Functional organization of the monkey visual cortex for stereoscopic depth

(1)

F

u n c t i o n a l

O

r g a n i z a t i o n

O

f

T

h e

M

o n k e y

V

i s u a l

C

o r t e x

F

o r

S

t e r e o s c o p i c

D

e p t h

.

Daniel Lewis Adams.

PhD. Thesis.

University College London.

(2)

INFORMATION TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed,

a note will indicate the deletion.

uest.

ProQuest 10017388

Published by ProQuest LLC(2016). Copyright of the Dissertation is held by the Author.

This work is protected against unauthorized copying under Title 17, United States Code. Microform Edition © ProQuest LLC.

ProQuest LLC

789 East Eisenhower Parkway P.O. Box 1346

(3)

Abstract

O ur own, and other previous studies of V3 and V3A (two visual areas

constituting the V3 complex) have show n it to contain m any orientation

tuned cells w ith strong binocular interactions (Zeki 1978b; Burkhalter and

Van Essen 1986; Felleman and Van Essen 1987). This, and the area's M

dom inated cortical input, led us to choose the V3 com plex as a likely

candidate for an area specialized in the processing of stereoscopic depth.

Thus, the aim of this study was to record from single cells in V3 and V3A

and determ ine their selectivity for stereoscopic depth. The responses of cells

in these tw o areas w ere exam ined and the distribution of disparity tuned

cells was com pared w ith that in area V2, which has all m odalities of vision

represented in it. The results show that the V3 complex contains a high

percentage of disparity tuned cells and th at these are also orientation or

direction tuned. Thus, these cells are detecting the horizontal disparity of

visual features. Previously defined classes of disparity tuned cell (Poggio

and Fischer 1977; Poggio et ah 1988) w ere found in both the V3 complex and parts of V2. In agreem ent w ith previous studies, (Burkhalter and Van Essen

1986; H ubei and Livingstone 1987) disparity tuned cells w ere found in the

thick stripes of V2.

The second p a rt of this stu d y was to examine brains injected w ith

anatom ical tracing agents to discover the cortical connections of the V3

complex and an area on the dorsal prelunate gyrus (DP) w hich receives an

input from V3A. Results show ed that the areas injected V3, V3A and DP

have extensive connections w ith each other and w ith areas dealing w ith

other attributes of vision. The results dem onstrate th at the V3 com plex

(7)

areas w hich use this inform ation for the analysis of m ore complex visual

(8)

Acknowledgements

This piece of work w ould not have been possible w ithout the help of a

num ber of people, primarily, my supervisor Professor Zeki, whose w isdom

encouragem ent and advice was invaluable. I w ould also like to express my

gratitude to Dr. Stuart Shipp for sharing his enthusiasm and expertise during

experim ental w ork as well as his critical reading of the m anuscript. G rant

W ray and John Romaya for technical and com puter assistance and M ark

Rayan and Anne Fitzpatrick for histological tutoring and help. I w ould also

like to thank m y colleagues, Konstantinos M outoussis, A ndreas Bork and

Ludovica Marini for being good company and sharing the work load during

experiments that usually lasted 24 hours a day for an entire week.

I am also grateful to the BBSRC, W ellcome Trust and latterly the

D epartm ent of Social Security for providing the funds necessary for the

(9)

In trod u ction

Sectionl

The perception of depth

The ability to judge depth is an im portant faculty of visual animals;

it provides them w ith know ledge of distance w ith respect to the organism

and distances of objects w ith respect to one another and hence prepares

them for m aking a p p ro p riate m otor responses. It is of ev o lu tio n ary

im portance because it greatly aids the hunting and capture of prey as well

as the safe navigation of anim als in their environm ent. Psychophysical

studies indicate th at the perceptual transform ation from tw o to three

dim ensions relies on tw o types of cues: cues for m onocular depth, and

stereoscopic cues for binocular disparity. A t distances greater than about

50m the retinal images from each eye are virtually identical, so d ep th

m u st be judged using m onocular cues. There are at least five types of

m onocular d ep th cue:

1. Previous familiarity. If w e know from experience som ething about the size of an object we can judge the object's distance.

2. Interposition. If one object is partially h id d en from view by another object we assum e the hidden object is further away.

3. Linear and size perspectives. Parallel lines ap p ear to converge w ith distance. The greater the convergence of lines, the greater the im pression

of distance. The visual system interprets the convergence as d ep th by

assum ing th at the lines rem ain parallel.

4. Distribution o f shadows and illumination. Patterns of light and d ark can give the im pression of d ep th . The sh ad in g on objects im plies

different inclinations to the light source and on the w hole, b rig h ter

(10)

5. Motion for monocular movement) parallax. As w e m ove ou r h e ad s from side to side, the image projected by an object in the visual field

moves across the retina. N earby objects seem to m ove quickly and in the

o p p o site direction to our m ovem ent, w hereas d ista n t objects m ove

slowly.

All of the above cues (except the fifth) are used by artists to endow

their paintings w ith a sense of depth. This im pression of d ep th can be so

co nvincing to the b ra in th a t if som e p ain tin g s (p articu larly th o se

c o n ta in in g a g rea t deal of lin e ar p e rsp e c tiv e cues) are v ie w e d

m onocularly, the visual system will m ake convergent and divergent eye

m ovem ents that m atch those th at w ould be m ade w hen view ing the real

three dim ensional scene (Enright 1987).

Objects view ed at distances sm aller th an ab o u t 50m project a

slightly different image onto each retina because the eyes are separated

horizontally. This difference (called disparity) can be used by the visual

system as a pow erful d ep th cue. The analysis of d isp arity to p roduce

d epth vision is called stereopsis.

Historical review of stereoscopic vision

The history of ideas about vision is docum ented back to the tim e of

the Greeks, w ho, follow ing an idea by Em pedocles (5th century BC),

believed th at light leaves the eye in the form of a cone m ade u p of

straight rays and these rays gain know ledge of the w o rld b y "feeling"

objects rath er like an invisible hand. The inform ation g ain ed is then

tran sp o rte d back to the eye and b rain to generate visu al sensations.

Recently it has been com m ented th at this idea of active v ision is m ore

(11)

The first breakthrough in visual optics was the w ork of an Arabic

scholar by the nam e of A lhazen (965-1040). H e firm ly rejected the

em anation theory of vision proposed by the Greeks and in his book "The

Book of Optics" (Kitab al-manazir) (Alhazen 1989). He proposed that the

visual scene is projected as a two dim ensional inverted im age onto the

retina. Realising th at the im age w ould be inverted and reversed, he

suggested th at it was corrected by a further refraction from the back of the

eye to form a corrected im age on the optic disc. M any years later,

Johannes K epler (1571-1630) concluded th at there w as no basis for the

second inversion proposed by Alhazen and suggested th at the image w as

rectified by a m ental process. The realization that the visual image w as

projected onto the retina as a tw o dim ensional rep resen tatio n caused

people to w onder how the flat image on the retina was transform ed into a

three dim ensional p recept of the visual w orld. This problem w as of

p articu lar concern to artists, w hose m ain preoccupation w as how to

rep resen t three dim ensions on a flat canvas. Since stereoscopic cues

cannot be used in paintings, the foci of these enquiries w ere based on

m onocular depth cues.

A lthough the rudim ents of binocular vision had been observed by

Euclid, w ho noticed that the eyes each have a slightly different view of

the w orld, and Aristotle, w ho noted that w hen one eye is pressed w ith a

finger, a double image is produced, binocular vision was not equated

w ith d e p th perception. Thus the question arose "why do w e have tw o

eyes?" For m any years (since the tim e of Alhazen) it w as believed th at

two eyes w ere useful because if one becomes dam aged, the other rem ains

intact. This is not to say th at the geom etry of binocular optics w as not

studied; A lhazen him self described the w ay in w hich objects placed in

(12)

that lie at, or close to, the fixation plane, i.e. on corresponding points,

are fused to form a single image. W ith this observation, A lhazen

described w hat w e now know as the horopter, i.e. the plane in space at

w hich objects can be seen as single fused images. A lthough A lhazen

realized that this plane w as not fronto-parallel, its precise geom etry w as

n o t stu d ied u n til 1818 by Vieth and M üller (Vieth 1818; M üller 1826).

Sim ilarly, A lhazen described a sm all area in space either side of the

fixation plane, at w hich objects can still be seen as single and fused, this

area is now know n as Panum 's fusional area, and its study had to w ait

until 1858 (Fanum 1858).

P erhaps the earliest description of stereopsis w as p ro d u ce d by

Leonado da Vinci (1452-1519); in his w ork Trattato della pittura (see Keele 1955; Strong 1979) he observed; "That a p a in tin g , th o u g h

conducted w ith the greatest art and finished to the last perfection, both

w ith regard to its contours, its lights, its shadow s and its colours, can

never show a relievo equal to that of the natural object, unless these be

view ed at a distance and w ith a single eye." taken from W heatstone

(1838). H ere Leonado im plied that tw o dim ensional pictures can only

truly p o rtray the real scene if that scene is devoid of stereoscopic d ep th

cues. H ow ever, Leonado did not show a full u n derstanding of the finer

points of stereopsis other than the idea that closing one eye transform ed

the visual w orld into an image m ore resem bling that on a canvas.

In 1613 Franciscus A guilon published his w orks on optics entitled

Opticorum libri sex. His 6 books followed the w ork of Euclid, A lhazen and others w ith the ad d en d u m that d ep th perception is im proved w ith

binocular vision. This point w as illustrated in A guilon's book w ith an

(13)

estimating the distance to an object held in front of him (reproduced in

figure 1).

Figure 1.

Rubens' engraving depicting a one-eyed man underestim ating the distance to an object held in front of him to portray the point that binocular vision is im portant for depth discrim ination. Produced in 1613, it w as com m issioned by Franciscus A guilon to illustrate his work O pticorum libri sex (Taken from Judson and van de V elde 1978).

Thus, the question "Why do we have two eyes?" was answered

nearly two centuries after it was asked. Aguilon studied the geometry of

binocular vision using the theorem s of Euclid and coined the term

"horopter" to describe the plane in space at which both fused and diplopic

images appear to lie. The word horopter comes from the Greek "horos", meaning space and "opter”, observer. The word horopter is still used today, but to describe a slightly different plane. Aguilon used the term to

describe a fronto-parallel plane at the same distance from the observer as

the fixation point. Today the horopter describes a plane in space, points

(14)

fronto-parallel plane, this produces a horizontal circle (the Vieth-Müller circle)

which passes through the centres of both eyes (see figure 2).

Eyes'

V ieth -M ü ller circle (H orop ter)

Figure 2.

I llu s t r a t io n o f th e V ieth-M üller circle or horopter. T h e f ig u r e r e p r e s e n t s a horizontal plane in clu d in g both ey e s and a fixation p oin t (F). Points A and B lie on the Vieth-M üller circle. From E uclidean geom etry, angles a = a' and b = b'. Thus the points A and B both project to corresponding retinal p o in ts and are see n as sin g le fu sed im ages, as do all points that lie on the circle.

The Vieth-Müller circle was first described in a paper by Vieth

(1818), although it was simultaneously studied by Müller (1826). They

showed that all points projecting to corresponding retinal points form a

circle that passes through the optic centres of both eyes. Vieth mistakenly

generalized his theory of corresponding points in three dimensions to a

sphere. His mistake was corrected by Prévost (1843) who showed that the

theoretical horopter is in fact a toroid formed by sweeping the Vieth-

Müller circle through the interocular axis.

The notion that binocular vision gives rise to depth perception was

not further pursued until the 19th century w hen Charles W heatstone

(the inventor of the W heatstone bridge) published a paper entitled:

(15)

remarkable, and hitherto unobserved. Phenomena of Binocular Vision"

(Wheatstone 1838). He showed an understanding of stereoscopic depth

perception and illustrated the phenom enon w ith the invention of the

stereoscope. With this instrument, shown in Figure 3, two photographs

of a scene taken 60-65mm apart, one from the position of each eye, are

m ounted onto a binocular like device such that the right eye sees only the

picture taken from the right position and the left eye sees only the left

picture. Fusion of the two images produces a three dimensional image.

Figure 3.

A diagram of W heatstone's stereoscopic apparatus. The tw o m onocular im ages (E' and E) are reflected by mirrors (A' and A) placed close to the eyes of the view er, so that each eye only sees one im age. The tw o im ages are therefore superim posed and the view er fuses them to produce a "solid" stereo image. (Taken from W heatstone 1838).

Thus the stereoscope vividly demonstrates stereopsis by enabling

convincing stereoscopic depth to be perceived from tw o-dim ensional

images. In W heatstone's own w ords, "The preceding experim ents

render it evident that there is an essential difference in the appearance of

objects w hen seen with two eyes, and when only one eye is employed,

and that the m ost vivid belief of the solidity of an object of three

dimensions arises from two different perspective projections of it being

(16)

d isp arity in W heatstone's stereoscopic pictures are exam ined it can be

seen th at corresponding elem ents in each picture are shifted tow ards or

aw ay from each other; the form er causes a shift in d ep th in front of, and

the latter behind, the view er's fixation plane.

In 1858, Panum described a range of disparities w ithin w hich tw o

sim ilar images, one in each eye, are perceived as a single image (Panum

1858). This plane (Panum 's fusional area) is an area in d ep th that extends

in front of and behind the fixation plane that corresponds to the region of

b in o cu lar single vision. The im ages from the tw o eyes of all p o in ts

w ithin this area are said to be 'fused'. The front to back size of the area in

d ep th is d ependent upon a num ber of factors: it increases in a geometric

w ay as the fixation plane becom e fu rth er aw ay from the eyes b u t the

visual angle subtended from the front of the area to the back rem ains

constant, and it tends to increase in size tow ards the p eriphery of the

visual field. It is also d ependent on the n atu re of the stim ulus and its

background (Panum 1858; Ogle 1964).

The m eeting of projections from the eyes

In order th at the tw o view s of the w orld be unified to p roduce a

single percept, inform ation from both eyes m ust arrive at single cells in

the brain. A lthough this concept seems obvious today, for a long tim e it

w as th o u g h t th at the m onocular images rem ained separate th ro u g h o u t

the visual system and w ere combined "by a m ental act" (Helm holtz 1893).

Ram on y Cajal stood alone in his belief that inputs from corresponding

parts of the tw o retinae converge on w hat he called "isodynamic cells" and

it is this th at forms the basis of unified binocular vision (Ramon y Cajal

1911). The location of the first binocular cells in the visual system has

(17)

binocular vision m ay be elucidated. N ot unreasonably, Galen (175) and

after him , A lh azen (1989), th o u g h t th a t the locus of b in o cu lar

com bination w as the optic chiasm . D uring the R enaissance, René

D escartes p roposed th at the optic fibres m ight converge on the pineal

gland for unification (Traité de l'Homme). In fact the first possible site for unification of inform ation from the tw o eyes is the lateral geniculate

nucleus (LGN), w here fibres from both eyes term inate having already

been com bined en-passage th ro u g h the optic chiasm . The LGN is a

n u c le u s c o m p risin g six lay ers. P h y sio lo g ica l a n d a n a to m ic a l

exam inations of the layers (Silva 1956) reveals them to contain cells that

receive m onocular inputs; layers 1, 4 and 6 receive projections from the

c o n tralateral eye and the oth ers are exclusively in n erv a te d by the

ipsilateral eye. Cells in these layers are also solely excited by their

respective eye. Extensive connections exist betw een the layers of the LGN

b u t these do n o t endow the cells w ith the p ro p e rty of b in o cu larity

(G uillery 1971). Thus, fibres leaving the LGN still carry m onocular

inform ation. The next stage in the visual p athw ay is the prim ary visual

cortex (striate cortex, area 17 or VI), more specifically layer 4c, the in p u t

layer. In 1959, H ubei and Wiesel proved Ramon y Cajal to be correct by

d e m o n stra tin g the existence of binocular cells in the cortex (Cajal's

"isodynamic cells"). These cells first occur in VI except in layer 4c. V I is

therefore the first locus w here fusion of the tw o m onocular im ages can

occur (H ubei and Wiesel 1959). Quite w hy cells retain their exclusive

m onocularity until this stage, w hen they have h ad am ple o p p o rtu n ity

for generation of binocularity at previous parts of the visual system , is

(18)

The random dot stereogram

H ow does the brain know w hich p a rts of the m onocular im age

correspond to the sam e object view ed by the other eye? This question

(the co rresp o n d en ce problem , w h ich w e w ill re tu rn to) rem ain s

unresolved today b u t a great advance w as m ade in 1960. Bela Julesz, a

ra d a r engineer, w as w orking w ith w artim e aerial p h o to g rap h s taken

b eh in d enem y lines. A technique used to spot cam ouflaged structures

w as to take tw o photographs from slightly different locations and view

them w ith a stereoscope this w ay objects taller th an their surro u n d in g s

stand o u t in binocular depth. At the time, the prevalent view about the

correspondence problem held by psychologists w as th at the b rain first

recognized an object in the m onocular images, and then paired the tw o

objects to produce single vision and stereoscopic depth. W orking w ith

stereoscopic images com prising few recognizable forms, Julesz knew this

n o t to be the case. To dem onstrate his observation he m ade images w ith

a com puter, sim ilar to his stereoscopic aerial photographs b u t containing

no recognizable forms. These "random dot stereogram s" (RDS) com prise

tw o random , yet correlated patterns of dots w ith a central portion of dots

being shifted by an integer m ultiple of the dot size in opposite directions

for each im age (Julesz 1960). Fusion of the tw o im ages p ro d u ces a

pow erful sensation of depth, the central portion being displaced in front

of, or behind the surrounding texture, depending on the direction of the

(19)

Figure 4.

An example of Julesz's original random dot stereogram. Diverging the eyes in such a way as to superimpose and fuse each monocular image produces the sensation of depth. In this figure, divergent viewing produces a smaller square displaced behind the plane of the paper.

These figures v iv id ly d em o n strate th a t, c o n tra ry to th e

p red o m in an t theory of the day, binocular d ep th perception does not

require m onocular form recognition and is an operation carried out quite

"early" in the visual system. This revelation shifted the em phasis of

n eu ro p h y sio lo g ical research from p ro b lem s in m o n o cu lar form

recognition to the search for binocular cells in the visual brain. It also

introduced the notion of local and global stereopsis.

Local and global stereopsis

One of the m ost im portant consequences of the invention of the

random dot stereogram was the revelation that m onocular form or d ep th

cues are not necessary for binocular fusion. Thus, the visual system

m u st be using a very m uch m ore sim ple m ethod of fusing binocular

images than previously supposed. Since the m any similar elem ents in a

random dot stereogram could easily give rise to false matches, the system

m ust be using m ore than just the form of the individual elem ents to

produce binocular fusion. Thus, global stereopsis is the m ethod the

visual system uses to examine large portions of the m onocular images

a n d u n d e rta k e s a cross correlation analysis, n o t only m atch in g

(20)

n eig h b o u rh o o d to disam biguate false m atches (Julesz 1971; Julesz and

O sw ald 1978). Local stereopsis refers to the m ore classical kind of

binocular fusion w here, w ithin Panum 's fusional area, unique features

in each m onocular im age are paired to resolve am biguity (Julesz 1971;

Julesz and O swald 1978).

Psychoanatom y

Since the d ep th features of random d o t stereogram s can only be

seen w hen m onocular inform ation from each eye has converged onto a

single cell, it can be said th at all the n eu ral m achinery u p to V I is

incapable of responding to stereoscopic depth. Thus, any perceptual

effects produced by view ing stimuli presented as random dot stereogram s

m u st be p ro d u ced in the cortex rath er th an the retina or LGN. The

lo c a liz a tio n of p e rc e p tu a l effect u sin g p sy c h o p h y sic s w as called

psychoanatom y (Julesz 1971). M any types of ran d o m dot stereogram s

(both static and dynam ic) have since been p ro d u ced , each designed to

exhibit a particular optical illusion or perceptual aftereffect. These studies

are n u m ero u s and m ostly draw the sam e conclusions, i.e. p ercep tu al

effects can be reproduced w ithout the need for m onocular features and

m u st th erefo re be p ro d u c e d w ith in th e v isu a l cortex. The first

phychophysicist to em ploy disparity instead of lum inance gradients for

the presentation of optical illusions w as Papert (1961). By presentation of

com m on optical illusions, such as the M üller-Lyer lines^, Papert show ed

th at the illusions p ersist w hen no m onocular cues are p resen t, th u s

dem onstrating that the locus of the illusion is cortical rather than retinal

or thalam ic.

(21)

O cular dom inance

The advent of single unit recording bro u g h t about a revolution in

the study of the visual system. First applied in the prim ary visual cortex,

cells w ere fo u n d th at resp o n d ed to stim uli p resen ted to the anim al

(H ubei and Wiesel 1959). Cats w ere used at first, and later, because of

their superior vision (especially colour vision) and the sim ilarity of their

visual system to hum ans, m onkeys w ere recorded from. The relative

influence of the two eyes on the response of a cell (ocular dom inance)

varies from cell to cell in VI and a scheme of classification w as devised

(H ubei and Wiesel 1962). The stim uli used to categorize cells into these

groups w ere presented to each eye in turn. Thus, it was not possible to

find out how the responses of the cell m ight change w hen both eyes are

stim u la te d sim ultaneously (binocular interaction). Cells in V I w ere

found to fit into one of seven groups: group 1 contains cells th at are

exclusively contralaterally driven, w hile group 7 cells are exclusively

ipsilaterally driven, w ith all gradations in between.

C ells in th ese g ro u p s w ere fo u n d to fit into a c o lu m n a r

o rg a n iz a tio n in the p rim a ry v isu al cortex, the ocular d o m in an ce

colum ns. These robustly organized colum ns are only p resen t in the

p rim ary visual cortex and com prise a patchw ork of m odules norm al to

the surface of th at area, each eye supplying irregular stripes of cortex

about half a m illim etre wide. Ocular dom inance colum ns rem ain one of

the m o st solid fin d in g s in the v isu al system and have since been

d em o n strated anatom ically and physiologically. The first anatom ical

dem o n stratio n w as achieved by m aking lesions in single layers of the

LGN an d looking at the p attern of degenerated axon term inals in the

striate cortex (H ubei and W iesel 1969). Since the layers of the LGN

(22)

m onocular ocular dom inance colum ns. Technical advances led to an

im p ro v e d anatom ical m eth o d of v isu alizatio n of ocular dom in an ce

colum ns involving the injection of a radio-labelled am ino-acid (tritiated

proline) into one eye. It w as first achieved in the m ouse (Grafstein and

Laureno 1973) and later em ulated in the m onkey (Wiesel et ah 1974). The am ino-acid is transported from the retinal ganglion cells to the LGN and

th en transneuronally to striate cortex. A utoradiography of sections taken

from striate cortex exhibit patches of dense transport of the amino acid in

layer 4, corresponding to the dom ains that contain m onocular cells. The

2 D eoxy-glucose (2DG) technique (Sokoloff 1977), w hen em ployed by

T ootell et al. (1988a), also dem onstrated, am ong other things, the p attern of ocular dom inance columns in the striate cortex of the m acaque.

The 2DG technique is a m ethod of show ing a p ictu re of functional

activity; radioactive glucose (^4(]_2_deoxy-d-glucose) is injected into the

anim al's circulation w hile it is presented w ith a visual stim uli. A fter

view ing the stim uli for a few m inutes in the paralysed and anaesthetized

state (to p rev e n t eye m ovem ents) the anim al is killed and its b rain

sectioned. Areas that were very active directly before the anim al's death

co n tain m ore of the radioactive glucose an d can be v isu alized by

autoradiography. Thus, full field m onocular stim ulation of the anim al

w ill p roduce pictures of the ocular dom inance colum ns in flat m ounted

sections, cut parallel to V i's surface. These techniques have established

th at the ocular dom inance colum ns form a series of parallel bands, each

0.25 to 0.5 m m wide. They spread through all layers of the cortex and in

each hem isphere a retinotopic m ap of the visual hem ifield is represented

twice, one w ithin the left eye colum ns and one w ith in the right. The

b ands tend to ru n 90° to the V 1/V2 border in iso-eccentric stripes', thus

m apping tw o circular visual fields onto the elliptical surface of the striate

(23)

H ubei and W iesel suggested th at the differing degrees of ocular

dom inance m ight, in some way, provide the basis for a depth sensitivity

m echanism , the dom inance characteristics being segregated so th at each

m onocular com ponent of an im age could be com pared to th at of the

other eye. This has so far proved not to be the case, and the functional

significance (if any) of ocular dom inance and its colum nar organization

in V I in som e (but not all^) anim als w ith stereoscopic vision rem ains

unclear.

O cular interaction

A side from H ubei and W iesel's ocular dom inance characteristics,

another type of binocular pro p erty exists for cells in the visual cortex.

Since H ubei and Wiesel stim ulated their cells m onocularly, they did not

find inform ation relating to the binocular interactions of cells. W hen

cells are d riv en binocularly their responses m ay change and are n o t

predictable from the m onocular responses. Zeki (1979) exam ined this

p ro p e rty an d d ev ised a m eth o d of classification of b in o cu lar cells

d e p en d in g u p o n their response d u rin g binocular stim ulation. Thus,

cells w ere divided into six groups according to w hether they w ere driven

by one eye only (two categories), dom inated by one eye (two categories),

e q u a lly w ell d riv e n by e ith e r eye (one categ o ry ) or d riv e n by

sim u ltan eo u s binocular stim ulation only (one category). U sing these

categories, so called "ocular interaction histogram s" could be plotted for

p o p u latio n s of cells. W hen these histogram s are p lo tted for p articu lar

visual areas one can see th at nearly all cells in prestriate cortex, i.e. all

cells beyond V I, are (i) binocular and (ii) produce differing degrees of

bin o cu lar interaction. Thus all these p restriate areas are, in theory.

(24)

candidates for analysing stereoscopic depth. However, since disparity was

n o t systematically varied in Zeki's study, cells that only respond at critical

disparities m ay have been missed. It is this property that enables cells to

be d ep th selective, and an area w ith a high proportion of these disparity

selective cells w ould be a far m ore likely candidate for an area of d ep th

specialization.

D isparity tuning

The receptive fields of binocular cells in VI w ere found to occupy

corresponding areas on the tw o retinæ and their response properties w ere

qualitatively identical for both eyes (Hubei and Wiesel 1962). Stereoscopic

vision requires m ore than the presence of binocular cells alone. In order

to get inform ation reg ard in g the relative d ep th s of objects from the

fixation plane the system needs to be capable of fusing the images from

b o th eyes an d an aly sin g the in fo rm atio n d eriv e d from th e sm all

differences betw een each image. To achieve this one w ould expect some

b in o c u la r cells to h av e rec ep tiv e field s th a t are slig h tly n o n

corresponding, this receptive field disparity being in the order of a few

m inutes of arc, thus allowing the cells to respond maxim ally w hen slight

disparities exist betw een the two m onocular images. This is w hat m akes

cells disparity tuned.

T he e x p e rim e n ta l s tim u la tio n of b o th re c e p tiv e fie ld s

sim u ltan eo u sly w ill yield inform ation reg ard in g binocular interaction

(facilitation or attenuation), b u t in order to stim ulate cells at specific

disp arities a m ore elaborate regim e is required. One solution in the

paralysed anim al is to superim pose both receptive fields on a screen using

(25)

q u a n tita tiv e ly v aried to stim u late the cell over a ran g e of retin al

disparities. A nother m ethod is to leave the receptive fields separated and

em ploy tw o stimuli, one for each receptive field (Henry et ah 1969). If the tw o stim uli are correctly positioned, the anim al should fuse them and its

visual system be presented w ith a single binocular stim ulus. By varying

the distance betw een the tw o stim uli one can change the disparity. The

advantage of both of these techniques is that one can stim ulate cells at

specific disparities and if a cell's responses are sensitive to small changes

in disparity, it shows that the cell is tuned to stim uli placed at a specific

distance relative to the horopter.

The m ethod of superim posing the receptive fields of a cell using

prism s w as em ployed by Pettigrew et al . (1968) in the striate cortex of the cat. Their m ajor finding w as th at cells had greatly facilitated responses

w h en stim u lated by a single m oving bar w ith th eir receptive fields

superim posed; furtherm ore, some cells were exquisitely sensitive to the

exact position of the receptive fields. Differences as small as 3 m inutes of

arc w ould greatly m odify the cell's responses, and if disparity (i.e. the

prism setting) w as set to non optim al values, some cells' responses could

be inhibited. Thus cells in cat striate cortex can be selective for disparity.

If the d isp arities these cells are tu n ed to deviate from zero one can

assum e they are doing m ore than fusing the tw o retinal im ages to form

the "cyclopean im age"^. A p o p u latio n of cells w hose responses are

m axim ized over differing disparities w ould in effect be tuned to different

positions in d ep th relative to the horopter.

(26)

In order to m easure differences in disparity tuning from cell to cell,

the very sm all eye m ovem ents that occur, even after complete paralysis

of the eye m usculature, m ust be stopped, or m easured and subtracted

fro m th e d ata. The m eth o d s of eye m o v em en t e lim in a tio n in

B ishop/P ettigrew 's group were to plot the optic disks of each eye onto a

screen and m onitor their m ovem ent. This m ethod is only possible in the

cat because it, unlike the monkey, has reflective retinae; shining a bright

light into the cat's eyes produces an image of its retinae on the tangent

screen. A dditionally, the eyes w ere im m obilized by gluing the scleral

m argins to brass rings. W ith these m easures in place the authors found a

spread of optim al disparities of ±1° (Nikara et al. 1968), m ost cells being tuned to points at or very close to zero disparity, i.e. the horopter.

U sing the prism m ethod of aligning receptive fields, H ubei and

Wiesel w ere also looking for binocular cells (Hubei and Wiesel 1970), b u t

in the m onkey visual cortex. In order to keep track of eye m ovem ents

they sim ultaneously recorded from a binocular V I "reference" cell in the

opposite hem isphere and repeatedly plotted its receptive fields, taking

note of any m ovem ents. Since the receptive fields of VI cells are sm all

and well defined, any m ovem ent of the positions of the reference cell's

receptive fields will correspond to eye m ovem ents, and can be rem oved

from the data obtained from the investigating electrode. Unable to find

binocular d ep th cells in area 17 (VI) they m oved onto the territory of the

prestriate cortex w hich at the time w as thought to comprise tw o areas (18

an d 19). H ere they found a large p ro p o rtio n of cells (43%) w hose

responses w ere critically d e p en d en t u p o n the prism setting and w ere

therefore disparity selective or "binocular depth cells". H ubei and Wiesel

n o ted th at although there w as variation in horizontal d isp arity am ong

(27)

v ary in g orientation selectivities. These groups of binocular d ep th cells

w ere segregated from cells not concerned w ith d e p th and tentatively

called d ep th columns. Hubei and Wiesel also noted that the disparity of a

cell w as related to its orientation selectivity, "The displacem ent of the

field in one eye, relative to the field in the other, is usually at right

angles to the receptive field orientation." (Hubei and Wiesel 1970). Thus

o n ly cells w ith n e ar vertical o rien tatio n tu n in g w ill be tu n e d to

h orizontal disparity. Retrospective exam ination of this paper show s th at

the au th o rs w ere recording from tw o areas, V2 and V3A; they found

b in o cu lar d e p th cells in both, alth o u g h they w ere u n aw are of this

subdivision w ithin the lunate sulcus at the time.

The sp read of optim al d isparities found by B ish o p /P ettig rew 's

g roup (Barlow et ah 1967; Nikara et ah 1968) in cat striate cortex prom pted H u b ei and W iesel to re-exam ine the stereoscopic m echanism s in th at

anim al (Hubei and W iesel 1973). H ubei and W iesel's recordings from

m o nkey striate cortex h ad show n there to be very little v ariatio n in

receptive field disparities in VI; the great m ajority of cells had their fields

in precise registration (only 5% show ing detectable disparities) and they

h ad found no convincing exam ples of binocular d ep th cells (H ubei and

W iesel 1970). In order to find out if the difference w as a genuine species

difference they recorded from the cat's striate cortex, controlling for eye

m ovem ents by keeping a striate reference cell in the opposite hem isphere

w ith w hich to com pare disparities. C ontrary to B ishop/P ettigrew 's data

they fo u n d very little variation in disparities and concluded th at the

m ain m echanism s subserving stereoscopic depth perception in the cat (as

in the m onkey) lie outside the striate cortex. The discrepancy betw een

these tw o groups is probably dow n to their stim ulation and field plotting

(28)

b e tw ee n th e receptive fields of a cell (taking eye m ovem ents into

account), b u t B ishop/P ettigrew 's group defined it as the prism setting

required to produce a maximal response of a cell. It is possible that a cell

m ay still n o t give a maximal response w hen its receptive fields have been

perfectly superim posed. Rodieck recognized this problem in term inology

(Rodieck 1971) and suggested th at the term "incongruity" be used for

receptive field d isp arity (i.e. non-corresponding locations of receptive

fields), keeping the term "disparity" for response-defined disparity only.

Since these early studies m any disparity tuned cells have been found in

cat and m onkey striate cortex (Nelson et al. 1977; Foggio and Fischer 1977; Von der H eydt et al. 1978).

Once isolated, the best way to study a binocular depth cell is to plot

a grap h of horizontal stim ulus disparity against the cell's response to that

stim ulus. For non-disparity tuned binocular cells such a graph will give

rise to a bro ad curve, show ing sim ply th at the cell responds better to

b in o cu lar stim u latio n th an m onocular. For a sm aller p ro p o rtio n of

binocular cells the curve will be a far more specific shape, show ing that

the cell is disparity tuned. D isparity tuned cells m odulate their responses

over a very narrow range of disparities and the type of m odulation they

show fits them into a schem e of classification devised by Foggio and

Fischer (1977). This classification scheme divides cells into tw o groups,

eith er n e a r /f a r (asym m etric) or tuned. Figure 5 show s the response

properties (in the form of disparity tuning curves) for their four types of

(29)

F ixation p lan e

E yes

TE

Far

N ear

D isp arity

Figure 5.

S h a p e s o f th e d is p a r it y -response curves for the four types of disparity tuned cell described by Foggio and Fischer in their 1977 paper. TE, Tuned excitatory, Tl, Tuned inhibitory. The N ear and Far cells fall into the asym m etric category, and the TE and TI cells into the tuned category. The positions of the curves on the x axis are relative to the position of the fixation p lane.

Poggio and Fischer's recording set-up had a num ber of advantages

over previous ones that enabled them to find and classify these cells:

firstly they recorded from the awake behaving monkey, trained to fixate a

particular point while its receptive fields were plotted and stim uli

presented. This carries the advantage that one always knows the position

of zero disparity and eye movement problems were eradicated. Also the

stimuli were presented in real depth relative to the depth of the fixation

spot (i.e. the stim ulus screen was moved tow ards or away from the

animal) rather than using disparate stimuli or superim posing receptive

fields, as had been used in the previous studies. An important feature of

(30)

fixation plane. Poggio and Fischer found th at m ost of their cells w ere

tuned to points w ithin ±0.5° of the fixation plane. This corresponds to a

real distance in depth of roughly ±3cm at their particular fixation distance,

a value roughly equal to Panum 's fusional area. They concluded that the

range of optim al disparities they found (For TE and TI cells) w as the

neural correlate of Panum 's fusional area (Poggio and Fischer 1977).

Since their 1977 study, Poggio et al. have found disparity tu n ed cells in both V2 and V3 (Poggio et al. 1985; Poggio et al. 1988). In their 1988 paper they enlarged their categories of disparity tuned cells to include the

tuned near (TN) and tuned far (TF) types: these neurones have response

profiles sim ilar to the tuned excitatory type, b u t their responses occur at

disparities betw een ±0.01° to 1.0° thus, they differ from TE cells in their

location, b u t not the w id th of their optim al response. These neurones

are in a m inority in VI (14/31), increase in V2 (21/30) and are com m on

in V3/V3A (23/26) (Poggio et al. 1988). The eccentricities of different types of d isp arity tu n ed cells w ere also studied by the authors b u t since the

m ajority of the cells they investigated in VI and V2 had their receptive

fields w ithin 4°, and all of their V3 cells were at eccentricities greater than

4°, th ey w ere n o t able to p ro v id e any strict co rrelatio n b etw een

eccentricity and stereotuning.

The p ro p o rtio n s of the various types of d isparity tu n ed cells in

different areas have also been the subject of investigation and it has been

show n to vary in different visual areas. The p ro p o rtio n of tu n ed cells

versus asym m etric cells changes as the level of hierarchical processing is

increased. In VI, V2 and V3 tuned cells and asymmetric cells are in equal

(31)

proportion of asymmetric cells increases to about 90% (Roy et al. 1992) and in the lateral intraparietal area (LIP) the figure is 100% (Gnadt and Mays

1991).

The functions of different types of disparity tuned cell

Since "tuned" cells (TE, TI, TN and TF cells) are often present at

the early stages of visual processing it has been suggested that they m ay be

p re d o m in a n tly involved in the stereo-m atching process to m ain tain

accurate stereo fusion w ithin Panum 's fusional area (Trotter 1995). The

w idth of tuning of TE/TI cells w ould therefore correspond to the lim it of

stereo acuity (Poggio and Fischer 1977). The asym m etric cells' function

m ay be m ore to do w ith the control of convergent and divergent eye

m ovem ents and to tell the visual system w hether an object ap p earin g

outside Panum 's fusional area is in front of or behind that plane (Poggio

and Fischer 1977). Indeed the near and far cells in LIP have been show n

to be m odulated by vergence (Gnadt and Mays 1991) and this area projects

to the su p erio r colliculus, a region assigned to oculom otor control

(Sparks 1986). Asymmetric cells, especially the near cells, m ay also have

a function in the control of h an d m ovem ents because they resp o n d to

stim uli nearer to the animal than the fixated object, and could be useful

for the direction of hand m ovem ents tow ard th at object. The changing

proportions of tuned and asym metric cells m entioned above suggests that

the fu rth er up the parietal sequence of areas one progresses, the greater

(32)

Other types of disparity

Disparity is a general term encompassing all types of difference

between the two monocular images of an object. These differences may

be further subdivided: position disparity is the shift in horizontal

position that arises from an object being displaced from the fixation plane;

orientation disparity is the perceived difference in orientation between

each m onocular image of a line which tilts in depth tow ards or away

from the observer (illustrated in figure 6).

1

Figure 6.

Illustration of the orientation difference betw een each im age on the retinae w h en a line tilted tow ards or aw ay from the observer is view ed binocularly. L and R represent the (exaggerated) im ages on the left and right retinae of a single line tilted in depth, the cube is added to aid perspective view ing. By analysing the angle 6 the inclination of the line can be calculated and depth perceived.

D isparity curvature is the difference in perceived cu rv atu re

between the two monocular images of a surface of a three dimensional

object. Shapes rotated about their vertical axis cause disparity of

horizontal width. Together these differences are analysed to produce a

(33)

position disparity allows one to know the distance of an object from the

fixation plane, b u t alone it gives no cues to the three dim ensional form

of an object. One of the constituents of three dim ensional form is tilt in

depth. Since we are capable of discrim inating w hether lines are rotated

about a horizontal axis, one m ight expect to find cells selective for tilt in

dep th in the visual cortex. In effect, a cell selective for tilt in depth could

com prise a cell whose two m onocular receptive fields are each tuned to a

slightly different orientation, hence the term "orientation disparity".

Since our eyes are separated in a horizontal direction it is only possible to

perceive tilt in d ep th w hen view ing vertically or near vertically oriented

lines (neglecting all other d ep th cues). Cells of this type w ere indeed

found in the cat's striate cortex (Blakemore 1970) and since in m onkey VI

(H anny et ah 1980).

So called "tilt in depth" cells give an optim al response w hen each

eye is p resen te d w ith a line slightly differing in o rientation. Eye

m ovem ents will cause problem s w hen looking at this type of cell because

cyclotorsion will m ake all cells appear to be selective for tilt in depth.

Blakemore (1970) elim inated this possibility by recording from m any cells

in a single anim al w hose eyes w ere held still (by suturing them to firm ly

h e ld m etal rings) and dem o n stratin g th at the difference in the cell's

preferred orientation betw een the two eyes varied significantly from cell

to cell (Blakemore 1970). Another feature of these cells was that they w ere

also selective for horizontal disp arity , and their responses could be

occluded if the optim al orientation disparity stim ulus w as not presented

at the correct horizontal disparity. The existence of cells selective for tilt

in d ep th in the m acaque was disputed by H ubei (Hubei and Wiesel 1977)

w h o m ain tain ed th a t the p ro p erties of each receptive field rem ain

(34)

found th at have opposite direction selectivities in each eye (Zeki 1982) it

is not unreasonable to suggest that VI cells m ay have slightly different

orientation preferences for each eye.

Some cells have been found to m ed iate an o th er a ttrib u te of

stereoscopic vision, th at is m otion in depth (Pettigrew 1973; Regan and

Beverley 1973; Zeki 1974a). Stim uli com prising a fixed d isp arity are

perceived at a fixed depth but if disparity is changed w hile the stim ulus is

b ein g v iew ed , m otion in d e p th is perceived. The d irectio n of 3

dim ensional m otion is dependent u pon both the direction of the images

in bo th eyes and the direction of change of disparity. By changing these

param eters, trajectories of any direction can be p roduced. Pettigrew

fo u n d cells in area 18 of the cat th at signalled changing d isp a rity

(Pettigrew 1973), and later C ynader et al (1978) found cells in area 18 of cat's visual cortex th at w ere selective for specific trajectories; p erhaps

unsurprisingly, the m ajority of cells found w ere selective for trajectories

th at w ould result in the stim uli coming close to, or actually hitting the

anim al (Cynader and Regan 1978). Zeki (1974a) found cells in V5 of the

m acaque th at had opposite direction selectivities or each eye and w ould

therefore respond m axim ally to m ovem ent tow ards, or aw ay from, the

(35)

,

---^ b '

/

XV

Figure 7.

Diagram to sh ow that w hen a point at A ,

having its retinal images at a

and a', is displaced to B,

having its images at b and b',

the d isp la c e m en t is in opposite directions in the two e y es (a ssu m in g no eye movements take place). Cells w ith o p p o site d irectio n selectivities in each eye have been found in V5 (Zeki 1974a).

H ow ever, the existence of such m otion in d ep th cells in m onkey

area V5 w as d isputed in 1983 (M aunsell and Van Essen 1983a) because,

alth o u g h th ey ap p eared to respond w ell to m otion in d e p th w h en

stim ulated at trajectories far from the unit's optim al fixed disparity, their

best response w as still to m otion in the fronto-parallel plane, i.e. at a

fixed disparity. This m ay be true for some trajectory cells, b u t cannot

stand for cells w ith opposite direction selectivities in each eye as reported

by Zeki in V5 (Zeki 1974a). It has since been suggested th at m otion in

d ep th is not analysed by single units, b u t by the change in o u tp u t from

m any units, each tuned to a fixed disparity (Gum m ing 1995). W hether

the o u tp u ts from these units w ould have to converge onto a single cell

for the m otion in depth to be perceived is unknow n.

Psychophysics

Stereopsis has been studied extensively by psychophysicists because

it is an accessible attribute of vision to study in this way. The advent of

the ran d o m d o t stereogram inspired a d elu g e of stu d ies because it

(36)

p e rc ep tu a l p ro p erties of stereopsis. The form s seen in ran d o m d o t

stereo g ram s are also invisible m onocularly and therefore, m u st be

constructed once the visual signals have reached the cortex.

V ario u s a u th o ritie s have su b d iv id e d h u m a n ste reo p sis into

categories; here I shall review each of these (som etim es overlapping)

divisions form ed m ainly on the basis of psychophysical studies.

Local/global

In order to prevent the false m atching of the small black and w hite

squares th at constitute random dot stereogram s, a type of stereopsis th at

exam ines m ore th an the local features of the visual scene m u st exist.

Julesz called this type of stereopsis "global" (Julesz 1971) because it has to

take inform ation from a large area of the visual field in order to prevent

m ism atching of the identical local features. Local stereopsis is therefore

the m ech an ism req u ire d to p erceiv e d e p th in the m ore classical

stereoscopic scenes w here there is no chance of false m atching because

each p a rt of the m onocular image is unique. Thus, the division of local

and global stereopsis m ay be looked upon as an absence and presence

respectively of interactions betw een different parts of the visual field at or

beyond the initial disparity processing level. A lthough this classification

of stereopsis w as p ro d u ced entirely on the basis of psychophysics it

corresponds well to a physiological p roperty of cells, i.e. their receptive

field size. Cells w ith large receptive fields w ould be able to u n d ertak e

global stereopsis because they can take inform ation from larger areas of

the visual field, w hereas, for exam ple, striate cells, w hich have very

sm all receptive fields, w ould be better suited to the analysis of local

(37)

Fine /coarse

The division of stereopsis into fine and coarse is one of m agnitude

of disparity. A "fine" system w ould be capable of transform ing sm all

disparities into fused percepts of three dim ensional objects. D isparities

th at exceed Panum 's fusional area cause double vision. H ow ever, even

w hen an image appears double due to its large disparity, it still carries a

dep th percept. Thus, the coarse system does not require fusion of the tw o

m o nocular im ages in order to gain inform ation about relative depth.

Physiologically, this concept could co rresp o n d to the responses of

different types of disparity tuned cell. Fine stereopsis could be signalled by

"tuned" cells w hich resp o n d specifically to d isp arities at or n ear to

Panum 's fusional area giving narrow response profiles. Large disparities

w ill n o t stim ulate "tuned" cells, b u t w ill cause "asym metric" cells to

respond because these cells are stim ulated over a large range of disparities

th at exceeds the limits of the horopter, either in front of (near cells), or

behind it (far cells).

C y c lo p e a n /n o n c yclo p e a n

Julesz coined the term cyclopean; it refers to visual stim uli th at

can only be perceived binocularly, for exam ple, the contours p resen t

b e tw ee n areas of d ifferen t d isp a rity in a ran d o m d o t stereo g ram .

Essentially, cyclopean stim uli do dot exist w hen view ed m onocularly.

The "cyclopean eye " is the concept of a system in the brain w here the

d ep th inform ation is extracted by processing disparity. The cyclopean eye

is therefore the p art of the brain able to 'see' depth contours defined by

disparity in random dot stereogram s. A non-cyclopean stim uli will still

appear on the cyclopean eye, b u t unlike the cyclopean image it will also

appear in m onocular images. This concept corresponds to the physiology

(38)

stim u la tio n , b u t w ho's resp o n ses are only tu n e d w h en th ey are

stim ulated binocularly at the correct disparity.

The correspondence problem

In order that two retinal images be fused, the brain has to decide

w h ich p a rts of each m onocular im age correspond to the sam e object

v iew ed by the other eye. This is the correspondence problem . From

Julesz's ran d o m -d o t-stereo g ram studies in the 1960s, w e know th at

binocular correspondence happens early in the visual system, i.e. before

object recognition, and th at the stim ulus requirem ents necessary for

c o rresp o n d en ce (as w ell as stereoscopic d ep th ) do n o t n eed to be

so p h is tic a te d . The exact stim u lu s re q u ire m e n ts n e ce ssa ry for

correspondence have been studied psychophysically for m any years.

Colour and stereopsis

Several lines of evidence have suggested that colour is not used by

the stereom atching system to achieve correspondence. An often used

m ethod of producing stereopsis is to view tw o im ages th ro u g h red and

green anaglyphs. One image, m ade up of red and w hite dots, is view ed

th ro u g h a green filter placed over one eye, producing a black and green

pattern, and the other image, m ade up of green and w hite dots is view ed

th ro u g h a red filter producing a black and red pattern. The tw o patterns

are superim posed, avoiding the need for divergence or crossing of the

eyes (as in Julesz's random dot stereogram s). W hen this is done, the

region of the p attern s containing d isp arate dots is seen stan d in g o u t

vividly in d ep th even though each eye sees it in a different colour. Thus

it w as suggested th at colour cannot be im p o rtan t for the stereoscopic

(39)

coloured filters to explore this phenom ena; polarized filtered random dot

stereogram s containing different coloured dots (red, green, blue and

yellow) w ere prepared and the coloured dots w ere arranged so th at no

spatially corresponding dots w ere of the same colour, the colours being

shifted in one direction. The result w as a stereogram that contained tw o

cues: colour and position; the view er could p a ir the dichoptically

presented dots on the basis of their colour or their position. In this case

co lo u r b ased m atch in g w as possible, b u t the m ore com m on a n d

preferable m atch w as based on the position of the dots, p ro d u cin g an

id e n tic a l sen sa tio n of d e p th as w h en p u re c o n tra st ran d o m d o t

stereogram s w ere used.

Isolum inance studies

Isolum inance is used by psychophysicists (and physiologists) as a

tool to p re se n t stim u li w hose contours are visible d u e p u re ly to

chrom atic (and not luminance) differences. Presenting various types of

stim u li at isolum inance is a m eth o d of disco v erin g w h a t role the

chrom atic contrast sensitive system in the brain plays in the perception of

the p a rtic u la r stim ulus. A stim ulus th at is u n d etectab le or w hose

detection is adversely affected w hen presented at isolum inance w o u ld

probably be analysed by a part of the brain w hich cannot draw know ledge

from chrom atic contrast alone. Various experim ents have been carried

o u t u sin g isolum inance and disparity, som e of w hich have p ro d u ced

conflicting results.

Lu and Fender (1972) found that stereoscopic d ep th perception is

difficult or absent in fine random dot stereogram s com posed of tw o

iso lu m in an t colours rather th an black and w hite. Their stim uli w ere