Bypassing the hippocampus:

thunderingaardvarkΤεχνίτη Νοημοσύνη και Ρομποτική

18 Νοε 2013 (πριν από 3 χρόνια και 10 μήνες)

309 εμφανίσεις










Bypassing the
hippocampus
:


Rapid
Neocortical
Acquisition of Long Term Arbitrary Associations via Fast Mapping


Tali Sharon



A THESIS SUBMITTED FOR THE DEGREE

“DOCTOR OF PHILOSOPHY”




University of Haifa

Faculty of

Social Sciences

Department of

Psychology





November
,
20
10






2








Bypassing the
hippocampus
:

Rapid
Neocortical
Acquisition of Long Term Arbitrary Associations via Fast Mapping



By:
Tali
Sharon


Supervised by:
Dr. Asaf Gilboa
and

Prof. Rachel Tomer



A THESIS SUBMITTED FOR THE DEGREE

“DOCTOR OF PHILOSOPHY”



University of Haifa

Faculty of Social Sciences

Departme
nt of Psychology




November, 20
10



Recommended by: __

Dr. Asaf Gilboa
______
_
__ Date: ___________________


(Advisor)


Recommended by: __
Prof. Rachel Tomer _______
Date:
___________________


(Advisor)


Approved by:
__ Prof. Rachel Tomer _______
Date: ___________________


(Chairman of Ph.D Committee)





3


ות תרכה ףד
טופישה ירחא בתכי הד

















































4

Table of Contents

ABSTRACT

................................
................................
................................
................................
..........................

7

LIST OF TABLES

................................
................................
................................
................................
................

13

LIST OF FIGURE
S

................................
................................
................................
................................
..............

14

INTRODUCTION

................................
................................
................................
................................
...............

15

A
NTEROGRADE
A
MNESIA AND MEMORY SU
BSYSTEMS

................................
................................
.............................

15

E
E
E
N
N
N
T
T
T

................................
................................
................................
................................
..............................

17

S
S
S
U
U
U
B
B
B

................................
................................
................................
................................
..............................

17

H
H
H
I
I
I

................................
................................
................................
................................
................................
.

17

A
A
A
T
T
T
H
H
H

................................
................................
................................
................................
................................
..

17

M
M
M
D
D
D

................................
................................
................................
................................
..............................

17

V
V
V
L
L
L

................................
................................
................................
................................
................................

17

F
F
F
O
O
O
R
R
R
N
N
N
I
I
I
X
X
X



&
&
&



M
M
M
B
B
B

................................
................................
................................
................................
................

17

P
P
P
P
P
P

................................
................................
................................
................................
................................

17

F
F
F
O
O
O
R
R
R
N
N
N
I
I
I
X
X
X

................................
................................
................................
................................
..........................

17

D
ECLARATIVE
M
EMORY

................................
................................
................................
................................
....

18

D
ECLARATIVE MEMORY SY
STEMS AND CONSOLIDAT
ION

................................
................................
...........................

20

N
OVEL SEMANTIC LEARNI
NG IN AMNESIA

................................
................................
................................
..............

23

N
EUROCOGNITIVE MECHAN
ISMS OF NEW LEARNING

IN AMNESIA

................................
................................
...............

28

The method of vanishing cues

................................
................................
................................
................

28

Errorless learning

................................
................................
................................
................................
...

29

Varied errorless learning

................................
................................
................................
........................

30

Anchoring new information to existing mental representations

................................
...........................

31

Common ground

................................
................................
................................
................................
....

33

S
UMMARY

................................
................................
................................
................................
......................

34

F
AST MAPPING

................................
................................
................................
................................
................

35

What is fast map
ping?

................................
................................
................................
...........................

35

Fast mapping in adults

................................
................................
................................
...........................

39

T
HE PRESENT STUDY

................................
................................
................................
................................
.........

43

PART I


F
AST MAPPING IN ADULT
S: A BEHAVIORAL STUD
Y

................................
................................
........

49

M
ETHODS

................................
................................
................................
................................
......................

53

Participants

................................
................................
................................
................................
............

53

Stimuli

................................
................................
................................
................................
....................

54

Procedure

................................
................................
................................
................................
...............

56

Hypotheses

................................
................................
................................
................................
.............

61

Task performance
................................
................................
................................
................................
...............

61

Memory performance

................................
................................
................................
................................
........

62

R
ESULTS

................................
................................
................................
................................
........................

64

Task performance

................................
................................
................................
................................
..

64

RT's

................................
................................
................................
................................
................................
.....

64

Accuracy

................................
................................
................................
................................
.............................

65

Memory performance

................................
................................
................................
............................

66

Free and cued recall

................................
................................
................................
................................
...........

66

Forced choice associative recognition

................................
................................
................................
................

74

D
ISCUSSION

................................
................................
................................
................................
....................

78

PART II

NEW LEARNING IN AMNE
SIA VIA FAST MAPPING
: A PATIENT STUDY

................................
...........

90

A
.

P
ILOT STUDY

................................
................................
................................
................................
...............

95

M
ETHODS

................................
................................
................................
................................
......................

95

Participants

................................
................................
................................
................................
............

95

Stimuli

................................
................................
................................
................................
....................

95

Procedure

................................
................................
................................
................................
...............

97


5

Hypotheses

................................
................................
................................
................................
...........

100

FM performance
................................
................................
................................
................................
...............

100

Memory performance

................................
................................
................................
................................
......

100

R
ESULTS

................................
................................
................................
................................
......................

102

FM performance
................................
................................
................................
................................
...............

102

RT's

................................
................................
................................
................................
................................
...

102

Choice of picture refe
rent

................................
................................
..........

Error! Bookmark not defined.

Memory performance

................................
................................
................................
..........................

103

D
ISCUSSION

................................
................................
................................
................................
..................

105

B
.

A

PATIENTS ST
UDY

................................
................................
................................
................................
......

108

M
ETHODS

................................
................................
................................
................................
....................

108

Participants

................................
................................
................................
................................
..........

108

Patients' MRI Volumetric analysis

................................
................................
................................
....................

109

Stimuli

................................
................................
................................
................................
..................

119

Procedure

................................
................................
................................
................................
.............

122

R
ESULTS

................................
................................
................................
................................
......................

127

Task performan
ce

................................
................................
................................
................................

127

RT's

................................
................................
................................
................................
................................
...

127

Choice of picture referent

................................
................................
......................

Error! Bookmark not defined.

Recognition

................................
................................
................................

Error!
Bookmark not defined.

D
ISCUSSION

................................
................................
................................
................................
..................

136

M
ETHODS

................................
................................
................................
................................
....................

147

Participants

................................
................................
................................
................................
..........

147

Stimuli

................................
................................
................................
................................
..................

147

Procedure

................................
................................
................................
................................
.............

148

MRI data acquisiti
on

................................
................................
................................
............................

152

Preprocessing and data analysis

................................
................................
................................
..........

152

R
ESULTS

................................
................................
................................
................................
......................

157

Behavioral data

................................
................................
................................
................................
....

157

Study performance

................................
................................
................................
................................
...........

157

Memory performance

................................
................................
................................
................................
......

157

Imaging
data

................................
................................
................................
................................
........

158

Novel
-
base line

................................
................................
................................
................................
.................

158

Novel
-
familiar

................................
................................
................................
................................
...................

162

Subsequent m
emory

................................
................................
................................
................................
........

164

D
ISCUSSION

................................
................................
................................
................................
..................

168

Novel
-
base line

................................
................................
................................
................................
.................

169

Novel
-
famili
ar

................................
................................
................................
................................
...................

172

Subsequent memory

................................
................................
................................
................................
........

174

GENERAL DISCUSSION

................................
................................
................................
................................
..

180

REFERENCES

................................
................................
................................
................................
..................

186

APPENDICES

................................
................................
................................
................................
..................

210

A
PPENDIX
1



EXPERIMENT
1,

LIST OF STIMULI

................................
................................
................................
....

210

A
PPENDIX
2



EXPERIMENT
1,

CONSENT FORM

................................
................................
................................
....

214

A
PPENDIX
3



EXPERIMENT
1,

FREE RECALL TASK FOR
M
,

NO CONTEXT VERSION

................................
..........................

215

A
PPENDIX
4



EXPERIMENT
1,

FREE RECALL TASK FOR
M
,

CONTEXT VERSION

................................
...............................

216

A
PPENDIX
5



EXPERIMENT
1,

CUED RECALL
TASK FORM
,

NO CONTEXT VERSION

................................
.........................

217

A
PPENDIX
6



EXPERIMENT
1,

CUED RECALL TASK FOR
M
,

CONTEXT VERSION

................................
..............................

219

A
PPENDIX
7



EXPERIMENT
2
A
,

STIMULI
................................
................................
................................
.............

222

A
PPENDIX
8



EXPERIMENT
2
A
,

ANALYSIS WITHOUT CON
SIDERATION OF THE ER
RONEOUS STUDY TRIALS

.........................

228

A
PPENDIX
9



EXPERIMENT
2
B
,

PATIENTS TABLE

................................
................................
................................
..

229

A
PPENDIX
10



EXPERIMENT
2
B
,

PILOT
E
NGLISH
FM

STUDY

................................
................................
..................

231

A
PPENDIX
11



EXPERIMENT
2
B
,

MEASURES OF CONFIDEN
CE RATING
S AND SEMANTIC CATEG
ORIZATION

........................

234

A
PPENDIX
1
2



EXPERIMENT
3,

EXPERIMENTAL PROTOCO
L
,

IMAGING STUDY

................................
.............................

236

A
PPENDIX
11



EXPERIMENT
3,

METAL SCREENING QUES
TIONNAIRE

................................
................................
........

241

A
PPENDIX
1
4



EXPERIMENT
3,

'
NOVEL
-
FAMILIAR
'

MASKED BY
'
NOVEL
-
BASE LINE
'

CONTRAST
,

FM

TASK
.........................

243


6

A
PPENDIX
1
5



EXPERIMENT
3,

SUBSEQUENT MEMORY EF
FECTS OF THE
FM

TASK UNDER P
<0.01

THRESHOLD

................

244








































7









Bypassing the
hip
p
ocampus
:


Rapid
Neocortical Acquisition of Long Term Arbitrary Associations via Fast Mapping




Tali Sharon



Abstract


Declerative memory,
defined as the conscious recollection of facts and events

(Squire, 1992)

is believed to be mediated by two separate

memory systems

(McClelland et
al., 1995;
O’Reilly and Rudy, 2001; Norman and O’Reilly, 2003
)
.
The initial process of
the
encoding

of

novel information into memory is
believed to be critically dependent on

the medial temporal lobe (MTL) memory system

and o
n

the hippocampus in particular.
After
rapid
initial
a
cquisition,
a slow process of
consolidation takes place
, which typically
takes few years in humans,

by which the
information
is gradually assimilated into the
second memory system, located in various
regions of the neocortex
.
The reason this
process is slow is because rapid changes in this syst
em would cause a "catastrophic

interference" (French, 1999) with the pre
-
existing memory structures.
According to this
theory,
it is
only after sufficient consol
idation
has occurred
that

the information
can
cease
to depend

on the hippocampus.

However, recent evidence from patients with specific
lesions to the hippocampus has posed a difficulty for this conceptualization, demonstrating

8

an ability to acquire some no
vel information postmorbidly

which seems to have
declarative
-
like
characteristics in the absence of an intact hippocampus

(e.g. Kitchener et
al., 1998; McCarthy et al., 2005).


Up to date, the mechanism mediating this novel declarative learning has not b
een
discovered. Studies attempting to create new learning
in hippocampal patients
have
implemented methods of slow repetitive learning which
r
esulted in the acquisition of
memories with characteristics of non declarative memory (rigid, specific and not
acc
essible to conscious recollection
,
Bayley and Squire, 2002
; Stark et al., 2005
). These
findings

have led some researches (e.g. Bayley and Squire, 2002) to conclude that extra
hippocampal declarative learning may not be possible afterall and that those repo
rted cases
of apparent learning were a result of the mediation
of

residual functioning tissue
in

the
hippocampus.

In contrast to this,
we demonstrate in this thesis a neurocognitive mechanism which
enable
s

rapid acquisition of declarative like novel information in the form of arbitrary
associations independently of the hippocampus. This mechanism is termed 'fast mapping'
(FM
, Carey and Bartlett, 1978
)
and
is known to support vocabulary
acquisition
in childr
en

as fast as after only a single exposure to the word
-
object association
.
The FM
mechanism
allows for
a mapping
to be created
between a word and it's referent is
rapidly
created
by
the child
based on logical hypothesis formation and in particular disjunct
ive syllogism.
The rational for
the choice to

explore FM as
a possible hippocampal
independent learning
mechanism
in this thesis
wa
s as follows:

first, FM has been studied as a mechanism which
enables learning in young children whom do not yet have a fully

functional hippocampal
memory system

(Bauer, 2008)
, thus there is evidence for it's possible hippocampal
independence. Second, the fact that FM induces an
arbitrary
associative type of learning is

9

predictive of it's possible role in the encoding of declar
ative
memories, as relational
memory

is considered a distinguishing mark of declarative memory (
Cohen and
Eichenbaum, 1993
). Third,
learning through FM is incidental and as such may be more
inclined to be encoded via an alternative memory system. Finally,
learning through FM is
created within a context, a parameter which may facilitate the assimilation of the newly
acquired information into memory (Tse et al., 2007).



The present thesis involves
three studies
constructed to provide

converging

evidence
from
behavioral, imaging and lesion
methods for the role of

FM
in extra hippocampal
declarative
memory formation
. For the purpose of this research, a FM paradigm
involving
pictures of rare and exotic fruit, vegetables, animals and flower and thei
r corresponding
labels
has been constructed and fitted for use with adult participants. In the first section of
this thesis, our goal was to characterize

the
different
behavioral
aspects of
FM
learning and
memory in
a healthy adult
population
.
Our hypothes
es were that adults will succe
ss
fully
learn and retain over time novel associations through FM. We also hypothesized that
context will facilitate the encoding of these association into memory. Finally, we
explored
different methods of memory retrieval

cues
: free and cued recall and recognition memory,
expecting to find better memory for retrieval via recognition. Our findings from this
first
study

lend support to these hypotheses. It was found that adult participants were able to
successfully map the correc
t referents for the novel labels presented to them through the
FM paradigm. They were also able to retain these associations for the long term and
retrieve them from memory via recognition after a 20 minute and a week's delay. After 3
months however, the p
articipants
'

performance on the memory test for these associations
no longer differed from chance.
The participants recall memory in contrast was very low
across all times of testing.
Thus, it appears that the FM mechanism is available also in
adulthood. T
hese

findings replicate previous studies using adult participants (Markson and

10

Bloo
m, 1997; Halberda et al., 2006
;
Ramachandra et al., 2010
),

however the innovation in
the present study is in its large number of overall associations presented. According to

our
results, t
hese
novel
associations
created
are accessible for retrieval via a recognition but
not a recall type of memory test. These findings also replicate previous studies testing

for
FM production ("recall memory"; Chapman et al., 2006;
Kay
-
Raining

Bird et al., 2004
)

and suggest that a single presentation of a novel picture
-
label association may suffice for
the creation of
a
crude

memory

representation accessible for retrieval

when presented with
sufficient retrieval cues, but perhaps more exposures

may be required to create a richer
more varied representation of the association. This notion has also been implied by Carey,
(1978) referred to as "slow mapping".
The
results

for the effect of a study context on the
subsequent memory for the associations

acquired via FM

showed that a context indeed
enhanced retrieval performance. A surprising finding in this regard was the especially high
rate

and long term retention

of memory for the associatins between the novel pictures and
the sentences used as the co
ntext itself. These findings have important implications for an
ongoing debate in the FM literature regarding this mechanisms role in learning processes
beyong the l
exical domain as they show that learning through this mechanism is not
specific for novel w
ords.


In the second part of the thesis,
we implement
ed

an adapted version of the FM task,
and a control
standard memorization task requiring explicit encoding with a group of
patients with cerebral lesions. Four patients with lesions to the hippocampus
, o
r lesions to
structures directly impairing hippocampal function

were recruited in order to examine the
necessity of this structure to the formation of declarative memory. It was our hypothesis
that FM learning would not depend on the integrity of the hippo
campus and that these
patients would demonstrate an ability to acquire novel associations through FM, but would
nonetheless fail to encode the same type of associations through a matched control

11

paradigm involving explicit encoding instructions (EE).
In ad
dition, a
s the ATL has been
depicted

as a convergent zone

for semantic
memory
(Martin and Chao, 2001, Damasio et
al., 2004)
or as an a
-
modal semantic 'hub'

(Patterson et al., 2007;

Rogers et al. 2004,
2006)
,

we
also
recruited
for this study
two
patients who suffered damage to the anterior
part of their lateral temporal lobe (ATL).
One of these patients had additional damage to
the hippocampus as well. These subject
s

were tested
in order to examine the
possible
involvement of this region in FM lea
rning.

Our hypotheses for th
ese patients
were that
they
would show the opposite pattern of performance, showing a failure to acquire novel
associations via FM but would nonetheless succeed in the standard EE task, given an intact
hippocampal system. Findin
gs from this study have confirmed these hypotheses showing a
pattern of a classical dissociation between FM and EE functions in hippocampal patients.
Surprisingly, their FM memory performance did not fall from those of a group of healthy
matched control pa
rticipants, despite poor performance on the EE task as
well as on
neuropsychological tests of associative memory
. Furthermore, they were able to retain
these associations over a week.
The characteristics
of the FM paradigm as well as the
participants'

perf
ormance
indicate that

the
memory representation acquired by our
hippocampal patients is

of a decl
arative or declarative like type of memory.
The
performance of the ATL lesioned patients also confirmed our hypothesis and pointed to the
crucial role of the n
eocortical anterior temporal region to learning via FM.

In the
third

part of the thesis, we

used functional

magnetic resonance imaging
(fMRI) in order to further explore the neuroanatomical correlates of the FM learning
mechanism. In this study an even
t related subsequent memory paradigm was implemented
with task (FM and EE) as a between subject factor. In this study, we expected to find
activations in neocortical regions known to mediate semantic memory correlated with FM
learning. In particular, in li
ght of the results of
the second

study we were interested in the

12

temporal polar (TP) region of the ATL.
Although no significant neural correlates were
found under a p<
0.001 statistical threshold for the subsequent memory analysis of the FM
task, a more permissive p<0.01 threshold indicated activations in the hypothesized TP
region. An analysis of the contrast between the neural activations of exposure to novel
associati
ons and a low level perceptual base line
revealed an interesting distinction
between the neural activations correlate with FM and standard EE learning.
In both
studies, medial and lateral pre frontal activations along with superior and medial posterior
par
ietal regions were found, however distinct areas within these regions were found for the
FM and EE tasks. Interestingly, th
i
s

distinction resembled

that found between the brain
regions mediating recollection and familiarity processes
of recognition memory

(Yo
nelinas
et al., 2005).


Overall, we demonstrate in this thesis converging evidence from behavioral,
imaging and lesion studies pointing to a possible role of the FM mechanism in
acquiring
novel declarative

knowledge independent of the hippocampus. Als
o, in contrast to current
theories of memory consolidation we show preliminary evidence for neocortical plasticity
in the rapid encoding of

novel arbitrary associations. We discuss these findings and their
important

implications for the understanding of
ne
uro
cognitive processes

of

memory

and
their potential implications for the rehabilitation of memory impaired patients.














13

L
ist of tables

Table 1. Experiment 1 demographic characteristics.

................................
...........................

54

Table

2. Experiment 2a demographic characteristics.

................................
.........................

95

Table 3. Volumetric analysis.

................................
................................
............................

113

Table

4
. Experiment 2b results.

................................
................................
........................

130

Table

5
. Experiment 2b results: Amnesic patients' recognition memory, FM task.

.........

132

Table 6. Experiment 3 Peak activations
in the 'novel
-
base line' contrast.

.........................

161

Table 7. Experiment 3 Novel
-
familiar condition
................................
...............................

164

Table 8. Eperiment 3 'subsequent memory' contrast, EE task

................................
...........

165



















14

List of figure
s


Figure 1. Diagramic representation of the primary structures involved in amneisa.
...........

17

Figure 2. A memory taxonomy

................................
................................
...........................

23

Figure 3. Experiment 1 examples of novel pictures

................................
............................

55

Figure 4. Experiment 1 Examples of stimuli pairing

................................
..........................

56

Figure 5. Experiment 1 study paradigms

................................
................................
.............

58

Figure 6. Experiment 1 recognition tests.

................................
................................
............

61

Figure 8. Experiment 1 results: recall memory for context.

................................
................

70

Figure 9. Experiment 1 results: free recall memory for

context by novelty.

.......................

71

Figure 10. Experiment 1 results: cued recall memory for context by familiaritynovelty.

..

74

Figure 11. Experiment 1 results: recognition memory.

................................
.......................

77

Figure 12. Experiment 2 examples of stimuli

................................
................................
.....

96

Figure 13. Experiment 2 FM study
paradigm.

................................
................................
....

98

Figure 15. Experiment 2a recognition test.
................................
................................
........

100

Figure 16. Experiment 2a results: RT'S.

................................
................................
............

102

Figure 17. Experiment 2a
results: recognition memory.

................................
...................

104

Figure 18. MRI scans of AD.

................................
................................
............................

115

Figure 19. MRI scans of ShB.

................................
................................
...........................

116

Figure 20. MRI scans of EC.

................................
................................
.............................

116

Figure 21. MRI scans of NS.

................................
................................
.............................

117

Figure 22. MRI scans of AA.

................................
................................
............................

118

Figure 23. MRI scans of KS.

................................
................................
.............................

119

Figure 24. Experiment 2b study paradigm.

................................
................................
.......

121

Figure 25. Experoment 2b recognition memory test.

................................
........................

122

Figure 26. Experiment 2b results: RT's.

................................
................................
............

128

Figure 27. Experiment 2 results: MTL patients'recognition performance

........................

134

Figure 28. Experiment 2 results: ATL patients' recognition performance.

.......................

135

Figure 29. Experiment 3 timeline for a single event.

................................
........................

149

Figure 30. Experiment 3 paradigm time line.

................................
................................
....

151

Figure 31. Experiment 3 recognition memory test.

................................
...........................

152

Figure 32. Experiment 3 results: recognition memory performance.

................................

158

Figure 33. Experiment 3 the 'n
ovel
-
base line' contrast

................................
......................

160

Figure 34. Experiment 3 Novel
-
familiar contrast.

................................
.............................

163

Figure 35. Experiment 3 'subsequent memory' contrast, EE task

................................
.....

167

Figure 36. An EPI image.

................................
................................
................................
..

177















15

Introduction

T
wo years ago, Henry Gustav Molaison passed away.
Up until the age of 27,
H
enry

had

suffered from
frequent
epileptic seizures
, intractable by anticonvulsant therapy and
depr
i
ving him from leading a normal life.
In 1953, his
doctors attempted a radical
, then

experimental,

surgical procedure
in which

they removed the part of his brain where the
seizures originated.

The surgery was
partially
successful in
reducing the seizures, but

its

results were far from allowing Henry to lead a normal life.
As reported in 1
954 by William
Beecher Scoville and Brenda Milner, an unexpected effect of surgery was

a grave loss
of

memory for all events following the surgery.
Although being lucid, with no damage to
general
intelligence

or to other cognitive domains such as perceptual or lingual functions,
Henry

could not create new memories.

In other words, Henry was stuck in time, or

as

he
described his state "like waking from a dream… every day is alone
in itself
..
." (Milner et
al.,

1
968, p
p
. 217).
The part of Henry's brain that was removed was
part of
the medial
temporal lobe bilaterally, including the hippocampus
, amygdala

and adjacent

neocortical

structures; the loss of memory
Henry suffered has been termed
'anterograde amnesia'

and

Henry has been since known as H.M., probably the most studied patient in the history of
neuroscience

(see Squire, 200
9
)
.

Anterograde Amnesia

and memory subsystems

Anterograde amnesia (AA) is defined as the inability to create new memories, or
acquire new
information since onset of brain injury
.

It is associated with pathology in two
main regions
:

the medial

temporal l
obe
(MTL)
and the medial diencephalon
.
The

relevant

structures forming the MTL include the dentate gyrus, hippocampus, subiculum,
presubiculum, parasubicu
lum

and entorhinal, perirhinal

and parahippocampal cortices
.
T
he
hippocampus is considered the critical structu
re involved in memory pathology
. It is
sit
uated at the end of the medial temporal lobe system and is a recipient of convergent

16

projections from each of the structures that precedes it in the hierarchy.
Some of these
areas are sometimes regarded as the "hippocampal formation" or the "extended
hippo
campal system". The meaning of the term with regard to the exact structures it
includes varies but largely refers to the CA1
-
3 fields of the hippocampus itself
("hippocampus proper"), the dentate gyrus,
the subicu
lar complex and the entorhinal
complex. It
sometimes also
encompasses

adjacent structures including the perirhinal and
parahippocampal cortices.
In humans, damage limited to the hippocampus itself is
sufficient to cause moderately severe
AA
.

One example for this is Zola
-
Morgan et al.'s
(1986) patient R.B.who suffered a bilateral lesion confined to a part of the hippocampus

(CA 1 field) and
exhibited m
a
rked
a
nterogr
a
de
amnesia
,

with no sign of any other
cognitive impairment.
However, the extra
hippocampal structures in the MTL are also
known to have a role in memory processing and clinical reports have shown that in some
cases, the more extensive the damage to the MTL, the more severe the memory
impairment (
e.g. Bayley and Squire, 2005
).

The me
dial diencephalon
and the basal forebrain have also been implicated in
memory disorders.

The medial diencephalon
comprises

the
thalamus
,

including the various
thalamic nuclei, the

hypothalamus

and the
mammillary bodies
.

Because the pattern
s

of
diencephalic and MTL amnesia resemble each other,
they raise

the question of w
h
ether
these two regions form an anatomically linked functional system (Aggelton, 2008
).
Indeed,
a
natomically, the MTL and medial diencephalon are strongly connected primaril
y via the
fornix (Aggleton,
et al.,

2005; Saunders et al., 2005). Specifically, the mammillary bodies
and the anterior thalamic nuclei receive direct, dense inputs from the
extended
hippocampal
system

(
Poletti and Creswell, 1977;
Saunders,
et al.,

2005
).

Y
et another region known to have widespread projections to the MTL and to
potentially be able to modulate its function (
Insausti et al., 1987
)
is the basal forebrain
,


17

which is the primary source of cholinergic innervation of the cortex.
The basal forebrain

includes the medial septal nucleus and the diagonal band of Broca which project to the
hippocampal formation mainly through the fornix, and the nucleus basalis, which
projects widely to frontal, parietal, and temporal cortices (Mesulam et al 1
983).






















Figure
1
.
A schematic

representation of
MTL dienchephalic pathways.


Coronal section of a human MRI image showing a simplified schematic representation of two MTL
-
diencephalic pathways. (1) Afferents from the Entorhinal cortex (Ent.) reach the hippocampus (Hi; CA fields
and dentate gyrus) via t
he perforant pathway (PP). Efferents from the hippocampus reach the subiculum
(Sub) and back to the Entorhinal cortex. The Entorhinal cortex projects to the dorsomedial thalamus (MD)
primarily through the perirhinal cortex (not depicted) as well as through

direct connections. (2) Efferents
from the hippocampal/subicular complex reach the anterior thalamic nuclei (Ath) through the fornix,
primarily through the mamillary bodies (MB) but also directly, together forming the ‘extended hippocampal
-
diencephalic sy
stem’. Return connections from the anterior thalamus to the hippocampal/subicular complex
travel through the cingulum bundle (not depicted).

;


H.M.
and additional patients suffering from
anterograde amnesia

have taught us a
great deal about human memory: it has given us insights as to the cognitive and neural
organization of memory. The pattern of spared and compromised memory functions in
these patients suggests there are important distinctions between diff
erent mnemonic
E
E
E
n
n
n
t
t
t



S
S
S
u
u
u
b
b
b



H
H
H
i
i
i



A
A
A
t
t
t
h
h
h



M
M
M
D
D
D



V
V
V
L
L
L



F
F
F
o
o
o
r
r
r
n
n
n
i
i
i
x
x
x



&
&
&



M
M
M
B
B
B



P
P
P
P
P
P



F
F
F
o
o
o
r
r
r
n
n
n
i
i
i
x
x
x



Fornix & MB

Forni
x

AT
n


VL

MD


PP

Sub

Hi


Ent



18

systems.

For examp
le, although in AA memories since
the
onset of amnesia are impaired,
m
emories from the past remain largely intact
(
although, usually
some loss of memory from
the time period before the surgery
is also involved
termed
limite
d
'retrograde amnesia').

Another example is that
although
AA patients
are
not able to acquire long term novel
memories,

they

are capable of maintaining information in mind in the immediate period
(
'short term memory', Baddeley, 1990
).
This pattern of prese
rved and lost capabilities have
led scientists to suggest the existence of a distinct memory

system called

the
declarative

memory

system
(Squire, 1992)
.

Declarative Memory

The distinction between declarative and non declarative memory has been
suggested by
Larry Squire

(1992)
.
While declarative memory supports the conscious

recollection of facts and events, non declarative

memory relates to memory that
manifests
itself throu
gh performance and is not available to conscious recollection.
The terms
‘explicit memory’ and ‘implicit memory’ are sometimes used as well and have
approximately the same meanings as declarative and nondeclarative memory, respectively.
However, they are u
sually used to describe the manner in which memory is assessed rather
than the underlying memory function. Thus, a memory test requiring no consciouss
recollection is termed implicit while memory tests specifically requiring conscious or
aware recollection

are termed explicit. Yet another variation has been termed the
'relational memory' theory (Eichenbaum et al., 1994). This theory holds that a division
should be made between memory for relations among items and memory for individual
representations. As me
mory for relations has been shown to be reliant on declarative
memory, this conceptualization is congruent with the declarative memory
conceptualization. In the following sections, the declarative
-
nondeclarative disctintion will
primarily be utilized.


19

Non

declarative memory includes abilities such as skills and habits

(also termed
procedural memory
)
, simple forms of conditioning

and the phenomenon of priming

(see
Figure
2
)
. It has been consistently found that AA involves the loss of the capacity to form
new declarative memory, but leaves non
-
declarative memory processes intact

(Tulving and
Schacter, 1990; Schacter and Buckner, 1998; Schacter et al., 200
4)
.

An early interesting
observation of such non declarative memory
in an amnesic patient
was made by Claparède
in 1911:

"... I tried the following experiment.., to see if she would better retain an
intense impression that set affectivity in
to play. I pricked her hand forcibly with a pin
hidden between my fingers. This little pain was as quickly forgotten as indifferent
perceptions and, shortly after the pricking, she remembered no more of it. However,
when I moved my hand
near hers again, she pulled her hand back in a reflex way
and without knowing why. If, in fact, I demanded the reason for the withdrawal of
her hand, she answered in a flurried way, 'Isn't it allowed to withdraw one's hand?'..
.
If I insisted, she would say to me, 'Perhaps there is a pin hidden in your hand'. To
my question, 'What can make you suspect that I would like to prick you', she
would take up her refrain, 'It is an idea which came int
o my head', or sometimes
she would try to justify herself with 'Sometimes pins are hidden in hands'. But she
never
recognized

this idea of pricking as a memory."

(Quoted

by M
accurdy, 1928
)

This
fascinating quote is actually a
description

of an intact conditioning learning
effect in amnesia which has since been examined and confirmed with other conditioning
experiments such as eye blinks (Weiskrantz and Warrington, 1979). Additional studies
have found that

amnesics

show

for example
a prese
rved ability to

show improvement over
multiple trials of

trac
ing

a figure reflected in
a
mirror ('
mirror tracing’
,

Milner, 1962;
Damasio et al., 1985; Tranel et al., 1994
, Cohen and Squire, 1980)
,

to

press a sequence of

20

keys

‘serial reaction time’ (Nissen an
d Bullemer, 1987
)
and

to perform
normally o
n
various motor learning tasks (rotary pursuit, bimanual tracking and tapping, Corkin, 1968)
.

As is demonstrated in Claparède's report, in all of these cases the preserved non declarative
memory
performance

is acc
ompanied by a lack of
conscious

awareness to the learning
episode
and an inability of the amnesics to explicitly report what they had learned. Another
interesting example of this dissociation
comes

from research on priming
. Priming has been
defined as "a n
on

conscious form of memory that involves a change in a person’s ability to
identify, produce or classify an item as a result of a previous encounter with that item or a
related item" (Schacter et al., 2004). It is commonly assessed using indirect or impli
cit
tests such as performance on word stem completion, or
measurement

of reaction times.
Performance on these measures is improved
in AA patients after being

previously
presented with the target item
s
, even without a conscious awareness
of the study episode
and despite severe impairment in explicit memory tests for the exact same items
(Hamann
and

Squire
, 1997
; Shimamura, 1986)
.

Declarative memory systems and consolidation

At the synaptic level, much is known about consolidation into lon
g term memory
.
Hebb
(1949)
and Konorsky
(1948)
proposed
that

the
synapse linking two cells is
strengthened if the ce
lls are active at the same time. In 1973 t
he
first such synapses were
identified

by Bliss and Lomo
in the hippocampus

and

t
he effect of
Long

term potentiation
(LTP)
was

described
.

The LTP model has since been regarded as the
primary experimental
model for investigating the synaptic basis for learning and memory.

As opposed to synaptic consolidation,
consolidation at the
system's
level
is

less
understood.

Current theories of declarative memory suggest the existence of two
complementary memory systems (Squire et al., 1
984; Teyler and DiScenna, 1986;
McClelland et al., 1995
; O’Reilly and Rudy, 2001; Norman and O’Reilly, 2003
): a

2
1

hippocampal
-
based system which specializes in rapid acquisition of specific events, and a
neocortical system which slowly learns through environmental statistical regularities and
requires the initial support of the hippocampal system.
The purpose of the s
low learning of
the neocortex is to be able to eventually extract and model the similarity structure in its
environment. The rapid learning of the hippocampal system, in contrast, sacrifices the
ability to generalize for the ability to rapidly learn arbitr
ary patterns of activity.
One
important characteristic of neocortical learning that t
his notion of two memory systems

addresses is

the phenomen
on

of 'catastrophic interferences'
.

C
onnectionist models of
memory
have demonstrated that
when new learning is
r
apid
,
models of “
neocortical


neural networks show near
-
complete forgetting of pre
-
existing
knowledge (French, 1999)

before the novel associations can be learned. Humans and animals on the other hand, show
a certain
extent

of interference but are able to m
aintain old knowledge in the face of new
learning
.
Thus, memories are believed to be encoded in a process mediated by the
hippocampus
, buffering the novel information and allowing interleaved activation of old
and new information, before they can be perman
ently stored

in the neocortex
. This
assimilation of novel information in the neocortex is
referred

t
o as

consolidation at

the
system level, or as
Müller
and

Pilzecker
(
1900)

defined consolidation

"
a time dependent
assimilation and storage process of an ex
perience to the extent that is sufficient for
retrieval
"
.

Probably the
most cited

evidence for

system's

consolidation is the
"Ribot
gradient" (Ribot, 1881) found in retrograde amnesia, a phenomen
on

in which memories for
recent periods are more susceptible to damage than remote memories, which presumably
have already been well consolidated in structures outside the hippocampus.

Drawing on
this, Bontempi et al. (1999) mapped the regional metabolic act
ivity in the brains of mice
correlated with retention of a spatial discrimination task tested at different
times
. They
demonstrated
that increasing the retention intervals resulted in decreased hippocampal

22

metabolic activity and a loss of correlation betwe
en hippocampal metabolic activity and
memory performance; while in the neocortex a recruitment of certain cortical areas was
observed following the longer delay
. Thus, Bontempi et al. (1999) demonstrated a

transitory interaction between the hippocampal f
or
mation and the neocortex in

mediating

the establishment of long
-
lived cortical memory representations.


Declarative memor
y
can

further

be

divi
d
ed into e
pisodic and semantic
memory

(
Tulving, 197
2
; s
ee
Figure
2
)
, and there is a debate in the literature whether these
constitute independent systems (Schacter and Tulving, 1994) or only sub
-
systems within
declarative memory (Squire, 2004)
. Episodic memo
ry refers to the ability to create new
memories for
specific events and episodes experienced

throughout an
individual's

life, or

information based on particular and concrete experiences located in personal time and
space
.

This form of memory is
similar to
William James's conceptualization

of memory

as
"the

kno
wledge of an event as fact...
with
the additional consciousness that we

have
thought or experienced it

before
"

(James,
1892
).

In contrast,

s
emantic memory
refers to the
memory of language, of facts abou
t the world, or about oneself ("personal semantic
memory", Kopelman et al., 1989), which is free of contextual information, independent of
personal time and space.


In standard
theories

of
system's
consolidation (Squire, 1992; Squire et al., 1984;
Dudai, 2
004; McGaugh, 2000) there is no distinction between episodic and semantic
memories and
these

are conceptualized to be dependent on the MTL until they have been
consolidated, after which they become independent of it. In these theories, consolidation is
mod
eled as a neocortical learning of patterns extracted from the hippocampus over time
.

Other theories offer distinctions between these two memory subsystems. The "Multiple
Trace Theory" for example (Nadel
and

Moscovitch, 1997;
Moscovitch
and

Nadel, 1999;
Nadel et al., 2000
) suggests that while semantic memory is extracted from episodic

23

memory in a 'semantization' process and can eventually exist independently of the MTL,
episodic memory remains dependent on it and the creation of multiple tra
ces of each
memory underlie the consolidation process
.

Yet another theory of memory transformation
(Cermak, 1984) further stresses the 'semantization' component as it posits that all
memories are actually encoded as episodic memories and eventually become
semanticized,
or 'decontextualized', through the process of extraction of gist or abstraction.
The specific
process of consolidation

is not yet fully known
.
Each
of t
he theories mentioned above
provide
s partial

explanation for
the

evidence from the neurobi
ological
and
neuropsychological literature, and they all still lack important estimates for consolidation,
as
the exact time course it follows and the specific parameters it
depends on (see Meeter
and

Murre, 2004). However,
consolidation theories basically

agree upon the principle by
which semantic memory is initially dependent on the MTL for acquisition and after a
process of some sort becomes independent of it.




Figure
2
. A memory taxonomy

The declarative


non
-
declarative memory distinction and it's subsystems.

Novel semantic learning in amnesia

I
nitially, evidence
in agreement with this notion
was fou
nd showing that
hippocampal
amnesics
are impaired in acquiring new episodic as well as semantic
memory
.

Verfaellie et al.
(
1995
)

for example studied knowledge of words that had entered
Memory

Declarative

Non
-

declarative

䕰楳潤Ec

Se浡湴楣m

S歩k汳…
䡡扩bs

P物浩湧

C潮摩o楯湩湧


24

the lexicon post
-
morbidly in a long

standing amnesic and reported no
evidence of
vocabulary acquisition. Gabrieli

et al.
(
1988
) examined patient HM and reported very
scant post morbid semantic acquisition
.
Evidence from n
europsychological
standardized
tests also revealed striking deficits in new learning. For example, it wa
s observed that
dense amnesic subjects tended to score zero correct across three trials of the six arbitrarily
paired associates of the Wechsler Memory Scale (Meyer
and

Yates, 1955; Cutting, 1978).
But d
espite these initial finding, more and more
studies o
f AA
now
present evidence that
there are exceptions to this rule. These studies have demonstrated

that although

hippocampal

amnesics
are
impaired in their capacity to form novel episodic memory since
onset of amnesia, they are able to acquire
at least some

novel semantic information post
morbidly (after becoming amnesic)
,

s
upporting old ideas

that amnesia represents a
selective loss of episodic memory ability (Kinsbourne and Wood, 1975; Schacter and
Tulving, 1982; Cermak, 1984).

Kitchener, Hodges, and McCar
thy (1998) report on patient RS, densely amnesic as
a result of a subarachnoid hemorrhage, who was found to have acquired information about
famous people, public events and new vocabulary during the 13
-

year period since he
became amnesic, despite having n
o measurable anterograde episodic memory function.
R
S

was able to explicitly report this knowledge giving information about
the famous people he
was presented with

and definition of
the novel
words
.

Verfaellie, Koseff and Alexander
(2000) describe patient
PS, who suffered a selective lesion of the hippocampus proper and
has gained a sense of familiarity of novel vocabulary and famous people
. McCarthy,
Kopelman
and

Warrington (2005) reported on patient RFR, who suffered from herpes
simplex encephalitis causi
ng a severe global amnesic syndrome. They report a

16 year
follow up on this patient in which they conclude, amongst other
s
, that despite his severe
anterograde amnesia, RFR showed "a limited but significant ability to acquire new word

25

meanings". In additi
on, he acquired sufficient information about famous faces and names
to support familiarity judgments.

Evidence for semantic acquisition in amnesia can also be
found in the developmental neuropsychological literature. Vargha
-
Khadem et al. (1997)
studied thr
ee patients with brain injuries occurring at a young age, involvin
g bilateral
hippocampal damage and
found that despite being profoundly amnesic for episodic
memory, their factual knowledge, or semantic memory was within the low average to
average range. I
t is important to mention
however
that
,

as these are
developmental

amnesic
patients
,

the possibilities that reorganization processes have taken place, or compensation
through learned strategies, should be taken into account (Squire et al. 2004).


What are
these studies telling us? The main conjecture raised
(e.g. Bayley
and

Squire, 2002; Tulving
,

1991;
Kitchener et al., 1998
) was

that
this
novel semantic learning
wa
s

being

mediated by some sort of non

declarative memory system directly in the
neocortex.
It
was hypothesized, in accordance with the computational models of
complementary memory systems (Mclelland et al., 1995), that
this
learning
wa
s a result of
small changes occurring very slowly directly in neocortical connections through repeat
ed
exposure to
the information and as such
these memories
sh
ould have the characteristics of
being rigid and

hyperspecific
. Indeed, studies attempting to create
such memories in
laboratory
paradigms
have tended
to
support this notion demonstrating

that in order for
new learning to occur, numerous repetitions of the information is r
equired and learning
results in limited, impoverished and shallow
memory not accompanied by conscious
awareness and with no ability to be generalized
(t
hese
studies are r
eviewed in the
following

section
).


One of the

problem
s with this approach is that it appears that
in many cases,
amnesic patients can consciously access these newly acquired memories, and explicitly
report them
.

Also, some cases have found the learning to

be flexible after all.
Thus,

26

declarative

like memories

do appear

to be
acquired somehow

after all
.
Westmacott
and

Moscovitch (2001)

for example

examined incidental learning of famous names and
vocabulary terms from different time periods since onset of th
e amnesia in patient K.C., a
person with extensive bilateral hippocampal damage. Results show that although K.C.'s
ability to acquire explicit semantic information was impaired relative to control
participants and appeared to be limited to simple represent
ations, K.C. had demonstrated
implicit as well as explicit semantic knowledge of information that had entered popular
culture after the onset of his amnesia.

This
raises another question: if the learning is indeed declarative, is it really taking
place
directly in the neocortex? An alternative explanation might be that the

learning is
being mediated by residual tissue in the MTL
.

Bayley
and

Squire (2002) argue that
semantic learning not mediated by the MTL is necessarily non
-
declarative by nature. They
argue that the declarative semantic information acquired by patients reported in the
literature, including Westmacott
and

Moscovitch's patien
t K.C., was probably mediated by
the MTL after all
-

by structures within the medial temporal lobe that remained intact in
these patients
.
They suggest that the characteristics of the knowledge acquired in these
patients, rather than being unusual and qual
itatively distinct, are what one might observe in
any individual who has only partly learned a body of material,
following partial
functioning of the MTL
. Bayley
and

Squire (2002) report a case study of patient E.P., a
patient with extensive damage to the
medial temporal lobe (more extensive than that of
K.C.) who demonstrated considerable learning of new information that, in accordance with
their theory, was not accompanied by conscious knowledge. Bayley
and

Squire propose
that
intact structures within the

medial temporal lobe are responsible for the learning
in
those cases when factual information
seems to be
acquired as
conscious
declarative
knowledge, by memory impaired patients, then. In contrast,
cases of

non
-
declarative

27

acquisition of
knowledge, as in

their study,
are mediated
directly
by structures
within the
neocortex.

Nonetheless, a study
conducted recently by O'Kane et al. (2004)
on patient H.M.
,

who has as mentioned absolutely no functioning hippocampus, revealed
that despite
prior
failures t
o
demonstrate new

learning
(Gabrielli et al., 1988),

H.M. is in fact capable of
post
-
morbid explicit learning.
I
n this study,
H.M.
demonstrated an ability to recall the
corresponding famous last name for 12 out of 35 postoperatively famous personalities,
w
hen their first names were provided as cues. What was fascinating is that t
his number
nearly doubled

when semantic cues were added and he was able to provide identifying
semantic facts for one third of these recognized names. This suggests

that
H.M. has
ac
quired a
semantic network in which these names were incorporated and that his
performance
was not
just a consequence of some sort of perceptual non

declarative
memory
. According to O'Kane et al. (2004), these results provide "robust, unambiguous
evidence t
hat some new semantic learning can be supported by structures beyond the
hippocampus proper" (p.417).

However, this notion still remains in dispute by researchers, and O'Kane et al.'s
study (2004) has already been challenged by Bayley
and

Squire (2005), wh
o report on two
deeply amnesic patients, E.P.
and

G.P., whose lesions involve virtually the entire MTL
bilaterally, that have not been able to learn nearly any new semantic knowledge at all since
onset of their amnesia. Other reported cases of this kind ar
e Verfaellie et al's. (2000)
patient S.S., having bilateral damage to the medial temporal lobes, whose performance did
not exceed chance on even tasks of familiarity of novel semantic information, and
Cipolotti et al.'s (2001) patient V.C.

In summary, it

appears that there is still much dispute regarding this issue and that
t
he

evidence is mixed and unclear (which is expected when dealing with case studies with

28

different etiologies resulting in unique lesions). It is still not clear whether explicit
-

declarative semantic learning can be supported by structures outside the MTL, however it
is quite clear that post morbid semantic learning can take place despite of

an extreme
impairment in episodic memory, which raises a question as to the neurocognitive
mechanism that is supporting this learning.

Neurocognitive mechanisms of new learning in amnesia

What is the mechanism supporting semantic learning in amnesia? So
far, the
answer to this question remains unknown. Different neurocognitive mechanisms, primarily
focusing on intensive repetitive exposure, have been proposed to explain this phenomenon
and researchers have tried by various methods to actively teach amnesi
c patients new
semantic information.



The method of v
anishing cues

G
lisky et al.

(
1986
) taught four memory impaired patients novel computer
-
related
vocabulary by

a method they termed "the method of vanishing cues". This method
involved giving subjects a c
ue, in the form of a word stem of the to
-
be
-
learned words and
systematicly reducing letter fragments across trials. Glisky et al. (1986) show that all
patients acquired a substantial amount of the vocabulary, however the
learning was slow
and strongly

depe
ndent on first
-
letter cues.

The patients

retained the vocabulary
they had
learned when tested after a

6
-
week interval and showed some transfer of the knowledge
they had acquired.

However, evaluations of this method

in comparison to a control
standard antic
ipation (or rote learning) method
and to other memory rehabilitation methods
revealed that this method had only a small and nonsignificant effect and did not have
advantage over standard learning

(
Hunkin and Parkin, 1995
; Kessels et al., 2003
).


29

Errorless l
earning

On the basis of studies
involving

the vanishing cues method and on a study
Hayman
and

Macdonald, (1990) performed on amnesic patient K
.
C
.
,
Tulving at el. (1991)
claimed that in order for learning to take place in patients with
AA
, an errorless trai
ning
method is crucial

-

a procedure in which no opportunity is given during training to make
errors
.
In their study with K
.
C
.
, Hayman
and

Macdonald (1990) taught
him

to interpret
ambiguous phrases in terms of disambiguating words that accompanied the phrases
(e.g., a servant in name only
--
BRIDESMAID). The presence of interfering responses to
the phrases was manipulated and the results showed that K.C. lear
ned and retained best
those phrase
-
word pairs that had no incorrect or competing word
responses and were not
t
ested during learning, and

his worst performance
was with phrase
-
word pairs that had
original

competing associations and
were tested in
the course

of

learning.

Tulving at el.
(1991) claimed that learning procedures that allow for incorrect

responding

create
interference
to the learning because the incorrect responses compete with the correct
responses.

The method of the vanishing cues might have bee
n effective, according to
Tulving et al. (1991),
due to its constraining

of response

with words other than the desired
target and thereby
creating reduction in

interference through response competition.
Furthermore, the authors propose that the lack of lea
rning exhibited by H
.
M
.

in Gabrieli et
al.'s study (1988) might be explained by the same mechanism as in their training method
on every learning trial,
H
.
M
.

had to select the correct response to a definition or sentence
frame from
a set of

words, and the
s
e
lection procedure for any given

target
word continued
until that target word was
selected
.

This procedure allows

for
a
great deal of incorrect
responding which may have interfered with the

learning.
Tulving et al., 1991 succeeded to
teach patient K
.
C
.

sem
antic knowledge in this method, which remained after a 12 months
period.



30

Based on this notion,
Bayley
and

Squire, (2002) employed an "errorless learning"
procedure
(see also Wilson et al., 1994)
with their severely amnesic patient E.P. The
procedure they employed included 24 study sessions across 12 weeks in which E.P. was
presented with novel three word sentences, and was explicitly asked to think about the
meaning of these sentences. Results of
this study show that E.P. indeed demonstrated
considerable learning of new information, which persisted partially for 3 months, however,
his performance was well below the level achieved by controls, his ability to complete the
three word sentences was not

accompanied by conscious knowledge, and he was not able
to generalize the knowledge he had acquired beyond the exact stimuli he had learned
(when the second word in the sentence frame was replaced by a synonym, while the
performance of controls did not di
ffer significantly, E.P.'s performance fell from 20.7%
correct to 3.4%). In this study, no less than 32 repetitions were needed for E.P. to exhibit
learning.

Varied errorless learning

The results of Bayley
and

Squire's experiment (2002) raise a theoretica
l problem: if
the learning achieved by amnesic patients is as they propose, a neocortical one
independent of the hippoca
mpal system, why is it so
hyper specific
? After all,
computational models simulating a neocortical network of learning show that this ne
twork,

as opposed to the MTL
network is constructed for generalization

(McClelland et al., 1995)
.
Stark et al. (2005)

attempted to settle this theoretical difficulty.

They proposed that the
methods used in the errorless learning procedure itself are at the

heart of the hyper
specificity in the observed behavior. They believed that by introducing variance into the
training set, the neocortically mediated learning achieved could be properly generalized.
Indeed, Stark et al. (2005) were successful in training
patient T.E. (a patient with severe
damage to the hippocampal region) to learn three word sentences at a more conceptual

31

level by employing a procedure of varied errorless learning (they varied the sentences by
using synonyms).
A more recent study (Stark e
t al., 2008) further showed that gradually
introducing the variant items into training may be the optimal strategy for training an
individual with severe amnesia to learn and generalize new semantic information.

However, despite overcoming the initial theo
retical difficulty, Stark et al.'s study (2005),
similar to Bayley
and

Squire's study (2002),
failed to produce learning that resembles
semantic memory acquired in naturalistic settings. First of all,

T.E.
and

E.P.'s learning was
not available to conscious

awareness
. Second, as in

Bayley
and

Squire's study (2002)
, this
laboratory induced
leaning required intense exposures to the semantic information.
In their
everyday lives, it is questionable whether amnesic patients are exposed to so many
repetitions of t
he semantic material they encounter.

Actually, McClelland et al. (1995) offer an explanation themselves for the hyper
specifity of new information acquired in amnesic patients through their computational
models, although they refer in their study to skill

learning. They predict that the spared
acquisition arises from accumulating small changes in connections among relevant neural
populations in the neocortical system, and that the information is often restricted to the
specific context in which it was acqu
ired simply because the learning takes place directly
within the connections among the neural populations that were activated during the
acquisition process itself alone.

Anchoring new information to existing mental representations

Another factor suggeste
d by Tulving et al. (1991) to facilitate semantic learning in
AA patients was meaningfulness. They suggested that memory can be enhanced when the
to
-
be
-
learned material is meaningful for the subject in as much as it represents something
the subject already

knows, or to the extent that the relation of the material is consistent
with existing concepts.


32

Skotko et al. (2004) suggested a
mechanism for semantic learning based on a
similar notion.

They proposed that new semantic knowledge can be acquired by amnesic
patients, at least temporarily, so long as it can be anchored to mental representations
established preoperatively. The researchers tested this hypothesis on patient H.M. in the
context
of crossword puzzle solving, an old time hobby of H.M. Skotko et al. (2004) tested
H.M. on 3 new crossword puzzles: one puzzle testing pre
-
morbid knowledge, another
testing post
-
morbid knowledge, and another combining the 2 by giving postoperative
semantic

clues for preoperative answers. Results show that in marked contrast to his
inability to show any learning on the post
-
morbid crossword puzzle, H.M. was able to
show significant learning when information could be linked to mental representations
establish
ed preoperatively (the combined crossword puzzle). It is noteworthy that Skotko
et al. (2004) had also employed an "errorless learning" procedure in this study, in as much
as H.M. was forced to correct all mistakes at the end of each puzzle. Skotko et al.'
s (2004)
hypothesis of anchoring new semantic information to pre