《认知神经科学》课程教学资源（参考文献）[Goldman-Rakic, P. S.（1987）]Development of cortical circuitry and cognitive function

Development of Cortical Circuitry and Cognitive Function Patricia S.Goldman-Rakic y include of s to th keys at vario esterod studies city t e end gs sugge nd that fully Pmcheeogenogpehlcgrtaten about t the un derlying neur 3 not neces absolutely nec the eme 1oegicalrgcientence. euroanatomy,n ophysiolog and neuro ny,neurophysiology study by its the ability to ept for biocher tive approaches w which ms carr iment to 5 euro nsmitt and ne led kn t ural and whe but that ci the ac t mamma eural ryology onc and not insigni the deve pment of the empty or cult that ed moin ity for unpre erime hudiestg still felt in many prominent departments of Goldman-Rakic.19821965). Uahn8em58aP8aeasHwofModctae,Yad heteCldDvlomenc

Development of Cortical Circuitry and Cognitive Function Patricia S. Goldman-Rakic Yale Unit>ersity School of Medicine GOLDMAN-RAHC, PATMCIA S. Development of Cortical Circuitry and Cognitive Function. GHILD DEVEIJOPMENT, 1987,58,601-622. Recent fiinctionai and anatomical studies in nonhuman primates have elucidated the basic neural circuitry underlying delayed-respohse function in adult nonhuman primates. Thus circuitry includes connections of the principal sulcus with other areas of parietal association and limbic cortex and projections to the caudate nucleus, superior colliculus, and other premotor centers. Anatomical tracing in primate fetuses and in monkeys at various stages of post￾natal development indicates tfiat these various classes of cortical connections begin to form by tiie second trimester of pregnancy. Electromicroscopic studies of the principal sulcus and odier areas of cerebral cortex show that the number and density of synapses in die cortex increase rapidly, reach￾ing and maintaining higher tiian normal adult values between 2 and 4 montfis postnatally, before slowly declining over a period of years to stable adult levels. The capacity to perfonn delayed￾response and/or AB at short delays emerges around 4 months of age, coinciding with the end of the period of highest synaptic density in the principal suleus. These findings suggest that a critical mass of cortical synapses is important for die emergence of this cognitive function, and that fully mature capacity may depend upon the elimination of excess synapses that occurs during adolescence and young adulthood. Knowledge of the neural basis of nonnal cognitive development may prove useful both to social and educational purposes as well as to understanding developmental disorders of cognition. Research on the development of behav￾ior has proceeded for decades without benefit of knowledge about the underlying neuro￾logical maturation that is correlated with and absolutely necessary for the emergence of increasingly complex functions with age. Historically, &e absence of hard facts about neurological development was due to many factors. First and foremost, the methods of neuroanatomy, neurophysiology, and neuro￾chemistry were relatively primitive and con￾cepts of developmental mechanisms were accordingly rather general. Developmental study by its nature requires the ability to quantify change, and, except for biochemical assays, until quite recently, genuine quantita￾tive approaches were not feasible for measur￾ing many parameters of neural maturation. A second impediment to developmental neuro￾logical studies was initially the lack of de￾tailed knowledge about the structural and functional organization of the adult mamma￾lian brain, but that circumstance no longer ex￾ists. A third and not insignificant fector un￾doubtedly was the antinativist climate and "empty organism" cult that permeated main￾stream psychology for almost half a century, still felt in many prominent departments of psychology even today. The "empty organ￾ism" has long since been filled wilii inten￾tionality and information-processing skills, but not necessarily with a central nervous sys￾tem. Circumstances for a genuine neurobio￾logical science of mammalian development have changed radically in the past 15 years. Mfgor technical advances have taken place in the fields of neuroanatomy, neurophysiology, and neurochemistry, and, moreover, the con￾ceptual and technical boundaries between diese once strictly parochial disciplines have merged together into the truly multidiscipli￾nary endeavor of neuroscience. These ad￾vances have allowed definition of the cir￾cuitry by which neural systems carry out specific functions, including discovery of the neurotransmitters, enzymes, and neuromodu￾Iators that are at the heart of intercellular com￾munication. Moreover, whereas studies of neural embryology were once associated largely with research in avian and amphibian species, the development of the technique of prenatal surgery has opened up the possibil￾ity for unprecedented experimental studies in mammals, including fetal primates (Raldc 6c Goldman-Rakic, 1982, 1985). Send request for reprints to Section of Neuroanatomy, C303 Sterling Hall of Medicine, Yale University School of Medicine, P.O. Box 3333 New Haven, CT 06510-8001. IChM Developmeat, 1987,58,601-622. O 1987 by the Society for R€ssearch in Ghild Development, Inc. All rights reserved. 000&O920/87/5803-0021*01.00]

602 Child Development The pr sent article will review recent nan,to learn and respond to situations or tud frontal cortex in rhesus m nkeys We able part in pretront cor the anima Sumcicnty weilanbas huma cter of an .1).Fo wing a d provide an ad quate basis loce Fu aye 19)Oe major has been to co onstnctaneooeaeiernanteherg vatepateper o光a o cued by 6 emerge. did on the previous trial. many basic The are of the function(s)tap ed by es thnt dehavedeponeic ks has been tion. hasized the sa al ratu FsteE dy brain other highlighted their 955,a 1972. Cu the view Delayed Regulatior ior by represe of an c ha Passin am. 1985). This type me ory i nse tasks (e.g ncet al. ore 197o:G0 simu e arlow also erely layed- ponse is that in 1). in the ad havior by a repre ciatic ap but see Fuster,1980,198 ep rethorced ogical processes that to the next trial.An in ceys are ere at spons hand,are absent at nil in the

602 Child Development The present article will review recent studies conducted in my laboratory that con￾cern the biologica] development of the pre￾frontal cortex in rhesus monkeys. We have been particularly interested in prefironta! cor￾tex because its preeminent role in cognitive functions makes it a part of the brain with special significance for human society and culture. Moreover, this is one of the few areas of association cortex that has been character￾ized sufficiently well anatomically, biochem￾ically, and functionally in adult monkeys to provide an adequate basis for developmental comparison (for review, see Fuster, 1980, 1985; Goldman-Rakic, 1987; Stuss & Benson, 1986). One major goal of our studies has been to construct a neurology for particular cogni￾tive functions and to determine die critical cellular events necessary for mature function to emerge. Knowledge of the normal se￾quence and mechanisms of neur^ maturation is necessary, in my opinion, for approaching many basic issues in cognitive and develop￾mental psychology. It is also fundamental for interpreting the pathological processes that occur both in severe developmental disorders such as mental retardation, childhood schizo￾phrenia, and autism as well as in milder con￾ditions like dyslexia and minimal brain dys￾function. SynopolB of Prefirontal Function in Adult Primates Delayed responses: Regulation of behav￾ior by representational knowledge.—In mon￾keys, a specific subregion of the prefrontal cortex, that surrounding the principal sulcus (shown in Fig. 2A and to be described further in a later section), mediates performance on delayed-response tasks (e.g., Goldman et al., 1971; Goldman & Hosvold, 1970; Goldman￾Rakic, in press; Harlow, Davis, Settiage, & Meyer, 1952; Jacobsen, 1936). Humans with bilateral prefrontal lesions are also severely impaired on these tasks (Freedman & Oscar￾Berman, 19^). Aldiough the dependence of delayed-response performance on prefrontal cortex is very well established, it's functional signiiicance is not usually fully appreciated (but see Fuster, 1980, 1985; GoIdman-Raklc, in press). It may therefore be helpful to con￾sider briefly the psychologicfd processes that are measured in the delayed-response para￾digm as a background for studies of cognitive development. The delayed-response task was designed by the comparative psychologist Walter Hunter (Hunter, 1913) to analyze and com￾pare the abilities of different species, includ￾ing m£«i, to learn and respond to situations on the basis of stored infonnation; it was rea￾soned that this ability is different and dissoci￾able from leaming and memory in situations in which all necessary information is present at the time of response. Accordin^y, in the classical visuospatial version of the test, the animal observes a food morsel being hidden in one of two locations, which are then blocked from view by the lowering of an opaque screen (Fig. 1). Following a delay pe￾riod lasting from 1 to several seconds, the sub￾ject is allowed to select one of the two loca￾tions. In a variation of die test, spatial delayed alternation, the subject is required to altemate between lefl and right food wells on succes￾sive trials that are separated by delay periods. In bodi tasks, die correct choice cannot be cued by any stimuli present at the time of response; ra^er, it must be guided by inter￾nalized knowlec^e of what die subject saw or did on the previous trial. The nature of die function(s) tapped by the delayed-response tasks has been die sub￾ject of coundess studies and much specula￾tion. Some authors have emphasized the sa￾lient mnemonic aspects of the tasks (Fuster, 1973, 1980, 1985; Jacobsen, 1936), while odiers have hi^i^te d their spatial nature (e.g., Mishkin & Pribram, 19^), and still others have focused on motor control fiinc￾tions (Stamm, 1979; Teuber, 1972). Gurrent thinking in die field supports the view that die delayed-response task indexes the ability of an oi:ganism to guide its behavior by stored information, that is, representational memory (Fuster, 1^0, 1985; Goldman-Rakic, 19S7; Passins^iam, 1985). This type of memory is more akin to working memory and is diflferent from die type of associative memory diat is required, for example, in leaming a simul￾taneous visual discrimination task. A major difference between discrimination and de￾layed-response tasks is that in die former, a discriminative stimulus is always present to g^de respon&ng, while in the latter, be￾havior is guided by a representation of a prior stimulus. Also, in associative learn￾ing, a given stimulus-response asscxriation is repeatedly reinforced and strengthened, whereas in working memory tasks like de￾layed response, inform^on fiom one trial is irrelevant to the next trial. An importfuit jKiint is that the delayed-response deficits produced by prefrontaJ lesions in monkeys are nearly as severe at delays of 1—2 sec between cue and response as at much longer delays and, on the other hand, are absent at nil delays (Goldman, 1971). This type of evidence indicates a sharp dissociation in the neural mechanisms

Patricia S.Goldman-Rakic 603 that mediate cates prefrontal cortex primarily in the former response and CUE to iect Perman kic 1983). of two spatiallys rated locations:ade in one tolows,and then Piaget,the lity and 840 has been human cognitive development is the same as an ay these task inistered to subjects.Mo import the seque than in de nse:in Ap an ent trial th ate ocation of the In the expe RESPONSE task the of the bait varl in a quas is minor 0 n the the ks are a we tho AB de as Diamon keys with mo esion contro groups on human Diamond. es child on tec ed est the animal's cor ent and the

Patricia S. Goldman-Rakic 603 CUE DELAY RESPONSE FIG. 1.—The three components of a delayed￾response trial. In the cue period (top panel), the monkey watches an experimenter hide a peanut half in one of two food wells and then cover bodi wells with identical cardboard plaques. A delay (middle panel) of 1 or more sec is Iben imposed by the lowering of an opaque screen that effectively blocks die animal from responding immediately. In the response phase of tiiie trial (lower panel), the screen is raised and the animal allowed to select one of the two food wells. Monkeys with lesions of the principal sulcus perform the task but do so at chance levels of accuracy. The delayed-response task is similar in all formal respects to Piaget's A 5 Test of Object Permanence (see text for furdier ex￾planation) (from Goldman-Rakic et al., 1983). that mediate representational and associative guidance of behavior, respectively, and impli￾cates prefrontal cortex primarily in the former. Equivalence of flayed response and Piaget's Stage IV AB Object Permanence Test.—The delayed-response task and its neural basis may be of particular interest to developmental psychologists _because of its formal similarity to Piaget's AB Stage IV Ob￾ject Permanence Task and our recent demon￾stration that AB performance, like delayed￾response ability, depends upon the integrity of prefrontal cortejc (Diamond & Goldman￾Rakic, 1983). In AB, as in delayed response, the child watches as a reward is hidden in one of two spatially separated locations; a delay of a few seconds follows, and then the child is allowed to find the reward. According to Piaget, die AB test reflects the origins of in￾tentionality and "formal operations" in chil￾dren's diought processes (Piaget, 1954), and it has been used worldwide to chart milestones in human cognitive development. Aldiough each trial in the AB paradigm is the same as an individual trial in the de￾layed-response task, in practice there are pro￾cedural difiFerences in the ways these tasks are administered to subjects. Most important, in AB the sequence of events is under the control of the experimenter to a^reater extent than in delayed response: in AB an object is hidden at one of two locations until the sub￾ject is correct on a given trial; on the subse￾quent trial, the location of the bait is deliber￾ately repeated or reversed (by the experi￾menter). In the classical delayed-response task, the position of the bait varies in a quasi￾random schedule from trial to trial. Because of this minor procedural difference in the way the two tasks are administered, we thought it would be important to test directly whether AB depends upon prefrontal cortex as does delayed response (Diamond & Goldman￾Rakic, 1983). Accordingly, adult rhesus mon￾keys with prefrontal lesions were compared to nonnal and lesion control groups on AB administered as nearly exactly as it is given to human infents (Diamond, 1985). Some species-related differences in_ testing were, however, unavoidable. For AB testing in hu￾mans, the child is lighdy restrained on its mother's lap, whereas monkeys, freely mov￾ing, are tested from a cage. For humans, a distraction technique is used to impose a de￾lay (Diamond, 1985); with monkeys, the delay is imposed by lowering an opaque screen be￾tween the animal's compartment and die test tray

604 Child Development a航ents wit ative response on a pde variety ke d that er h play infan s.For consin ort Test (Mine ple human infants and pre 1982 Knight,Hillyar the same that was 1981 sly rewarded bu Task (Shallic (D )and tasks that p uire on,Buc They have a strong to ndency ided by ass ples The Ho y rev a data hat d1 er ed theB error.Dur atal de ermr ility may be in the ssively d t 4 are ace Com anng the 1987. m man and that br out this ab d ad have at the le hun t at th d thos lesi CO behave as do that the sity Eeoee yatt出 on wo ld be m beha be pre ta on ativ aved nse events Th ure a ront be c may ere why tone, onse rat than ex ognition tha 1a6 aPP Di time ,1975 the ele also explain why ntal ca e outside rld and to b I0 tests (e.g. Hebb 939 nt(For excellent dis ide ion ance of shortterm life. The be avior o he of see ter,1980.1985 pro foll e ha the man infa to behave of hu with pretr e t or on monkeys long after it has 1985 1987；Jacob 6).By virtue of the ems to re tations for the time n ces observe a re information about the location of coul

604 ChUd Dev^opment Our results estaUished that AB, like de￾layed response, is impaired selectively by dorsolateial prefronttJ lesions in adult rhesus monkeys, and that die pattem of erTors dis￾played by the operated monkeys closely re￾sembles that exhibited by human infants. For example, human in&nts and i^efrontal mon￾keys are nearly always correct when the ob￾ject is hidden in the same location that was previously rewarded but perfimn near chance when the position is reversed (Diamond, 1985; Diamond & Goldman-Bakic, 1983). They have a strong tendency to repeat the previously rewarded response and are strongly guided by associative principles. The selection of a previously reiOTirded response rather than one glided by updated visual data is temied the J ^ error. During postaatal de￾velopment, progressively longer delays be￾tween hiding and response are required to elicit this error in children (Diamond, 1985). Gomparing the human and monkey data, we were able to demonstrate diat the brain￾lesioned adult monkeys behave at the level of 7V'2-9-month-old human in^ts, making the AB error widi delays of 2-5 sec. Nomial mon￾keys and those with lesions of die posterior parietal cortex behave as do 12-month-old hu￾man infente, who perform correctly at these delays. On die basis of their common depen￾dence on prefrontal cortex Mid on what we know and can dieorize about prefrontal cor￾tex, we conclude that bodi ddayed response and AB tasks measure a common jmwess— die emergence of representational memory. This capacity may be considered a building block, if not a comerstone, of cognitive devel￾opment in man. Moreover, "object perma￾nance"—the recognition that an otgect has continuity in time and space when not in tjteuj—must depend on die elemental capac￾ity to form representations of the outside world and to base responses on those repre￾sentations In die absence of the objects diey represent. (For excellent discussion of die re￾levance of short-term memory processing to the temporal o^jaaization of behavior, see Fuster, 1980, 1985.) Representational processes disaUow in￾appropriate responses.—A cardinal feature of die behavior of human infents and of hunians and monkeys with prefrontal lesions is the strong tendency to persevenUfi a particul^ re￾sponse long after it has ceased to be ^propri￾ate. For example, a monkey with a prefroatal lesion, unlike a nramal monkey, seems to re￾spond to the first food well &at catches its eye or to die just previously reinforced position radier than on the basis of recendy observed informadon about the location of the reward. Patients with frc»ital lobe damage are simi￾larly perseverative, exhibiting impairments on a wide variety of tests that require suppres￾sion of prept^nt responses (or responses to distracting stimulation), including the Wis￾consin Gard Sort Test (Milner, 1963, 1964), tests of selective attention (Kni^t, Hillyard, & Neville, 198?* «ni ^ Hillyard, Woods, & Neville, 1^0, 1981), die Stroop Test (Perret, 1974), the Tower of London Task (Slu^ice, 19^), and tasks that require inhibition of pre￾potent responses (Guitton, Buchtel, & Doug￾las, 1982, 1^5). Some have considered per￾severation as a primary deficit attiibuteble to damage to the orbital prefrontal cortex (Fus￾ter, 1985; Mishkin, 1964). However, I have argued feat perseveratiw behavior and dis￾tractability may be secondary consequences of file more fundamental impairment in the mechanism by which symbolic rei»esenta￾tions are accessed and held "on line" to guide a response (Goldman-RaJdc, 1987). It seems possible diat witfiout diis ability, an organism would be virtually compelled to respond refexively to stimuli present in the environ￾ment at the time of resp<mse or on the basis of prepotent response tendMicies. Further, it is easy to envision that die intensity and sa￾lience of stimulalion would be important fac￾tors goveming b^avior under such circum￾stances. In acMition, behavior mi£^t be regulated largely by associative c(mditioning to events that occurred in close temporal proximity. The sparing of associative memory after prefrontal irqury may help to explain why human ii^iits diaracteristically repeat a previously rewarded response rather dian ex￾ecute a new more apEHK^riate response in Piaget's AB paaK^m (Diamond, 1^ ; Uz￾giris & Hunt, WfS). It may also explain why frontal lobe p(#^ s can have an average store of feotual iralc^iw&on suid perform wdl on conventional IQ tests (e.g., Hebb, 1939) but be grossly deficient in how they utilize dieir representational knowfedge to guide behav￾ior in everyday life. "Hie behavior of such pa￾tients, and to some extent of hum^i infiunts, is fragmented, la^b g clear direction and ejffies￾sively controH^ by extemal stimulafeon. It foUows diat die prefrontal ccatex may be nec￾essary to overricte the teiwiency to behave stricdy on die basis of reinibrcemeat or on the basis of stirou^rtkm pre&eirt at the mome^ of response (Fuster, 1985; Goldman-R^c, 1987; ]acdbsen, 1936). By virtue o£&€ sii^u￾lar process of wcffkiag memory, die jjirefrontal cortex nuiin^iiis access to internalized tepre￾sentsrtiCTis for die time necessary to craniate a response or response sequence, tlie emer￾gence of this capacity in postnatal life could

Patricia S.Goldman-Rakic 605 mplish the dual purpose from the frontal pole toward the co per rgin of the media ted by centers of asso iative con The pr al otor. to guished fo of thef unterpa t).have ompo Bai 947 0 by which the ns can be the six-l ered orga ation of its ce h ssion IV. which is vir calization of function domain bsent in motor co tex d poorly develo the princ tational kno e 0 stellate the of th aeiayodre the latt se tasks (Gol 9 an svold neurons whose of of ste and thu contr has a l arge than for fea ar dis y )of o tinctive pre al ning,1978.7 ones Mishkin 1972).The th is that en fore re ombling more e prima cialized for not and irtue its Both aman-t elicits ions of the body ture, be 98 1964 Pribram ada 8&. cord or motor nuclei moto nost studie of fr in ma e the body tsover which thev haw to one e，iti tem tor nucle and conp p05 e to dissociate these icits by exp in a in the ther ze the be its n of spe of P efro system. the and moto lat ed to as hibi ciat tion of behavior. the in mos of the inc motor and motor cortices (Fig.2A) The three ig.2A-D) the ne its anato sto see Fig. 41 The prin the netr

Patricia S. Goldman-Rakic 605 accomplish the dual purpose of initiating and guiding correct responses and per force disal￾lowing or inhibiting incorrect ones that would be mediated by centers of associative condi￾tioning. In nonhuman primates, die prefrontal cortex, in general, and the principal sulcus, in particuUir (and its human counterpart), have multiple connections with the neural centers by which these motor control Binctions can be exerted (see below and Goldman-Rakic, 1987, for friller discussion). Localization of function and domain￾specific memory.—As mentioned, the princi￾pal sulcus is the cortical focus for regulation of behavior by representational knowledge of the location of objects. Monkeys lacking prin￾cipal sulcal cortex exhibit deficits only on spa￾tial delayed-response tasks (Goldman & Ros￾vold, 1970; Goldman, Rosvold, Vest, & Galkin, 1971) while leaving memory for the features of objects unaffected. In contrast, monkeys with lesions of the inferior convexity cortex exhibit deficits on tasks requiring working memory for visual features (e.g., color, shape) of objects rather than their loca￾tion (Passingham, 1972, 1975; Mishkin & Manning, 1978; Jones & Mishkin, 1972). The conclusion from these studies is that different subdivisions of prefrontal cortex may be spe￾cialized for working memory in different in￾formational domains. Both spatial and nonspatial working memory impairments have been reported in human patients with prefrorrtal lesions (Ghorover & Gole, 1966; Gorkin, 1965; Freed￾man & Oscar-Berman, 1986; Milner, 1963, 1964, 1965; Pribram, Ahumada, Hartog, & Ross, 1964; Semmes, Weinstein, Ghent, & Teuber, 1963). This is predictable, since in most studies of frontal lobe injury in man, the site of damage is not confined to one func￾tional subdivision but more commonly ex￾tends over several. In monkeys, however, it is possible to dissociate these deficits by experi￾mentally induced selective lesions and, fur￾ther, to analyze the circuit and cellular basis of specific functions. SynomU of Prefrontal Anatomy in Adult Nonhnman Primates Gross morphology of the prefrontal asso￾ciation cortex.—The prefrontal cortex in monkeys lies in the anterior-most position in the cerebral hemispheres in front of the pre￾motor and motor cortices (Fig. 2A). The three areas together comprise the cortex of the fron￾tal lobe, widi die prefrontal cortex occupying the largest fraction (see Fig. 2A). The princi￾pal sulcus (PS) is a deep fissure extending an￾teriorly from the frontal pole toward die con￾cavity formed by the arcuate sulcus, which lies at the posterior margin of the prefrontal cortex. The prefrontal, premotor, and motor subdivisions of the frontal lobe can be distin￾guished from one another by their cellular composition (Bonin & Bailey, 1947; Brod￾mann, 1909, 1925). Thus, although each sub￾division is considered neocortex by virtue of the six-layered organization of its cells, the prefrontal cortex is distinctive in having a well-developed layer IV, which is virtually absent in motor cortex and poorly developed in premotor cortex. All neocortex contains a mixture of pyramidal and stellate (star￾shaped) cells; the former are the projection neurons of the cortex that send their axons out of the cortex to distant structures; the latter are neurons whose axons form intrinsic or lo￾cal connections within die cortex. Layer IV is fonned largely of stellate cells, and thus pre￾frontal cortex has a larger complement of in￾trinsic neurons than does premotor and motor cortex. The granular fourth layer is a dis￾tinctive feature of prefrontal cortex only in primates; in all other mammalian species studied so far, the "prefix)ntal" cortex is agran￾ular (Akert, 1964), therefore resembling more the premotor areas of the primate. Prefrontal cortex also differs from pre￾motor and motor areas by virtue of its connec￾tivity and functions (Goldman-Rakic, 1987). Briefly, electrical stimulation of motor and premotor areas elicits contractions of the body musculature, a fact that can be explained by direct connections of these areas with the spinal cord or brain-stem motor nuclei. Both the premotor and motor cortices contain somatotopic representations that correspond to the body parts over which they have juris￾diction. In contrast, prefrontal cortex has no direct connections to the spinal cord or brain￾stem motor nuclei and consequendy does not participate in a direct way in the execution, fine tuning, or metrics of motor output As might be expected, its influence over volun￾tary motor behavior is achieved throu^ more indirect connections with the primary motor system, including the premotor and motor cortex, and may be characterized as regula￾tory—related to initiation, timing, and inhibi￾tion of behavior. The principal sulcus.—Since the princi￾pal sulcus is the primary fissure in the pre￾frontal cortex (Fig. 2A-D) and the neural focus for delayad-response function, under￾standing its anatomy and physiology should provide insight into the neural substrate for regulation of behavior by stored or represen-

Cd subdivisi of the g th :IS,in ate g CD TH, OTS mporal

PRI PREFRONTAL FIG. 2.—A, Diagram of the three major subdivisions of the frontal lobe in nonhuman primates. The major sulci are the principal sulcus (PS) in prefrontal cortex, tiie arcuate sulcus (AS) that separates the prefrontal and premotor areas, and the central sulcus (CS) that is the boundary between the frontal and parietal lobes. Furdier subdivisions within prefrontal cortex are indicated: (a) inferior convexity— Brodmann's area 12; (b) cortex surrounding the principal sulcus—Brodmann's area 46; and (c) the frontal eye field—Brodmann's area 8. The mi^jor focus of the present article is Brodmann's area 46. B, Summary of areas in the posterior parietal cortex dmt process visuospatial and somatic-related infonnation (7m, lip, la, and 7b) and the distribution of their projections in die principal sulcus as revealed by correspondence in a zip-a-tone pattem (from Cavada & Coldman-Rakic, 1985), C, Summary of direct and indirect circuits linking the principal sulcus with structures involved in memory—the hippocampal formation. D, Summary ofmajor projections from the principal sulcus to areas of the brain diat are involved in motor control. Other abbreviations: IS, intraparietal sulcus; CA, Ammon's hom of the hippocampus; CC, corpus callosum; CD, caudate nucleus; CING and RS, cingulate and retrosplenial cortices; CML, caudomedial lobule; CS, collateral sulcus; DG, dentate gyms; PSUB, presubiculim; TH, TF, and 28, parahippocampal areas; OTS, occipitotemporal sulcus; SC, superior colliculus; and SMA, supplementary motor cortex

Patricia S.Goldman-Rakic 607 nation The primacy of the al nyed of evidence: on necessary ster the location of input to the ical stud (Am 00贤 arietal rozoski Pe ides Par 1984 studies 3 hich is the corticai 1980 Lynch. onal respons one the principal sulcus should act rent l ines of e tructure that 1984a 1987 (Co ctions the duce a dou ections un d le from 1970 ed hand the motor control 63 stin h5o0s cross es h tion o inhi out some Fig. ven both 1977a K5 important,prir ndet, t ed re and 84 Go ,1972b,1972c uitry exists components s of c onnections).sto circuits) and (pre examp increase connecti (or right) Columnar and laminar org 0 bee The archite die an of principe sulcus make 0阳 modular type o until the response is made. e cortex cus as dis major comnec eredetail.to the point th ete in width sulc sized col s of n be nections and principa pro The as has beer example, the in organize into six layers,and

tationa] information. The primacy of the prin￾cipal sulcus for delayed-response function has been established by at least five separate lines of evidence; (1) ablation studies (Blum, 1952; Butters, Pandya, Stein, & Rosen, 1972; Goldman & Rosvold, 1970; Mishkin & Man￾ning, 1978); (2) pharmacolo^cal studies (Am￾sten & Goldman-Rakic, 1985; Brozoski, Brown, Rosvold, & Goldman, 1979); (3) mi￾crostimulation studies (Stamm, 1969); (4) electrc^hysiological recording (Fuster, 1973, 1980; Kubota & Niki, 1971; Niki, 1974a, 1974b, 1974c; Kojima & Goldman-Rakic, 1982, 1984); and (5) functional inactivation, for example, hypodiermic iractivation (Alex￾ander & Goldman, 1978; Fuster & Alexander, 1971). These different lines of research have been extensively reviewed (Fuster, 1980; Goldman-Rakic, 1984a, 1987; Goldman￾Rakic, Isseroff, Schwartz, & Bugbee, 1983). Briefly, remold of the princi{»l sulcus is sufficient to produce a profound delayed￾response deficit, whereas equal-sized lesions in other portions of the front^ lolw have little effect (Goldman & Rosvold, 1970). Delayed￾response performance is disrupted in die first few seconds of the delay when electrical stimulation is applied across the principal sul￾cus (Stamm, 1969). Electrc^hysiological re￾cording in animals performing delayed￾response tasks has established that principal sulcal neurons can be driven both by auditory and visual stimuli (e.g., Azuma & Suzuki, 1984; Kojima, 1980; Niki, 1972a; Suzuki & Azuma, 1985). Most important, principal sul￾cal neurons respond in time-locked foshion to the xnaior events of a delayed response (Fus￾ter, 1973; Kojima & Goldman-Rakic, 1982, 1984) or delayed alternation (Kubota & Niki, 1971; Niki, 1972b, 1972c) trial. Of particular relevance is a class of cell that increases its firing selectively during the delay period of the task and in relation to die position of the stimulus. For example, a neuron may increase its dischfuge during the delay only if die cue had been on the left (or ri^t) (for review, see Fuster, 1980, 1985; also Golcbnan-Rakic, 1984a, 1987). This and other response charac￾teristics of principal sulcus neurons make them excellent candids^s for holding visuo￾spatial information "on line" until die correct response is made. Connectivity of the principal sulcus.— Over the past decade, die rasgor connections of the principal sulcus have been worfted out in considerable detail, to the point that rela￾tions between specific connections and specific functions can be seen and ap￾preciated (for reviews see Goldman-R^c, 1984b, 1984c, 1987). For example, die in￾Patricia S. Goldman-Rakic 607 volvement of the principal sulcus in delayed response implies access to visuospatial infor￾mation necessary to register the location of the food reward. Indeed, the major cortical input to the posterior two-thirds of die princi￾pal sulcus originates in the posterior parietal cortex (Fig. 2B; Goldman-Rakic & Schwartz, 1^2; Petrides & Pandya, 1984; Schwartz & Goldman-Rakic, 1984), which is the cortical center for spatial information processing (Mountcastle, Motter, Steinmetz, & Dufiy, 1984; Lynch, Mountcastle, Tdbot, & Yin, 1977). Given the mnemonic aspect of de￾layed-response functions, one mi^ t presume that the principal sulcus shotild interact with die hippocampus—the m^or subcortical structure that is crucial for certain forms of memory (Gohen, 1984; Squires & Butters, 1984). In support, multiple direct and indirect connections connect the principal sulcus with the hippocampus, and these connections un￾doubtedly play a m£^or role in assessing memories from long-term storage for use in the task at hand (Fig. 2C; Goldman-iUkic, Selemon, & Schwartz, 1984). The anatomy of motor control by which principal sulcus neurons participate in the selection or inhibi￾tion of responses has also been examined and worked out in some detail (Fig. 2D). This anatomy includes projections from the jainci￾pal sulcus to the caudate nucleus (Goidman & Nauta, 1977a; Selemon & Goldman-Rakic, 1985; Yeterian & Van Hoesen, 1978), connec￾tions with the motor thalamus (Goldman￾Rakic & Porrino, 1985; Ilinsky, Jouandet, & Goldman-Rakic, 1985), and with the deep "motor" layers of the superior colliculus (Fries, 1984; Goldman & Nauta, 1976). Thus, all circuitry exists for the components of de￾layed-response performance—visuospatial in￾put (pariet^-prefrontal connections), storage/ recall mechanisms (prefrontal-hippocMnpal circuits), and motor commands (preftontal connections with motor structures). Columnar and laminar organisation of the prifuHpal sulcus.—The microarchitecture of the principal sulcus has also been studied to some degree, and several studies have re￾vealed that diis cortex has a modular type of organization. For example, fibers originating in the parietal cortex (so-celled associational fibers) terminate in the principal sulcus as dis￾tinct vertically oriented columns about one￾half millimeter in width (Goldman & Nauta, 1977b). These columns altemate with equal￾sized colunms of fibers from the contralateral principal sulcus (so-called callosal pro￾jections) (Fig. 3, Goldman-Rakic & Schwartz, 1982). The cortex, as has been mentioned, is organized into six layers, and both callosal

608 Child Development and associational fibers tend to terminate review,see Rakic Goldman-Rakic,1984) Ou focus has been on of axon eigf detal) y r yer areas that comprise the sys nce of the phic methods for studvine The sign ayering odoropetonted re t ml ets of inco of afferents have y in the dis ctive proje it w iographic ca vas the ava n in fig mad layer III pro fetal monkeys ore this metho eoiwgfst ng br ons for ause the e ca te silver) ade m 100 sbcortical structures;neu sin layer VE pp to de ping anim ect selectively to the tha (Fig. vival time aptcachihe ure and ngortiea d ser nt and system (Leo a 1973 lity of the Finally. the re because one could ation of con for ex aseat adult function ca It is wChodtdsooveredwihsihwerimpregna6e out that By cont radio phic tech sulcus an I that ga ioactive isot nt res ini cted size of re at the site begin the ds o th t6 irth or at any other arbi品 eins ancy,the processes and he proteins are rion of the to dg of cell bncamRoebe,ecd ceden logy,the precise survival times optimal for t in my labdy O of ten ina field liest embryor es that i dist on of target 6caeial r ev mpts tha s by whi the axo ns reach their targ e pla 6 at or a birth in smal rlier in the the brain rk:the Sodman-Rakic 1982).It here to develop nniques for 小12 pre nt any giv

608 Child Development and associational fibers tend to terminate selectively in layers IV and I, while associa￾tional fibers additionally terminate in layer VI (see Fig. 3 for details). The cortex also has a radial or horizontal geometry due to its distinct six-layered or￾ganization. The significance of the layering pattem is that cells in the specific layers that are targets of incoming callosal or associa￾tional afferents have distinctive projection targets in subcortical and/or other cortical structures. For example, as shown in Figure 3, pyramidal neurons in layer III project primarily to other cortical areas, bodi widiin the same (associational neurons) and opposite (callosal neurons) hemispheres. While over 80% of callosal and associational neurons originate in layer III, layer V is the msyor source of projections to the caudate nucleus and putamen, to the colliculus, and to other subcordcal structures; neurons in layer VI project selectively to the thalamus (Fig. 3). Knowing these facts, one can begin to ap￾preciate the synaptic architecture and wiring diagrams underlying a specific cortical func￾tion. Prenatal Development of Prefrontal Connections Background for prenatal studies.—It is reasonable to assume that adult function can emerge only if the proper connections and circuitry necessary to carry out that fimction are established. The principal sulcus and its interconnections provide a natural model sys￾tem for the study of the biological basis of die particular type of cognitive capacity indexed by delayed-response tasks, namely, the guid￾ance of behavior by stored representations. Although in theory it is possible to begin the developmental study of structure-function re￾lations at birth or at any other arbitrary point in infancy, the processes and mechanisms of development are complex, onierly, and syn￾chronized and can best be appreciated only by comprehensive analysis of antecedent causes. Accordingly, the study of prefrontal cortex development in my laboratory includes analysis of the earliest embryonic ages that it is feasible to study in primates. It is important to realize that the critical cellular events that take place at or after birth in small mammals, for example, the rat, occur much earlier in the primate, well before birth (for fuller discus￾sion of these events, see Goldman-Rakic, 1986; Rakic & Goldman-Rakic, 1982). It was therefore necessary to develop techniques for prenatal surgery and prenatal intervention that could be applied to the primate brain (for review, see Rakic & Goldman-Rakic, 1984). Our focus has been on the outgrowdi of axons and the establishment of synaptic connections between the key areas that comprise the sys￾tem underlying delayed response. Autoradiographic methods for studying prefrontal connections.—Very few studies have ever been conducted on the develop￾ment of connectivity in the primate brain, and it was the availability of the autoradiographic method in conjunction with prenatal surgery that made it feasible to trace connections in fetal monkeys. Before this method, connec￾tions were traced by performing rather large lesions of structures, and after an appropriate survival period, examining brain sections for degenerating axons and terminals (that could be visualized because they selectively im￾pregnate silver). This mediod had several drawbacks that made it particularly difficult to apply to developing animals. The optimal sur￾vival time for visualizing the products of de￾generation had to be worked out empirically and separately for every age point and neural system (Leonard, 1973). The reliability of the silver impregnation varied with investigators. Finally, the resolution of connections was in￾trinsically poor because one could describe projections only between large areas. Golum￾nar distribution of connections, for example, was not discovered with silver-impregnation methods. By contrast, the autoradiographic tech￾nique was a highly reliable method that gave clear and consistent results. Small quantities of radioactive isotopes are injected into desig￾nated areas; the size of the area injected can be controlled by the volume of isotope in￾jected. Neurons at the site of injection absorb die mdiolabeled amino acids into their cell bodies. In the cell body, the amino acids are converted into proteins by normal protein synthesis, and these proteins are transported from the cell dirough its axon to the terminal region of die cell. From knowledge of cell biology, the precise survival times optimal for transport to occur can be accurately predicted to optimize visualization of terminal fields (the distribution of labeled axons in target stmctures) and for determination of die path￾ways by which the axons reach their targets. Sections cut throu^ the brain are mounted on slides that are dipped in photographic emulsion in the dark; the emulsion is exposed only by the beta emissions radiating from labeled cells, axons, and terminals that may be present within any given section. After 8— 12 weeks, the slides are developed and the exposed areas appear as darkened grains

PRINCIPAL SULCUS CONTRA IPSI CONTRA IPS b WM CORPUS CALLOSUM▲ CORTICO-THALAMIC of the a ps in 、of mod in th mitive white er

PRINCIPAL SULCUS CONTRA IP S1 CONTRA IP S1 CORTICAL-STRIATAL CQRTICQ-THALAMIC FIG. 3.—Simplified diagram of the modular architecture of the principal sulcus with particular refer￾ence to the location of callosal (black triangles) and associational (white) neurons that give rise to axons destined for cortical areas in die opposite (contralateral) and same (ipsilateral) hemisphere, respectively; stippling represents the tenninals of callosal (darker) and associational (lighter) reciprocal prqjectiGns to tiie principal sulcus. The major points illustrated are: (a) both types of corticocortical neurons reside in the same layers of cortex (III and V); (b) columns of same-hemisphere inputs altemate with columns of opposite-hemisphere inputs; and (c) columns of high-density callosal neurons receive callosal inputs; columns with hi^-density associational neurons receive associational inputs. Thus we may view the prefrontal cortex as made up of modules that are devoted primarily to eidier intrahemispheric or in￾terhemisiAeric processing. Integration of the information processed in the two types of modules is inte￾grated at another level and undoubtedly underlies the integration of the two hemispheres. The precise organization of die cortico-cortical connections in die principal sulcus encourages belief that cognitive fimction can ultimately be understood at a synaptic and molecular level. Roman numerals indicate layers; WM, white matter

610 Child Development In surgery ing fror 169 em (E69)to the average gestat 165 es were er similar to postr be visual zed as on the the spective princ sulcus the fetal tis sufficient tim Dat nves- These aut adiog the rth of cortico callos eric)and prisingly.pe on on opmen av ons and wide limits can be d d from analysis of stissue prepar ith con ed the mode of deve ent of the co de d his a h ma o the bir ay ergntweeo977 Ral Tho of the feus 68,1 present in the s (Fig A z the callosum fr the in alread

610 Child Development above the structure in which radioactivity was present (for examples, see Goldman & Nauta, 1977a, 1977b; Goldman-Rakic, 1981a). Development of prefrontal efferents.— In die studies mentioned, prenatal surgery was performed on timed-pregnant rhesus monkeys carrying fetuses ranging from 69 em￾bryonic days (E69) to 152 embryonic days (E152) and on several newbom monkeys. In diis species, the average gestation lasts 165 days. The fetuses were exposed dirough an incision in the uterus and were operated on in a manner similar to postnatal neurosurgeiy. Although the primary fissures do not develop fully until the last third of gestation, the prin￾cipal sulcus can be visualized as a small in￾dentation on the dorsoktera] surface of the frontal lobe; in younger fetuses, the iiyection was placed in die center of the dorsolateral part of the lobe. After microinjections of radioactive amino acids were made in the pro￾spective principal sulcus, the fetal tissues were sutured and the fetus returned to the uterus, which was also sutured. The fetuses survived 1-3 days, sufficient time for trans￾port to occur in die padiways under inves￾tigation. These autoradiographic Investigations were designed to provide basic information on timetables for the outgrowth of cortico￾cortical (callosal and intrahemispheric) and selected cortico-subcortical connections. Sur￾prisingly, perhaps, precise information on the development of connectivity was not avail￾able for any species, since only gross approxi￾mations and wide limits can be deduced from analysis of nervous tissue prepared with con￾ventional stains. Another basic question con￾cemed the mode of development of the co￾lumnar pattem of callosal innervation characteristic of the adult pathways. We wondered whether diis input was initially diffusely distributed in the cortex before be￾coming segregated into distinct colmnnar territories in a manner similar to the bii^asic development of afferent territories in the vi￾sual cortex (Hubel & Wiesel, 1977; Rakic, 1976). Thorough examin^on of the youngest fetus injected, E69, provided no evidence that callosal or intrahem^pheric axons were yet present in the vicinity of their target struc￾tures (Fig. 4). Although labeled axons could be seen entering the callosum from the in￾jected hemisphere, none reached and were present in the opposite hemisphere at this age. In contrast, labeled fibers were already evident in the caudate nucleus, which is part of the basal ganglia and also in die thalamus E69-E70 P4-PII FIG. 4.—Gallosal fibers at different embryonic (£) and postnatal (P) ages. The first number be￾neatii each drawii^ indicates the age in days when Uie animal was injected witii radioactive tracers, and the second number of tihe pair indicates the ^ e in days when it was sacrificed. Callosal fibers do not reach the contralateral hemisphere until after E70 and do not enter the cortica] plate in significwit numbers before E123. Callosaj axons invade the cortex between E123 and E152. At around E135 (not shown) and E152, they are already well seg￾regated into the modular pattem chaiacteristic of the neocortex in the postnatal monkey (from Gold￾man-Rakic, 1981b)