."

/'i'

Separate visual pathways tor perception and action

Accumulating neuropsychological, electrophysiologicaland behavioural evidence suggests thai the neuralsubstrates of visual perception may be quite distincttram those underlying the visual control of actions. Inother words, the set of object descriptions thai permitidentification and recognition may be computed inde-pendently of the set of descriptions thai allow anobserver to shape the hand appropriately to pick up anobject. We propose thai the ventral stream of projectionstram the striate cortex to the inferotemporal cortex Playsthe major role in the perceptual identification of objects,while the dorsal stream projecting tram the striate cortexto the posterior parietal region mediates the requiredsensorimotor transformations tor visually guidedactions directed at such objects.

Me/vyn A. Gooda/e isat the Oept of

Psych%gy,Universily of Western

Ontatio, London,Ontatio N6A 50,

Canada, and A. OavidMi/ner is at the Oept

of Psych%gy,Universily of St

Andrews, St AndrewsKY169JU, UK.

In an influential article that appeared in Science in1969, Schneider! postulated an anatomical separationbetween the visual coding of the location of a stimulusand the identification of that stimulus. He attributedthe coding of the location to the ancient retinotectalpathway, and the identification of the stimulus to thenewer geniculostriate system; this distinction rep-resented a significant departure trom earlier mono-lithic descriptions of visual function. However, thenation of 'localization' failed to distinguish between themany different patterns of behaviour that vary withthe spatial location of visual stimuli, oniy same ofwhich turn out to rely on tectal mechanisms2-4.Nevertheless, even though Schneider's original pro-posal is no longer generally accepted, bis distinctionbetween object identification and spatial localization,between 'what' and 'where', hag persisted in visualneurosclence.

Two cortical visual systemsIn 1982, für exarnple, Ungerleider and Mishkin5

concluded that 'appreciation of an object's qualitiesand of its spatiallocation depends on the processing ofdifferent kinds of visual information in the inferiortemporal and posterior parietal cortex, respectively.'They marshalled evidence from a number of elec-trophysiological, anatomical and behavioural studiessuggesting that these two areas receive independentsets of projections from the striate cortex. Theydistinguished between a 'ventral strearn' of projec-tions that eventually reaches the inferotemporalcortex, and a 'dorsal strearn' that terminates finally inthe posterior parietal region. The proposed functionsof these two strearns were inferred largely frombehavioural evidence derived from legion studies.They noted that monkeys with legions of the infero-temporal cortex were profoundly impaired in visualpattern discrimination and recognition6, hut lessimpaired in solving 'landmark' tasks, in which thelocation of a visual cue determines which of twoalternative locations is rewarded. Quite the oppositepattern of results was observed in monkeys withposterior parietallesions 7-9.

So, according to Ungerleider and Mishkin's 1982version of the model of two visual systems, theinferotemporal lesions disrupted circuitry specialized

20

Melvyn A. Goodale and A. David Milner

tor identifying objects, whiie the posterior parietallegions interfered with neural mechanisms underlyingspatial perception. Thus, within the visual domain,they made much the same functional distinctionbetween identification and localization' as Schneider,hut mapped it onto the diverging ventral and dorsalstreams of output from the striate cortex. Since 1982,there has been an explosion of information about theanatomy and electrophysiology of cortical visualareas1O, 11 and, indeed, the connectional anatomyamong these various areas largely confirms theexistence of the two broad 'streams' of projectionsproposed by Ungerleider and Mishkin (see Fig.1)12,13.

It hag recentiy been suggested14 that these twostreams can be traced back to the two main cyto-logical subdivisions of retinal ganglion cells: Olle ofthese two subdivisions terminates selectively in theparvocellular layer, while the other terminates in themagnocellular layer of the lateral geniculate nucleus(LGN)14-16. Certainly, these 'parvo' and 'magno'subdivisions remain relatively segregated at the levelof the primary visual cortex (VI) and in the adjacentvisual area V2. They also appear to predominate,respectively, the innervation of area V 4 and themiddle temporal area (MT), which in turn provide themajor visual inputs to the inferotemporal and posteriorparietal cortex, respectively. However, it is becomingincreasingly clear that the separation between magnoand parvo information in the cortex is not as distinct asinitially thought. For example, there is recent evidencetor a parvo input into a subset of MT neurones17 asweil as tor a large contribution from the magnopathway to V4 neurones18 and to the 'blobs' in VI(Ref. 19). In short, it now appears that the dorsal andthe ventral streams each receive inputs from both themagno and the parvo pathways.

Two visuomotor systems: 'what' versus 'how'Dur alternative perspective on modularity in the

cortical visual system is to place less emphasis oninput distinctions (e.g. object location versus objectqualities) and to take more account of output require-ments20,21. It seems plausible from a functionalstandpoint that separate processing modules wouldhave evolved to mediate the different uses to whichvision can be put. This principle is already generallyaccepted in relation to 'automatic' types of behavioursuch as saccadic eye movements22, and it is possiblethat it could be extended to other systems tor a rangeof behavioural skills such as visually guided reachingand grasping, in which close coordination is requ~dbetween movements of the fingers, bands, upperlimbs, head and eyes.

It is also our contention that the inputs andtransformations required by these skilled visuomotoracts differ in important respects from those leading towhat is generally understood as 'visual perception.'Indeed, as has been argued elsewhere, the functionalmodules supporting perceptual experience of theworld may have evolved much more recently thanthose controlling actions within it21. In this article, it is

TINS, Vol. 15, No. 1, 19920166 - 2236/92/$0500

proposed that this distinction ('what' versus 'how') -rather than the distinction between object vision andspatial vision ('what' versus 'where') - captures moreappropriately the Cunctional dichotomy between theventral and dorsal projections.

Dissociation between prehension andapprehension

Neuropsychological studies of patients with damageto Olle projection system hut not the other have alsobeen cited in support of the model proposed byUngerleider and Mishkin5.23. Patients with visualagnosia following brain damage that includes, fürexample, the occipitotemporal region, are often un-ahle to recognize or describe common objects, faces,pictures, or abstract designs, even though they cannavigate throUgh the everyday world - at least at alocal level - with considerable skil124. Conversely,

patients suffering from optic ataxia following damageto the posterior parietal region are unable to reachaccurately towards visual targets that they have nodifficulty recognizini5. Such observations certainlyappear to provide support in humans für an occipito-temporal system mediating object vision hut notspatial vision, and a parietal system mediating spatialvision hut not object vision.

Closer examination of the behaviour of suchpatients, however, leads to a different conclusion.Patients with optic ataxia not only have difficultyreaching in the right direction, hut also in positioningtheir fingers or adjusting the orientation of their bandwhen reaching toward an object that can be orientedat different angles25. Such patients may also havetrouble adjusting their grasp to reflect the size of theobject they are asked to pick up.

Visually guided grasping was recently studied in apatient who bad recovered from Balint's syndrome, inwhich bilateral parietal damage causes profound dis-orders of spatial attention, gaze and visually guidedreachini6. While this patient bad no difficulty inrecognizing line drawings of common objects, herability to pick up such objects remained quite im-paired. For example, when she reached out für a smallwooden block that varied in size from trial to mal,there was little relationship between the magnitude ofthe aperture between her index finger and thumb andthe size of the block as the movement unfolded. Notonly did she fail to show normal scaling of the graspingmovement; she also made a large number of adjust-ments in her grasp as she closed in on the object -

adjustments rarely observed in normal subjects. Suchstudies suggest that damage to the parietal lobe canimpair the ability of patients to use information aboutthe size, shape and orientation of an object to controlthe hand and fingers during a grasping movement,even though this same information can still be used toidentify and describe the objects. Clearly, a 'disorderof spatial vision' fails to capture this range ofvisuomotor impairments.

There are, of course, other kinds of visuospatialdisorders, many of which are associated with parietallobe damage, while others are associated with tem-porallobe lesions27.28. Unfortunately, we lack detailedanalyses of the possible specificity of most suchdisorders: in many, the deficit may be resmcted toparticular behavioural tasks. For example, a recentlydescribed patient with a parietal injury performed

TINS, Vol. 15, No. 1, 1992

PG Cortex

Rostral STS

MSTc

~

Fig. 1. The 1982 version of Ungerleider and Mishkin's5 model of two visualsystems is illustrated in the small diagram of the monkey brain inset into thelarger box diagram. In the original model, V1 is shown sending a dorsal streamof projections to the posterior parietal cortex (PG), and a ventral stream ofprojections to the inferotemporal cortex (TE). The box diagram illustrates oneof the most recent versions of the interconnectivity of the visual cortical areas,showing that they can still be broadly segregated into dorsal and ventralstreams. However, there is crosstalk between the different areas in the twostreams, and there may be a third branch of processing projecting into therostral superior temporal sulcus (STS) that is intimately connected with boththe dorsal and ventral streams. (This is illustrated in both the brain and boxdiagrams.) Thus, the proposed segregation of input that characterized thedorsal and ventral streams in the original model is not nearly as clear cut asonce was thought. (Moditied, with permission, tram Ret. 11.)

poorly on a task in which visual guidance was neededto leam the correct route throUgh a small ten-choicernaze by moving a hand-held stylus23. However, hewas quite unimpaired on a locomotor rnaze task inwhich he was required to move his whole bodythrough space when working from a two-dirnensionalvisual plan. Moreover, he bad no difficulty in recallinga complex geometrical pattern, or in carrying out atask involving short-term spatial memory29. Suchdissociations between performance on different'spatial' tasks show that after parietal damagespatial information may still be processed quite wenfür some purposes, hut not für others. Gf course, thefact that visuospatial deficits can be fractionated inhumans does not exclude combinations of suchimpairments occurring after large legions, nor would itexclude possible selective input disorders occurringafter smaller deafferentation lesions close to wherethe dorsal stream begins.

21

Adimensions of the objects she was

11 . about to pick up, even though shePatient (DF) appeared to be unable to 'perceive'

10 those dimensions.A similar dissociation was seen

9 in her responses to the orientationof stimuli. Thus, when presented

8 with a large slot that could be5.0 4.5 4.0 3.5 3.0 2.5 placed in Olle of a number of

Width of object (cm) different orientations, she showed

great difficulty in indicating theorientation either verbally or

10 manually (i. e. by rotating her handPatient (DF) or a hand-held card). Neverthe-

9 less, she was as good as normalsubjects at reaching out and placing

8 her hand or the card into the slot,turning her hand appropriatelyfrom the very onset of the move-

7 5.0 4.5 4.03.5 3.02.5 ment30.31.

Width of object (cm) These disparate neuropsycho-logical observations lead us to

Fig. 2. In both (A) the manual matching task and (8) the grasping task, five white plaques (each propose that the visual projectionwith an overall area of 25 cm2 on the top surface, hut with dimensions ranging tram 5 x 5 cm to2.5 x 10 cm) were presented, one at a time, at a viewing distance of approximately 45 cm. Oiodesemitting in fra red light (IREOs) were attached to the tips of the index finger and thumb of the fighthand and were tracked with two infrared-sensitive cameras and stored on a WA TSMART computer(Northern Oigitallnc., Waterloo, Canada). The three-dimensional position of the IREOs and thechanging distance between them were later reconstructed off fine. (A) In the manual matchingtask, OF and two control subjects were instructed to indicate the width of each plaque over aseriesof randomly ordered trials by separating the index finger and thumb of their fight hand. In OF,unlike the controls (CG and 0), the aperture between the finger and thumb was not systematicallyrelated to the width of the target. OF also showed considerable trial to trial variability. (8) Incontrast, when they were instructed to reach out and pick up each plaque, OF's performance wasindistinguishable from that of the control subjects. The maximum aperture between the index .finger and thumb, which was achieved weil be fore contact, was systematically related to the width Dorsal and ventral systems Inof the plaques in both OF and the two control subjects. In interpreting all these graphs, it is the the monkeyslope of the function that is important rather than the absolute values plotted, since the placement How weil do electrophysiologicalof the IREOs and the size of the hand and fingers varied somewhat from subject to subject. 8ars studies of the two projection sys-represent means .:t SE. (Modified, with permission, from Ref. 31.) tems in the visual cortex of the

monkey support the distinction weare making? While any correlations

between human neuropsychology and monkey neuro-physiology should only be made with caution, it islikely that humans share many features of visualprocessing with Dur primate relatives - particularlywith the Old World monkeys in which most of theelectrophysiology hag been carried out. Furthermore,legion studies of the two projection systems in themonkey should show paralleis with the results of workdüne on human patients.

It was noted earlier that although there aredifferences in the major retinal origins of inputs to thedorsal and ventral systems in the monkey brain, thereis subsequently a good deal of pooling of information.Moreover, there are convergent simiIarities in what isextracted within the two systems. For example, bothorientation and disparity selectivity are present inneurones in both the magno and parvo systems withincortical areas VI and V2 (Ref. 15).

Nevertheless, there are special features in theproperties of individual neurones in the posteriorparietal cortex (and in its major input areas V3A andMT) that are not found in the ventral system. Themost striking feature of neurones in the posteriorparietal region is not their spatial selectivity (indeed,like those of inferotemporal cells, their receptivefjelds are typically large), hut rather the fact that their

E~Q)uc:(\IÜ;'50w

~

$.E

12Control (CG)

11

10

95.04.54.03.53.02.5

Width of object (cm)

8

7

6

5

12 10Control (CG) Control (CJ)

11 9

10 8

9 75.0 4.5 4.0 3.5 3.0 2.5 Width of object (cm) Width of object (cm)

Complications also arise on the opposite side of theequation (i.e. in relation to the ventral stream), whenthe behaviour of patients with visual agnosia is studiedin detail. The visUal behaviour of one patient (DF) whodeveloped a profound visual-form agnosia followingcarbon monoxide poisoning was recently studied.Although MRI revealed diffuse brain damage con-sistent with anoxia, most of the damage in the corticalvisual areas was evident in areas 18 and 19, with area17 apparently remaining largely intact. Despite herprofound inability to recognize the size, shape andorientation of visual objects, DF showed strikinglyaccurate guidance of band and fin~er movementsdirected at the very same objects30. 1. Thus, whenshe was presented with a pair of rectangular blocks ofthe same or different dimensions, she was unable todistinguish between them. When she was asked toindicate the width of a single block by means of herindex finger and thumb, her matches bore no relation-ship to the dimensions of the object and showedconsiderable trial to trial variability (Fig. 2A). How-ever, when she was asked simply to reach out andpick up the block, the aperture between her indexfinger and thumb changed systematically with thewidth of the object, just as in normal subjects (Fig.2B). In other words, DF scaled her grip to the

22

Control (CJ)

system to the human parietal cor-tex provides action-relevant infor-mation about the structural charac-teristics and orientation of objects,and not just about their position.On the other hand, projections tothe temporal lobe may furnish ourvisual perceptual experience, andit is these that we postulate to beseverely damaged in DF.

TINS, Vol. 15, No. 1, 1992

responses depend greatly on the concurrent be-haviour of the animal with respect to the stimulus.Separate subsets of cells in the posterior parietalcortex have been shown to be irnplicated in visualfixation, pursuit and saccadic eye movements, eye-hand coordination, and visually guided reachingmovements32. Many teIls in the posterior parietalregion have gaze-dependent responses; i. e. wherethe animal is looking determines the response ampli-tude of the tell (although not the retinallocation of itsreceftive field)33. In reviewing these studies, Ander-sen3 emphasizes that most neurones in this area'exhibit both sensory-related and movement-relatedactivity.' In a particularly interesting recent develop-ment, Taira et al.34 have shown that same parietalteIls are sensitive to those visual qualities of an objectthat determine the posture of the band and fingersduring a grasping movement. They studied neuronesselectively associated with band movements made by Visual and attentional requirements torthe monkey in reaching and picking up solid objects. perception and actionMany of these teIls were selective tor the visual As DeYoe and Van Essen15 have suggested,appearance of the object that was to be manipulated, 'parietal and temporal lobes could both be involved inincluding its size and in several tages its orientation. shape analysis hut associated with different compu-

The posterior parietal cortex may receive such tational strategies.' For the purposes of identification,form information from Olle or both of the areas V3 or learning and distal (e.g. social) transactions, visualV4, both of which project to area MT35. Other visual coding often (though not always44.46) needs to beinputs pass throUgh area MT and the adjacent medial 'object-centred'; i.e. constancies of shape, size,superior temporal (MST) area, both of which contain colour, lightness, and location need to be maintainedteIls variously selective tor object motion in different across different viewing conditions. The above evi-directions, including rotation and motion in depth32. dence from behavioural and physiological studiesThus, the posterior parietal cortex appears to receive supports the view that the ventral stream of pro-the necessary inputs tor continually updating the cessing plays an important role in the computation ofmonkey's knowledge of the disposition and structural such object-specific descriptions. In contrast, actionqualities of objects in its three-dimensional ego-spate. upon the object requires that the location of the objectAlso, many motion-sensitive teIls in the posterior and its particular disposition and motion with respectparietal cortex itself appear to be weil suited tor the to the observer is encoded. For this purpose, codingvisual monitoring of limb position during reaching of shape would need to be largely 'viewer-centred'49,behaviour'36; in contrast, motion-sensitive cells in the with the egocentric coordinates of the surface of thetemporal lobe have been reported not to respond to object or its contours being computed each time thesuch self-produced visual motion37. As für the output action occurs. We predict that shape-encoding cells inpathways, the posterior parietal region is strongly the dorsal system should predominantly have thislinked with those pre-motor regions of the frontal property. Nevertheless, certain constancies, such ascortex directly irnplicated in ocular controI33.38, size, would be necessary tor accurate scaling of graspreaching movements of the limb39, and grasping aperture, and it might therefore be expected that theactions of the hand and fingers39. visual properties of the manipulation cells found by

Thus, the parietal cortex is strategically placed to Taira et al.34 in the posterior parietal region wouldserve a mediating role in the visual guidance and have this property.integration of prehensile and other skilled actions (see It is often suggested that the neuronal properties ofRef. 40 für a detailed account of this argument). The the posterior parietal cortex qualify it as the primeresults of behavioural analyses of monkeys with mediator of visuospatial attentionso. Certainly, manyposterior parietal damage support this further. Like cells (e.g. in area 7a) are modulated by switches ofpatients with optic ataxia, such animals faiJ to reach attention to different parts of the visual field51.correctly für visual targets41, and they also have (Indeed, the 'landmark' disorder that follows posteriordifficulty in shaping and orienting their bands when parietal damage in monkeys may be primarily due to aattempting to retrieve food42.43. Their reaching irn- faiJure to attend or orient rather than a faiJure topairment is, therefore, Olle symptom of a wider localize9.40.52.) However, it is now known that atten-visuomotor disorder, and most of the deficits that tional modulation occurs in neurones in many parts ofhave been reported on 'maze' tasks following pos- the cortex, including area V 4 and the inferotemporalterior parietal damage may also be visuomotor in region within the ventral stream53.54. This mightnature9.40. explain the occurrence of landmark deficits after

Nonetheless, neurones in the dorsal stream do not inferotemporal as weIl as posterior parietal damage7.8.show the high-resolution selectivity characteristic of In general terms, attention needs to be switched toneurones in the inferotemporal cortex, which are particular locations and objects whenever they are thestrikingly sensitive to form, pattern and colourl0. In targets either tor intended action51.55 or fürthis and in neighbouring temporal lobe areas, same identification54. In either case, this selection seemsceIls respond selectively to faces, to bands, or to the typically to be spatially based. Thus, human subjectsappearance of particular actions in others44. There- performing manual aiming movements have a

TINS, Val. 15, No. 1, 1992

fore, it is unsurprising that monkeys with inferotem-para! legions have profound deficits in visua! recog-nition; however, as noted by Pribram45, they remainhighly adept at the visually demanding skill of catchingflies!

A further peculiarity of many visual cells in thetemporal cortex is that they continue to maintain theirselective responsiveness over a wide range of size,colour, optical and viewpoint transformations of theobject44.46. Such cells, far trom providing the momen-tary information necessary für guiding action, specifi-cally ignore such changing details. Consistent withthis, behavioural studies have shown that by lesioningthe inferotemporal cortex (but not the posteriorparietal cortex), a monkey is less ahle to generalize itsrecognition of three-dimensional shape across viewingconditions47.48.

23

--

predilection to attend to visual stimuli that occur with-in the 'action space' of the hand56. In this instancethe attentional facilitation might be mediated by mech-anisms within the dorsal projection system; in otherinstances it is probably mediated by the ventralsystem. Indeed, the focus of lesions causing thehuman attentional disorder of 'unilateral neglect' isparietotemporal (unlike the superior parietal focus füroptic ataxia25) , as is the focus für object constancyimpainnents57. We conclude that spatial attention isphysiologically non-unitary55, and may be as muchassociated with the ventral system as with the dorsal.

A speculation about awarenessThe evidence from the brain-damaged patient DF

described earlier suggests that the two corticalpathways may be differentiated with respect to theiraccess to consciousness. DF certainly appears tohave no conscious perception of the orientation ordimensions of objects, although she can pick them upwith remarkable adeptness. It may be that informationcan be processed in the dorsal system withoutreaching consciousness, and that this prevents inter-ference with the perceptual constancies intrinsic tomany operations within the ventral system that doresult in awareness. Intrusions of viewer-centredinformation could disrupt the continuity of objectidentities across changing viewpoints and illuminationconditions.

If this argument is correct, then there should beoccasions when normal subjects are unaware ofchanges in the visual array to which their motorsystem is expertly adjusting. An example of such adissociation hag been reported in a study on eye-handcoordination during visually guided aiming58. Subjectswere unable to report, even in forced-choice testing,whether or not a target bad changed position during asaccadic eye movement, although correction saccadesand manual aiming movements directed at the targetshowed near-perfect adjustments tor the unpredict-ahle target shift. In other words, an iIIusory percep-tual constancy of target position was maintained in theface of large amendments in visuomotor control. Inanother re cent example, it has been reported that thecompelling illusion of slowed motion of a movingcoloured object that is experienced at equiluminancedoes not prevent accurate ocular pursuit under thesame conditions (see Ref. to Lisberger and Movshon,cited in Ref. 59). Such observations may illustrate theindependent functioning of the ventral and dorsalsystems in normal humans.

We do not, however, wish to claim that the divisionof labour we are proposing is an absolute one. Inparticular, the above suggestion does not imply thatvisual inputs are necessarily blocked from awarenessduring visuomotor acts, although that may be a usefuloption to have available. Rather, we assume that thetwo systems will often be simultaneously activated(with somewhat different visual information), therebyproviding visual experience during skilled action.Indeed, the two systems appear to engage in directcrosstalk; tor example, the posterior parietal andinferotemporal cortex themselves interconnect33.60and both in turn Eroject to areas in the superiortemporal SuiCUsll- 3. There, cells that are highlyform selective lie close to others that have motionspecificity«, thus providing scope tor cooperation

24

between the two systems (see Fig. 1). In addition,there are many polysensory neurons in these areas,so that not only visual hut also cross-modal interactionbetween these networks may be possible. Trus mayprovide same of the integration needed tor theessential unity and cohesion of most of Dur perceptualexperience and behaviour, althoUgh overall control ofawareness may ultimately be the responsibility ofsuperordinate structures in the frontal cortex61.Nevertheless, it is feasible to maintain the hypothesisthat a necessary condition tor conscious visual experi-ence is that the ventral system be activated.

Concluding remarksDespite the interactions between the dorsal and

ventral systems, the converging lines of evidencereviewed above indicate that each stream uses visualinformation about objects and events in the world indifferent ways. These differences are largely areflection of the specific transformations of inputrequired by perception and action. Functional modu-larity in cortical visual systems, we believe, extendstrom input fight through to output.

Selected references1 Schneider, G. E. (1969) Science 163,895-9022 Ingle, D. J. (1982) in Analysis of Visual Behavior (Ingle, D. J.,

Goodale, M. A. and Mansfield, R. J. W., eds), pp. 67-109,MIT Press

3 Goodale, M. A. and Murison, R. (1975) Brain Res. 88,243-255

4 Goodale, M. A. and Milner, A. D. (1982) in Analysis of VisualBehavior (Ingle, D. J., Goodale, M. A. and Mansfield, R. J. W.,eds), pp. 263-299, MIT Press

5 Ungerleider, L. G. and Mishkin, M. (1982) in Analysis ofVisual Behavior (Ingle, D. J., Goodale, M. A. and Mansfie!d,R. J. W., eds), pp. 549-586, MIT Press

6 Gross, C. G. (1973) Prog. Physiol. Psychol. 5, 77-1237 Pohl, W. (1973) J. Comp. Physiol. Psychol. 82, 227-2398 Ungerleider, L. G. and Brody, B. A. (1977) Exp. Neurol. 56,

265-2809 Milner, A. D., Ockleford, E. M. and Dewar, W. (1977) Cortex

13,350-36010 Desimone, R. and Ungerleider, L. G. (1989) in Handbook of

Neuropsychology, Vol. 2 (Boiler, F. and Grafman, J., eds), pp.267-299, Elsevier

11 Boussaoud, D., Ungerleider, L. G. and Desimone, R. (1990)J. Comp. Neurol. 296, 462-495

12 Morel, A. and Bullier, J. (1990) Visual Neurosci. 4, 555-57813 Baizer, J. S., Ungerleider, L. G. and Desimone, R. (1991)

J. Neurosci. 11, 168-19014 Livingstone, M. and Hubei, D. (1988) Science 240,740-74915 DeYoe, E. A. and Van Essen, D. C. (1988) Trends Neurosci.

11,219-22616 Schiller, P. H. and Logothetis, N. K. (1990) Trends Neurosci.

13,392-39817 Maunsell, J. H. R., Nealy, T. A. and De Priest, D. D. (1990)

J. Neurosci. 10, 3323-333418 Nealey, T. A. and Maunsell, J. H. R. (1991) Invest. Ophthal-

mol. Visual Sci. 32 (Suppl.) 111719 Ferrera, V. P., Nealey, T. A. and Maunsell, J. H. R. (1991)

Invest. Ophthalmol. Visual Sci. 32 (Suppl.) 111720 Goodale, M. A. (1983) in Behavioral Approaches to Brain

Research (Robinson, T. E., ed.), pp. 41-61, Oxford UniversityPress

21 Goodale, M. A. (1988) in Computational Processes in HumanVision: An Interdisciplinary Perspective (Pylyshyn, Z., ed.), pp.262-285, Ablex

22 Sparks, D. L. and May, L. E. (1990) Annu. Rev. Neurosci. 13,309-336

23 Newcombe, F., Ratcliff, G. and Damasio, H. (1987) Neuro-psychologia 25, 149-161

24 Farah, M. (1990) Visual Agnosia, MIT Press25 Perenin, M- T. and Vighetto, A. (1988) Brain 111, 643-67426 Jakobson, L. S., Archibald, Y. M., Carey, D. and Goodale,

M. A. (1991) Neuropsychologia 29,803-80927 Milner, B. (1965) Neuropsychologia 3,317-338

TINS, Vol. 15, No. 1, 1992

28 Smith, M. L. and Milner, B. (1989) Neuropsych%gia 27,71-81

29 Ettlinger, G. (1990) Cortex 26,319-34130 Milner, A. D. eta/. (1991) Brain 114, 405-42831 Goodale, M. A., Milner, A. D., Jakobson, L. S. and Carey,

D. P. (1991) Nature 349, 154-15632 Andersen, R. A. (1987) in Higher Functions of the Brain, Part

2 (The Nervous System, Va/. V, Handbook of Physi%gy,Section 1) (Mountcastle, V. B., Plum, F. and Geiger, S. R.,eds), pp. 483-518, American Physiological Association

33 Andersen, R. A., Asanuma, C., Essick, G. and Siegel, R. M.(1990) J. Camp. Neuro/. 296, 65-113

34 Taira, M., Mine, S., Georgopoulos, A. P., Murata, A. andSakata, H. (1990) Exp. Brain Res. 83, 29-36

35 FeIleman, D. J. and Van Essen, D. C. (1987) J. Neurophysio/.57,889-920

36 Mountcastle, V. B., Motter, B. C., Steinmetz, M. A. andDuffy, C. J. (1984) in Dynamic Aspects of Neocortica/Function (Edelman, G. M., Gall, W. E. and Cowan, W. M.,eds), pp. 159-193, Wiley

37 Perrett, D. I., Mistlin, A. J., Harries, M. H. and Chitty, A. J.(1990) in Vision and Action: The Contro/ of Grasping(Goodale, M. A., ed.), pp. 163-180, Ablex

38 Cavada, C. and Goldman-Rakic, P. S. (1989) J. Camp.Neuro/. 287, 422-445

39 Gentilucci, M. and Rizzolatti, G. (1990) in Vision and Action:The Contro/ of Grasping (Goodale, M. A., ed.), pp. 147-162,Ablex

40 Milner, A. D. and Goodale, M. A. Prog. Brain Res. (in press)41 Bates, J. A. V. and Ettlinger, G. (1960) Arch. Neuro/. 3,

177-19242 ~~ugier-Gr!m~u~, ~:; ~~e.no!s:._C. and Stein, D. G. (1978)42 Faugier-Grimaud, S., Frenois, C. and Stein, D. G. (1978)

Neuropsycho/ogia 16, 151-16843 Haaxma, R. and Kuypers, H. G. J. M. (1975) Brain 98,

239-260

Control 01 neuronal

Marian Joels and E. Ronald de Kloet

The rat adrenal hotmone corticosterone fan cross theblood-brain barrier and bind to two intracellularreceptor populations in the brain - the mineralocorticoidand glucocorticoid receptors. Recent studies have re-vealed thai the corticosteroid hotmones are abte tores tore changes in neuronal membrane propertiesinduced by current or neurotransmitters, probablythrough a genomic action. In general, mineralocorti-raid receptors mediate steroid actions that enhancecellular excitabiliiy, whereas activated glucocorticoidreceptors fan suppress temporarily raised neuronalactiviiy. The steroid-mediated control of excitabiliiy andthe imPlications foT information processing in the brainare reviewed in this article.

It hag been acknowiedged für many years that adrenalcorticosteroid hormones that are reieased into thebiood circulation can cross the biood-brain barrier andbind to intracellular receptors in the brain (see Refs1-3). During the 1960s, McEwen and co-workersshowed with the help of radioligand binding andautoradiography that (3H]corticosterone, adrninisteredto adrenalectomized (ADX) rats, is retained by intra-cellular receptors in some brain structures, particu-iariy in the hippocarnpus4. The steroid-receptor com-piex displays increased affinity für the cell nuciearcompartment; it can bind to the genome and act as atranscription factor für specific genesl-3.

The iocalization of intracellular corticosteroid hor-mone receptors in brain structures naturaily raised

TINS, Val. 15, No. 1, 1992

AcknowledgementsThe authors aregrateful to D. rarer,L. Jakobson andD. Perrett For theircomments on a draftoF this paper.

44 Perrett, D. I., Mistlin, A. J. and Chitty, A. J. (1987) TrendsNeurosci. 10, 358-364

45 Pribram, K. H. (1967) in Brain Fundion and Learning(Lindsley, D. B. and Lumsdaine, A. A., eds), pp. 79-122,University of California Press

46 Perrett, D. I. et a/. (1991) Exp. Brain Res. 86, 159-17347 Humphrey, N. K. and Weiskrantz, L. (1969) Quart. J. Exp.

Psychol. 21, 225-23848 Weiskrantz, L. and Saunders, R. C. (1984) Brain 107,

1033-107249 Marr, D. (1982) Vision, Freeman50 Goldberg, M. E. and Colby, C L. (1989) in Handbook of

Neuropsychology (Vol. 2) (Boiler, F. and Grafman, J., eds),pp. 301-315, Elsevier

51 Bushnell, M. C, Goldberg, M. E. and Robinson, D. L. (1981)J. Neurophysiol. 46, 755-772

52 Lawler, K. A. and Cowey, A. (1987) Exp. Brain Res. 65,695-698

53 Fischer, B. and Boch, R. (1981) Exp. Brain Res. 44,129-13754 Moran, J. and Desimone, R. (1985) Science 229, 782-78455 Rizzolatti, G., Gentilucci, M. and Matelli, M. (1985) in

Attention and Performance XI (Posner, M. I. and Marin,O. S. M., eds), pp. 251-265, Erlbaum

56 Tipper, 5., Lortie, C and Baylis, G. J. Exp. Psychol. HumanPercept. Perform. (in press)

57 Warrington, E. K. and Taylor, A. M. (1973) Cortex 9,152-164

58 Goodale, M. A., Pelisson, D. and Prablanc, C (1986) Nature320, 748-750

59 Sejnowski, T. J. (1991) Nature 352,669-67060 Cavada, C and Goldman-Rakic, P. S. (1989) J. Camp.

Neuro/. 287, 393-42161 Posner, M. I. and Rothbart, M. K. (1991) in The Neuro-

psychology of Consciousness (Milner, A. D. and Rugg, M. D.,eds), pp. 91-111, Academic Press

ex dtability by corticosteroid hormones

the question as to whether cellular activity, andparticularly the electrical properties of neurons, couldbe affected by the hoffilones. An early study by Pfaffet al. showed that in hypophysectornized rats thatreceived a peripheral injection of cortisol (the adreno-cortical steroid found in humans and primates), spon-taneous single-unit activity in the hippocampus wasreduced with a delay of approximately 30 min (Ref. 5).However, subsequent extracellular recording ofhippocampal, forebrain and hypothalamic neuronsrevealed a disparity in the effects of corticosteroidhoffilones, which were excitatory or inhibitory orwhich exerted no changes at a116-9.

In retrospect, two important factors may havecontributed to the variability in results. Olle factorrelates to the background electrical activity of thetissue exposed to the corticosteroid hoffilones. Theeffects of corticosteroid may weil be voltage depen-dent and derive their excitatory or inhibitory naturefrom the prevailing level of excitability. The extracel-lular recording methods used in vivo in the studiesmentioned above do not allow control of the back-ground electrical activity, in contrast to methodsdeveloped tor use in vitra over the past decades. Thesecond factor sterns from the realization over the pastsix years that corticosterone in the rat brain binds totwo intracellular receptor populations: the mineralo-corticoid receptor (MR), which binds corticosteronewith high affinity and is discretely localized, par-ticularly in neurons of lirnbic structures; and the

Marian Joels is at the

OeptofExperimentalZoology, University ofAmsterdam, 1098SM Amsterdam, TheNetherlands, andE. Ronald de Kloetis at the Center for

BiophannaceuticalSciences, LeidenUniversity, 2300 RALeiden, TheNetherlands.

250166 - 2236/92/SQ500