Proc. Nati. Acad. Sci. USA
Vol. 91, pp. 5637-5641, June 1994
Neurobiology
The animal model of human amnesia: Long-term memory impaired
and short-term memory intact
PABLO ALVAREZ*t, STUART ZOLA-MORGAN*I, AND LARRY R. SQUIRE*tt
Departments of *Psychiatry and tNeurosciences, University of California at San Diego, La Jolla, CA 92093; and WVeterans Affairs Medical Center, San Diego,
CA 92161
Contributed by Larry R. Squire, March 8, 1994
ABSTRACT Normal monkeys and monkeys with lesions of
the hippocampal formation and adjacent cortex (the H+ lesion)
were trained on the delayed nonmatching to sample (DNMS)
task with a delay of 0.5 s between the sample and the choice.
The animals with H+ lesions learned the task normally at this
short delay and also exhibited the same pattern of response
latencies as normal monkeys. This finding contrasts with
previous observations that initial learning of the DNMS task
with delays of8-10 s is impaired after H+ lesions. The absence
ofan impairment at a delay of0.5 s indicates that the H+ lesion
does not affect short-term memory. In contrast, when monkeys
with H+ lesions were tested at longer delays (>30 s), an
impairment was observed. This selective impairment occurred
when the delays were presented sequentially (from 0.5 s to 10
min) and also when delays were presented in a mixed order (1
s, 1 min, and 10 min). The data indicate that the H+ lesion
produces a selective impairment in long-term memory, in the
absence of a detectable deficit in short-term memory or perception.
Accordingly, the rmdings firm the long-standing
idea, based primarily on studies of humans, that short-term
memory is independent of medial temporal lobe function. The
findings thereby establish an important parallel between memory
impairment in monkeys and humans and provide additional
support for the validity of the animal model of human
amnesia in the monkey.
During the last decade, a model of human amnesia was
developed in the nonhuman primate (1, 2). Bilateral damage
to the medial temporal lobe reproduces many of the features
of memory impairment in human amnesia. The structures
that when damaged produce amnesia are the hippocampus
and the adjacent, anatomically related perirhinal, entorhinal,
and parahippocampal cortices (3). One behavioral task that
has been used extensively to measure visual recognition
memory in monkeys is delayed nonmatching to sample
(DNMS). Performance on DNMS is sensitive to medial
temporal lobe damage in both humans and monkeys (34). In
this task, the animal first sees an object, and then after a
prescribed delay the animal is given a choice between the
previously seen (sample) object and a novel one. The animal
must choose the novel object to obtain a food reward.
Typically, animals are trained with a short (8-10 s) delay until
they reach a criterion level of performance (90% correct for
100 trials). At this point, the delay is progressively lengthened.
Monkeys with medial temporal lobe lesions exhibit an
impairment that appears to become more severe as the
interval between the sample and choice increases (1, 3). Poor
performance when the delay between sample and choice is
long suggests that the impairment is due to a loss of memory
function similar to that encountered in human amnesia: the
object cannot be held in memory for more than afew seconds.
Yet, it is also the case that the impairment in performing
DNMS is usually accompanied by an impairment in learning
the task initially-that is, animals with medial temporal lobe
lesions need more trials than normal animals to learn the task,
even at the short (8-10 s) delay that is typically used for
training. There are at least three possible explanations forthis
impairment. First, the delay used for training (i.e., 8-10 s)
may be too long for animals with a memory impairment to
bridge effectively, with the result that more trials are needed
to learn the task. It is also possible that medial temporal lobe
lesions produce some cognitive impairment other than memory
(e.g., a perceptual or attentional impairment) that makes
the task difficult to learn. Finally, damage to the medial
temporal lobe might produce both impaired long-term memory,
reflected in poor performance at long delays, and
impaired short-term memory, reflected in slower learning of
the task.
As discussed previously, this issue is fundamental (5-8).
Human amnesic patients exhibit intact short-term memory
(9, 10). If medial temporal lobe lesions impair short-term
memory in monkeys, then the medial temporal lesion cannot
provide a valid animal model of human amnesia.
To address the issue directly, one needs to be able to test
animals with very short delays between sample and choice.
We developed an automated testing apparatus, based on the
one described by Murray et al. (11), which can be used to test
monkeys on DNMS with delays as short as 0.5 s. A finding
of normal performance at a 0.5-s delay by monkeys with
medial temporal lobe lesions would suggest that short-term
memory is intact and that the 8-s delay used in the standard
version of the task is too long-i.e., it is beyond the limit of
short-term memory. Conversely, a finding of impaired performance
at a 0.5-s delay would suggest that the lesion
produces a short-term memory impairment or some cognitive
impairment other than or in addition to memory.
The issue under study was earlier the subject of a commentary
(7), which included a preliminary report of some of
the data presented here.
METHODS
Subjects. The subjects were nine young adult cynomolgus
monkeys (Macacafascicularis) weighing between 4.1 and 6.8
kg at the start of this study. Based on weight-and-age tables
(12, 13), these monkeys were estimated to be between 4 and
7 years of age (young adults). The nine animals consisted of
two groups: four monkeys with bilateral lesions involving the
hippocampus proper, the dentate gyrus, the subiculum, the
posterior half of entorhinal cortex, and the parahippocampal
cortex (group H+) and five unoperated control monkeys
(group N). Fig. 1 shows a coronal section from the brain of
a monkey in the H+ group. Prior to the present study, all
animals had participated in a pilot study of visual object
discrimination learning. The H+ animals and animals N4 and
Abbreviation: DNMS, delayed nonmatching to sample.
5637
The publication costs ofthis article were defrayed in part by page charge
payment. This article must therefore be hereby marked "advertisement"
in accordance with 18 U.S.C. §1734 solely to indicate this fact.
5638 Neurobiology: Alvarez et al.
FIG. 1. Thionin-stained coronal section midway through the
lateral geniculate from one monkey (H+3) in the operated group. This
animal sustained near total bilateral ablation of the hippocampal
region (i.e., the cell fields of the hippocampus proper, the dentate
gyrus, and the subiculum), with only slight sparing of the most
anterior portion of the uncal region of the hippocampus. The parahippocampal
cortex and the posterior half of entorhinal cortex were
extensively damaged bilaterally. There was some sparing of the
anterior entorhinal cortex bilaterally, but cells in layer II, which
project to the ablated hippocampal cell fields, were completely
eliminated through retrograde degeneration. The amygdaloid complex
was entirely spared. Overall, this damage was similar to that
described previously in monkeys with H+ lesions prepared in our
laboratory (35). In addition, perirhinal cortex sustained slight and
asymmetrical damage. Inferotemporal cortex (area TE) sustained
moderately severe damage on the left side and moderate damage on
the right side. The posterior region of area TE on the right side was
compressed by a moderate enlargement of the lateral ventricle.
There was also slight and asymmetrical damage to the anterior
ventral portion ofvisual cortex. Histological examination ofthe brain
from a second monkey (H+4) indicated that damage was similar to
what was observed in monkey H+3, except that damage to area TE
and the visual cortex was minimal. Monkey H+4 required more trials
than any ofthe other monkeys to reach performance criterion during
training on the basic task. This monkey experienced motivational
problems throughout testing and was not tested on mixed delays.
Analysis of magnetic resonance images from the two other operated
monkeys indicated bilateral and extensive damage to the hippocampal
region and adjacent entorhinal and parahippocampal cortex.
(x 1.57.)
N5 had also participated in a formal study of memory for
simple object discriminations (4). In addition, animals N1-N3
had been tested on a battery of five tasks, including the
standard, manually administered version of DNMS (animals
N4-N6 in ref. 14). None of the animals had had any experience
with the automated test apparatus described below.
Test Apparatus. An 80286-IBM compatible computer operated
a touch-sensitive monitor screen (Microtouch Systems,
Woburn, MA) and an automatic pellet dispenser (11).
Individual stimuli displayed on the monitor were composed
of two superimposed ASCII characters of different shape,
size, and color. A contact by the monkey within a 5 x 6 cm
rectangular area centered on the stimulus counted as a
response. Correct responses were followed by the delivery of
a reward (190-mg banana-flavored monkey pellet; P. J.
Noyes, Lancaster, NH) into a small hopper located 5 cm
directly below the center of the monitor.
DNMS. Monkeys were initially pretrained in four stages to
use the apparatus. They were first trained to retrieve rewards
from the hopper, then to touch a stimulus when it was
presented in the center of the monitor screen, then to touch
the stimulus within 60 s to obtain a reward, and finally to
touch a stimulus on the left, center, or right of the screen
within 30 s.
Initial Learning. A sample stimulus was presented in the
center of the screen for 20 s (sample phase). If the monkey
touched the stimulus, the screen became blank for 0.5 s and
then the choice phase began. The sample stimulus was not
rewarded. Ifthe monkey did not touch the sample, the screen
blanked, and the next intertrial interval (20 s) began. For the
choice phase, the sample stimulus and a novel stimulus
appeared on the left and right sides ofthe screen at a distance
of 8.2 cm from the center. Whether the sample appeared on
the left or right side varied according to a predetermined,
pseudorandom schedule (16). If the monkey responded correctly-i.e.,
it touched the novel stimulus within 20 s-the
screen blanked, two banana pellets were delivered, and the
intertrial interval began. Ifthe monkey did not respondwithin
20 s or responded incorrectly-i.e., it touched the sample
stimulus-the screen blanked, no reward was delivered, and
the intertrial interval began. Monkeys were trained until they
reached a learning criterion of 90%o correct performance or
better within five consecutive sessions (200 trials). The
animal's responses and response latencies were recorded by
the computer.
Sequential Delay Testing. After the learning criterion had
been reached, the delay interval was increased consecutively
to 4 s, then to 8 s, 15 s, 60 s, 3 min, and 10 min. Two hundred
trials (40 trials a day for 5 days) were given at each ofthe 4-,
8-, and 15-s delays. One hundred trials were given at the 60-s
delay (20 trials a day for 5 days) and the 3-min delay (10 trials
per day for 10 days). Fifty trials were given at the 10-min
delay (5 trials per day for 10 days).
Mixed Delay Testing. Monkeys were next tested using
delays of 1 s, 1 min, and 10 min presented in a mixed order.
Specifically each delay was presented for a block of 3 days
before moving to the next delay, and the order in which the
delays were presented was balanced within and between
monkeys. For the 1-s and 1-min delays, 60 trials per block
were given (20 trials per day); for the 10-min delay, 15 trials
per block were given (5 trials per day). Monkeys received a
total of two blocks at the 1-s and 1-min delays and a total of
four blocks at the 10-min delay. Only four normal and three
H+ monkeys were given mixed delay testing.
RESULTS
Animals with H+ lesions learned at the same rate as N
animals to work with the apparatus and to obtain rewards
during the four stages of pretraining. A two-way ANOVA
(group by stage, as measured by the number ofdaily training
sessions required to complete each stage) revealed no significant
effects (F < 1.0, P > 0.10).
Initial Learning. Table 1 and Fig. 2 show the number of
trials needed by the monkeys in group N and group H+ to
reach 90% correct performance with the 0.5-s delay. The
groups performed similarly, as measured by trials to criterion
[N = 2011, H+ = 2837; t(7) = 0.64, P > 0.50] and errors to
criterion [N = 721, H+ = 943; t(7) = 0.46, P > 0.50]. Fig. 3
shows the mean latencies for responses during the sample
and choice phases, averaged across all training trials. A
two-way ANOVA (group x response type) indicated that
responses on the choice trial (N = 1.2 s, H+ = 1.0 s) were
faster than responses to the sample [N = 2.8 s, H+ = 2.3 s;
F(1, 7) = 47.2, P < 0.001]. There was no effect ofgroup [F(1,
7) = 2.31, P > 0.10], and no interaction between group and
response type [F(l, 7) = 1.1, P > 0.10]. Additional analyses
indicated that the latency to respond at the start of training
(first 100 trials; N = 3.6 s, H+ = 2.3 s) was greater than when
criterion performance was reached (last 100 trials; N = 1.2 s,
H+ = 1.3 s; P < 0.05). There werenodifferences between the
groups and no interaction between group and stage oftraining
(P> 0.10). In summary, as measured by trials and errors to
criterion as well as by response latencies, the H+ group was
Proc. Natl. Acad. Sci. USA 91 (1994)
Proc. Nati. Acad. Sci. USA 91 (1994) 5639
Table 1. Learning and performance scores
Initial learning Sequential delay testing Mixed delay testing
Animal Trials Errors 0.5 s 4 s 8-15 s 30 s-1 min 3-10 min 1 s 1 min 10 min
N1 701 244 92 87 85 82 77 95 86 75
N2 3700 1509 92 84 79 78 67 N3
1631 419 93 97 89 79 68 96 89 60
N4 2625 959 90 80 70 66 67 98 88 62
N5 1398 476 93 77 80 81 67 94 87 67
Mean N 2011 721 92 85 81 77 69 96 88 66
H+1 517 144 93 91 83 68 69 98 86 60
H+2 1539 567 90 78 75 71 62 98 79 59
H+3 2830 804 90 68 75 65 59 94 82 55
H+4 6460 2258 91 79 70 70 53 Mean
H+ 2837 943 91 79 76 69 61 97 82 58
Scores are given for the monkeys in the N and H+ groups on the DNMS task for initial learning, for the performance test
when delays were presented sequentially (Sequential delay testing) , and for the performance test when delays were
presented in a mixed order (Mixed delay testing). The Trials and Errors scores are the number oftrials and errors required
toreachlearningcriterion during initiallearning. The scores fordelays are the percent correct scores forthe indicated delays.
Animals N5 and H+4 were not tested on the mixed delay paradigm.
indistinguishable from the N group in learning the DNMS
task at the 0.5-s delay.
Sequential Delay Tesing. Table 1 shows the performance of
the H+ and N groups as the delay intervals were increased
sequentially from 0.5 s to 10 min. The data for the 0.5-s delay
were not included in the analyses because these data simply
represent the final trials of training. Except for the 4-s data,
the data from adjacent delay intervals were averaged as
indicated in Table 1. An ANOVA involving two groups and
four delays (4 s, 8-15 s, 30 s-1 min, and 3-10 min) revealed
a significant effect of group [F(1, 7) = 5.49, P = 0.05] and
delay [F(2, 14) = 23.59, P < 0.001] and no significant group
x delay interaction [F(2, 14) = 0.62, P > 0.50]. Separate
comparisons between the N group and the operated group at
each of the delay intervals revealed that the H+ group was
unimpaired at the short-delay intervals [4 s: N = 85% correct,
H+ = 79% correct, t(7) = 1.0, P > 0.30; 8-15 s: N = 81%
correct, H+ = 76%correct; t(7) = 1.2, P > 0.20] and impaired
at the longer-delay intervals [30 s-i min: N = 77% correct,
H+ = 69% correct, t(7) = 2.7, P < 0.05; 3-10 min; N = 69%
correct, H+ = 61% correct, t(7) = 2.2, P = 0.05].
Response latencies during sequential delay testing (Fig. 4)
were analyzed with a three-factor ANOVA (group x response
type x four delays). There was an effect of response
7000
.
6000
to0
0~c5000
e 4000
s-
o
,,' 3000
c 2000
a
1000
0
N H+
FIG. 2. Initial learning ofthe DNMS task with a delay of0.5 s for
normal monkeys (N) and monkeys with bilateral lesions of the
hippocampal region, the posterior entorhinal cortex, and the parahippocampal
cortex (H+). Symbols show scores for individual ani-
mals.
type (sample vs. choice) [F(1, 7) = 47.9, P < 0.001] and delay
[F(1, 7) = 66.0, P < 0.001] but no effect of group (F < 1.0)
and no interaction involving the group factor (F < 1.0). The
significant response type x delay interaction [F(3, 21) = 9.6,
P < 0.001] indicated that as the delay increased the response
latency for the sample phase increased more than the response
latency for the choice phase.
MixedDelay Testing. Fig. S shows the performance ofthe N
and H+ groups when the delays were presented in the mixed
block design. As in the case when delays were presented
sequentially, an analysis of variance (two groups and three
delays) revealed an effect ofgroup [F(1, 5) = 8.42, P < 0.05]
and delay [F(2, 10) = 142.52, P < 0.001] and no group x delay
interaction [F(2, 10) = 2.43, P > 0.10]. Separate comparisons
between the normal group and the H+ group at each of the
three delay intervals indicated that the H+ group was unimpaired
at the 1-s delay [N = 96%o correct, H+ = 97%o correct;
t(5) = 0.61, P > 0.50] and impaired at the 1-min delay [N =
88% correct, H+ = 82% correct; t(5) = 2.8, P < 0.05]. At the
10-min delay, the performance of the H+ group was numerically
worse than that ofthe normal group, but this difference
did not reach significance [N = 66% correct, H+ = 58%
correct, t(5) = 1.9, P = 0.11], presumably due to the unusually
large variance in the N group.
4
- 3
U
0
U,
0
c
2
U,
C
0
a: 1
N Ht N HI
Sample Choice
FIG. 3. Mean response latencies to the sample and choice trials
during initial learning of the DNMS task for normal monkeys (N,
shaded bars) and monkeys with H+ lesions (open bars). Error bars
show the standard error of the mean.
Neurobiology: Alvarez et al.
5640 Neurobiology: Alvarez et al.
12 r
10
-
'4
2
0
Chas
.5s 4s 8s-15s 30s-lm
DelaY
3m-10m
FIG. 4. Mean response latencies to the sample phase (solid lines)
and choice phase (dashed lines) during performance of the DNMS
task by normal monkeys (N, *) and monkeys with H+ lesions (o).
The delays were increased sequentially from 0.5 s to 10 min.
To compare directly the findings from the short (1-s) delay
and the two long (1 min and 10 min) delays, an ANOVA was
also carried out involving two groups and two delays (1 s and
the average of the 1-min and 10-min delays). This revealed a
significant effect of delay [F(1, 5) = 316.9, P < 0.001] and a
significantgroup x delay interaction [F(1, 5) =
8.61, P < 0.05].
The effect ofgroup approached significance [F(1, 5) = 5.33, P
=
0.07]. The H+ group was significantly impaired at the long
delays [70% correct vs. 77% correct; t(5) = 3.3, P < 0.05].
Finally, the mixed delay data were also evaluated with
signal detection analyses, using an unbiased measure of
discriminability (d') (17). For the application of d' scores to
data from the DNMS test, see Ringo (5). An analysis of
variance based on d' scores (two groups x three delays)
yielded a significant effect of delay [F(2, 10) = 148.3, P <
0.001] and a group x delay interaction that fell just short of
significance [F(2, 10) = 3.65, P < 0.07]. The results were
similarwhen the datafrom the two long delays were averaged
and the ANOVA was repeated. There was an effect of delay
[F(1, 5) = 192, P < 0.001] and a marginal group x delay
interaction [F(1, 5) = 6.05, P < 0.06].
The response latencies during mixed delay testing were
similar for the two groups (P > 0.10).
100 r
901
I
I
80o
70
N (4)
(3)
60
50
Is 1 min
Defty
10mn
FIG. 5. Mean percent correct performance for normal monkeys
(N, *) and H+ monkeys (o) on the DNMS task when the delay
intervals were presented in a mixed order. Error bars show the
standard error of the mean.
DISCUSSION
Monkeys with bilateral lesions of the hippocampal formation
(the H+ lesion) learned the DNMS task as well as normal
monkeys when the delay between sample and choice was
very short (0.5 s). In addition, during the sample and choice
phases ofthe training trials, H+ monkeys exhibited the same
pattern of response latencies as normal monkeys.
In contrast to their intact performance at the 0.5-s delay,
the H+ monkeys were impaired when the delay between
sample and choice was lengthened to 30 s or more. The
impairment was observed at long delays both when the delay
trials were presented sequentially and when the delays were
presented in a mixed order. The normal monkeys exhibited
forgetting as the delay was increased, but the H+ monkeys
exhibited more forgetting than the normal monkeys. The
mixed delay condition was important because of the possibility
that, when delays are presented sequentially, a deficit
at the long delays might be due to some factor other than the
length of the delay. For example, as the delays were increased
and animals gained more experience with the task,
normal animals might have acquired strategies that enhanced
their performance while H+ animals could not. The mixed
delay condition avoids this problem. Finally, the response
latencies were virtually identical for H+ and normal monkeys
during the sample and choice phases of the test trials.
Thus, an impairment appeared selectively at long retention
intervals but not at very short retention intervals. The
impairment cannot be attributed to some perceptual or cognitive
deficit, because such a deficit should have impaired
performance at short delays as well as at long delays. Rather,
the H+ lesion appears to reproduce in monkeys a key feature
of human amnesia-namely, intact short-term memory and
impaired long-term memory. This same conclusion was
reached in an earlier study in which monkeys with large
medial temporal lobe lesions performed normally on DNMS
when training was given preoperatively and postoperative
testing was carried out at very short delays (8).
The idea that short-term memory and long-term memory
can be dissociated in experimental animals, whether by
lesions or by other methods, has been questioned recently (5,
6). This is surprising, as elegant demonstrations of this
distinction can be found in earlier work with rodents, pigeons,
and monkeys (18-21). Moreover, the distinction between
short-term and long-term memory has been well established
in the literature of human neuropsychology during
the past four decades (9, 10, 22, 23). In humans, bilateral
damage to medial temporal lobe or diencephalic structures
produces global amnesia, which spares short-term memory
but impairs the ability to establish a usable long-term mem-
ory.
It is worth considering further why the idea that medial
temporal lobe lesions selectively impair long-term memory
might seem peculiar. As Horel (ref. 6, p. 9) wrote, ".
long-term memory is part of the cortical areas where the
information to be remembered is processed and perceived."
If one takes this view, which is widely accepted (1, 23, 24),
then it might seem inconsistent to propose that the medial
temporal lobe is involved in long-term memory but not in
processing or perceiving. Stated differently, neocortical lesions
cause domain-specific information-processing deficits
(e.g., prosopagnosia, aphasia) and corresponding domainspecific
memory deficits (e.g., forfaces orwords). Moreover,
lesions of specific neocortical areas should always impair
performance at very short delays as well as at long delays,
within the domain of information processing for which that
area ofneocortex is specialized. In contrast, medial temporal
lobe lesions appear to cause global memory impairment
without any corresponding impairment in information proProc.
Natl. Acad. Sci. USA 91 (1994)
Proc. Natl. Acad. Sci. USA 91 (1994) 5641
cessing, and medial temporal lobe lesions impair performance
at long delays but not at very short delays.
The cognitive effects of medial temporal lobe lesions can
be understood in terms of the kind of computation that this
brain region appears to carry out. Perception and short-term
memory are thought to depend on coordinated activity within
the neocortex. It has been suggested that, at the time of
learning, the medial temporal lobe system establishes functional
connections with widely distributed areas of neocortex,
based on synaptic changes within this system that occur
as a part of learning (23). Medial temporal lobe structures
need to operate in concert with neocortex, if short-term
activity in neocortex is to be transformed into long-term,
permanent memory (25-28, 36). By this scenario, medial
temporal lobe damage spares short-term memory because
short-term memory can be supported by the neocortex, and
an impairmentis produced thatappears selective tolong-term
memory. The function of the medial temporal lobe is to
provide for an evolutionarily late cognitive ability-the ability
to store, retrieve, and operate on declarative knowledge
(23, 29). Because ofhow the computation is organized in the
brain, anatomically and functionally, damage to the medial
temporal lobe produces a syndrome that can be appropriately
described in terms of memory problems.
If the medial temporal lobe can be understood as having
memory functions, it should be possible to distinguish its
contribution from that ofother cortical areas. As Horelwrote
(ref. 6, p. 5), there should be "measures that differentiate [its
function] from other functions." Indeed, a review of the
behavioral effects of medial temporal lobe damage identifies
three defining features of medial temporal lobe function.
First, as shown in the current study and in an earlier one (8),
the lesion dissociates short-term and long-term recollection.
This distinction arises naturally from the kind ofcomputation
that the medial temporal lobe is involved in.
A second definingfeature ofmedial temporal lobe function
is that it is involved in memory for a limited period of time
afterlearning. Gradually overtime memory is reorganized (or
consolidated), and storage in neocortex eventually becomes
independent ofthe medial temporal lobe system. As a result,
a lesion within this system that is sufficiently delayed after
learning does not produce retrograde amnesia. Fourprospective
studies in monkey, rat, and mouse have demonstrated
this effect-i.e., temporally graded retrograde amnesia (4,
30-32). In contrast, there is no evidence for temporally
graded retrograde amnesia following a lesion in neocortex.
A third defining feature ofmedial temporal lobe function is
that damage produces memory impairment that is multimodal-i.e.,
memory is globally impaired regardless of the
sensory modality in which information is presented (15, 33).
In contrast, the memory problems associated with neocortical
lesions are domain-specific-i.e., they are specific to the
kind of material that is ordinarily processed by the damaged
area.
In summary, the present study confirms a key feature of
medial temporal lobe function in the nonhuman primate. It is
essential for long-term memory but not for short-term memory.
Following damage to the medial temporal lobe in the
monkey, short-term memory is intact,just as itis after similar
damage in humans.
This research was supported by the Medical Research Service of
the Department of Veterans Affairs, National Institutes of Health
Grant NS19063, the Office of Naval Research, the McKnight Foundation,
and the McDonnell-Pew Center for Cognitive Neuroscience.
1. Mishkin, M. (1982)Philos. Trans. R. Soc. LondonB298,85-92.
2. Squire, L. R. & Zola-Morgan, S. (1983) in The Physiological
Basis ofMemory, ed. Deutsch, J. A. (Academic, New York),
pp. 199-268.
3. Squire, L. R. & Zola-Morgan, M. (1991) Science 253, 1380-
1386.
4. Zola-Morgan, S. & Squire, L. R. (1990) Science 250, 288-290.
5. Ringo, J. L. (1991) Behav. Brain Res. 42, 123-134.
6. Horel, J. (1994) Cortex, in press.
7. Alvarez-Royo, P., Zola-Morgan, S. & Squire, L. R. (1992)
Behav. Brain Res. 52, 1-5.
8. Overman, W. H., Ormsby, G. & Mishkin, M. (1991) Exp. Brain
Res. 79, 18-24.
9. Baddeley, A. P. & Warrington, E. K. (1970) J. Verb. Learn.
Verb. Behav. 9, 176-189.
10. Cave, C. & Squire, L. R. (1992) J. Exp. Psychol. Learn. Mem.
Cognit. 18, 509-520.
11. Murray, E. A., Gaffan, D. & Mishkin, M. (1993) J. Neurosci.
13, 4549-4561.
12. Szabo, J. & Cowan, W. M. (1984) J. Comp. Neurol. 222,
265-300.
13. Hartley, L. H., Roger, R., Nicolosi, R. J. & Hartley, T. (1984)
J. Med. Primatol. 13, 183-189.
14. Zola-Morgan, S., Squire, L. R., Amaral, D. G. & Suzuki,
W. A. (1989) J. Neurosci. 9, 4355-4370.
15. Murray, E. A. & Mishkin, M. (1984)J. Neurosci. 4,2565-2580.
16. Gellerman, L. W. (1933) J. Gen. Psychol. 42, 207-208.
17. Green, D. M. & Swets, J. A. (1966) Signal Detection Theory
and Psychophysics (Wiley, New York).
18. Kesner, R. P. & Novak, J. M. (1982) Science 218, 173-175.
19. Wright, A. A., Santiago, H. C., Sands, S. F., Kendrick, D. F.
& Cook, R. G. (1985) Science 229, 287-289.
20. Winocur, G. (1985) Behav. Brain Res. 16, 135-152.
21. Nagahara, A. H. & McGaugh, J. L. (1992) Brain Res. 591,
54-61.
22. Scoville, W. B. & Miner, B. (1957) J. Neurol. Neurosurg.
Psychiatry 20, 11-21.
23. Squire, L. R. (1987) Memory and Brain (Oxford Univ. Press,
New York).
24. Damasio, A. R. (1989) Cognition 33, 25-62.
25. Squire, L. R., Shimam A. P. & Amaral, D.. G. (1989) in
Neural Models of Plasticity, eds. Byrne, J. & Berry, W.
(Academic, New York), pp. 208-239.
26. Halgren, E. (1984) in The Neuropsychology ofMemory, eds.
Squire, L. R. & Butters, N. (Guilford, New York), pp. 165-
182.
27. Teyler, T. J. & Discenna, P. (1986) Behav. Neurosci. 100,
147-154.
28. Milner, P. (1989) Neuropsychologia 27, 23-30.
29. Sherry, D. F. & Schacter, D. L. (1987) Psychol. Rev. 94,
439-454.
30. Kim, J. J. & Fanselow, M. S. (1992) Science 256, 675-677.
31. Cho, Y. H., Beracochea, D. & Jaffard, R. (1993) J. Neurosci.
13, 1759-1766.
32. Winocur, G. (1990) Behav. Brain Res. 38, 145-154.
33. Suzuki, W. A., Zola-Morgan, S., Squire, L. R. & Amaral,
D. G. (1993) J. Neurosci. 13, 2430-2451.
34. Zola-Morgan, S. & Squire, L. R. (1990) Ann. N. Y. Acad. Sci.
608, 434-450.
35. Zola-Morgan, S., Squire. L. R. & Amaral, D. G. (1989) J.
Neurosci. 9, 898-913.
36. Alvarez, P. & Squire, L. R. (1994) Proc. Nati. Acad. Sci. USA
91, in press.
Neurobiology: Alvarez et aL