© 1992 SCIENTIFIC AMERICAN, INC

Working Memory and the Mind

Anatomic and physiological studies of monkeys are locating the neural
machinery involved in forming and updating internal representations of the
outside world. Such representations form a cornerstone of the rational mind

by Patricia S. Goldman-Rakic

T
he seeming simplicity of everyday life tive memory acquires facts and figures and
belies the enormously complex ongoing holds them in long-term storage. That knowl­
operations of the mind. Even routine edge is of no use, however, unless it can be ac­
tasks such as carrying on a conversation or cessed and brought to mind in order to in­
driving to work draw on a mixture of current fluence current behavior.
sensory data and stored knowledge that has Working memory complements associative
suddenly become relevant. The combination memory by prOviding for the short-term acti­
of moment-to-moment awareness and instant vation and storage of symbolic information,
retrieval of archived information constitutes as well as by permitting the manipulation of
what is called the working memory, perhaps that information. A simple activity involving
the most Significant achievement of human working memory is the carry-over operation
mental evolution. It enables humans to plan in mental arithmetic, which requires tempo­
for the future and to string together thoughts and ideas, rarily storing a string of numbers and holding the sum of
which has prompted Marcel Just and Patricia Carpenter of one addition in mind while calculating the next. More com­
Carnegie Mellon University to refer to working memory as plex examples include planning a chess move or construct­
"the blackboard of the mind." ing a sentence. Working memory in humans is considered
Until recently, the fundamental processes involved in such fundamental to language comprehension, to learning and to
higher mental functions defied description in the mechanis­ reason.
tic terms of science. Indeed, for the greater part of this cen­

N
tury, neurobiologists often denied that such functions were umerous lines of evidence indicate that the opera­
accessible to scientific analysis or declared that they be­ tions of working memory are carried out in a part of
longed strictly to the domain of psychology and philosophy. the brain known as the prefrontal lobes of the cere­
Within the past two decades, however, neuroscientists have bral cortex. (Cortex derives from the Latin word meaning
made great advances in understanding the relation between bark; the cerebral cortex consists of an outer rind of so-called
cognitive processes and the anatomic organization of the gray matter neurons surrounding the cerebrum.) Much of the
brain. As a consequence, even global mental attributes such evidence identifying this structure as the center for working
as thought and intentionality can now be meaningfully stud­ memory comes from observations of the effects of injuries
ied in the laboratory. to the prefrontal part of the hemispheres. For example, pa­
The ultimate goal of that work is extraordinarily ambi­ tients having frontal lobe damage exhibit gross deficiencies
tious. Eventually researchers such as myself hope to be able in how they use knowledge to guide their behavior in every­
to analyze higher mental functions in terms of the coordi­ day situations. Nevertheless, they often retain a full store of
nated activation of neurons in various structures in the information and may continue to score well on conventional
brain. It should also be possible to identify the cells that tests of intelligence.
mediate the activity of those structures. Such research will Although most fully developed in humans, some elements
help explain the origin of mind. It may also lead to more of working memory exist in other animals, especially in oth­
complete descriptions of baffling mental disorders such as er primates; if their prefrontal cortices are damaged, those
schizophrenia. animals develop symptoms much like the ones seen in hu­
For many years, insight into the operation of the brain was mans. Neuroscientists have therefore turned to monkeys
stymied by the misconception that memory is a single entity in their efforts to explore the nature of working memory.
that could be traced to a single structure or location. Since
the 1950s, neuroscientists have increasingly come to appre­
ciate that memory consists of multiple components con­
PATRlCIA S. GOLDMAN-RAKlC has devoted her academic ca­
structed around a distributed network of neurons. Accord­ reer to studying the neurobiology of memory and cognition. She
ing to present thinking, a form of memory known as associa- received a Ph.D. from the University of California, Los Angeles,
in 1963. Two years later she joined the Intramural Research Pro­
gram of the National Institute of Mental Health. In 1979 she
moved to the Yale University School of Medicine, where she is a
WORKING MEMORY enables a human to retrieve stored sym­
professor of neuroscience. Goldman-Rakic sits on several national
bolic information, such as the bowings and fingerings of a
advisory boards and is a member of the National Academy of
memorized piece of music, and to translate that information Sciences. She recently served as president of the Society for Neu­
into a controlled set of motor activities. Studies of similar but roscience. Her current research focuses on identifying the neural
simpler information processing performed by primates is re­ mechanisms that carry out higher cortical functions in primates.
vealing the structure of working memory.

SCIENTIFIC AMERlCAN September 1992 III

© 1992 SCIENTIFIC AMERICAN, INC
 WORKING MEMORY TASK ASSOCIATIVE MEMORY TASK animal receives a reward, usually food
or juice.
Delayed-response tests tap working
memory processes because the animal
must retain the memory of the location
of the stimulus during the period of
the delay. The proper response at the
CUE CUE AND end of the delay is indicated not by ex-
RESPONSE
ternal stimuli but by the memory of
what the subject saw on the previous
trial. Furthermore, the correct response
may differ from one trial to the next, de­
pending on new information presented
to the subject in each trial. Correct re­
sponses in working memory tasks, as in
human affairs, are guided by memory
rather than by immediate sensory infor­
mation, and they depend on constant
\!)"" ;.e;""""",,,,,,,,,, updating of the relevant information.

D
elayed-response tests resemble
very closely the object-perma­
nence task, developed in the
DELAY DELAY early part of this century by the French
child psychologist Jean Piaget, that is
widely used to chart the cognitive de­
---- - - - .---- - -- _ . _ - - - - - - - - - --- velopment of young children. For Pia­
get's task, a child is shown two boxes,
one of which contains a toy. The boxes
are then closed. After a brief wait, dur­
ing which the child is purposely dis­
tracted, the child is asked to pick which
box contains the toy. Once the child
gives several consecutive correct re­
sponses, the toy is switched into the
other box while the child watches. The
experimenter then continues the test to
find out whether the child will change
his or her response in accord with the
RESPONSE CUE AND updated information.
RESPONSE A series of studies has demonstrated
that performance on the object-perma­
nence task, like the ability to conduct
delayed-response activities, depends on
the degree of maturity of the subject's
prefrontal cortex. Human infants less
than about eight months old (whose cor­
tices have not yet acquired adult cir­
WRONG
cuitry) perform poorly on these tasks,
as do monkeys whose prefrontal re­
MEMORY TASKS help to assess the workings of the mind. In the classical working
gions have been surgically ablated. In
memory task (left), a monkey briefly views a target stimulus-in this case, a morsel
both cases, the subjects' responses are
of food. Only after a delay is the animal allowed to retrieve the food. The experi­
menter randomly varies the location of the food between trials, so that each re­ guided by habit and by reflex rather
sponse tests only the animal's short-term retention of visual and spatial informa­ than by representational principles. In­
tion. An associative memory task (right), in contrast, follows a consistent pattern fants and brain-injured monkeys tend
throughout. Here a plus sign always indicates the correct response. The task there­ to repeat the response that previously
fore measures the animal's ability to retain long-term rules. was reinforced (for example, choosing
the box on the right even after they have
seen that the toy was transferred to the
Such exploration has been aided by the on information immediately present in box on the left) rather than change their
design of repeatable tests of working the environment. In the prototypical de­ response to agree with newly presented
memory functions. layed-response test, an animal receives information. Both humans and monkeys
Working memory is being assessed a brief visual or auditory stimulus that act as if "out of sight" is "out of mind."
in monkeys by means of tasks known is then hidden or taken away. After a Such behavior implies that the mech­
as delayed-response tests, which evalu­ delay of several seconds, the animal is anism for guiding behavior by repre­
ate an organism's ability to react to sit­ given a signal that tells it to respond to sentational knowledge is destroyed in
uations on the basis of stored or in­ the location where the stimulus had ap­ monkeys having prefrontal lesions and
ternalized representations rather than peared. If the response is correct, the not yet developed in human infants. In

1 12 SCIENTIFIC AMERICAN September 1992

© 1992 SCIENTIFIC AMERICAN, INC
support of that notion, I, along with Joaquin M. Fuster of the University of end of a delay of three to six seconds,
Jean-Pierre Bourgeois and Pasko Rakic, California at Los Angeles, along with Ki­ the central light, or fixation spot, switch­
also at Yale University, have examined sou Kubota and Hiroaki Niki of the Kyo­ es off, instructing the animal to move its
the rate at which neural connections to Primate Center in Japan, performed eyes to the location where the stimulus
form in the prefrontal cortices of juve­ the first experiments of how individual was seen before the delay. If the re­
nile monkeys. neurons behave in the prefrontal cortex. sponse is correct, the animal is reward­
The time of most rapid synapse for­ The researchers introduced fine elec­ ed with a Sip of grape juice. Because the
mation in the animals' prefrontal region trodes into the prefrontal cortices of animal's gaze is locked onto the fixation
occurs when the animals are roughly monkeys trained to perform simple de­ spot, each stimulus activates a speCific
two to four months old, the same age at layed-response tasks and then recorded set of retinal cells. Those cells, in turn,
which the monkeys acquire the capaci­ the animals' neuronal activity in relation trigger only a certain subset of the vi­
ty to perform delayed-response tasks. to the events in the task. Those studies sual pathways in the brain.
The concept that an object exists con­ revealed a range of responses among Using the eye-movement experiment,
tinuously in space and time even when the neurons in the prefrontal cortex. we have demonstrated that certain neu­
out of view and, more generally, the Some cells showed heightened electri­ rons in the prefrontal cortex possess
ability to form abstract concepts may cal activity when information was pre­ what we call "memory fields": when a
depend on a fundamental capacity to sented, whereas others became active particular target disappears from view,
store representations of the outside during the delay period, when the ani­ an individual prefrontal neuron switch­
world and to respond to those repre­ mals were remembering the informa­ es into an active state, producing elec­
sentations even when the real objects tion. A third set of neurons responded trical signals at more than twice the
are not present. most strongly when the animals began baseline rate. The neuron remains acti­
The studies described above raised their motor response. vated until the end of the delay peri­
the enticing possibility of identifying At Yale, Shintaro Funahashi, Charles ]. od, when the animal delivers its re­
more precisely the brain structures as­ Bruce and I have used the single-neuron sponse. A given neuron appears always
sociated with delayed-response activi­ technique in conjunction with a delayed­ to code the same visual location. For
ties and representational memory. Much response experiment that tests spatial example, some neurons fire only if the
of the progress toward that goal has memory. For our experiment, a monkey stimulus appears at the nine o'clock
derived from experiments that monitor is trained to fix its gaze on a small spot position on the television screen; the
electrical activity in �ingle neurons in in the center of a television screen. A vi­ cell does not respond to visual stimuli
monkeys' prefrontal cortices while the sual stimulus, typically a small square, that appear elsewhere in the monkey's
animals perform tasks that depend on appears briefly in one of eight locations visual field. Other neurons code for oth­
speCific delayed-response skills. on the screen and then vanishes. At the er target locations in working memory.

TIME 1 TIME 2
r
TIME 3
r---'
, ,
1 ,
L __ ..J

x x

CUE DELAY RESPONSE

DELAYED-RESPONSE TASK has been used to study working off, the animal moves its eyes to look where the target ap­
memory in monkeys. While a monkey fixes its gaze on a cen­ peared (right). Measurements of electrical activity show that
tral spot, a target flashes on the screen (left), then vanishes. certain neurons in the prefrontal cortex react to the appear­
During a delay of several seconds, the monkey keeps a mem­ ance of the target, others hold the memory of it in mind and
ory of the spot "in mind" (center). When the central spot turns still others fire in preparation for a motor response.

SCIENTIFIC AMERICAN September 1992 113

© 1992 SCIENTIFIC AMERICAN, INC
NEURONAL CIRCUITRY connects the prefrontal cortex is part of an elaborate
prefrontal cortex to the sensory, limbic network of reciprocal connections be­
and motor systems in a monkey brain tween the principal sulcus and the ma­
(top). Anatomic studies show that neu· jor sensory, limbic and premotor areas
ral projections from the parietal lobe to
of the cerebral cortex. That particular
the prefrontal cortex exhibit a modular
network seems to be dedicated to spa­
pattern, as seen in this frontal cross sec·
tial information processing. The net­
tion (middle). Radioactive tracers reveal
the metabolic activity in a frontal cross work's structure probably follows the
section of the brain of a monkey per­ same baSic plan as do other similarly
forming a delayed-response task (bot­ organized networks that draw on mul­
tom). The distribution of activity close­ tiple parts of the brain and are dedicat­
ly conforms to the anatomic links. ed to other cognitive functions-object
recognition, language production and
comprehension, and mathematical rea­
soning, for example.
The neurons capable of retaining the As previously noted, delay-response
visual and spatial coordinates of a stim­ experiments demonstrate that neurons
ulus (in other words, of keeping its loca­ in the principal sulcus are sensitive to
tion "in mind" after it vanishes) appear the speCific location of visual stimuli.
to be organized together within a spe­ Those neurons must therefore have ac­
cific area of the prefrontal cortex. These cess to visual and spatial information
neurons collectively fonn the core of the originating elsewhere in the brain. The
spatial working memory system. If the principal sulcus does in fact receive sig­
activity of one or more of these neurons nals from the posterior parietal cortex,
falters during the delay period-if the where the brain processes spatial vi­
animal is distracted, for example-the sion. Clinical studies have documented
animal will probably make an error. that damage to the parietal cortex in
humans causes spatial neglect, a loss of

T
he activation of prefrontal neu­ awareness of the body and its relation
rons during the delay period of a to objects in the outside world.
delayed-response task depends Given that working memory depends
neither on the presence of an external on accessing and bringing to mind in­
stimulus nor on the execution of a re­ formation that is stored in long-term
sponse. Rather the neural activity corre­ memory, one might presume that the
sponds to a mental event interposed principal sulcus also interacts with the
between the stimulus and the response. hippocampus, the neuronal structure
Monkeys whose prefrontal cortices have that controls associative, or learned,
been damaged have no difficulty in memory. Researchers have used radio­
mOving their eyes to a visible target or active amino acids to trace direct con­
in reaching for a desired object, but they nections between the principal sulcus
cannot direct those motor responses by and the hippocampus.
remembering targets and objects that My colleague Harriet Friedman, also
are no longer in evidence. at Yale, and I have used a remarkable
Because the prefrontal cortex func­ technique known as autoradiography
tions as an intermediary between mem­ to measure brain metabolism. Our work
ory and action, one can imagine that shows that the hippocampus and the
damage to the prefrontal cortex could principal sulcal areas of the cortex are
spare knowledge about the outside often simultaneously active during de­
world yet destroy the organism's ability layed-response tests. My co-workers and
to bring that stored knowledge to mind I think t�at the primary role of the hip­
and to utilize it. Indeed, monkeys whose pocampus is to consolidate new as­
prefrontal cortices have been damaged, with major sensory and motor control sociations, whereas the prefrontal cor­
as well as many humans with similar centers. Various researchers have found tex is necessary for retrieving the prod­
injuries, exhibit no difficulty learning that the part of the cortex near the ucts of such associative learning (facts,
sensory-discrimination tasks. All forms princigal sulcus, a large groove in the events, rules) from long-term storage
of associative, or long-term, learning prefrontal cortex, is critical for the vi­ elsewhere in the brain for use in the
are preserved as long as the subject sual and spatial working memory func­ task at hand.
can still find the familiar environmen­ tions. I have focused my research on A particularly useful version of auto­
tal stimuli associated with certain con­ this particular region in the belief that radiography, called the 2-deoxyglucose
sequences and expectations [see "The an in-depth neurobiological analysis of method, has made it possible to ob­
Biological Basis of Learning and Indi­ one major subdivision of the prefrontal serve directly which parts of the brain
viduality," by Eric R. Kandel and Robert cortex could serve as a starting point are activiated during specific tasks. In
D. Hawkins, page 78]. for analysis of the other subdivisions this technique, developed by Louis Sok­
Over the past decade, improved tech­ of the brain and help lead the way to oloff of the National Institute of Men­
niques for investigating the anatomy of development of a unified theory of the tal Health, animals are injected with the
the brain have provided for the first function of the entire prefrontal cortex. compound 2-deoxyglucose, a molecule
time an accurate and detailed picture Studies of direct and indirect neuro­ that appears chemically identical to glu­
of how the prefrontal cortex connects nal linkages in the brain reveal that the cose, the sugar that cells consume to

114 SCIENTIFIC AMERICAN September 1992

© 1992 SCIENTIFIC AMERICAN, INC
provide energy. The more active a cell increasingly being used to monitor ac­ asked them to perform tasks that de­
is, the more 2·deoxyglucose it takes in. tivation patterns in the human brain pended on comparing current auditory
Unlike normal glucose, however, 2-de­ and to identify which neurons are en­ stimuli with recently presented ones in
oxyglucose cannot be broken down by gaged during specific mental tasks. One, order to detect whether they are the
metabolic activity, so it accumulates in known as positron emission tomogra­ same or different. Frontal lobe patients
the cell. Sokoloff uses a radioactive ver­ phy (PET ), resembles autoradiography displayed patterns of electrical activity
sion of the compound. The concentra­ in that the subject takes in a radioac­ quite unlike those of healthy subjects
tion of radioactivity in each part of the tive compound that exposes changes performing the same task, suggesting
brain is therefore directly proportional in blood flow to a given region of the that the patients do not store recent in­
to how active the cells there have been. brain, indirectly displaying that region's formation in memory the same way as
For our studies, a monkey trained to degree of metabolic activity. Another do normal people.
perform the delayed-response tasks re­ way to record human brain activity is In one study, subjects were exposed
ceives an intravenous injection of ra­ to measure the changing electrical po­ to steady patterns of low and high
dioactive 2 -deoxyglucose. Immediately tentials on the scalp in response to tones and occasional, unexpected audi­
after completing the task, the animal is controlled sensory stimulation, a pro­ tory stimuli. Healthy people developed
sacrificed and its brain is dissected into cedure called electroencephalography positive electrical potentials on their
thin slices that are placed on photo­ (EEG). Neither PET scans nor EEGs can cortices within one third of a second of
graphic film. Radioactivity darkens the provide anything close to the resolution hearing the anomalous sound. Patients
film, so each exposure serves as a snap­ possible in 2 -deoxyglucose studies in who had lesions in their prefrontal
shot of the activity of the cells in one animals, but they are invaluable tools cortices showed no such response, al­
particular slice of the brain. for monitoring the human brain during though they reacted normally to the fa­
My colleagues and I have found that mental activity. miliar background tones. These data
the prefrontal cortex, as well as many of A series of PET studies at Hammer­ are consistent with the notion that the
the areas with which it is connected (for smith Hospital in London and at Wash­ prefrontal cortex temporarily stores in­
example, the hippocampus, the bottom ington University examined subjects formation against which current stim­
portion of the parietal cortex and the performing tasks that required them to uli are judged.
thalamus), exhibits a high level of meta­ keep a mental record of recently pre­ The ultimate function of the neurons
bolic activity during delayed-response sented lists of words. Another PET ex­ in the prefrontal cortex is to excite or in­
performance. The same areas are no­ periment by the Washington University hibit activity in other parts of the brain.
tably less active when the monkey per­ group required subjects to generate an In this way, information processed in
forms associative memory tasks that appropriate verb to accompany a noun the principal sulcus can direct neurons
do not depend on short·term, rapid up­ flashed in front of them on a card. The in the motor centers that in turn car­
dating of information. participants in all three tests displayed ry out movements of the eyes, mouth,
heightened neuronal activity in the pre­ hands and other parts of the body.

T
hese results confirm anatomic frontal cortex while performing their Whole-brain studies tell only part of
studies of the connections be­ tasks, all of which engaged working the story; to understand the details of
tween the prefrontal cortex and memory. how signals pass to and from the pre­
other parts of the brain. More signifi­ In a complementary study, Robert frontal cortex, one must scrutinize the
cantly, they also reveal the degree to T. Knight of the University of Califor­ brain on a cellular scale.
which various parts of the brain are en­ nia at Davis looked at EEGs of patients When viewed through a conventional
gaged in certain discrete memory whose frontal lobes were injured. He microscope, the cerebral cortex appears
tasks. The studies also hint at how the
prefrontal cortex organizes the many
different kinds of information that
must flood through it. In fact, patterns
of brain activity appear distinctly differ­
ent depending on whether the task calls
up memories of location or of at­
tributes of objects.
I think the prefrontal cortex is divid­
ed into multiple memory domains, each
specialized for encoding a different kind
of information, such as the location
of objects, the features of objects (col­
or, size and shape) and additionally,
in humans, semantic and mathematical
knowledge. Recently Fraser Wilson and
James Skelly in my laboratory at Yale
have begun to define an area below
the prinCipal sulcus in monkeys where
neurons respond preferentially to com­
plex attributes of objects rather than to
their locations. They have found neu­
rons there that increase their rate of PYRAMIDAL NEURON (left) in the prefrontal cortex is thought to modulate signals
firing when a monkey is remembering to and from the prefrontal cortex. Each such neuron is covered with thousands of
a red circle but not when calling up a spines, bulblike projections where synaptic connections occur. Synapses have differ·
memory of a green square, for example. ent morphologies depending on whether they are excitatory or inhibitory (right).
Noninvasive imaging techniques are The dopamine-containing connections in the cortex are of the inhibitory type.

SCIENTIFIC AMERICAN September 1992 115

© 1992 SCIENTIFIC AMERICAN, INC
 RE C ALLED PROJECTION NEURON to be divided into six layers of vary­
VISUAL STIMULUS
ing cellular composition and denSity.
Cells in each layer form their own set
of connections within the brain. One
class of cell, which reSides in the fifth
layer of the cortex, projects to areas be­
yond the cortex, including the caudate
nucleus and putamen (which regulate a
variety of motor activities) and the su­
PRERESPONSE POST RESPONSE perior colliculus (which specifically pro­
SIGNAL SIGNAL cesses visual motor functions). Neurons
in the sixth layer of the cerebral cor­
tex project into the thalamus, through
which sensory inputs from the brain's
periphery travel to reach the cortex.

T
he prefrontal cortex probably can­
not independently trigger motor
responses. Nevertheless, it may
regulate motor behavior by initiating,
programming, facilitating and cancel­
ing commands to brain structures that
are more immediately involved in di­
recting muscular movement. Such COIn­
mands are transmitted via an elaborate
L AYER 6 set of chemical pathways in the brain.
Neuroscientists and biochemists around
the world have been racing to learn
more about these chemicals and how
they regulate the operation of the brain.
A number of researchers studying
rodent brainS, including Anne Marie
Thierry and Jacques Glowinski of the
College of France in Paris, Brigitte Ber­
INHIBITORY ger of Pitie Salp�triere Hospital, also in
Paris, and Tomas Hbkfelt of the Karo­
linska Institute in Sweden, along with
many colleagues, find that the prefron­
tal cortex abounds in catecholamines,
a family of compounds that prepare
SUBST A N TIA MEDIODORSAL the body for a stressful situation. Those
NIG RA T H AL AMUS compounds also act as neurotransmit­
ters, substances that transmit neuronal
INHIBITORY EXCITATORY impulses in the brain. My co-workers
and I have discovered a similar abun­
dance of catecholamines in the prefron­
+
tal cortices of nonhuman primates. One
of the most familiar catecholamines,
dopamine, regulates how neurons react
to stimuli and seems to play a central
SUPERIOR
COLLICULUS EYE MOVEMENT role in schizophrenia.
� A growing body of evidence suggests
EXCITATORY that dopamine is one of the most im­
portant of the chemicals that regulate
cell activity associated with working
memory. An imbalance in the abun­
dance of dopamine in the prefrontal
I cortex can induce deficits in the work­
ing memory similar to those resulting
from lesions in the prinCipal sulcus
region of the prefrontal cortex. For ex­
ELABORATE FLOW OF NEURAL SIGNALS is involved in producing a memory-guid­
ample, aged monkeys whose prefrontal
ed eye movement. A neuron in the fifth layer of the prefrontal cerebral cortex trans­
cortices are deficient in dopamine and
mits signals along a chain of neurons in the striatum, the substantia nigra and the su­
perior colliculus, where they trigger motor response in the eyes. Impulses from the norepinephrine (a chemical relative of
substantia nigra travel to the mediodorsal thalamus and back to the cortex, indicat­ adrenaline) perform poorly in delayed­
ing the completion of the motor response and signaling the prefrontal neuron to re­ response tests. Injecting the aged ani­
turn to a baseline level of activity. The graphs show the electrical activity of the neu­ mals with the deficient neurotransmit­
rons; inverted triangles indicate the nearly instantaneous travel of the signals. ters restored their memory function

116 SCIENTIFIC AMERICAN September 1992

© 1992 SCIENTIFIC AMERICAN, INC
so that they tested roughly as well as analysis of the physical and chemical Harvard University have shown that
younger, healthy monkeys. interactions between pyramidal cells schizophrenic subjects exhibit impaired
Many of my colleagues and I are striv­ and other neurons in the cerebral cor­ performance on working memory tasks
ing to learn which cells respond to tex should clarify how dopamine and much like those my colleagues and I
dopamine and how they affect working other neurotransmitters influence cog­ have used to study working memory
memory. Within the past several years, nition by stimulating or repressing the in rhesus monkeys. Conversely, Martha
we have collected evidence showing that cellular responses of cortical neurons. MacAvoy and Bruce of Yale have dem­
neurons in certain layers of the cere­ onstrated that monkeys with lesions in

I
bral cortex contain a great abundance nvestigations of the workings of the relevant portions of the prefrontal
of D) receptors, one of the chemical the prefrontal cortex are revealing cortex exhibit the same type of predic­
sites where dopamine binds to a cell. In­ not only how the mind operates tive tracking disorder that has long been
terestingly, the neurons that are rich in but also what goes wrong when it mal­ considered a marker of schizophrenia
D1 receptors are those that project to functions. Medical researchers have im­ in humans.
the thalamus, the brain structure that plicated dysfunction of the prefrontal Perhaps researchers should begin to
relays information to the cortex. cortex as the cause of many neurolog­ think of schizophrenia as a breakdown
Csaba Leranth, John Smiley and F. ical and psychiatric disorders, includ­ in the processes by which representa­
Mark Williams of Yale are examining ing Parkinson's disease and especial­ tional knowledge governs behavior. In
the cellular structures that enable dopa­ ly schizophrenia. The abnormal mental my view, neural pathways in the pre­
mine to modulate responses to senso­ attributes associated with schizophre­ frontal cortex update inner models of
ry inputs in the cerebral cortex. The re­ nia strongly resemble those caused by reality to reflect changing environmen­
searchers use an antibody developed by physical damage to the prefrontal cor­ tal demands and incoming information.
Michel Geffard of the Institute of Cellu­ tex: thought disorders, reduced atten­ Those pathways guide short-term mem­
lar Biochemistry and of Neurochemistry tion span, inappropriate or flattened ory and moment-to-moment behavior.
of the National Center of Scientific Re­ emotional responses and lack of initia­ If they fail, the brain views the world
search in Bordeaux, France, to label the tive, plans and goals. Schizophrenic pa­ as a series of disconnected events, like
neurons and their axonal projections tients, like frontal lobe patients and a slide show, rather than as a continu­
that contain dopamine. They then scru­ monkeys afflicted with prefrontal lobe ous sequence, like a movie. The result
tinize those cells under an electron mi­ leSions, retain a normal ability to per­ is schizophrenic behavior, excessively
croscope. The team looked in particular form routine procedures or habits but dominated by immediate stimulation
at the points of contact between dopa­ exhibit fragmented, disorganized behav­ rather than by a balance of current, in­
mine-releasing cells and the neuronal ior when attempting to perform tasks ternal and past information.
spines, small protuberances where the involving symbolic or verbal informa­ At present, theories describing the
cells receive incoming signals. Spines tion [see "Major Disorders of Mind and fundamental causes of schizophrenia
are discrete sites where calcium ions Brain," by Elliot S. Gershon and Ronald are inadequate, much as knowledge of
can enter and activate cellular mecha­ o. Rieder, page 126). the functioning of the working memory
nisms involving information processing SchizophreniC patients taking tests system remains frustratingly sketchy.
and modulation of neuronal responses. such as the Wisconsin Card Sort test Fortunately, neurobiological research
In most cases, the dopamine-releasing tend to repeat a previous response has been advanCing at a breathless pace
cells make symmetric contact with the even when it is clear that it is no long­ in the past few years. Such research
spines-that is, the cell projections on er the correct one; normal subjects, in should lead to a greater understanding
either side of the synaptic cleft show contrast, shift hypotheses much soon­ not only of schizophrenia but of the
roughly the same density. Such symmet­ er after making an error. SchizophreniC prefrontal cortex and how it shapes
ric contacts are thought to have an in­ individuals are also severely impaired short-term memory and the broader
hibitory effect: when the postsynaptic both on spatial delayed-response tasks working of the rational mind.
site is activated, the cell's normal, spon­ and on a variety of tests of problem
taneous electrical activity is dampened. solving, abstraction and planning.
A large proportion of the spines of pyr­ Studies of cerebral blood flow by FURTHER READING
amidal cells-the major class of neuron David H. Ingvar of University Hospital
WORKING MEMORY. Alan Baddeley. Ox­
that projects out of the cortex-receive in Lund, Sweden, and by Daniel R. ford University Press, 1986.
asymmetric contacts from the axons of Weinberger, Karen F. Berman and oth­ CIRCUITRY OF PRIMATE PREFRONTAL COR­
another cell whose point of origin has ers at the National Institute of Men­ TEX AND REGUIATION OF BEHAVIOR BY
not yet been identified but which is tal Health, as well as measurements REPRESENTATIONAL MEMORY. P. S. Gold­
thought to carry signals from other of local cerebral metabolism made by man-Rakic in Handbook of Physiology,
cortical areas. Those asymmetric con­ Monte S. Buchsbaum of the University Section 1, Vol. 5: Higher Functions of
the Brain, Part 1. Edited by Fred Plum.
tacts probably have an opposite, excita­ of California at Irvine, show that schizo­
Bethesda, Md., American Physiological
tory effect. phrenic patients have below-average
SOCiety, 1987.
Pyramidal cells receive the major sen­ blood flow into their prefrontal cortic­ MNEMONIC CODING OF VISUAL SPACE
sory or informational signals arriving es, indicative of a depressed level of ac­ IN THE MONKEY's DORSOLATERAL PRE­
at the cerebral cortex. The network of tivity in that part of the brain. Schizo­ FRONTAL CORTEX. Shintaro Funahashi,
excitatory and inhibitory synapses, or phrenic subjects often suffer from im­ Charles ]. Bruce and Patricia S. Gold­
connections, noted by the Yale group paired ability to move their eyes to track man-Rakic in Journal of Neurophysiolo­
provides a mechanism by which dopa­ and project the forward trajectories of
gy, Vol. 6 1, No.2, February 1989.
PREFRONTAL CORTICAL DYSFUNCTION
mine could alter the way that various moving targets, further evidence that
IN SCHIZOPHRENIA: THE RELEVANCE OF
classes of pyramidal neurons respond the disorder involves malfunctions in a WORKING MEMORY. Patricia S. Goldman­
to integrate such signals across thou­ posterior part of the prefrontal cortex, Rakic in Psychopathology and the Brain.
sands of spines in their dendrites. In where the eye-movement centers in­ Edited by Bernard]. Carroll and James
this way, dopamine may regulate the volved in predictive tracking are located. E. Barrett. Raven Press, 199 1.
overall output of the cortex. Further Sohee Park and Philip S. Holzman of

SCIENTIFIC AMERICAN September 1992 117

© 1992 SCIENTIFIC AMERICAN, INC