An investigation of implicit and explicit memory in college students and healthy older adults

MISSING IMAGE

Material Information

Title:
An investigation of implicit and explicit memory in college students and healthy older adults
Physical Description:
vii, 132 leaves : ill. ; 29 cm.
Language:
English
Creator:
Rediess, Sharilyn, 1959-
Publication Date:

Subjects

Subjects / Keywords:
Memory -- Adult   ( mesh )
Memory -- Aged   ( mesh )
Aging   ( mesh )
Clinical and Health Psychology thesis Ph.D   ( mesh )
Dissertations, Academic -- Clinical and Health Psychology -- UF   ( mesh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 1989.
Bibliography:
Bibliography: leaves 123-131.
Statement of Responsibility:
by Sharilyn Rediess.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001073673
oclc - 20896125
notis - AFF8296
System ID:
AA00009081:00001

Full Text











AN INVESTIGATION OF IMPLICIT AND EXPLICIT MEMORY
IN COLLEGE STUDENTS
AND HEALTHY OLDER ADULTS








By

SHARILYN REDIESS


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1989














TABLE OF CONTENTS


page


ACKNOWLEDGEMENTS . .

ABSTRACT . . .

CHAPTERS

ONE INTRODUCTION . .

TWO REVIEW OF THE LITERATURE. . .

Memory without Awareness in Amnesia .
Implicit and Explicit Memory
in Normal Aging . .
Summary of Implicit
and Explicit Memory Literature .
Experimental Rationale and Hypotheses .


iii


THREE METHODS . .

Lexical Decision Experiment .
Item Recognition Experiment .

FOUR RESULTS . .

Lexical Decision Experiment .
Item Recognition Experiment .
Comparisons of Lexical Decision
and Item Recognition .
Implicit and Explicit Memory
for Single Words .
Shopping List Test .


FIVE DISCUSSION


Priming for New Associations .
Priming of Studied Old Associations
Priming of Non-Studied Old
Associations . .
Summary and Future Directions .

REFERENCES . .

BIOGRAPHICAL SKETCH . .


. 46

. 46
. 60

. 64

. 64
. 70

. 80

. 84
. 86

. 88


88
111

113
119


. 123


132


.














ACKNOWLEDGEMENTS
There are many people who have been instrumental in my

accomplishing this goal. I thank my teacher, advisor, and

friend, Russell Bauer. I credit his support, guidance, and

teaching for the conceptualization of this research project,

as well as my development as a clinician, and my fascination

with neuropsychology. The most important element in

learning is a teacher who is excited about his topic, who

challenges his students to think, and who gives his students

the support and confidence to grow. I have been most

fortunate to have had such a teacher in Rus.

This study also reflects ideas generated under the

influence of Robin West's lectures on memory and aging. She

deserves many thanks for sharing her wealth of knowledge in

a manner that stimulates exciting research ideas. I feel

quite fortunate to have had the benefit of her careful and

thoughtful comments in the writing of this paper.

I am also grateful to Ira Fischler for his comments on

the paper and his always difficult questions. Someday, I

hope to answer them. Jacquelin Goldman also deserves many

thanks for her contributions to this project, as well the

important role she has played in my clinical training.

iii







Thanks also to Edward Valenstein for always redirecting my

attention to the "neuro" part of neuropsychology.

Nathaniel Martin deserves my gratitude as well as a

medal for providing the computer programming for this study

and for helping me clarify my ideas. His patience and

emotional support has kept me afloat.

Finally, I wish to thank my parents, Herman and Sharon

Rediess for so many things that are difficult to put into

words. All my successes belong to them.














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

AN INVESTIGATION OF IMPLICIT AND EXPLICIT MEMORY
IN YOUNG COLLEGE STUDENTS AND
HEALTHY OLDER ADULTS

By

Sharilyn Rediess


May 1989

Chairperson: Russell M. Bauer, Ph.D.
Major Department: Clinical and Health Psychology


Traditional measures of memory rely on explicit

remembering, or the conscious recollection of information.

Recently, attention has been focused on implicit memory, or

learning which can be demonstrated indirectly without

conscious retrieval. The implicit/explicit memory

distinction is of particular relevance to the study of

memory and aging, given this literature's emphasis on age-

related impairments in conscious, strategic retrieval.

Implicit memory for pre-existing representations in memory

(e.g., words or idioms) has been well demonstrated in a wide

variety of clinically amnestic patients and in older adults.

Models developed from this literature have contributed to







our understanding of the relationship between encoding and

retrieval processes. Implicit memory for newly learned

associations, which have no pre-existing representation in

memory, is not well accommodated by these models, but

nevertheless is of great theoretical importance.

The present study investigated the effect of age on

implicit and explicit memory for newly learned associations

and old associations. Lexical decision and item recognition

priming were employed as measures of implicit associative

memory; item recognition accuracy and cued recall were used

as explicit memory measures. In addition, the influence of

automatic versus strategic processing at retrieval was

investigated by varying the time available for memory search

in the two associative priming tasks.

Age differences in implicit and explicit memory tasks

varied in the predicted direction, indicating that older

adults were more sensitive to the manipulations of retrieval

demands than younger adults. The manipulation of time

allowed for search also affected older adults

differentially, suggesting that age differences were

manifested as a disruption of strategic memory search.

Finally, older adults demonstrated a bias toward processing

well-learned semantic relationships, at the expense of

processing novel information.









The specificity hypothesis is proposed to accommodate

these data, and to provide a framework for future

investigations of implicit and explicit memory distinction.

This hypothesis states that the demonstration of learning is

the result of a dynamic interplay between the specificity of

the memory representation, the specificity of the retrieval

cue, and the specificity required in the subject's decision

or response.


vii












CHAPTER ONE
INTRODUCTION


Experimental neuropsychologists and cognitive

psychologists are fascinated by the many facets of memory.

Various dissociable dimensions of memory functioning in

normal subjects and patient groups are observed, measured,

simulated, and modeled in the investigation of the normal

and abnormal process. Understanding the nature of these

dissociations in memory and the conditions under which they

are masked or amplified is the challenge of the memory

researcher.

The amnestic syndrome is particularly interesting to

memory theorists because it accentuates dissociations that

are much less apparent in normal memory. Anterograde

amnesia is characterized as much by the profound disruption

in new learning as it is by the sparing of immediate memory

span, remote memory, general knowledge, and previously

learned skills (Squire, 1982). In addition, recent studies

indicate that amnestic patients show normal or near-normal

ability to demonstrate learning of certain kinds of

material, under certain conditions (Cohen, 1984; Graf,

Squire and Mandler, 1984, Kinsbourne and Wood, 1975). These

dissociations in memory, seen most clearly in acquired brain







disease, allude to the structure, processes, and

organization of normal memory. Although there continues to

be considerable controversy over the theoretical and

anatomical implications of preserved learning in amnestics,

most researchers agree that the investigation of spared and

impaired memory abilities is central to understanding the

nature of human memory.

Within this endeavor, research examining the role of

conscious awareness in mnemonic processing has had

considerable influence on our present conceptualizations of

human memory. Traditional memory measures have relied on

tasks requiring explicit conscious recollection of

information acquired during a specific learning event. Word

list learning and story recall are typical examples of

explicit memory tasks. In contrast, new learning can also

be demonstrated using tasks which do not require conscious

remembering. These implicit tasks reflect prior learning in

the form of savings across learning trials, alterations in

response biases, and decreased response latency to

previously presented stimuli.

The importance of awareness as a component of memory is

highlighted in reports of dissociations between implicit and

explicit memory in amnestic patients. These patients can

demonstrate normal learning and retention on a variety of

implicit memory tasks, while explicitly denying memory of

the event in which this material was learned. Dissociation

between memory with and without awareness has been reported







in amnestic and memory disordered patients of various

etiologies.

The existence of such a dissociation suggests that

implicit and explicit tasks involve distinct memory systems

or processes which may be differentially vulnerable to the

neuropathological changes occurring in memory disorders.

In addition, fundamental questions are raised regarding the

role of awareness in memory and the expression of knowledge.

The study of this phenomenon, and its implications for

the neuropsychology of memory, has been pioneered entirely

by investigations of clinically amnestic patients. However,

the study of normal age-related changes in memory and

cognition is being recognized as a valuable approach to to

test theories of spared and impaired memory function. While

healthy older adults do not exhibit the severe memory

disorder seen in clinical amnesia, two important

commonalities exist. First, the pattern of spared and

impaired memory functions in older adults bears striking

similarity to that observed in true amnesia. Immediate

memory span or primary memory, fund of knowledge, and

previously learned skills remain stable across the lifespan

(Craik, 1977; Howard, 1983; Heaton, Grant, and Matthews,

1986). Considerably less resilient, however, is the type of

mnemonic processing required on typical laboratory memory

tests. Older adults perform well below younger controls

when learning is incidental or intentional, under

elaborative or nonmeaningful encoding conditions, and under







free recall, cued recall or recognition (Craik, 1977;

Perlmutter and Mitchell, 1982; Poon, 1985). Paralleling

studies of amnestic patients, several studies have recently

shown that implicit memory performance is not affected by

aging (Light, Singh, and Capps, 1986, Moscovitch, 1982;

Howard, 1987; Rabinowitz, 1986; Light and Singh, 1987;

Moscovitch, Winocur, and McLachlan, 1986). The existence of

this dissociation between explicit and implicit memory in

this population as well as other memory disorders suggests a

fundamental vulnerability of explicit memory processes to

the neuropathological changes occurring in both normal aging

and various amnestic syndromes.

A second area of convergence is suggested in the

neurobiology of memory dysfunction in aging and amnesia.

Senile changes (cortical plaques, neurofibrillary tangles,

and granuovacular degeneration) which characterize dementia

of the Alzheimer's type, are present to a lesser extent in

the brains of cognitively intact elderly (Bondareff, 1977).

These cellular abnormalities as well as areas of cell loss

are found primarily in the hippocampus, prefrontal and

superior temporal regions of neocortex (Bondareff, 1977;

Tomlinson, Blessed, and Roth, 1968). Memory changes

occurring in normal aging may reflect gradual dysfunction of

the corticolimbic system, which must be more severely

affected to produce clinically significant amnesia.

This paper considers the contribution of the study of

memory and awareness to current conceptualizations of







mnemonic processes, and tests hypotheses derived from this

literature within the context of age-related changes in

memory performance. Specifically, Chapter Two will review

the implicit/explicit memory literature, discuss prominent

models derived from existing experimental work, present the

applications of this work to the study of memory and aging,

and outline the pertinent research questions addressed by

the present study. Chapter Three details the procedure used

in the Lexical Decision and Item Recognition experiments,

and Chapter Four presents the results obtained. The

discussion in Chapter Five focuses particular attention on

the retrieval demands in the two experimental procedures and

the interaction between well-learned associations and new

associations. A framework for conceptualizing the results

obtained, as well as previous work, will be presented.












CHAPTER TWO
REVIEW OF THE LITERATURE

Memory Without Awareness in Amnesia

At a time when psychoanalytic views of unconscious

memories and functional amnesia dominated the psychiatric

literature, Korsakoff (1889) and Claparede (1911) provided

the earliest anecdotal reports of memory without awareness

in organic amnesia. As part of a conditioning experiment,

both physicians subjected their amnestic patients to a

noxious stimulus (an electric shock and a pin stick to the

hand). At a later meeting, Korsakoff's patient expected

his doctor to "electrify him" (Schacter, 1987, p. 504);

Claparede's patient was reluctant to shake hands with him,

declaring that "sometimes pins are hidden in people's hands"

(Schacter and Tulving, 1982, p. 25). Neither patient could

recall the episode in which these events occurred, and they

could only provide a confabulated explanation for their

apprehension. Korsakoff proposed that memory traces were

formed after the onset of the amnestic syndrome. Although

these traces were unavailable to consciousness, Korsakoff

believed that they affected ongoing behavior (Korsakoff,

1889, as cited in Schacter, 1987).







Procedural Learning

The first clear experimental demonstrations of memory

without awareness are found in studies of the bitemporal

amnestic patient, H.M. Despite this patient's profound

anterograde amnesia, he clearly showed savings in mirror

tracing (Milner, Corkin, and Teuber, 1968), tactual maze

learning (Corkin, 1965), rotary pursuit performance,

perception of Gollin Figures (Corkin, 1968; Milner et al.,

1968), and complex puzzle solutions (Cohen and Corkin,

1981). While H.M. required more trials to criterion than

control subjects and performed more slowly on motor tasks,

he demonstrated savings which stood in marked contrast to

his inability to explicitly recall the experience of

learning. Milner et al. (1968) also described examples of

H.M.'s residual learning in everyday life. For example,

fourteen years after the surgery, H.M. had learned the

arrangement of rooms and furniture in his house. Milner et

al. observe that, "just as in the test situations, these

achievements appear to depend on frequent repetition of the

items and their embedding in a constant framework" (pg.

232). They did not indicate whether this everyday example

of residual learning capability was reported explicitly by

H.M. or inferred indirectly through his behavior at home.

Subsequently, motor, perceptual, and cognitive skill

learning have been reported in profoundly amnestic patients

of various etiologies, including Korsakoff's disease,

encephalitis, anoxia, ruptured anterior communicating artery







aneuyerism, and closed head injury (Brooks and Baddeley,

1976; Cermak, Lewis, Butters and Goodglass, 1973; Cohen,

1984; Cohen and Squire, 1980; Moscovitch, 1982; Wood, Ebert

and Kinsbourne, 1982). Many of these skills were

maintained over extended periods of time, and, typically,

the patients did not remember the practice sessions in which

the skills were acquired (Cohen, 1984).

Direct or Repetition Priming

Extending these observations of preserved memory

function in amnesia, Warrington and Weiskrantz (1968, 1970,

1974) reported that amnestic patients exhibited retention of

words and pictures using a variation of a cued recall test.

In their experiment, subjects first studied a list of words

or a set of pictures. Later, the memory test consisted of

the initial letters of studied words or studied items in

degraded form which subjects were required to identify.

Control subjects and amnestic patients produced an

equivalent number of correct responses using this technique,

while dramatic differences were obtained using standard free

recall and recognition tests.

Graf, Squire, and Mandler (1984) determined that the

demonstration of preserved memory using partial information

hinges upon a subtle instructional manipulation. In their

experiments, subjects were given the first three letters of

a previously studied word which they were asked to complete

with the first word that comes to mind. Word stem

completion for control subjects and amnestic patients were







compared with cued recall performance in which subjects used

similar word stems as cues to aid recall of studied items.

The results indicated that allowing the amnestic patient to

use unconscious retrieval (by asking them to complete word

stems with the first word that comes to mind) was critical

to demonstrating preserved memory performance.

This direct or repetition priming has been observed in

amnestic patients in the form of increased probability to

complete word stems with previously studied items (Graf,

Squire and Mandler, 1984), alterations in free association

responses (Shimamura and Squire, 1984), facilitation of

reaction time to previously presented words in a lexical

decision task (Gardner, Boller, Moreines and Butters, 1973;

Scarborough, Cortese and Scarborough, 1977; Moscovitch,

1982), reductions of reading speed for previously studied

items (Moscovitch, et al., 1986), alteration of spelling

bias by presenting homophones (Jacoby and Witherspoon,

1982), and modifying preference judgements for previously

presented melodies (Johnson, Kim and Risse, 1985). Amnestic

patients demonstrated learning at normal or near normal

levels when memory was assessed using these implicit

measures, while they performed well below controls on

explicit memory tests for the same material.

These findings raised questions regarding the

relationship between explicitly and implicitly demonstrated

learning. Several investigators (Graf, Mandler, and Haden,

1982; Graf and Mandler, 1984; Jacoby and Witherspoon, 1982)







reported that manipulations that typically affect explicit

recognition and recall (e.g., semantic encoding) do not

influence performance on implicit memory tests such as word

stem completion or perceptual identification. These

findings have lead some to postulate that the implicit and

explicit memory represent two independent forms of memory

(Cohen, 1984; Schacter, 1987; Tulving, 1972 and 1983).

Others have emphasized that, in normal subjects, implicit

memory contributes to explicit remembering (Jacoby, 1983 a

and b; Mandler, 1980; Shimamura and Squire, 1984).

The interaction between implicit and explicit memory

was demonstrated by Graf, Squire, and Mandler (1984). They

reported that amnestic patients' recognition performance,

while considerably lower than normal subjects, was above

chance. The decay of recognition performance for both

amnestic and normal subjects paralleled the decay of word

completion, with the amnestic subjects reaching chance level

in both word completion and recognition after two hours.

This indicated that priming contributed to recognition

performance in both groups, but this effect was short-lived,

and normal subjects were able to augment their performance

with explicit remembering.

Temporary repetition priming is also thought to account

for amnestic subjects' ability to learn strongly related

paired associates. Shimamura and Squire (1984) reported

that amnestics showed near-normal explicit memory for

recently presented related word pairs at immediate testing;







however, this learning decayed to baseline over two hours.

Control subjects maintained the associations beyond the two

hour interval. In addition, a recent study by Mayes,

Pickering and Fairbairn (1987) suggested that amnestic

patients' susceptibility to proactive interference in a A-B,

A-C paired associate learning paradigm may be the result of

temporary priming. In their experiment, subjects rated A-B

pairs, then A-C pairs. In a free association test,

amnestics and normals produced the same number of B list

intrusions in response to A items. In explicit recall,

amnestics produced the same number of intrusions while

control subjects produced fewer intrusions.

These studies suggest that if the systems or processes

underlying implicit and explicit memory are fundamentally

different, they work in tandem in the normal individual and

may account for quantitative and qualitative differences

across memory tasks in amnestic patients.

Indirect Priming and Priming of Existing Representations

The implicit memory findings reported thus far share a

common property in that most demonstrate priming effects

within the same modality of presentation and all depend on

perceptual processing the stimulus at learning and

retrieval. Priming can also occur across modalities and in

the absence of perceptual encoding. Graf, Shimamura and

Squire (1985) reported that priming of word stem completion

(with visually presented word stems) occurred in both normal

and amnestic subjects regardless of whether the target words







were learned auditorily or visually. Within modality

priming was significantly larger than cross modality priming

in both groups. This indicated that specific perceptual

information contributed considerably to the priming effect;

nevertheless, the cross modality priming provided evidence

that a higher order representation of the stimulus was

activated as well. Additional evidence for nonperceptual

factors in priming was provided in a study by Shimamura and

Squire (1984). Shimamura and Squire (1984) presented words

(e.g., baby) for study in an incidental learning task. They

then had subjects free associate to words that were related

to studied words (e.g., child). The probability of

producing a previously presented and related word was

significantly higher than baseline free association for both

control and amnestic patients. These data suggest that the

presentation of a word activates its semantic associates

even though semantic associations were not presented or

emphasized at encoding.

Priming of New Associations

Many interpretations of repetition priming in amnestics

considered this spared function a manifestation of the

activation of previously stored representation (Cermak,

1984; Diamond and Rozin, 1984; Tulving, 1983). Strong

evidence for this view was reported in a study of priming

for nonwords (Cermak, Talbot, Chandler and Wolbarst, 1985).

Normal subjects show priming for previously studied nonwords

(which can be assumed to have no prior representation in







memory), while Korsakoff patients do not. These findings

fit neatly within the view of preserved implicit memory in

amnestics as a temporary activation of unimpaired semantic

memory (Cermak, 1984; Tulving, 1972 and 1983).

However, this view was challenged by Schacter and Graf

(1986). In their experiment, subjects studied unrelated

word pairs in paired associate format (e.g., window--

reason). Subjects then completed word stems which were

presented with cues that either were studied with the target

word (window--rea_ ) or not studied with the target

(mold--rea ). When the subjects were told to complete

the stems with the first word that came to mind, both normal

and amnestic patients responded with targets more often

under the same cue condition (window--rea_) than under the

different cue condition (mold _). In contrast, when their

instructions were to complete the same stems with a word

that was studied previously, amnestic patients showed no

effect of prior study or study context and performed well

below normals. These data indicated that their amnestic

subjects showed implicit memory for word pairs that had no

preexisting association but were recently associated in a

paired-associate task.

Schacter (1987) and others (Schacter and Graf, 1986;

Graf and Schacter, 1987; Shimamura, 1986) argued that these

data indicate that implicit memory in normals and amnestic

patients extends beyond the activation of preexisting

representations in memory. However, the nature of priming







for newly learned associations has remained unclear. Graf

and Schacter identified features of associative word

completion that set it apart from word completion for single

words. First, elaborative encoding strategies aimed at

forming a meaningful association between the two unrelated

words was required to elicit priming for new associations.

Second, Graf and Schacter's (1987) reexamination of their

data revealed that only mild to moderately impaired amnestic

patients exhibited priming for newly learned associations.

This stands in contrast to priming of preexisting

associations (single words and old associations), which can

be demonstrated in severely amnestic patients.

Could implicit memory for information which has no

preexisting representation in memory be a reflection of

explicit memory? Schacter and Graf (1986) and Graf and

Schacter (1987) argued against this interpretation by

pointing out that the memory-impaired subjects performed

much poorer under explicit cued recall instructions than

under implicit word completion instructions. Thus, these

patients apparently were not using explicit remembering to

complete the word stems. In addition, normal subjects

denied using explicit strategies to complete the word stems,

and they did not show higher completion rates during the

second half of the task (which might indicate that subjects

were "catching on" and using explicit memory to aid their

completion).







Schacter and Graf also point out ways in which word

completion for new associations is unlike explicit cued

recall performance for the same associations. For example,

priming for new associations is insensitive to the type of

elaborative encoding (generating a sentence relating the two

words versus rating a sentence regarding how well it related

the two words), or to proactive or retroactive interference.

In summary, priming for newly formed associations shares

qualities with both explicit memory for new associations and

implicit memory for single words.

Shimamura (1986) and Graf and Schacter (1987) propose

that priming for newly formed associations represents a

different type of implicit memory. Shimamura suggests that

forming a new association requires a minimal level of

elaborative encoding, but once this has occurred, the

association can be accessed via implicit retrieval

procedures in memory impaired subjects and normals. Normals

would be expected to augment this encoding and under

explicit retrieval conditions show good recall. Presumably,

the basic level of encoding which mild to moderately

impaired amnestics are capable of performing is still

insufficient for reliable explicit remembering.

Schacter (1987) and Graf and Schacter (1987) support

the view that the processes involved in the priming of new

associations are fundamentally different from those

underlying repetition priming or explicit memory. Thus, the

minimal level of elaborative encoding required to







demonstrate priming for new associations is not simply a

weakened normal elaborative encoding. They point out that

the demonstration of implicit memory for new associations

requires that some portion of the target response be

presented. For example, if subjects study window-reason,

and the association between the two words is later tested

via word completion, subjects must be shown window--rea

in order to generate the correct target; window-- does

not produce priming. A similar constraint to this "new

learning" is seen when amnestics are taught new associations

via the method of vanishing cues (Glisky, Schacter, and

Tulving, 1986). In each case some part of the target item

must be presented to elicit the target response. Thus, the

encoding underlying these new associations is

"hyperspecific". Schacter (1987) accounts for this

hyperspecificity by asserting that priming of new

association must rely to some degree on the activation of a

preexisting representation.

These findings, then, suggest that (1) implicit memory

for new information, unlike priming for single words or

previously acquired representations, requires a minimal

level of elaborative processing as well as incomplete

disruption of neurobiological substrate of memory; (2) the

retrieval of this information through implicit means is not

sensitive to certain factors (interference, type of

elaborative encoding) which are known to influence explicit

retrieval; (3) the formation of new associations in memory







impaired individuals is "hyperspecific", relying heavily on

the retrieval environment to activate preexisting

representations (Schacter, 1985).

Theories of the Implicit/Explicit Memory Dissociation

While the significance of the dissociation between

implicit and explicit memory performance is widely accepted,

the theoretical basis for its existence continues to be

debated. Many researchers argue that the dissociation

reflects separate memory systems, for example episodic

versus semantic memory (Cermak, 1984; Tulving, 1972 and

1983), procedural versus declarative memory (Cohen, 1984;

Cohen and Squire, 1980), or perceptual versus

autobiographical memory (Jacoby and Dallas, 1981). Each

multiple memory system model points out an interesting and

important aspect of implicit/explicit performance

differences, but none has been able to account for the

existing data adequately (Schacter, 1987). Much of the

inadequacy in existing theories can be attributed to the

failure to appreciate differences between various implicit

memory tasks (Moscovitch, 1984; Moscovitch et al., 1986).

Various investigators have recognized this deficiency and

have attempted to approach the implicit/explicit memory

distinction by more carefully analyzing the processing

demands of various implicit and explicit memory tasks (Graf

and Mandler, 1984; Graf and Schacter, 1987; Jacoby, 1983 a

and b; Moscovitch et al., 1986; Shimamura, 1986; and

Schacter, 1985 and 1987). The common element of these views







is the assumption that implicit and explicit memory tasks

reflect a single internal representation of the memory

event. Different cognitive processes (and neurobiological

systems) are assumed to contribute fundamentally different

components to the representation. While a consensus has

not been reached regarding the parameters which modulate the

formation and access of the representation, the current

approach appears to have broader implications for memory

theory in general.

Activation and elaboration. Studies demonstrating

direct or repetition priming suggest that preexisting mental

representations can be "activated" during learning making it

more likely that studied items would be produced or

identified when partial information was provided at test

(Rozin, 1976). An activation explanation of normal priming

in amnesia is supported by studies demonstrating that normal

subjects exhibit priming for recently studied nonwords

(which presumably have no previous representation in

memory), while amnestic subjects do not (Cermak, 1984;

Diamond and Rozin, 1984).

Extending Mandler's (1979, 1980) model of recognition

processes, Graf and Mandler (1984) interpret implicit and

explicit memory task differences in terms of two types of

processing which can occur at encoding: activation and

elaboration. In their view, activation is an automatic

process by which the internal organization of an existing

representation (or schema) is further integrated or







strengthened, increasing its accessibility. This intraitem

integration focuses on the "perceptual, featural, and

intrastructural aspects of the event... independent of its

relations to other events and representations" (Mandler,

1980, p.255). Conscious elaboration, on the other hand,

makes a representation or schema more retrievable by

establishing meaningful relations with other mental events,

thereby creating an identifiable context. Elaboration

allows for the generation of new as well as the

reinforcement of old retrieval paths.

At encoding these two processes contribute to different

components of the internal representation of the memory

event. Activation and its consequence, integration, create

the perceptual component of the memory representation, and

elaboration produces the semantic or conceptual component.

According to this model, "the dissociation between recall

and word completion performance is due to the utilization of

different kinds of information [at retrieval] derived from a

single underlying representation" (Graf and Mandler, 1984,

p. 553). Thus, in normal subjects, the experimenter must

mask the explicit memory component of the implicit task in

order to engage the subject in processing which will be

sensitive primarily to perceptual component of the memory

representation. The neuropathology of amnesia, however,

eliminates explicit recollection as a means of retrieving an

item from memory by disrupting the elaborative process and

hence the formation of the conceptual/contextual components.







Studies of indirect priming indicate that viewing

implicit memory in terms of activation of primarily

perceptual attributes of a memory representation is too

restricted. Semantic properties of a representation must be

activated as well. This criticism does not invalidate the

activation/elaboration hypothesis, but would require a

modification to allow for the activation of general meaning

information. This modification does not seem unreasonable

given that amnestic subjects can access the meaning of a

word when it is presented, and that basic meaning

information can be accessed rapidly without specific

contextual referents (e.g., in reading).

Another limitation of the activation hypothesis is its

failure to allow for temporary perceptual modifications of a

representation via the intraitem integration process. The

automatic integration process, as described by Graf and

Mandler (1984) only allows for the strengthening of existing

features of a representation; modification of features is

not accommodated by the model. Roediger and Blaxton (1987)

have demonstrated that the temporary influence of highly

specific featural information about a stimulus (e.g.,

typeface) can be demonstrated via perceptual identification.

They reported that words presented at a brief exposure

duration are identified more readily if they are printed in

the typeface that was used at study. Information regarding

typeface would not be expected to be contained in a

preexisting representation of a word. Clearly, this







indicates that highly unique and specific perceptual

features are being incorporated into a representation.

Since this incorporation of specific featural information

decays rapidly (unless repeated), it would not be

maladaptive to the maintenance of a stable store of

information.

The activation/elaboration approach also has difficulty

accounting for implicit memory for newly learned

associations. Since elaborative encoding is required to

demonstrate priming for new associations, it is unclear why

this encoding would not allow for explicit retrieval. One

possibility, perhaps accommodated by the activation-

elaboration model, would consider priming of new

associations in mild amnestic as a reflection of a weakened,

but not completely disabled, elaboration process. Thus, a

poorly elaborated association relies heavily on the

supportive retrieval environment characteristic of implicit

memory tasks. Even so, this explanation cannot account for

Graf and Schacter's results; differences in their mild

amnestic patients' word stem completion performance was

attributable to a manipulation of instructions, not the type

of retrieval cue.

Schacter (1987) proposes that priming of new
associations is aided by, although not entirely dependent

on, automatic activation of preexisting representations.

According to this view, the requirement that some part of

the target response must be provided for priming of new







associations to occur, indicates that an activation process

is supporting the effect. Schacter does not clearly resolve

the critical question of how preexisting representations and

new associations interact. Neither does he provide a

convincing explanation of the manner in which implicit

versus explicit instructional manipulations determine

priming effects in memory disorder subjects. The nature of

implicit memory for new associations and its relationship to

explicit memory processes and preexisting associations needs

to be investigated further.

Processing approach. Jacoby and colleagues (Jacoby

and Dallas, 1981; Jacoby, 1983a and b), and Roediger and

Blaxton (1987) support a model that also focuses on the

processing demands characterizing implicit and explicit

memory tasks. Like Graf and Mandler, proponents of the

processing approach interpret performance differences on

implicit and explicit tasks as reflecting the interaction of

encoding and retrieval demands. The nature of this

interaction determines how a single underlying

representation in memory is formed and expressed.

Jacoby (1983b) elegantly demonstrated this interaction

in a study of implicit and explicit memory in normals. In

his experiment, he varied encoding conditions by having

subjects read a target word alone (e.g., cold), read a

target word in the context of its antonym (e.g., hot--

cold), or generate a target word as an antonym of a stimulus

word (hot ???). Processing at retrieval was manipulated by







probing memory with a typical recognition test and a

perceptual identification task. Perceptual identification

involves presenting target words and nonstudied words at a

brief exposure duration; the dependent variable is

proportion of studied versus nonstudied words that are read

by the subject.

Jacoby reported that in recognition testing, words

learned in the generate condition were remembered better

than those encoded in the context condition, which in turn

were better remembered than words encoded alone. The

opposite results were obtained in the perceptual

identification task: words encoded alone were identified at

a higher probability than words presented in context, and

words presented in context were identified at a higher

probability than generated words. In fact, the probability

of perceptually identifying a word which was generated was

not significantly different from the probability of

identifying a new word.

Jacoby proposes that encoding and retrieval operations

vary along a continuum of data-driven and conceptually

driven processing. Data-driven processing involves the

formation and access of a perceptual code in the

representation of an episode. Conceptually driven

processing establishes and retrieves the conceptual and

contextual codes within the same representation. At

encoding these processes contribute to different components

of a representation of a memory event. According to Jacoby,







performance on a memory task depends on both the nature and

match of processing occurring at both encoding and

retrieval. The performance differences on implicit and

explicit tasks reflect the degree to which data-driven or

conceptually driven processing is engaged at encoding and

retrieval.

Unlike Graf and Mandler, Jacoby rejects the notion that

priming depends on the activation of an existing schema or

representation in memory. Instead, he argues that highly

specific episodic representations are formed during

encoding. At retrieval, implicit (data-driven) tasks are

more sensitive to the perceptual components of the

representation, while explicit (conceptually driven) tasks

are more sensitive to the contextual and meaningful

components. Roediger and Blaxton (1987) emphasize that

data-driven and conceptually driven processing should be

considered a continuum with many memory tasks involving both

operations to greater or lesser degrees.

This processing approach is recognized as an

elaboration of Tulving and Thompson's (1973) concept of

"encoding specificity" and Morris, Bransford and Franks'

(1977) notion of "transfer appropriate processing". As

such, there is no distinction between types of processing

that result in "good" or "bad" memory, but instead,

successful demonstration of learning is determined by the

degree to which the activity at retrieval emulates the

activity at encoding.







While data-driven processing accounts for the highly

specific and episodic nature of a memory representation

(Roediger and Blaxton, 1987), Jacoby's restricted use of the

perceptual identification task has resulted in a limited

conceptualization of implicit memory (Shimamura, 1986).

Like Graf and Mandler's activation process, Jacoby limits

data-driven processes to the perceptual components of a

memory representation; indeed, the perceptual component is

most likely to be accessed by the perceptual identification

task. This type of task, however, would not be sensitive to

a minimal level of semantic or "conceptual" analysis which,

according to studies of indirect priming, occurs

automatically at encoding as well (i.e., data-driven

conceptual processing).

Possibly a more damaging criticism of Jacoby's model is

its exclusive emphasis on the episodic nature of the memory

representation, i.e., that each memory experience forms a

specific representation rather than activating an abstract

representation in memory (Schacter, 1987). Studies that

have demonstrated the sensitivity of perceptual

identification to subtle alterations in physical

characteristics of the stimuli (Roediger and Blaxton, 1987)

certainly indicate that unique, episodic information can be

encoded and retrieved implicitly. However, this assertion

does not adequately recognize the critical role of

preexisting representations in priming. If a unique,

episodic representation is formed by data-driven processes,







it also contains perceptual and semantic components derived

from the activation of a previously stored representation

which may include semantic/conceptual information.

The processing approach's emphasis on the formation of

a detailed episodic representation of a memory event might

accommodate the priming of newly learned associations.

However, amnestics lack of priming for nonwords is more

problematic, since it is assumed that this failure results

from the absence of a representation in memory. Thus, again

Jacoby's model must account for the strong evidence that

preexisting representations may contribute to the formation

of new representations in memory.

In summary, Graf and Mandler (1984), Graf and Schacter

(1987), Schacter (1987) Jacoby (1983 a and b, 1984) and

Moscovitch et al. (1986) agree that implicit and explicit

retrieval draw upon the same underlying representation.

Dissociations between implicit and explicit memory tasks

reflect the complex multicomponent nature of the

representation and processing demands with differing

sensitivities to the various components. Less clear is the

nature of this representation and the interaction between it

and the processes which form and access it (i.e., types of

encoding, implicit and explicit retrieval demands). In

addition, the nature of implicit memory for new associations

remains unclear. Its occupies a puzzling borderzone between

typical repetition priming and explicit remembering.







Our efforts to understand the implications of the
implicit/explicit distinction rests on research efforts to

(1) examine the processing demands of various implicit and

explicit memory tasks; (2) examine the nature of implicit

memory for preexisting and new associations; and (3) to

continue to explore these processes in various forms of

memory dysfunction.

Implicit and Explicit Memory in Normal Aging

Studies conceptualizing older adults' memory
difficulties in terms of the distinction between implicit

and explicit memory are a relatively recent contribution to

the literature. Perhaps because there exists a fundamental

process underlying all forms of anterograde memory

dysfunction, the findings reported to date in older adults

parallel those reported in amnestics. In addition, theories

of implicit and explicit memory discussed above share many

of the same constructs with existing conceptualizations of

age-related memory changes.

Direct or Repetition Priming in Older Adults

Given that profoundly amnestic patients demonstrate
normal priming for single words (repetition priming), memory

and aging researchers anticipated that older adults would

exhibit this type of priming as well. This expectation has

been supported. Older adults show normal benefit from prior

exposure on implicit tasks such as perceptual identification

(Light and Singh, 1987), word fragment completion (Light,

Singh and Capps, 1986), and lexical decision (Moscovitch,







1982), while their explicit memory for the same material is

impaired relative to younger adults. These findings have

been used to support age-related impairments in specific

contextual encoding (Light and Singh, 1987) and effortful,

deliberate, self-initiated retrieval processes (Craik, 1985;

Howard, 1987). The integrity of older adults' semantic

memory preexistingg representations) is also demonstrated by

intact repetition priming (Howard, 1987).

Implicit Memory for New Associations

More recent efforts have focused on implicit memory for

new associations for several reasons. Such priming effects

reflect the formation of a new representation in memory or a

substantial and specific modification of existing

representations. Studies of associative priming in

amnestics suggest that priming magnitude for new

associations is affected by some factors that affect

explicit memory (e.g., elaborative encoding, degree of

memory impairment), but not others (e.g., type of

elaborative encoding, retroactive and proactive

interference) (Schacter and Graf, 1986; Graf and Schacter,

1987). Older adults' well-documented deficiency in using

(or benefitting from) elaborative encoding strategies and in

forming new associations under explicit memory conditions

raises the question of whether age differences would be

found in implicit memory for new associations.

The two studies that have examined associative priming

in elderly subjects applied paradigms developed by McKoon







and Ratcliff (1979) and Ratcliff and McKoon (1978) for
studying associative priming in young college students.

McKoon and Ratcliff (1979) had subjects study related and

unrelated word pairs for later cued recall. One group of

subjects then performed word-nonword decisions on target

words of prime-target pairs, some of which had been

previously studied. The other group performed a recognition

judgement on target words of the same prime target pairs.

McKoon and Ratcliff reported that subjects in both lexical

decision and item recognition conditions showed priming

between newly learned word pairs.

Drawing upon this study and other work by the same

authors (Ratcliff and McKoon, 1981), Rabinowitz (1986)

reasoned that the magnitude of priming in implicit

associative memory might reflect the degree to which older

adults had integrated the two members of a pair in memory.

If older adults are impaired in integrating novel

associations, then large age differences in priming

magnitude would be expected for unrelated studied pairs, and

small or absent age differences in priming would be expected

for related studied pairs. In this study, younger and older

subjects studied related and unrelated word pairs for later

cued recall testing. Each pair was presented for 5 sec.

Before cued recall testing, subjects were shown single items

and they were required to respond (as quickly as they could)

"yes" if the item had been studied and "no" if it had not

(item recognition). Words in this item recognition priming








task were either primed (preceded by the word that it was

studied with) or unprimed (preceded by a word that it was

not studied with). If an association has been formed the

primed trials should lead to faster decisions than the

unprimed trials. This difference between the primed and

unprimed trial is the priming effect.

Rabinowitz found that older adults' explicit memory

(recognition accuracy and cued recall) performance was

significantly worse than younger adults for both related and

unrelated word pairs. However, there were no age

differences in the amount of priming for either related or

unrelated pairs. (Both younger and older subjects exhibited

larger priming effects for related than unrelated pairs).

Rabinowitz argued that the absence of age differences

in priming effects is inconsistent with the view that older

adults fail to adequately integrate novel information in

memory. Instead, these results support a deficit in

deliberate, effortful retrieval and a "conscious evaluation

stage", along with a corresponding increase in their

reliance on automatic processes.

A study by Howard, Heisey, and Shaw (1987) evaluated

priming of new associations by presenting unrelated word

pairs for study within sentences (e.g., the dragon sniffed

the fudge). This method allowed them to determine whether

older adults (like younger adults) encode sentences as

nondirectional propositions by testing forward and backward

priming. More relevant to the present discussion, they







manipulated study time between groups, allowing 15 sec. per

sentence in the short study condition and an additional 10

sec. presentation in the long study group. As in

Rabinowitz's (1986) study, an item recognition task was used

to examine associative priming and explicit memory was

tested via cued recall. Howard et al., reported that older

adults did not exhibit priming for studied pairs under short

study conditions. With extended study, both younger and

older subjects exhibited equivalent priming in terms of

magnitude and bidirectional (forward and backward) effects.

Howard et al. concluded that older adults require more

study time to establish new associations than younger

controls. Given this additional time, however, older adults

exhibit automatic activation of the new association in a

similar manner as younger subjects.

Theories of Age-Related Memory Deficits

Interpretations of dissociations between implicit and

explicit memory in normal aging are couched within existing

conceptualizations of age-related changes in memory. The

views presented here are not only influential in the current

thinking of memory and aging, but they are also particularly

relevant to the study of implicit memory phenomenon in

general.

Automaticity of processing and retrieval. One popular

conceptualization of age-related memory changes is that

distinguishing automatic and effortful or strategic

processing (Hasher and Zacks, 1979). Automatic processing








occurs rapidly outside of our awareness, does not demand

attentional resources, and is not affected by aging.

Effortful or strategic processing is slower, deliberate,

requires attentional resources. Automatic processing is

sufficient for encoding the "flow of information" such as

physical characteristics of a stimulus or frequency of

occurrence information, but effortful processing is required

for meaningful encoding, bringing strategies to bear on a

memory problem or deliberate retrieval of specific

information. Proponents of this view argue that older

adults become less able to engage in this type of processing

or, alternatively, this type of processing becomes less

efficient with age.

The essential difference in retrieval demands of

implicit and explicit memory tasks has been characterized in

terms of a distinction in the degree to which conscious

effortful processes are required at retrieval (Howard, 1987;

Rabinowitz, 1986). This interpretation is largely based on

a two-process model of retrieval (e.g., Baddeley, 1982;

Klatsky, 1984). Retrieval can occur through a passive,

automatic or "direct match" process in which attributes of a

stimulus rapidly access the representation in memory.

Alternatively, a slower, deliberate search strategy can be

initiated following conceptually organized associations. As

described above, Howard (1987) proposed a general deficiency

in this deliberate search process with aging, affecting







retrieval of both well learned (semantic) and newly learned

material.

The automatic-effortful processing continuum has been
an influential construct within the memory and aging

literature. This processing continuum is implied, if not

directly stated, in the process/activation theories of

implicit and explicit memory as well. Activation or data-

driven processing is viewed as a rapid, automatic process

which is determined entirely by the stimulus and its

representation in memory, rather than by a strategy adopted

by the subject. Elaboration or conceptually driven

processing is slow, deliberate, requiring attentional

resources and is subject to the individuals' (or

experimenters') imposed strategies for processing

information. Investigators studying preserved memory

function in amnestics often acknowledge the apparent

significance of conscious awareness in characterizing the

implicit/explicit memory distinction; however, they do not

emphasize automaticity and effort as a useful explanatory

construct. This difference in emphasis between the amnesia

literature and the memory and aging literature most likely

reflects the intuitive appeal within the latter field of a

age-related reduction in "mental energy" paralleling the

reduction in physical energy (Craik and Byrd, 1982).

Although this construct has wide appeal in the aging
literature, there has been considerable disagreement over

the criteria used to independently classify tasks as







"automatic" or "effortful". This difficulty may reflect the

tendency to view processing demands as a dichotomy rather

than a continuum, as well as the problem of circular

reasoning (i.e., characterizing age-related memory deficits

as impaired effortful processing, and in turn defining

effortful tasks as those which older adults show

impairment). Despite these problems, many investigators

(e.g. Shiffrin and Schneider 1977; Craik and Byrd, 1982)

have recognized that degree of automaticity characterizes a

fundamental aspect of mnemonic processing. Whether this

construct has more than descriptive value remains to be

determined. The implicit/explicit memory paradigm provide

the memory and aging research with a new method of examining

this issue.

Semantic memory and general versus specific encoding.

Studies comparing older and younger adults with respect to

word association (Howard, 1983), semantic priming (Howard,

McAndrews, and Lasaga, 1981), and knowledge of scripts

(Light and Anderson, 1983) attest to the stability of the

contents and organization of semantic memory over the

lifespan. Indeed, older adults often perform better than

younger adults on tests of vocabulary and general

information, reflecting the benefit of their rich semantic

networks. They do not, however, benefit from encoding

procedures which emphasize semantic attributes or

relationships in to-be-remembered stimuli. This presents

an apparent paradox: older adults, despite their superior







fund of semantic knowledge, exhibit a particular impairment

in semantic encoding.

Rabinowitz and Ackerman (1982) argue that this paradox

is resolved when considering the types of errors older

adults commit on memory tasks. On a recognition test, older

adults will choose semantically related distractors more

often than younger adults, whereas false positives for

unrelated distractors occur equally in younger and older

adults (Rankin and Kausler, 1979). In similar studies

Perlmutter (1979) and Rabinowitz and Ackerman (1982) had

subjects generate words in response to target words. Later,

younger adults showed a clear advantage in recall when they

were cued with words they had previously generated. Older

adults showed no advantage of their own cues over general

category labels or cues derived from published word-

association norms. Rabinowitz and Ackerman conclude that

older adults encode global or general semantic attributes of

new information, while younger adults encode specific,

unique features of the stimulus. The effect of encoding

specific features allows younger adults to discriminate a

memory representation from other similar representations in

memory. It can be argued that discriminability is an

essential attribute of a retrievable memory representation.

Older adults' tendency toward general semantic encoding

may occur by virtue of a bias toward processing information

in terms of similarity or overlap with past experiences.

Older adults may process new information in a manner







redundant with existing knowledge, rather than distinctive

from existing knowledge. This routinized inflexibility of

processing is consistent with the stability of crystalized

intelligence and the decrease in fluid intelligence over the

lifespan (Horn, 1982).

Although Rabinowitz and Ackerman interpret this problem

of nonspecific processing as occurring at encoding, it is

clear that this same processing bias could be active at both

encoding and retrieval. Nonspecific encoding would create

nondistinctive memory representations; nonspecific retrieval

would create interference among many related representations

in memory.

This conceptualization has not been applied to the

implicit/explicit memory distinction; however, aspects can

be understood within implicit/explicit models developed in

the amnesia literature. First, a shared assumption is that

explicit remembering is highly dependent on distinctive,

contextually relevant components of a memory representation.

Second, processing which is assumed to be inaccessible or of

limited value in explicit remembering is characterized as

relying heavily on a temporary activation of preexisting

information. This activation contributes to components of a

memory representation which (because they are relatively

nondistinctive) can only be reliably accessed via implicit

memory measures. Older adults' encoding of information in a

general manner might be viewed as analogous to the basic

level semantic analysis that can occur via activation or







data-driven processing. Thus, normal implicit memory in

older adults may tap this processing bias toward preexisting

representations which occurs at the expense of contextually

distinctive information.

Questions Raised by the Aging and Memory Literature

In summary, older adults, like mild amnestics, can

exhibit "normal" priming for newly learned associations,

despite their impairment in demonstrating this new learning

explicitly. The Howard et al. data imply that older adults

may not encode these associations in an entirely normal

fashion since they require a longer study time to show

priming for new associations. However, both Rabinowitz

(1986) and Howard et al. (1987) emphasize the role of an

automatic/unaware process in older adults' spared implicit

memory performance and interpret explicit memory deficits as

manifestations of impaired conscious, deliberate retrieval.

Several interesting issues are raised by their

conclusions. First, both investigators based their

interpretations on an item recognition paradigm which, like

explicit recognition testing, requires a conscious memory

decision. Although the dependent variable (reaction time)

is an indirect measure of association between the two items,

this measure also reflects a conscious evaluation regarding

the target item's status as a previously studied word.

Given their emphasis on conscious retrieval, a more careful

investigation of implicit associative priming in older

adults would utilize a task in which the response does not







require a conscious memory decision. Second, the fact that

these investigators were able to demonstrate priming using

the item recognition procedure raises the question of why a

failure in conscious retrieval would affect typical yes/no

recognition testing, but not item recognition priming (which

requires the same response but under speeded conditions).

Successful application of the implicit/explicit paradigm to

the study of memory and aging will depend on more critically

evaluating processing demands and their interactions using

various implicit and explicit memory tasks.

Summary of Implicit and Explicit Memory Literature

The body of research reviewed in this paper may raise

as many questions as it answers. However, the existing data

permit certain conclusions to be drawn: (1) Implicit and

explicit memory tasks access different components of a

single underlying representation of a memory event; (2)

One type of encoding/retrieval process, termed activation or

data-driven, forms/accesses components of a new memory

representation which are characterized by perceptual

information about the stimulus, and a basic level of

featural and semantic information that is influenced by

preexisting representations in memory; (3) A second type of

encoding/retrieval process, elaboration or conceptually

driven, forms/accesses distinctive semantic and conceptual

components of the representation (e.g., contextual cues,

retrieval plans); (4) The first process is obligatory,

occurring rapidly and automatically on presentation of a







stimulus, the second process may or may not occur, depending

on the deliberate use of strategies; (5) The priming of

new associations appears to require both types of processes,

as it shares characteristics of more typical forms of

implicit and of explicit memory; (6) On any memory

measure, the manifestation of learning will depend heavily

on the degree to which the retrieval processes match or

emulate the processes that occurred at encoding; (7) Task

demands may restrict both the type of processing engaged by

the individual as well as the degree of match occurring at

encoding and retrieval; (8) Healthy younger adults use

both activation/data-driven and elaboration/conceptually-

driven processes, often simultaneously; (9) Patients with

substantial disruption of the corticolimbic memory system

are able to perform the first process normally, but this

process typically does not yield components of a

representation that are accessible to explicit memory; (10)

Older adults make errors in memory which suggest that their

representation of a memory event is based heavily on

preexisting representations, with minimal distinctive

contextual information; (11) Studies of implicit memory in

older adults suggest that they can form memory

representations which incorporate new relationships, but

these representations, like those of mild amnestics, are

inflexible and highly dependent on the retrieval

environment; (12) Mild amnestics and normal elderly are

able to perform the activation/data-driven process normally,







but their errors in memory also reflect inefficient

elaborative processing by virtue of an incomplete lesion to

the corticolimbic memory system.

Theoretical questions raised by the implicit/explicit

memory literature emphasize how subtle retrieval processing

variations can elucidate aspects of a memory representation

and the processes which may have occurred at encoding

(Schacter, 1987; Shimamura, 1986; Moscovitch, 1984;

Moscovitch, et al., 1986). Further investigation of

priming of new associations may be a particularly useful

vehicle for understanding age-related memory failures and

the implicit/explicit memory distinction. Research to date

examining this phenomenon has been limited. As mentioned

above, the only studies of associative priming in older

adults used a paradigm which requires a conscious, explicit

memory decision (item recognition). Current

conceptualizations of age-related memory dysfunction

(general versus specific encoding, automatic versus

effortful retrieval, the role of semantic memory or

preexisting representations in the processing of new

information) are well suited for evaluation using the

implicit memory methodology. Comparing implicit memory for

preexisting associations and new associations using tasks

which vary in their retrieval demands will allow for a more

careful analysis of older adults' processing deficits.








Experimental Rationale and Hypotheses

The application of the implicit/explicit memory
paradigm to the study of age-related memory changes is new.

The present study was conceptualized as a broad exploratory

investigation which allowed for the evaluation of several

issues raised by the literature. Four conceptual domains

were explored: (1) the nature of age differences and

similarities in implicit and explicit memory for new

associations; (2) the role of varying retrieval demands in

contributing to age differences and similarities in implicit

and explicit memory tasks; (3) the influence of automatic

and strategic processing at retrieval in determining age

differences and similarities; and (4) the specific

contribution of old, well learned associations to

information processing in older and younger subjects.

The present investigation examined retrieval demands by

using lexical decision and item recognition priming

procedures and cued recall. These memory probes were chosen

for the following reasons. First, the encoding demands and

stimuli can be equated (i.e., subjects study the same list

of related and unrelated word pairs). Second, in both

priming tasks, a prime word would be expected to facilitate

a decision regarding the target word if an association

exists between the two words. Third, the degree to which

explicit retrieval processes are required to. access this

association differs across the three tasks. Lexical

decision requires subjects to decide whether the target word







is a real word or a nonsense word. Thus, facilitation

between a prime and target reflects the association between

the two words, without a conscious memory decision. In

contrast, in item recognition, subjects must decide if the

target is a previously studied word. In this case, the

facilitation between a prime and target is still an indirect

measure of memory since the dependent variable is reaction

time not accuracy; however, the reaction time does reflect

the contribution of a conscious memory decision. Item

recognition accuracy reflects a direct measure of memory;

although the retrieval demands are limited since the target

word is fully presented for evaluation by the subject. Cued

recall requires the highest degree of explicit remembering

since the subject must search and retrieve an item in

response to a cue. The comparison of lexical decision

priming and item recognition priming allowed for a more

careful analysis of age-related processing differences

occurring at retrieval.

Automatic and strategic processing in retrieval was

examined more closely within these priming procedures by

controlling the time between the onset of the prime word and

the onset of the target word, or stimulus onset asynchrony

(SOA). A short SOA (150 msec.) would be expected to prevent

the subject from initiating a search of memory for the

target in response to the prime word. A long SOA (900

msec.) would allow for such a search to be initiated, and

the potential existed for subjects to correctly anticipate







the target if the search was successful. Comparing these

two conditions was proposed as a method to elucidate age

differences in automatic and strategic search of memory.

There has been considerable debate regarding the efficacy of

SOA manipulation in controlling processing; the manipulation

of SOA was nevertheless included with these considerations

in mind.

The effect of age on the processing of new and old

associations was examined by using semantically related and

unrelated words to form the paired associate word pairs and

corresponding prime-target pairs. The priming procedures

allowed for the examination of newly formed associations and

well learned associations under varying retrieval demands

and under automatic and strategic processing conditions.

Age was expected to effect the processing of related and

unrelated word pairs differently with respect to these

proposed age sensitive variables.

Measures of implicit and explicit memory for single

words was also included in the experiment. Implicit memory

for single words is a robust finding in severe amnestics as

well as older adults. If older adults failed to demonstrate

priming for newly learned associations but showed intact

implicit memory for single words, then conclusions could be

drawn regarding specific age-related impairments in the

formation of associations. Perceptual identification was

employed as the implicit measure and two-item forced choice

was chosen as the explicit memory measure.







Hypotheses

Hypothesis One: Age differences in the formation of

new associations were expected to vary with the retrieval

demands required by the tasks. Older and younger adults

were expected to perform similarly in the Lexical Decision

experiment because the demonstration of memory was not

dependent on explicit retrieval. Older adults were expected

to show smaller priming effects than younger adults in the

Item Recognition experiment because younger adults would use

conscious retrieval effectively to facilitate their reaction

times. Item recognition accuracy would show moderate age

differences, again reflecting younger adults' more effective

explicit memory. The largest age differences were expected

in the cued recall task reflecting older adults' inefficient

explicit search and retrieval of a specific item in memory.

Hypothesis Two: Age differences were expected to be

manifested as a disruption of strategic memory search (in

decreased accuracy or reaction time) when, under the long

SOA conditions, older adults are given sufficient time to

initiate a search. When the SOA is short, promoting an

automatic retrieval process, age differences were expected

to be small or nonexistent.

Hypothesis Three: Age differences in processing of old

associations were expected only in the condition in which

old associations and new associations were in conflict.

Older adults were predicted to show priming for the old

associations (supporting their general semantic encoding





45

bias), while younger adults would not show priming (because

they would be expected to form more distinctive new

associations).

Hypothesis Four: Based on the weight of previous

findings, age differences would not be expected for

perceptual identification of single words; however, age

differences would be expected for recognition of single

words.













CHAPTER THREE
METHODS

Lexical Decision Experiment

Subjects

Younger adults. Fifteen University of Florida

undergraduate students (8 females and 7 males) were each

paid $10.00 to participate in the experiment. Each subject

was right handed and reported English as their native

language. Mean age was 22.6 years with a range of 19 to 27

years. Mean years of education and WAIS-R Vocabulary scaled

scores were 14.3 and 12.6 respectively.

Older adults. Fifteen community dwelling older adults

(6 males and 9 females) were also paid $10.00 for their

participation in the study. They were recruited through a

newspaper advertisement and met the following criteria: (1)

high school education, (2) negative history for neurological

disease, alcohol abuse, head injury, seizure disorder,

stroke, significant coronary artery disease, and

uncontrolled hypertension; and (3) self report of good

current health status. Mean age was 67.2 years with a range

of 61 to 77 years. The older adults' mean years of

education and mean WAIS-R Vocabulary scaled score were not

significantly different from that of the younger adults







(13.0 and 12.6 respectively). Although older adults show a

trend toward lower Verbal Fluency Scores than younger

adults, the differences were not significant. Table 3-1

summarizes these subject characteristics.



Table 3-1. Mean Years of Age and Education, WAIS-R
Vocabulary Scale Scores, and Verbal Fluency Scores for
Younger and Older Adults in Lexical Decision Experiment.

Group Age Education Vocabulary Fluency


Younger 22.6 14.2 12.5 46.0
(19-27) (13-17) (10-16) (33-68)

Older 65.0 13.1 11.3 43.4
(60-72) (12-17) (8-18) (27-50)


Note: Verbal Fluency Scores are the sum of correct
responses generated in 60 seconds for each F, A, and S
letter stimulus.

Design

Lexical decision priming paradigm. The experimental

design was a 2 (group) by 4 (priming condition) by 2 (SOA)

mixed factorial design. Group was the only between subjects

variable. Priming condition and SOA were within subject

variables. The four priming conditions were: (1) Studied

Old Associations, (2) Studied New Associations, (3) Non-

Studied Old Association and (4) Unprimed. Word-nonword

prime-target pairs were included as well, but reaction times

for these decisions were not included in the analyses.

Table 3-2 presents a sample study and test trial to

illustrate the four pair types. SOA was varied randomly









Table 3-2. Sam
Priming.
Study Pairs
(n=10)


METAL

CANNON

HATE

OPERA

PRISON

BREAD

CAPTAIN

VICIOUS

PARLOUR

SPORT


Lple Study-Test Trial for Lexical Decision

Prime-Target (Test) Pairs
(n=18)


BENCH

ASSAULT

LOVE

VIOLIN

BLADE

BUTTER

LONELY

TORTURE

CHAIR

BEACH


METAL

PARLOUR

OPERA

SWEET

ATTEMPT

TABLE

INVENT

BARK

BREAD

FAME

KNIFE

MONDAY

HATE

CANNON

NEEDLE

SPORT

NOBLE

GRIN


DATCH

BEACH

VIOLIN

POVE

LONELY

CHAIR

TORTURE

VICIOUS

BUTTER

PRISON

BLADE

POVENT

LOVE

ASSAULT

SHREAT

FIME

CAPTAIN

BONCH


(N)

(F)

(2)

(N)

(4)

(3)

(4)

(F)

(1)

(F)

(3)

(N)

(1)

(2)

(N)

(N)

(F)

(N)


Note: 1=Studied Old Associations, 2=New Associations,
3=Non-Studied Old Associations, 4=Unprimed, F=Filler, and
N=No responses.


across test lists with the constraint that half of the test

lists utilized a 150 msec. SOA and half used a 900 msec.







SOA. SOA was defined as the time from the onset of the

prime word to the onset of the target word. Each subject

studied 20 lists of 10 word pairs. A test list of 18 prime-

target pairs immediately followed each study list. The

pairs and lists were randomized for each subject.

Each of the four priming conditions were represented twice

in each test list. This yielded twenty potential

observations for each priming condition at each SOA for each

subject.

Cued recall test. Every set of five study-test trials

was separated by a cued recall test. This cued recall test

used ten pairs in the five previous study trials (two from

each study list). Five of these pairs were from the Studied

Old Asscociations, and five were from the Studied New

Associations priming condition. Thus, half were

semantically related and half were unrelated, and all pairs

chosen were paired at both study and test. The right hand

word of the pair was presented on a sheet of paper followed

by a blank line. The subject was required to respond with

the word that was paired with presented word in the study

list. Subjects were strongly encouraged to guess if they

were uncertain. Items in the cued recall tests were

randomized (as they were based on the randomization of the

study-test lists) for each subject.

Single item memory tasks. Following the last cued

recall test each subject performed a perceptual

identification task and a two-item forced choice recognition







test. Stimuli for each of these tasks consisted of forty

targets and forty distractors. None of the words chosen had

been used in the Cued Recall test. Two left hand words of a

pair (one from a semantically related pair and one from an

unrelated pair) were selected from each study list to form

the pool of target items. The pool of distractor items were

words never presented during the experiment. Stimuli for

the perceptual identification task and the two item forced

choice task were counterbalanced across subjects such that

each item was represented equally in each test.

Materials

Three hundred and sixty word pairs were formed. All

items were common words between three and eight characters

in length. McKoon and Ratcliff's (1979) published prime-

target pairs constituted two hundred and eighty of the word

pairs used. The remaining 72 pairs were chosen from

Thorndike and Lorge (1944) and Webster's Pocket Dictionary,

and were similar in length, commonality and ease of

associativity. These extra pairs were used as fillers or

restricted to primes of word-nonword trials which were not

part of the main analyses of interest. Nonwords were

created from filler items by replacing vowels or consonants

with randomly chosen vowels or consonants. Of the three

hundred and sixty word pairs, one hundred and twenty were

semantically related and two hundred and forty were pre-

experimentally unrelated. As determined by McKoon and

Ratcliff (1979), the unrelated pairs were easily associated








with study, but the preexperimental association between them

was not sufficient to produce priming between the first and

second words without prior study. The materials also

included a set of extra words to be used in the multiple

choice and perceptual identification tests, as well as

pronounceable nonwords. The extra words were chosen to

conform to the length and commonality characteristics of the

main pool of word pairs.

Study and test list construction, randomization,

stimulus presentation, and data collection were controlled

by an IBM PC/XT microcomputer. The main pool of word pairs

and the pronounceable nonwords were divided into groups

(Primacy, Studied Old Associations, Studied New

Associations, Non-Studied Old Associations, Unprimed,

Nonwords, Fillers, Recency). Unrelated word pairs that

made up the Studied New Associations and the Unprimed

priming conditions were represented equally in the two

conditions, as were the related pairs that formed the

Studied Old Associations and the Non-Studied Old

Associations. Words in the non-experimental groups

(Primacy, Recency, Filler, and Nonwords) remained in these

groupings for all subjects.

Words for a given trial were chosen in the following

manner. One item from the Primacy group and one item from

the Recency group were chosen randomly without replacement

and assigned to position one and ten in the study list. Two

items from each of the four priming conditions were chosen







randomly without replacement and placed randomly in the

study list. Items in the Non-Studied Old Association and

the Unprimed conditions were linked to items in the Recency,

Fillers, and Nonwords groups to recombine word pairs to

conform to the specific priming condition. For example, the

pair "PARLOUR CHAIR" was linked with the Recency pair "SPORT

BEACH" such that on the test list the pairs "TABLE CHAIR",

"PARLOUR BEACH", and "SPORT FIME" were created.

The primary constraint on the construction of the test

lists was that prime-target pairs placed in the first

position of the test list were always "no" responses (i.e,

word-nonword pairs). This constraint was used so a "yes"

response (which would be used in the analyses) would not be

in the Primacy or Recency portion of the test list. The

eight experimental test list pairs and their linked four

word-nonword and four filler word-word pairs were placed

randomly in positions 2 through 17. Thus, the test list

consisted of twelve "yes" responses and six "no" responses.

While a equal number of "yes" and "no" responses would have

been desirable, it was decided that increasing the number of

test items to achieve this balance would have proved

overwhelming for subjects. As the primary analyses of

interest concerned only correct yes responses, it was felt

that this imbalance would not adversely effect the results.

Refer to Table 3-2 for a sample study-test list

construction. For a given subject twenty study-test lists

were constructed.







Procedure

Each subject was tested individually. Following a

brief explanation of the experiment and an inquiry regarding

health history, each subject read and signed an informed

consent form. The subject sat facing an Amdek A 300 amber

monitor and was shown the right and left shift keys which

were labeled "yes" and "no" respectively. A Yes/No practice

trial began each test session. The subject was given the

following instructions: "To get you accustomed to

responding on the computer you will see the word 'yes' or

'no' printed in the center of the screen. When you see the

word 'yes' press the right key, labeled 'yes', with your

right index finger. When you see the word 'no' press the

left key labeled 'no' with your left index finger. You are

to respond as quickly and accurately as you can. Rest your

fingers on the keys so you can respond quickly. Ready?"

Twenty such yes/no trials were presented.

The practice study-test list followed the Yes/No

practice. Subjects were given the following instructions:

"This is a study of memory. You will be studying pairs of

words. In the center of the screen you will see word pairs

presented one pair at a time. Read the two words out loud

and then do whatever you can to try to remember the two

words together. You can say them to yourself over and over,

or make a sentence in your mind, or think of the two words

as pictures and relate them together. Do whatever will help

you remember them together as a pair, because later I will







give you the first word and want you to come up with the

second word. You will see 10 such pairs in one study list.

Let's go through a practice study list. Remember say the

words out loud and try to remember them together as best you

can." The word "READY?" is presented in the center of the

screen. The examiner hit the "return" key to begin the*

study list presentation. Ten word pairs, printed in capital

letters, were presented in the center of the screen for five

seconds each with a one second interstimulus interval.

Subjects were corrected for misreading a word and prompted

for forgetting to read a word aloud.

After studying the practice study list, subjects were

given the following instructions: "Now you will see a test

list. In the center of the screen you will see a first word

followed immediately by a second word. If the second word

is a real word, respond 'yes' by hitting the 'yes' key. If

the second word is a nonsense word, respond 'no' by hitting

the 'no' key. Respond 'yes' or 'no' as quickly and

accurately as you can. Remember you only have to respond to

the second word and you have to decide if that word is a

real word or a nonsense word. After we go through this

practice test list you will see what I mean. Rest your

fingers on the 'yes' and 'no' keys to be ready to respond.

Ready?"

The words "Ready for Test?" were presented in the

center of the screen and the examiner hit the 'return' key

to begin the test list presentation. The prime was







presented in the center of the screen for 100 msec.,

followed by either a 50 or 800 msec. blank screen, then

followed by the target word or nonword presented in the

center of the screen. The target remained on the screen

until the subject made a response. Once the response was

made the next prime was presented 1 sec. later. Eighteen

such prime-target pairs were presented. Subjects'

performance was monitored during the practice session and

correction was provided if necessary. It was stipulated

that subjects would take the practice test again if it was

apparent that they did not understand the instructions;

however, this was not necessary with any subject. Data was

not collected during the practice trial.

After this practice trial the following instructions

were given: "As you noticed some first and second words

were ones you studied together and some were not. Also some

of the words made sense together and some did not. Your

job, though, was to respond yes or no as fast and accurately

as you could whether the second word was a real word or a

nonsense word. At different points during the experiment,

I will also be testing your memory for the word pairs on a

paper and pencil test. You will be given the first word and

you will need to write down the word that you studied with

it. So it is very important that you try to remember the

word pairs. Now, there are twenty study-test lists like the

one you just did. We will take a break after every five

study-test lists so you can rest. Ready for the first study







list? Remember, read the words out loud and try to remember

them together as best you can." The experimental study-test

lists were presented in the same manner as the practice

list. Prior to each test list the subject was reminded to

place his/her fingers on the response keys if necessary.

The computer recorded the subject's reaction time and

response. Subjects were not given feedback regarding their

responses.

After every five study-test trials, subjects were given

a cued recall test. Ten word pairs (one related pair and

one unrelated pair from each of the five study lists) were

selected by the computer and printed out. The right hand of

each pair was printed in capital letters followed by a line.

Subjects were given the following instructions: "This paper

has first words from pairs in the lists you just studied.

Write down the words that went with them. If you can't

remember one please guess because many times guesses are

correct. So try to write a word down for each one even if

you are unsure or just guessing." After he/she was finished

and allowed to stand and stretch, the next five study-test

trials were presented. After the second cued recall test

(at the halfway point) subjects were encouraged to take a 10

minute break and walk about. No subject took a break

lasting longer than ten minutes.

After the final cued recall task the subjects performed

a perceptual identification task. They were instructed as

follows: "This is a test of perceptual speed or how fast







your brain can understand what you see. In the center of

the screen you will see a row of hatch marks. The hatch

marks will be replaced by a word and the word will be

quickly crossed out by a row of "X's". All you have to do

is to try to read the word. Now, the word will be presented

for a very brief time so you will need to pay close

attention to the screen. If you cannot read the word try to

make a guess. Sometimes people can figure out part of the

word and then make a guess. Please make a guess even if you

are not sure. Ready?" Eight number symbols (########)

were presented to the center of the screen for five seconds.

A word then replaced the number symbols for a specified

length of time (described below) followed immediately by

eight "X"'s (XXXXXXXX) for two seconds. The first ten words

were distractors (not previously presented words) presented

in a descending series from 110 msec. to 20 msec. The

eleventh through fiftieth words were both targets and

distractors presented for 20 msec. each in a pseudorandom

order. The subject's responses were recorded down by the

examiner. This protocol for exposure duration was standard

for all Young Adults. In a pilot screening of the procedure

it was apparent that many older adults would not be able to

perform the task at such rapid exposure durations.

Therefore two alternate exposure duration series (60 or 90

msec.) were employed for older adults. The sixty msec.

exposure duration started with the first ten filler words

presented in descending series from 170 to 60 msec. If the







older adult was able to identify five of these first ten

descending series items they continued with the 60 msec.

exposure duration for the remaining perceptual

identification test items. If they identified less than

five of the first ten, the procedure was aborted and the 90

msec. exposure duration was employed. For this exposure

duration, the first ten fillers were presented in descending

series from 200 msec. to 90 msec. with the remaining test

items presented at 90 msec. One subject was excluded from

the study due to an inability to report any of the items

presented at the 90 msec exposure duration.

Following the perceptual identification task, subjects

completed a two-item forced choice recognition test.

Subjects were given a sheet of paper with 20 word pairs.

One member of each pair was a studied word and one and

nonstudied word. Subjects were given the following

instructions: "This is a test to see how well you remember

the word you studied today. One word in each pair is a word

you studied and one is one you did not see today. Circle

the one you recognize. If you are uncertain, make a guess.

Remember only one word in each pair is one you studied

today."

This recognition test completed the main part of the

experiment. Three additional measures were obtained. The

first was the Shopping List Test. Subjects were given the

following instructions: "We are interested in 'everyday

memory'-- the memory tasks people do naturally. We have








been trying to put together an "average person's" shopping

list. We need to figure out what grocery store, department

store, and drug store items people buy most often. To help

us develop this list could you please tell me 15 things you

most oftee buy or a friend or relative buys for you at the

store." Responses were written verbatim. If subjects

provided a general category, for example "meat", they were

asked to give a specific item from the category, for example

"hamburger". Only specific items were recorded. After

approximately 20 minute filled delay interval, subjects were

asked to give the exact shopping list they provided earlier.

Subjects were prompted to give 15 items even if this

required guessing. Responses were recorded verbatim.

During the delay interval the remaining tests were

administered. The WAIS-R Vocabulary Subtest was

administered in the standard manner (Wechsler, 1981).

Controlled Oral Word Association Test (Benton, Hamsher,

Varney, and Spreen, 1983; Lezak, 1983) was then

administered. Subjects were given a letter and asked to say

as many words as they could that begin with that letter in

one minute. The letters F, A, and S were used in all

subjects. Responses were recorded verbatim. Subjects were

also asked to name as many kinds of animals as they could in

one minute. Again, responses were recorded verbatim. The

entire experimental session took approximately two hours.









Item Recognition Experiment
Subjects

Younger adults. Fifteen University of Florida

undergraduate students (6 females and 9 males) were each

paid $10.00 to participate in the experiment. Each subject

was right handed and reported English as their native

language. Mean age was 21.5 years with a range of 19 to 26

years. Mean years of education and WAIS-R Vocabulary scaled

scores were 14.4 and 12.2 respectively.

Older adults. Fifteen community dwelling older adults

(7 males and 8 females) were each paid $10.00 for their

participation in the study. They were recruited through the

same newspaper advertisement as subjects in the Item

Recognition Experiment and met the same criteria. Mean age

was 67.2 years with a range of 61 to 77 years. The older

adults' mean years of education and mean WAIS-R Vocabulary

scaled score were not significantly different from that of

the Younger Adults (14.0 and 13.7 respectively). Table 3-3

summarizes subject variables for this experiment.

Design

The experimental design was identical to the Lexical

Decision Experiment. The same prime-target word pairs

conditions and SOA conditions were used. See Table 3-4 for

a Lexical Decision sample study-test list.







Materials

The same set of 360 word pairs were used in this

experiment. In the test lists, however, extra word pairs

replaced six word-nonwords in target positions.



Table 3-3. Means Years of Age and Education, WAIS-R
Vocabulary Scale Score, and Verbal Fluency Score for Younger
and Older Subjects in Item Recognition Experiment.

Group Age Education Vocabulary Fluency

Younger 21.4 14.4 12.2 46.5
(19-26) (13-18) (9-16) (36-61)
Older 67.2 14.1 13.6 44.8
(61-77) (12-18) (9-18) (30-53)

Note: Verbal Fluency Score is the sum of correct responses
generated in 60 seconds for each F, A, and S letter
stimulus.


Procedure

Subjects performed the yes/no practice trial and study

practice trial in an identical fashion as subjects in the

Lexical Decision Experiment. At the end of the study list

the following instructions were given: "Now there will be a

test list. In the center of the screen you will see a first

word followed immediately by a second word. If this second

word was in the list you just studied, respond 'yes' by

hitting the 'yes' key. Respond 'no', by hitting the 'no'

key, if the second word was not in the list. You do not

have to decide whether the first and second word were








Table 3-4. Sample Study-Test Trial for Item Recognition
Priming Paradigm.
Study Pairs Prime-Target (Test) Pairs
(n=10) (n=18)


METAL

CANNON

HATE

OPERA

PRISON

BREAD

CAPTAIN

VICIOUS

PARLOUR

SPORT


BENCH

ASSAULT

LOVE

VIOLIN

BLADE

BUTTER

LONELY

TORTURE

CHAIR

BEACH


METAL

PARLOUR

OPERA

SWEET

ATTEMPT

TABLE

INVENT

BARK

BREAD

FAME

KNIFE

MONDAY

HATE

CANNON

NEEDLE

SPORT

NOBLE

GRIN


LATCH

BEACH

VIOLIN

SOUR

LONELY

CHAIR

TORTURE

VICIOUS

BUTTER

PRISON

BLADE

EVENT

LOVE

ASSAULT

THREAD

TIME

CAPTAIN

BENCH


(N)

(F)

(2)

(N)

(4)

(3)

(4)

(F)

(1)

(F)

(3)

(N)

(1)

(2)

(N)

(N)

(F)

(N)


Note: l=Studied Old Associations, 2=New Associations,
3=Non-Studied Old Associations, 4=Unprimed, F=Fillers,and
N=No responses.

studied together, only whether the second word was in the

list at all. Respond 'yes' or 'no' as quickly and

accurately as you can. Remember you only have to respond to







the second word and you have to decide if that word was in

the study list. After we go through this practice test list

you will see what I mean. Rest your fingers on the 'yes'

and 'no' keys to be ready to respond. Ready?"

Subjects performed the practice test list and then

proceeded on with the experimental procedure as as described

above. Thus, the only differences between the Item

Recognition Experiment and the Lexical Decision Experiment

were the use of additional fillers place of nonwords in the

test list, and the decision which the subject made regarding

the target word.













CHAPTER FOUR
RESULTS

Lexical Decision Experiment

Priming Paradigm

Outlier observations were defined as reaction times

exceeding three standard deviations from the overall group

mean reaction time. Responses below 200 msec. were also

discarded. Reactions times for correct "yes" responses

only were included in main analyses. Error rates were quite

low, averaging 1 to 6 percent. Mean reaction times for each

Priming Condition at short and long SOA were calculated for

each subject. Group means of these subject means are

presented in Table 4-1.

Subject mean reaction times were submitted to a 2

(Age) X 4 (Priming Condition) X SOA (2) analysis of variance

(ANOVA). Main effects were obtained for Age, F(1,28) =

12.19, p < .0019, Priming Condition, F(3,84) = 55.82, p <

.0001, and SOA, F(1,28) = 8.30, p < .0075. Mean reaction

times for younger and older adults were 581 msec. and 699

msec. respectively.

The significant main effect of SOA reflects the faster

response times exhibited by all subjects in the long SOA

condition as compared to the short SOA. The Age X SOA







interaction approached significance, F(1,28) = 2.82, p =

.10. As Table 4-1 indicates, older adults showed a strong


Table 4-1. Mean Reaction Times (in milliseconds) for
Correct "Yes" Lexical Decisions.

Group SOA Priming Condition
Studied New NonStud. Unprimed
Old Assoc. Assoc. Old Assoc.

(n=296) (n=295) (n=289) (n=296)
150 535 596 600 613
[9] [11] [10] [10]
{1%} (2%} {4%} {1%}
Younger
Adults
(n=297) (n=293) (n=291) (n=292)
900 490 573 613 624
[10] [13] [11] [11]
{1%} (2%} (3%} {3%)

(n=297) (n=294) (n=293) (n=292)
150 671 736 709 762
[10] [10] [9] [10]
{1%} {2%} (2%} {3%}
Older
Adults
(n=296) (n=297) (n=283) (n=290)

900 606 678 689 738
[ll] [11] [11] [10]
{1%} {1%} {6%) {3%}

Note: Number of observations is in parentheses, standard
error is in brackets; percent error (including outliers) is
in braces.


benefit (a reduction in reaction times) in the long SOA

condition over the short SOA condition, by an average of 42







msec. Younger adults exhibited a smaller advantage (11

msec), most likely due to younger adults' generally fast

reaction times (leaving less room for an advantage to be

realized).

Priming effect was obtained by subtracting reaction

time for a primed condition from the unprimed condition.

Table 4-2 presents the priming effects for younger and older

adults at the two SOA conditions. As the overall ANOVA

yielded a significant main effect of Priming Condition and

interaction between SOA and Priming Condition, F(3,84) =

9.49, p < .0001, as well as an interaction between Age and

Priming Condition which was marginally significant,

F(3,28)=2.03, p=.10, planned comparisons of each primed

condition with the unprimed condition were carried out at

each SOA.

Priming of studied old associations. In the Studied

Old Associations condition, at the short SOA, younger adults

showed facilitation of 78 msec., F(1,28) = 20.22, p =

.0003, and older adults showed facilitation of 91 msec.,

F(1,28) = 27.68, p < .0001. At the long SOA, larger priming

effects were obtained with this priming condition for

younger, F(1,28) = 91.32, p < .0001, and older adults,

F(1,28) = 10.82, p = .003. Younger adults showed a 133

msec. facilitation and older adults showed a nearly

identical 132 msec. facilitation. Priming for these old

associations was confirmed at the individual level as well.

Fourteen of 15 younger subjects and 12 of 15 older subjects










Table 4-2. Priming Effects (in milliseconds) for Young And
Older Adults in Lexical Decision Experiment.
Group SOA Primed Condition
Studied New NonStudied
Old Assoc. Assoc. Old Assoc.


150 78* 17 13
Younger (11) (10) (15)
Adults
900 133* 50* 10
(9) (16) (9)



150 91* 24 53*
Older (22) (18) (18)
Adults
900 132* 60* 49*
(18) (14) (17)

Note: Priming effect is the difference in reaction time
between each primed pairtype condition and the unprimed
pairtype condition. Standard error is in parentheses. An
asterisk (*) indicates p < .005.


showed facilitation at the short SOA, and all subjects

showed facilitation at the long SOA. Thus, the priming of

old associations effect was quite robust, occurring in both

groups at both short and long SOA.

Priming of newly learned associations. In contrast,

priming in the New Associations condition occurred only at

the long SOA. Younger adults showed 17 msec facilitation

in this condition at the short SOA, but this was not

significant (F < 1); older adults' priming effect was

slightly larger, but still non significant, F(1,28) = 2.95,







p = .10. At the long SOA, younger adults showed 50 msec.

facilitation, F(1,28) = 10.82, p = .003, and older adults

showed 60 msec facilitation, F(1,28) = 15.76, p = .0007.

Examination of the priming effect at the individual level,

revealed that 12 of 15 younger subjects and 13 of 15 older

subjects showed facilitation at the long SOA. Thus, younger

and older adults exhibited similar pattern of priming for

newly learned associations, with significant faciliation of

reaction time occurring only at the long SOA.

Priming of nonstudied old associations. An interesting

difference between older and younger adults is seen in the

Non-Studied Old Associations condition, however. Older

adults showed priming in this condition at both short SOA,

F(1,28) = 9.61, p = .005, and long SOA, F(1,28) = 12.82, p =

.0016, while younger adults did not (F < 1). This

difference is particularly striking at the individual level.

In the short and long SOA conditions only half of the

younger subjects exhibit facilitation; however, all of the

older subjects show priming in the short SOA condition and

11 of 15 show priming at the long SOA condition. This is a

very interesting difference as it suggests that pre-existing

semantic relationships between words were more influential

in older adults' processing of prime-target pairs than in

younger adults.

The analyses, then, reveal that (1) older and younger

adults showed equivalent priming for old associations that

were recently studied; (2) this priming for old associations







occurred at both short and long SOA, although the priming

effect was larger in the long SOA condition; (3) older and

younger adults showed priming of newly learned associations

at long SOA but not at short SOA; (4) while younger adults

did not show priming for old associations when the target

had been previously studied with an unrelated word, older

adults did; (5) this semantic priming effect in older

adults occurs at both short and long SOA.

Cued Recall

Comparisons of percent correct responses on the cued

recall test for older and younger adults are presented in

Table 4-3. Subject responses consisting of exact target

items or derivations of target items were accepted (eg.

crackers for cracker) as correct responses.


Table 4-3. Mean Percent Correct for Related and Unrelated
Pairs in the Cued Recall Task.

Group Pairtype Total

Old Assoc. New Assoc.

Younger 88% 70% 79%
Adults (10) (17) (13)


Older 83% 38% 60%
Adults (12) (22) (16)
Note: Standard deviations are in parentheses.

It is apparent that older adults performed best when

the pairs were old associaitons, with a mean percent correct

of 83% compared to younger adults mean correct of 88% For

new associations, however, older adults recall half the







number that younger adults recall (38% and 70%

respectively). The number of correct cued recall responses

for old and new associations was submitted to a 2 (Age) X 2

(Pairtype) repeated measures ANOVA. Significant main

effects were obtained for Age, F(1,28) = 12.06, p < .002,

and Pairtype, F(1,28) = 154.58, p < .0001. Reflecting older

adults good recall of old associations and poor recall of

new associations, the Age by Pairtype interaction was

significant, F(1,28) = 21.56, p = .0002. Newman-Keuls post

hoc means comparisons for total cued recall, cued recall of

old associations and new associations yielded significant

differences (p < .05) between older and younger adults for

new associations and overall cued recall performance. The

difference between older and younger adults' recall of old

associations was not significant, however. When interpreted

in the context of the reaction time data, these results

suggest that older adults showed normal implicit memory for

newly associated word pairs, although their explicit memory

for the same pairs was well below that of younger controls.

Item Recognition Experiment

Priming Paradigm

Outlier observations were defined as reaction times

exceeding three standard deviations from the overall group

mean reaction time. Responses below 200 msec. were also

discarded. Reaction times for correct "yes" responses

only were included in main analyses. Error rates varied

with Priming Condition and with Age, averaging 5 to 38%.







Error rate analyses will be presented in detail below. Mean

reaction times for each Priming Condition at short and long

SOA were calculated for each subject. Group means of these

subject means are presented in Table 4-4.


Table 4-4. Mean Reaction times for Younger and Older
Subjects for Correct "yes" Responses in Item Recognition
Priming Task.

Group SOA Priming Condition
Studied New NonStud. Unprimed
Old Assoc. Assoc. Old Assoc.

(n=280) (n=285) (n=237) (n=256)
150 629 683 778 779
[18] [15] [17] [15]
{6%} {5%} {21%) {15%}
Younger
Adults
(n=275) (n=282) (n=244) (n=266)
900 536 609 745 732
[14] [16] [20] [14]
{8%) (6%} {18%) (11%}



(n=265) (n=277) (n=199) (n=237)

150 834 859 978 970
[20] [17] [25] [21]
{12%) {8%} {34%} {21%}
Older
Adults
(n=262) (n=260) (n=187) (n=220)
900 775 866 981 960
[20] [22] [29] [22]
{13%) {13%) {38%) {27%}

Note: Number of observations is in parentheses, standard
error is in brackets; percent error (including outliers) is
in braces.







Subject mean reaction times were submitted to a 2 (Age)

X 2 (SOA) X 4 (Priming Condition) repeated measures ANOVA.

Older adults' mean reaction times were significantly longer

than younger adults' (903.42 msec and 686.48 msec

respectively), F(1,28) = 18.17, p = .0004. Significant main

effects were also obtained for SOA, F(1,28) = 8.05, p =

.008, and Pairtype, F(3,84) = 85.85, p < .00001.

Stimulus onset asynchrony. The effect of SOA on

reaction time is complex and must be considered in light of

the significant interaction between SOA and Priming

Condition, F(3,84) = 3.18, p = .027, and the marginally

significant interaction between SOA and Age, F(3,84) = 3.07,

p = .087. Consulting Table 4-4, it is apparent that younger

adults respond more quickly under the long SOA condition in

all Priming Conditions. This facilitation with long SOA may

reflect a benefit obtained when the longer SOA provided

younger adults with time to initiate a search of memory in

response to the prime word. Having initiated this search

and done so successfully (finding the correct target item in

memory), younger adults were able to evaluate the target

item more quickly. Examining the older adults response

times with respect to SOA suggests process differences.

Older adults show the same benefit under the long SOA

condition when the prime-target pairs are old associations

that were studied. However, when the pairs are newly

learned, this benefit is not seen, nor is it seen when the

prime-target pairs are old associations that were not







studied as pairs. The effect of long SOA on the priming of

new associations may be to allow older adults time to

initiate a search of memory. However, unlike younger

adults, this search is not successful, and, in fact, appears

to impair their ability to evaluate the target as a

previously presented word relative to the short SOA

condition. The short SOA condition may prevent this search

process from being engaged.

Priming of old associations. Priming effect is defined

as the difference in reaction time between a primed

condition and the unprimed condition. These differences are

presented for each group at each SOA in Table 4-5. Planned

comparisons of reaction times for each primed conditions

with the unprimed condition were performed for both groups

at each SOA. As suggested in the analyses of SOA, strong

priming effects were obtained for studied old associations

at the short and long SOA for both younger and older adults

(p values < .0001). This robust priming effect was

confirmed at the individual level with 14 of 15 younger

adults and 14 of 15 older subjects showing priming at the

short SOA, and all subjects in both groups showing priming

at the long SOA.

Priming of newly learned associations. In the analyses

of priming for newly learned associations, both groups show

priming effects at both SOA (p values < .0001). Examining

Table 4-5, the priming effects obtained for newly learned

associations suggested that younger adults show larger










Table 4-5. Priming Effect for Younger and Older Subjects in
Item Recognition Experiment.

Group SOA Primed Condition
Studied New NonStudied
Old Assoc. Assoc. Old Assoc.


149* 96* 1
Younger 150 (19) (12) (13)
Adults

900 196* 123* -13
(20) (14) (15)


Older 150 137* 111* -8
Adults (22) (22) (30)

900 185* 94* -21
(31) (23) (31)

Note: Priming effect is calculated by subtracting reaction
time for a primed condition from the unprimed condition.
Asterisks (*) indicate significance at the p < .0001 level.
condition. Standard errors are in parentheses.


priming effects under the long SOA condition compared with

the short SOA condition (123 msec. and 96 msec.

respectively). However, older adults show the opposite

pattern: larger priming effects are obtained in the short

SOA condition than in the long SOA condition (111 msec. and

94 msec. respectively). Although these Age by SOA

differences are in the direction indicated by the analyses

of reaction time and SOA discussed above, they do not reach

significance when analyzed as priming effect (p = .21).

Thus, older adults exhibited priming for newly learned







associations at both SOA's, but the differences in the size

of the priming effect in the two SOA conditions were not

large enough to reach significance.

At the individual level, all younger subjects exhibited

priming for new associations at both the short and long SOA.

Similarly 14 or 15 older adults showed priming at the short

SOA. Consistent with other subtle differences seen at the

long SOA, only 12 of 15 older adults showed priming at the

long SOA.

Priming of nonstudied old associations. As can be seen

in Table 4-5, the Non-Studied Old Association condition did

not produce facilitate in reaction time for either group, F

< 1. Thus, neither younger or older adults showed priming

between old associations when the target item had been

previously studied with an unrelated word. The absence of a

priming effect in this condition does not reflect a simple

failure of the prime stimulus to facilitate processing of

the target. Instead, a considerable disruption of

processing is evident in the large variance and high error

rate (Table 4-4). The variance and error rate are

considerably higher in the Non-Studied Old Association

condition than that obtained in the Unprimed condition.

(These two conditions are comparable in that both have prime

words that were not studied and target words that were

studied; the only difference between the two is that one

contains prime target pairs that are related and one the

other condition does not). Thus, the conflict between old







and new associations inherent in this condition resulted in

disruption of processing beyond that attributable to being

not primed. This disruption is confirmed at the individual

level with only half the subjects in each group showing

priming at either SOA.

Item Recognition Accuracy

Table 4-6. presents the proportion correct for younger

and older adults item recognition responses. After arcsin

transformation (Winer, 1971), proportion correct was

submitted to a 2 (Age) X 2 (SOA) X 4 (Priming Condition)

repeated measures ANOVA. Significant main effects were

obtained for Age, F(1,28) = 17.7, p = .0005, SOA, F(1.28) =

5.1, p = .03, and Priming Condition, F(3,84) = 60.7, p <

.0001. There was no significant interaction between Age and

SOA or SOA and Priming Condition. However, a significant

Age by Priming Condition interaction was obtained, F(3,84)

=2.7, p = 05. Although a significant three-way interaction

was not obtained, interesting differences in accuracy were

apparent when examining simple effects for each Priming

Condition at each SOA.

Older adults' accuracy was equivalent to younger

adults' for the Studied Old Association condition at short

SOA (F < 1), but the differences between the two groups

approached significance at the long SOA, F(1,28) = 2.9, p =

.10. As anticipated, age-related differences in accuracy

were more apparent for targets of newly learned associates.







At the long SOA significant age differences in accuracy were

obtained, F(1,28) = 4.69, p = .04. At the short SOA,



Table 4-6. Item Recognition Accuracy for Younger and Older
Adults.

Group SOA Priming Condition
Studied New NonStud. Unprimed
Old Assoc. Assoc. Old Assoc.


Younger 150 .94 .96 .81 .90
Adults (.02) (.02) (.03) (.02)

900 .93 .94 .83 .91
(.02) (.01) (.03) (.02)


Older 150 .90 .92 .66 .81
Adults (.03) (.03) (.05) (.03)

900 .88 .88 .65 .76
(.02) (.03) (.04) (.02)
Note: Item recognition accuracy is expressed as proportion
correct; standard error is in parentheses.


however, this difference was only marginally significant,

F(1,28) = 3.03, p = .09. Thus, in these two priming

conditions, the majority of older adults errors occurred

under the long SOA conditions. The difference in older

adults' accuracy for New Associations at the short and long

SOA is particularly apparent in Table 4-4, as the error

rates reported in this table include outlier responses as

well as incorrect responses. When these outlier responses

are included, younger adults have an average error rate of

5% in the short SOA condition and 6% in the long SOA







condition. In marked contrast, older adults have an average

error rate of 8% in the short SOA condition and 13% in the

long SOA condition. This difference in older adults' error

rates in the short and long SOA is not as striking in the

other primed pairtypes or in the unprimed condition.

Dramatic differences in accuracy were obtained in the

Non-Studied Old Associations condition. While both groups

made more errors in this condition, older adults' error rate

was significantly higher than the younger adults for both

short, F(1,28) = 7.91, p = .009, and long SOA, F(l,28) =

18.75, p = .0004. Older adults were also much less likely

than younger adults to correctly identify a previously seen

item in the unprimed condition (when the prime and target

were neither related or studied as pairs). This was

evident in both short SOA, F(1,28) = 9.34, p = .005, and

long SOA, F(1,28) = 25.04, p = .0001.

Cued Recall

Table 4-7 presents comparisons of percent correct

responses on the cued recall test for older and younger

adults. Subjects responses consisting of exact target

items or derivations of target items (e.g., crackers for

cracker) were accepted as correct responses.

It is apparent that older adults performed best when
the pairs were semantically related with a mean correct of

82% compared to younger adults mean correct of 92%. For the

Unrelated pairs, however, older adults recalled half the







number that younger adults recalled (32% and 66%

respectively). The number of correct cued recall responses



Table 4-7. Mean Percent Correct for Related and Unrelated
Pairs in the Cued Recall Task.

Group Pairtype Total

Related Unrelated

Younger 92% 66% 78%
Adults (7) (20) (13)


Older 82% 32% 58%
Adults (8) (14) (10)
Note: Standard deviations are in parentheses.


for related and unrelated pairs was submitted to a 2 (Age) X

2 (Pairtype) repeated measures ANOVA. Significant main

effects were obtained for Age, F(1,28) = 26.10, p < .0001,

and Pairtype, F(1,28) = 220.76, p < .0001. Reflecting older

adults good recall of related pairs and poor recall of

unrelated pairs, the Age by Pairtype interaction was

significant, F(1,28) = 21.56, p = .0002. Newman-Keuls post

hoc means comparisons for total cued recall, cued recall of

related and unrelated pairs yielded significant differences

(p=.05) between older and younger adults for both pairtypes

and overall cued recall performance. Thus, even though

older adults remembered 16 of 20 related cued recall items,

younger adults performed significantly better.







Comparisons of Item Recognition and Lexical Decision

Analyses for comparing lexical decision and item

recognition were performed by submitting priming effects for

each of the three primed conditions to a 2 (Age) X 2

(Experiment) analysis of variance. SOA was included in the

analysis for priming of New Associations only, since the age

by SOA interaction was marginally significant in this

condition only.

Priming for Old Associations

As anticipated, priming of old, well-learned
associations is a robust phenomenon. Given a brief period

of study, older and younger adults in both lexical decision

and item recognition experiments demonstrated large priming

effects for these semantically related word pairs. Table 4-

8 presents a comparison of the priming effects obtained for

this Studied Old Associations condition in the two

experiments. Collapsing across SOA, priming effects were

larger in the item recognition task than in the lexical

decision task F(1,116) = 16.60, p = .0002, with no


Table 4-8. Priming Effects for Studied Old Associations in
Lexical Decision and Item Recognition Experiments.

Group Experiment
Lexical Decision Item Recognition

Younger 105 173
Adults (9) (14)

Older 111 161
Adults (14) (19)
Note: Standard errors are in parentheses.









significant interaction between Age and Paradigm Type (F <

1). Newman-Keuls post hoc means comparisons confirmed this

lack of interaction by indicating that both younger and

older adults showed larger priming effects in the item

recognition experiment (p < .05).

Priming for New Associations

Supporting the hypothesis that older adults can

demonstrate implicit memory for newly formed associations,

both older and younger adults showed priming between

previously unrelated word pairs that had been studied

experimentally. Comparing the two priming tasks (Table 4-

9), however, age differences in processing are suggested.

At the short and long SOA, both younger and older

adults show larger priming effects in the Item Recognition

experiment than Lexical Decision (p values < .05). As

mentioned above, the facilitation observed for newly

associated word pairs at the short SOA in the Lexical

Decision task is not significant in either group. When the

time between the prime and target is lengthened in the long

SOA condition, younger adults still show larger priming

effects in the Item Recognition task, (p < .05) while older

adults do not. Older adults not only respond more slowly in

this condition, but they also commit more errors. Thus, in









Table 4-9. Priming effects for Newly Learned Associations
in Lexical Decision and Item Recognition Experiments.

Group SOA Experiment
Lexical Decision Item Recognition

150 17 96
Younger (10) (12)
Adults
900 50 123
(16) (14)

150 25 111
Older (18) (20)
Adults
900 60 94
(14) (23)
Note: Standard errors are in parentheses.


the demonstration of implicit memory for new associations,

older adults perform like younger adults in the Lexical

Decision experiment, when a conscious decision regarding the

status of a target item in memory is not required. However,

when a memory decision is required and older adults are

given more time to process the prime word (as in the long

SOA condition of the Item Recognition experiment), subtle

but consistent age differences are found in the form of

longer reaction times and increased error rates.

Priming for Nonstudied Old Associations

Clear age and paradigm differences were apparent when

examing the priming effects for old associations that were

not studied (Table 4-10). SOA did not differentially affect

priming effect in this condition so comparisons were

performed collapsing across this variable. In contrast to







performed collapsing across this variable. In contrast to

the other primed conditions, larger priming effects were

obtained in the Lexical Decision task than in the Item

Recognition task, F(1,116) = 9.32, p = .003. As Table 4-10

shows, older and younger adults perform differently in the

Lexical Decision task and similarly in the Item Recognition

task. This interaction between Age and Paradigm Type

approaches significance, F(1,116) = 2.98, p = .08. Newman-

Keuls post hoc means comparisons indicate that younger



Table 4-10. Priming Effects for Non-Studied Old
Associations in Lexical Decision and Item Recognition
Experiments.

Group Experiment

Lexical Decision Item Recognition

Younger 12 6
Adults (14) (14)


Older 51 14
Adults (18) (30)
Note: Standard errors are in parentheses.


adults fail to demonstrate priming in either experiment (p >

.05). Older adults clearly show priming for these old

associations in the Lexical Decision task but not in the

Item Recognition task (p < .05). Thus, the processing

differences in the item recognition and lexical decision

tasks are highlighted when comparing older adults

performance in this priming condition.







Implicit and Explicit Memory for Single Words

Lexical Decision Experiment

Implicit memory for single words was tested via
perceptual identification. The probability of correctly

identifying previously studied words versus nonstudied words

in the perceptual identification task is presented in Table

4-11. Arcsin transformations of these probabilities were

submitted to a 2 (Age) X 2 (Item Status) ANOVA. A

significant main effect for Item Status (studied versus

nonstudied) was obtained, F(1,28) = 49.74, p < .0001,

indicating that studied words were identified more

frequently than distractors. The main effect for Age and

the interaction of Age and Item Status were not significant

(F < 1). Thus, both older and younger subjects demonstrated

implicit memory for previously studied items.

In contrast, age differences were found in the explicit

recognition test, F(1,28) = 6.00, p = .02. Younger adults

recognized an average of 95% (+/-3%) compared to 88% (+/-

9%). Thus, while older adults explicitly recognized fewer

studied words than younger adults in the recognition test,

the two groups identified an equivalent number of studied

words in the implicit task.









Table 4-11. Probability of Perceptual Identification as of
Function of Prior Study for Younger and Older Subjects.

Group Item Status

Studied Not Studied


Younger .65 .42
Adults (.19) (.23)


Older .60 .32
Adults (.24) (.21)
Note: Standard deviations are in parentheses.


Item Recognition Experiment

As in the Lexical Decision experiment, priming or

implicit memory for single words that were studied was

determined by comparing the number of studied and nonstudied

words identified in the perceptual identification task.

Probabilities of perceptual identification of targets and

distractors are presented in Table 4-12. Analysis of these

probabilities revealed a significant effect of prior study,

F(1,28) = 59.00, p < .0001. Main effect for Age approached

significance, F(1,28) = 3.73, p = .06, owing to older

adults' identifying more items overall than younger adults.

This most likely reflects the use of longer exposure

durations (60 msec. or 90 msec.) in older adults and the

variability in the procedure with this group. More

importantly, however, the interaction between Age and Item

Status (studied versus not studied) was not significant (F <







1). Thus, both older and younger adults identified more

studied words than not studied words.



Table 4-12. Probability of Perceptual Identification as of
Function of Prior Study for Younger and Older Subjects.

Group Item Status

Studied Not Studied


Younger .58 .34
Adults (.24) (.22)


Older .73 .46
Adults (.17) (.15)
Note: Standard deviations are in parentheses.



Due to a ceiling effect, significant age differences

were not found on explicit recognition testing. Younger

adults correctly recognized 93% and older adults recognized

90%. This ceiling effect prevents meaningful conclusions

regarding a dissociation between implicit and explicit

memory for single words in older adults.

Shopping List Test

The Shopping List Test was included as a everyday

memory measure of episodic and semantic memory (West and ?,

198?). Subjects are required to selectively retrieve a

subset of information from semantic memory that was

retrieved in a recent episode. The dependent variable was

the number of shopping list items retrieved by the subject

which were items they reported eariler as their most







frequent shopping list items. Total possible correct was 15

items. Mean correct shopping list items did not vary with

across the two experiments, thus the data were collapsed

with respect to this variable. Younger adults retrieved an

average of 14.03 (+/- .65) items while older adults

retrieved an average of 12.97 (+/- .80) items. This

difference was significant at the p=.05 level. Older adults

retrieved fewer shopping list items than younger adults

eventhough these items had been originally generated from

their own familiar shopping list.













CHAPTER FIVE
DISCUSSION

As anticipated, a dissociation between implicit and

explicit memory was demonstrated in older adults'

performance. Both younger and older adults showed priming

for single words, while older adults recognized fewer single

words than younger adults. Since priming for single words

has been reported by other investigators (Light and Singh,

1987; Light, Singh, and Capps, 1986; Moscovitch, 1982), the

present discussion will focus primarily on the associative

priming effects and their theoretical implications.

Priming effects for newly associated words will be

discussed first; the specificity hypothesis will be advanced

as a conceptual framework for explaining the present results

and for directing future investigations of implicit and

explicit memory. Next, the priming effects for old

associations will be discussed. The interpretation of these

results will focus on the influence of overlearned semantic

relationships on the processing of novel information.

Priming for New Associations

Lexical Decision

Age differences in priming of newly formed associations

were not found using a lexical decision task. Both older







and younger adults exhibited convincing and equivalent

priming effects for recently learned word pairs under the

long SOA condition. Facilitation was also obtained at the

short SOA, but this effect did not reach significance in

either group.

The sensitivity of priming for new associations to SOA,

suggests that, although an association was formed, it was

not activated in a passive automatic fashion. Instead, the

long time lag between the presentation of the prime and the

presentation of the target allowed both older and younger

adults to intentionally anticipate the target item. Despite

the fact that the decision itself was not based on the

target item's status in recent memory (but rather on the

target item's status in the lexicon), both groups appeared

to use an intentional search of recently formed associations

to improve their response times.

Previous studies of associative priming in a lexical

decision task have debated whether automatic facilitation

occurs for new associations. While McKoon and Ratcliff

(1986) obtained priming of newly learned associations at SOA

as short as 150 msec., suggesting that new associations can

be automatically activated, many others have not (den Heyer,

1986; Neely and Durgunoglu, 1985; Durgunoglu and Neely,

1987). Durgunoglu and Neely (1987) methodically examined a

variety of factors which might influence priming for new

associations at a short SOA (e.g., intermixing short and

long SOA within a block of priming trials, including







nonwords in the study lists, excluding semantically related

words from the priming trials). They concluded that, in a

lexical decision task, automatic priming for new

associations occurs only under highly circumscribed

conditions. Thus, the absence of priming at the short SOA

found in the present experiment is not unexpected. More

importantly, if a short SOA is assumed to promote automatic

processing, it is clear that this manipulation did not

provide a specific advantage for older adults. Thus, the

claim that older adults' intact implicit memory performance

is the result of the automatic, unaware or passive

activation of an underlying memory representation must be

rejected as overly simplistic.

More interesting then, is the use of strategies by

both older and younger adults to facilitate lexical

decisions. Durgunoglu and Neely (1987) proposed that when a

lexical decision task contains an item from an immediately

preceding paired associate learning task, subjects recognize

that the episodic information is useful in making lexical

decisions. When a prime is presented, recently formed

associations allow subjects to retrieve the representation

of the studied word pair, and thus prepared themselves to

respond "word" if the target is a match. If the presented

target is a word that does not match the representation

retrieved by the subject, the response is slower. "Nonword"

responses are slower still. Priming of lexical decisions

for newly learned associations, then, most likely reflects a







strategic process. Contrary to the hypothesized age-related

decline in effortful processing, older adults appeared to

use this strategy in the same manner as younger adults.

Item Recognition

In the Item Recognition experiment, both groups

demonstrated priming for new associations at both short and

long SOAs. Given that the item recognition task required an

evaluation and explicit decision regarding the status of an

item in memory, it may seem counterintuitive that priming

was obtained at the short SOA (which presumably promotes

automatic processing) in this task when it was not in the

Lexical Decision experiment. However, close examination of

the data reveals interesting age and task differences.

First, the priming obtained at short SOA in this task

cannot be interpreted as automatic facilitation. Reaction

times at the short SOA in both groups were quite long,

indicating that processing continued for up to 700 msec.

after the presentation of the target. The failure of the

short SOA to restrict processing to automatic activation

exemplifies the limitations in using SOA manipulations to

delineate automatic and strategic processing. Particularly

in more difficult tasks, automatic activation will be masked

by other strategic components.

Although the SOA manipulation did not provide

information regarding automatic versus strategic processing,

it does reveal age differences in the process of retrieval.

Younger adults' reaction times were facilitated considerably







in the long SOA condition over the short SOA condition.

This benefit in the long SOA condition may reflect younger

adults initiating a search of memory in response to the

prime. As their search was likely to be successful they

correctly anticipated the target and were able to respond

more quickly. The short SOA did not allow sufficient time

for this search to be initiated. In this case the prime and

target most likely formed a single unit which was then

evaluated regarding its familiarity as a compound cue

(Ratcliff and McKoon, 1988).

Older adults, however, failed to demonstrate this

benefit in the long SOA condition. Older adults still

showed priming in the long SOA condition; however, they

clearly did not benefit from the long SOA (relative to the

short SOA) in the same manner as younger adults. In
addition, older adults were significantly less accurate

than younger adults in the long SOA condition, but the

difference in accuracy was not significant in the short SOA

condition. These findings suggest that, in the long SOA

condition, older adults also initiated a search of memory in

response to the prime. However, this search was less

successful than it was in younger adults. Older adults were

not able to efficiently and reliably identify the correct

target item in memory; thus, when the target was presented,

they did not realize the benefit of searching memory and

anticipating a response. Their higher error rate in the







long SOA condition suggests that alternative responses may

have been retrieved resulting in interference.

Under the short SOA condition, older adults were not

given time to initiate a search in response to the prime.

As with younger adults, the close presentation of the prime

and target allowed the two words to form a single unit which

was then evaluated regarding its familiarity. Older adults

may have responded quickly and as accurately as younger

adults for two reasons: alternative response may not have

been generated in a search process; and the prime and target

together form a stronger and more specific representation

than either alone, decreasing the degrees of freedom in the

match process.

The two previous studies of implicit memory for new

associations (Rabinowitz, 1986; Howard et al., 1987) did not

address the effect of SOA on reaction times. Rabinowitz did

not control SOA. Howard et al., reported that half of her

participants were assigned to a 150 msec SOA condition and

half to a 450 msec SOA condition. They failed to find

differences related to the SOA manipulation. Two factors

may have conspired against finding significant SOA

differences in their study. First, the between subjects

manipulation of this variable may have reduced the power for

detecting an effect. Second, a 450 msec SOA may not have

been long enough to show an advantage in younger adults or,

more likely, the lack of advantage in older adults. Third,

differences in processing propositions (as in the Howard et