4
Brief Communication 731 Form and motion coherence activate independent, but not dorsal/ventral segregated, networks in the human brain O.J. Braddick*, J.M.D. O’Brien*, J. Wattam-Bell*, J. Atkinson* and R. Turner There is much evidence in primates’ visual processing for distinct mechanisms involved in object recognition and encoding object position and motion, which have been identified with ‘ventral’ and ‘dorsal’ streams, respectively, of the extra-striate visual areas [1–3]. This distinction may yield insights into normal human perception, its development and pathology. Motion coherence sensitivity has been taken as a test of global processing in the dorsal stream [4,5]. We have proposed an analogous ‘form coherence’ measure of global processing in the ventral stream [6]. In a functional magnetic resonance imaging (fMRI) experiment, we found that the cortical regions activated by form coherence did not overlap with those activated by motion coherence in the same individuals. Areas differentially activated by form coherence included regions in the middle occipital gyrus, the ventral occipital surface, the intraparietal sulcus, and the temporal lobe. Motion coherence activated areas consistent with those previously identified as V5 and V3a, the ventral occipital surface, the intraparietal sulcus, and temporal structures. Neither form nor motion coherence activated area V1 differentially. Form and motion foci in occipital, parietal, and temporal areas were nearby but showed almost no overlap. These results support the idea that form and motion coherence test distinct functional brain systems, but that these do not necessarily correspond to a gross anatomical separation of dorsal and ventral processing streams. Addresses: *Visual Development Unit, Department of Psychology, University College London, Gower Street, London WC1E 6BT, UK. Wellcome Department of Cognitive Neurology, Institute of Neurology, University College London, 12 Queen Square, London WC1N 3BG, UK. Correspondence: O.J. Braddick E-mail: [email protected] Received: 2 February 2000 Revised: 27 March 2000 Accepted: 12 May 2000 Published: 2 June 2000 Current Biology 2000, 10:731–734 0960-9822/00/$ – see front matter © 2000 Elsevier Science Ltd. All rights reserved. Results and discussion The form and motion coherence stimuli that we used have been found to give thresholds that are independent in normal observers [7] and dissociated in normal and abnormal development [6,8]. In the ‘coherent form’ con- dition, subjects viewed arrays of randomly oriented lines in which a concentric region of 100% coherence (Figure 1) appeared and disappeared. The fMRI signal was con- trasted with the incoherent condition, in which all lines were randomly oriented. Detection of this pattern coher- ence has some aspects in common with the detection of structure in Glass patterns [9] and alignments of pattern elements [10]. Concentric pattern organization is known to stimulate many neurons in macaque area V4 [11]. A number of brain areas respond to moving visual stimuli [12,13]. Among these we have identified extra- striate areas that are selectively activated by coherent com- pared to incoherent motion ([14] and O.J.B., J.M.D.O., J.W-B., J.A., and T. Hartley, unpublished observations). This contrast has been tested in the present study, to allow direct comparison with form coherence. The form coherence test showed consistent differential activation of distinct bilateral regions in the fusiform/lingual gyri (FG/LG), middle occipital gyrus (MOG), intraparietal Figure 1 Form coherence stimulus. The 100% coherent form stimulus consisted of an array of line segments, in which those within 4° of the central fixation point were aligned tangential to the circular fixation point, with the remainder oriented randomly.

Form and motion coherence activate independent, but not dorsal/ventral segregated, networks in the human brain

Embed Size (px)

Citation preview

Brief Communication 731

Form and motion coherence activate independent, but notdorsal/ventral segregated, networks in the human brain O.J. Braddick*, J.M.D. O’Brien*, J. Wattam-Bell*, J. Atkinson* and R. Turner†

There is much evidence in primates’ visual processingfor distinct mechanisms involved in object recognitionand encoding object position and motion, which havebeen identified with ‘ventral’ and ‘dorsal’ streams,respectively, of the extra-striate visual areas [1–3]. Thisdistinction may yield insights into normal humanperception, its development and pathology. Motioncoherence sensitivity has been taken as a test of globalprocessing in the dorsal stream [4,5]. We have proposedan analogous ‘form coherence’ measure of globalprocessing in the ventral stream [6]. In a functionalmagnetic resonance imaging (fMRI) experiment, wefound that the cortical regions activated by formcoherence did not overlap with those activated bymotion coherence in the same individuals. Areasdifferentially activated by form coherence includedregions in the middle occipital gyrus, the ventraloccipital surface, the intraparietal sulcus, and thetemporal lobe. Motion coherence activated areasconsistent with those previously identified as V5 andV3a, the ventral occipital surface, the intraparietalsulcus, and temporal structures. Neither form normotion coherence activated area V1 differentially. Formand motion foci in occipital, parietal, and temporal areaswere nearby but showed almost no overlap. Theseresults support the idea that form and motion coherencetest distinct functional brain systems, but that these donot necessarily correspond to a gross anatomicalseparation of dorsal and ventral processing streams.

Addresses: *Visual Development Unit, Department of Psychology,University College London, Gower Street, London WC1E 6BT, UK.†Wellcome Department of Cognitive Neurology, Institute ofNeurology, University College London, 12 Queen Square, LondonWC1N 3BG, UK.

Correspondence: O.J. BraddickE-mail: [email protected]

Received: 2 February 2000Revised: 27 March 2000Accepted: 12 May 2000

Published: 2 June 2000

Current Biology 2000, 10:731–734

0960-9822/00/$ – see front matter © 2000 Elsevier Science Ltd. All rights reserved.

Results and discussionThe form and motion coherence stimuli that we usedhave been found to give thresholds that are independentin normal observers [7] and dissociated in normal and

abnormal development [6,8]. In the ‘coherent form’ con-dition, subjects viewed arrays of randomly oriented lines inwhich a concentric region of 100% coherence (Figure 1)appeared and disappeared. The fMRI signal was con-trasted with the incoherent condition, in which all lineswere randomly oriented. Detection of this pattern coher-ence has some aspects in common with the detection ofstructure in Glass patterns [9] and alignments of patternelements [10]. Concentric pattern organization is known tostimulate many neurons in macaque area V4 [11].

A number of brain areas respond to moving visualstimuli [12,13]. Among these we have identified extra-striate areas that are selectively activated by coherent com-pared to incoherent motion ([14] and O.J.B., J.M.D.O.,J.W-B., J.A., and T. Hartley, unpublished observations).This contrast has been tested in the present study, toallow direct comparison with form coherence. The formcoherence test showed consistent differential activation ofdistinct bilateral regions in the fusiform/lingual gyri(FG/LG), middle occipital gyrus (MOG), intraparietal

Figure 1

Form coherence stimulus. The 100% coherent form stimulus consistedof an array of line segments, in which those within 4° of the centralfixation point were aligned tangential to the circular fixation point, withthe remainder oriented randomly.

sulcus (IPS), and (in two out of four subjects) the superiortemporal sulcus (STS) (see Table 1, Figure 2).

The same subjects showed five distinct regions differen-tially activated by coherent motion, as reported previously.One corresponds to the location of V5 [12,13,15] andanother may be V3A [16] (although identification is uncer-tain without retinotopic flat-mapping [17]). Other acti-vated regions were found on the posterior ventral surface(FG/LG), the IPS and STS (see Table 1 and Figure 2). Inseveral cases, form and motion foci were in the samegeneral anatomical region. However, only a small fractionof voxels activated in either test were found to be acti-vated in both (0%–8% in different subjects, depending onthe overall extent of activation for the two contrasts).

Figure 3a shows areas responding to motion coherence inthe region designated as ‘V3A’ and the adjacent response toform coherence in the middle occipital area. Area V5(motion coherence) lies anterior to the section illustrated;the middle occipital form coherence area, seen primarily in

the left hemisphere of this subject, lies between the V5 and‘V3A’ motion activation areas, overlapping with them mini-mally. Activations on the ventral occipital surface, in theLG/FG, are shown in Figure 3b. Form activation was pos-terior and lateral to the ventral-surface motion response,and medial/posterior to V5, without overlapping either; thisrelationship, illustrated schematically in Figure 2, was con-sistent across subjects even though locations of the focivaried (see Supplementary material). Form and motioncoherence both gave differential, but non-overlapping,responses in IPS (Figure 3c). All four subjects showed amotion response in STS, but the form response in thisregion was less consistent (see Supplementary material).

Our results lead to three main conclusions. First, severalextra-striate cortical areas show much stronger activationby coherent global form than by the same elements ran-domly arranged. Second, area V1 shows no such differentialactivation, suggesting that its response is determined by thelocal spatial elements. Third, areas activated by form andmotion coherence are almost completely non-overlapping.

732 Current Biology Vol 10 No 12

Table 1

Regions differentially activated by the contrasts ‘coherent motion minus noise’ and ‘coherent form minus incoherent form’.

Subject AW JO JW KY Mean

Coordinate Coordinate Coordinate Coordinate Coordinate

x y z z-score x y z z-score x y z z-score x y z z-score x y z

Form coherence

Middle occipital 29 –96 17 7.4 26 –94 20 7.4 30 –95 12 8.1 31 –90 23 4.6 29 –94 18

Ventral surface 28 –83 –20 5.9 38 –77 –22 7.2 35 –74 –14 7.0 44 –75 –15 4.6 36 –77 –18

IPS 26 –64 50 5.6* 37 –45 54 7.1 36 –54 56 6.1 39 –52 55 4.5 35 –54 54

STS/Lateral sulcus 53 –39 6 6.1 64 –19 18 7.6 58 –20 26 4.7† ‡ ‡ ‡ ‡ 58 –26 17

Motion coherence

V5 50 –73 –2 8.2 41 –72 –3 7.6 45 –58 –5 8.2 48 –64 0 8.0 46 –67 –3

V3A? 28 –82 36 7.1 18 –84 22 6.4 27 –84 26 7.3 17 –81 29 7.7 23 –83 28

Ventral surface 22 –75 –17 5.6§ 12 –86 –12 5.0* 30 –70 –16 6.3 7 –60 –2 7.2 18 –73 –12

IPS 33 –36 53 7.5¶ 32 –36 54 7.4 35 –44 55 7.4 31 –35 64 7.3 33 –38 57

STS 59 –44 14 7.8 62 –39 10 7.3 52 –46 2 7.5† 63 –36 9 7.4 59 –41 9

Coordinates are quoted as the mean Talairach coordinate for themost significant voxel in a given cluster. Left hemisphere coordinates(negative x) are transformed to right hemisphere coordinates (positivex). Clusters are groups of adjacent voxels, all of which have z-scoresabove 3.09 (p = 0.001). The z-scores, quoted in the fourth column foreach subject, are for the most significant voxel in a pair of (left andright) clusters. These are all significant at the level p < 0.05(corrected), except where indicated. §Activation in the left hemispherenot significant. ¶Activation in the right hemisphere not significant.*Focus only found in the left hemisphere. †Focus only found in theright hemisphere. ‡No focus in either hemisphere. SPM96 andSPM99 (Wellcome Department of Cognitive Neurology, London,http://www.fil.ion.ucl.ac.uk/spm) were used for fMRI data preprocessingand statistical analysis. Each image volume was realigned to the firstvolume, spatially normalized to the space of Talairach and Tournoux [23]

with subsampling to a resultant 2 × 2 × 2 voxel size. The resultant imagevolumes were spatially smoothed with a 4 mm FWHM Gaussian kernelprior to statistical analysis [23,24]. Effects were estimated according tothe linear model at every voxel, after specifying the appropriate designmatrix. Hypotheses about regionally specific condition effects weretested by comparing the estimates using linear contrasts whichgenerated an SPM{t} for each contrast, transformed to the unit normaldistribution SPM{Z} and thresholded at 3.09 (or p = 0.001uncorrected). The resulting foci were then characterised in terms ofspatial extent (k) and peak height (u). All foci listed met criteria of overallsignificance, according to the theory of Gaussian fields, in terms of theprobability that a region of the observed number of voxels (or bigger)could have occurred by chance or that the peak height observed(or higher) could have occurred by chance, over the entire volumeanalysed (that is, a corrected p-value).

These findings support the idea that form coherence testsglobal pattern processing mechanisms beyond V1, inde-pendently from mechanisms for global motion processing.The lack of overlap between responses indicates thatthey are specific to the two types of processing, and do

not simply reflect non-specific activation linked to visualattention or visual organization.

Areas activated by form and motion coherence clearly donot divide between dorsal and ventral locations in termsof gross anatomy. Human anatomical relationships arelikely to differ from the ventral/dorsal division in themacaque, since the topography of the visual areas showsmarked differences [17]. However, it is striking that formand motion responses lie close together, but distinct,within several widely distributed posterior brain regions.For example, the parietal lobe is thought to be the targetof the dorsal stream, in humans and monkeys, based onlesion evidence [5]. However, our data show parallel fociof form and motion activation in the IPS. Since eyepursuit gain falls steeply above 1 Hz [18], it is unlikelythat the distinct IPS activation by motion (which reversedthree times per second) is an artefact of induced eyemovements. The temporal lobe also showed nearby focifor form and motion coherence. Thus, the two kinds ofglobal processing are represented, but segregated, in bothparietal and temporal cortex.

The most consistent foci for both form and motioncoherence were in extra-striate occipital cortex. One ofour motion foci corresponds to V5 and the other is consis-tent with V3a, both well established as motion-sensitive.The motion focus in FG/LG may include parts of V2 andVP [18]. A separate focus of activation by form coherence

Brief Communication 733

Figure 2

Schematic illustration of regions activated by form coherence (shownin green) and by motion coherence (shown in red). Foci that aredistinct on individual subjects may overlap in projection to the cerebralsurface even though they are non-overlapping voxels, and so are notwell depicted in such rendered views.

Intraparietal sulcus

V3A

V5

Lingual/fusiform V2/VP?

Main foci of activation for:

(schematic)

Motion coherenceForm coherence

Middle occipital gyrusV3?

Current Biology

Figure 3

Current Biology

(a) (b) (c)

Differential activation by form- and motion-coherence across regionsfor individual subjects. (a) Activation in the middle occipital region.This coronal slice is taken from y = –84 for subject JW. This clearlyshows activation by form coherence (shown in green) on the leftventral surface, as well as the middle occipital region. The position ofthe motion focus (shown in red) is consistent with the reportedlocation of V3A. Both form and motion show activation over aconsiderable anterior-posterior region. (b) Ventral surface activation

by form coherence (green) and motion coherence (red). Thistransverse slice is taken from z = –12 for subject JW. The mostlateral motion foci are part of the V5 area of activation. The formcoherence focus is lateral to the motion coherence focus on theventral surface. (c) Activation in intra-parietal sulcus. This coronalslice is taken from y = –46, from subject JO. The form focus (green)is much larger than the motion focus (red), which only appears in theleft hemisphere in this slice.

was lateral to this FG/LG motion focus but was posteriorand ventral to the LO complex, where responses tostructured versus unstructured images have beenreported [19,20]. The MOG activation we observed wasalso clearly posterior to this area, but in a more dorsallocation, possibly corresponding to V3.

Activation by organised visual texture has been reported inLG/FG [21], but at locations anterior and medial to ourLG/FG focus. Our form coherence stimulus has a numberof spatial properties — contour alignment, recognizableshape, spatial frequency content, and large-scalesymmetry — and exploring these separately may help tounderstand the relation to other experiments [9,10,19,21].In summary, the well-segregated patterns of activationshow that form and motion coherence are processed by dis-tinct yet tightly integrated visual brain systems. Furtherwork is needed to relate those to other measures of visualcortical organisation.

Materials and methodsFour subjects with emmetropic (normal) vision were tested, with func-tional MRI (T2* weighted) scans for a BOLD signal [22] covering 32slices of 3 mm voxels. Stimuli were presented in a blocked design, withten 38 second blocks of 12 scans in each run. Visual stimuli wereviewed passively, with subjects fixating a central spot. Each subjectcompleted three runs.

The form coherence stimulus consisted of an array of 1024 line seg-ments, in a display of 12.8 × 12.8°. Lines were oriented tangential toconcentric circles within an 8° central region and were random else-where. In ‘coherent’ blocks, such patterns alternated at 2 per sec withpatterns of randomly oriented lines. These blocks alternated with inco-herent blocks containing a sequence of unrelated, randomly orientedarrays, at the same rate of 2 per sec.

The motion coherence sequence was similar, with blocks of reversingcoherent motion alternating with blocks of dynamic noise. The displaywas a 128 × 128 array of 0.1° high contrast random black and whitepixels, drifting vertically at 5°/sec, and reversing direction 3 times/sec.The dynamic noise pattern was composed of similar pixels in asequence of unrelated random arrays at 50 frames/sec. SPM96 andSPM99 (Wellcome Department of Cognitive Neurology, London) wereused for preprocessing and statistical analysis (Table 1).

Supplementary materialSupplementary material including including fuller details of methods,analysis and the regions of activation is available at http://current-biology.com/supmat/supmatin.htm.

AcknowledgementsThis research was supported by programme grant G79088507 from theMedical Research Council and by the Wellcome Trust through its supportfor the Wellcome Department of Cognitive Neurology and the LeopoldMuller Functional Imaging Laboratory.

References1. Young MP: Objective analysis of the topological organization of

the primate cortical visual system. Nature 1992, 358:152-155.2. Mishkin M, Ungerleider L, Macko K: Object vision and spatial vision:

two central pathways. Trends Neurosci 1983, 6:414-417.3. Milner AD, Goodale, MA: The Visual Brain in Action. Oxford: Oxford

University Press; 1995.4. Newsome WT, Paré EB: A selective impairment of motion

perception following lesions of the middle temporal visual area(MT). J Neurosci 1988, 8:2201-2211.

5. Britten KH, Shadlen MN, Newsome WT, Movshon JA: The analysis ofvisual motion: a comparison of neuronal and psychophysicalperformance. J Neurosci 1992, 12:4745-4765.

6. Atkinson J, King JA, Braddick OJ, Nokes L, Anker S, Braddick F:A specific deficit of dorsal stream function in Williams’ syndrome.Neuroreport 1997, 8:1919-1922.

7. Braddick: Motion processing: from elementary signals to globalorganisation. Perception 1998, 27:S1.

8. Atkinson J, Braddick OJ, Lin MH, Curran W, Guzzetta A, Cioni G:Form and motion coherence: is there dorsal stream vulnerabilityin development? Invest Opthal Vis Sci 1999, 40:S395.

9. Wilson HR, Wilkinson F: Detection of global structure in Glasspatterns: implications for form vision. Vision Res 1998,38:2933-2947.

10. Field DJ, Hayes A, Hess RF: Contour integration by the humanvisual-system — evidence for a local association field. Vision Res1993, 33:173-193.

11. Gallant JL, Braun J, Van Essen DC: Selectivity for polar, hyperbolicand Cartesian gratings in macaque visual cortex. Science 1993,259:100-103.

12. Watson JDG, Myers R, Frackowiak RSJ, Hajnal JV, Woods RP,Mazziotta JC, et al.: Area V5 of the human brain: evidence from acombined study using positron emission tomography andmagnetic resonance imaging. Cereb Cortex 1993, 3:79-94.

13. Dupont P, Orban GA, De Bruyn B, Verbruggen A, Mortelmans L:Many areas in the human brain respond to visual motion.J Neurophysiol 1994, 72:1420-1424.

14. Braddick OJ, Hartley T, O'Brien JMD, Atkinson J, Wattam-Bell J,Turner R: Brain areas differentially activated by coherent visualmotion and dynamic noise. Neuroimage 1998, 7:S322.

15. Zeki S, Watson JDG, Lueck CJ, Friston KJ, Kennard C, FrackowiakRSJ: A direct demonstration of functional specialization in humanvisual cortex. J Neurosci 1991, 11:641-649.

16. Tootell RBH, Mendola JD, Hadjikhani NK, Ledden PJ, Liu AK, Reppas JB:Functional analysis of V3A and related areas in human visualcortex. J Neurosci 1997, 17:7060-7078.

17. Tootell RBH, Dale AM, Sereno MI, Malach R: New images fromhuman visual cortex. Trends Neurosci 1996, 19:481-489.

18. Carpenter RHS. Movements of the Eyes. London: Pion; 1978.19. Malach R, Reppas JB, Benson RR, Kwong KK, Jiang H, Kennedy WA,

et al.: Object-related activity revealed by functional magneticresonance imaging in human occipital cortex. Proc Natl Acad SciUSA 1995, 92:8135-8139.

20. VanOostende S, Sunaert S, VanHecke P, Marchal G, Orban GA: Thekinetic occipital (KO) region in man: an fMRI study. Cereb Cortex1997, 7:690-701.

21. Beason-Held LL, Purpura KP, Van Meter JW, Azari NP, Mangot DJ,Optican LM: PET reveals occipitotemporal pathway activationduring elementary form perception in humans. Vis Neurosci 1998,15:503-510.

22. Turner R, Howseman A, Rees GE, Josephs O, Friston KJ: Functionalmagnetic resonance imaging of the human brain: data acquisitionand analysis. Exp Brain Res 1998, 123:5-12.

23. Talairach J, Tournoux P: Co-planar Stereotactic Atlas of the HumanBrain. New York: Thieme Medical; 1988.

24. Friston KJ, Ashburner J, Poline JB, Frith CD, Heather JD, FrackowiakRSJ. Spatial realignment and normalization of images. Hum BrainMapp 1995, 3:165-189.

734 Current Biology Vol 10 No 12