Area V5 of the Human Brain:Evidence from a Combined StudyUsing Positron EmissionTomography and MagneticResonance Imaging

In pursuing our work on the organization of human visualcortex, we wanted to specify more accurately the po-sition of the visual motion area (area V5) in relation tothe sulcal and gyral pattern of the cerebral cortex. Wealso wanted to determine the intersubject variation ofarea V5 in terms of position and extent of blood flowchange in it, in response to the same task. We thereforeused positron emission tomography (PET) to determinethe foci of relative cerebral blood flow increases pro-duced when subjects viewed a moving checkerboardpattern, compared to viewing the same pattern when itwas stationary. We coregistered the PET images fromeach subject with images of the same brain obtained bymagnetic resonance imaging, thus relating the positionof V5 in all 24 hemispheres examined to the individualgyral configuration of the same brains. This approachalso enabled us to examine the extent to which resultsobtained by pooling the PET data from a small groupof individuals (e.g., six), chosen at random, would berepresentative of a much larger sample in determiningthe mean location of V5 after transformation into Ta-lairach coordinates.

After stereotaxic transformation of each individualbrain, we found that the position of area V5 can varyby as much as 27 mm in the left hemisphere and 18mm in the right for the pixel with the highest significancefor blood flow change. There is also an intersubjectvariability in blood flow change within it in response tothe same visual task. VS nevertheless bears a consistentrelationship, within each brain, to the sulcal pattern ofthe occipital lobe. It is situated ventrolateralry, just pos-terior to the meeting point of the ascending limb of theinferior temporal sulcus and the lateral occipital sulcus.In position it corresponds almost precisely with Flech-sig's Fold 16, one of the areas that he found to bemyelinated at birth.

J. D. G. Watson,1-2 R. Myers,2 R. S. J. Frackowiak,2

J. V. Hajnal,3 R. P. Woods/1 J. C. Ma2ziotta,4

S. Shipp,1 and S. Zeki '

1 Department of Anatomy, University CollegeLondon, London WC1E 6BT, United Kingdom,2 MRC Cyclotron Unit, Hammersmith Hospital,London W12 OHS, United Kingdom,3 GECMarconiLtd., Hirst Research Centre, Wembley HA9 7PP andthe NMR Unit, Hammersmith Hospital, LondonW12 OHS, United Kingdom, and 4 Division ofNuclear Medicine, Department of RadiologicalSciences and Department of Neurology, UCLASchool of Medicine, Los Angeles, California 90024

In a previous study, we identified the positions ofareas V4 and V5 of human visual cortex, using thetechnique of positron emission tomography (PET)and relatively simple visual stimuli that emphasizedcolor or motion, respectively (Zeki et al., 1991). Inextending our work on area V5, we used new anddifferent motion paradigms in other groups of sub-jects, only to notice a certain degree of variability inthe position of that area outside the V1/V2 complexon the occipitotemporal border that was activated andthat we presumed to be area V5. It is conceivable thatthe new motion paradigms we used, being slightlydifferent from the one we had used originally to iden-tify the position of area V5, might have activated otherareas contiguous to V5, and not V5 itself since, in themonkey at least, V5 is known to be surrounded bysatellite areas whose cells are also responsive to visualmotion though in a more complex way (Zeki, 1980;Desimone and Ungerleider, 1986; Tanaka et al., 1986;Wurtz et al., 1990). On the other hand, it was equallyplausible that the position of area V5 itself may notbe constant from one individual to the next Thisseemed likely, especially in view of the fact that thecalcarine sulcus, which constitutes one of the mostdistinctive landmarks of the human brain, is alsomarkedly variable in position (see Talairach et al.,1967, p 211; Belliveau et al., 1991). Such variabilitiesraise important questions, especially when imagingstudies concentrate on single subjects, as in condi-tions of disease. We therefore undertook the presentstudies with two main aims in mind. One was relatedto the intersubject variability in the position of areaV5 and the level of blood flow changes in it in re-sponse to the same task. With sufficient numbers ofsubjects, such a study also affords a comparison ofresults between subgroups, or between individualsand a group. The latter is not trivial since there aremany disease conditions in which the number of sub-jects available is necessarily limited. Thus, implicit inour approach was our second aim, which was to de-termine the extent to which conclusions reached bystudying single subjects would be representative ofthose obtained by studying groups.

The above aims naturally made it important to ob-tain a more accurate picture of the position and rel-ative blood flow increase of area V5 in each individual,by repeating our previous paradigm on a larger num-

Cercbral Conex Mar/Apr 19933 79-94; 1O47-3211/93/M 00

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

ber of subjects and with an improved PET camera, ofincreased sensitivity. However, even improved PETscans would have been insufficient to solve one ofour major problems, and indeed a major problem ofcerebral cartographic studies, which is to relate theposition of a functionally defined cortical area, in ourcase area V5, to the sulcal and gyral anatomy of thebrain, since PET images are low in spatial resolutionand do not highlight such cerebral landmarks. Wethus set out to coregister our PET images with imagesof the same brain obtained with magnetic resonanceimaging (MRI),the latter technique being of far great-er value in revealing the details of cortical anatomy.Coregistration methods (Pelizzari et al., 1989; Evanset al., 1991) have been described and used to revealthe position of regions that are usually otherwise welldefined, such as primary sensory or motor cortex(Grafton et al., 1992). We wanted to go beyond andask whether a functionally defined conical area lyingwell outside the primary visual cortex (VI), in whatused to be known as visual "association" cortex, wouldalso have a consistent relationship to the anatomy ofthe cortex, just as area VI has a definite relationshipto the calcarine sulcus. Coregistering our PET images,obtained from repeated PET measurements within in-dividual subjects, with very high-resolution MR im-ages obtained from the same subjects allowed us notonly to examine the variation in sulcal and gyral anat-omy in the area of interest, but also to relate thisvariability to the position of area V5 as defined by PETresults.

Materials and MethodsWe studied 14 normal volunteers and collected tech-nically satisfactory data from 12 subjects and therefore24 hemispheres. The results from these 12 are re-ported here. Of the 12, nine were male and threefemale; their ages ranged from 21 to 70 years (mean,43 ± 17 years). Two wrote with the left hand. Thedegree of hand preference was scored with a shortquestionnaire based on the Edinburgh MRC hand-edness scale (Oldfield, 1971; Schachter et al., 1987)The scale ranges from —100, indicating complete leftdominance, to +100, for complete right dominance.The two lefthanders scored —65 and 0, one right-hander scored 55, while the remaining nine scored80 or above.

All subjects gave informed written consent. Thestudies were approved by the Hammersmith HospitalMedical Ethics Committee, and permission to admin-ister radioactivity was obtained from the Administra-tion of Radioactive Substances Advisory Committeeof the Department of Health, UK.

Experimental DesignAll subjects underwent 12 sequential scans over thecourse of a single 3 hr session, each scan providingmeasurements of relative regional cerebral blood flow(rCBF). Changes in rCBF were used as an index ofthe local synaptic activity elicited by the presentationof specific visual stimuli (Raichle, 1987). The rCBFwas compared in two states of stimulation, using thefollowing stimuli.

Stimulus ASubjects' eyes were open, with the subject viewing ahigh-resolution Amiga monitor (Commodore Busi-ness Machines Inc., West Chester, PA), at a distanceof 37 cm. The screen was rectangular and at this dis-tance the display covered the central 30° of the visualfield vertically and 40° horizontally. The display con-sisted of a stationary, random, array of approximately600 small black squares each subtending 1°, displayedon a white background.

Stimulus BSubjects' eyes were open, with the same display mov-ing coherently in one of eight directions that changedrandomly every 5 sec in 45° steps from 0° to 315°. Thesmall squares moved en bloc at a speed of 6.2 squares/sec in the horizontal and vertical directions, and 5.8squares/sec in the diagonal directions. The choice ofthe moving stimulus was derived from our experiencewith the physiology of area V5 in the macaque mon-key, which is characterized by a heavy concentrationof direction-selective cells, most of which respondoptimally to small spots or squares moving in theappropriate direction (Zeki, 1974). In particular, wearranged that the stimulus motion should be in dif-ferent directions so as to stimulate as many of thedirectionally selective cells in V5 as possible. All mov-ing stimuli were translated by one pixel per frame (ata 50 Hz screen refresh rate), which was the smoothestmotion that could be generated with our raster-baseddisplay unit.

When the display was stationary subjects were askedto fixate a nominated small square at the center of thescreen. For the moving display, the subjects fixated asmall stationary square at the center. The two stimuliwere alternated from scan to scan and, to avoid anypossible order effects, the series commenced withone stimulus in half the subjects and with the otherstimulus in the other half. Both stimuli were identicalin terms of brightness and contrast; the latter was wellabove 90% and as such would have been expected tostimulate both the magno- and parvocellular systems(Kaplan and Shapley, 1982; Tootell et al., 1988).

Data AcquisitionrCBF was measured by recording the distribution ofcerebral radioactivity following the intravenous in-jection of the freely diffusible positron-emitting "O-labeled tracer H2"O. Any increase in rCBF entails anincrease in the amount of radioactivity recorded fromthat region (Mazziotta et al., 1985; Fox and Mintun,1989).

The measurement of local radioactivity was carriedout by scanning the brain with a CTI 953B PET scan-ner (CTI Inc., Knoxville, TN). The inter-detector col-limating septa, which are used conventionally to limitthe detection of scattered radiation, were removed toincrease the acceptance angle and hence the numberof photons recorded. The inevitable increase in noisedue to scattered photons is more than adequatelycompensated for by the more efficient use of the ad-ministered radioactivity (Townsend et al., 1991) • Spe-

80 Location of Human Area V5 • Watson et al

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

cifically, the point source sensitivity is six to seventimes higher than with a standard machine like theCTI 931 scanner recording with the septa in place(Bailey et al., 1991a). In practice, there is a threefoldincrease in useful counts over the whole brain and a6vefold increase at the center of the field of view(Bailey et al., 1991b). An advantage of this is that lessradiation need be administered per scan and there-fore more scans can be performed in each subject.The averaging of more scans with each stimulus re-sults in improved signal to noise, producing data ofa quality sufficient for the identification of activatedregions in the brains of individual subjects, and isclearly an improvement of substance in the study ofindividual patients. To date, the sensitivity of con-ventional PET scanning has been such that meaning-ful data could not usually be obtained from singlesubjects, especially if the changes in rCBF were notlarge, and this necessitated the averaging of activa-tions from groups of subjects. Here we show that withour more sensitive technique, we can obtain reliabledata from single subjects.

The scanner collects data from 16 rings of crystaldetectors covering an axial field of view of 10.65 cm.The emission data were corrected for the attenuatingeffects of the tissues of the head by using measure-ments made from a transmission scan collected priorto the activations. The corrected emission data werethen reconstructed as 31 axial planes by filtered backprojection with a Hanning filter of cutoff frequency0.5 cycles/pixel. The resolution of the resulting im-ages was 8.5 x 8.5 x 4.3 mm at full-width half-max-imum (FWHM) (Spinks et al., 1992). Each plane wasdisplayed in a 128 x 128 pixel format, with a pixelsize of 2.0 x 2.0 mm. The 31 original planes weretransformed by interpolation to 43 planes, to produceimages with approximately cubic voxels.

Each rCBF measurement began with a backgroundscan lasting 1 min. Ten seconds prior to the end ofthis scan, the subjects opened their eyes and startedto view the computer display. A second 3 min scanstarted immediately after the background scan, andat its beginning an H2

15O infusion was started. Theinfusion was at 10 ml/min and continued for 2 min,and was followed by a 30 sec flush of nonradioactivenormal saline. This procedure differs from our pre-vious technique of generating circulating H2

HO inwhich subjects inhaled C"O2 (Zeki et al., 1991). Theprocess was repeated 12 times in each subject with12 min between scans to allow for the decay of ra-dioactivity to background levels. The integrated countsaccumulated over the 3 min of the second scan, cor-rected for background activity (first scan), were usedas an index of rCBF. On average, each subject re-ceived 1014 MBq of H2"O for each of the 12 scans.

Image TransformationsAll calculations and image manipulations were car-ried out on Sun 3/60 and SPARC computers (SunComputers Europe Inc., Surrey, UK), using ANALYZEversion 5 image display software (BRU, Mayo Foun-dation, Rochester, MN) and PROMATLAB (MathWorks

Inc., Natick, MA). Statistical maps of significant bloodflow change were then derived using SPM software(MRC Cyclotron Unit, London, UK). The PET scanswere analyzed in three ways: as a group in transformed(Talairach) coordinates, as individuals in the samespace, and as individuals untransformed, in their ownanatomical space defined by MR imaging. Figure 1sets out the steps in image manipulation and analysis,in diagrammatic form.

Anatomical StandardizationIn all three analyses the first step was to correct forhead movement between scans by aligning them allwith the first one, using Automated Image Registra-tion (AIR) software specifically developed for the pur-pose (Woods et al., 1992). To analyze the results ob-tained from groups of subjects the 12 realigned imagesfrom each subject were averaged and the intercom-missural (AC-PC) line, linking the anterior to the pos-terior commissure, identified. All the images from allsubjects were then transformed into the standard an-atomical space of the stereotaxic atlas of Talairachand Tournoux (1988), which uses this line as its ref-erence point. This transformation was done by usinglinear proportions and nonlinear resampling algo-rithms (Friston et al., 1991a) using some additionalinformation derived from MRI scans. In this stereo-taxic space each pixel measures 2 x 2 mm, with aninterplane distance of 4 mm. With individual brainsan identical procedure was employed for image re-alignment and stereotaxic normalization prior to sta-tistical manipulation on a subject-by-subject basis,rather than as a group. This was undertaken so thatthe position of activated regions could be reportedin Talairach coordinates for comparison between sub-jects and with previous and future results. For thepurposes of PET to MR image coregistration in indi-viduals, no anatomical transformation was undertak-en

Smoothing of PET ImagesThe PET images (group and individual) were all fil-tered with a low-pass Gaussian filter (FWHM of 5 x5 x 3 pixels, 10 x 10 x 12 mm) to smooth the datain three dimensions (Friston et al., 1990). This servedto increase the signal-to-noise ratio by attenuating thehigh-frequency noise in the images, and by aug-menting the effects of image averaging across sub-jects, when performed.

The filter that we used for individual and groupanalysis was considerably smaller than the filter wehad used with the poorer-resolution and lower-sen-sitivity PET camera in previous group studies (Zekiet al., 1991). The better resolution and sensitivity ofthe PET camera used in this study improved signal-to-noise characteristics sufficiently to permit the useof a smaller filter with equal effect.

Statistical AnalysisWe used the technique of statistical parametric map-ping (SPM) for data analysis (Friston and Frackowiak,1991).

Cerebral Cortex Mar/Apr 1993, V 3 N 2 81

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

PET MRI

PET scans: 12 for eachsubject, 6 tn each

experimental condition

MRI scans foreach subject

subject I subject n

automated realignment ofeach subject's PET scansto the subject's first scan

~lautomated realignment to

the AC-PC planestereotnctlc normalization

image smoothing

ANCOVA (group)

average of all 12 PETscans for optimalanatomical detail

Ininge smoothing

ANCOVA (Individual)

SFMHI Ullagesfor group

Image segmentedto remove brain

coverings

ANCOVA (Individual)

image segmentedand Interpolatedto cubic voxels

automated PET toMRI coregbrranon

SPM|l| linagesfor Individual

SPM|t| Imagesfor Individual

LSPM|t| Imagesrealigned toMR Image

transformadonparameter!

SPM|t| and MR Images superimposed,then displayed as sections

and surface renderings

Figure 1 . Row diagram illustrating the stages of data acquisition, image manipulation, and statistical analysis used in this series of experiments. Note that on the toft, nsidethe broken ouilme, all the images have been transformsd into the sterectaxic coordinate system of Talairach and Toumotn (198B]. On the right, the PET and MRI images werenot thus transformed.

For GroupsWhen activation leads to a change in rCBF, the changemay be confounded by differences in global flow be-tween subjects, or within each subject between scans.We corrected for this confounding effect by perform-ing a pixel-based analysis of the covariance (AN-COVA) of rCBF against relative global CBF, the latterbeing treated as the confounding covariate (Fristonet al., 1990). The ANCOVA is used to calculate, foreach pixel, the mean values of rCBF across subjects(with the global CBF adjusted to 50 ml/dl/min), to-gether with the associated error variance. This is per-formed separately for each of the sequence of 12 scansby pooling the data for each comparable scan acrossall subjects.

To compare the activity elicited in the brain by thestatic and moving stimuli, the difference between thesix mean values of rCBF obtained for each of the twovisual conditions was evaluated, again for each pixel,by use of the / statistic, transformed to the normaldistribution. This generated a statistical parametricmap (SPM{f}) of the areas of significant rCBF changeassociated with the difference in the tasks. On suchan SPM-U}, only pixels whose significance values ex-ceeded a certain threshold were displayed The levelof the threshold is set to correct for the effective num-ber of independent tests constituting the SPM, whichis less than the actual number of pixels because neigh-boring pixels are not truly independent (the theoryand practical aspects of setting the statistical thresh-

olds are to be found in Friston et al., 1991b). Pixelsexceeding threshold were then displayed on coronal,sagittal, and transverse views of the brain as projectionmaps. The stereotaxic coordinates of the most signif-icant sites of change were determined and correla-tions with anatomical areas made by reference to thestandard atlas (Talairach and Tournoux, 1988).

The pixels with maximally significant stimulus-re-lated activation were used to estimate the relative sizeof the changes in rCBF. These values represent ad-justed, average rCBF from spherical regions of ap-proximately 10 mm in diameter centered on the cho-sen coordinates.

For IndividualsA procedure similar to the one described above wasused for analyzing the pattern of rCBF change in in-dividual subjects. The SPM{f} was noisier than for thegroup analysis (six scans from the same individualwere averaged and compared with six others, as op-posed to 72 scans from 12 subjects compared against72). Hence, a secondary Gaussian smoothing filter of4 mm FWHM was applied to the SPM{f} in the x- andy-dimensions. The effect of this filter was a smallchange in the in-plane resolution, on average from9 9 mm FWHM to 10.7 mm.

The average location of significant areas of changeobtained from the group analysis of the 12 subjectswas used to direct the search for the location of areaV5, that is, for a bilateral prestriate area in the an-

82 Location of Human Area V5 • Watson et al

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

terolateral parts of the occipital lobe in which corticalactivation by visual motion might be expected in eachindividual. We were interested in this part of the oc-cipital lobe because (1) our previous work had sug-gested that the zone of maximal rCBF change with amotion stimulus would occur here (Zeki et al., 1991);(2) the evidence from a study of the patterns of cal-losal connectivity and myelination in the human brain,and its relation to the same pattern in the monkeybrain, suggested this region as the most plausible site(Clarke and Miklossy, 1990); and (3) this is the gen-eral region implicated in cerebral akinetopsia (Zihlet al., 1983, 1991), even if the lesion in that patientwas relatively large.

In more precise terms, we first identified the po-sition of area V5 in the group of 12 brains, by deter-mining the region of maximal rCBF change in thispart of the occipital lobe. With V5 so defined, we nextdefined the distance of the search radius, that is, theregions from the center of V5 that we were preparedto consider as belonging to V5 in any given individual,which we set at 15 mm, that is, 1.5 times the FWHMof the primary smoothing filter. This search distanceis arbitrary: we felt that it should be larger than theprimary smoothing filter, as a narrow search based onthe location of V5 determined by image averagingacross all subjects would by definition exclude anindividual site of V5 that was far enough away fromthe mean to contribute little or nothing to that meanposition. On the other hand, too large a distance wouldlead to the inclusion of areas such as the V1/V2 com-plex that we would not consider as candidates for thelocation of V5 in individuals. Finally, we accepted anactivation as significant and as belonging to V5 if itoccurred on at least three contiguous axial planes.

Thus, by limiting the number of pixels interrogat-ed, we could afford to use less stringent statisticalthresholding. Thresholding of the SPM{ t) was carriedout with decreasing degrees of harshness, in arbitrarysteps determined by the SPM software, starting withp < 0.05, corrected for multiple nonindependentcomparisons (a Z score of about 3.8). The thresholdwas then lowered top < 0.001, without a correctionfor multiple comparisons (a Z score of 3.09), if nosignificant activation could be found, and finally to athreshold of p < 0.01 (a Z score of 2.33) until anactivation could be found in the searched region. Thelocation of the pixel with the most significant Z scorewas taken to be V5. Because the primary filtering wascarried out in all three dimensions, the coordinatesof this point lay close to the center of the area ofsignificant change within the appropriate plane, andthis plane was at or near the center of the contiguousplanes when considered in the z-dimension In otherwords, this point lay at or very close to the center ofmass of each V5.

PET-MRI CoregistrationFor the coregistration of SPM and MR images obtainedfrom individual brains, the steps of image realignmentto the intercommissural line and anatomical stan-dardization were omitted. However, the subsequent

filtering, followed by ANCOVA and the generation ofa thresholded SPM{r}, were identical. For each in-dividual, the SPMU} was then coregistered with thesubject's own MRI scan. Such superimposition al-lowed us to determine the position of the region ofmaximal rCBF change in relation to the gyral andsulcal pattern of that brain, with the hope of learningwhether there is any consistent relationship betweenthe two.

The MRI scans were obtained with a 1 tesla PickerHPQ Vista system using a radiofrequency (RF) spoiledvolume acquisition that is relatively spin-lattice re-laxation time (Tl) weighted to give good gray/whitecontrast and anatomical resolution [repeat time (TR)24 msec; echo time (TE) 6 msec; nonselective exci-tation with a flip angle of 35°; field of view in plane25 x 25 cm; 192 x 256 in plane matrix with 128secondary phase-encoding steps oversampled to 256;resolution 1.3 x 1.3 x 1.5 mm; total imaging time 20min]. After reconstruction, the MR images were alsoaligned parallel with the intercommissural line, andinterpolated to yield a cubic voxel size of 0.977 x0.977 x 0.977 mm, which permitted coregistrationwith the PET images.

A PET image with the best possible anatomicaldetail was constructed by averaging the 12 realignedPET scans from each individual. A rigid body coreg-istration with the MRI scan was carried out using theAIR software originally developed for PET to PET re-alignment with adaptations for this purpose (Woodset al., 1992). In separate validation studies, the meanerror in the alignment of the two images was approx-imately 1 mm, with a maximum error of 2 mm (R. P.Woods, unpublished observations). The reorientationparameters in terms of translations in x, y, and z androtations about these axes were calculated. These pa-rameters were saved and used subsequently to co-register the statistical parametric maps of significantrCBF change (which contain little anatomical infor-mation) with the subject's cerebral anatomy as de-scribed by the MRI scan.

ResultsWe describe under separate headings the results ob-tained from the entire group of 12 subjects analyzedas a single group, from selected subgroups of six sub-jects and from individual subjects analyzed as indi-viduals. Implicit in this subdivision is one of the mainaims of this work, namely, the extent to which resultsobtained from small groups or single individuals, interms of location of a functional area and the extentof blood flow change in it, are valid for a larger pop-ulation of brains.

Group Results

V5: The Motion AreasOur first step was to confirm that there is an areasituated on each lateral occipital surface associatedwith the perception of visual motion. We have indeedbeen able to do this: Figure 2 and Table 1 documentthe highly significant foci of increased rCBF induced

Cerebral Cortex Mar/Apr 1993, V 3 N 2 83

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

Tablt 1Cortical areas associated with the perception of visual motion' grouped data from 12subjects comparing visual motai with a static visual image

Region

LeftV5Right V5

Mean

V1/V2

Mean

rCBf (ameans;

Static

54.852.353.6

73.368.470.9

tQpJSlcQ fH'Hrpml/dl/min)

Motion

57454,856.1

78.072.675.3

. Talairachcoordinates

- 4 4 . - 7 0 ,+40 . -68 ,

+2.-88,+4,-88,

00

0+4

/score

7.839,05

11.0311.03

The points recorded are those with the highest / scores within each area. The twopoints given for V1/V2 share the same /score of 11.03, the highest significancevalue that could be accommodated by the SPM program. The coordinates for V5 maybe compared with the group data from the previously reported three sub|ects, that,when reanalyzed in the same fashion as the present experiment, gave locations forV5 of - 4 6 , - 6 2 . + 8 on the left and + 4 2 , - 6 6 , - 4 on the right (Zeki et al,

in our subjects by looking at the dynamic rather thanstatic visual display. As before, we conclude that thisrepresents the motion area V5 in man. The averageincrease in rCBF in a 10 mm diameter spherical re-gion of interest centered on the pixels of most highlysignificant change was 4.7%. The anatomical coordi-nates of the points of maximally significant changeare comparable to those obtained from an earlier groupof three subjects studied on a different PET scanner(Zeki et al., 1991). The present results are more ac-curate because we have since been able to identifythe position of the AC-PC line more accurately usingthe coregistration of PET and MR images. This hasresulted in an anatomically more reliable transfor-mation into stereotaxic space. In fact, when we usedthe improved stereotaxic transformation obtained fromthe present results to reanalyze the stereotaxic posi-tion of area V5 obtained from our earlier group ofthree, the position of V5 in the two studies is quitesimilar (see Table 1). The average blood flow changewas less in the study of the present group comparedto the earlier and smaller group of three (4.7% vs6 1%).

As previously, our results expressed in Talairachand Tournoux coordinates show that human area V5

has a ventrolateral location, at the confluence of theoccipital and temporal lobes, and at the junction ofBrodmann's areas 19 and 37 inferiorly.

Other AreasIn addition to the activation in VI and V2 (see Table1; the difficulty of separating the two areas is discussedin Zeki et al., 1991), the improved resolution andstatistical power of the present experiments revealedseparate areas of activation, not seen in our previousstudy (see Figs. 2, 4). One zone of activation lies inthe cuneus and extends laterally onto the surface ofthe brain superiorly. It may correspond to parts ofareas V3 and V3A identified in the monkey by a com-bination of anatomical and physiological criteria(Cragg, 1969; Zeki, 1969, 1978). This is a topic thatwe shall address in a subsequent report.

Subgroup AnalysisOur experiment with 12 subjects provided seventy-two measurements in each of the two states, and thusrepresents a comparatively large sample size. Moststudies use smaller groups. This made it interestingto ask how representative results obtained from small-er groups would be. One way of approaching the issuewas to take different samples of six subjects from thewhole group and contrast the location and extent ofrCBF changes of human area V5 in them.

There are 924 possible combinations of 6 from 12.It was not practicable to test all of these with ourcurrent software. Various subgroups of six were there-fore chosen in such a way as to maximize the potentialvariability in position and blood flow change of V5,to obtain information about limiting cases. For ex-ample, a subgroup of six individuals was chosen inwhich left V5 was more anteriorly placed; hence, thecomplementary subgroup of the remaining six sub-jects had more posterior locations for left V5. Othersubgroups of six were chosen with the aim of maxi-mizing or minimizing the rCBF increases for V5. Theresults are given in Table 2, which shows that betweentwo groups of six subjects, drawn from a populationof 12, the position of the most significant pixel of V5in stereotaxic coordinates may vary, in the limitingcase, by as much as 13 mm. Group blood flow in-

Table 2Subgroup analysis' venation in the location of the most significant pixel of V5 and the rCBf increases at this pom for selected samples of six sub|ects

Group

12345678

Indrvrtual feature

Left V5. more postenorLeft V5, more anteriorR jht V5, more oitacuuediaJR jht V5, more superolateralLeft V5, greater rCBF increasesLeft V5, less rCBF increasesRight V5. greater rC8f increasesRight V5. less rCBf increases

TalakachcoorrJroies

- 4 0 , - 7 2 . - 4-44, -62,+4+40,-64. 0+48.-B6.+4-42, -66, 0-46, -62.+4+ 38,-66, 0+ 38,-60. 0

rCBF |ad|ustml/dl/min)

State

55.954.652.945.754.656.551.649.3

ed group means;

Mouon

58.257.955.348.158.058.454951.1

RelativerCBfincrease1*1

4.26.04.55.16.33.46.33.7

/so

5.886.837.336.426.135.567.81623

84 Location of Human Area V5 • Watson et al

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

Motion vs. stationary — group

«##**•#

— 12 mm 0 mm 12 mm 24 mm 38 mm

sagittal coronal

transverse

Figure 2 . Data averaged from the entire group of v are transverse images of the brain after stereotaxic normalization, with the distances from the AC-PCplane indicated. A, Anatomical features obtained by averaging an DIOOO now scans from all subiects. S, The arithmetical difference between adjusted mean blood flows for movingand stationary stimuli. C. The SPMU'r values derived from the formal pixel-by-pixel comparison of the adjusted mean blood flows and variances for each of the two conditions.The color scale on the right applies only to row C and reflects the I value of each pixel in the SPM{ t] images. 0. The orthogonal proiections of the statistical comparison at athreshold of p < 0.000001, corrected for multiple comparisons (a mean /value of 6.1). The areas showing significant increases in blood flow are on the lateral occipital surfacesat the junction of areas 19 and 37 of Brodmann (V5). the V1/V2 complex, and an area arising in the cuneus on each side and extending superiorly.

creases recorded in response to a moving stimulusmight range from 3.4% to 6.3%.

Individual ResultsWe wanted to extend our studies to single subjects,since information obtained from a single subject maybe crucial, especially in cases of disease where thenumber of subjects maybe very limited. For example,to date only one good example of cerebral akinetopsiahas been described (Zihl et al., 1983, 1991).

Position of Area V5In the left hemisphere, the position of the most sig-nificant pixel corresponding to V5 in Talairach spacevaried more than on the right (Fig. 3, Table 3). Therange was 27 mm on the left and 18 mm on the right.We could discern no relationship between the indi-vidual locations and age, sex, or handedness score.

Blood Flow Changes in V5There was an average rCBF increase of 39 ml/dl/minin the left V5 areas associated with the perception of

visual motion (range, 1.4-5.4 ml/dl/min). In the righthemisphere, the mean increase was 3-7 ml/dl/min(range, 2.3-5.7 ml/dl/min). In relative terms, themean increases were 7.1% on the left (range, 2.6-9.6%) and 7.2% on the right (range, 4.2-11.5%). Norelationship was found between blood flow changesand age, sex, or handedness. The higher individualblood flow increases were associated with greater sta-tistical significance.

Seventeen of the 24 V5 areas were significant atthe harshest threshold of p < 0.05 (corrected for mul-tiple comparisons), with Zscores from 3.86 to as highas 8.52. Six areas were significant in the next lowerrange, with Z scores of 3.09-3-66. In only one casewas the threshold lowered further in order to discernarea V5 (a threshold of 0.001 < p < 0.01, with a Zscore of 2.46). As the area of search for V5 was con-strained by a knowledge of its location in the averagedgroup results, even this result may be considered sig-nificant. The fact that 17 V5 areas were identified atthe threshold of p < 0.05 (corrected for multiplecomparisons) means that they could have been reported in their own right, with no prior knowledge

Cerebral Cortex Mar/Apr 1993, V 3 N 2 86

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

Tabla 3Locaums for the most significant pud of V5 in the 12 individual subjects, with then associated rC8F increases

Left V5 RtgtaV5

Sub-ject

n163n164n172n180n185n191n192n197n206n210n216n221

MeanSE

Tafa»achcoordinates

- 4 4 . - 6 2 . + 8- 3 8 . - 6 6 . 0- 4 6 , - 6 2 . + 4- 4 2 , - 7 4 , - 4-36 , -72 , 0-44.-62.+S-46, -68 , 0- 4 6 , - 7 0 . - 4-48. -64 . 0-40 . -72 . 0- 3 0 . - 8 2 . - 4-34. -68 , 12

-41. -69.+25.6, 6.0. 5 3

rCBf

Stat-ic

49.353.656.353.753.656256.058.157.755.454.047.0

54.23.3

Mo-tion

52.755.061.756.956.361.161.461.962.05B.759.2502

58.13.9

In-crease

3.41.45.43.22.74.95.43.84.33.35.232

3.91.2

In-crease

1*1

6.92.69.66.05.08.79.66.57.56.09.66.8

7121

Zscore

3.332424.13539424.34.5448.542

Tata irachcoordinates

+46,-66, 0+38,-66,-4+46,-64,+4+44,-66, 0+40,-68, 0+40,-78,+4+40,-68, 0+34,-64, 0+42.-60.+4+42,-72,+4+40.-68.+4+38,-62.+8

+41.-67.+23.7. 4.7. 32

rC8F

Stm-1C

51.757.349552.353546.649.749.655.442254.445.1

5064.4

Mo-tion

54.963.053.657.156.149.955.452.959244.756.747.8

54.35.0

In-crease

3.2574.1482.63.35.73.33.82.52.32.7

3.712

In-crease

1%)

6210.08.3924.971

11.56.76.95.9426.0

7221

Zscore

2.54.44.05.73.94.54.34.15.33.13.33.3

The rCBF measurements are expressed in ml/dl/min, adjusted to a mean whole bram btood flow of 50 ml/dl/min for each subject

of their likely positions anywhere in the brain (Fristonet al., 1991b).

Area V1/V2Our confidence in all the above results was reinforcedby the control built in to the experiment, namely, theactivation of areas VI and V2, from which V5 receivesits input. We have discussed elsewhere the difficultiesof separating VI and V2 in these PET images (Zekiet al., 1991), but we nevertheless expected that inevery subject, and in groups of subjects, the activityin V1/V2 must lie along the calcarine sulcus, one ofthe most conspicuous landmarks in the cerebral cor-tex and whose position is easily determined from theMR images. The position of the calcarine sulcus ishighly variable in individuals, even after normaliza-tion to the stereotaxic framework of Talaraich andTournoux (Talairach et al., 1967, p 211; Belliveau etal., 1991). Significant activation was detected in allindividual subjects at atlas coordinates consistent withthe location of the calcarine sulcus, or along the lin-gual gyrus beneath it. Coordinate locations alone werenot then sufficient to determine the precise anatom-ical location of this functional activation, but the in-dividual PET/MR coregistrations were able to showthat this activation center was truly centered alongthe course of the calcarine sulcus in all subjects, andthus was felt to represent area V1/V2.

We were also interested to note that, in additionto the rCBF changes in V5 and the cuneus, there werealso changes in the parietal lobe, corresponding toBrodmann's area 7 (Fig 4). This was found in all foursubjects for whom the head positioning in the PETcamera was such that we were able to scan as high asthe vertex.

Coregistration of Individual PET Results urttb MRIThe coregistration of PET SPM{/} images with theMRI scans obtained from the same brain allowed us

to examine the anatomical location of the activatedareas in relation to the sulcal and gyral pattern of theoccipital lobe more precisely than had been previ-ously possible. Figure 6 shows surface renderings ofboth hemispheres from four subjects. The zones ofactivation were not exclusively superficial, but havebeen projected onto the surface of each hemispherein order to illustrate their relation to the overall cor-tical topography. Figure 5 shows a series of slicesthrough the brain shown in the top row of Figure 6.The bilateral sites of activation corresponding to V5can be traced through at least five or six slices. Theirprecise size depends on the level of filtering and sta-tistical thresholding, but each is comparable in extentto the width of a gyrus. The focal points of activationare recessed from the cortical surface. In the righthemisphere, at least, the focus occupies the posteriorbank of a relatively deep fissure that, by reference toFigure 6, is seen to be a vertically oriented sulcus,one that might be identified as the posterior contin-uation of the inferior temporal sulcus. The same maybe true for the focus in the left hemisphere, thoughthe judgment is rather more equivocal. As elsewherein the human cerebrum, the disposition of sulci inthe occipital lobe is highly variable and there is acorresponding imprecision in the terminology. Wethus postpone to the Discussion the question of aconsistency in the relationship of a functionally de-fined cortical area, V5, which we were able to identifyin all 24 hemispheres, to a gyral configuration that israther less predictable.

DiscussionThe group results from 12 subjects revealed areas ofthe human visual cortex that were differentially andcommonly activated by the visual motion stimulusthat was used. These included the area that we hadpreviously identified as the human equivalent to mon-key V5, V1/V2, and other sites not reported before in

86 Location of Human Area V5 • Watson et al

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

Figure 3. Data from three individual sublets to show the maximum range in the position of V5 on each side, after stereotaxic normalization. A, C. and E, Transverse imagesof blood flow for each subject to show anatomical detail, obtained by averaging the 12 scans from the subject. Distances from the AC-PC plane are indicated. B. 0, and F.SPfvUf} images obtained for each subject. The locations of the most significant pixel for left V5 are. in B, - 4 8 , - 6 4 , 0 ; 0. - 4 4 , - 6 2 , + 8 ; and F. - 3 0 . - 8 2 , - 4 . For rightV5. the locations are + 4 2 . - 6 0 . + 4 in ft + 4 0 . - 7 8 . + 4 in 0. and + 4 0 . - 6 8 . + 4 in F. These locations are indicated by the M e circles. The distance separating left V5in D and F is 27 mm, while the separation between the right V5 of B and 0 is 18 mm. The three sets of SPMif } images are scaled to the same range of Zvalues, indicatedby the color scale on the right.

terms of a specific motion stimulus, nor yet charac-terized in terms of homology with monkey occipitaland parietal cortex. Second, we were able to find thesesame regions of significant rCBF change in individuals—particularly in area V5, located bilaterally in theanterolateral part of the occipital lobe. Third, thisallowed us to gauge the variability between individ-uals undertaking the same task, both in the positionof area V5 and in the magnitude of its rCBF change.Finally, we used high-resolution MRI scans to deter-mine how far the variable location of V5 could be

ascribed to individual variation in patterns of gyralanatomy, as distinct from inconsistencies in stereo-taxic normalization.

72«? Identification of Area V5The results reported here, based on a study of 12subjects and using an improved PET camera with ahigher sensitivity, confirm the results we obtainedfrom our previous study (Zeki et al., 1991) of threesubjects scanned in an older PET camera (CTI-Sie-mens 931) (Table 1). The present results allowed us,

C e r e b r a l C o r t e x M a r / A p r 1 9 9 3 , V 3 N 2 I 7

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

Figure 4 . Data from a different single subject to those shown in Figui ' -is on transverse planes parallel with the AC-PC plane obtained by averagingthe 12 blood flow scans, after stereotaxic normalization, fi, The SPMfr } images tor this subject. The images share a constant colorscs/efor the /values in each pixel, indicatedon the right. C, The orthogonal projections of the SPM{r} at a threshold of p < 0.001, uncorrected for multiple comparisons. V5 can be identified on both sides, as can areasof increased blood flow in the V1/V2 complex and both cuneate areas, extending superiorly. Also seen are areas of inaeased blood flow in the superior parietal lobes, from theborder of the occipital lobe forwards, corresponding to Brodmann's area 7.

however, to obtain a clearer picture of the variabilityin the position of area V5. Moreover, the use of thenew camera made it possible to coregister the PETimages with the MR images and thus relate the po-sition of area V5 to the sulcal and gyral anatomy ofthe occipital lobe.

Our previous study, perhaps because of the lowersensitivity of the camera, revealed only one candidateprestriate area that was consistently present when mo-tion was the critical stimulus. In addition to the resultitself, there was more than one line of reasoning thatsuggested that the active area in our previous studymust be V5. Its location was closely similar to thatindicated by Clarke and Miklossy (1990) from their

anatomical studies. They used two criteria: one relat-ed the position of the prestriate areas to the patternof callosal distribution within that zone of cortex,since cortical areas have a definite relationship to stripsof callosally connected cortex (Zeki, 1977a,b), andthe other related to the pattern of myelination withinthe prestriate cortex. These anatomical studies showedthat V5 must be located fairly ventrally on the lateralsurface of the occipital lobe. It was this very territorythat was compromised in the akinetopsic patient de-scribed by Zihl et al. (1983, 1991), although the le-sions went beyond the confines of the area that wenow define as V5. It thus seems more than plausibleto suppose that the area we describe here and else-

I I Location of Human Area V5 • Watson et al

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

Figure 5. Data from a single subject in which the MR images and SPM:/:- images have been reregistered and superimposed. Axial slices at 4 mm intervals are depicted,parallel with the AC-PC plane, which is indicated by ACPC underneath the relevant slice. The SPM { ( ( images share a common color scale for their pixels' I values, indicated onthe right. At the AC-PC level there is activation along the calcarine sulcus. reflecting activity in the V1/V2 complex. V5 is present on each side and is indicated by arrows: inparticular, the relationship with underlying suici is seen. On both sides there is also significant activation inferiorly in the lingual gyri and superiorly, arising from the cuneus andcontinuing upward.

where is human V5. As far as we can tell from theabstract published by Miezin et al. (1987), an areaoverlapping with the one reported here but extendingto the parietal cortex was among the areas identifiedusing a visual motion stimulus.

The Variability in the Position of Area V5in IndividualsUsing the same general strategy of image realignmentto the intercommissural plane, followed by stereotax-ic normalization, we found that the previously re-ported study on V5 in three subjects, analyzed in thesame way as the present 12 subjects (using the samestereotaxic transformation), would place V5 on theleft more anterosuperior to the location here deter-mined, by some 12 mm (Table 1). On the right thedifference is only some 5 mm. Our supposition wasthat such differences might be due to the variation inthe precise position of V5 in different individuals,rather than to any systematic difference introducedby the new PET machine. Closer analysis of the resultsconfirms this.

The location of V5 in Talairach coordinates wasdetermined for both hemispheres in each individual

from SPM{f} maps, following the same image realign-ment and stereotaxic normalization used for the groupanalysis. When the 12 subjects were analyzed as in-dividuals, the extreme variation in position of V5 was27 mm on the left and 18 mm on the right (Table 3).The analysis of the subgroups of six presented inTable 2 also demonstrates that such selections of sub-groups, tested with an identical experimental paradigm, can yield mean locations for V5 that differedfrom each other by as much as 13 mm.

A similar analysis of variability can be carried outfor changes in rCBF. The maximum relative individ-ual increase in rCBF for V5 in the present study was11.5% and the minimum was 2.6% (Table 3). Theaverage of all the individual values was 7.1% on theleft and 7.2% on the right. However, the increasesobtained when the 12 subjects were incorporated intoa single group prior to analysis were 4.7% on the leftand right (Table 1). The analysis of subgroups of six(Table 2) shows that relative increases in blood flowreported for V5 could vary from 3-4% to 6.3% on theleft and 35% to 58% on the right. The rCBF changesderived from the averaged data are smaller than thoserecorded in individuals because of the combination

Cerebral Cortex Mar/Apr 1993. V 3 N 2 81

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

Figure 6 . The cerebral hemispheres from four subiecis, showing each V5 area as defined by the PET activation experiments, superimposed on the individual's own MR image.Each subject occupies a row: the iirsi row is the subject shown in Figure 5. The images were derived from slice data such as that presented in Figure 5, but are now displayedas surface-rendered objects viewed at rotations of 90° and 50° from the occipital pole, to allow the patterns of sulci and gyri to be seen. In each subject the PET SPMfr ' , imagewas edited to leave only V5. The statistical threshold was lowered so that the PET image was contiguous with the cortical surface of the MRI after the PET and MR imageswere coregistered and superimposed, so that the process of surface rendering (to a depth of some eight pixelsl does not falsely locate the site of V5 in terms of surface features.The PET activation sites were rendered as red areas on the final images.

of spatial filtering (which reduces the peaks of rCBFchange) and the variation in the location of V5 acrossindividuals such that the anatomical transformationinto Talairach coordinates fails to superimpose thefoci from different individuals onto one another. Witha set filter size, as more subjects are added to a group,the average increase recorded in rCBF will tend todecline (Raichle et al.. 1991), at least down to a certain point.

Group versus Individual StudiesWe consider here the relative merits of studying groupsand single individuals. Each type of study providesimportant information but has problems attached toit. The necessity for using groups in PET experimentsarose from the simple problem that, at the permitteddoses of administered radiation, the signal-to-noiseratio was generally too low to inspire confidence inresults obtained from single subjects. Because theyincrease the signal-to-noise ratio, grouped data aremore reliable. They extract the most significant andcommonplace activation sites even if they sacrificethe individual patterns of activation; some of the lat-ter, although divergent from the mean, may never

theless contain important lessons. In addition,grouped results in reference to standard stereotaxiccoordinates allow an easier comparison of results ob-tained with different paradigms, in different labora-tories, and with different groups of subjects. Themethod has its drawbacks, however, especially in re-lating the site of activation to the anatomy of the ce-rebral cortex. In addition, the method is awkward touse when single patients, manifesting a rare condi-tion, are the subject of study. These drawbacks canbe compensated for by the use of more advancedtechniques, such as we have used, which allow therepeated scanning of single subjects at the same doseof radiation and give results with greater statisticalconfidence. Because single subjects are used, the PETimages can be coregistered with MR images. The siteof activation can thus be directly related to the anat-omy of the cerebral cortex. Nevertheless, the study ofsingle subjects inspires greater confidence when theresults obtained from it are, so to speak, validated bygroup results that show that a similar area is activewhen the same task is used in a group of subjects.For example, in this study we used the position of V5as defined in the analysis of the group of 12 subjects

SO Location of Hi. VS • Watson et al.

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

as the basis on which to search for V5 in each indi-vidual. We thus tend to use both approaches simul-taneously, preferring the former for the screening ofthe cerebral areas involved in the execution of a taskand the latter for a more detailed study of the area inrelation to the anatomy of the brain.

PET studies often use small groups of about six toeight subjects. This small size, along with the individ-ual biological variation, the characteristics of low-passfiltering and the statistical methods, will all combineto make the reporting of results in terms of standardstereotaxic coordinates uncertain. The subgroup anal-ysis performed in this study showed that intersubjectvariability alone can allow a reported activation siteto vary by close to the width of a gyrus. Put moresimply, even standardized coordinates are uncertainEven if they were certain, relocating them onto thesingle brain used in the Talairach and Tournoux sys-tem will only be valid as far as that brain, measuredin the postmortem and fixed state, reflects the actualanatomy of the individuals that made up the PET studygroup.

V5 Areas in Individuals, Coregistered withMSI ScansHaving identified the location of area V5 on the SPM{/}maps for each individual subject after stereotaxic nor-malization, we used the same statistical approach tofind the same areas for each subject, but this timewithout anatomically transforming the PET images.In other words, the locations were now expressed inthe unique reference framework of each subject's ownbrain, and not in terms of a standard reference frame-work. It was then possible to coregister the individualPET results with the corresponding MR images, whichin turn allowed us to locate areas such as V5 withrespect to the more detailed anatomy of the brain interms of sulci and gyri. In gyrencephalic brains, afocus of activation might fall on the surface of thecerebral cortex, or it might be buried deep within asulcus and thus not be visible on the surface. Becauseof this, one needs both two- and three-dimensionalreconstructions for the process of anatomical local-ization (Figs. 5, 6). Two-dimensional reconstructionsare vital for identifying structures lying within sulci(as V5 may be), while three-dimensional reconstruc-tions are instrumental in identifying the detailed sur-face configuration of the brain in terms of sulci andgyri, and the relationship of an activated area to them.Hence, an added advantage of using MR images withvoxel dimensions less than 1.5 mm was the ability tomake such three-dimensional reconstructions. Thelatter helped us to identify the pattern of sulci andgyri in the occipital lobes of all our subjects (24 hemi-spheres) and thus make a statement about the rela-tionship of area V5 to the sulcal and gyral pattern ofthe occipital lobe of the human brain.

The Pattern of Gyri and Sulci in theOccipital Lobe of the Human BrainTopographical-functional relationships in the humancerebral cortex have been known to exist for many

years. Chief among these is the relationship of theprimary visual cortex to the calcarine sulcus of theoccipital lobe (Henschen, 1893; Flechsig, 1905). Ittherefore seemed plausible that such relationshipsmight be found elsewhere in the occipital lobe. Thereis, however, a great deal of variability in the patternof fissuration of the occipital cortex. Perhaps for thisvery reason, the literature contains a scanty and un-satisfactory description of the gyral anatomy of thelateral occipital lobe. There are nevertheless somesulcal patterns that seem to bear a more or less con-sistent relationship to the position of human V5. Wetherefore provide a brief description of the occipitallobe with the hope of specifying the position of hu-man V5 with as little ambiguity as possible.

Most descriptions of the lateral occipital lobe di-vide it into gyri. Some consider that there are threegyri—the upper, middle, and lower occipital gyri (Ta-lairach and Tournoux, 1988); others divide it into justtwo, the lateral and superior gyri, while admitting thatthe lateral gyrus may have further subdivisions (Cun-ningham, 1902; Toldt, 1908). These descriptions relyon a horizontally oriented sulcus, the lateral occipi-tal, of which there are sometimes two lying parallelwith each other (Ono et al., 1990). By contrast, theInternational Anatomical Nomenclature Committee(1989) recognizes only two sulci, the transverse oc-cipital and the variable lunate, both of which areangled dorsoventrally and hence do not subdivide theoccipital lobe into gyri stacked upon one another. Wefound that the most useful landmark for identifyingthe position of area V5 is the posterior continuationof the inferior temporal sulcus. This is present in thegreat majority of hemispheres, though it adopts a va-riety of forms. Normally it runs dorsoventrally at theborder between the temporal and occipital lobes; itis so identified by Cunningham (1902), who refers toit as the ascending limb of the inferior temporal sul-cus, and by Toldt (1908), who calls it the anterioroccipital sulcus. As the latter term has also been ap-plied to other nearby sulci (Ono et al., 1990), weprefer the former term, which we abbreviate to ALITS.V5 lies at the junction of ALITS with the lateral oc-cipital sulcus. Where the two do not actually intersect,as commonly happens, V5 is to be found near theirinterpolated meeting point (e.g., Fig. 6, first row,right). The site of activation may fall within the sulcus,at least in part (Fig. 5). In fact, the inferior temporalis a highly interrupted sulcus that is not always easilyidentified. In some hemispheres its posterior contin-uation is better identified as a separate, transverselyoriented sulcus, often extending over the inferola-teral margin of the occipital lobe (see Fig 7). Again,V5 is situated posterior to this sulcus, near its dorsaltermination (e.g., Fig. 6, first and third rows, left; Fig.8).

One or the other of these patterns sufficed to de-scribe the location of V5 in 16 of our 24 hemispheres.In the remainder, the patterns were suggestive butless clear cut. On the basis of its dense myelination,Clarke and Miklossy (1990) located an area that theyconsidered to be area V5 in a similar region. In the

Cerebral Cortex Mar/Apr 1993, V 3 N 2 81

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

H g a n 7. Line diagrams of the left hemispheres of three sub|ects to show variation ai the surface anatomy of the occipital lobs. The iflper row corsairs lateral views, whilethe lower nm has views with a 30° rotation, i, ascending limb of nlenor temporal sulcus. also known at anterior ocdphai o i n n ; b, toterai occipital t u b u ; c. tip of cafcarinesuteus.

sketch of the brain they provide, the inferior temporalsulcus runs more or less continuously into the lateraloccipital, and lacks an obvious ascending limb. De-spite this somewhat different sulcal anatomy, the lo-

B

Rgnro B. A. The position of human area V5 in an individual brain, 8S determinedm this study. B. Rating's diagram of the myetoarchrtBCture of the human brain.Compare the position of FsU 16 with that of V5 in A.

cation they illustrate is quite consistent with the lo-cation of area V5 derived from our studies.

Not the least interesting feature of a location forV5 just posterior to ALITS is that it coincides almostexactly with Flechsig's Feld 16 in what he calls thegyrussubangularis (Flechsig, 1920) (see Fig. 8). ThisFeld 16, which is classed among the Pramature Rin-denfelder, thus belongs to cortex which is myelinatedat birth, though Flechsig states that it is not as heavilymyelinated at birth as the other areas comprising thisgroup, among which he numbers the calcarine cortex(area VI). According to Flechsig, the myelinationspreads postnatally to include Feld 27, lying just an-terior. If area V5 is well demarcated at birth, by virtueof its early myelogenesis, it is not inconceivable thatherein lies the link between function and gyral mor-phology—the arrival of afferents to this zone perhapspromoting an infolding of the cortical sheet. We notein the same context that Bailey and von Bonin (1951,their Fig. 7) show the "anterior occipital sulcus" al-ready beginning to form in the 30 week embryo.

Other AreasWe suggest provisionally that the areas of activationfound bilaterally in the cuneus and the lingual gyriare part of one functional area that corresponds toarea V3 in the monkey. On topographical grounds, itis reasonable to expect V3 to lie juxtaposed with theVI/V2 complex, both from the anatomical findings inmacaque and from the recent arguments based onlesion studies in man (Horton and Hoyt, 1991). Fur-thermore, in all subjects scanned high enough, therewas a ribbon-like area of activation in parietal cortex,extending forward from the parieto-occipital fissure.

12 Location of Human Area V5 • Watson et al

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

This is Brodmann's area 7, a region that in the monkeyat least is known to receive input from V5, and otherprestriate areas (Maunsell and Van Essen, 1983; Un-gerleider and Desimone, 1986; Cavada and Goldman-Rakic, 1989; Andersen et al., 1990; Zeki, 1990). It isof interest that a previous study, intended to chart theretinotopic organization of striate cortex, also re-vealed a somewhat similar distribution of activatedregions in prestriate cortex—although V5 was not spe-cifically identified (Foxetal., 1987). In fact, that studyemployed an alternating checkerboard stimulus. Thelatter lacks coherent motion but is otherwise similarto ours in its spatiotemporal properties.

In summary, we show that methodological im-provements in PET scanning, and the coregistrationof PET and MR images, allow the organization of vi-sual pathways to be studied in single subjects. Thisapproach complements studies in groups of subjectsthat are essential to an identification of the commoncortical areas related to particular functions, especial-ly when the blood flow changes elicited are small.Studies in individuals will be invaluable and of crucialimportance in those informative but rare patients withselective loss of cortical function.

NotesWe are most grateful to all our volunteers, who provided somuch time and effort for this study. We thank Professor GM. Bydder, Dr. W. Curati, Professor I. R. Young, Mrs. S. White,and Miss K. Williams of the NMR Unit at the Royal Post-graduate Medical School, Hammersmith Hospital, for theirencouragement and their help in performing the MRI scans.From the MRC Cyclotron Unit we thank Mr. G. Lewington,Ms. C. Taylor, and Ms. A. Williams for carrying out the PETscans, Dr. J. Clarke and the radiochemistry section for de-veloping the water generator, and Dr. T. Jones, Mr. D. L.Bailey, Dr. T. J. Spinks, and Ms S. Grootoonk for the intro-duction of three-dimensional PET scanning. We also thankMr. J. Romaya, from the Department of Anatomy, UniversityCollege, London, for developing the computer displays ofvisual stimuli.

Correspondence should be addressed to S. Zeki, De-partment of Anatomy, University College London, LondonWC1E 6BT, UK, or to R. S. J. Frackowiak, MRC CyclotronUnit, Hammersmith Hospital, London W12 OHS, UK.

ReferencesAndersen RA, AsanumaC, Essick G, Siegel RM (1990) Cor-

tico-cortical connections of anatomically defined subdi-visions within the inferior parietal lobule. J CompNeurol296-65-113.

Bailey DL, Jones T, Spinks TJ (1991a) A method for mea-suring the absolute sensitivity of positron emission to-mographic scanners. Eur J Nucl Med 18:374-379.

Bailey DL, Jones T, Spinks TJ, Gilardi M-C, Townsend DW(1991b) Noise equivalent count measurements in a neu-ro-PET scanner with retractable septa. IEEE Trans MedImaging 10:256-260.

Bailey P, von Bonin G (1951) The isocortex of man. Ur-bana: University of Illinois.

Belliveau JW, Kennedy DN, McKinstry RC, Buchbinder BR,Welsskoff RM, Cohen MS, Vevea JM, Brady TJ, Rosen BR(1991) Functional mapping of the human visual cortexby magnetic resonance imaging Science 254:716-719.

Cavada C, Goldman-Rakic P (1989) Posterior parietal cor-tex in rhesus monkey: I Parcellation of areas based ondistinctive limbic and sensory corticocortical connec-tions. J Comp Neurol 287:393—421.

Clarke S, MiklossyJ (1990) Occipital cortex in man: or-ganization of callosal connections, related myelo- andcytoarchitecture, and putative boundaries of functionalvisual areas. J Comp Neurol 298:188-214.

Cragg BG (1969) The topography of the afferent projec-tions in the circumstriate visual cortex of the monkeystudied by the Nauta method. Vision Res 9:733-747.

Cunningham DJ (1902) Text-book of anatomy. London:Pentland.

Desimone R, Ungerleider LG (1986) Multiple visual areasin the caudal superior temporal sulcus of the macaque.J Comp Neurol 248.164-189.

Evans AC, Marrett S, Torrescorzo J, Ku S, Collins L (1991)MRI-PET correlation in three dimensions using a vol-ume-of-interest (VOI) atlas. J Cereb Blood Flow Metab11:A69-A78.

Flechsig P (1905) Gehirnphysiologie und Willenstheo-rien. Paper presented at Fifth International PsychologyCongress, Rome Reprinted in: Some papers on the ce-rebral cortex (von Bonin G, trans), pp 181-200. Spring-field, IL: Thomas, I960

Flechsig P (1920) Anatomie des menschlichen Gehirnsund Ruckenmarks. Leipzig. Thieme

Fox PT, Mintun MA (1989) Noninvasive functional brainmapping by change-distribution analysis of averaged PETimages of H2

15O tissue activity. J Nucl Med 30.141-149.Fox PT, Miezin FM, Allman JM, Van Essen DC, Raichle ME

(1987) Retinotopic organisation of human visual cortexmapped with positron emission tomography. J Neurosci7:913-922.

Friston KJ, Frackowiak RSJ (1991) Imaging functional anat-omy. In: Brain work and mental activity. Quantitativestudies with radioactive tracers (Lassen NA, Ingvar DH,Raichle ME, Friberg L, eds), pp 267-279. Copenhagen:Munskgaard.

Friston KJ, Frith CD, Liddle PF, Dolan RJ, Lammertsma AA,Frackowiak RSJ (1990) The relationship between glob-al and local changes in PET scans. J Cereb Blood FlowMetab 10:458-466.

Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ (1991a)Plastic transformation of PET images. J Comput AssistTomogr 15:634-639.

Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ (1991b)Comparing functional (PET) images: the assessment ofsignificant change. J Cereb Blood Flow Metab 11-690-699.

Grafton ST, Mazziotta JC, Woods RP, Phelps ME (1992)Human functional anatomy of visually guided fingermovements. Brain 115-565-587.

Henschen SE (1893) On the visual path and centre. Brain16:170-180.

HortonJCHoytWF (1991) Quadrantic visual field defects:a hallmark of lesions in extrastriate cortex (V2/V3). Brain114:1703-1718.

International Anatomical Nomenclature Committee (1989)Nomina anatomica, 6th ed. London: Churchill Living-stone.

Kaplan E, Shapley RM (1982) X and Y cells in the lateralgenlculate nucleus of macaque monkeys. J Physiol (Lond)330:125-143.

Maunsell JHR, Van Essen DC (1983) The connections ofthe middle temporal visual area (MT) and their relation-ship to a conical hierarchy in the macaque monkey. JNeurosci 3:2563-2586.

Mazziotta JC, Huang SC, Phelps ME, Carson RE, MacDonaldNS, Mahoney K (1985) A non-invasive positron com-puted tomography technique using oxygen-15 labeledwater for the evaluation of neurobehavioral task batteries.J Cereb Blood Flow Metab 5:70-78.

Miezin FM, Fox PT, Raichle ME, Allman JM (1987) Localized responses to low contrast moving random dot pat-terns in human visual cortex monitored with positronemission tomography. Soc Neurosci Abstr 13:631.

Oldfield RC (1971) The assessment and analysis of hand-

Cerebral Cortex Mar/Apr 1993, V 3 N 2 93

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from

edness- the Edinburgh inventor)'. Neuropsychologia 9:97-113.

Ono M, Kubik S, Abernathey CD (1990) Atlas of the ce-rebral sulci. Stuttgart: Thieme.

Pelizzari CA, Chen GT, Spelbring DR, Weichselbaum RR,Chen CT (1989) Accurate three-dimensional registra-tion of CT, PET, and/or MR images of the brain. J ComputAssist Tomogr 13:20-26.

Raichle ME (1987) Circulatory and metabolic correlates ofbrain function in normal humans. In: Handbook of phys-iology: the nervous system, Vol V (Plum F, ed), pp 643-674. New York: American Physiology Society-Oxford UP.

Raichle ME, Mintun MA, Shertz LD, Fusselman MJ, MiezenF (1991) The influence of anatomical variability on thefunctional brain mapping with PET- a study of intrasub-ject versus intersubject averaging. J Cereb Blood FlowMetab 11:S364.

Schachter SC, Ransil BJ, Geschwind N (1987) Associationsof handedness with hair color and learning disabilities.Neuropsychologia 25:269-276

Spinks TJ, Jones T, Bailey DL, Townsend DW, Grootoonk S,Bloomfield PM, Gilardi MC, Casey ME, Sipe B, Reed J(1992) Physical performance of a positron tomographfor brain imaging with retractable septa. Phys Med Biol8.1637-1655.

TalairachJ, Toumoux P (1988) Co-planar stereotaxic atlasof the human brain. Stuttgart. Thieme.

Talairach J, Szikla G, Tournoux P, Prossalentis A, Bordas-Ferrer M, Covello L, Iacob M, Mempel E (1967) Atlasof stereotaxic anatomy of the telencephalon. Paris- Mas-son.

Tanaka K, Hikosaka K, Saito H, Yukie M, Fukada Y, Iwai E(1986) Analysis of local and wide-field movements inthe superior temporal visual areas of the macaque mon-key. J Neurosci 6:134-144

Toldt C (1908) Anatomischer Atlas fur Studierende undArzte. Berlin: Urban and Schwarzenberg.

Tootell RBH, Hamilton SL, Switkes E (1988) Functionalanatomy of macaque striate cortex IV: Contrast and mag-no-parvo streams. J Neurosci 8:1594-1609.

Townsend DW, Geissbuhler A, Defnse M, Hoffman EJ, SpinksTJ, Bailey DL, Gilardi M-C, Jones T (1991) Fully three-dimensional reconstruction for a PET camera with re-tractable septa. IEEE Trans Med Imaging 10.505-512.

Ungerleider LG, Desimone R (1986) Cortical connectionsof visual area MT in the macaque J Comp Neurol 248.190-222.

Woods RP, Cherry SR, Mazzlotta JC (1992) A rapid auto-mated algorithm for accurately aligning and reslicingpositron emission tomography images J Comput AssistTomogr 16:620-633.

Wurtz RH, Yamasaki DS, Duffy CJ, Roy J P (1990) Func-tional specialization for visual motion processing in pri-mate cerebral cortex. Cold Spring Harbor Symp QuantBiol 55:717-727.

Zeki SM (1969) Representation of central visual fields inprestriate cortex of monkey. Brain Res 14:271-291

Zeki SM (1974) Functional organization of a visual area inthe posterior bank of the superior temporal sulcus of therhesus monkey. J Physiol (Lond) 236 549-573

Zeki SM (1977a) Colour coding in the superior temporalsulcus of the rhesus monkey visual cortex. Proc R SocLond [Biol] 197:195-223.

Zeki SM (1977b) Simultaneous anatomical demonstrationof the vertical and horizontal meridians in areas V2 andV3 of rhesus monkey visual cortex Proc R Soc Lond [Biol]195:517-523

ZekiSM (1978) The third visual complex of rhesus monkeyprestriate cortex. J Physiol (Lond) 277:245-272.

Zeki S (1980) The response of cells in the anterior bankof the superior temporal sulcus in macaque monkeys. JPhysiol (Lond) 308:85P

Zeki S (1990) The motion pathways of the visual cortexIn Vision, coding and efficiency (Blakemore C, ed), pp321-345. Cambridge- Cambridge UP

Zeki S, Watson JDG, Lueck CJ, Friston KJ, Kennard C, Frac-kowiakRSJ (1991) A direct demonstration of functionalspecialization in human visual cortex. J Neurosci 11:641—649.

Zihl J, von Cramon D, Mai N (1983) Selective disturbanceof movement vision after bilateral brain damage. Brain106:313-340.

Zihl J, von Cramon D, Mai D, Schmid Ch (1991) Distur-bance of movement vision after bilateral posterior braindamage Further evidence and follow up observations.Brain 114:2235-2252

Location of Human Area V5 • Watson et al

at University C

ollege London on D

ecember 23, 2011

http://cercor.oxfordjournals.org/D

ownloaded from