Implicit memory in Alzheimer

2
Implicit memory in Alzheimer' s disease
thesis submitted in accordancewith the requirements of the University of
Liverpool for the degree of Master of Philosophy by Ioannidou Despina
May/2003

3
CONTENT TABLE
Abstract ............................................................................................................. 1
INTRODUCTION ................................................................................................ 2
Division of memory .......................................................................................... 2
Methods of memory testing .............................................................................. 3
Alzheimer' s disease .......................................................................................... 4
Implicit memory in Alzheimer' s disease .......................................................... 6
LITERATURE REVIEW ...................................................................................... 8
META-ANALYSIS .............................................................................................. 25
Categories of meta-analysis combination ......................................................... 26
i.sample size ................................................................................................. 26
ii. difference between the mean ages of the AD and the control group.......... 26
iii. dementia severity .................................................................................... 26
iv.number of exposures of stimuli ................................................................ 30
v.orienting task ............................................................................................. 30
Methodology ...................................................................................................... 32
RESULTS .............................................................................................................. 33
Calculation of the average group-comparison size effects ................................ 33
Calculation of the average AD priming size effects .......................................... 33
i.sample size.................................................................................................... 34
ii. difference between the mean ages of the AD and the control group ....... 34
iii. dementia severity ..................................................................................... 35
iv.number of exposures of stimuli ................................................................. 35
v.orienting task .............................................................................................. 35
DISCUSSION ........................................................................................................ 36
CONCLUSIONS .................................................................................................... 37
GENERAL DISCUSSION ...................................................................................... 39
Explicit memory of controls .............................................................................. 39
Associations with other tests ............................................................................. 40

4
AD and the transfer-appropriate processing ...................................................... 42
Semantic memory in AD .................................................................................. 43
Nature of stem-completion ............................................................................... 46
EXPERIMENT .................................................................................................... 51
Memory in healthy aging .................................................................................. 54
Methodology ..................................................................................................... 59
---Participants ................................................................................................... 59
---Design ........................................................................................................... 60
---Materials ....................................................................................................... 61
Procedure .......................................................................................................... 61
RESULTS ............................................................................................................ 63
Stem-completion ............................................................................................... 63
---young-old controls ........................................................................................ 63
--- matched old controls-AD patients ............................................................... 65
Cued-recall ...................................................................................................... 67
Analysis of responses by word-frequency .................................................... 69
DISCUSSION ....................................................................................................... 74
Young-old controls ........................................................................................... 74
Matched old controls-AD patients .................................................................... 77
CONCLUSIONS ................................................................................................... 80

5
Implicit memory in Alzheimer's disease
by Despina Ioannidou
Abstract
One of Alzheimer's disease main characteristics is gradual loss of memory. Although
AD patients have a well established memory deficit when performing in explicit
memory tasks, they have been repeatedly found to exhibit intact performance in
several implicit memory tasks, in comparison to healthy control subjects. Since many
studies that investigated AD performance in the implicit memory test of stem-
completion have produced contradictory findings, a review of the relevant literature
takes place, resulting in a meta-analysis of published data. Our main goal was to
examine the influence of several factors on the performance of AD patients, as
suggested by other researchers. In the experiment, time pressure during test-phase is
examined as a variable that could manipulate the perceptual or conceptual nature of
the stem-completion test and affect the performance of the subjects. An additional
variable is also examined; that of the concreteness or abstractness of the stimuli used,
in order to investigate the existence of a material-specific phenomenon; a post-hoc
analysis of word-fluency also takes place. The three groups of participants are: a group
of healthy young subjects, a group of healthy old subjects and a group of AD patients.
The existence of an age effect is also investigated as resulting from the direct
comparison between young-old subjects in stem-completion ( implicit memory ) and
cued-recall ( explicit memory ).

6
INTRODUCTION
In his book “Essentials of Human Memory” ( 1999 ), Baddeley refers to human
memory and its biochemical basis in a way analogous to that of a building and the
structure of the material used for its creation. In a same way that an architect faces
several constraints when using certain materials, human memory is affected
accordingly by its biochemical structure. Baddeley, with consideration about the
complexity of the molecular basis of memory and the ambiguity of any findings
coming from this area, presented the essence of recent work concerning
neurophysiology of memory. This brief description will enable us to comprehend
several aspects of human memory that have been examined thus far.
Division of memory
According to Baddeley ( 1999 ), it has been more than a century since it was first
suggested that there should be a division of memory into separate parts. Since then,
and especially in the last thirty years, relevant findings enhanced such a suggestion.
The first division between long-term and short-term memory was widely accepted by
the early 1970s, even though the fractionation of the human memory was considered
unnecessary by many scientists a decade earlier. However, intensive research on the
subject indicated that there was a need for further fragmentation of human memory.
Richard Atkinson and Richard Shiffrin ( 1968 ), led by the theoretical current of that
time, suggested the existence of a model that included three kinds of memory: sensory,
short-term and long-term memory.
Sensory memory is responsible for the entrance of stimuli coming from the
environment and their brief registration. Such information can be obtained by either
visual, auditory or haptic ways and is stored for a short period of time until it fades
away, or remains for further manipulation.
This further manipulation is achieved by short-term memory, a kind of working
memory with great importance that can elaborate on stored material through coding,
comprehension and reasoning. Due to the processes of short-term memory there could

7
be a successful link between storage of new material and recollection of old
information.
While information is kept in short-term memory for a very limited period of time, its
storage becomes permanent, or at least of a long duration, when it is transferred to
long-term memory from which it is subsequently recollected.
Long-term memory can be divided in at least two different ways. Endel Tulving (1972)
introduced the first distinction between episodic and semantic long-term memory. He
suggested that episodic memory includes the remembrance of particular incidents in
one’s life. Being able to recollect our experiences of our actions during the days and
the years is part of our episodic memory. On the other hand, our ability to store
information about the world and to preserve generalized knowledge of the world
beyond our personal experience, is based on our semantic memory.
Another, more recent distinction was that between explicit and implicit memory. Graf
and Schachter ( 1985; 1986; 1987 ) have used the term explicit memory to refer to the
memory illustrated by the subjects when they are asked to deliberately retrieve
information about themselves and their actions. On the other hand, they have used the
term implicit memory ( Graf and Schachter, 1985; 1986; 1987 ) to refer to memory
expressed by subjects that subconsciously show facilitation of performance when they
are asked to perform a task that involves information studied in a previous episode
where it appeared not to have any connection to the present task.
The difference between these two types of long-term memory, can be more easily
delineated if we refer to the particular tests used for their examination.
Methods of memory testing
Explicit memory requires from subjects to voluntarily remember information from a
previous studied episode. For example, they can be asked to recall words they had been
asked to read earlier, or to recollect as many details of a story they heard from the
researcher, or they could be asked to draw a picture they were shown earlier. In all
cases, subjects are directed to perform memory tasks examining directly their ability
to retrieve information.
On the contrary, if a researcher wishes to illustrate one’s implicit memory they would
pursue more indirect methods of testing, such as the ones below:

8
Stem-completion: the subject, at a study-phase, is presented with lists of words. At a
later phase he/she is asked to fill in strings of three letters in order to complete a whole
word (e.g. MOT-). Half of those trigrams are the first three letters of the words the
controls have studied earlier. Although subjects are unaware of the connection between
the two different tasks, they tend to show a preference for producing words they were
presented with earlier; subjects also appear to be less reluctant when they produce
studied words rather than unstudied, as they make use of less time in order to provide
studied items rather than unstudied. This facilitation of performance, known as
priming, is considered to be indicative of i m p l i c i t memory, a kind of memory
that does not require intentional recollection of information and is different from e x
p l i c i t memory, which asks from the subjects to reproduce previous experience.
Explicit memory is examined through a task known as graphemic cued recall, were
subjects are asked to fill in the trigrams with the words they have studied earlier.
Similar to stem-completion is the word-fragment test, where the subject is asked
during the test-phase to complete fragments of words (e.g. –ss-ss—for assassin) and
priming is also demonstrated by completing the fragments with studied words more
often than with others.
Another test designed to illustrate implicit memory, is the perceptual identification
task. Subjects are presented with word-material and at test-phase they are asked to
identify words briefly shown to them (e.g. in a few ms.). Priming is revealed by their
tendency to identify words they have seen earlier.
Implicit memory can be investigated through lexical decision. Priming is shown by the
reduced latency of subjects to identify studied words from strings of letters presented
to them, rather than when the strings include unstudied words.
At this point we would like to underline the fact that people can reveal aspects of
implicit memory not only with lexical material but also in cases where the stimuli are
pictures, objects, sounds or when their motor skill ability is tested.
Alzheimer’s disease
Alzheimer's disease, the most common type of dementia in the elderly today, was
originally described in 1907 by a famous neuropathologist called Alois Alzheimer.
He was the first to describe the clinical history of a 51 year-old woman, patient of his,

9
and the pathological findings of her brain. There were findings of fibrils in cells that
were tangled together ( tangles ) and numerous miliary foci due to the deposition of a
special substance ( amyloid plaques ). These are the classical pathological findings in
Alzheimer's disease, the degenerative, generalized cognitive impairment, known by
the name of the scientist ever since.
Today Alzheimer’s disease is considered to be the most common of dementing
diseases and although it has gathered the interest of many scientists around the world
for the last decades, it remains the most difficult, among the illnesses of this kind, to
diagnose and confront with any medication.
The most important risk factors are increasing age and a family history of dementia.
At the age of 60, the risk of developing AD is estimated to be 1%, doubling every five
years to reach 30-50% by the age of 85.
Intensive research aiming to investigate the reasons that cause such an extensive
impairment is being conducted all around the world. The neurophysiological nature of
the disorder is described by the existence of neuritic plaques, synaptic and neuronal
loss, and neurofibrillary tangles that occur in areas of the hippocampus and cerebral
cortex. Aggregation of beta amyloid, oxidative stress, blood-brain barrier leakage,
inflammation, and the inability of neurons to send nutrients and hormones to axons
lead to neuronal death. The beta amyloid forms fibrils that are rather large,
extraneuronal spherical bodies, termed neuritic plaques. Neurofibrillary tangles, on the
other hand, are intracellular proteins that occur within limbic and cortical neurons. The
neuropathology occurs in a specific pattern starting in the medial temporal lobe and
spreading through the cortical association areas and back to the nucleus basalis in the
forebrain, the primary location of cholinergic cell bodies and axons projecting to the
hippocampus, parietal, frontal, and occipital cortexes. The temporal and middle
temporal areas, and to a lesser extent, the parietal-associated and frontal cortexes, are
the location of neurofibrillary tangles and neuritic plaques.
It is not yet well understood why these plaques and tangles conquer healthy brain
tissue in parts of the brain connected to several aspects of intellectual action, and
hence, lead those brain areas to devastation.
At its onset the illness makes its appearance through symptoms of forgetfulness. The
patient begins to lose his orientation ability and doesn’t seem to remember recent
events. In addition, personality changes such as impatience, loss of self-control and
agitation, can take place.

10
As the disease progresses the patient faces growing difficulties in completing simple
tasks that involve abstract thinking and judgement. At the same time memory problems
become more severe and so is disorientation. These problems often result in
depression, as the person feels really confused and begins to withdraw.
At the final stages of the disorder, the patient is unable to move and express him/herself
or to recognize even the closest relatives. Death comes from complications of the
dramatic deterioration the disease causes.
Implicit memory in Alzheimer’s disease
Two of the most common symptoms of AD are severe memory loss and gradual
decline of language functions. Although the main characteristic of the dementia is
extensive impairment in the ability to intentionally recollect experience from the past,
there has been growing interest in the maintenance or not of a more indirect form of
memory, implicit memory that was mentioned earlier.
The scientific concern of this particular area was initiated by research from several
experimenters of implicit memory in amnesic patients.
At first we should specify as amnesics, the patients that as a result of a brain injury can
no longer retain new information for a long period of time. In other words, although
brain injury has left them with almost no impairment in other cognitive abilities, they
have great difficulties remembering new experiences for longer than a short-term
memory period ( Roediger , 1990 ).
The belief of the researchers, that the failure of amnesics to retain information was due
to their inability to transfer verbal information from short-term to long-term memory,
began to change around 1970. At that time, the first experiments examining implicit
memory in amnesics were conducted and the results showed that the patients were able
to exhibit intact implicit memory with verbal material, for periods of time wider than
those of short-term memory.
After numerous experiments, existence of amnesics’ intact priming in several implicit
tests was established and scientific interest in other memory impaired populations
began to rise. In the article " Implicit memory: a selective review " by Shachter, Chiu
and Ochsner ( 1993 ), there were evidence from experiments ( Cermak et al,
1985; Graf et al, 1984; Moscovitch, 1982; Warrington & Weiskrantz, 1974; Graf et

11
al, 1985; Schacter, 1985; Shimamura & Squire, 1984; Shimamura et al, 1987; Ober &
Shenaut, 1988; Keane et al, 1991; Martin, 1992; Cave & Squire, 1992; Schacter et al,
1991b; Gabrieli et al, 1990 ) on patients who suffered from disorders such as
Huntington’s disease, Parkinson’s disease and Alzheimer’s disease, showing that those
patients were able to exhibit intact priming in several implicit memory tests, inspite
their great inability to perform on explicit memory tests.
In the next part, we are going to present a review of all the research that has been
conducted so far regarding the performance of AD patients in the implicit memory test
of stem-completion. The purpose of this review is to cite the results of many
experimenters who attempted to illuminate the ambiguity among the findings on this
particular test, that caused great controversy in comparison to other memory tasks.

12
LITERATURE REVIEW
In 1987, Shimamura, Salmon, Squire and Butters, were the first to report a study of
patient populations measured in stem-completion priming. The three groups tested
were patients with alcoholic Korsakoff’s syndrome ( KS ), patients with Huntington’s
disease ( HS ) and patients with Alzheimer’s disease ( AD ). All three groups,
consisting of subjects with established extended disability in explicit memory,
demonstrated impaired performance in retention tasks ( free recall and recognition )
that required intentional recollection of word-material. More specifically, KS patients
exhibited performance similar to that of AD patients in all the explicit memory tests,
whereas HD patients, although impaired, did better in both free recall and recognition
tasks. When the three groups were tested in the implicit test of stem-completion, only
the AD group showed impaired priming. During study-phase, where subjects were
presented with the words on which they would be later tested, the experimenters used
a "rating likeability orienting task" ( subjects are instructed to rate each word according
to how much they like it ). By the term of "orienting task" is implied the certain kind
of manipulation that each investigator requires from participants to make on the
stimuli. The researchers suggested that the results could reflect failure in AD patients’
ability to activate lexical representations, leading, hence, to problematic semantic
memory, a prominent cognitive symptom of Alzheimer’s disease.
In the above report there was a reference to another study which was conducted the
same year by Salmon, Shimamura, Butters and Smith and was reported next year in
1988. Likewise, the patient groups consisted of AD patients, HD patients and KS
patients. The procedure followed was the same, except from the fact that target-stimuli
were presented twice and were the same stimuli to be used for assessing stem-
completion, recall and recognition as well. The results replicated those of the previous
study. Even though all patient groups performed poorly on the explicit memory tests (
recall and recognition ), only the AD group failed to show intact priming in the implicit
memory test of stem-completion. A potential explanation given by the authors, was,
that because of the disrupted organisation of AD semantic memory, there could be a
failure in the automatic activation of traces of previously presented words. In the
General Discussion part of both the above studies, evidence was provided that

13
impaired priming observed in AD patients, could not be attributed to global cognitive
impairment, since they were able to follow the instructions in order to perform the test.
Lexical priming in AD was also assessed by Heindel, Salmon, Shults, Walicke and
Butters ( 1989 ). The patient groups tested, were HD patients, AD patients and
Parkinson’s disease, demented and non-demented patients. The orienting task used
at study-phase was rating the likeability of words. AD patients were impaired on the
stem-completion test replicating all previous results reported, while HD and PD
patients primed normally. The fact that AD patients showed intact priming in a motor-
skill learning task although impaired lexical priming, whereas HD patients exhibited
the opposite pattern, prompted the authors to conclude that this double dissociation
was dependent upon distinct neuroanatomical systems.
Until 1990, all published data referring to AD performance in stem-completion, were
indicative of impaired priming. However, there had been some studies reporting equal
priming between patient populations and normal groups, when tested in other implicit
memory tests, such as reading speed ( e.g. Moscovitch, Winocur and McLachlan, 1986
). At that time the first intact-reporting studies for Alzheimer's patients in stem-
completion made their appearance.
In 1990, Grosse, Wilson and Fox, sought to investigate the extent to which the rating
likeability encoding process, could ensure semantic elaboration of target-stimuli by
the AD patients. At this point, we would like to refer to several findings coming from
studies on normal subjects that illustrated the significant role of the encoding process
in memory performance. What is the encoding process? By this term we mean the
procedure that subjects follow each time they are presented with certain stimuli, in
order to encode those stimuli and allow their entrance in their memory. The encoding
process is deeply dependent upon the way the stimuli are presented to the subjects. For
example, in the case that several words are presented by just being read by the subjects
one after another, it is more possible that subjects will follow a less profound way of
encoding these words, in the way of being more affected by the format of the words.
In the opposite case where subjects are led to manipulate the words by exploring their
meaning, they tend to proceed to an encoding process where they commit to the
semantic elaboration on the words. It has been found that the encoding process the

14
controls follow each time at the study-phase of an experiment, has a direct effect on
their performance in different memory tests. More specifically, it has been found that
when controls are prompted to use a profound encoding process by elaborating
semantically on the words, they tend to perform better at explicit memory tests ( e.g.
Jacoby and Dallas, 1981 ). However, when they follow a presentation of stimuli where
they are more affected by the perceptual characteristics of the words, they tend to be
more efficient at implicit memory tests ( e.g. Roediger et al, 1989 ). Accordingly,
when Grosse, Wilson and Fox ( 1990 ) studied all previous reports of impaired stem-
completion priming in AD patients, they suggested that the rating-likeability task used
by other experimenters, was not sufficient enough to ensure semantic elaboration on
words, since AD patients are well known for the growing deterioration of semantic
memory. They used, instead, a sentence-frame task in order to assure that AD subjects
would fully understand and adequately elaborate on the stimuli. The sentence-frame
task required the subjects to complete phrases where the missing words were the
target-stimuli. In this way, the experimenters could be sure that AD subjects fully
understood the meaning of the words, since they were able to meaningfully use them
in phrases. The findings revealed intact priming for the AD patients and an explanation
given was that the sentence-frame orienting task provided the AD group with rich
contextual encoding which activated the defective semantic network of AD patients.
Partridge, Knight and Feehan ( 1990 ), also manipulated the encoding process in
attempting to explain the failure of other experimenters to demonstrate intact
performance of AD patients in a stem-completion test. In their study, they, too,
suggested that the rating-likeability orienting task could not ensure the full
comprehension of words, when it comes to assessing implicit memory in patients with
a wider range of deficits, like AD patients. Accordingly, during study-phase they
adopted a process in which subjects had to provide the meaning of the target-words.
Their findings exhibited intact priming for the AD group, leading them to conclude
that results were due to full registration of words during study-phase, by following the
meaning orienting task. In accordance, Partridge et al ( 1990 ) suggested that the
particular manipulation at encoding was successful in assuring the subjects' attention
during test.

15
In 1990, Landrum and Radtke for the first time addressed the question of whether
severity of dementia could affect implicit memory performance of AD patients. In
previous studies, potential influence of this variable on the results has been considered,
but was evaluated to be rather unlikely. In their study, Landrum and Radtke ( 1990 )
tested in stem-completion and recognition, two groups of patients categorised in terms
of severity of dementia, as either mildly or moderately demented. The encoding
process took place by attending to each word for 5 secs, while at test-phase each stem-
completion problem included the first two letters of the word and three or four empty
boxes indicating the remaining letters. In addition, target-stimuli were divided into two
groups, of high and low-frequency respectively, and were counterbalanced across
study list. The results exhibited intact priming for the mildly demented group and
impaired priming for the moderately impaired group, while additionally, the
moderately impaired group showed less priming for low-frequency words. At the
recognition test both groups showed diminished performance. The authors concluded
that patients with severe cognitive impairment showed similarly affected implicit and
explicit memory performance, whereas moderately impaired patients could exhibit
intact priming in an implicit task ( stem-completion ), in spite of their explicit memory
deficit.
Scott, Wright, Rai, Exton-Smith and Gardiner ( 1990 ) used a two-conditioned
encoding process in their study. Subjects had to either provide to each word a
meaningfully related word, or to pronounce a letter within the word according to its
numerical position. By following this procedure, the experimenters meant to contrast
a deep processing with a less profound one, and to discover the magnitude of influence
that manipulation of encoding process had on explicit and implicit memory tests. The
findings indicated that despite poor performance of AD patients on the recognition
test, those same patients showed little impairment in stem-completion when compared
to the normal group. Manipulation of level had an effect only on the recognition test,
with semantic process leading to better performance for both groups. The authors
claimed, that results supported the idea of separate memory systems underlying
conscious recollection and implicit memory. They also speculated that the relative
stages of the dementia could play a significant role, suggesting that as the disease
progresses the patient' s performance deteriorates as well, and that could explain
previous contradiction in data.

16
In a study by Randolph ( 1991 ), encoding process followed rating likeability and
stimuli were presented three times to two patient groups ( consisting of HD patients
and AD patients, respectively ) and to the normal group. Randolph ( 1991 ) believed
that explicit memory performance measured by cued-recall should be correlated with
lexical priming in implicit memory tests. That was the reason for the rating-task
repetition, in order to avoid floor effects in explicit recall by the AD group. Randolph
( 1991 ) reported impaired priming in stem-completion for the AD group whereas the
HD group performed even better than the controls in the same test. His findings were
also indicative of a correlation between performances of explicit and implicit tests for
both patient groups. In addition, the fact that for the AD group there was also a
correlation between a weakening of semantic associations and episodic memory, led
Randolph ( 1991 ) to conclude that in Alzheimer' s disease, explicit, implicit, and
semantic memory functions may decline in parallel.
In 1991, Bondi and Kaszniak conducted four experiments testing AD patients and PD
patients on several implicit and explicit memory tasks. Their predictions concerning
stem-completion expected from PD subjects to show intact priming, whereas AD
patients were predicted to show impaired priming. The encoding process followed
rating the pleasantness of words. The results indicated impaired stem-completion
priming only for the AD group. Bondi and Kaszniak ( 1991 ) attributed AD group’s
poor performance to disrupted conceptual relationships within semantic memory.
They noted, however, that there had been other investigators ( e.g. Nebes, Brady
and Huff, 1989; Nebes et al, 1984 ) who reported different demonstrations, concluding
that the reasons for discrepant findings were unclear.
In 1991, Keane et al conducted a number of experiments aiming to demonstrate that
failure of AD patients to show intact priming on certain tasks, was not due to the nature
of stimuli but due to specific demands of the task. For that purpose, they tested the
same group of AD patients on two different lexical implicit tests, perceptual-
identification and stem-completion. They were assuming that perceptual-identification
demands a perceptual memory process which was localised to occipital lobe regions,
that seemed to be relatively spared in AD. Stem-completion, on the other hand,
according to Keane et al ( 1991 ), made use of a more conceptual processing, localised

17
to temporoparietal regions that were considered to be compromised in AD. During
study-phase of both tests, words of high as well as low-frequency were used, visually
presented and read aloud by the experimenter, one or three times. The results
demonstrated intact priming in perceptual identification and impaired priming in stem-
completion for the AD group. Although both groups provided more completions of
high frequency words than of low frequency ones, there was no effect of the frequency
variable on subjects' priming. The results in the recognition test replicated all previous
findings of impaired explicit memory in AD. The investigators supported the existence
of two components, a perceptual and a conceptual one, based upon neuroanatomically
distinct regions, intact and impaired, respectively.
In 1992, Christensen et al examined the hypothesis that cognitive impairment observed
in AD is attributed to deficient cholinergic neurotransmission ( Coyle et al, 1983 ), as
it was indicated by evidence showing depletion of cholinergic enzymes in AD (
Kopelman and Corn, 1988 ). Administration of certain drugs, such as scopolamine, are
known to induce “blockade” to central cholinergic neurons, leading to memory
impairment in normal young ( Nissen et al, 1987; Broks et al, 1988; Kopelman and
Corn, 1988; Rusted and Warburton, 1988, 1989 ) and older ( Sunderland et
al, 1987; Huff et al, 1988 ) subjects, as well as exacerbation of memory problems in
AD patients ( Sunderland et al, 1987; Huff et al, 1988 ). Since it was not yet clear
whether the amnesia and the cognitive deficits caused by scopolamine administration
were qualitatively different to those observed in AD, Christensen et al ( 1992 ) tested
three groups of mildly, moderately and severely impaired AD patients and a group of
healthy controls that were administered with scopolamine. All subjects were tested in
a variety of memory tasks, some of which were known to result to impaired
performance following administration of scopolamine in young subjects, and others
that remained unaffected. Stem-completion was included in the group of tasks
unaffected by scopolamine in healthy controls. Christensen et al ( 1992 ) suggested
that if cholinergic depletion was responsible for the deficits in the early stages,
impairments should be evident on tasks affected by scopolamine but not on tasks
unaffected by scopolamine in young subjects. During study-phase subjects were
instructed to make sentences by using given target-stimuli. The results indicated that
moderately and severely impaired patients showed impaired priming, whereas the
mildly impaired group exhibited priming quite similar to that of the normal group. The

18
major finding was that AD patients in general performed in a way that could not
support the "cholinergic model ", although the performance of mildly demented
patients mimicked that seen in cholinergic "blockade" in young subjects. Christensen
et al ( 1992 ) suggested that a number of variables affected priming in AD, such as
elaboration of encoding process or explicit memory intrusion to the performance of
normal participants. The findings also designated a direct relation of the severity of
dementia to the performance of patients on word-completion.
In a study of Perani et al in 1993, a group of mildly demented AD patients was
neuropsychologically evaluated in several aspects of memory functioning ( including
a stem-completion test ) along with a measurement of regional cerebral glucose
metabolism with [18F] fluoro-deoxyglucose ([18F]FDG) and PET. The encoding
process in stem-completion demanded that subjects learned a list of words auditorilly
presented every 2 secs. According to the investigators, half of AD subjects exhibited
impaired stem-completion priming. The multivariate regression analysis of the
Positron emission tomography demonstrated metabolic values of left thalamus, lateral,
occipital, and parietal areas, that, according to the experimenters, entered the best
predictive model for priming effect. Perani et al ( 1993 ) suggested that the brain
regions where the metabolic values were located, revealed involvement in lexical-
semantic processing ( left parietal cortex ), whereas activation of lateral occipital
cortex and left thalamus, could be interpreted as a correlate of more proficient
processing of the primed stimuli. The experimenters concluded that these results, in
combination with the activation of frontal associative cortex, basal ganglia and
cerebellum during the skill learning testing( areas clearly involved in motor control ),
support a model of partially independent memory components, distinct between them
in a neuroanatomical basis.
In 1993, Bondi et al tested a group of AD patients, three patients with a ruptured
anterior communicating artery ( AcoA ) aneurysm and one patient with bilateral
temporal lobe damage sustained from herpes encephalitis ( HE ), on several explicit
and implicit memory tasks. In stem-completion, AD patients and the HE patient were
predicted to show impaired lexical priming, because of damaged temporoparietal
cortex, which was thought to be involved in such a test. ACoA patients, on the other
hand, were expected to show intact lexical priming in the same test, because of primary

19
limitation of damage to basal forebrain structures. At study-phase, patients and their
matched control subjects were instructed to use the rating likeability task in order to
encode target-stimuli. As it was predicted, AD patients and the HE patient performed
poorly on the stem-completion test, whereas ACoA patients exhibited intact lexical
priming. Bondi et al ( 1993 ) concluded that the failure of AD patients and the HE
patient to demonstrate intact performance on the stem-completion test, was due to
temporal lobe cortex damage. Summarising, they supported Heindel’s ( 1989 )
conclusion about multiple implicit memory systems that could be differentially
impaired by disorders affecting distinct neuroanatomic circuits.
In 1994, Burke et al, seeking to clarify the discrepancies among results concerning AD
performance in stem-completion, conducted an experiment where patients were
compared with their matched controls in stem-completion, free recall and recognition.
The investigators aimed to illuminate the role of encoding process, by contrasting two
different orienting tasks: rating likebility versus an encoding process where subjects
were instructed to provide definitions of the target words. Furthermore, Burke et al (
1994 ) meant to investigate the stochastic relationship between the implicit test of
stem-completion and the explicit tests of recall and recognition, and the way in which
the latter tasks could influence performance of AD patients on the former test. The
assessments revealed impaired performance of the AD group on stem-completion and
failed to detect any interaction between the encoding process and the resulting
performance. Moreover, there was no significant correlation between the scores of
stem-completion and those of explicit tests; there was, nevertheless, a significant
correlation between free recall and recognition. The authors concluded that impaired
performance on stem-completion is a general characteristic of AD patients. They did,
however, make an attempt to explain the controversy caused by different findings.
Referring to Partridge’s research ( 1990 ), they attributed his findings of intact AD
performance, to the fact that the controls used in that study performed poorly compared
to controls of other studies reporting impaired AD priming. Consequently, Burke et
al ( 1994 ) assumed that this could have influenced the general appearance of results.
Gabrieli et al reported three experiments in a study, in 1994. In the first experiment, a
group of AD patients, two non-demented amnesic (AMN) patients and a control group
were measured for lexical priming in stem-completion. They were, additionally,

20
administered with recall and recognition tests. In the Introduction part of the study,
Gabrieli et al ( 1994 ) argued that stem-completion demands more of a conceptual
process than of a perceptual one, and that could be the cause of impaired AD
performance. They added that all studies reporting impaired performance, required
from subjects to follow semantic encoding of target stimuli; consequently, they
suggested that AD patients could show preserved intact lexical priming if the test was
conducted in a more perceptual way. Accordingly, the investigators compared three
different orienting tasks, each of them considered to demand different encoding
processes. The first one instructed all participants to locate a certain letter within the
word and that was the non-semantic condition, whereas the second one asked them to
answer questions semantically related to each word and that was the semantic
condition. A third condition was included, an intentional one, where the subjects were
instructed to remember each word shown to them. The results demonstrated impaired
performance of AD patients on all tests, regardless of the encoding task used, whereas
AMN patients showed lexical priming similar to that of controls. The orienting task
was significantly correlated with performance in all groups on recall and recognition,
with the letter condition linked to poorer performance than the other two conditions.
Gabrieli et al ( 1994 ) suggested that their findings indicate failure of the lexical
learning processes in AD to render a word more accessible after this word has been
recently processed. They concluded that stem-completion deficit could reflect a kind
of anterograde amnesia in AD.
In 1994, a study by Russo and Spinnler was published, consisting of two experiments.
In the first experiment, a stem-completion test was administered to AD patients and
normal subjects, along with a cued recall test and some other fluency tasks. Russo and
Spinnler ( 1994 ) sought to elucidate the impact that orienting task had on subjects’
lexical performance, questioning to what degree AD subjects fully registered the
meaning of words when they were asked to rate them according to how much they
liked them. Therefore, during study-phase, they instructed all subjects to use the rating
likeability orienting task for one list of words and to provide the meaning of the words
presented to them in a second list. The assessment exhibited intact priming for the AD
group in both encoding conditions, whereas the impaired performance patients showed
in the explicit memory test ( cued recall ) was affected by the nature of encoding task.
This finding, according to Russo and Spinnler, could challenge the views of Partridge

21
et al ( 1990 ), Grosse et al ( 1990 ) and Christensen et al ( 1992 ) who supported the
idea of preserved lexical priming when the encoding task involved more effective
semantic processing. In the second experiment, Russo and Spinnler tested the same
AD group in perceptual-identification, replicating all previous findings of preserved
priming in that test. The investigators went further by conducting a meta-analysis of
all previously published data, that concerned AD performance on stem-completion and
perceptual-identification. After combining the results of their study in stem-
completion with ten further studies that examined stem-completion in AD, Russo and
Spinnler ( 1994 ) showed impaired priming. After conducting a similar meta-analysis
to word-identification studies with AD patients, the results pointed to a preserved
repetition priming in AD. That dissociation between stem-completion and word-
identification was further supported after those memory tasks were directly contrasted.
Accordingly, they were led to the conclusion that those two implicit memory tests
might apply to different cerebral areas, suggesting that stem-completion in particular,
could involve more anterior areas than word-identification, and underlined the need
for further investigation.
In 1994, Huberman, Moscovitch and Freedman reported a study where two groups of
PD and AD patients respectively and their matched controls were tested in two
different implicit memory tests. The first test was reading transformed script, a skill
learning test where AD patients were predicted to prime normally, whereas PD patients
were expected to show impaired performance. The second test was a stem-completion
task where PD patients were predicted to show intact priming, followed by a slightly
if at all poorer performance from AD patients. At stem-completion study-phase,
subjects were asked to read and attend each word presented to them for 3 seconds. The
actual test-phase took place 5-7 minutes after the study-phase. In the meantime the
subjects were presented with irrelevant to the studied items trigrams and were asked
to provide the first person's name that came to mind; the same procedure was followed
for other irrelevant trigrams where the subjects had to provide names of places. PD
patients were found to perform normally on both implicit tests, whereas AD patients
performed normally on the skill-learning test and showed nearly intact priming in
the stem - completion test. Huberman et al ( 1994 ) supported the idea that, in
their study, stem-completion was more of an implicit memory test, because it was well
disguised due to the distractor tests and the prolonged interval between study and test-

22
phase. Accordingly, normal subjects with adequate memory would not notice any
relation between the studied words and the stem completion test-phase, and, hence,
were prevented from using explicit strategies. According to the experimenters, that
could serve as a factor affecting the results of other studies where study-phase was
immediately followed by the test-phase and where AD patients were found to perform
poorly compared to the controls. Huberman et al ( 1994 ) suggested that AD patients
could exhibit intact stem-completion performance under optimal conditions, such as
when their dementia is mild, the words are common and normal subjects are prevented
from using explicit memory. They concluded that preservation of priming in AD
patients is dependent, among other factors, mostly upon the integrity of the neocortex
regions and the conceptual or perceptual demands of the test.
In 1994, a report by Deweer aimed to assess several aspects of AD patients’ memory
capacity. The test battery included a stem-completion test, in which the investigators
made use of the rating likeability orienting task. The patients were divided into two
groups in terms of their being institutionalised or outpatients. Deweer et al ( 1994 )
found intact lexical priming for both AD groups and went on attempting to account for
the contradiction among the results of several studies. At first, they referred to the fact
that preservation of the physical characteristics of stimuli during study and test-phases
might have influenced the patients’ performance. They claimed that in their study,
where the format of words remained the same, AD subjects primed normally, whereas
in studies that had reported opposite findings, the format of words was changed from
study to test-phase. The authors continued by adding the factor of the perceptual or
conceptual nature of the task. They suggested that the orienting task and the physical
similarity of the stimuli, prompted the subjects to draw upon a more perceptual
analysis of items. Deweer ( 1994 ) et al ended up by referring to the great variability
observed on performances of both AD groups and the elderly controls, and called
attention to the need for further investigation of memory abilities in elderly normal
subjects.
In 1995, a study by Carlesimo et al was published, which consisted of a parallel
evaluation of a stem-completion test along with assessments of explicit memory tests,
lexical-semantic evaluation tests and general intellectual efficiency tests. The subjects
involved were AD patients, multi-infarct demented patients ( MID ) and their controls.

23
The aim of the study was to investigate the degree to which AD subjects’ lexical
priming is dependent upon explicit memory deficits or upon deterioration of lexical-
semantic knowledge. The orienting task used was rating the likeability of words every
5 seconds. Carlesimo et al ( 1995 ) found impaired priming for AD patients, whereas
after a multiple regression analysis they established strong correlation between lexical
performance of AD patients on stem-completion, and their performance on graphemic
cued recall. The authors argued that their results were compatible with the transfer-
appropriate approach, which supports the idea of the division of memory tasks,
according to whether they draw upon conceptual or perceptual processing. Carlesimo
et al ( 1995 ) opposed to the idea of a multiple memory-system theory ( Graf and
Mandler, 1984; Rozin, 1976 ) which suggests that repetition of words activates
previously established representations in lexical-semantic memory, and thus leads to
repetition priming. They argued that their results failed to detect the power of lexical-
semantic evaluation tests to predict AD performance on stem-completion. The
investigators, however, pointed out the weakness of both theory models to account for
the strong dissociation between amnesic patients who had been found to prime
normally in stem-completion, and AD patients who performed poorly on the same test,
although both patient groups exhibited similar mnemonic problems when they were
assessed on explicit memory tests. Carlesimo et al ( 1995 ) suggested that, that
dissociation could be interpreted as follows: amnesics could have an automatic but not
deliberate access to memory traces and thus could succeed in priming normally,
whereas AD patients could not have access in either way. The authors concluded that
this differentiation stems from a neuroanatomical distinction between the lesions of
the two patient groups.
In 1995, Randolph et al examined directly the possibility that failure of AD patients to
show intact priming in stem-completion, was due to their explicit memory deficit. In
order to investigate the potentiality of that explanation, Randolph et al ( 1995 ) equated
the AD group and their matched controls in terms of explicit memory performance, by
manipulating the number of exposures of the stimuli. In other words, subjects were
instructed to read aloud and rate for likeability a list of words presented to them once
and a list presented to them four times; the four-times exposure was thought to achieve
equivalent assessments of explicit memory for both groups. The results revealed
similar performance for both groups and on both the explicit test ( cued recall ) and the

24
implicit test ( stem-completion ), but only for words presented four times to the AD
patients. Randolph et al ( 1995 ) suggested that the findings didn't support the idea of
a selective implicit memory deficit in AD but were rather indicative of a single
memory system underlying performance on implicit or explicit tasks. Alternatively,
they introduced as a less parsimonious explanation the existence of two memory
systems, an implicit and an explicit one, equally impaired in AD but also equally
responsive to encoding manipulations.
In 1996, Downes et al reported a study that described three experiments, all testing AD
performance on a stem-completion test. Subjects were examined not only for lexical
priming but also for cohort priming. Cohort priming was designated as the tendency
subjects showed to fill in the trigrams with words articulated in the same manner as
the target-stimuli. That is, even when subjects didn't fill in the trigrams with target-
stimuli, they tended to use words that “sounded” like the target-stimuli. In the first
experiment subjects were told that they should try to remember the words presented to
them and then they were instructed to rate for likeability each word presented to them
for no more than 5 seconds. The results showed impaired lexical and cohort priming
for AD patients, whereas the controls showed intact priming of both types. In the
second experiment subjects were presented with non-words and then were asked to
read each word and rate it according to word-likeness; they were also instructed to try
to remember as many as they could. The results revealed intact cohort priming for both
groups. In the third experiment subjects were told that they should not have to try and
remember words, only to pick out the correct one from two alternatives. Then were
instructed to either read each word aloud and try to commit it to memory or to rate it
according to how much they liked it. Controls exhibited normal priming under both
encoding conditions, whereas AD patients primed normally only after using the
“reading aloud” orienting task. Downes et al ( 1996 ) interpreted the findings as
providing evidence that stem-completion was not so much a conceptually-driven task
which led AD patients to perform poorly, as was a task which included two basic steps:
first the stem was translated into some phonological form and subsequently a selection
of the appropriate lexical entry took place in the phonological representation system.
The experimenters argued that it was obvious by the ability of the same AD patients
to show intact performance following the reading orienting task and not the rating

25
likeability task. Downes et al ( 1996 ) concluded that according to their approach, stem-
completion was a phonological-driven rather than a conceptually-driven test.
In 1997, a report by Fleischman et al was published. In the Introduction part the authors
presented a table which included most of the studies examining AD performance on
stem-completion. Fleischman et al ( 1997 ) pointed out the fact that in most cases
where AD subjects primed normally, generation conditions were followed ( words
were generated by the subjects at the study-phase ), whereas impaired performance
was observed following a reading or rating likeability orienting task. The investigators
compared the two encoding processes, referring to the latter one as a task which
maximises perceptual processing of words and to the former as a task which
maximises conceptual processing of words. Accordingly, at study-phase, subjects
were instructed to either read aloud words or to generate them by meaning. The
assessments revealed significant priming for both groups and under both encoding
conditions. There was, however, greater magnitude of priming following the reading
rather than the rating likeability condition for controls and AD subjects as well.
Fleischman et al ( 1997 ) examined a number of factors that could contribute to
ambiguity in the results of several studies, none of which could account in a
satisfactory way for the contradiction that had been observed. The authors concluded
that the question of whether AD patients could show intact priming in stem-completion
or not, should no longer be critical, and drew attention to the need for a more complex
theory which could serve as an explanation for both kinds of findings.
In 1998, Koivisto et al demonstrated intact stem-completion priming in a group of AD
patients, after using a reading aloud orienting task for words presented to them every
4 secs. The authors, at the beginning of the article, suggested that failure of AD patients
to exhibit intact stem-completion priming in certain studies, could be attributed not to
their failure of automatic memory but rather to contamination of results by controlled
uses of memory from the normal subjects. In order to examine that potential
explanation, some researchers ( Jacoby 1991; Jacoby et al, 1993; Toth et al, 1994 )
had used two variations of the stem-completion procedure, an inclusion and an
exclusion one, known as "the process dissociation procedure". In the inclusion task,
controls were asked to complete each stem with a studied word, whereas in the
exclusion task they had to provide to each stem a word they had not encountered

26
before. The idea was that in the inclusion stem-completion task, both the controlled
and the automatic memory processes work together in order to increase the probability
for correct responses. In the exclusion task, on the other hand, these two memory
processes are in opposition, as the automatic memory increases the probability and
controlled memory decreases the probability of completion with studied items.
According to some investigators ( Jacoby 1991; Jacoby et al, 1993; Toth et al, 1994 ),
two equations, after being assessed, can estimate the values of the above memory
processes. More specifically, the controlled recollection (R) can be estimated by
subtracting the probability of providing a studied word in the exclusion task from the
probability of providing a studied word in the inclusion task: R=Inclusion-Exclusion.
Automatic memory (A), on the other hand, can be estimated by the following equation:
A=Exclusion/(1-R). Koivisto et al ( 1998 ) tested their subjects in the stem-completion
task and both the inclusion and the exclusion procedures. Although the AD patients
were found to prime normally on stem-completion compared to the normal subjects,
differences between the two groups appeared when the equations were estimated. After
assessing the controlled memory of each group by subtracting the completion rates
for studied words in the exclusion task from the completion rates for studied words in
the inclusion task, AD patients were found to have impaired controlled memory
compared to the normal subjects. Automatic memory, on the other hand, was preserved
in AD patients, similarly to the normal participants, as the estimation of the respective
equation revealed. The authors suggested that the reason AD patients failed to show
intact priming in other studies, was contamination of the results by the explicit or
controlled memory of normal subjects. They suggested that in AD there was preserved
automatic and impaired controlled memory and that dissociation led them to conclude
that different memory processes were mediated by distinct neural systems.
In the third experiment of a study by Park et al ( 1998 ), a group of AD patients and
their matched controls were tested in picture naming and stem-completion. The
materials used in stem-completion study-phase were the same pictures and words used
in the picture-naming test. The investigators illustrated intact priming for AD patients
and put forward as a consideration that at least at the beginning of disease, AD patients
analysed the perceptual characteristics of words and pictures, in the same manner that
normal population did.

27
In 1999, Fleischman et al sought to examine in which way could severity of
Alzheimer’s disease and the potential reduction of priming in healthy old controls
could have an effect on the discrepant results found in stem-completion. In other
words, Fleischman et al ( 1999 ) investigated the possibility that in several reports
where AD patients demonstrated intact performance in stem-completion, the results
were somehow “altered” by the poorer performance of elderly subjects in comparison
to younger controls. Consequently, they tested large samples of young, old and AD
subjects in stem-completion, examining thoroughly the effects that age, cognitive
status and encoding had on their priming abilities. There were three orienting tasks
used: subjects read words aloud, or read words aloud and rated them for likeability, or
generated words from a short definition. Fleischman et al ( 1999 ) showed intact
priming similar to that of older normal controls, only for very mild patients, i.e. those
who had scores of 26 or more in a Mini-Mental State Examination ( MMSE ). They
also came up with more findings. Magnitude of priming was reduced for healthy old
normals as the age became greater. There was a reduction of priming in the AD group
correlated with the severity of disease. Finally, there was a disproportionate reduction
for priming following the generate encoding condition but only for healthy old
participants. The authors attributed the controversial results of several studies to the
fact that the samples used contained patients of different stages of the disease. They
suggested that, according to their findings, there was great influence on lexical
priming as the disease progressed. They also referred to the fact that in their
study old controls aged 75 years and more, showed lexical priming similar to that
shown by AD patients of early stages. Thus, they noted that non-significant group
effects might occur under similar conditions where AD patients were very mildly
demented and healthy participants very old.

28
META-ANALYSIS
It is obvious according to the previous review of articles, that the discordant results caused a
debate. A number of explanations were introduced by several experimenters in order to delineate
the main factors that contributed to the ambiguity of results. Therefore, we conducted a meta-
analysis of the published data in order to investigate the power of the suggested variables that
could be accounted for the variability of the results, according to the theoretical approaches of the
researchers.
Meta-analysis is a general term to describe statistical techniques that allow us to analyze the
pattern of findings from a number of studies. The main objective of meta-analysis is to assess the
strength of relationships over a range of studies by combining them, if possible, into a single
indicator of the relationship, and to assess the influence of various characteristics of the studies on
the strength of the relationships found ( Howitt, Cramer, 2003 ).
Effect size is the central concept in meta-analysis and indicates the amount of relationship between
two variables, the size of the effect that one variable has on another. The two most common
measures of the effect size are the Pearson correlation coefficient and the Cohen's d. Because less
calculation are involved in the estimation of Pearson correlation coefficient from the minimal
information the researchers supply than the Cohen's d, in our meta-analysis we used the Pearson
formula. More specifically, Cohen's d is the difference between the mean of one group of
participants and the mean of the other group adjusted by dividing by the standard deviation of the
scores, whereas the Pearson correlation coefficient ( r ) is a number between -1 and 1 that can be
calculated by using the test of significance reported in the analysis of each study. The larger the
correlation coefficient between two variables, the larger the effect of one variable on the other (
Howitt, Cramer, 2003 ).
As we mentioned earlier, several investigators pointed out certain study characteristics that, in
their opinion, could be responsible for affecting the results in various manners. In order to evaluate
these aspects we categorized the combination of effect sizes according to the main issues that were
mentioned by the researchers thus far.

29
Categories of meta-analysis combination
i.sample size:
In all studies the number of AD subjects varied from 7 to 91. Therefore we divided the reports into
two categories: a) those that included up to 10 AD subjects, and b) those including more than 10
AD subjects.
ii.difference between the mean ages of the AD and the control group
Very few references have been made to the distance between the mean ages of controls and AD
patients throughout the reports ( e.g. Burke et al, 1994 ).
As we can see in the Review Table, from a total of 25 studies, in 16 cases the samples of AD
patients were older than the controls. In one case ( Christensen et al, 1992 ) where the three patient
groups represented different stages of Alzheimer' s disease, only the severely impaired group was
younger than the control group. Finally, in 8 cases normal subjects were older than the patients
and in one case the mean ages of both groups were equal.
Because one can suggest that the differences between the priming of patients and their matched
controls are due to age differences that reflect not only the dementia deterioration but also the
aging deficit on the priming ability, the two categories we used were: a) studies where the two
groups compared had an age difference more than 3 years, and b) studies where the age difference
between groups was less than 3 years.
iii. dementia severity
The role of severity of the AD dementia was the subject of debate in several published studies.
Some investigators merely referred to the potential influence that severity of AD could have on
the priming ability of patients in stem-completion. Others went further by examining whether
different stages of the disease resulted in different performances among the samples.
Shimamura, Salmon, Squire and Butters ( 1987 ) who tested a group of mild to moderate AD
patients, strongly suggested that impaired priming could not be attributed to a global intellectual
or cognitive impairment of the patients, since they were able to follow the instructions and perform
the test successfully. Furthermore, Shimamura et al ( 1987 ) claimed that the baseline guessing
rates were normal comparing to the control subjects, implying that AD patients did not produce

30
unusual words as responses to the stems. Finally, in that study ( 1987 ), another group of patients
suffering from Huntington' s disease, even though was matched in terms of severity of dementia
with the AD group, managed to demonstrate intact priming.
A similar pattern of arguments was followed by Salmon, Shimamura, Butters and Smith ( 1988 )
accounting for impaired performance in a sample of AD patients. Two other patient groups
consisting of HD and KS patients respectively, exhibited intact priming, despite the fact that KS
subjects were similarly to AD subjects impaired in explicit memory.
Heindel, Salmon, Shults, Walicke and Butters ( 1989 ) shared the same position. Although the
performance on the Dementia Rating Scale ( Drs scores ) of both the HD and the AD groups were
almost equal ( HD: 120,6; AD: 118,3 ), only the former group of patients managed to prime
normally.
Partridge, Knight and Feehan ( 1990 ), on the other hand, who demonstrated intact priming for AD
subjects, also disregarded dementia severity as being responsible for discrepant results. They
attributed their findings to the manipulation of orienting task, and not to other reasons such as
differences within samples. They mentioned that the patients tested were older and more demented
than the subjects used in other, impaired-reporting studies.
Grosse et al ( 1990 ) who also reported intact priming, based, likewise, their findings on the
manipulation of encoding process. They considered unlikely ( although didn't completely reject )
the fact that results could be due to patient differences, since the dementia of the subjects was
comparable to the dementia of subjects used in other studies of opposite findings. They did,
however, mention that preserved priming could be limited to a subgroup of AD patients whose
neuropathology was less extensive.
Fleischman et al ( 1997 ) examined stem-completion priming in a moderately demented AD sample
and reported intact priming. The study-phase consisted of two conditions: subjects had to either
read aloud the words or to generate them by their meaning. Fleischman et al ( 1997 ) did not find
any correlation between the MMSE scores and the lexical priming of patients; dementia severity
was not significantly correlated with the priming results in either the first ( R2=0.10, P>0.09 ) or
the second ( R2=0.06, P>0.22 ) condition. In an attempt to analyze the discordance of opinions in
all previously published data, among other factors they referred to the severity of disease as a
potential variable affecting the assessments. They assumed that it could not apply for
discrepancies, because of lack of adequate evidence towards this direction. They added that there

31
had been impaired priming demonstrated by mildly demented patients, but also intact priming
demonstrated by severely demented patients. Furthermore, they claimed that in a few studies where
correlation between dementia severity and the performance of patients was investigated, it was
either significant or non-significant. Thus, the authors emphasized the need for systematic and
longitudinal studies that would employ bigger samples of patients and could provide more
compelling evidence about the role of this variable. They concluded that it is possible that a
specific cognitive deficit accounts for impaired stem-completion priming and not global
intelligence.
There were, however, some investigators who established a correlation between the severity of
dementia and the stem-completion performance.
Landrum and Radtke ( 1990 ) were the first to report such a correlation in a study where patients
were divided into two groups according to severity of dementia, as either mildly or moderately
demented. As we have already mentioned, however, the patients suffered from various causes of
dementia such as organic brain syndrome, senile dementia or Alzheimer’s disease. Landrum and
Radtke ( 1990 ) reported intact priming for mildly demented patients whereas moderately
demented patients were impaired, particularly for low-frequency words. The researchers exhibited
positive correlation between the MMSE scores of the patients and the performances on both stem-
completion and recognition, although correlation was stronger for the recognition test. The
investigators concluded that implicit memory is less affected by the cognitive state of subjects,
underlining, however, the fact that mildly demented patients prime normally in a stem-completion
test, whereas moderately impaired patients do not.
Scott et al ( 1991 ) found little impairment and concluded that the ambiguity of previously reported
results, could be indicative of different stages of the disease. They suggested that AD patients
could prime normally, at least at the initial stages of disease and before a greater reduction of
cognitive abilities took place.
Christensen et al ( 1992 ) employed three groups of AD patients in a study examining the
cholinergic hypothesis. The groups consisted of mildly, moderately and severely impaired patients
who were tested on several aspects of memory. In stem-completion, intact lexical priming was
demonstrated for mildly demented subjects, whereas the other two AD groups were impaired.
More specifically, AD patients who scored above 20 on the MMSE followed a performance similar
to that of normal participants.

32
Huberman et al ( 1994 ) directed attention to certain optimal conditions under which they believed
AD patients could exhibit relatively intact lexical priming. Among others, they referred to the case
when dementia of AD patients was mild, suggesting, therefore, that severity of disease affects the
performance of patients.
Gabrieli et al ( 1994 ) used a Spearman rank order analysis between the BDS scores of AD subjects
and their total priming scores and revealed strong negative correlation, which indicated that greater
severity of dementia led to greater reduction in stem-completion priming. Gabrieli et al ( 1994 )
supported the idea of preserved presemantic perceptual representation system(s) for objects and
words, during the early stages of the disease.
Park et al ( 1998 ) who tested AD patients in picture naming and stem-completion and reported
intact priming, suggested that patients at the early stages of the disease proceeded to the perceptual
analysis of an object in a way rather abstract than specific, similar to that used by normal people.
Koivisto et al ( 1998 ) maintained that neural systems which were responsible for automatic uses
of memory, did not seem to lose their power at the early stages of Alzheimer' s disease.
The most systematic study that has been conducted thus far in order to examine the influence of
dementia severity on stem-completion performance, was that of Fleischman and Gabrieli ( 1999 ).
They used a sample of 91 AD patients whose severity of dementia ranged from very mild to
moderate. The assessments revealed that AD patients whose MMSE scores were greater or equal
to 26, demonstrated lexical priming similar to that of old normal participants, whereas patients that
scored less than 26 exhibited reduced lexical priming. The investigators concluded that AD
patients of very mild dementia could prime as normally as control subjects of late old age.
According to the above studies, there were some experimenters who demonstrated variability of
results according to differences in the severity of disease. They came to the conclusion that the
magnitude of demonstrated priming was related to the severity of dementia the AD subjects were
suffering from. On the other hand, there were other investigators who argued that when AD
patients were compared to patient groups, such as HD, only AD subjects were impaired in stem-
completion, even though their cognitive abilities were similar and so was their explicit memory
performances. It is difficult to attempt a definite conclusion about the degree to which severity of
disease affects the results. It seems, though, that it is a variable which calls for greater examination
and should be investigated while considering the effect that different stages of the disease have on
the way patients cope with the nature of the stimuli. In other words, gradual loss of semantic

33
memory might be reflected by differences in the difficulty patients face in exhibiting intact priming
during different stages of the disease. Nevertheless, this is a question directly connected to the
question about the nature of stem-completion which will be discussed later.
The two study categories we compared were:
a) studies that employed AD patients of mild to moderate severity and,
b) studies that employed AD subjects whose severity ranged from mild to severe.
iv.number of exposures of stimuli:
As seen in the Review Table the stimuli during the study-phase in some cases were presented
more than once to the participants. Therefore we compared the effect sizes from studies that : a)
were conducted with 1 exposure of the stimuli during the study-phase, and b) included more
than 1 exposure of the stimuli during the study-phase. In one case ( Randolph et al, 1995 ) two
effect sizes were attained as a result from a direct comparison between 1 and 4 exposures.
v.orienting task
As we can see at the Review Table, seven out of eleven studies exhibiting impaired priming, made
use of the rating-likeability orienting task. In two further impaired-reporting studies the
experimenters compared the rating likeability task to other, more profound or more superficial
encoding processes. However, priming was demonstrated as impaired, regardless of the orienting
task used. Nevertheless, most of the studies that reported impaired AD performance made use of
the rating likeability task.
In only one out of nine reports that exhibited intact priming, was rating likeability used, whereas
in one more report, the same task was contrasted with an encoding process where subjects had to
provide the meaning of each word; lexical priming was intact under both conditions. In the
remaining seven studies, there were a variety of orienting tasks used. The studies in which more
than one encoding condition were used, reported intact priming for all conditions compared, except
for one ( Fleischman, Gabrielli and Rinaldi, 1997 ), where reading aloud produced
greater priming than generating by meaning.
There is a third group of studies which reported mixed results, according either to the severity of
disease ( meaning there was intact priming for mild or very mild AD patients and impaired for
more demented patients ) or according to other experimental manipulations. Rating likeability was

34
used along with other orienting tasks, and only in one study ( Downes et al, 1996 ) did different
encoding conditions cause mixed results. More particularly, in Experiment 2 the AD group primed
normally under the reading aloud process, whereas performance following rating likeability was
impaired.
Our opinion is that it is very difficult to make clear-cut assumptions about the role of orienting
task or whether it is a significant variable which could account for the controversy. Rating
likeability was used in nine out of eleven of the studies that reported impaired priming ( in two
studies there were other encoding conditions used as well ). There has been a great debate about
whether this particular orienting task can be considered to be followed by a “deep” or
“shallow”encoding process and whether it can ensure the full registration of the meaning of each
word on behalf of AD patients. We suggest that rating likeability is a quite dubious task that is
difficult to be categorized as either a condition where subjects follow a superficial encoding
process of the words or either a condition when subjects follow a more profound encoding process.
Normal controls and AD patients could differ individually according to factors that affect their
liking or not of the words. More specifically, some could be influenced by the “shallow”
characteristics of each word such as if it consists of many vowels or consonants, or whether it is a
noun or a verb, or by the general sound of its pronunciation, whereas others could proceed to a
more “deep”, conceptual processing of the word, such as by linking it to certain habits, or certain
events of their lives or even elaborate on the word by imagining pictures of it in a certain setting.
Maybe this is the reason why in at least three studies, intact priming was reported for AD patients
under the rating likeability condition. Consequently we categorized the studies as those that : a)
made use of the rating likeability orienting task during the study-phase, and those that b) followed
any other studying condition.
Studies where rating likeability was contrasted with other orienting tasks without reporting the
results separately for each orienting condition ( Russo & Splinler, 1994; Downes et al, exp. 3,
1996 ) were excluded.

35
Methodology
There were two ways to report the findings of the relevant studies. We had to either combine the
effect sizes resulting from the priming ability of the AD group or the effect sizes given by the
comparison between the priming scores of the two groups, the normal participants and the AD
participants. The priming of AD patients was provided throughout the studies by the statistical
analysis performed between the target-stems and the baseline-stems completed by the patients,
whereas the group comparison is translated into the statistical comparison between the
performances of the AD group and the normal participants. From the body of 25 studies, only 7
provided the required information that would allow us to calculate effect sizes for AD priming,
whereas in the second case, 15 studies provided us with the information needed. In one case
( Bondi & Kaszniak, 1993) there were no statistical comparisons performed and another one
( Landrum and Radtke, 1990 ) was excluded because of the heterogeneity of the patient group
regarding the causes of the disorders. Therefore, we calculated an overall effect size for AD
priming and another for the group comparison, and continued to additional combinations between
studies by using the group comparison size effects. The formulas used to calculate the effect sizes
from the t and F values respectively were:
r =
dft
t
2
2
and r =
N
z
where z is the converted significance level by using standardized tables ( Howitt, Cramer, 2003 ).
Following, in order to combine these effect sizes from all studies, we converted each effect size
correlation r obtained by this procedure into a Fisher zr, averaged them and converted the average
zr back to an effect size. The overall significance of the combined studies is given by each
significance level turned into the corresponding z-score. These z-scores are then summed and
divided by the square root of the number of significance levels used ( Howitt, Cramer, 2003 ).
Tables 1 & 2 present the studies used in estimating the effect sizes for group comparison and AD
priming respectively.

36
RESULTS
Calculation of the average group-comparison size effect
The average zr was 0.272. This value of zr corresponds to an average of the effect sizes of
0.27.
The value of z found as z = 4.21 is converted back to a significance level; thus the combined
significance level is 0.001.
Calculation of the average AD priming size effect
The average zr = 0.751 corresponds to an average of the effect sizes of 0.64.
The value of z = 5.54 results to the combined significance level of 0.001.
EFFECT
,63,50,38,25,130,00-,13-,25-,38-,50
group-comparison size effects
5
4
3
2
1
0
Std. Dev = ,36
Mean = ,24
N = 18,00

37
i.sample size:
a)up to 10 AD subjects:
The average zr = 0.480 corresponds to an average of the effect sizes of 0.45, at the overall
significance level of z=4.065=0.001.
b)more than 10 AD subjects:
The mean effects of these categories do not differ statistically (
t=1.223,df=8.490,p=0.254 )
ii.difference between the mean ages of the AD and the control group:
a) more than 3 years
significance level of z= 4.590=0.001.
b) less than 3 years
EFFECT
,700,675,650,625,600,575,550
AD priming size effects
2,5
2,0
1,5
1,0
,5
0,0
Std. Dev = ,06
Mean = ,633
N = 7,00

38
The mean effects of these categories do not differ statistically (
t=1.660,df=5.774,p=0.150 )
iii.dementia severity
Difference between
a) studies that employed AD patients of mild to moderate severity
b) studies that employed AD subjects whose severity ranged from mild to severe
The mean effects of these categories do not differ statistically ( t=-
0.002,df=8.955,p=0.998 )
iv.number of exposures of stimuli:
a) 1 exposure
significance level of z=3.177= 0.1.
b) more than 1 exposure
The mean effects of these categories do not differ statistically ( t=-
0.816,df=10.555,p=0.4333 )
v.orienting task:
a) rating likeability
significance level ofz=3.702=0.01.
b) any other task

39
The mean effects of these categories differ statistically ( t = 3.706, df = 10.765 , p =0.004 ),
with the mean effect 0.52 (=0.578) of category a) being significantly higher than the mean effect
0.06 (=0.057) of category b).
DISCUSSION
As we have mentioned before, correlation coefficient r is a number between -1 and 1. Its negative
or very low ( close to zero ) values are given from the F or t values that exhibit the lack of an effect
between variables. Consequently, in our case, studies that managed to report a statistically strong
dissociation between the priming scores of the normal and the AD participants, would result to
higher effect sizes, whereas the lower effect sizes would come from studies that exhibited no
difference between the two groups of subjects. The overall combined effect size of group-
comparison was 0.27, which is regarded as a medium-power effect size. The AD priming
combined effect size, on the other hand, was 0.64, a value which is considered to be of quite a
large power, and shows a statistically great dissociation between the target and the baseline
completions of the AD patients. Unfortunately, all 7 studies that participated in this priming effect
size combination, where studies that reported such a dissociation ( see Table 2 ). There were other
studies that reported no statistical significance between target and baseline completions but they
did not provide the relevant information in order for size effects to be calculated ( eg.Carlesimo
et al, 1995 ). Nevertheless, three of the studies entered in the priming effect combination were
studies that reported significant dissociation between the performances of AD patients and controls
( see Table 2 ).
The lack of statistical significance between the two categories of sample size show that the number
of AD patients participating in the experiments does not affect significantly the outcome of studies.
However, the effect size (0.45) of the first category ( groups consisting of up to 10 AD subjects )
is larger than the average effect size (0.26) of the second category ( groups consisting of more than
10 AD subjects ), indicating a tendency towards greater dissociation between groups when the
patient sample is more limited.

40
A similar assumption can be made about the combination of studies where the AD patients had a
mean age difference of more than 3 years with their controls and studies where the mean difference
between the two groups was less than three years. No statistical significance can account for
variability between the two categories, although greater age difference indicates greater
dissociation between groups, as shown by the mean effect sizes of the two categories( 0.45 and
0.19 respectively ).
As far as dementia severity is concerned, not only there was no statistical significance between
studies that employed AD subjects of mild and moderate severity and those that also included
severely demented patients, but also the average size effects were similar (0.31).
Statistical indifference occurred between studies that used a single presentation of words during
study phase and those that made use of more exposures of the target stimuli. Nevertheless, although
one would expect the opposite pattern, the mean effect sizes of the first (0.21) and the second
(0.37) category, suggest a tendency towards greater dissociation between groups after following
more than a single exposure.
The only statistical significance that occurred was that for the orienting task combination of
studies. Studies that employed the rating likeability as the encoding condition during study-phase
managed to exhibit a great dissociation between the performances of the AD and the normal
participants ( as illustrated by the large average effect size of 0.52). The mean effect size (0.06 )
of the second category of studies where other orienting tasks were used, reached zero indicating
that the priming scores of the two groups were similar ( there was no group effect established ).
CONCLUSIONS
There are two ways to approach the issue of the priming ability of AD patients in stem-completion.
The first one is to examine the difference of their priming magnitude when they are compared to
normal participants and the other is to estimate their general ability of stem-completion priming
by statistically analyzing the difference ( if any ) between targets and baselines. As mentioned
above, the average effect sizes throughout the reports for both aspects indicate that although stem-
completion priming is present for AD patients, is significantly lower than the stem-completion
priming exhibited by normals.

41
Our suggestion is that since AD patients have been found to prime normally in other implicit tasks
( such as word-identification ) that apply to cerebral areas less affected by the disease, the results
of this meta-analysis reflect the twofold nature of stem-completion. Therefore we suggest that
stem-completion consists of two components, a perceptual and a conceptual one, and the amount
of priming lost when the patients are compared to normal participants corresponds to the
conceptual demands that this particular memory tests makes on the subjects. Compatible with this
assumption is the significant dissociation that occurred between the outcomes of studies employing
the rating likeability orienting task and the others that made used of different encoding conditions.
We have already referred to rating likeability as an encoding condition characterized by ambiguity
concerning the demands it makes on the participants. In the case where it draws upon conceptual
processing of stimuli it might be that it restrains AD subjects from priming normally, at least more
than another orienting condition that would allow the patients to follow a less profound
manipulation of targets. We should mention, however, that we were not provided with statistical
findings regarding the direct comparison between rating likeability and other orienting tasks within
the same group of patients. Our opinion is that a meta-analysis of such direct comparisons among
studies would contribute in illuminating this matter.
Accordingly, direct comparisons between the same sample of AD patients and two groups of
normals ( one consisting of exact age-matching controls and another one with a lower mean age
than the patients ) should be useful. In other words, they would help us reassure whether the greater
mean effect size for studies with more than 3 years age difference, reflects a tendency for greater
dissociation between performances.
A last issue that we wish to refer to is the combination of studies involving different stages of
dementia. In the majority of studies the experimenters reported the mean scores of DRS and
MMSE tests the patients had taken, without reporting the range of the scores. Therefore, we were
not sure of the different stages that the AD participants could be categorized as being at. The rest
of the studies that were used in the combination had to be categorized as involving patients from
either the first two stages ( mild and moderate ) or from all three stages ( including severe ). We
suggest that it would be really interesting to compare separately AD patients from all three
different stages of dementia, and investigate if there is any correlation between the severity of the
disease and the difference in performances of AD and control subjects.

42
GENERAL DISCUSSION
Before presenting our experiment we wish to refer to certain issues regarding stem-completion and
the priming exhibited by AD patients in this particular memory test.
Explicit memory of controls
Some experimenters questioned the possibility that normals might have used explicit memory
processes when they completed the stems, and thus, contaminated the results by causing the AD
performance to be considered impaired when compared to the “up-lifted” performance of controls.
This is a variable that can be really difficult to control experimentally, mainly because subjects
could use either intentional or unintentional explicit memory ( according to Schachter, 1987 ). By
this term ( or voluntary/involuntary ), Schachter describes the phenomenon during which normal
subjects either fill in stems that they consciously remember to be linked to words they have studied
earlier ( especially if they had followed a deep elaboration on the material ) or seem to make use
of unintentional explicit strategies by subconscious use of familiar material when completing the
test, avoiding, thus, to provide the first word that comes to their mind.
Burke et al ( 1994 ) introduced an alternative explanation of Randolph’s ( 1991 ) conclusion that
in AD, explicit, implicit and semantic memory decline in parallel. According to Burke ( 1994 ),
Randolph’s ( 1991 ) conclusion was depending on findings which led him to believe that certain
memory tasks were linked with particular memory processes. Those findings resulted from the fact
that the tests of stem-completion and cued recall were significantly correlated between them for
all subjects. Randolph ( 1991 ) also showed that the significant difference between the AD subjects
and their matched controls in stem-completion was eliminated when the cued recall scores were
used as a covariate. However, as Burke et al ( 1994 ) argued, other investigators showed that
implicit memory tasks were often explicitly processed. Burke et al ( 1994 ) went on by giving an
alternative explanation to Randolph’s ( 1991 ) conclusion. They suggested that Randolph’s ( 1991
) idea to use the cued recall scores of the subjects as a covariate, may have removed from the stem-
completion scores the contamination from the use of explicit memory. As a consequence,
Randolph's ( 1991 ) findings could be the result of the overlap in the processes used by all subjects

43
when performing in an explicit ( cued recall ) and an implicit ( stem-completion ) memory test.
Burke et al ( 1994 ) in their study, however, failed to detect a significant correlation between stem-
completion and explicit memory scores. In the Discussion part they referred to the experiment by
Partridge et al ( 1990 ) where intact priming was reported for the AD group; they claimed that
in that study performance of controls on stem-completion was almost equal to that shown by AD
patients and much poorer than that shown by controls of other studies reporting impaired priming.
Burke et al ( 1994 ) concluded that in Partridge’s ( 1990 ) study controls made almost no use of
explicit memory processes, and that could be a contributing factor to discrepancies among reports.
Huberman et al ( 1994 ) reported intact priming for the AD subjects they tested. They suggested
that data resulted from the fact that they managed to "disguise" stem-completion by keeping a
delay of 5 to 7 minutes between study and test-phase. In the meantime the subjects were engaged
in providing names and places to irrelevant to the studied items trigrams, failing, hence, to detect
any relation between the studied words and the test-phase. According to the investigators, controls
were prevented from using explicit memory strategies to complete the stems, in contradiction to
other experimenters who proceeded to test-phase immediately after study-phase and reported
impaired AD priming. Huberman et al supported ( 1994 ) the idea that when explicit memory
contribution was removed statistically or when proper experimental manipulations were used for
that purpose, intact AD performance could be demonstrated.
Randolph et al ( 1995 ) included either one or four exposures of the target-stimuli at study-phase;
they supported the idea that stem completion measured something more than pure implicit
memory, so they matched AD patients’ explicit memory with that of normal subjects by presenting
the words either once or four times during study-phase. The data supported their view.
Associations with other tests
In some of the experiments there were associations or correlations observed between stem-
completion and other tests administered to the same AD patients.
Randolph ( 1991 ) tested AD and HD patients in stem-completion and cued-recall and reported
impaired priming in both tests for the AD group whereas the HD group was impaired only in the
explicit test. He examined the relationship between the two tests for all groups and found a

44
significant correlation. After running an ANCOVA on stem-completion performance covarying
out the performance on explicit testing, he found that AD patients were not differentially impaired
in lexical priming compared to controls. In other words, he suggested that, in opposition to HD
patients, impaired lexical priming that AD patients exhibited stemmed from a general impairment
affecting explicit and implicit memory in a proportionate manner. Additionally, when Randolph
( 1991 ) tested the AD group in a free-association task, he found a strong correlation between poor
performance in that task and their episodic memory, in accordance with the correlation he
established between cued-recall and stem-completion. Those results prompted the experimenter to
conclude that explicit, implicit and semantic memory decline in parallel in AD.
Keane et al ( 1991 ) tested a group of AD patients in a stem-completion test, a perceptual
identification test and a category-fluency task. Patients showed impaired priming in stem-
completion which was correlated with their performance on the category-fluency task, whereas a
similar correlation between intact perceptual-identification priming and category-fluency
performance was not established.
Burke et al ( 1994 ) failed to demonstrate significant correlation between impaired performance of
patients in stem-completion and their scores at explicit memory tests ( measured by free-recall and
recognition ). They did, however, reveal a strong correlation between word-completion and cued-
recall, after reanalysing the data from Partridge’s study ( 1990 ). That correlation, although
significant for the AD group, was non-significant for the control group.
Carlesimo et al ( 1995 ) also found that impaired stem-completion priming in the AD group was
completely accounted for by performance on the graphemic-cued recall task.
Therefore, there were some indications that impaired stem-completion performance was linked to
explicit memory and word fluency. We should make the observation, though, that similar
correlations between different memory tasks were not investigated in intact-reporting studies.
AD and the transfer-appropriate processing
The first reports that examined AD patients in stem-completion and reported impaired priming,
attributed the results to a disruption within the semantic network. According to those researchers
( Shimamura, Salmon, Squire and Butters, 1987 etc ) that disruption led to inability by AD subjects

45
to semantically associate trigrams with the stimuli presented to them during study-phase.
According to several experimenters ( Shimamura, Salmon, Squires and Butters, 1987; Salmon,
Shimamura, Butters and Smith, 1988, etc ) the patients could not achieve successful access to
semantic information, resulting, thus, in poor performance.
Other investigators ( Grosse, Wilson and Fox, 1990; Partridge et al, 1990, etc ) argued that the
encoding process used was not appropriate in order for patients to fully understand the meaning of
words. Therefore they demonstrated intact priming by using “deeper” orienting tasks which
enhanced conceptual processing of stimuli. However, there were experimenters ( Russo and
Spinnler, 1994; Downes et al, 1996, etc ) who compared different orienting tasks according to the
level of processing that each one required, and were led to contradictory results as well.
The greater debate was developed about whether stem-completion is a memory task that makes
conceptual or perceptual demands and to what degree. This debate stems from the transfer-
appropriate processing theoretical model according to which, different memory tests of explicit
and implicit memory, require analogous encoding processes during the study-phase, in order for
the performance of subjects to be facilitated in test-phase. In accordance, when a test asks from
subjects to be occupied with the perceptual characteristics of the words
later to be tested on, encoding process at study-phase should be based upon the perceptual analysis
of the material used. It has been established by many findings from research on normal subjects
( Bowers and Shachter, 1990; Graf and Ryan, 1990; Roediger et al, 1992; etc ) that in explicit
memory tests participants make use of conceptual, semantic encoding of words in order to retrieve
them, whereas implicit tests require more of a perceptual analysis of stimuli. Explicit memory
performances are usually facilitated by conceptual analysis of the material, also known as
conceptually-driven memory process, whereas most common implicit tests usually depending on
perceptual encoding of the stimuli are known as depending on data-driven memory process.
We should note, however, that it is possible for explicit tests to be based upon data-driven memory
processing and for implicit memory tests to draw upon conceptually-driven processes.
However, not all of implicit or explicit tests share the exact same amount of conceptual or
perceptual involvement. Consequently, there were investigators ( Bondi and Kaszniak,
1991; Keane et al, 1991 etc ) who argued that stem-completion is mainly based upon conceptual
memory processes, found to be defective in AD patients. Others ( Gabrieli et al, 1994 etc ) claimed
that stem-completion would involve a strong perceptual component if subjects were instructed

46
during the study-phase to follow an encoding condition ( e.g. reading aloud ) that would allow
them to focus on the superficial characteristics of the material.
According to several experimenters ( Keane et al, 1991; Gabrieli et al, 1994; Monti et al, 1994,
Keane et al, 1995, etc ) memory tests can be divided into two categories. The first one includes
memory tasks focused on the perceptual analysis of material and appeal to occipital cerebral areas
that have been found to be relatively spared in AD. Tests that belong to this category are more
sensitive to changes in the format or the modality of the material from study to test-phase. The
other category consists of memory tasks characterised by conceptual elaboration of the stimuli, a
procedure localised to temporoparietal regions which are seriously damaged in AD. These tests
are thought to be quite resistant to modality or format changes from study to test-phase.
Although perceptual identification is considered to be one of the most representative implicit tests
of the first category, the nature of stem-completion is still the subject of great debate.
Semantic memory in AD
In a review article ( 1989 ) Nebes raised two major questions regarding semantic deterioration in
AD. The first one examined whether the semantic deficit observed in AD is due to a difficulty in
accessing stored information or whether it is due to an actual loss of semantic information. As it
was noted earlier, there were investigators who attributed impaired performance of AD subjects in
stem-completion to either of the two proposed explanations. The second question rises as to which
degree are several memory-theoretical units capable of producing satisfactory explanations about
the variability of difficulties AD patients confront when performing on different memory tasks.
At first, Nebes ( 1989 ) refers to the loss of word-finding ability that AD patients exhibit during
the early stages of the disease. This is evident, according to several experimenters ( Nicholas,
Obler, Albert and Helm-Estabrooks, 1985 ) from the tendency to use indefinite terms by giving a
descriptive meaning of words, rather than the actual concept. AD patients also provide relevant
but incorrect descriptions and seem to make use of a gradually limited vocabulary ( Bayles, 1982
). However, Nebes ( 1989 ) notices that there was adequate evidence ( Bayles, 1982; Hier et al,
1985; Nicholas, Obler, Albert and Helm-Estabrooks, 1985 ) for relevant preservation of the
semantic ability to form phrases. Another aspect of word-finding impairment in AD, is illustrated

47
when patients are tested on verbal fluency. As the disease progresses they demonstrate gradual
difficulty in generating words from a given category, which could be either a letter category
( subjects are required to provide as many words as possible beginning with a certain letter ) or a
semantic category. Nebes ( 1989 ) pointed out a study by Ober, Dronkers, Koss, Delis and
Friedland ( 1986 ) as “...the most comprehensive examination of verbal fluency”. The
experimenters showed that even mildly demented patients provided half of words generated by
controls. In addition, they came up with a significant observation: AD patients did not produce any
less low-frequency items compared to normal subjects. This means that they did not seem to have
greater difficulty in accessing low-dominant words of a semantic category, since they tended to
provide words which gathered more or less typical semantic characteristics, similarly to controls.
The authors came to the conclusion that the fluency deficit of patients is not due to general
slowness in finding words, but is rather the outcome of following a more random way of searching
than controls.
AD patients were also found to demonstrate reduced performance in object- naming. In Nebes’ (
1988 ) article, it is underlined that this deficiency is strongly correlated with the progress of the
disease. According to his review the impairment is suggested to stem from three different sources.
The first and less acceptable one is, that patients misperceive the perceptual characteristics of
objects, and are, thus, led to misname them ( Rochford, 1971; Barker and Lawson, 1968 ). The
second one, which seems to account for the fluency deficit as well, is that patients are impaired in
the ability to match the semantic characteristics of objects with their names, and thus, have lost the
semantic information required ( Bayles and Tomoeda, 1983; Flicker et al, 1987; Huff et al, 1986;
Martin and Fedio, 1983 ). The third explanation suggests that semantic representations of the
objects are intact, but there is a disruption in the connection between semantic representations and
lexical ones. One really interesting line of evidence for the third explanation comes from results (
Barker and Lawson, 1968; Kirshner et al, 1984; Skelton-Robinson and Jones, 1984 ) showing that
the less common the name of an object is, the less possible it is for the patient to provide it. This
could mean that patients have a problem with accessing low-frequency words when they are shown
pictures of these objects.
A final area of word-finding problems, according to Nebes ( 1988 ), is naming to definitions.
Subjects are required to provide words in response to a given definition. In a study by Rissenberg
and Glanzer ( 1987 ), normals tended to perform better when the target words were concrete rather

48
than abstract. The AD subjects were much worse in providing words than controls, but were also
much more affected by the concreteness or abstractness of words; they faced serious difficulty in
naming abstract words compared to concrete ones.
Nebes ( 1988 ), in his review, also refers to the fact that AD patients retain knowledge of concept
meaning. More specifically, he maintains that their vocabulary follows a steady decline as the
disease progresses. In addition, he cites evidence from studies ( Warrington, 1975; Martin and
Fedio, 1983 ) suggesting that AD patients preserve some general information about given concepts,
but have difficulty in retaining the specific attributes of those concepts. Nebes ( 1988 ) also
provides evidence from a study ( Nebes and Brady, 1988 ) showing that even when AD subjects
retain knowledge about the semantic attributes, they get confused when they are required to access
this information by self-directed search. Patients demonstrate similar difficulties in searching a
semantic category for its members, although they are able to access general information from the
semantic category a given concept belongs to. He concludes that it is harder for patients to find
specific information about a concept when they are asked, rather than to generate the first word
that comes to mind.
Nebes ( 1988 ) also refers to the amount of semantic priming that AD patients are able to
demonstrate. At first he attributes semantic priming to activation of the given concept nodes; when
a concept is presented, there follows activation of relevant nodes that begins to spread, according
to the network theory of semantic memory ( Collins and Loftus, 1975 ). Semantic priming can
be illustrated by various indirect tests and AD patients are capable of exhibiting semantic priming
as well. Nebes ( 1988 ) suggests that since this kind of priming is due to automatic activation
of the stimulus nodes, AD patients are facilitated by tasks that enhance this automaticity rather
than more difficult tasks, like verbal fluency, that demand a greater magnitude of attention. Nebes
( 1988 ) supports this view by providing results of a study of his where AD subjects were engaged
in word-naming ( Nebes, Martin and Horn,1984 ) and managed to exhibit intact priming compared
to the normals although they were slower than them. Nebes et al ( 1984 ) interpreted their results
as indicative of the semantic priming AD patients are able to show when the memory task does
not make heavy demands from the patients in terms of attentional capacity. Following, we are
going to examine whether the nature of stem-completion as a memory test could be categorized or
not as a memory task that could enhance the automaticity of word-searching during the testing

49
procedure or constitutes a testing procedure that is considered highly difficult to facilitate the
automaticity mentioned.
Nature of stem-completion
In some parts of Nebes’ ( 1988 ) article, as well as conclusions at the end, the author, referring to
the review of the relevant reports, suggests that AD patients benefit from tasks demanding that
they provide answers through an “automatic” procedure. By contrast, when patients are required
to draw upon intentional memory search by analysing specific semantic information, they seem to
face great difficulties.
At this point, however, we would like to examine two issues.
Nebes’ ( 1988 ) review repeatedly referred to the nature of semantic deterioration that AD patients
exhibit. Since failure of these patients to demonstrate intact priming in several studies was
attributed to semantic deficiency, we consider rather essential that we should be familiar with the
way patients function and disease affects their semantic memory. The two issues we tend to refer
to are the following: How much of a semantic test is stem-completion and to what extent does it
make use of “automatic” memory search that Nebes ( 1988 ) mentioned. Stem-completion is an
indirect memory test that in the past was considered by some investigators as mainly a perceptual
task and by others as mainly conceptual. The most acceptable hypothesis, currently, is that it
includes both a strong perceptual and a strong conceptual component. The encoding of words
exposed during the study-phase could be achieved by following either “deep” or “shallow”
processing by the subjects. But could the test-phase be performed in a way that would enhance the
perceptual component of the stem-completion task?
During the test-phase of a stem-completion test, subjects are presented with trigrams and are
expected to show a tendency to complete them with studied words which are thought to have
enhanced entry due to prior exposure. A word is a stimulus that follows a series of representations
in human memory. At first the subject perceives an optical representation of the word, which
allows them to proceed in a perceptual analysis of its features. Almost simultaneously there is a
phonological representation of the word by the act of reading. Retention of the word is facilitated
when the word is read with a certain amount of attention. After the word is perceptually processed,

50
the subject is ready to initiate a search in semantic memory, aiming to match the perceptual
characteristics of the word with the relevant semantic information stored. The lexical and
phonological representation of the word plays the role of an index useful to help in semantic
searching. Before approaching specific information, the word firstly comes through a feeling of
familiarity ( if it was ever encountered ) and then is enhanced with the general attributes of its
category. Consequently, it can be designated as a well established semantic unit or remain with the
sense of familiarity or the subject can only provide general information of some kind. The AD
subjects are more likely to fail in attempting to terminate the procedure in full success.
During the test-phase of stem-completion the above procedure is followed. Trigrams serve as an
index for initiating a search in the lexical representation system. Once the potential correct answers
are detected, an item is then selected, and in cases where priming takes place, the selection is of a
word that has been recently enhanced by its prior exposure at study-phase. The enhancement of a
target word is additionally achieved by its specific phonological form. We should note, however,
that in some languages phonological factor is not as significant for lexical priming as it is in other
laguages. In Greek, for example, words, with very few exceptions, are pronounced exactly the
way they are written, whereas in English there are a number of irregular words pronounced
differently to the way they would if they were regular. For example, cohort priming effect of the
word BLAME ( Downes et al, 1996 ) would be illustrated by the alternative completions to the
stem with the words BLATENT or BLAME, rather than BLANK or BLACK. In Greek, though,
cohort priming would be difficult to illustrate, since almost all words are read the way they are
written. Nevertheless, we believe that pronouncing words usually enhances their retention.
At this point we would like to underline the fact that a word is being picked-out because it is there.
This means that if the word did not exist as an established lexical representation with full
registration of its meaning in subject' s memory, it is unlikely that it would serve as a produced
answer in stem-completion. The trigram matches the particular response because it has already
been processed semantically; and here lies the conceptual component of the test. In a word-
identification test, subjects are not required to produce words. They are only asked to read briefly-
presented words, without having to understand their meaning. This is obvious from the ability to
demonstrate priming even when the stimuli are non-words ( Bentin and Moscovitch, 1988; Salasoo
et al, 1985; Rueckl, 1990, etc ). There has been priming observed for non-words not only for

51
normal but also for patient groups ( Cermak et al, 1985; Gabrieli and Keane, 1988; Haist et al,
1991; Musen and Squire, 1991, etc ).
In a stem-completion test, subjects have to give the first word that comes to mind, so it is
understandable that they should choose an answer they possess, especially since testing in most
studies requires the responses in self-paced time, enhancing intentional, nonautomatic self-
searching.
And here lies the question about whether stem-completion is a test that draws upon memory
searching which facilitates AD performance by providing “automaticity” or not. We wish to
explore certain aspects of the test towards this direction, that could be designed in order for patients
to enhance their priming scores.
The first suggestion we make is that there should be a careful selection of the words that serve as
target-stimuli. In Nebes’ ( 1988 ) review, it was noted that AD patients are dramatically impaired
when asked to provide the meaning of abstract words ( Rissenberg and Glanzer, 1987 ). Since
they exhibit such a great difficulty, it could be that some part of the failure they demonstrated in
certain studies was due to the proportion of abstract words that each different study-list included.
Nebes ( 1988 ) also referred to the difficulty that AD patients show in naming the names of objects
especially when they are not common . He did, however, mention that patients do not provide
responses differently in terms of frequency, compared to the controls, in a category-fluency task (
Ober et al, 1986 ). We wish to add, though, the fact that the frequency of words has not been
examined thoroughly as a variable that could play a role in the formation of AD priming results.
As was noted earlier in the review, there were only three studies where manipulation of word-
frequency took place. In the study by Partridge et al ( 1990 ), the orienting task required from
subjects to produce the meaning of each word. To make it easier for the patients, the experimenters
used only high-frequency words and came up with intact priming. A question lies to whether the
magnitude of priming was not due to the encoding process followed, but to the material, which
became more accessible to patients because of its frequency.
In the study by Landrum and Radtke ( 1990 ) both high and low-frequency words were used. Intact
priming was found for mild AD patients and impaired for moderate AD patients. Moderate patients
showed less priming than the other two groups, especially for low-frequency words. This could be
indicative of the fact that since the progress of disease is followed by gradual reduction of

52
vocabulary, patients of early stages were more capable of accessing more words and words of
lower frequency, than patients of later stages.
And what about the “automaticity” of stem-completion? When subjects are required to produce
the first word that comes to mind, in most of the studies there was no time-limit. This provides the
subjects with a greater chance to make use of intentional or unintentional explicit memory when
they fill in the trigrams and, hence, contaminate the results. AD patients, on the other hand, are
tested in a far less “automatic” way compared to perceptual-identification, where time-limit forces
patients to automatically repeat words shown to them, without having to engage in any
conceptual processing. In stem-completion and under the conditions where the test usually is being
carried out, patients draw upon intentional self-searching in semantic memory, as they try to match
the trigrams with words they fully possess. If the test-phase was conducted under time-pressure, it
could be possible that patients would be more able to provide words they had earlier been exposed
to as responses to stems. Moreover, even if patients are dealing with target words not easy to access
because of loss of their semantic attributes, they would not have the time or anxiety to make sure
they possess specific, semantic characteristics of each word.
If the test-phase of stem-completion were designed so as to be conducted in a more perceptual
way, there should be an enhancement of the perceptual nature of study-phase as well. Words
should be presented as quickly as possible for subjects to read, without having to make any
judgement, or without having to draw upon conceptual elaboration on the words. By this way
normals should be restrained from semantically processing words, a process which could lead to
enhancement of explicit strategies used in the completion of stems. Consequently, we would
instruct the subjects to read each word as fast as they could and immediately proceed to the
following one.
Regarding the test-phase, our initial suggestion was to introduce a stem-completion test-phase
where subjects should produce words to stems as fast as possible. Accordingly, the stem-
completion test procedure would follow a pattern of design similar to that usually followed during
the test-phase of a word-identification task. In a word-identification test-phase, the subjects have
to recognize words briefly presented to them ( e.g. a few ms ). In our stem-completion test-phase,
the subjects would have to produce words to trigrams briefly presented to them ( e.g. a few secs ).

53
Subsequently, a time period of 5 secs was selected, as a time limitation that not only would create
the time-pressure that could affect the performance of the participants but also would be enough
in order for the responses to be provided.
Another suggestion is that in order to define the exact time-limitation during the filling of each
stem, an individual pre-testing of all participants could be conducted, on their ability to provide
words to stems as briefly as possible. Consequently, the final time-limits would be determined
from those data.
EXPERIMENT
The main idea and purpose of this experiment was to design a stem-completion testing procedure
that should be capable of facilitating the AD subjects. We attempted to include certain factors that
were indicated by other researchers as to have a positive effect on the priming abilities of AD
patients.
In other words, we aimed to conduct an experiment under such conditions that could enhance the
performance of patients.

54
At this point we would like to return to the stem-completion Review Table and refer to the columns
of time-limits during study-phase. From the eleven studies reporting impaired priming, only two (
Keane and Gabrieli, 1991; Carlesimo, Fadda, Marfia and Caltagirone, 1995 ) made use of time-
limitation during study-phase.
More specifically, in Keane and Gabrieli’s ( 1991 ) study, presentation of each word took place
every 4 secs, whereas in Carlesimo, Fadda, Marfia and Caltagirone’s ( 1995 ) experiment,
exposure of each word was lasting for 5 secs. There was also a report by Fleischman and Gabrieli
( 1999 ) which included a time-limit of 30 secs but because of its long duration it could not affect
the encoding procedure in any way.
From nine reports demonstrating intact priming for the AD group, five included time-limits at
study-phase ( Scott et al, 1991; Russo and Spinler, 1994; Huberman and Moscovitch, 1994;
Koivisto et al, 1998 ). Four of them used time-limits of 5 secs, 5 secs, 3 secs and 4 secs respectively.
There was also a study ( Fleischman, Gabrieli and Rinaldi, 1997 ) that made use of a 30 secs time-
limit.
From the mixed-results group of reports, there were 3 out of 6 studies that used time-limits during
study-phase. More specifically, Landrum and Radtke ( 1990 ) used a 5 secs exposure for each word
and found intact priming only for mild patients. Perani et al ( 1993 ) who came up with intact
priming for half of the patients, made use of a 2 secs auditory presentation of stimuli. Finally, in
the first experiment of the study by Downes et al ( 1996 ), where rating likeability was used, there
was also time-limitation of 5 secs; impaired priming was demonstrated. In the second experiment
there was no time-limit for the reading aloud encoding process, but there was a time-limit of 4 secs
for the rating likeability task. Priming was found to be intact following the first orienting task and
impaired following the second.
Looking at the Review Table we realize that time-limitation during test-phase was rarely used.
More accurately, none of the studies reporting impaired priming made use of such a time-limit.
From the intact-reporting studies, there was one ( Russo and Spinnler, 1994 ) including a 5 secs
limit. From the mixed-results group, Landrum and Radtke ( 1990 ) used a 30 secs time-limit which
could not have any effect on the procedure due to its extended duration. Finally, in the second
experiment of the study by Downes et al ( 1996 ) that reported intact priming only for the reading
aloud condition, there was a test-limitation of 10 secs.

55
In this experiment, we are going to include time-limits of different durations in test-phases. Our
purpose is to discover whether pressure of time has any effect, not only on the way subjects encode
the stimuli at study-phase, but also on the cognitive procedure they follow during test-phase.
At study-phase, we are aiming to verify whether there is any truth to the suggestion that once there
is limited time, encoding process of the material proceeds in a more “perceptual” way. In other
words, whether subjects avoid deeper elaboration of the material - which is considered to be a
conceptual process- because the given time is adequate only for the encoding of perceptual
characteristics of words. For this purpose, we are going to employ the orienting task of reading
aloud as fast as possible -in order for the participants to also benefit, besides the visual stimulus,
from the phonetic trace of each word-
As far as the control group is concerned, time pressure is aiming to limit the entrance of explicit
memory to completions of stems. Lack of ample time to elaborate on the material during study-
phase, in combination with time pressure during test-phase, could result in preventing explicit
strategies from intruding.
For AD patients, on the other hand, we are going to investigate whether the whole procedure could
be beneficial for their performance, by limiting the conceptual component. During study-phase,
they are going to be asked to read each word as quickly as possible and immediately proceed to
the next one. Inspite of their extended neuropathology, AD patients seem to preserve the ability to
read without facing substantial difficulties, at least before the ultimate stages of the disorder. Our
purpose is for them to focus on the perceptual characteristics of words, which are the visual and
phonetic-auditory forms. We have referred to the findings underlining severe alteration of areas
related to conceptual functions, and to others describing as “relatively spared” cerebral areas
responsible for perceptual functioning. As far as the test-phase is concerned, we should examine
whether AD subjects, are going to handle stem-completion as a task that, at some point, resembles
word-identification. Instead of employing themselves to intensive self-searching of semantic
memory they could perform in a more “automatic” way. More specifically, under time-pressure,
they could be able to provide words, familiar from prior exposure, without having to enter a
procedure of recognizing each and every one of the responses as well-established semantic units.
Consistent with the idea of “automaticity” is the use of oral completion of stems compared to the
written completions. In the Review Table we can see that from the eleven studies reporting
impaired priming, five made use of written completion, whereas only two out of nine studies

56
reporting intact priming made use of written completion. The group of mixed-results exclusively
employed oral completion of stems. We suggest that oral production could provide a kind of time-
pressure on its own compared to the written, because it is less time-consuming and more
spontaneous, in contrast with written production of words that is part of a more sophisticated,
conceptual procedure. Accordingly, we presume that AD subjects who participate in a stem-
completion test, regard it a more “relaxing” and less “intimidating” task, once their responses are
required orally and not in a written way.
Another variable we were interested in, was the concreteness or abstractness of target words and
whether it could affect the performance of the subjects. As we mentioned, in a study by Rissenberg
and Glanzer ( 1987 ), AD subjects were dramatically worse at providing abstract words to given
definitions compared with concrete words. Normal young and old participants also faced greater
difficulty with the production of abstract words, but not to the extent the AD patients did. Since
there has been no investigation of whether concreteness of studied words could contribute as a
factor to the creation of a material-specific phenomenon in stem-completion, we decided to include
two categories of word-stimuli. The first one consisted of high-concreteness words and the second
one of high-abstractness words. In this way, we aimed to investigate whether the nature of the
material used so far, has played any role to the confliction between the reported results.
All subjects were also tested in a cued recall test, that consisted of two conditions designed
similarly to stem-completion. In other words, there were two cued recall tasks where the test-phase
took place at first under no time pressure and subsequently under time pressure. Our purpose was
to examine the explicit memory performance of the same groups that were tested for their priming
ability, in a task that assimilates the corresponding implicit memory task in terms of the procedure
used. Accordingly, a direct comparison between the explicit and implicit memory of the same
subjects could be possible. Furthermore, we could investigate whether time pressure has any effect
on an explicit memory test, in direct comparison to an implicit memory task, and among the three
groups. Finally, the selection of the items used in the study-phase followed the same pattern used
in stem-completion, in terms of the concreteness and abstractness of the words. Thus, we could be
able to examine whether there was any effect on the results, coming from the word-type of the
materials.

57
Memory in healthy aging
In this experiment we decided to include also a sample of young subjects and have it compared to
the total sample of older subjects. The purpose was to investigate whether aging constitutes a
decreasing factor for memory and also whether it affects explicit and implicit memory similarly.
Baddeley, in his book " Essentials of human memory " ( 1999 ), dedicates a chapter to " memory
and aging ". He refers, separately, to different kinds of explicit memory and evidence coming from
relevant research. Aging has been found to constitute a decreasing factor to the working memory
- by this term is implied the memory responsible for the retention and at the same time
manipulation of information - of older people. According to Baddeley ( 1999 ), there are findings
since the 1960s suggesting that aging affects several aspects of working memory, such as linguistic
( Salthouse, 1992 ), visuo-spatial ( Rabbitt, 1989 ) and language production ( Kemper,
1990 ). In addition, older subjects have been found to perform poorly compared to younger ones
in a long-term memory study by Baddeley, Emslie, & Nimmo-Smith, ( 1994 ). Prospective
memory has also been proved to be affected by aging ( Cockburn and Smith, 1991; Craik, 1992 ).
Finally, a study by Bowels and Poon ( 1985 ), illustrated the fact that semantic memory is
influenced by aging in a way that is translated into gradual failure to access stored information, in
addition to greater inability to prohibit irrelevant information from appearing.
Thus, according to Baddeley ( 1999 ) older people have been found to perform poorly in several
aspects of explicit memory. But what about implicit memory?
A Fleischman and Gabrieli article ( 1998 ) by the title of " Repetition priming in normal aging and
Alzheimer's disease : a review of findings and theories ", includes an extensive literature review
of research conducted on priming tasks performed by normal aging subjects and AD patients.
Although some variability was observed among the results, it seems that implicit memory in older
people remains rather unaffected by aging in opposition to explicit memory. Fleischman and
Gabrieli ( 1998 ) described three groups of different kinds of implicit memory tests, according
to the demands that each group of tests made on subjects. The group referred to as containing
perceptual priming tasks, included the following tests:
Word-nonword identification task ( Abbenhuis et al, 1990, Exp. 2; Light et al, 1992, Exps 1 & 2;
Shachter, Church and Osowiecki, 1994, Exps 3-5 ), lexical decision ( Karayanidis, Andrew,
Ward & McConaghy, 1993; Ober, Shenaut, Jagust & Stillman, 1991; Rabbit, 1984, Exp 1; Balota

58
and Ferraro, 1996; Nillson et al, 1989 ), object decision ( Schachter, Cooper and Valddiserri, 1992,
Exps 1 & 2 ), naming task for pictures ( Mitchell, 1989; Mitchell, Brown & Murphy, 1990;
Sullivan, Faust & Balota, 1995 ) and words-nonwords ( Light and Prull, 1995, Exps 1 & 2; Ober
et al, 1991, Exps 1, 3 & 5; Wiggs and Martin, 1994, Exps 1 & 2; Hashtrudi, Chrosniak & Scwartz,
1991, Exps 1 & 1a; Light et al, 1996, Exp. 1, Light et al, 1995, Exp. 2; Wiggs and Martin, 1994 ),
degraded naming for words ( Hashtrudi et al, 1991 , Exps 2 & 2a; Prull, Light, Collet & Kennison,
1997, Exp. 2; Light and Singh, 1987, Exp. 3 ) and pictures ( Russo and Parkin, 1993, Exp. 1;
Ergis, Van der Linden & Deweer, 1995 ), word-fragment completion ( Light et al, 1986; Rybash,
1994; Winocur et al,1996; ) and anagram solution ( Java, 1992 ). The majority of the findings
illustrated age invariance.
The group of conceptual priming tasks included the following: word association ( Mc Evoy,
Holley and Nelson, 1995; Java, 1996; Grober, Gitlin, Bang and Buschke, 1992 ), category
exemplar generation ( Isingrini, Vazou & Leroy, 1995; Light, Prull & Kennison, 1997; Monti et
al, 1996; Light and Albertson, 1989; Maki and Knopman, 1996 ), and fact completion ( Small et
al, 1995; Rastle and Burke, 1996 ). In most of these studies older people were found to prime
normally, similarly to the younger participants.
Finally, the group of unclassified priming tasks ( rhyme exemplar generation, category exemplar
verification, homophone spelling, sentence puzzles, reading speed and text re-reading ) provided
mixed results.
Fleischman and Gabrieli included stem-completion in the group of perceptual priming tests. They
mentioned 21 experiments carried out with old participants, 15 of item priming and 4 of associative
priming.We are going to concentrate directly on the findings of the research concerning item
priming of stem-completion. The picture here is of greater variability compared to the rest of
priming tasks. Nine experiments failed to demonstrate a significant age effect in stem-completion,
although in most of them ( Light and Singh, 1987, exps 1 & 2; Dick, Kean and Sands, 1989, exp
1; Java and Gardiner, 1991, exps 1 & 2; Gibson et al, 1993, exps 2b & 5; Friedman, Snodgrass and
Ritter, 1994 ), except for two ( Nyberg, Backman, Erngrund, Olofsson and Nilsson, 1996; Park
and Shaw, 1992 ), the results favoured the performance of the younger participants. Six other
experiments managed to detect the existence of an age effect (Chiarello and Hoyer, 1988; Hultsch,
Masson and Small, 1991, accounting for 2% of the variance; Small, Hultsch and Masson, 1995,
accounting for 1.5% of the variance) in some cases under certain specifications: Davis et al ( 1990

59
) found an age effect in stem-completion, after the age of 70. Winocur et al ( 1996 ), demonstrated
age invariance between young and community-dwelling old subjects, but he reported limited
priming for institutionalized participants, when they were compared to young controls. Shachter
et al ( 1994 ), in conclusion, referred to reduced performance of old subjects, in voice-specific,
auditory stem-completion.
In a chapter by Light et al ( 2000 ) in Perfect, T.J. & Maylor, E.A. (Eds.) : " Models of Cognitive
Aging ", there is a review of literature about age differences in direct and indirect measures of
memory. Light et al ( 2000 ) refer to recollection and familiarity, according to contemporary
theories of memory, as the two processes subserving recall and recognition. The writers consider
the existing evidence as indicative of relatively preserved familiarity in aging, whereas memory
deficits in aging mainly stem from decreased recollection efficiency. Light et al ( 2000 ) suggest
that the existence of age-related deficits in direct measures of memory that require recollection,
cannot be doubted. In general, they suggest that adults over 60 are reported to be facing memory
problems ( Cutler and Grams, 1988; Dixon and Hultsch, 1983 ). These memory problems are
illustrated by the lower scores of older people in laboratory based tasks such as recall and
recognition ( Craik and Jennings, 1992; Light, 1991 ), in everyday life memory based tasks ( West
et al, 1992; Kirasic et al, 1996 ) and in standardized tests used in neuropsychological assessment
( Salthouse et al, 1996 ). Light et al ( 2000 ) report that aging and recollection tasks of episodic
memory are negatively correlated not only in adults over 50 but even in adults under 50 (
Verhaeghen and Salthouse, 1997 ), and also refer to the poorer memory for contextual information
aging adults have ( Kausler 1991; Light 1991, 1996 ). A few paradigms are offered from the writers
in order to illustrate the lack of contextual memory aging people are reported to show ( Hashtrudi
et al, 1990; Kliegl and Lindenberger, 1993, Mc Intyre and Craik,1987; Koriat et al, 1988; Brigham
and Pressley, 1988 ). At this point, Light et al ( 2000 ) suggest that a reasonable assumption could
be that memory for contextual information mostly contributes to recall, whereas the familiarity
component is deeply involved in recognition and in indirect measures of memory. They also cite
evidence from meta-analyses of direct ( Verhaeghen et al, 1993; La Voie and Light, 1994; Spencer
and Ranz, 1995; ) and indirect ( La Voie and Light, 1994 ) measures of memory and findings from
many other studies ( Cohen and Faulkner, 1989; Schacter et al, 1997; Koutstaal and Schacter,
1997; Norman and Schacter 1997 Tun et al, 1998; Bartlett et al, 1991; Dywan and Jacoby, 1990;

60
Jennings and Jacoby, 1993; Multhaup, 1995 ), supporting the hypothesis for relatively spared
familiarity and impaired recollection ability in aging.
Light et al ( 2000 ), also conducted a meta-analysis of 95 published studies that examined the
priming performance of older participants compared to younger participants. There was a mean
weighted effect size of 0.185 with a confidence interval ranging from 0.133 to 0.237, suggesting
that aging subjects show less priming than younger subjects. The investigators, however,
underlined the fact that, according to the findings of other meta-analyses ( La Voie and Light,
1994; Spencer and Raz, 1995; Verhaeghen et al, 1993 ) these aging differences in priming tasks
are of far less magnitude than the age differences in recall and recognition, as they were reported
in the above studies.
Finally, we are going to refer to the research conducted by Fleischman et al ( 1999 ) who tested 58
old healthy subjects, 25 young controls and 91 AD patients, and to some of the conclusions made
about the performance of old participants. The main finding was that advancing age caused a
reduction of priming in stem-completion. Furthermore, performance of the mildly demented AD
patients was similar to that of healthy old controls. Fleischman et al (1999 ) did not attribute this
age effect to explicit memory factors, but they discussed the potential explanation of attentional
resources. They suggested, that is, that reduced priming in aging people, could be due to limitations
of attention, especially in a test sensitive to study-phase attentional requirements such as those
made by stem-completion. Fleischman et al ( 1999 ) supported their view by using evidence from
the study of Winocur et al ( 1996 ), aiming to strengthen their suggestion that declining priming in
old participants could stem from individual differences in attentional capacity. In that study, as it
was mentioned earlier, an age effect was obvious for institutionalized old controls, but not for
community-dwelling ones when they were compared to young subjects. All participants were
tested in a word-fragment test besides stem-completion, a test that Fleischman et al ( 1999 ) felt it
was less sensitive to study-phase division of attention, in contrast with stem-completion. Their
suggestion was that the former test mostly depends on the identification of a pattern of letters
whereas the latter one draws upon the production of a response. Moreover, stem-completion scores
were found to correlate with those of an independent measure of complex attention, something
that did not happen with the word-fragment scores. In the stem-completion test of their study,
Fleischman et al ( 1999 ) observed that generation encoding process seemed to cause greater
reduction of priming to AD patients and old controls aged more than 75, than did the reading and

61
rating conditions. The researchers attributed this finding to the fact that the generation orienting
task was considered more semantic and less perceptual than the other two tasks, resulting, hence,
to less benefited priming.
According to the above references, aging affects memory performance although it seems to have
a greater impact ( in the way that causes greater reduction ) on explicit memory measures rather
than implicit memory measures. Therefore, we compared healthy young and older participants in
stem-completion ( an implicit memory task ) and cued-recall ( an explicit memory task ). Our
prediction was that an age effect would occur, although less extended in stem-completion than in
cued-recall. Additionally, we examined the degree to which older subjects are influenced or not
by time pressure and material characteristics ( concreteness-abstractness, word-fluency ), or
whether they differ in any way compared to younger controls.
Methodology
---Participants---
Three groups of participants took part in the experiment: A group of 20 young healthy control
subjects ( YCS ), a group of 20 old healthy control subjects ( OCS ) and a group of 8 patients
diagnosed as probable Alzheimer' s disease ( ADP ).
All 20 young subjects were recruited from first and second year students attending courses at the
University of Liverpool, whose ages ranged from 18 to 24 years. The OCSgroup was recruited
from the panel of volunteers maintained at the Department of Psychology, in the University of
Liverpool, and were paid a small honorarium to cover expenses. Their ages ranged from 62 to 70.
The AD patients were recruited from the Community Old Age Psychiatry Service on the Wirral,
Merseyside and their ages ranged from 59 to 88. All patients had probable Alzheimer’s disease
and in the view of the consultant psychiatrist ( Dr. Ferran ) and the community team, this was of
mild to moderate severity, such that reliable data could be collected from cognitive testing.
Diagnosis was based on clinical interview, in some cases CT scan which showed evidence of
atrophy, particularly affecting the temporal lobe regions, and some limited neuropsychological

62
testing. Neuropsychological testing included Folstein’s mini-mental state examination, the logical
memory subtest from the Wechsler Memory Scale – Revised which gives scores for immediate
recall and delayed recall of two prose passages, each scored out of 50. These tests give some
indication of the relative severity of dementia – all the patients would appear to fall in the mild-
moderate severity range. The table that follows describes the means and ranges of the patient
characteristics:
Table 3. Means and ranges of the patient characteristics
months since
diagnosis
MMSE logical memory I logical memory II category
fluency
17.37
( 8-25 )
17.5
( 15-21 )
7.81
( 5-10.5 )
1.56
( 0-4 )
6.62
( 3-12 )
For the comparison between the group of old controls and the DAT patients, a subset of 10 healthy
old participants from the OCS group was used, selected in terms of their scores on NART test, in
order to be matched to the corresponding scores of the patients on the same test. The following
table describes the means and standard deviations of the scores of the matched-OCS and ADP in
the NART test:
Table 4. Means and standard deviations of the matched-OCS and ADP scores in the NART test
GROUP N NART .
matched OCS 10 102.20(3.68)
ADP 8 101.38(8.18)
The age of the matched-OCS group ranged from 62 to 70.
The mean ages and standard deviations for each group are following:
Table 5. Mean ages and standard deviations of all groups

63
GROUP N AGE .
YCS 20 19.3 (1.658)
OCS 20 66.2 (2.441)
matched OCS 10 65.9 (2.283)
AD patients 8 77.875 (9.125)
---Design---
A mixed design was used, with a single between subjects factor ( group:either YCS-OCSor matced
OCS-ADP ), and either two ( cued recall ) or three ( stem-completion ) within-subjects factors.
For both the cued recall and stem-completion tests, two test versions were used and word type (
concrete, abstract ) was varied. An additional factor in the stem-completion tests was the status of
the stems used ( derived either from studied or unstudied words ). All subjects completed
all four memory tests in the same fixed order.
---Materials---
Experimental stimuli were selected from Paivio, Yuille and Madigan' s ( 1968 ) list of 925 nouns,
which provides normative ratings for concreteness on a 7-point scale. 192 words were selected and
were divided into two sets, designated "concrete" and "abstract", such that all words in the former
set had concreteness ratings above 4.5, and all those in the latter set had concreteness ratings up to
and including 3.5.
Two additional criteria used in the selection of words were: (a) that there should be at least ten
other words beginning with the first three letters; and (b) that each target word was not of the
highest frequency among the words that shared the same trigram.
From these two sets of words, eight lists of 24 words each were constructed, each containing equal
numbers of concrete and abstract words. Care was also taken to ensure that the lists were matched
in terms of mean word length and mean word frequency. The mean frequency of the words as
defined by Kucera and Francis' s ( 1967 ) ratings, was: 37.67, ranging from 1 to 275. Lists were
randomly assigned to form two groups of four, and for each subject, one group of lists was used
as targets for the word-stem completion and cued recall tests, with two lists of the remaining group
serving as distractors in the stem-completion tests. Across subjects, each group of lists served

64
equally often as targets and distractors. Word-stem completion and cued-recall tests were given in
a fixed order ( see Procedure ), and lists from the designated target group were rotated across
ordinal positions, giving eight possible list orderings in total. As far as possible, equal numbers of
participants from each subject group were given the eight list orderings.
Procedure
Stem-completion 1 ( no time pressure condition ): In the study-phase, at first subjects were
informed that their reading ability would be tested. Then they were presented with the 24 target
stimuli and were asked to read each word aloud and as fast as they could. As soon as a subject had
finished reading a word the next one was immediately exposed.
Once the study-phase was completed, subjects were required to name all the animals they could
remember in 2 mins. That was the first category fluency distraction task.
The test-phase then followed, in which subjects were presented with 48 three letter strings ( 24
derived from the target words and 24 from the distractor set ) and were instructed to say the first
word that came to mind beginning with those three letters. They were also informed that they
would be allowed up to 20 secs to generate a response for each trigram. The 20 secs were
considered as adequate time period for responses to be given without creating a sense of time
pressure to either the normal participants or the AD patients.
Stem-completion 2 ( time pressure condition ): The instructions for the study-phase of stem-
completion 2 were identical to those of stem-completion 1. The second category fluency distractor
task that was taking place between the study and test-phase of stem-completion 2, required from
the participants to name all the countries they could remember in 2 mins. In the test-phase, the
participants were also asked to provide the first word that came to mind in response to three letter
trigrams, but this time they were instructed that they would be allowed only 5 secs for each
response and this would generate the time pressure condition.
Cued-recall 1 ( no time pressure condition ): The study-phase was identical to that used in the
stem-completion tests, but in this case the subjects were explicitly instructed to try and remember
the words for a subsequent memory test. The test-phase, unlike the stem-completion tests, followed
immediately after the presentation of the target stimuli. The distractor tasks, were only used during

65
the stem-completion testing in order to prevent the controls from relating the two phases of each
condition and make use of explicit strategies in the completion of stems. Since the cued-recall
tasks are explicit memory tests and all participants were informed at the beginning of the true
nature of the tasks, there was no need for distractor tasks to be taken. Subjects were informed that
24 three-letter trigrams would appear on the computer monitor, and that each corresponded to the
first three letters of the words they had just seen. They were told to give a response only if they
could remember having read it earlier. The adequate no time pressure period of 20 secs was
allowed for each response.
Cued-recall 2 ( time pressure condition ): The study and test phases were identical to cued recall
1, the only difference being that only 5 secs was allowed for subjects to generate responses to each
of the trigrams and this was considered to be the time pressure condition. Following the four
versions of the two memory tests, all participants were tested in the NART.
RESULTS
Stem-completion
Data were entered in a 2( group )x 2( test: stem-completion 1; stem-completion 2 )x 2( word-type:
concrete; abstract )x 2( stem-type: studied; unstudied ) mixed ANOVA, where stem-completion 1
was the no time pressure condition. The same pattern was followed twice, the first time for the
comparison between the YCS and OCS groups, and the second time for the comparison between
the AD and matched OCS groups.
Young-old controls
There was a main effect of group, F(1,38)= 6.236 p<.017, showing that overall, the old subjects
generated fewer completions (.204) than did the young (.254). By the term "completions" , all the
words correctly answered as responses to the trigrams are implied, regardless if they were studied
or unstudied words. The main effect of test was non-significant, F(1,38)= 2.000 p=.165, showing
that, although overall, performance on stem-completion 1 which was the no time pressure

66
condition (.219) was numerically lower than that on stem-completion 2, the time pressure
condition (.239), the difference was unreliable. The main effect of word-type was significant,
F(1,38)= 14.716 p<.0005, indicating that, overall, concrete words were associated with higher
target completion rates (.255) than abstract words (.203). Finally, as expected, there was a main
effect of stem-type, F(1,38)= 220.994 p<.0005, confirming that target completions were more
likely for studied (.373) than for unstudied (.085) stems.
Of the interaction terms, only the group * stem-type: F (1,38)= 8.765 p = .005, interaction was
significant showing that, overall, the magnitude of priming for the OCS group ( studied items:
.319, unstudied items: .089 ) was lower than that observed for the YCS group ( studied items: .426,
unstudied items: .081 ). Importantly, the data show that this was not due to an overall reduction in
target completion rates since the two groups were matched for the unstudied items. A second,
related interaction approached significance, that involving group, test, and stem-type, F(1,38)=
3.334=.076. Interpretation of this suggests that the magnitude of priming exhibited by the YCS
and OCS groups was not equivalent across the two stem-completion tests. Thus, priming increased
from the first to the second stem-completion test for the YCS group ( stem-completion 1:
targets:.384, baselines: .082; stem-completion 2: targets: .464, baselines:.079 ), and decreased
slightly for the OCS group ( stem-completion 1: targets: .317, baselines: .090; stem-completion 2:
targets:.313, baselines:.091 ).
Table 6. Stem-completion priming for the young and the old controls
group overall
stem-completion priming
stem-completion 1
priming
stem-completion 2
priming
YCS .426 .384 .464
OCS .319 .317 .313
Of the remaining interaction terms the lack of significance for those involving group and word-
type factors showed that there was no evidence that the word-type factor differentially affected
stem-completion performance across the two groups: group * word-type: F(1,38)=.097 p=.757;
group * word-type * stem-type: F(1,38)=.392 p=.535. Nor was there evidence supporting the

67
conclusion that priming magnitude was reliably different for the concrete and abstract word sets,
F(1,38)= 1.850 p= .182.
The remaining interaction terms are listed below:
Group * test: F(1,38)= 1.918 p=.174, meaning that there was no difference in the way that both
groups performed on stem-completion 1 and stem-completion 2.
Test * word-type: F(1,38)=.496 p=.485, showing no interaction between the two types of tests and
the concreteness or abstractness of their material.
Test * stem-type: F(1,38)= 2.528=.120, revealing no interaction between the two versions of test
and the priming levels.
Group * test * word-type: F(1,38)=.582 p=.450, indicating that for both groups, the levels of
performance on stem-completion 1 and 2 was not differentially affected by the word-type of the
material they were tested on.
Word-type * stem-type: F(1,38)= 1.850 p=.182, showing no interaction between the levels of
priming and the word-type of the material.
Test * word-type * stem-type: F(1,38)=.126 p=.725, indicating that the priming levels throughout
both tests were not affected by the concreteness or abstractness of the material.
Group * test* word-type * stem-type: F(1,38)=.001 p=.971, showing that young and old controls
prime in a similar pattern on both stem-completion tests, regardless of the word-type of the
material.
Matched old controls-AD patients
For this comparison, the main effect of group was significant, F(1,16)= 15.324 p<.010, showing
that overall, the AD patients achieved fewer completions (.129) than did the old controls (.208).
Also, as anticipated, there was a significant main effect of stem-type, F(1,16)= 49.665 p <.000,
showing that target completions were more likely for studied (.252) than for unstudied ( .085 )
stems. Neither the test main effect, F(1,16)= .069 p=.796, nor the word-type main effects, F(1,16)=
2.557 p=.129, were reliable, showing that, overall, completion rates were equivalent for the two
tests ( stem-completion 1:.166; stem-completion 2:.171 ), and that the completion rates for

68
concrete words ( .185 ), although numerically higher than that for abstract words (.153 ), were not
so as to reach significance.
There was no evidence that the two groups differed to a significant degree in terms of priming
magnitude, although the group * stem-type interaction did approach the criterion value, F(1,16)=
3.778 p=.070. Thus, although the significant group main effect shows that completion rates were
lower overall for the AD group, the presence of this marginal interaction suggests that effect was
not equivalent for the studied and unstudied words. In other words, there was evidence that the
magnitude of priming was different for the two groups ( OCS-matched: studied:.315,
unstudied:.102 ; AD studied:.190, unstudied:.069 ).
Table 7. Stem-completion priming for the matched old controls and the AD patients
group overall
stem-completion priming
stem-completion 1
priming
stem-completion 2
priming
matched
OCS .315 .312 .318
ADP .190 .192 .188
As with young-old comparison, there was no evidence that the concreteness factor differentially
affected stem-completion performance across the two groups: word-type * group: F(1,16)= .292
p=.596; word-type * stem-type * group: F(1,16)= .811 p=.381.
The interaction terms were as follows:
Test * group: F(1,16)=.126 p=.728, indicating similarity in the way both groups performed in stem-
completion 1 and 2.
Test * word-type: F(1,16)= 1.094 p=.311, showing that the two types of stem-completion were in
no way affected by the word-type of their material.
Group * test * word-type: F(1,16)=.702 p=.414, revealing that the levels of performance of old
and AD subjects for both tests, were not differentially affected by the concreteness or abstractness
of the items used.

69
Test * stem-type: F(1,16)=.020 p=.890, showing that the levels of priming are similar for the two
types of stem-completion.
Test * stem-type * group: F(1,16)=.288 p=.599, indicating that the priming levels of both groups
followed similar pattern on the two types of testing.
Word-type * stem-type: F(1,16)= 1.004 p=.331, revealing no interaction between the amount of
priming and the word-type of the material that has been used.
Test * word-type * stem-type: F(1,16)=.498 p=.490, indicating that priming levels of both tests
were similarly affected by the concreteness or abstractness of the words.
Group * test * word-type * stem-type: F(1,16)=.279 p=.605, showing that old and AD subjects
presented no difference to the way that their priming interacted with the word-type of the items
on both tests.
Cued-recall
Data were entered in a 2( group) x 2( test: cued-recall 1; cued-recall 2 ) x 2( word-type: concrete;
abstract ) mixed ANOVA.
As with the analyses performed on the stem-completion data, two separate analyses, involving
YCS and OCS, and OCS-matched and AD groups, were performed.
Young-old controls
There was a significant main effect of group, F(1,38)= 24.879 p<.000, showing that overall, the
old subjects recalled fewer correct responses to the stems ( 5.425 ) than did the young ( 7.288 ).
The two remaining main effects, test, F(1,38)= 3.390 p=.068, and word-type, F(1,38)= 1.704
p=.194, were both non-significant, although, as can be seen, the first of these approached
significance. Thus, there was some evidence that recall performance improved from the first
version of no time pressure ( 6.013 ) to the second one of time pressure( 6.112 ).
Table 8. Cued-recall hits for the young and the old controls

70
group overall
cued-recall hits
cued-recall 1
hits
cued-recall 2
hits
YCS 7.288 6.825 7.750
OCS 5.425 5.200 5.650
None of the interaction terms ( listed below ) approached significance. Thus, although the old
subjects recalled reliably fewer words than the young, this effect was equivalent for both tests, and
both word types.
Group * test: F(1,38)=.405 p=.526, showing that there was no difference in the way that both
groups performed in cued-recall 1 and 2.
Group * word-type: F(1,38)=.055 p=.815, indicating that the two groups were not differentially
affected by the word-type of the items.
Test * word-type: F(1,38)=.324 p=.570, showing that there was no interaction between the levels
of performance on both tests and the word-type of the material.
Group * test * word-type : F(1,38)=.252 p=.616, showing that the levels of performance of both
groups and in both tests, were similarly affected by the word-type of the items.
As with the old-young comparison, there was a significant main effect of group, F(1,14)=13.773
p=.000, showing that overall, the AD subjects generated significantly fewer correct answers (
2.583 ), than did their matched controls ( 4.950 ). Neither of the other two main effects
approached significance. Thus, overall, recall performance was equivalent for tests 1 ( 3.750 ) and
2 ( 3.783 ), F(1,14)=.003 p=.958, and recall of concrete words ( 3.950 ) was not reliably different
from that of abstract words ( 3.583 ), F(1,14)=.331 p=.568.
Table 9. Cued-recall hits for the matched old controls and the AD patients

71
group overall
cued-recall hits
cued-recall 1
hits
cued-recall 2
hits
matched
OCS 4.950 5.000 4.900
ADP 2.583 2.500 2.665
As with the old-young comparison, none of the interaction terms ( listed below ) approached
significance, again showing that the observed recall deficit, in this case for the AD group, was not
differentially affected by word-type.
Group * test: F(1,14)=.044 p=.835, showing that old controls and AD patients did not performed
differentially to the two types of the cued-recall test.
Group * word-type: F(1,14)=.699 p=.407, indicating that the two groups were not differentially
affected by the concreteness or abstractness of the words.
Test * word-type: F(1,14)=1.205 p=.277, revealing no interaction between the performance on the
two types of test and the word-type of the material.
Group * test * word-type: F(1,14)=.098 p=.755, indicating that both groups and in both tests
demonstrated a performance similarly affected by the word-type of the items.
Analysis of responses by word-frequency
In order to examine whether groups differed in terms of the normative frequency of words
generated in response to the stems on the four memory tests, responses were analysed as follows.
First, they were classified as either target or non-target responses. In this case, "target" corresponds
to a studied word generated to a stem either in the implicit or explicit tests. Non-target responses
therefore included all other valid words given in response to the studied and nonstudied stems.
More specifically, the stem-completion nontargets and the cued-recall nontargets were all the
correct responses given to the stems by the subjects, and were not included in the study-lists ( false
positives ). For both response types, the mean word frequency was estimated ( using Kucera and
Francis norms ) for each subject, which then formed the basis for the one-way ANOVAs which

72
follow. Only the old-young comparison for non-target completions on the stem-completion test
approached significance.
young-old controls
stem-completion targets: F(1,38)= 1.505 p=.227
stem-completion nontargets: F(1,38)=3.883 p=.056, suggesting that, approaching significance, the
old controls provided words of greater frequency ( 94.7880 ) than those of the young controls (
75.5620 ).
cued-recall targets: F(1,38)= 2.185 p=.148
cued-recall nontargets: F(1,38)=.272 p=.606
matched old controls-AD patients
stem-completion targets: F(1,16)= 2.169 p=.160
stem-completion nontargets: F(1,16)=.105 p=.750
cued-recall targets: F(1,14)=.233 p=.637
cued-recall nontargets: F(1,14)= 1.313 p=.271
The mean frequencies of the target and nontarget responses of the two groups in both the stem-
completion and the cued-recall tests were as follow:
Table 10. Mean frequencies of the target and nontarget responses of matched-OCS and ADP
groups in stem-completion and cued-recall
group Stem-completion
targets
Stem-completion
nontargets
Cued recall
targets
Cued recall
nontargets
matched
OCS
40.82 92.56 38.92 72.17
ADP 53.08 98.66 34.02 101.81
Here are the bar charts for the word-frequency of targets and nontargets that the patients and their
matched controls produced:
1 1 0
1 0 0
9 0
8 0

73
Bar chart 1. Meanfrequency of stem-completion(sc) targets and stem-completion nontargets
of the AD patients and their matched controls
GROUP 1: matched-OCS GROUP 2 : AD patients
Bar chart 2. Meanfrequency of cued-recall(cr) targets and cued-recall nontargets of the AD
patients and their matched controls
GROUP 1: matched-OCS GROUP 2 : AD patients
G R O U P
21
M e
a n
1 2 0
1 0 0
8 0
6 0
4 0
2 0
c rta rg e ts
c rn o n ta rg e ts

74
The mean scores of the frequency of words that all three groups provided were:
Table 11. Mean scores of word-frequency of all groups
Stem-completion
targets
Stem-completion
non-targets
Cued-recall
targets
Cued-recall
non-targets
40.54 87.42 35.00 80.02
The relevant bar chart illustrates data as follows:
c rn o n ta rg e tsc rta rg e tss c n o n ta rg e t
s
s c ta rg e ts
M e
a n
9 0
8 0
7 0
6 0
5 0
4 0
3 0
2 0

75
Bar chart 3. Meanfrequency of stem-completion targets, stem-completion nontargets, cued-
recall targets and cued-recall nontargets for all groups
The mean frequencies of words that all three groups provided to the two separate categories of
targets and non-targets of each test were as follows:
Table 12. Mean frequencies of targets and non-targets for each group
Group Stem-completion
targets
Stem-completion
nontargets
Cued recall
targets
Cued recall
nontargets
YCS 35.14 75.56 32.12 71.95
OCS 40.92 94.78 38.16 80.76
AD 53.08 98.66 34.02 101.81
The relevant bar charts follow:
G R O U P
321
M e
a n
1 2 0
1 0 0
8 0
6 0
4 0
2 0
s c ta rg e ts
s c n o n ta rg e t
s

76
Bar chart 4. Meanfrequency of stem-completion targets and stem-completion nontargets for
each group
GROUP 1: YCS GROUP 2: OCS GROUP 3: AD
Bar chart 5. Mean frequency of cued-recall targets and cued-recall
nontargets for each group
GROUP 1: YCS GROUP 2: OCS GROUP 3: AD
Paired-samples t-test for targets*nontargets was significant for stem-completion ( t=-9.28,df=57,
p=.000 ) and cued-recall ( t=-5.472, df=53, p=.000 ).
DISCUSSION
Young-old controls
G R O U P
321
M e
a n
1 2 0
1 0 0
8 0
6 0
4 0
2 0
c rta rg e ts
c rn o n ta rg e ts

77
As we saw from the analysis of results, in both memory tests, there was a main effect of group
suggesting, overall, that young subjects generated more target "hits" to the stem-completion tests
and more correct responses to the cued-recall tests, than did the old subjects. Additionally, there
was a significant interaction between prime and group in the stem-completion analysis,
demonstrating that old controls exhibited limited priming when they were compared to young
participants. More specifically, although the two groups were similar at the production of baseline
words, they differed at the generation of target "hits".
According to several studies in aging memory that we referred to earlier ( p 54 ), there seems to
be a concordance in studies that examined explicit memory in old persons. Most of the researchers
established the existence of an age effect, that acts as a decreasing factor to explicit memory
performance of older subjects. In our study, old participants were found to perform more poorly
to cued-recall in comparison to young subjects.
Another finding was the existence of a main effect of test, approaching significance, indicating
that the second version of cued-recall conducted under the time pressure of 5 secs generated a
tendency towards better performance for both groups compared to the first version of 20 secs that
was not considered to create a sense of time pressure to the participants. Specifically, the mean
score of young controls was 6.825 for cued recall 1 and 7.750 for cued recall 2, and the mean score
of old subjects was 5.200 for cued recall 1 and 5.650 for cued recall 2. It seems, therefore, that
healthy subjects, young and old, when tested on explicit memory under time-pressure, tend to
provide greater scores.
Nevertheless, two alternative interpretations could ascribe for such a finding. The first is translated
into practice effects. It is possible that at the second version of cued-recall that followed the first
one, subjects were "trained" at performing the task in a more effective way. As a consequence the
scores of the normal subjects improved from the first version to the other. The second interpretation
that could be given is that the second version of cued-recall is performed in a "quicker" way, as
the subjects are required to proceed to the following trigram every 5 secs. Consequently, as a result
of a less delay during the testing procedure, the scores could have been improved due to less
forgetting compared to cued-recall 1.
The findings of implicit memory in aging are not consistently similar to those of explicit memory
studies, especially in the case of stem-completion.

78
In our experiment old subjects were found to prime poorly compared to young participants, in both
types of stem-completion. Therefore, our findings support the view that aging could act as a
decreasing factor in implicit memory as well.
Another finding that approached significance, was the group * test * stem-type interaction. Young
subjects seemed to have been affected differentially by the two versions of test as far as priming
levels are concerned, whereas old subjects were not. More specifically, as we saw, the relevant
data demonstrated the young participants' tendency to provide greater priming levels under time
pressure in stem-completion 2,-similarly to cued recall 2-, whereas old participants were not
affected in any way by the two versions of test. Therefore, there is an indication that testing under
time pressure could positively affect the performance of healthy young subjects in explicit as well
as implicit memory. Healthy old subjects, on the other hand, appeared to be affected by time
pressure only in explicit memory, -although to a more limited degree compared to young subjects-
, and were not affected by time-pressure in the stem-completion task. In this case one might suggest
that during cued-recall testing, practice effects or the "quick" second version can also account for
improvement of performance. Nevertheless, similar tendencies did not occur during the stem-
completion testing procedure, whereas young controls exhibited the same pattern of improvement
in both memory tasks.
Finally, there was no word-type effect of the stimuli on the performance of both groups, on either
memory test. Although young and older participants provided significantly more concrete words
as target hits than abstract ones in stem-completion, priming levels were not affected in any way.
As we mentioned earlier in this study, previous experimental findings indicated ( Rissenberg
and Glanzer, 1987 ) that normal subjects tended to produce more concrete words as responses to a
naming to definition task. In stem-completion, additionally, where subjects, both young and older
were instructed to produce responses to stems, there was a similar effect of concreteness. An
explanation could be that concrete words are "easier" compared to abstract ones, because it is
easier for someone to capture their meaning and bring their image in mind . It seems, though, that
explicit memory is more "sensitive" than implicit memory as far as word-type is concerned. In a
cued-recall test participants are from the beginning aware of the fact that they would be tested on
their ability to remember every word that appears on the screen. Consequently, it is possible that
a tendency is generated by subjects to handle every word in a similar way, meaning they invest
equal amounts of attention to each of the items, regardless of their word-type. In stem-completion,

79
on the other hand, participants are simply asked to read a series of words without having to
memorize any of them. During test-phase they tend to provide more concrete target "hits" than
abstract, maybe because the ability to easily visualize concrete words, strengthens their mnemonic
traces. In an explicit memory test like cued-recall, on the other hand, words are treated "equally"
by the participants during their encoding process, because of knowledge of the testing procedure
that will follow. Therefore, similar effort in memorizing the words is made by each participant,
resulting to similar memory for concrete and abstract words.
Post-hoc analysis of word-frequency, on the other hand, is indicative of certain conclusions. As
we saw earlier, there was no significant difference between the three groups regarding the
frequency of the words they provided. Nevertheless, even though all groups produced words of
similar frequency they tended to complete the stems of target "hits" of both the implicit and
explicit tests with words of lower frequency than the words they provided in every other case. In
opposition to the other characteristic of the items, concreteness or abstractness, word-frequency
seems to influence the three groups similarly, for both aspects of memory, implicit and explicit as
well. A speculation could be that word-frequency practices more power on subjects than
concreteness, by forcing them to be in a situation of promptness to recall or prime with words of
medium to low frequency. Higher-frequency words are more possible to be produced when the
first three letters are given, but when a prior exposure of stimuli takes place it seems that for all
participants, regardless of neuropathology, medium to low-frequency words leave more intense
traces.
Another explanation lies simply to the fact that high-frequency words are more likely to be
produced when subjects are asked to use trigrams without any prior exposure to studying-lists. But
if a study-phase is preceded then low-frequency words enter memory in a more decisive manner
and become more likely to be produced than in the first occasion. As a result there is a difference
in the frequency of words provided to stems after their presentation during a studying phase and
those provided to stems not belonging to any studied items.

80
According to the results, there was a main effect of group in both memory tests, demonstrating
that old participants completed more targets in all versions of stem-completion and cued-recall.
As it is well-established by now, although explicit memory has been found to deteriorate in healthy
aging, it is, though, far better than the explicit memory of AD patients. In this experiment, in spite
of the fact that old controls performed more poorly than young subjects, they did much better in
completing the cued-recall stems with correct answers comparing to AD patients.
There was no other significant main effect or significant interaction, showing that the two groups
were in no way affected by the experimental manipulation of time pressure. There was no
influence by the word-type of stimuli on the performance of neither group and for neither testing,
although concrete words provided by the subjects, numerically exceeded the abstract ones (
stem-completion: concrete:.185, abstract:.153, cued recall: concrete: 6.600, abstract: 6.112 ).
As we saw at the young-old controls comparison, there was a main effect of test for cued-recall
that approached significance, indicating a bias towards better performance under the second
version of time pressure. In the old controls-AD patients comparison, the OCS sample consisted
of 10 healthy old subjects matched to the AD patients in terms of NART scores. We compared the
two sub-groups of the OCS sample and found no significant main effect of group ( F(1,18)=
1.400p=.252 ). Nevertheless, it seems that matched OCS performed more "equally" than the rest,
in the two versions of cued recall ( matched OCS: cued recall 1: 5.000, cued recall 2: 4.900,
remainingOCS: cued recall 1: 5.400, cued recall 2: 6.400 ).
Analysis by ANOVA that sought to investigate any differences in stem-completion between the
two subgroups, indicated no significant main effect of group ( F(1,18)=.101p=.754 ),
or significant group * test interaction ( F (1,18) = .184 p =.673 ), showing that the two
subgroups behaved similarly towards the two versions of stem-completion.
As it was anticipated, AD patients were dramatically impaired in the two cued-recall tests,
replicating all previous findings that determined dramatic deterioration of explicit memory as the
main characteristic of Alzheimer's disease.
As we mentioned in the meta-analysis, the researchers use two ways to report their findings
regarding AD performance in stem-completion. Either by comparing the priming magnitude
between the patient and the control group or by comparing the target and the baseline hits generated
by the AD subjects. In our study the difference in performances between patients and healthy
participants did not reach significance, although it approached the criterion value F(1,16)= 3.778

81
p=.070. Therefore, a t-test paired analysis was conducted in order to evaluate AD priming as the
difference between target and baseline hits, showing almost significant priming ( t=-
2.243,df=7,p=.06 ).
In both stem-completion tests were unaffected by any of the conditions. Time-pressure in stem-
completion had no impact on the performance of either the patients or the matched old controls.
According to the data, it seems that testing under pressure caused different results from what we
predicted. Our prediction was that high pressure conditions would prevent healthy subjects from
using explicit strategies during completion of stems. Consequently, we would not expect they
could exhibit greater performance at the second versions of tests. Nevertheless, the tendencies that
were generated proved otherwise, especially for young subjects. Before discussing any possible
explanations, we should refer to the failure by AD subjects to exhibit any similar tendency towards
the same direction, as they were predicted to. The initial suggestion was that high pressure during
stem-completion testing, would create conditions of "automaticity" that would lead the patients to
perform in a highly perceptual manner, similar to that used when they participate in word
identification. According to the data, different results between the two versions of test did not
occur. The question lies as to whether the idea of "automaticity" can really have an impact on AD
performance or whether the time limitation used in this study was actually in the position to
produce the pressure needed.
There was also no significant main effect of word-type, showing that the concreteness or
abstractness of stimuli was not associated with the magnitude of "hits" that either group achieved
in stem-completion. There was no significant group * word-type interaction { F(1,16)=.292p=.596
}. We analysed the data between the two subgroups of healthy old subjects, but found no
significant group * word-type interaction ( F(1,18)= 2.260p=.150 ) to indicate that the two
subgroups behaved differentially towards items of certain word-type.
As we mentioned earlier, Nebes ( 1988 ) referred to a study ( Rissenberg and Glanzer, 1987 )
where AD patients were found to be dramatically worse than the controls in naming to definition
with abstract words rather than concrete. That prompted us to include the word-type variable in
our experiment, in search of a material-specific phenomenon that could affect the performance of
patients. Unexpectedly, it was demonstrated that there was no influence on performance on either
memory task. It seems, therefore, that the mechanisms underlying the production of words in stem-
completion and naming to definition are not similar or in any way connected.

82
Finally, we are going to refer to the post hoc analysis of word-frequency. We saw earlier that AD
patients did not differ from their matched controls when they provided responses to both memory
tests; they, too, produced words of much lower frequency as targets than as nontargets.
As it was mentioned in the description of the materials, all words were pseudorandomly assigned
in terms of word-type and word-frequency, and the mean frequency of all targets used was : 37.67.
Looking back at Tables 1, 2 & 3 we realize that, overall, the words given by the subjects as targets
are of medium to low-frequency, whereas nontargets are of quite higher frequency.
Investigators who examined the role that word-frequency plays on several priming tasks, have
found contradictory findings. There were researchers ( MacLeod, 1989 ) who reported that low-
frequency words produced greater priming in word-fragment completion than high-frequency
words, whilst others ( Tenpenny and Shoben, 1991 ) claimed the opposite pattern and yet others (
Roediger et al, 1992 ) who underlined the existence of this effect in word fragment completion but
its limitation in stem-completion.
In our experiment, high-frequency words do not seem to be preferred by subjects when they
complete stems with target " hits ", but we have to admit that we did not, initially, include word-
frequency as a potential variable. A post hoc analysis simply showed that there is no significant
difference, in terms of word frequency, between the three groups in the way they provide answers
as targets and nontargets.
In the experiment of verbal fluency by Ober, Dronkers, Koss, Delis and Friedland, ( 1986 ) AD
patients did not perform differentially compared to healthy subjects in providing words of low or
high frequency. The same assumption agrees with the general picture of our data. Nevertheless, in
the case where the difference between the mean frequency of targets and non-targets is due to the
fact that stems belonging to unstudied items are more likely to be completed by high frequency
words as we have mentioned in the young-old subjects comparison (p75) the following argument
should be considered. If normal subjects tend to better remember studied low-frequency words and
this results to the difference in target and nontarget frequencies, then for the AD patients that
follow the same pattern in word production regarding frequency this reflects a certain amount of
memory.
CONCLUSIONS

83
In general, the data appeared as follows:
Old subjects were found to perform poorly in comparison to young subjects. According to several
studies, explicit memory in aging people has repeatedly been found to deteriorate.
In the implicit memory test of stem-completion where findings are more obscure, our data seem
to be compatible with the views that growing age also affects memory performance in implicit
memory as well.
AD patients appeared to perform dramatically worse than old controls in cued recall, as it has been
replicated thus far in many studies. However, although they were found to be slightly impaired
compared to their matched controls, priming occurred. We begin by discussing the last of the above
findings.
Gabrieli ( 1998 ) reviewed the cerebral areas considered to be responsible for explicit and implicit
memory. Episodic memory is mediated, according to relevant research ( Scoville and Milner,
1957; Cohen and Squire, 1980, Gabrieli et al, 1988 ) by medial-temporal areas and diencephalic
systems. Priming, on the other hand, is thought to be mediated by different cerebral areas; we
notice, though, that perceptual priming is thought to use separate neural circuits compared to
conceptual priming. Priming is mediated by neocortical areas; perceptual priming is focused at
modality-specific cortical regions such as auditory and somatosensory neocortices ( Schacter et al,
1996; Squire et al, 1992 ), whereas conceptual priming is focused at amodal language areas such
as frontal and temporal lobes ( Demb et al, 1995; Gabrieli et al, 1996b; Wagner et al, 1997; Raichle
et al, 1994; Blaxton et al, 1996 ).
Today, Alzheimer' s disease is known to cause extended damage to association neocortices in the
frontal, parietal and temporal lobes, resulting, hence, to profoundly impaired explicit memory and
conceptual priming. On the other hand, the disease leaves modality-specific cortices, such as
primary visual, somatosensory, auditory and motor, relatively spared. This means that perceptual
priming remains intact in most AD patients, according to several studies on many tasks ( Balota
and Duchek, 1991; Fleischman et al, 1995; Park et al, 1998 ). Although intact priming on
perceptual tasks such as word-identification, lexical decision, object naming etc, is often reported,
stem-completion remains a memory task that reveals contradictory findings. In our opinion, this
particular memory task is difficult to categorize as either a perceptual or conceptual priming test.
In this experiment, we sought to investigate whether stem-completion could be designed to follow

84
a procedure as perceptual as possible, in order to appeal to cerebral areas which are responsible
for the perceptual analysis of the stimuli and have been found to remain relatively spared in AD
patients. Since the procedure of stem-completion study-phase was repeatedly investigated by
several researchers as to which degree represents the perceptual component of the test, we
attempted to manipulate the test-phase in a way that could limit the conceptual component of the
test and enlarge the role of the perceptual one.
The main idea of the experiment was to conduct, according to indications that stemmed from
relevant research, a stem-completion testing which would follow such an experimental procedure
that would create ideal conditions for the facilitation of the patients' performance. At first, words
that served as stimuli were carefully selected in terms of concreteness and frequency and
thoroughly counterbalanced across study lists. We included the variable of concreteness in order
to examine whether abstract target words could become a "blocking" effect on the priming ability
of the patients. Our intention was to investigate whether AD subjects would produce more concrete
target words as responses to stems, in a way that could affect their priming. If that would be the
case, it could serve as a material-specific phenomenon, applying, as a contributing factor, to the
discrepancies among the results. The second measure was to use time limitation as means of
diminishing the conceptual component during test-phase, which was thought to be enhanced
through intense searching of semantic capacities. Specifically, we suggested that under time
pressure they would simply provide words that met the perceptual criterion of the first three letters,
without having to establish particular traits of each word. That was the idea of " automaticity " as
a method to avoid the troubling situation of self-directed searching in order to provide well-known
semantic units.
The patients, on the contrary, did not treat stem-completion under pressure similarly to the way
they treat word-identification where it has been found that they prime normally ( Fleischman et al,
1995; Keane et al, 1991; Keane, Gabrieli, Growdon & Corkin, 1994; Koivisto, Portin & Rinne,
1996, e.t.c. ). Our initial speculation was that patients do not necessarily fully possess the semantic
characteristics of each target word they identify during the test-phase of a perceptual-identification
test. The main idea was to create in stem-completion such conditions, that patients would react in
a way highly perceptual as the one encompassed in a word identification test-phase.
In word-identification, patients are instructed to try and identify words briefly presented to them.
This particular instruction mainly appeals to their perceptual abilities rather than their semantic

85
structure. Consequently, patients are able to prime normally, since the disease leaves primary
visual, somatosensory, auditory and motor cortices, relatively spared. Accordingly, our prediction
was that if experimental manipulations in stem-completion managed to lead patients to perform
by appealing to the same cerebral areas, this would reflect intact performance.
One explanation for the failure of the initial prediction is that the conceptual component of the
stem-completion test-phase is resistant to perceptual influences. The AD subjects usually selected
to participate in such a memory task, are at first stages of the disease in order to ensure
comprehension of the instructions. During first stages the patients maintain intact certain aspects
of their lexical abilities, such as visually identifying and reading words. The perception, therefore,
of stimuli is a process they continue to possess without facing serious problems. Nevertheless, they
start to face difficulties in other aspects of lexical and semantic abilities, such as attributing
characteristic traits to each word they come across to, establishing, hence, every word as a well-
known semantic unit. The procedure where participants are required to provide words to trigrams,
could become highly conceptual for a person with awareness of its lexical deficit. It could be,
though, that completion of stems, which for a normal subject is not considered such a demanding
task and can be done with a certain amount of "automaticity", could result, for an AD subject, to a
task highly resistant to any attempted perceptual manipulations.
Another explanation is that the well-known memory deficit of the Alzheimer disease, prevented
the patients from performing at a state of automaticity, simply because they would not feel time
pressure. More specifically, it could be that patients, as the second version of stem-completion was
in progress, forgot about or didn't pay attention to the instruction of limited time and treated the
second version similarly to the first. In this case, difficulties in designing a time-pressure testing
could not be overcome, since "spontaneity" of the procedure could not be achieved by following
constant reminding and interventions.
Regarding the word-type variable, data failed to support the existence of a material-specific
phenomenon. AD patients were totally unaffected by the concreteness of material; by contrast, in
cued recall, data favoured numerically the production of abstract words. Nevertheless, in all other
cases, concrete target words exceeded numerically, without, though, affecting the performances of
the subjects.
One speculation is that in the report by Rissenberg and Glanzer,( 1987 ) where patients faced great
difficulties in providing abstract words to definitions, the answers lies in word-frequency. This

86
means that in the case where the items were not counterbalanced in terms of frequency and
concreteness as well, and the experimenters simply made an observation, the dramatic deficit of
patients' production of abstract words could be due to low-frequency.
The post-hoc analysis of word-frequency, on the other hand, offered clearer assumptions regarding
the tendencies that all three groups demonstrated. All participants tended to produce words of
average frequency as target hits for the conditions of both memory tasks, whereas frequency was
twice as great for nontarget words. It is interesting that AD patients did not follow a different
pattern of words in terms of frequency compared to normal subjects. However, since we did not
initially include word-frequency as a variable, we are in no position to make any assumptions as
to whether it interacts with the performance of the subjects or whether it generates different effects
between the three groups.
Following, we are going to refer to the performance by healthy old subjects. As Baddeley ( 1999
) points out, memory appears to be affected by aging in many ways, even by reduction in levels
of attention. Gabrieli et al ( 1999 ) supported the idea of lack of attention as a contributing factor
to the decrease of memory in healthy aging. Baddeley ( 1999 ) also refers to two biological factors
that seem to influence human memory as age progresses. The first one involves maintaining
normal blood sugar levels by following a certain nutritional diet ( Manning, Hall and Gold, 1990
) and the second concerns the effect of competition stress on performance ( Backman and
Molander, 1986 ). We suggest that the problems aging people face as they get older, reflect a
certain degree of cognitive and memory reduction.
There were studies ( Partridge et al, 1990 ) where it was speculated that intact priming in AD was
a product of impaired performance by the old subjects. We suggest that unless matched controls
were of great age ( 75 or more ) it is unlikely that general appearance of the old group was affected
by a pre-clinical AD state that had not been yet diagnosed, as was suggested by Gabrieli et al (
1999 ). Nevertheless, there have been studies demonstrating intact priming for patients, even when
their matched controls were younger. More specifically, there were five studies to demonstrate
intact priming for patients whose age was greater than the controls and only one study consisted
of normal subjects that aged more than 73.4 ( Scott et al, 1991; Deweer et al, 1994; Huberman
and Moscovitch, 1994; Fleischman et al, 1997; Koivisto et al, 1998 ).

87
Even though AD patients were not affected by any of the experimental manipulations, there were,
however, some indications that healthy subjects were influenced by time pressure, especially the
young ones. The relevant data, that approached significance, were as follows:
In cued recall there was a main effect of test, showing that, overall, all subjects, young and old,
performed better under time pressure.
In stem-completion, there was a group * test * stem-type interaction, indicating that, although old
controls performed similarly in both versions of stem-completion, the young ones exhibited greater
priming under time pressure.
Young subjects, thus, were found to be influenced by limitation of time during test-phase, in an
explicit and an implicit memory test as well, whereas old controls exhibited similar tendency only
in cued recall test. Additionally, when the sub-group of matched OCS was subtracted from the
OCS sample, the remaining subjects improved their performance. Our initial prediction was that
normal subjects not only would remain unaffected by manipulation of time, but also that this
manipulation should be a measure of limiting the use of explicit strategies.
Taking into consideration the fact that distractor tasks were used during both stem-completion
versions, we doubt the possibility that the results concerning the improvement of young controls'
performance under time-pressure were due to use of explicit memory. The small tendencies
observed in the two normal groups were unexpected and no possible explanation was found.
In attempting to elucidate the findings, we are going to refer to individual differences among the
controls. In some cases, subjects were unaffected by time pressure and provided similar results to
both test versions, whereas in other cases they improved during the second version. Taking a closer
look at the target hits the subjects provided, we see the following:
In cued-recall there were 40 performances ( 20:young, 20:old ) under the no pressure-condition to
be compared with another 40 performances under high-pressure. In 27 cases, the two performances
were quite similar to each other, with a difference in the production of correct responses between
cued-recall 1 and 2 being from none to three items. In 2 cases, both belonging to the OCS sample
where variability was greater than the YCS sample, performance on cued-recall 1 was better by 4
items than cued-recall 2. In the 11 cases left, the difference in the production of correct responses
between the two conditions was from 5 to 9 items, in favor of the second version. Consequently,
a 27.5 percent of normal subjects were particularly benefited from the high-pressure condition,
whereas only 5 percent of the sample exhibited similar tendency under the no-pressure condition.

88
We speculate that some of the normal subjects may improve their memory performances under
time-pressure; the conditions under which this might happen or the reason why time pressure left
the healthy OCS group unaffected during stem-completion, remains unknown. At this point, we
underline the fact that most of the subjects provided their responses within the first seconds;
consequently, we suggest that greater limitation of time should take place, in order to investigate
whether it could generate greater tendencies to normal subjects or similar tendencies even to AD
patients.
The effect size of this study as a coefficient correlation r that estimates the difference between
performances of AD and normal subjects has a value of 0.34, whereas as Cohen's d is 0.72. The
effect size of AD priming magnitude has a value of r=0.64 or d=1.67. Looking back at the mean
effect sizes for group-comparison and AD priming, ( 0.27 and 0.64 respectively ), we realize that
the values of the effect sizes of this study are quite similar to mean effect sizes obtained by the
meta-analysis. Therefore, after combining them the mean effect sizes were not altered in any way.
Ending this report, we suggest that results, overall, in addition to the results of the meta-analysis,
are indicative of the existence of AD priming although it is impaired in comparison to healthy
controls. This might reflect a strong conceptual component in stem-completion, quite resistant to
any attempts of limitation in favor of the perceptual component. In our opinion, there should be
more intensive research concerning the differences in the performances among patients that are at
different stages of the disease. Stem-completion is a memory task that has been repeatedly
investigated as to what seems to be the cause for all the contradictory findings reported thus far.
The whole matter calls for a closer look on the basis of its complexity compared to other tasks;
additional research should also examine the ways in which severity of AD disease in several
samples is translated into controversial results. In this study, 8 patients participated in the
experimental procedure. Greater samples should be recruited in order for more thorough research
to be completed that would examine different individual tendencies among the patients.
Time-pressure on the other hand, is another factor that could be interesting to be examined more
carefully, especially in terms of experimental design. This means that in a future attempt, it should
be more useful to test individually all subjects in their general ability to briefly complete stems,
and then create a time-threshold that would generate time-pressure even to subjects particularly
"quick" in completing stems.

Implicit memory in Alzheimer

More Related Content

Similar to Implicit memory in Alzheimer (20)

Implicit memory in Alzheimer