BEHAVIOURAL

BRAIN
ELSEVIER Behavioural Brain Research 66 (1995) 29-36
RESEARCH

Memory and amnesia

Andrew R. Mayes*
Department ~f Clinical Neurology, University of SheffieM, N-Floor, Royal Hallamshire Hospital, Glossop Road, SheffieM SIO 2JF, UK

Accepted 15 August 1994

Abstract

Brain damage can cause several distinct disorders of explicit memory as well as several disorders of implicit memory. Organic amnesia
is the best studied explicit memory disorder. It is a syndrome that can be caused by lesions in (a) the medial temporal lobes, (b) the
midline diencephalon, or (c) the basal forebrain. It remains unresolved whether one or several functional deficits underlie the syndrome,
how these deficits should be characterised, and what is the exact location of the causal lesions. There is good evidence that amne-
sics encode information normally so their deficit(s) must be of storage or retrieval processes. If storage is disrupted, then one would
expect item-specific implicit memory for certain kinds of novel information to be disrupted in amnesics. Current evidence is unable
to indicate conclusively whether or not this prediction is met mainly because indirect memory test performance depends on explicit
as well as implicit memory. Storage deficits should also result in accelerated forgetting in amnesic patients. Studies are described which
reveal accelerated loss of free recall, but not recognition, for stories and semantically organised word lists in amnesics at delays between
15 s and 10 min. This suggests that amnesia involves a storage deficit for complex contextual associations that possibly occurs in
conjunction with one or more other functional deficits.

Key words: Anterograde amnesia; Explicit memory; Implicit memory; Forgetting rate

1. Introduction are believed to cause deficits in various forms of skill

learning and memory.
Damage in different brain regions can cause memory to It has been argued that the first four groups of memory
break down in a number of distinct ways. The resultant syndromes involve what is known as explicit or declara-
memory syndromes can be divided into the following tive memory whereas the last four groups of syndromes
groups: (a) disorders of immediate memory, (b) disorders involve what is known as implicit, procedural or non-
of previously well-established, primarily semantic memory, declarative memory (for example, see [6] and [11]). Ex-
(c) memory disorders caused by damage to the prefrontal plicit memory is present when the rememberer is aware
association neocortex, (d) the amnesic syndrome, (e) dis- that an item was encountered on one or more occasions
orders of skill learning and memory, (f) disorders of clas- in the past as applies when items are recalled or recogn-
sical conditioning, (g) disorders of non-associative kinds ised. In contrast, implicit memory is demonstrated by
of memory, such as habituation, and perhaps (h) disorders changes in behaviour and/or processing efficiency in rela-
of item-specific implicit memory (ISIM) (see [6]). This tion either to previously experienced specific items or situ-
categorisation of kinds of memory disorder assumes that ations, or to items similar to previously experienced items,
the disorders within each group have more in common without an accompanying awareness that these things have
with each other than they do with disorders in other groups. been previously encountered. Explicit memory retrieval is
For example, there is evidence to suggest that dissociable believed to be an effortful process whereas implicit memo-
deficits in immediate memory for different types of infor- ries are believed to be retrieved using relatively automatic
mation are caused by lesions to adjacent association neo- processes.
cortical regions that share similar neural architecture and It seems likely, therefore, that memory syndromes sup-
this architecture differs radically from that of the basal port the view that there are two very distinct kinds of
ganglia and cerebellum where differently located lesions memory, explicit and implicit, and these can be subdivided
into major subvarieties (such as skill memory and con-
* Corresponding author. Fax: (44) (742) 760095. ditioning), which can in turn be subdivided into further

0166-4328/95/$9.50 O 1995 Elsevier Science B.V. All rights reserved

SSDI 0 1 6 6 - 4 3 2 8 ( 9 4 ) 0 0 1 2 0 - 0
30 A.R. Mayes / Behavioural Brain Research 66 (1995) 29-36

subvarieties (such as motor, perceptual and cognitive skill imply that selective storage deficits should not occur be-
memory, and motor and affective conditioning). Not only cause if storage is disrupted so also should be encoding
will these different kinds of memory be mediated by dis- and retrieval. Organic amnesia constitutes a challenge to
tinct brain regions, but they may also depend on encod- the strong form of this dogma because many researchers
ing, storage and retrieval processes that are qualitatively believe that it arises from a storage deficit. Different theo-
distinct. All kinds of memory comprise these three pro- ries of the deficit(s) that underlie the disorder have pos-
cesses and the effectiveness of memory depends on their tulated selective failures of encoding, storage or retrieval
efficiency. as well as combinations. The major problem is deriving
With respect to explicit memory for facts and events, distinct predictions from the various hypotheses about the
incoming information has to be rapidly represented in the pattern of memory deficit that should be found in the
brain by encoding operations, an important feature of syndrome.
which is the meaningful interpretation of the input in terms
of what has already been stored by the individual. In other 2. The amnesic syndrome
words, forming a rich and meaningful representation of
incoming information involves encoding operations in As indicated above, organic amnesia is only one of the
which other information is retrieved. Variations in the ra- kinds of memory disorder that result from brain damage.
pidity and efficiency with which meaningful interpretations Like the others, it may prove to be dissociable into sev-
of inputs are achieved is a major determinant of how good eral subdisorders. Pure forms of the syndrome have four
explicit memory is in normal humans. The finding that characteristic features. The first and best known of these
intelligence correlates with explicit memory can be related is anterograde amnesia. This is an impairment in the abil-
to the likelihood that more intelligent people can process ity to recall and recognise post-morbidly experienced facts
information faster and more efficiently (see [5]). Superm- and events. In other words, it is an explicit memory deficit
nemonists have been shown to differ from those with more for facts and events experienced post-morbidly. Some re-
normal memories primarily in the richness of their encod- searchers believe that amnesics have a selective deficit in
ing of new information (see [5]). episodic memory (memory for personally experienced
It is only possible to store what has been encoded, but events) and that their difficulty in learning new facts results
little is known about the psychological factors that influ- from this, but there is no dispute about the claim that their
ence whether, and how well, encoded information is ability to learn new facts is impaired. The second feature
stored. Factors such as attention and arousal may be im- of amnesia is retrograde amnesia. This is exactly like an-
portant because there is evidence that higher levels of terograde amnesia except that the deficit in the ability to
arousal during learning are associated with slower forget- recall experienced facts and events applies to memories
ting (see [5]). Variations in attention and arousal do, how- that were acquired pre-morbidly, sometimes up to decades
ever, also influence what is encoded so it is hard to show pre-morbidly. One aspect of retrograde amnesia that has
convincingly that they also affect the efficiency with which excited considerable interest is the relative preservation of
encoded information is stored. The characteristics of stor- older pre-morbid memories in comparison with more re-
age at the neuronal level are being intensively explored, but cent ones. The remaining two defining features of the syn-
it remains unknown whether kinds of memory that are drome relate to functions that are preserved. Thus, the
mediated by distinct brain systems depend on different third feature is the preservation of intelligence and the
neuronal storage mechanisms. fourth feature is the preservation of immediate memory as
Retrieval involves searching for stored information and shown by patients' normal ability to repeat back a string
may involve planning because the search often requires of spoken digits in the correct order (see 11). The fact that
finding cues that automatically activate the memory. The intelligence in patients may be unaffected suggests that
cues that automatically activate a memory are those en- they should be able to process information as effectively
coded when the memory was originally acquired and re- as non-amnesic people, and their preserved immediate
encoding them may involve a difficult search with false memory implies that encoded information is retained nor-
trails. It is believed that patients with frontal association mally at least for a few seconds. It would, therefore, seem
neocortex lesions have planning deficits which can not unlikely that deficient encoding is the major cause of the
only disrupt elaborative encoding, but also impair retrieval syndrome.
when 'automatic' cues are not readily available (see [6]). Amnesia can be caused by brain lesions in any one of
It is widely believed that information is stored where it three brain regions. The critically affected regions include:
is represented in the brain. This dogma acts as a healthy (a) the medial temporal lobes, (b) the midline diencepha-
corrective to the view that there are specialised storage Ion, and (c) the basal forebrain (see [6,11,13]). It has also
regions in the brain, but, in its strongest form, seems to been argued by Mishkin on the basis of monkey models
 A.R. Mayes / Behavioural Brain Research 66 (1995) 29-36 31

of amnesia that lesions of the ventromedial frontal cortex lised brain damage and also by investigating the effects
can cause amnesia (see [6]). Currently, there is no evi- mainly on recognition memory of specific brain lesions in
dence in humans to support this claim, perhaps because monkeys (and sometimes other animals like rats). Most
lesions rarely selectively destroy this large frontal site. work has been done with medial temporal lobe amnesia.
Many frontally lesioned patients do not, however, show Research with monkey models of amnesia has had a com-
amnesia although they sometimes have different kinds of plex history, but there is now a growing consensus which
memory deficits (see [6]). suggests that: (a) severe recognition memory deficits are
There is still much controversy about the exact location caused by lesions to the perirhinal and parahippocampal
of the critical lesions within each of the key regions. Pre- cortices; (b) selective amygdala lesions do not contribute
cise knowledge is important if the syndrome is to be simu- to amnesia although they do cause other kinds of memory
lated in a heuristically valuable way by neural network deficit; and (c) hippocampal lesions do not cause severe
models that have architectures constrained by the neu- amnesia, but may cause a milder version of the syndrome
roanatomy of the syndrome. Precise knowledge of the (see [ 11,13]). The last conclusion is the most controver-
syndrome's neuroanatomy is also important in helping to sial because it has been claimed recently that visual re-
determine whether the syndrome dissociates into several cognition memory is unaffected over delays of between a
subdisorders caused by lesions to separate brain regions. half and three and a half minutes following excitotoxic
If the structures, damage to which causes amnesia, are lesions of the hippocampus [8]. The evidence from hu-
serially linked, then it is likely that the syndrome is unitary mans is much scantier because of the rarity of selective
even if different parts of the critical circuit perform differ- lesions, but there is evidence that perirhinal and parahip-
ent operations because damage to any part of the circuit pocampal lesions cause very severe amnesia whereas am-
should disrupt the whole circuit's function in a qualita- nesia is relatively mild in patients who have damage largely
tively similar way. Conversely, if the structures, damage to confined to the hippocampus (see [ 10,13]).Very few cases
which causes amnesia, are not closely interconnected, then with selective amygdala damage have ever been reported,
it is more likely that the syndrome is functionally hetero- but the few case studies available do suggest that bilateral
geneous. Evidence that amnesia can be caused by lesions amygdala damage may disrupt recall and recognition of
to structures such as the fornix which interconnect the non-verbal materials (see [12]). If this is true, then one
main regions implicated in amnesia is clearly important in would expect that amnesic patients with otherwise sym-
helping to discriminate between these possibilities (see metrical lesions and amygdala damage should show more
[6,11,13]). severe explicit memory deficits for non-verbal materials.
Research into organic amnesia has focused on three key Knowledge of the precise location of lesions in the mid-
and related issues. The first of these issues is whether the line diencephalon and basal forebrain that cause amnesia
syndrome comprises one or several functional deficits. The is currently less good. This is partly because the midline
second issue concerns the precise nature of this or these thalamic nuclei are small and close together. Nevertheless,
functional deficits, and the third issue relates to the exact there is evidence that anterior thalamic lesions cause am-
location of the lesions that cause the syndrome or differ- nesia both in humans and other mammals, and lesions to
ent components of it. If the syndrome comprises several other midline thalamic nuclei, particularly those that re-
subdisorders, then it should be possible to find dissociat- ceive connections via the internal medullary lamina (see
ing patterns of symptoms in patients with lesions in dif- [11,13]) have also been implicated in amnesia. It remains
ferent locations. For example, it has been suggested that unclear to what extent mammillary body lesions disrupt
anterograde amnesia and at least the severe form of ret- explicit memory although a recent case study suggests that
rograde amnesia with a prolonged temporal gradient are free recall may be impaired, but not recognition [2]. In the
caused by different underlying functional deficits. If so, it basal forebrain, there is evidence that lesions of the sep-
should be possible to find patients with selective retro- tum and pathways between the septum and the hippoc-
grade amnesia and other patients who primarily show an- ampus can cause amnesia [1]. Lesions to the structures
terograde amnesia accompanied by a retrograde amnesia that interconnect midline diencephalon and the medial
that only affects memories acquired up to a short period lobe temporal lobes also appear to cause mild amnesia as
before brain damage occurred (see [6,11 ]). it has been reported that lesions of both the fornix and the
retrosplenial cortex (which reciprocally connects anterior
thalamus and hippocampus) cause mild memory deficits
3. Some comments on the neuroanatomy of amnesia (see [6,11]).
Amnesia can only result from a single functional defi-
The neuroanatomy of amnesia has been investigated by cit if the above structures can be linked into a serial cir-
trying to find amnesics who have very selective and loca- cuit. One major problem in doing this is explaining why
32 A.R. Mayes / Behavioural Brain Research 66 (1995) 29-36

parahippocampal and perirhinal cortex lesions cause a One kind of hypothesis about amnesia that has been
much more severe anmesia than do hippocampal lesions. popular is that patients do not encode inputs so richly and
One explanation of this puzzle is that the following circuits meaningfully as do normal people. As is known from nor-
are implicated in amnesia: (a) the best known circuit runs mal people, the effect of this will be poor explicit memory.
from the parahippocampal and perirhinal cortices to the There are a number of problems for this type of view. First,
hippocampus, from there via the fornix to the anterior it is unlikely that poor semantic encoding can cause such
nucleus of the thalamus both directly and also indirectly severe deficits in explicit memory as are seen in some
via the mammillary bodies. The circuit may be completed amnesics like H.M. Second, it is unclear why patients with
back to the medial temporal lobes using several routes preserved intelligence should fail to access the meaning of
including the retrosplenial cortex, the cingulate cortex and incoming information as well as normal people. This kind
possibly another pathway; (b) a circuit running from the of problem may arise following lesions to the frontal lobes
parahippocampal and perirhinal cortices via the hippoc- of the brain, but such lesions do not usually cause amnesia
ampus and then back to these polsensory cortices; (c) a (see [6]). Third, there is no reason why an encoding
circuit running from the parahippocampal and perirhinal problem should cause explicit memory deficits for memo-
cortices to midline thalamic nuclei that are not in the ries acquired up to decades before amnesia developed.
hippocampal circuit, and from there perhaps via the fron- This difficulty might be overcome if a semantic process-
tal cortex back into the medial temporal lobe area; (d) the ing deficit were postulated which affected retrieval as
above circuits may be modulated by reciprocal interac- well.
tions with basal forebraln structures. Currently, the most popular kind of account of amne-
In most amnesic patients two or more of these circuits sia is that it is caused by some kind of storage failure.
will be disrupted so their explicit memory may be dis- Storage accounts may differ from each in two kinds of
rupted in similar ways. If the multiple circuit account is ways. First, the kind of storage deficit postulated may
correct, however, there may be up to three kinds of func- differ. If fact and event memories are largely stored in
tional deficit underlying amnesia if circuits (a) to (c) have neocortical locations, the structures involved in amnesia
measurably different functions. Basal forebraln lesions could play a role in modulating the storage processes that
should decrease the efficiency with which these circuits occur at these locations. This view receives some support
operate and disrupt each of their functions to varying from the finding that Korsakoff amnesics who have struc-
degrees. Nevertheless, the multiple circuit account is not tural damage in the midline diencephalon show a wide-
necessarily correct. For example, parahippocampal and spread reduction in metabolic activity in the neocortex [9].
perirhinal cortex lesions may cause more severe amnesia A more influential kind of storage deficit account postu-
than hippocampal lesions because they destroy the back- lates that a key aspect of storage occurs first in the struc-
projections of these cortices to association neocortex tures implicated in amnesia, and particularly the hippoc-
whereas hippocampal lesions will only disturb what these ampus, but that with the passage of time, storage processes
polysensory cortices backproject. If this is correct, then are reorganised so that they come increasingly under neo-
amnesia would only involve one neural network with le- cortical control and independent of the hippocampus and
sions only disrupting one function. unaffected by amnesia (see [ 11 ]). The second kind of way
in which storage deficit accounts can differ relates to ex-
actly what information they postulate the storage deficit
4. Theories of amnesia and the predictions they make applies. Although earlier hypotheses seemed to imply that
all aspects of fact and event information were affected
The lesion evidence does not, therefore, clearly discrimi- equally, more recent hypotheses have suggested that some
nate between whether one or several functional deficits fact and event information may be stored normally, and
underlie amnesia. Evidence from the pattern of memory that the storage deficit affects the formation of particular
disruption is also not conclusive although there is some kinds of association such as that between attended items
evidence that retrograde and anterograde amnesia can be (targets) and the context in which they were experienced
at least partially dissociated from each other (see [6]) and [3,6,11].
some amnesics seem to show free recall deficits, but rela- The final kind of hypothesis about the functional defi-
tively normal recognition (for example, see [2]). Theories cit underlying amnesia is that patients encode and store
about the functional deficit underlying amnesia have information just like normal people, but have a deficit of
tended to ignore the possibility that there may be several retrieval that prevents them from achieving explicit me-
deficits confounded with each other. The theories can be mory of facts and events. One factor that makes this kind
grouped according to whether they postulate encoding, of hypothesis relatively implausible is that the processes
storage, or retrieval deficits. involved in encoding and retrieval show a considerable
 A.R. Mayes / Behavioural Brain Research 66 (1995) 29-36 33

amount of overlap with each other so that a selective does an impairment in their ability to encode associations
deficit of one or the other is unlikely. between targets and features of their encoding context
The three kinds of hypotheses make distinctive predic- such as the targets' spatial location. In order to test these
tions in several areas. First, only encoding deficit hypoth- possibilities, my colleagues and I [7] presented a group of
eses predict that patients will not show preserved ability amnesics of several different aetiologies with a series of
to answer questions about briefly presented, complex ma- coloured pictures for 6 s each. Every picture contained
terials, such as pictures, immediately after presentation. drawings of six objects in different positions on a card.
This prediction is based on the assumption that even nor- Each object had an unpredictable colour and size. Imme-
mal people cannot process all aspects of the stimuli when diately after each picture was withdrawn, subjects were
they are briefly presented and that answers to questions asked one unpredictable question related to (a) the colour
given immediately after presentation are held in an imme- of one of the objects, (b) the relative size of the object
diate memory which is normal in amnesics. drawings, (c) the location of the objects, or (d) semantic
The second prediction is based on the distinction be- features of the objects. The duration of object exposure
tween ISIM and explicit memory. It is possible to con- had been titrated so that normal subjects made quite a few
struct tests that enable one to compare explicit and im- errors. The performance of the amnesics was identical to
plicit memory for the same material. Although it is possible that of their controls indicating that they were encoding
that explicit and implicit memory typically tap memory for the tested semantic, colour, size, and spatial location fea-
slightly different aspects of the material, it seems highly tures as well as the normal subjects. These results fit with
likely that explicit memory for the information tapped into earlier work which found that the recognition memory of
by the implicit memory tests would be impaired in pa- amnesic patients improves as much as, but no more than
tients. It also seems highly likely that explicit and implicit that of normal people when subjects are instructed to
memory for the same information will be based on the encode semantic features of various materials as opposed
same memory representations. It therefore follows that if to encoding the materials in the way that they do sponta-
amnesics fail to encode or store specific information, then neously (see [6]). Both sets of results suggest that amne-
their implicit as well as their explicit memory for that new sics encode facts and events normally.
information should be disrupted. In contrast, if retrieval of
explicit memories is selectively disrupted in amnesics, then 4.2. Prediction two." implicit memory
they should show preservation of implicit memory for all
forms of novel information. There is good evidence that amnesics show preservation
Third, if storage is impaired, the consolidation of infor- of several forms of implicit memory including classical
mation into long-term memory should not be occurring conditioning for eye blink responses, and skill learning and
normally in patients. As consolidation takes some time to memory for various kinds of motor, perceptual and cog-
complete during which time memory performance is pre- nitive responses (see [6,11 ]). But there is reason to believe
sumably becoming progressively more dependent on long- that memory for these forms of implicit memory is medi-
term storage, then one would expect amnesics' explicit ated by the basal ganglia and cerebellum, structures which
memory to become progressively more impaired. One are not implicated in amnesia. In other words, these are
would therefore expect amnesics to show accelerated for- forms of memory that are stored in regions of the brain
getting for those kinds of information which they fail to which are not involved in storing memories about facts
store normally. The period of time over which this occurs and events. But there is also evidence that amnesics show
would depend on the time course of the affected consoli- normal ISIM for information that was already familiar
dation processes, which is unknown. This prediction before the training episode (see [4,6,10). This effect is
should apply not only if it is postulated that amnesia af- found not only with previously familiar verbal material like
fects storage modulation, but also if it is postulated that words, but also with previously familiar non-verbal mate-
aspects of fact and event information are initially stored rial like the faces of famous people. But some researchers
in structures like the hippocampus because if these struc- believe that ISIM for such items depends on the prolonged
tures are damaged, then consolidation will (at the very activation of already existing memories rather than on the
least) be impaired. creation of new memories (see [4,6,11 ]). If they are cor-
rect, then the preservation of implicit memory for already
4.1. Prediction one." encoding familiar items does not provide evidence for normal stor-
age processes in amnesics. Preservation of ISIM for in-
Although amnesics' preserved intelligence makes it un- formation that was novel prior to training would do so, but
likely that they fail to process inputs in as rich and mean- the evidence that amnesics show normal implicit memory
ingful way as normal people, this remains a possibility as for such information is much more equivocal. Indeed,
34 A.R. Mayes / Behavioural Brain Research 66 (1995) 29-36

some researchers now believe that amnesics may not show a storage deficit for some or all aspects of fact and event
normal ISIM for novel associations, particularly when information.
these involve semantic representations (see [4] passim). If
they are correct, then amnesics do not store such asso- 4.3. Prediction three:forgetting rate in amnesia
ciations normally.
It is best to regard the current literature involving ISIM Studies of the rate at which amnesics forget have been
for novel information in amnesia as inconclusive for the quite numerous, but nearly all of them have focused on
following reason. Whereas explicit memory is tapped in recognition and examined forgetting between delays of 10
tasks that make a direct reference to the assessment of min or more after learning and 1 week or more (see [ 6,11 ]).
memory, ISIM is tapped using indirect memory tasks in Interpretation of the findings from these studies is made
which no reference is made to memory. An attempt is difficult because the studies all suffer from an artefact that
usually also made to disguise the fact that test items have underestimates the rate at which amnesics forget relative
been encountered before during the earlier stage of the to their controls (see [6]). Nevertheless, there is no good
task. Although a few years ago it was assumed that per- evidence that amnesics, regardless of where their lesions
formance on indirect memory tasks depended solely on are, lose the ability to recognise recently experienced ma-
implicit memory, this position has now been challenged by terials at delays between 10 min and 1 week or longer. If
workers who believe that performance also depends on a this conclusion were correct and generalised to other forms
contribution from explicit memory that may be either fa- of explicit forgetting, then one would have some difficulty
cilitatory or inhibitory (see [11 ]). As normal people's ex- defending the view that amnesics fall to store some or all
plicit memory is superior to that of amnesics, it is hard to aspects of fact and event information. The extant studies
know whether amnesics' performance on indirect memory are, however, limited not only because they are all con-
tasks reflects impaired or normal implicit memory. founded by an artefact, but also because they (a) fall to
In order to obtain interpretable results it is important to examine the rate at which amnesics lose the ability to free
use several approaches designed to minimise the contri- recall recently experienced materials, and (b) fail to exam-
bution of explicit memory or to control for it. If these ine forgetting rate at delays less than 10 min.
approaches depend on different assumptions and the re- In unpublished work, Isaac and I have examined the
suits found using them are consistent, then one can have rate at which a mixed group of patients with relatively mild
confidence in one's conclusions. Four approaches are ap- amnesia lost the ability to free recall and recognise short
propriate: (a) the use of indirect memory tasks where ISIM stories at delays between 15 or 20 s and 10 min. In order
is revealed by reduction in reaction times to repeated items. to avoid floor, ceiling and scaling effects in this kind of
This measure of ISIM is unlikely to be influenced by ex- study, it was necessary to match amnesic and control
plicit memory because this is believed to be a relatively memory levels at the shortest delay used by giving the
slow process. (b) Dividing attention at test can reduce the patients more learning opportunity. We achieved this by
contribution of explicit memory which is effortful, but has reading the stories five times on average to the amnesics
a minimal impact on explicit memory, which is believed to and only once to their control subjects. Relatively mild
depend on automatic retrieval processes. (c) Autonomic amnesics were used because it would have been impos-
responses can be used to reveal distinctive responses to sible to match the memory of severe amnesics to that of
repeated stimuli. These may be provide a good measure of normal people after a 15-s delay. To prevent rehearsal
ISIM because such responses are believed to be automati- during the 15- or 20-s delay (these different delays were
cally produced. In unpublished work, Diamond and I have used in two separate experiments) subjects were required
found that a mixed group of amnesics showed normal to count back by threes from a large number as fast as they
levels of autonomic discrimination for repeated words de- could. Similar filler activities were also used to prevent
spite showing very impaired recognition for the same rehearsal at the longer delays. Each delay was tested with
words, which suggests that autonomic measures should a different story and recognition was tested using different
provide a good means of checking whether amnesics show stories from those used for testing free recall. Recognition
preservation of ISIM for novel associations. (d) Finally, for each story was assessed using 12 forced-choice four-
a procedure, developed by Jacoby (see [11 ]) for directly alternative questions. Pilot work was used to ensure that
assessing the contribution of explicit and implicit memory the stories were of equal difficulty.
to task performance can be modified to give estimates of We succeeded in matching amnesic and control recog-
response bias as well as memory strength. If amnesics nition and recall levels at the shortest delay used. After 10
seem to show normal levels o f l S I M for novel information, min, the amnesics showed a normal rate of loss of recog-
and novel associations in particular, with all the above nition, but they showed markedly accelerated loss of the
approaches, then it would be hard to argue that they have ability to free recall the stories. This effect was found to
 A.R. Mayes / Behavioural Brain Research 66 (1995) 29-36 35

a similar degree in amnesics regardless of where their le- dered so that words from different categories were inter-
sions were, although there was some evidence that the mixed. The third kind comprised words similarly grouped
Korsakoff patients (who have midline diencephalic le- into a few semantic categories and organised so that words
sions) showed their accelerated loss of free recall ability from the same category were next to each other.
earlier than did amnesics with medial temporal lobe or The results were striking. The amnesics showed accel-
basal forebrain lesions. An additional experiment showed erated free recall loss of the last two kinds of word list in
that the apparently accelerated amnesic loss of free recall which there was some degree of meaningful organisation,
ability was not an artefact of the matching procedure. Nor but they showed a normal rate of loss of free recall for the
did it arise from surreptitious rehearsal by the normal kind of word list in which the words were unrelated to
subjects during the ten minute delay because a further each other. Unlike normal subjects, amnesics did not show
control experiment showed that when the delayed free slower forgetting of meaningfully organised word lists than
recall test was given as a surprise to the normal subjects, lists of unrelated words. Cued recall in which subjects
they recalled as much as did the normal subjects in the were given either the first three letters of words as cues or
main experiments. It might be argued that immediate me- told to what semantic category they belonged was lost at
mory, which is preserved in amnesics, was making a small a normal rate by the amnesics. The finding of a normal rate
contribution to patients' free recall after a 15 second delay, of loss of free recall and recognition ability for unrelated
but that recall at the 10-min delay was entirely mediated lists of words replicated what has already been reported
by long-term memory at which they are impaired. If this (see [ 11 ]).
argument were correct, then the accelerated forgetting
shown by the amnesics would be a trivial consequence of
the fact that immediate memory contributed to free recall 5. Conclusion
at the shortest delay but not the longest. To test this pos-
sibility, three severely amnesic patients with normal im- To the extent that they have been examined, tests of the
mediate memory were read stories five times and tested three predictions suggest several things about the func-
after the 15-s delay. These patients recalled practically tional deficit(s) that underlie amnesia. First, the memory
nothing at this delay, which indicates that immediate me- deficits are very unlikely to be caused by a disruption of
mory makes a negligible contribution to recall of stories encoding processes so the underlying deficit(s) must con-
after a filled 15-s delay. cern either storage or retrieval or both. Second, the cur-
Accelerated loss of free recall ability in the amnesics rent evidence about ISIM for novel information is equivo-
does not necessarily support the view that they fail to store cal and cannot be used to discriminate between storage
some aspects of fact and event information, because the and retrieval deficit views of amnesia so more systematic
phenomenon could result from excessive sensitivity to in- studies need to be undertaken. Third, results with experi-
terference. If the retrieval process is deranged so that it is ments that examine amnesic forgetting rates over short
less good at discriminating between rival memories, then delays suggest that patients may be suffering from a stor-
amnesics might show accelerated forgetting as interference age problem for some aspects of fact and event informa-
from new material builds up following learning. In order tion. Recall of both stories and organised lists of words
to test this possibility, subjects were read three very similar involves the retrieval of complex associations between sev-
stories to the target story either before or after presenta- eral discrete items and their encoding context.. Recall of
tion of the target story. When this was done, both free lists of unrelated words involves the retrieval of simpler
recall and recognition were slightly impaired both in the associations between individual items and their encoding
patients and their control subjects, but the effect was not context because normal subjects find it very hard to form
bigger in the patients than in the normal subjects. It would links between unrelated words after a brief learning expo-
seem, therefore, that the accelerated loss of free recall sure. Recognition of all these materials only requires the
ability in the amnesics is likely to have been caused by a retrieval of simple associations between target items and
storage deficit. their encoding context. It may be, therefore, that amnesia
We wished to determine whether the accelerated loss of involves a storage deficit that affects complex associations
free recall effect occurred with other kinds of material as involving more than two items which may obligatorily in-
well as stories so we examined loss of free recall, cued clude the encoding context. This hypothesis needs to be
recall and recognition for different kinds of word list. Three systematically explored.
kinds of word list were used. The first kind comprised 20 The hypothesis is not, however, a complete account of
words each drawn from a different semantic category. The amnesia because there are two features of the syndrome
second kind comprised words that were grouped into a that it cannot explain. The first feature is that most am-
few semantic categories. These words were randomly or- nesics (but perhaps not those with selective lesions to the
36 A.R. Mayes / Behavioural Brain Research 66 (1995) 29-36

h i p p o c a m p u s or other parts of the hippocampal circuit ity in the association neocortex and cause a retrieval defi-
downstream from the hippocampus [10]) show impaired cit for pre-morbid and probably post-morbid memories.
recall and recognition for simple as well as complex as- Future research that examines the memory of patients
sociations after filled delays of 15 s or less. The second with very selective brain lesions will be needed to test
feature is that m a n y amnesics showed temporally graded whether these hypothetical deficits exist.
retrograde amnesia that extends back over several de-
cades.
One explanation of these features is that the storage
deficit hypothesis just outlined is wrong. An alternative References
explanation is that amnesia is not a unitary deficit and
[ 1] Cramon, D.Y. von and Schuri, U. The septo-hippocampal path-
comprises at least three independent deficits. The first ways and their relevance to human memory: a case report, Cortex,
kind of deficit would be a storage failure for complex as- 28 (1992) 411-422.
sociations that is not apparent immediately after learning [2] Dusoir, H., Kapur, N., Byrnes, D.P., McKinstry, S. and Hoare,
R.D., The role of diencephalic pathology in human memory dis-
but develops over a period of perhaps 10 min. There is
order: evidence from a penetrating paranasal brain injury, Brain,
some evidence that explicit memory deficits may not be 113 (1990) 1695-1706.
present until a minute or so has passed following learning [3] Eichenbaum, H., Otto, T. and Cohen, N.J., Two functional com-
in patients with mammillary body lesions [2] and selective ponents of the hippocampal memory system, Behav. Brain Sci., in
press.
hippocampal lesions associated with temporal lobe epi- [4] Graf, P. and Masson, M.E.J., ImplieitMemory, Lawrence Erlbaum,
lepsy [10]. It may be, then, that damage to the hippo- Hillsdale, NJ, 1993.
campal circuit is responsible for this kind of deficit. The [5] Lloyd, P., Mayes, A.R., Manstead, A.S.R., Meudell, P.R. and
Wagner, H.L., Introduction to Psychology: An Integrated Approach,
second kind of deficit might be described as an interme- Fontana, London, 1984.
diate m e m o r y deficit for single items or simple associa- [6] Mayes, A.R., Human Organic Memory Disorders, Cambridge Uni-
tions. If such an intermediate m e m o r y exists and its neural versity Press, Cambridge, 1988.
[7] Mayes, A.R., Downes, J.J., Shoqeirat, M., Hall, C. and Sagar,
substrate is disrupted, one would expect to see an explicit
H.J., Encoding ability is preserved in amnesia: evidence from a
m e m o r y deficit after a very few seconds that would not be direct test of encoding, Neuropsyehologia, 31 (1993) 745-759.
further exacerbated as time passes. Interestingly, it has [8] Murray, E.A. and Mishkin, M., Effects of excitoxic amygdalo-
recently been suggested that the parahippocampal and hippocampal lesions on visual recognition in rhesus monkeys, Soc.
Neurosci. Abstr., 19, (1993) 438.
perirhinal cortices may be the neural substrate for an in- [9] Paller, K.A., Richardson, B.C., Shimamura, A.P., Plaisant, O.,
termediate m e m o r y system [3]. The third kind of deficit Reed, B.R. and lagust, W.I., Neurophysiological substrates of
would be a disruption o f the storage of premorbidly es- human memory impairments: altered regional cerebral glucose uti-
lization in alcoholic Korsakoffs syndrome and Alzheimer's dis-
tablished memories. If this kind of deficit exists, its neu- ease, as measured by positron emission tomography (PET), Soc.
ral underpinnings are a matter of speculation but probably Neurosci. Abstr., 19 (1993) 1078.
include the association neocortex o f the lateral temporal [ 10] Pigott, S. and Milner, B., Memory for different aspects of complex
visual scenes after unilateral temporal- or frontal-lobe resection,
cortex and, just possibly, the polysensory cortices of the
Neuropsychologia, 31 (1993) 1- 15.
medial temporal cortex. The first kind of storage deficit [11] Squire, L.R., Knowlton, B. and Musen, G., The structure and
may also be associated with some retrograde amnesia, but organization of memory, Annu. Rev. Psychol., 44 (1993) 453-495.
this would probably only extend back a few weeks or [12] Tranel, D. and Hyman, B.T., Neuropsychological correlates of
bilateral amygdala damage, Arch. Neurol., 47 (1990) 349-355.
months into the pre-morbid period. Finally, it is also pos- [ 13] Zola-Morgan, S. and Squire, L.R., The Neuroanatomy of Amne-
sible that some midline diencephalic lesions disturb activ- sia, Annu. Rev. Neurosci., 16 (1993) 547-563.