

of

the

Qptical

Society

of

America

A,2, 1220-1230.

Geisler.

W.

S. &

Albrecht,

D.

G. (1995).

Bayesian

lysisof identification

performance

inmonkey ual cortex: Nonlinear mechanisms and stimulus

certainty.

Vision

Research,

35,

2723-2730.

Georgeson,

M.A. &

Sul]ivan,

G.

D.

₍₁₉₇₅₎

Contrast

eonstancy: Deb]urring inhuman vision by spatial

frequency

channels.

foumal

of

RhJ,siology

(Lond.),

252,

627-656.

Gilbert,C,D,,Das, A. ILo,M.,Kapadia, M. &

theimer,

G. (1996).

Spattal integrationand cortical

dynamics.

I]roceedings

of

the

Adetional

Academy

of

Sciences,

[ISA,

93,

615-622,

Gilchrist,A.,Kessyfidis, C.,Bonato, F.,Agestini,T.

Cataliottj,

J.,

Lj,

X.,

Spehar,

B.,

Annan,

V. &

mou,

E. (1999).

An

anchoring theory of

lightness

perceptien.

RsychologicaJReview, 106, 795-834.

Grossberg,

S.,

&

Mingolla,

E. (1987).

Neural

dynarnics

of surface perception: Boundary webs, illuminants,

and shape-from-shading.

Computer

Vlsion,

ics,and image Precessing,37,116-165,

Grossberg,

S.,

&

Todorovic, D.

(1988}.

Neural

ics

of 1-D and 2-D brightness _perception:A unified

rnodel of classical and recent _phenomena.

tionand RsychoPdysics,43,241-277,

Hejnemann, E.G.

(1955),

Simu]taneeus brightness

duction

as a function of

inducing

and test-fie]d

luminances,

fou"ial

of

ExPerimental Flsychology,50,

89-96,

Heinemann, E.

G. {1972).

Sirnultaneous

brightness

duction.

In

Jameso,

D,

and

Hurvic,

L. M,

{Eds.),

dbooh

of

Sensoew

Physiology,

V[IL4

Wszaal

phJ,sics.

Springer-Ver]ag

(Berlin).

Helson.

H. (1963}.

Studies

of anomalous contrast and

assimilation.

Iburnal

of

the

QPtical

Society_qfAmeF

ica,53,179-I84･.

Kingdom, F.A.A. & Moulden, B.

_(1988).

Border

(12)

106

The

Japanese

Journal

of Psychonomte Science Vol.28, No. 1

fects

on

brightness:

A review of findings,models

and issues.Spatial Vision,3(4),225-262.

Kingdom, F.

A.

A. (1997).

Simultaneous

contrast:

The

legacies

of

Hering

and

Helmholtz.

.Percoption,

26,

673-677.

Kingdorn, F.A. A.,Blakeslee, B.

&

McCourt,

ME.

(1997).

Brightness

with and without _perceived

transparency:

When

docs

it

make a difference?

fercePtion,26,

493-506.

Lamme,

V. A,

F.,

&

Roe]fsema,

P.R.

(2000).

Thc

tinctmodes of vision offered by fecdforward and

recurrent processing. Trends

in

IV2uroscience,23,

11,571-579.

Land, E.H. & McCann,

J.

J. (197

1 ).

The

rctinex theory

of vision

foumal

of

lhe

QPtical

Sociely

ofAmerica,

61, 1-11.

Lennie, P. & D'Zmura, M.

_0988).

Mechanisms of

color vision.

CRC

Criticat

Reviews

in

Neurobiology,

3(4),

333-400.

McCourt, M. E.

_(1982).

A spatial frequency dependent

grating-induction effect.

Vision

Research,

22,

134.McCourt,

M E,

(1994),

(}ratinginduction:A new

planation for statienary visual _phantoms. Vision

Research, 34, 1609-1618,

Mou]den,

B. &

Kingdom,

F.A,A.

(1989),

White's

effect:

A

dual

mechanism,

Vision

Research,

29,

1245-1259.

Moulden, B. & Kingdom,

F.AA,

(l991).

The

local

border

mechanism

in

_grating

induction.

Vision

search,

31,

1999-2008.

Paradiso,M.A,,& Hahn, S,

_(1996).

Filling-in_percept$

produced by

]uminancc

moclulation. Vision

search,

36,

2657'2663.

Paradiso, M. A.,& Nakayama. K.

_Q991).

Brightness

perception and

filling-in,

VisionResearch,31.

1236.

Pessoa,

L.,

Mingolla,

E. &

Neumann,

H. (1995).

A

contrast- and luminance-driven mu]tjseale

work model of

brightness

_perception. Vision

search, 35,2201'2223.

Pokorny,

_J.,

SheveLl,

S.K.,

&

Smith,

V.

C.

_(1991).

our appearance and colour constancy. In

Gouras,

P.

(Ed.),

Vision

and

Visual

Llyshinction

(Vol.

6,

The

Perception

of

Co]or,

pp.

43-61),

Boca

Raton, FL:

CRC Press.

Reid,

C,

R. &

Shapley,

R. C1988),

Brightness

induction

by localcontrast and the spatial dependence of

assimi]ation. VisionResearch, 28{1),115-132.

Ross, W. D.,& Pessoa, L.

(2000).

Lightness irorn

trast:A selective integration model. Rgrceptionand

RsJ)choph)'sics,62(6),1160-1181.

Rossi,A.

F. &

Paradiso,M. A.

(1996).

Temporal limits of brightness induction and mechanisms of

ness perception. VisionResearch, 36, 1391-1398.

RossL

A.

F. &

Paradiso,

M.

A. (1999),

Neural

corre-lates

of perceived

brightness

in

the retina,

lateral

geniculate nucleus, and striate cortex.

fournal

of

Arleuroscience,

19,6145-6156,

Rossi,

A.F.

Rittenhouse,

C.D.

&

Paradiso,

M.A.

C1996}.

The

representation ot

brightness

in

pri-mary visual cortex. Science,273, l104-1107.

Shapley,

R.

& Enroth-Cuge]],

C,

<1984).

Vi$ual

adap-tation

and retinal gain-controls.

thogress

in

Retinal

Research,

3,

263--346.

Shapley, R.& Reid,R.C.

_(1985).

Contrast and

assimi-lationinthe _pcrception of

brightness.

Proceedings

of

theIVational

Academy

of

Science

USA,

82,

5983L

5986.

Smith,

V,

C. Jin,

P.

Q. &

Po]<orny,

J. (2001).

The role of spatial frequency incolor induction.

Vision

Re-search, 4L IO07-1021.

Spehar,

B.

Gi]chrjst,

A,

L. & Arend, L.E.

(1995).

The

critical role of relative luminance relations in

tc'seffect and _grating

induction.

Vision

Research,

35,

2603-･2614.

Taya, R.,Ehrcnstein, W.,& Cavonius, C,

_<1995).

ing the strength of the Munker-White effect by

stereoscopic vievtTing. Percoption,24,

685-694.

Todorovic, D.

_{1

997).

Lightness and

junctions,

tion,

26,

379--39s.

White,

M.

&

White, T.

(1985>.

Counterphase lightncss

induction. vasionResearch, 25, 1331.-1335,

Whil/e,

M. <1979).

A

new effect oC _pattern on

ceived

]ightness.

Perception,

8,

413-416.

Whitc, M.

(198".

The cffect of the nature of the

surround on the _perceived lightness of _grey bars

within square-wave test _gratings.

_I'ercoption,

_10,

215-230.

WyszeckL

G. _Cl986).

Color

appearance, InBoff,K.R.

Kaufman L.,

&

ThDmas,

J.B.

(Eds.),

Hdndbooh

of

Percoption

and

Human

Rei:formance

(Vol.

1,

Sensory

Processcs

and

Perception},

John

Wiley

and

Sons

(New

York).

Wyszecki,

G.,

&

Stiles,

W.S.

{1982).

Color Science:

Concepts and Mbthods.

Quantitative

Data and

mulae

(pp.

487),

John

Wilcy

&

Sons,

New

York.

Yund,

E.W.,

&

Armington.

J.C.

(1975).

Color and

brightness

contrast effects as a

functien

of spatial

variables. Iijsion

Research,

15,

917-929.

Yund, E,W.

Snodder]y,

D.M., ll'epler,

N,K.,

&

Valois,

R.

L. (1977).

Brightness

contrast effects

in

rnonkey

lateral

gcniculate nucleus. Sensompi

esses, 1,260-271.

Zaidi,

Q. (1989).

LocaL and distal

factors

in

visual

grating induction.

Vision

Research,29,

691-697.

Zaidi,

Q.,

Spehar,B. &

Shy,

M.

(1997).

Induced effects

of background and

foregrounds

in

sienal configurations: The role of T-junctions.

caption, 26,395-408.