Proc. Natl. Acad. Sci. USAVol. 92, pp. 5969-5972, June 1995Neurobiology

Positron-emission tomography studies of cross-modality inhibitionin selective attentional tasks: Closing the "mind's eye"

(human brain activity/regional cerebral blood flow/attention/deactivation)

RYUTA KAWASHIMA, BRENDAN T. O'SULLIVAN, AND PER E. ROLANDDivision of Human Brain Research, Department of Neuroscience, Karolinska Institute, S-171 77, Stockholm, Sweden

Communicated by Seymour S. Kety, National Institutes of Health, Bethesda, MD, March 14, 1995 (received for review April 20, 1994)

ABSTRACT It is a familiar experience that we tend toclose our eyes or divert our gaze when concentrating attentionon cognitively demanding tasks. We report on the brainactivity correlates of directing attention away from potentiallycompeting visual processing and toward processing in anothersensory modality. Results are reported from a series ofpositron-emission tomography studies of the human brainengaged in somatosensory tasks, in both "eyes open" and"eyes closed" conditions. During these tasks, there was asignificant decrease in the regional cerebral blood flow in thevisual cortex, which occurred irrespective ofwhether subjectshad to close their eyes or were instructed to keep their eyesopen. These task-related deactivations of the association areasbelonging to the nonrelevant sensory modality were inter-preted as being due to decreased metabolic activity. Previousresearch has clearly demonstrated selective activation of cor-tical regions involved in attention-demanding modality-specific tasks; however, the other side of this story appears tobe one of selective deactivation of unattended areas.

Regional cerebral blood flow (rCBF) changes in the humanbrain have now been extensively studied by positron-emissiontomography (PET) methods (1). Regional changes in rCBF aremonotonically related to regional changes in cerebral meta-bolic rate, in particular the metabolic demands of maintainingthe transmembrane ionic gradients of active neurons (1-4).Brain structures that actively participate in the performance ofspecific cognitive tasks can be identified as discrete regions ofincrease in rCBF. During specific cognitive tasks, our attentionis selectively focused on the relevant sensory modality orsubmodality from which information is necessary to success-fully perform the task. In behavioral terms, stimuli within thefocus of attention are generally discriminated more quickly andaccurately, are registered more vividly in awareness and mem-ory, and exert greater control over behavior than do unat-tended stimuli (5, 6). In physiological terms, selective attentionis described as changes in the excitability of cortical neuronsthat are limited to, or focused on, a specific sensory modalityor submodality (1, 7). Increases in rCBF associated with thesechanges in synaptic metabolic activity can then be identified inmodality-specific tasks (8, 9). Behavioral models of selectiveattention generally accept the notion that attention has alimited capacity, which must be flexibly distributed amongcompeting processes (5). The question arises, therefore, whatis happening neurobiologically to cortical areas of anothersensory modality which is not being attended during theperformance of a task? In particular, is there evidence ofinhibition or deactivation in these regions which are not beingattended during the performance of a cognitive task?To answer this question, we performed two PET studies in

nine normal volunteers while they were engaged in somato-sensory tasks [roughness discrimination (10) and tactile shape

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

matching (11)]. The two tasks were matched in terms of theirperformance difficulty and regional changes in CBF wereidentified by standard image subtraction techniques in whichthe task state is compared to a similar control state. One of thestudies (tactile shape matching) was performed with thesubjects' eyes open in both the task and control states so thatthe "eyes open" condition was matched before subtracting theimages. The other somatosensory study (roughness discrimi-nation) was performed with the eyes closed in both the taskand control states, so that this condition also was matchedbefore image subtraction. The two somatosensory tasks werechosen to specifically examine whether hypothesized decreasesin rCBF in nonattended areas such as the visual cortical areasduring somatosensory tasks would still occur irrespective of thegeneral level of visual input (and blood flow values).

MATERIALS AND METHODSSeveral points of methodology were critical to the perfor-mance of these studies. First, it was necessary to obtainquantitative data on absolute rCBF changes (in ml per 100 gof brain tissue per min) since normalization procedures rou-tinely used in many other PET studies could artefactually leadto increases and reciprocal decreases in rCBF values. Absolutemeasures of rCBF required brachial arterial cannulation andcontinuous arterial sampling to define input curves after eachbolus injection of the radioisotope. High-resolution functionalimages (full width at half maximum, 4.5 mm) were obtained byusing an 8-ring (15 slice) PET camera (PC2048-15B), which hasan interslice interval of 6.7 mm (12). The freely diffusible trac-er [15O]butanol (13) was used in preference to H2150. Ana-tomical standardization and accurate functional localizationwas obtained by coregistration with magnetic resonance im-aging (MRI)-defined anatomy by using a computerized brainatlas program (14, 15), which corrects interindividual differ-ences in brain shape and size by both linear and nonlinearparameters in order to reformat all individual images intostandard atlas anatomy.The roughness discrimination task reported here was per-

formed with the eyes closed, while the tactile matching taskwas performed with the eyes open. Both tasks were performedwith the right hand. In roughness discrimination, nine subjectsperformed a series of two alternative forced-choice discrimi-nations of quantified roughness stimuli of known wavelength,amplitude, and stimulus energy (16) and indicated with athumbs-up response only if the stimulus presented second wasperceived as "rougher" than the first. The control state was restwith eyes closed. The tactile matching task required othersubjects to match a spherical ellipsoid presented to their righthand, with one member of a linear array of similar ellipsoidshidden from view behind a white curtain in front of them. Theright hand was used for all palpations, and they indicated theirchoice of a matching stimulus by pointing with their right index

Abbreviations: rCBF, regional cerebral blood flow; PET, positron-emission tomography; MRI, magnetic resonance imaging.

5969

5970 Neurobiology: Kawashima et al.

finger. The subjects' eyes were open throughout this study andfixated on a cross drawn on the curtain in front of them. In thecontrol state the subjects lay quietly without moving or speak-ing and fixated the cross with no specific task to perform.Throughout the PET measurements, the electroencephalo-gram and the electrooculogram were monitored continuously.The subjects' behavior was also monitored continuously withvideo cameras. The tests started at the start of the injection ofisotope and continued for 180 sec.

Subjects had their PET studies as well as MRI tomographsperformed with the same bite-fixation stereotaxic helmets inorder to correlate functional and anatomical data. Anatomicalstructures of each subject's MRI were fitted interactively usingthe computerized brain atlas (14, 15). All PET and MRIimages were then transformed into the standard atlas brainanatomy using these parameters. The precision of the refor-mation process has a SD of 2-3 mm in the localization of theinner and outer brain surfaces (15). Subjects received 70 mCi("1 mCi/kg; 1 Ci = 37 GBq) of [150]butanol as a bolusintravenous injection at the commencement of each PET run,which lasted for 80 sec. Images were reconstructed with aHanning filter of 4 mm and were displayed with a pixel size of2.01 x 2.01 mm. There was no further filtering of images. Thearterial radiotracer concentration was measured continuouslyand the arterial partial pressures of 02 and CO2 (Po2 andPco2) were measured repetitively. The rCBF was calculated bythe dynamic approach on the data sampled between 0 and 80sec after the start of the injection in frames of 5 sec each (17,18). The rCBF images were corrected according to the arterialPco2 measured during each PET procedure. The correctionwas done before test minus control subtractions to the arterialPco2 level of the control (18). Increases in rCBF were calcu-lated by voxel-by-voxel subtractions of the control state imagesfrom the corresponding images of each task state. Decreasesin rCBF were calculated by voxel-by-voxel subtractions of thetask state images from the corresponding images of eachcontrol state. All images of the brain were anatomicallystandardized as described above, and mean change images,variance images, and descriptive Student's t images werecalculated voxel by voxel. The procedures and statistical anal-ysis were described extensively in a recent report (18). Thecriterion used for accepting rCBF changes in adjacent clus-tered voxels as activations gives an average probability of 0.9of finding one false-positive cluster (and <0.08 of findingmore) in the three-dimensional space representing the brain.The descriptive t image was limited by accepting only voxelshaving t > 1.83 and occurring in clusters of size 12 and above;all other voxels were set to 0. This image is called a clusterimage. In this image only clusters of size 12 and above havingt > 1.83 are shown and considered regions of changed rCBF.In Table 1, the cluster sizes of the activated fields are shownin cm3. One voxel had a volume of 44.03 mm3. Clusters of 12or more contiguous voxels in the t images were consideredareas of significant physiological change in rCBF on the basisof previous analysis of noise images (18). Areas of significantrCBF change were identified from the final cluster images andanatomical localization made by calling structures from theCBA program. The values of rCBF change (ml per min per gof tissue) and the volume of brain in which this occurred (cm3)were calculated by voxel-by-voxel analysis of regions in meansubtraction CBF images, which had been identified fromcorresponding regions of significant change in cluster images.Further details of the image analysis procedure have beendescribed (18).

RESULTSFig. 1 illustrates that activation (shown in red) of the con-tralateral sensorimotor strip is associated with diffuse de-creases (shown in blue) in visual cortical and other cortical

FIG. 1. Increases (red) and decreases (blue) in the rCBF are shownfor roughness discrimination (Left) and for tactile matching (Right).Corresponding sections are shown to illustrate somatosensory activa-tion associated with decreases in activity in visual areas. The atlaspositions of the central, postcentral, and parietooccipital sulci areshown. Right in the image is left in the brain. Roughness discriminationwas performed with the eyes closed while tactile matching wasperformed with the eyes open. Both tasks were performed with theright hand. The two somatosensory tasks were of comparable difficultywith a percentage correct rate of 74% and 82%, respectively. Therewas no significant difference in global blood flow in test and controlconditions in either experiment.

areas in both somatosensory tasks (roughness discrimination,Left; tactile shape matching, Right). For the purposes ofdiscussion, we regarded visual areas as the whole of theoccipital lobe. The decreases in rCBF were most marked innonprimary visual regions, in particular in areas of the occip-ital lobe near the border of the occipital lobe and the parietallobe. In addition, there were decreases in rCBF in the precu-neus, the posterior part of the superior parietal lobule, and theposterior part of the angular gyrus, which are putative remotevisual areas (1). Some areas in the frontal lobe also showeddecreases in activity. The function of much of the frontalcortex remains uncertain, but it is possible in this case thatthese decreases in activity may have been related to the highlysignificant decreases in activity identified in visual corticalregions. There were no visual regions at all that showed anyincrease in activity while performing the somatosensory tasks.Similar decreases occurred both when the subjects' eyes wereclosed (i.e., roughness discrimination) and when the subjects'eyes were open during the task performance (i.e., tactile shapematching).The subjects fixated the cross hair in the open eyes condi-

tions (tactile shape matching and control condition, fixation)and had their eyes closed in the closed eyes condition (rough-ness and its control). The number of eye movements presentfor the measurement period for control and shape matching,respectively, and the number of eye blinks did not differ(frequency of blinks and small eye movements 0.79 ± 3.3 and0.87 ± 3.7 Hz; mean ± SD). There were no systematic changesin arterial Pco2 between the test conditions and their respec-tive control conditions (P> 0.1, P > 0.5; paired comparison,t test for tactile matching versus control and roughness versuscontrol) and the global blood flow did not change (P> 0.9, P

Proc. Natl. Acad. Sci. USA 92 (1995)

Proc. Natl. Acad. Sci. USA 92 (1995) 5971

> 0.5; paired comparison t test) for tactile matching versuscontrol 54.5 ± 2.7 and 57.1 ± 3.0 and roughness versus control60.8 ± 4.0 and 60.9 ± 3.3.

Fig. 2 shows the mean values of decreased rCBF in primaryversus nonprimary visual areas in the occipital cortex of eachhemisphere, as well as the value of increased rCBF in the leftprimary and nonprimary somatosensory cortex. Table 1 pro-vides details of the specific anatomical location of significantrCBF decreases including Talairach coordinates of the centerof each field, field volume, and the actual value of the rCBFchange. No specific fields of significant change were identifiedin the primary visual cortex in either study. This is in contrastwith the number ofbroad cortical fields that showed significantdecreases in activity that were distributed across the nonpri-mary visual regions.

DISCUSSION AND CONCLUSIONSThe results of this study prompt a further examination of themeaning of rCBF changes in PET studies. The underlyingmechanisms for the tight coupling of rCBF changes withchanging synaptic metabolic demands in active neuronal pop-ulations are still poorly understood (1, 20, 21). The highestenergy requirements in the brain are thought to be for activetransport of Na+ through ATP-dependent ion channels torestore membrane potentials after depolarization (3, 7, 22, 23).Depolarization may take the form of action potentials orexcitatory postsynaptic potentials (EPSPs). There is alreadygood evidence that EPSPs, without action potential formation,make a sizeable contribution to this metabolic load of the brainas reflected in rCBF studies (1, 7, 22, 23). It has been shownthat inhibition of the Na+ pump leads to a marked decrease inrCBF (23). The best explanation of decreased rCBF wouldappear to be that it reflects a net decrease in regional synapticmetabolic demand (20, 23-25). Increase of GABAAergic localinhibition in the neocortex decreases rCBF and most likely alsothe local metabolism (24, 26). We call the decrease in rCBFseen in the visual areas deactivation. This deactivation, for thereasons described above, most likely was due to a decreasedexcitation of these areas or a local net increase in synapticinhibition. The present results do not give any clues concerninghow the deactivation of visual areas is controlled by other

Eyes closed

..6 *..

--w

uz co > ;> > ;>vL Z vL A. z z

8.64.2.0

.

-2.-4.

Eyes open

4 ..

vn Z > > R >

FIG. 2. Histograms of increases and decreases in mean rCBF in mlper 100 g per min by region: primary and nonprimaryvisual cortex (PVand NV) and primary and nonprimary somatosensory cortex (PS andNS). Lt, left; Rt, right. Primary visual cortex was defined as the areaof cortex lining the calcarine fissure visualized directly from MRI;nonprimary visual cortex was defined as the remainder of cortex in theoccipital lobe. Primary somatosensory cortex was defined as theanterior half of the postcentral gyrus as visualized on MRI; nonpri-mary somatosensory cortex was defined as the posterior half of thepostcentral gyrus. Changes in mean rCBF are shown in ml per min per100 g of brain tissue ISE. Level of significance is indicated (*,significant at P = 0.05). Note that these regions are larger than thoselisted in Table 1.

structures in the brainstem or thalamus, since we were unableto find any concomitant change in rCBF in these structures.

In studying both eyes open and eyes closed tasks withmatched control states, we found that decreases were inde-pendent of the level of rCBF in the visual cortical regions inthe control states-that is, independent of whether visualinformation reached the visual cortices or not during thesomatosensory tests. The functional significance of this ob-served decrease in cortical activity in regions not directlyinvolved in performing a specific task may be understood as ameans of reducing the probability of interference of informa-tion from other sensory modalities. Deactivation of corticalregions not associated with the task at hand may therefore bean essential component of selective attentional processesplaying a complementary role to activation of those areas of

Table 1. Anatomic location of significant rCBF decreasesDecrease,

Talairach coordinate rCBF, ml ml perVol, per 100 g 100 g per

Region X Y Z cm3 per mm mi t valueTactile shape matching (eyes open)

Superior occipital gyrus (left) 31 -68 32 1.63 61.6 (3.1) 14.1 (0.8) 2.62 (0.12)Middle occipital gyrus (left) 33 -74 19 1.01 67.9 (2.5) 14.0 (0.7) 3.03 (0.18)Middle occipital gyrus (left) 46 -77 4 0.84 47.5 (2.5) 12.9 (1.0) 2.25 (0.06)Inferior occipital gyrus (left) 27 -79 -11 0.97 54.6 (3.6) 11.1 (0.5) 2.59 (0.14)Inferior occipital gyrus (left) 41 -83 -8 0.62 51.6 (2.3) 13.8 (0.9) 2.42 (0.13)Middle occipital gyrus (right) -27 -78 11 0.84 53.8 (2.3) 12.3 (0.6) 2.19 (0.08)Inferior occipital gyrus (right) -26 -89 -10 1.58 57.1 (2.2) 11.6 (0.7) 2.70 (0.21)

Roughness discrimination (eyes closed)Middle occipital gyrus (left) 35 -61 22 1.58 62.2 (2.2) 7.1 (0.3) 1.86 (0.05)Middle occipital gyrus (left) 48 -63 7 1.12 43.0 (1.8) 7.3 (0.5) 1.94 (0.08)Middle occipital gyrus (right) -43 -52 9 0.88 61.0 (2.4) 9.3 (0.8) 2.07 (0.11)Inferior occipital gyrus (right) -40 -65 -2 1.76 47.6 (1.8) 7.1 (0.4) 2.19 (0.09)

Talairach coordinates (19) of these identified structures were calculated as follows. First, the center of gravity of each cluster was identified.Second, the stereotaxic coordinates ofthe center ofgravity ofeach cluster were measured relative to the midpoint between the anterior and posteriorcommissures. Third, these coordinates were transformed into Talairach coordinates by using linear parameters in order to correct differences insize and baseline angle between the standard brain of the computerized brain atlas system and the original Talairach standard brain. The Talairachcoordinates (X,Y,Z) of each cluster were calculated in millimeters measured from the midpoint between the anterior and posterior commissures.Coordinates are given in the order X (width), Y (anterior-posterior), and Z (height). rCBF decrease values are actual mean decreases in ml per100 g per min (SE). t values are calculated by Student's t test.

Neurobiology: Kawashima et al.

I -cn_lwm %J.6.-

?.N 4 n

0m 2 'C06 0.

E -20

C#') -4 -

-6.

5972 Neurobiology: Kawashima et al.

cortex that are required to be active during a given task. Thisfinding may be related to the observations of unaltered globalblood flow and metabolism in the brains of subjects doing teststhat are associated with increases of the rCBF and regionalcerebral metabolism (25, 27). Earlier Seitz and Roland (25)reported decreases of rCBF in parietal, superior temporal, andprefrontal regions in a somatosensory task, and Haxby et al(28) very recently reported significant decreases of normalizedrCBF (most likely concurrent with true rCBF decreases) inauditory and prefrontal regions during a visual task. Thesestudies and the present study thus strongly suggest that asso-ciated cross-modal decreases in cortical activity in areas notdirectly required in performing a given task may be a frequentconsequence of the regulation of attention. Decreases in visualareas occurred in a somatosensory task in which the eyes wereopen but also in a similar task in which the eyes were closed.Put simply, in performing cognitive demanding somatosensorytasks, even if you do not tend to close your eyes, the "mind'seye" will close.

1. Roland, P. E. (1993) Brain Activation (Wiley, New York), pp.195-236.

2. Greenberg, F., Hand, P., Sylvestro, A. & Reivich, M. (1979)ActaNeurol. Scand. Suppl. 72, 12-14.

3. Mata, M., Fink, D. & Gainer, H. (1980) J. Neurochem. 34,213-215.

4. Juliano, S., Hand, P. & Whitsel, B. (1981) J. Neurophysiol. 46,1260-1282.

5. Kahneman, D. & Treisman, A. (1984) in Varieties ofAttention,eds. Parasuraman, R. & Davies, D. (Academic, New York), pp.929-961.

6. Posner, M. & Petersen, S. (1990)Annu. Rev. Neurosci. 13,25-42.7. Roland, P. E. (1981) J. Neurophysiol. 46, 744-754.8. Roland, P. E. (1982) J. Neurophysiol. 48, 1059-1078.9. Corbetta, M., Miezen, F. & Dobmeyer, S. (1991) J. Neurosci. 11,

2383-2402.10. O'Sullivan, B., Roland, P. E. & Kawashima, R. (1994) Eur. J.

Neurosci. 6, 137-148.

11. Kawashima, R., Roland, P. E. & O'Sullivan, B. (1992) Soc.Neurosci. Abstr. 8, 1420.

12. Litton, J., Holte, S. & Eriksson, L. (1990) IEEE Trans. Nucl. Sci.37, 743-748.

13. Berridge, M., Adler, L., Nelson, D., Cassidy, E., Muzic, R.,Bednarczyk, E. & Miraldi, F. (1991) J. Cereb. Blood Flow Metab.11, 707-715.

14. Bohm, C., Greitz, T., Blomqvist, G., Farde, L., Forsgren, P.,Kingsley, D., Sjogren, I., Wiesel, F. & Wik, G. (1986)Acta Radiol.369, 449-452.

15. Seitz, R., Bohm, C., Greitz, T., Roland, P., Eriksson, L.,Blomqvist, G., Rosenqvist, G. & Nordell, B. (1990) J. Cereb.Blood Flow Metab. 10, 443-457.

16. Roland, P. E. & Mortensen, E. (1987) Brain Res. Rev. 12, 1-42.17. Koeppe, R., Holden, J. & Ip, W. (1985) J. Cereb. Blood Flow

Metab. 5, 224-234.18. Roland, P. E., Levin, B., Kawashima, R. & Akerman, S. (1993)

Hum. Brain Mapp. 1, 3-19.19. Talairach, J., Szikla, G., Tournoux, P., Prossalentis, A., Bordas-

Ferrer, M., Covello, L., Jacob, M. & Mempel, E. (1967) Atlasd'Anatomie Stereo-taxique du Telencephale (Masson, Paris).

20. Sokoloff, L. & Kety, S. S. (1960) Physiol. Rev. 40, 38-44.21. Lassen, N. A. (1991) in Brain Work and Mental Activity, eds.

Lassen, N. A., Ingvar, D. H., Raidke, M. L. & Friberg, L.(Munksgaard, Copenhagen), pp. 68-77.

22. Creutzfeldt, 0. (1975) in Brain Work, eds. Lassen, N. A. & Ingvar,D. H. (Munksgaard, Copenhagen), pp. 21-46.

23. Astrup, J., Blennow, G. & Nilsson, B. (1977) Brain Res. 177,115-126.

24. Roland, P. E. & Friberg, L. (1988) J. Cereb. Blood Flow Metab. 8,314-323.

25. Seitz, R. J. & Roland, P. E. (1992)Acta Neurol. Scand. 86,60-67.26. Gjedde, A. (1993) in Functional Organization oftheHuman Visual

Cortex, eds. Gulyas, B., Ottoson, D. & Roland, P. E. (Pergamon,Oxford), pp. 291-306.

27. Sokoloff, L., Mangold, R., Wechsler, R. L., Kennedy, C. & Kety,S. S. (1955) J. Clin. Invest. 34, 1101-1108.

28. Haxby, J. V., Horwitz, B., Ungerleider, L. G., Masiog, J. M.,Pietrini, P., Grady, C. L. (1994) J. Neurosci. 14, 6336-6353.

Proc. Natl. Acad. Sci. USA 92 (1995)