The
Medial
Temporal
Lobe
Memory
System
LARRY
R.
SQUIRE
AND
STUART
ZOLA-MORGAN
Studies
of
human
amnesia
and
studies
of
an
animal
model
of
human
amnesia
in
the
monkey
have
identified
the
anatomical
components
of
the
brain
system
for
memory
in
the
medial
temporal
lobe
and
have
illuminated
its
function.
This
neural
system
consists
of
the
hippocampus
and
adjacent,
anatomically
related
cortex,
including
en-
torhinal,
perirhinal,
and
parahippocampal
cortices.
These
structures,
presumably
by
virtue
of
their
widespread
and
reciprocal
connections
with
neocortex,
are
essential
for
establishing
long-term
memory
for
facts
and
events
(de-
clarative
memory).
The
medial
temporal
lobe
memory
system
is
needed
to
bind
together
the
distributed
storage
sites
in
neocortex
that
represent
a
whole
memory.
How-
ever,
the
role
of
this
system
is
only
temporary.
As
time
passes
after
learning,
memory
stored
in
neocortex
gradu-
ally
becomes
independent
of
medial
temporal
lobe
struc-
tures.
IN
1957,
SCOVILLE
AND
MILNER
DESCRIBED
A
PROFOUND
and
selective
impairment
in
human
memory
after
bilateral
surgical
removal
of
the
medial
temporal
lobe
(1).
Comprehen-
sive
neuropsychological
evaluation
of
one
patient
from
that
series
(patient
H.M.)
established
the
fundamental
principle
that
the
ability
to
acquire
new
memories
is
a
distinct
cerebral
function,
separable
from
other
perceptual
and
cognitive
abilities
(2).
Memory
impair-
ment
has
also
been
linked
to
medial
temporal
lobe
damage
in
cases
of
viral
encephalitis
(3),
posterior
cerebral
artery
occlusion
(4),
and
Alzheimer's
disease
(5).
However,
the
medial
temporal
lobe
is
a
large
region
that
includes
the
hippocampal
formation
(6),
the
amygdaloid
complex,
and
adjacent
cortical
areas
(Fig.
1).
It
has
therefore
been
difficult
to
determine
from
human
cases
precisely
which
structures
and
connections
within
the
medial
temporal
lobe
are
important
for
memory.
Human
Memory:
Anatomical
Findings
Since
Patient
H.M.
Initially,
attention
was
drawn
to
the
hippocampal
region
(6)
because
patients
who
underwent
temporal
lobe
surgery
developed
memory
impairment
only
when
the
removal
extended
far
enough
posteriorly
to
include
the
hippocampus
and
the
parahippocampal
gyrus
(1,
7).
Several
single-case
reports
also
linked
memory
impair-
ment
to
hippocampal
lesions
(8).
However,
in
these
cases
memory
functions
were
usually
assessed
informally,
and
the
damage
in
most
cases
was
not
confined
to
the
hippocampus.
The
picture
was
also
complicated
for
a
time
by
findings
from
behavioral
studies
of
rats,
monkeys,
and
other
animals
with
hippo-
campal
lesions,
which
were
initiated
after
patient
H.M.
was
first
described
but
which
could
not
be
easily
interpreted
in
terms
of
impaired
memory
(9).
In
1978,
two
alternatives
were
proposed
to
the
view
that
the
hippocampus
itself
is
critical
for
memory
functions.
Horel
proposed
that
memory
functions
were
disrupted,
not
by
hippocampal
damage,
but
by
damage
to
temporal
stem
white
matter,
which
lies
near
the
hippocampus
just
above
the
lateral
ventricle.
He
suggested
that
the
temporal
stem
would
necessarily
have
been
damaged
by
the
surgical
procedure
used
to
remove
the
medial
temporal
lobe
in
humans
and
monkeys
(10).
An
alternative
view
was
presented
by
Mishkin
on
the
basis
of
findings
from
monkeys
with
extensive
medial
temporal
lobe
ablations,
similar
to
Fig.
1.
(A)
Ventral
view
of
a
mon-
B
key
brain
showing
the
components
HIPPOCAMPUS
of
the
large
H+A+
lesion
that
first
established
an
animal
model
of
hu-
man
amnesia.
This
view
shows
the
ENTORHINAL
I
amygdala
(A)
and
hippocampus
CORTEX
(H),
both
in
crosshatching,
and
adjacent
cortical
regions
(+)
in-
cluded
in
the
H+A+
surgery.
Blue,
CORTEX
CAMPALRCORTEX
perirhinal
cortex
(areas
35
and
36);
yellow,
periamygdaloid
cortex
(area
51);
pink,
entorhinal
cortex
(area
UNIMODAL
AND
POLYMODAL
28);
green,
parahippocampal
cortex
ASSOCIATION
AREAS
(areas
TH
and
TF).
(B)
A
schemat-
(rnatmoaadprea
oe
ic
view
of
the
medial
temporal
lobe
memory
system.
The
entorhinal
cortex
is
the
major
source
of
projections
to
the
hippocampus.
Two-thirds
of
the
cortical
input
to
the
entorhinal
cortex
originates
in
the
adjacent
perirhinal
and
parahippocampal
cortices,
which
in
turn
receive
projections
from
unimodal
and
polymodal
areas
in
the
frontal,
temporal,
and
parietal
lobes.
The
entorhinal
cortex
also
receives
other
direct
inputs
from
orbital
frontal
cortex,
cingulate
cortex,
insular
cortex,
and
superior
temporal
gyrus.
All
these
projections
are
reciprocal.
SCIENCE,
VOL.
253
The
authors
are
at
the
Veterans
Affairs
Medical
Center,
San
Diego,
CA
92161
and
in
the
Department
of
Psychiatry,
University
of
California,
San
Diego,
La
Jolla,
CA
92039.
1380
on December 9, 2008 www.sciencemag.orgDownloaded from
the
surgical
lesion
sustained
by
patient
H.M.
He
proposed
that
conjoint
damage
to
both
the
hippocampus
and
amygdala
is
required
to
produce
severe
amnesia
in
monkeys
and
humans
(11).
This
proposal
was
consistent
with
the
fact
that
all
the
human
surgical
cases
with
amnesia
had
damage
to
both
the
hippocampus
and
amygdala
(1,
7).
In
1986,
a
role
for
the
hippocampus
itself
in
human
memory
was
established
from
a
case
of
amnesia
(12).
Patient
R.B.
developed
memory
impairment
in
1978
at
the
age
of
52,
after
an
episode
of
global
ischemia.
He
survived
for
5
years,
during
which
time
he
had
significant
memory
impairment
in
the
absence
of
other
cognitive
dysfunction.
Upon
his
death,
histological
examination
of
the
brain
revealed
a
circumscribed
bilateral
lesion
involving
the
entire
rostro-
caudal
extent
of
the
CA1
field
of
the
hippocampus.
Studies
of
animal
models
of
global
ischemia
in
the
rodent
have
confirmed
the
vulner-
ability
of
CA1
neurons
in
the
hippocampus
to
ischemia
and
implicated
a
mechanism
for
this
selective
lesion
that
involves
the
excitotoxic
effects
of
a
glutamate
neurotransmitter
(13).
Findings
from
patient
R.B.
allowed
two
conclusions.
First,
the
hippocampus
itself
appeared
to
be
a
critical
component
of
the
medial
temporal
lobe
memory
system.
This
idea
was
later
supported
by
an
additional
case
in
which
memory
impairment
was
associated
with
a
bilateral
lesion
confined
to
the
hippocampus
(14).
Second,
because
R.B.
was
not
as
severely
amnesic
as
H.M.
(15),
other
hippocampal
regions
in
addition
to
the
CAI
field,
or
structures
outside
the
hippocampus,
must
be
important
for
memory
functions.
Further
confirmation
for
these
ideas
has
come
from
improve-
ments
in
magnetic
resonance
imaging
(MRI),
which
make
it
possible
to
obtain
anatomical
information
from
living
patients
(Fig.
2,
A
and
B).
With
the
use
of
high-resolution
MRI,
the
hippocampal
region
(defined
in
those
studies
as
the
fimbria,
dentate
gyrus,
hippocampus
proper,
and
subiculum)
was
found
to
be
shrunken
and
atrophic
(57%
of
normal
size)
in
four
amnesic
patients
(16,
17).
The
temporal
lobe
was
of normal
size.
Because
these
patients
were
not
as
memory-impaired
as
patient
H.M.,
the
findings
with
MRI
reinforce
Fig.
2.
(A)
A
Ti-weighted
MR
coronal
image
from
a
healthy
normal
subject
(R.C.)
at
the
level
of
the
left
hippocampal
formation.
(B)
An
image
at
the
same
level
as
in
(A)
in
a
severely
amnesic
patient
of
the
same
age
(W.I.).
Note
the
markedly
shrunken
hippocampal
formation
in
the
patient.
(C)
A
coronal
image
of
a
monkey
brain
showing
three
CuSO4-filled
beads
(white
circles)
anchored
to
the
skull,
which
were
used
as
landmarks
to
establish
stereotaxic
coordinates
for
making
hippocampal
lesions.
(D)
An
image
of
the
same
monkey
as
in
(C)
after
the
hippocampal
lesion.
The
image
is
in
a
plane
perpendicular
to
the
long
axis
of
the
hippocampus.
[Reprinted
from
(17)
and
(47)
with
permission,
1990
Oxford
University
Press]
20
SEPTEMBER
1991
MEMORY
DECLARATIVE
(EXPLCIT)
NDECRATIVE
(MPC
FACTS
EVENTS
SILLS
PRIMING
SMPLE
NONASSOCIATIVE
AND
CLASICAL
LEARNING
HABETS
COND(TONING
Fig.
3.
Classification
of
memory.
Declarative
(explicit)
memory
refers
to
conscious
recollections
of
facts
and
events
and
depends
on
the
integrity
of
the
medial
temporal
lobe
(see
text).
Nondeclarative
(implicit)
memory
refers
to
a
collection
of
abilities
and
is
independent
of
the
medial
temporal
lobe
(60).
Nonassociative
learning
includes
habituation
and
sensitization.
In
the
case
of
nondeclarative
memory,
experience
alters
behavior
nonconsciously
without
providing
access
to
any
memory
content
(19,
20).
the
conclusion
that
structures
in
addition
to
the
hippocampus
itself
are
likely
to
be
important
for
memory
functions.
Development
of
the
Animal
Model
In
the
early
1980s,
an
animal
model
of
human
amnesia
became
available
in
the
nonhuman
primate
(11,
18).
The
animal
model
made
it
possible
to
investigate
systematically
which
anatomical
structures
are
important
for
memory.
The
research
strategy
is
straightforward.
Monkeys
can
be
prepared
with
bilateral,
circumscribed
lesions
limited
to
a
particular
structure
or
combination
of
structures.
The
effects
of
the
removals
on
memory
are
then
determined
quantita-
tively
by
evaluating
the
performance
of
these
monkeys
on
tasks
that
are
identical
to,
or
in
some
cases
analogous
to,
tasks
used
to
detect
memory
impairment
in
human
patients.
To
determine
whether
the
impairment
in
monkeys
is
long-lasting,
as
it
is
in
humans
after
medial
temporal
lobe
damage,
monkeys
can
be
tested
several
years
after
surgery.
One
of
the
reasons
that
it
took
so
long
to
develop
an
animal
model
of
human
amnesia
was
that,
until
the
early
1980s,
human
amnesia
itself
was
incompletely
understood,
and
it
was
therefore
unclear
what
parallels
should
be
looked
for
in
experimental
animals.
It
is
now
appreciated
that
memory
is
not
a
single
faculty
but
is
composed
instead
of
multiple
separate
systems,
only
one
of
which
is
impaired
in
amnesia
(19).
Accordingly,
only
certain
kinds
of
mem-
ory
tests
are
appropriate
for
demonstrating
an
impairment
in
animals
that
corresponds
to
human
amnesia.
Human
amnesia
impairs
the
ability
to
acquire
information
about
facts
and
events
(declarative
memory)
but
spares
the
capacity
for
skill
learning,
certain
kinds
of
conditioning
and
habit
learning,
as
well
as
the
phenomenon
of
priming
(Fig.
3).
Declarative
memory
is
accessible
to
conscious
recollection
and
available
to
multiple
re-
sponse
systems.
Nondeclarative
(implicit)
memory
includes
several
kinds
of
abilities,
all
of
which
are
nonconscious
and
expressed
through
performance.
These
abilities
provide
for
cumulative
changes
in
perceptual
or
response
systems
and
for
the
development
of
new
skills
and
habits.
The
information
is
not
accessible
to
conscious
recollection,
and
it
is
inflexible,
that
is,
it
has
limited
access
to
systems
not
involved
in
the
initial
learning.
The
idea
that
the
medial
temporal
lobe
is
involved
in
only
one
kind
of
memory
also
developed
in
the
animal
literature
(20,
21),
and
the
findings
from
humans
and
experimental
animals
are
now
in
substantial
agreement
that
limbic
lesions
have
selective
and
specific
effects
on
memory
(22).
The
development
of
the
animal
model
in
the
monkey
began
with
large
bilateral
lesions
of
the
medial
temporal
lobe
that
approximated
ARTICLES
1381
on December 9, 2008 www.sciencemag.orgDownloaded from
Table
1.
Characteristics
of
human
amnesia
that
have
been
produced
in
monkeys
with
large
bilateral
lesions
of
the
medial
temporal
lobe
(the
HWA+
lesion).
Characteristics
References
1.
Memory
is
impaired
on
several
tasks
including
ones
(25,
27,
56)
identical
to
those
failed
by
amnesic
patients.
2.
Memory
impairment
is
exacerbated
by
increasing
the
(11,
25)
retention
delay
or
the
amount
of
material
to
be
learned.
3.
Memory
impairment
is
exacerbated
by
distraction.
(25)
4.
Memory
impairment
is
not
limited
to
one
sensory
(32)
modality.
5.
Memory
impairment
can
be
enduring.
(25)
6.
Memory
for
events
prior
to
the
onset
of
amnesia
(57)
can
be
affected
(retrograde
amnesia).
7.
Skill-based
memory
is
spared.
(21,
58)
8.
Immediate
memory
is
spared.
(59)
the
damage
sustained
by
amnesic
patient
H.M.
(11,
23-25).
The
lesion
was
produced
by
means
of
a
direct
surgical
approach
through
the
ventral
surface
of
the
temporal
lobe,
and
the
damaged
area
included
the
hippocampal
formation
(that
is,
the
dentate
gyrus,
the
hippocampus
proper,
the
subicular
complex,
and
the
entorhinal
cortex),
the
amygdala,
and
the
surrounding
perirhinal
and
parahip-
pocampal
cortices
(Fig.
1A).
This
has
been
termed
the
H+A+
lesion
(26),
where
H
refers
to
the
hippocampus,
A
to
the
amygdala,
and
+
to
the
cortical
regions
adjacent
to
the
hippocampus
and
the
amygdala.
The
H+A+
lesion
in
monkeys
reproduced
many
important
fea-
tures
of
the
memory
impairment
in
patient
H.M.
and
in
other
amnesic
patients
(Table
1).
Like
human
amnesic
patients,
monkeys
were
severely
impaired
on
a
number
of
memory
tasks
but
were
entirely
normal
at
acquiring
and
retaining
skills.
Although
many
tasks
have
been
used
to
demonstrate
impaired
memory
in
monkeys
(27),
the
one
most
widely
used
has
been
the
delayed
nonmatching
to
sample
task.
In
this
test
of
recognition
memory,
a
single
object
is
presented
and
then
after
a
delay
two
objects
are
presented,
the
original
object
and
a
novel
one.
The
animal
must
choose
the
novel
object
to
obtain
a
food
reward.
Unique
pairs
of
objects
are
used
on
every
trial.
In
an
early
comparison
of
H+A+
lesions
with
lesions
of
the
white
matter
that
forms
the
temporal
stem,
only
the
H+A+
lesions
impaired
memory
(23).
This
result
suggested
that
the
structures
of
the
medial
temporal
lobe
were
critical
for
recognition
memory,
not
the
white
matter
that
connects
lateral
temporal
neocortex
to
dien-
cephalic
and
other
subcortical
targets
(28).
Contributions
of
Hippocampus,
Amygdala,
and
Adjacent
Cortex
A
lesion
of
the
posterior
medial
temporal
lobe
(the
H'
lesion)
also
impairs
memory
(Fig.
4),
albeit
less
severely
than
the
H+A+
lesion
(11,
25,
29-31).
The
H'
lesion
was
limited
to
the
hippo-
campal
formation
and
the
parahippocampal
cortex,
and
it
spared
the
anterior
portion
of
the
entorhinal
cortex
together
with
the
amygdala
and
other
cortex
adjacent
to
the
amygdala.
Monkeys
with
H'
lesions
that
were
first
trained
before
surgery
and
then
tested
after
surgery
(11,
32)
performed
better
than
monkeys
that
were
both
trained
and
tested
after
surgery
(29-31,
33).
Nevertheless,
the
H'
lesion
produced
statistically
significant
memory
impairment
in
mon-
1382
keys,
even
after
preoperative
training.
Moreover,
in
human
patients
(12,
14)
even
more
restricted
lesions
can
produce
clinically
mean-
ingful
memory
impairment.
There
are
three
possible
explanations
for
why
memory
in
monkeys
was
more
impaired
after
HWA'
lesions
than
after
H'
lesions.
As
the
H'
lesion
was
extended
anteriorly
to
include
the
A'
component,
the
memory
impairment
could
have
been
exacer-
bated
either
by
the
amygdala
damage
(A),
the
damage
to
cortex
adjacent
to
the
amygdala
('),
or
the
combination
of
amygdala
and
cortical
damage
(A').
The
experimental
evidence
favors
the
second
of
these
alternatives,
that
is,
the
additional
impairment
results
from
damage
to cortex
adjacent
to
the
amygdala
(Fig.
4).
An
important
clue
came
from
a
group
of
monkeys
with
stereotaxic
lesions
of
the
amygdaloid
complex
that
spared
the
surrounding
cortex
(the
A
lesion).
Monkeys
with
the
A
lesion
performed
normally
on
the
delayed
nonmatching
to
sample
task
and
on
three
other
memory
tasks,
all
of
which
are
sensitive
to
the
effects
of
H'
or
HWA'
lesions
(34).
Moreover,
monkeys
with
bilateral
lesions
of
the
hippocampal
formation,
made
conjointly
with
circum-
scribed
lesions
of
the
amygdaloid
complex
(the
HWA
lesion),
were
impaired
on
these
same
four
memory
tasks,
but
the
impairment
was
no
greater
after
HWA
lesions
than
after
H'
lesions
(34).
Thus,
amygdala
damage
alone
did
not
impair
memory
nor
did
it
exacerbate
the
memory
impairment
associated
with
hippocampal
formation
lesions.
Experiments
with
rodents
have
led
to the
same
conclusion.
Hippocampal
lesions,
or
lesions
of
anatomically
related
structures,
impair
performance
on
a
variety
of
spatial
and
nonspatial
memory
tasks.
However,
amygdala
lesions
do
not
impair
performance
on
these
same
tasks,
nor
does
the
addition
of
an
amygdala
lesion
to
a
lesion
of
the
hippocampal
system
exacerbate
the
deficit
(35).
The
findings
for
rats
and
monkeys,
taken
together,
suggest
that
the
severe
memory
impairment
associated
with
HWA'
lesions
should
be
attributed,
not
to
the
addition
of
the
amygdala
to
the
H'
lesion,
but
instead
to
the
addition
of
cortical
regions
that
lie
adjacent
to
the
amygdala
and
that
are
necessarily
damaged
when
the
amygdala
is
removed
by
the
conventional
surgical
approach.
A
10or
NMTS-1
90h
8
&
I
B
'°°r
NMTS-2
90
h
801
70-
60
N
7
'
A=3
i
WA=3
'
H
=3
H
=3
H§,A
M
=3
80-
A=3
N=7
H
*A=3
H
-3
70
h
60
50
50
H
=3
8
15
60
10mn
8
15
60
1
Omin
Delay
(s)
Deby
(s)
Fig.
4.
(A)
Performance
on
the
delayed
nonmatching
to
sample
task
6
to
8
weeks
after
surgery
by
seven
normal
monkeys
(N)
and
five
groups
of
monkeys
with
lesions:
A,
circumscribed
lesion
of
the
amygdala
sparing
surrounding
cortex;
H',
lesion
of
the
hippocampal
formation
and
parahip-
pocampal
cortex;
H+A,
a
combination
of
these
two
lesions;
HA+,
lesion
of
the
hippocampus,
the
amygdala,
and
adjacent
cortical
regions,
that
is,
perirhinal,
entorhinal,
and
parahippocampal
cortex;
H',
lesion
of
the
hippocampus
and
the
same
cortical
regions
damaged
in
the
H+A+
lesion.
(B)
Performance
of
the
same
normal
monkeys
as
in
(A)
and
four
of
the
five
groups
with
lesions
on
the
delayed
nonmatching
task
1
to
2
years
later.
At
the
10-min
delay
in
(A)
and
(B),
the
SEM
ranged
from
2.3%
to
4.7%.
SCIENCE,
VOL.
253
on December 9, 2008 www.sciencemag.orgDownloaded from
The
identity
of
the
critical
cortical
structures
was
suggested
by
a
reexamination
of
the
histological
material
from
an
early
study
(23)
in
monkeys
with
HWA'
lesions,
which
showed
that
the
perirhinal
and
entorhinal
cortex
adjacent
to
the
amygdala
had
sustained
substantial
damage.
Neuroanatomical
evidence
suggests
why
these
cortical
regions
could
be
important
for
memory
(Fig.
1B).
The
perirhinal
cortex
and
the
parahippocampal
gyrus
(areas
TH
and
TF)
provide
nearly
two-thirds
of
the
cortical
input
to
the
entorhinal
cortex
(36).
The
entorhinal
cortex
in
turn
is
the
source
of
the
perforant
path,
the
major
efferent
projection
to
the
hippo-
campus
and
dentate
gyrus.
Thus,
these
cortical
regions
collectively
provide
the
principal
route
by
which
information
in
neocortex
reaches
the
hippocampus.
We
therefore
reasoned
that
extending
the
H'
lesion
forward
to
include
anterior
entorhinal
cortex
and
the
perirhinal
cortex,
but
sparing
amygdala
(the
H`+
lesion),
should
exacerbate
the
deficit
produced
by
the
H'
lesion
and
impair
memory
as
much
as
it
is
impaired
by
the
HWA'
lesion.
This
idea
has
been
confirmed.
Monkeys
with
bilateral
H`+
lesions
were
as
severely
impaired
on
the
delayed
nonmatching
to
sample
task
as
those
with
HWA'
lesions,
and
the
severity
of
the
impairment
remained
stable
for
more
than
1
year
after
surgery
(Fig.
4)
(37).
In
two
testing
sessions
and
with
three
retention
delays
(15
s,
60
s,
and
10
min),
monkeys
with
H+
lesions
were
more
severely
impaired
than
either
monkeys
with
H'
lesions
(P
<
0.05)
or
monkeys
with
H+A
lesions
(P
<
0.05).
We
also
reasoned
that
memory
should
be
severely
disrupted
by
a
lesion
limited
to
the
perirhinal
and
parahippocampal
cortex
(the
PRPH
lesion),
sparing
the
hippocampus,
the
amygdala,
and
the
entorhinal
cortex
(38).
Monkeys
who
sustained
the
PRPH
lesion
were
severely
impaired
on
all
three
amnesia-sensitive
tasks
that
were
administered.
Overall,
they
appeared
about
as
impaired
as
monkeys
with
H+A+
lesions.
The
findings
with
H`+
and
PRPH
lesions
indicate
that
damage
to
the
perirhinal
cortex,
and
not
damage
to
the
amygdala,
contrib-
utes
to
the
severe
memory
impairment
associated
with
H+A+
lesions.
This
conclusion
requires
that
damage
to
perirhinal
cortex
is
substantial
in
monkeys
with
H`+
lesions
or
PRPH
lesions
but
minimal
in
monkeys
with
H'
or
H+A
lesions.
To
address
this
issue,
we
measured
planimetrically
the
areas
of
the
entorhinal,
perirhinal,
and
parahippocampal
cortices,
and
the
area
of
the
inferotemporal
cortex
(area
TE)
(37).
The
perirhinal
cortex
was
the
only
area
where
the
H`+
and
PRPH
monkeys
sustained
more
damage
than
the
H'
and
the
H+A
monkeys.
In
the
monkeys
with
H'
or
H+A
lesions,
less
than
25%
of
the
perirhinal
cortex
was
damaged
bilaterally.
In
the
monkeys
with
H`+
lesions,
however,
approximately
74%
of
the
perirhinal
cortex
was
damaged
bilaterally.
In
addition,
in
the
monkeys
with
PRPH
lesions,
approximately
68%
of
the
perirhinal
cortex
was
damaged.
These
investigations
permit
the
main
components
of
the
medial
temporal
lobe
memory
system
to
be
identified.
The
memory
system
consists
of
the
hippocampal
formation,
including
entorhinal
cortex,
Fig.
5.
Dissociation
of
emotion
(E)
and
memory
(M)
by
lesions
of
the
amygdala
and
the
hippocampal
formation.
The
emotion
score
was
obtained
by
measuring
the
responsiveness
of
monkeys
to
seven
inanimate
stimuli
that
could
elicit
investigatory
or
consummatory
behavior.
The
memory
score
is
the
score
on
the
delayed
nonmatching
to
sample
task
at
a
10-min
delay.
Normal,
performance
by
15
normal
monkeys;
A
and
H,
performance
by
two
groups
of
monkeys
with
damage
to
both
the
hippocampal
formation
and
the
amygdala
(total
n
=
10);
A,
performance
by
two
groups
of
monkeys
with
damage
to
the
amygdala
but
not
the
hippocampal
formation
(total
n
=
7);
H,
performance
by
two
groups
of
monkeys
with
damage
to
the
hippocampal
formation
or
associated
cortical
areas
or
both,
but
not
the
amygdala
(total
n
=
10).
Error
bars
show
the
SEM.
[Reprinted
from
(42)
with
permission,
X
1991
Churchill
Livingstone]
together
with
the
adjacent,
anatomically
related
perirhinal
and
parahippocampal
cortices
(Fig.
1B).
Importantly,
the
cortex
adja-
cent
to
the
hippocampus
is
not
simply
a
conduit
for
connecting
neocortex
to
hippocampus.
Thus,
both
the
H`+
lesion
and
the
PRPH
lesion
produced
more
severe
memory
impairment
than
the
H'
lesion.
Accordingly,
the
cortex
that
was
damaged
in
the
H`+
and
PRPH
lesions,
but
undamaged
in
the
H'
lesion,
must
also
be
important
for
memory
function.
The
implication
is
that
information
from
neocortex
need
not
reach
the
hippocampus
itself
in
order
for
some
memory
storage
to
occur.
These
considerations
explain
why
memory
impairment
is
increased
by
damage
to
cortical
structures
in
addition
to
the
hippocampus.
Another
important
conclusion
is
that
the
amygdaloid
complex
is
not
a
component
of
the
medial
temporal
lobe
memory
system
and
does
not
contribute
to
the
kind
of
(declarative)
memory
that
depends
on
this
system
(39).
Experiments
with
both
rats
and
monkeys
suggest
that
the
amygdala
is
important
for
other
functions,
including
conditioned
fear
and
the
attachment
of
affect
to
neutral
stimuli
(40).
The
amytdala
may
also
have
a
broader
role
in
establishing
links
between
stimuli,
as
in
making
associations
among
sensory
modalities
(41).
Additional
evidence
also
supports
the
distinction
between
the
functions
of
the
hippocampal
formation
and
the
amygdala.
For
six
different
groups
of
monkeys
that
were
given
memory
tests,
quanti-
tative
ratings
of
emotional
behavior
were
obtained
(42).
Partial
or
complete
damage
to
the
amygdaloid
complex
caused
marked
alter-
ations
in
emotional
behavior.
Specifically,
monkeys
were
less
fearful
than
normal
and
were
unusually
willing
to
touch
and
otherwise
interact
with
novel
stimuli.
However,
unless
there
was
also
damage
to
the
hippocampal
formation
or
adjacent
cortex,
memory
was
intact.
Conversely,
all
groups
with
damage
to
hippocampal
formation
or
associated
cortex
had
memory
impairment.
Yet,
unless
there
was
also
damage
to
the
amygdala,
emotional
behavior
was
normal
(Fig.
5).
In
summary,
these
findings
emphasize
the
importance
for
memory
of
the
hippocampal
formation
and
the
perirhinal
and
parahippo-
campal
cortex
of
the
medial
temporal
lobe.
Other
findings
in
monkeys
and
humans
are
consistent
with
this
proposal
(43).
Medial
temporal
lobe
cortex
presumably
participates
in
memory
functions
by
virtue
of
its
extensive
reciprocal
connections
with
putative
memory
storage
sites
in
neocortex.
Thus,
disconnection
of
infer-
otemporal
cortex
(area
TE)
from
the
limbic
system
severely
impaired
memory
(44).
In
contrast,
damage
to
the
fornix,
which
is
a
major
subcortical
efferent
projection
of
the
hippocampal
formation,
and
damage
to
a
major
diencephalic
target
of
the
fornix,
the
mammillary
nuclei,
had
only
mild
effects
on
memory
(30,
45,
46).
NORMAL
0
CD
L..
0
0
L)
C,)
A
L)
0
0
C1)
0
0
C)
01
b..
A
and
H
0
e)
L-
0
U)
Ul)
H
80
°)
0
0
70
60
2
0
a-
0
I
0
10
Jso0
20
SEPTEMBER
1991
ARTICLES
1383
on December 9, 2008 www.sciencemag.orgDownloaded from
The
Hippocampus
Until
recently,
it
has
been
difficult
to
determine
in
monkeys
whether
the
hippocampus
itself
makes
an
essential
contribution
to
memory
function.
A
direct
surgical
approach
to
the
hippocampus
necessarily
damages
adjacent
cortex
(that
is,
entorhinal
and
parahip-
pocampal
cortex),
and
a
stereotaxic
approach
based
on
a
standard
monkey
brain
atlas
is
problematic,
because
the
size
and
shape
of
the
hippocampus
can
vary
considerably
from
monkey
to
monkey.
We
have
combined
stereotaxic
neurosurgery
with
MRI
to
im-
prove
the
accuracy
of
surgical
lesions
within
the
hippocampus.
Prior
to
surgery,
MR
images
were
obtained
from
four
monkeys
with
small
radio-opaque
landmarks
anchored
to
the
skull
(47)
(Fig.
2,
C
and
D).
The
landmarks
were
then
used
to
make
bilateral
hippocampal
lesions
(the
H
lesion).
Monkeys
sustained
damage
to
the
hippocam-
pus,
dentate
gyrus,
and
subiculum,
but
the
perirhinal,
entorhinal,
and
parahippocampal
cortices
were
almost
entirely
spared.
At
delays
of
10
min
in
the
nonmatching
to
sample
task,
monkeys
with
H
lesions
were
as
impaired
as
monkeys
with
H'
lesions.
On
two
other
Fig.
6.
Schematic
drawing
of
primate
neocortex
plus
the
components
of
the
medial
temporal
lobe
memory
system
believed
to
be
important
in
the
transition
from
perception
to
memory.
The
networks
in
cortex
show
representations
of
visual
object
quality
(in
area
TE)
and
object
location
(in
area
PG).
If
this
distributed
activity
is
to
develop
into
a
stable
long-term
memory,
neural
activity
must
occur
at
the
time
of
learning
along
projections
from
these
regions
to
the
medial
temporal
lobe
(first
to
the
parahippocampal
gyrus
(TF/H),
perirhinal
cortex
(PR),
and
entorhinal
cortex
(EC),
and
then
in
the
several
stages
of
the
hippocampus
[the
dentate
gyrus
(DG)
and
the
CA3
and
CAl
regions].
The
fully
processed
input
eventually
exits
this
circuit
by
way
of
the
subiculum
(S)
and
EC,
where
widespread
efferent
projections
return
to
neocortex.
The
hippocampus
and
adjacent
structures
bind
together
or
otherwise
support
the
development
of
representations
in
neocortex,
so
that
subsequently
memory
for
a
whole
event
(for
example,
represented
in
TE
and
PG)
can
be
reactivated
even
from
a
partial
cue.
[Reprinted
from
(61)
with
permission,
©
F.
K.
Schattauer
Verlagsgesellschaft
mbH]
tasks
(retention
of
object
discriminations
and
eight-pair
concurrent
discrimination
learning),
monkeys
with
H'
lesions
were
significant-
ly
more
impaired
than
H
monkeys
(47).
These
results
provide
direct
support
for
the
long-standing
idea
that
the
hippocampus
itself
is
important
for
memory.
They
also
emphasize
the
importance
of
the
cortical
regions
adjacent
to
the
hippocampus
(specifically,
the
ento-
rhinal
and
parahippocampal
cortices
that
are
damaged
in
the
H'
lesion
but
not
in
the
H
lesion)
(48).
We
have
also
examined
the
effects
of
hippocampal
damage
by
using
a
noninvasive
procedure
to
produce
global
ischemia
in
the
monkey
(the
ISC
lesion).
This
method
consistently
produced
sig-
nificant
bilateral
cell
loss
in
the
CAl
and
CA2
fields
of
the
hippocampus
and
the
hilar
region
of
the
dentate
gyrus
(49).
Cell
loss
was
greater
in
the
posterior
portion
of
the
hippocampus
than
in
its
anterior
portion.
On
the
delayed
nonmatching
to
sample
task,
four
monkeys
with
ISC
lesions
performed
similarly
to
the
three
monkeys
with
H'
lesions.
Moreover,
the
impairment
was
long-lasting.
On
two
other
tasks,
monkeys
with
ISC
lesions
performed
better
than
monkeys
with
H'
lesions
(P
=
0.06)
and
similar
to
monkeys
with
H
lesions.
These
findings
from
ISC
monkeys
suggest
that
even
incomplete
damage
to
the
hippocampus
is
sufficient
to
impair
memory
in
monkeys,
just
as
in
humans
(patient
R.B.).
It
is
possible
that
additional
neural
damage
occurred
in
the
ISC
monkeys
and
in
R.B.
and
that
such
damage
was
not
detected
in
histological
examination.
Yet,
the
ISC
monkeys
obtained
about
the
same
memory
scores
as
H
monkeys
and
better
memory
scores
than
H'
monkeys.
Accordingly,
it
seems
reasonable
to
suppose
that
the
ISC
animals
(and
by
extension,
patient
R.B.)
did
not
have
substantial
neuropathological
change
affecting
memory
that
was
not
detected
histologically.
The
damage
produced
by
the
ISC
lesion
to
a
critical
component
of
hippocampal
circuitry
(that
is,
the
CA1
region)
may
be
as
disruptive
to
hippo-campal
function
as
more
extensive
damage
(the
H
lesion).
Results
from
a
rodent
model
of
global
ischemia
are
consistent
with
this
view
(50).
Function
of
the
Medial
Temporal
Lobe
Memory
System
The
medial
temporal
lobe
memory
system
performs
a
critical
function
beginning
at
the
time
of
learning,
in
order
that
represen-
tations
can
be
established
in
long-term
memory
in
an
enduring
and
usable
form.
Coordinated
and
distributed
activity
in
neocortex
is
thought
to
underlie
perception
and
immediate
(short-term)
memory
(44,
51).
These
capacities
are
unaffected
by
medial
temporal
lobe
damage.
However,
if
distributed
cortical
activity
is
to
be
trans-
formed
into
stable
long-term
memory.,
then
the
hippocampus
and
related
structures
must
be
engaged
at
the
time
of
learning
(Fig.
6).
The
hippocampus
and
related
structures
may
serve
as
a
device
for
forming
conjunctions
between
ordinarily
unrelated
events
or
stim-
ulus
features,
which
are
processed
and
represented
by
distinct
cortical
sites
(52).
In
this
sense,
the
hippocampal
system
is
a
storage
site
for
a
simple
memory,
a
summary
sketch,
or
an
index
(53).
As
long
as
a
percept
is
in
view
or
in
mind,
its
representation
remains
coherent
in
short-term
memory
by
virtue
of
mechanisms
intrinsic
to
neocortex.
However,
a
problem
potentially
arises
when
attention
shifts
to
a
new
percept
or
a
new
thought,
and
one
then
attempts
to
recover
the
original
memory.
We
propose
that
the
capacity
for
later
retrieval
is
achieved
because
the
hippocampal
system
has
"bound
together"
the
relevant
cortical
sites
that
together
represent
memory
for
a
whole
event.
The
role
of
this
system
can
be
further
described
in
two
important
ways.
First,
the
hippocampal
system
is
crucial
for
the
rapid
acquisi-
tion
of
new
information
about
facts
and
events,
which
are
then
SCIENCE,
VOL.
253
1384
on December 9, 2008 www.sciencemag.orgDownloaded from