Upload
khangminh22
View
2
Download
0
Embed Size (px)
Citation preview
Low Involvement Processing: Effects of Stimulus Exposure and Repetition
on Implicit Memory, Explicit Memory and Affect
Michelle P. Y. Lee
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Rotman School of Management Univers~ty of Toronto
43 Copyright by Michelle P. Y. Lee (2001)
National Library I*i of Canada Bibliotheque nationale du Canada
Acquisitions and Acquisitions et Bibliographic Services setvices bibliographiques
395 Wellington Street 395, me Wellington Ottawa ON KIA ON4 Ottawa ON K 1 A ON4 Canada Canada
The author has granted a non- L'auteur a accorde une Licence non exclusive licence allowing the exclusive pennettant a la National Library of Canada to Bibliotheque nationale du Canada de reproduce, loan, distn'bute or sell reproduire, preter, distribuer ou copies of this thesis in microform, vendre des copies de cette these sous paper or electronic formats. la forme de microfiche/film, de
reproduction sur papier ou sur format electronique.
The author retains ownership of the L'auteur conserve la propriete du copyright in this thesis. Neither the droit d'auteur qui protege cette these. thesis nor substantial extracts fiom it Ni la these ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent etre imprimes reproduced without the author's ou autrement reproduits sans son permission. autorisation.
Low Involvement Processing: Effects of Stimulus Exposure and Repetition on Implicit Memory, ExpIicit Memory and Affect
Michelle P. Y. Lee Dactor of Philosopher (2001) Rotman School of Management University of Toronto
ABSTRACT
Involvement is widely regarded as an important construct in consumer behaviour. particularty in
terms of its role in moderating effects of advertising exposure on memory and affect. Effects and processes
that characterize the higher end of the low involvement continuum. however, are much better understood in
comparison to lower levels. This dissertation attempts to contribute to an understanding of the latter with
three experiments that examine the effects of stimulus exposure and repetition on memory (experiments 1
and 2). as well as on affect (experiment 3). under conditions of involvement that may be described as 'very
low'.
In the first two experiments, using 3 selective attention procedure, subjects were incidentally
exposed to advertising messages. Evidence that these exposures had an impact on memory was sought
using two types of memory measures - direct and indirect measures. believed to tap explicit and implicit
memory respectively. The results suggest that. despite minimum amounts of attention devoted to
processing the advertising messages, there was implicit memory for the advertised brand names as
evidenced by an increased propensity to respond with those brand names on the indirect measures of brand
name generation (experiment i ) , consideration set formation and choice (experiment 2). A concomitant
finding was poor explicit memory for the same brand names. These results were, however, moderated by
repetition status (1 exposure vs. 3 exposures) of the brand names at exposure.
The third experiment examined. within the context of the mere exposure effect, the effects of
perceptual tluency and conceptual fluency on liking for words encountered under low involvement
conditions. In particular. the proposition that both types of tluency can give rise to the mere exposure effect
was tested. The results support that proposition and also indicate that perceptual and conceptual tluency
operated independentiy in generating the observed effects. Moreover. the effects on recognition bore
similarity to that found for affect. Repetition of the wurds was also found to have a moderating influence.
ACKNOWLEDGMENTS
My heartfelt thanks to Carolyn Yoon. for taking a chance with me and for being a friend and a
mentor since the first day. I would like to also thank Fred Feinberg, for the invaluable advice he has
dispensed over the years and for bringing a little (or perhaps not so little) Grazzi to my life. A note of
thanks also goes to Scott Hawkins. whose encouraging words I have always appreciated. and to my fellow
ductoral studentddinner buddies/X-blast rivals. Edward. Seh-Woong and Kwanho, for beins a phone call
away whenever help is needed. I would like to especially thank Andy Mitchell. for his patience and his
guidance and most importantly. for never allowing me to think that I was not up to the task.
To my parents, Irene and Roger. and my parents-in-law, Gloria and Chandra, my deepest gratitude
tor their unwavering love and support and for taking care of me when I was too busy to take care of myself.
And if my cats. Winnie and Tigger. could read thls. thanks to them too, for taking away the blues just
simply by being their puny selves. Last but not least. I am grateful to Mucus for the things he did for me,
which would rival the length of this dissertation, if enumerated.
All of these people have helped make the destination worth the journey.
iii
TABLE OF CONTENTS
Chapter 1
I. 1. Introduction
1.2. Low vs. high involvment processing
1.3. Low involvement processing in consumer behaviour
1 .A. Low involvement processing: Effects on memory
1.5. Low involvement processing: Effects on affwt
1.6. Synopses of experiments 1 -3
Chapter 2
2. I . Implicit Memory: A bit of history
2.1.1. Korsakoff
2.1.2. Claparkde
2.1.3. Janet and Freud
2.1.4. Ebbinghaus
2.1.5. W d n g t o n and Weiskrantz
2.2. Explicit vs. Implicit Memory
2.3. Attention and Memory
Chapter 3: Experiment 1
3.1. Design
3.3. Subjects
3.3. Materials and Procedure
3.4. Dependent Measures and Manipulation checks
3.5. Results
3.6. Discussion
Chapter 4: Experiment 2
4.1. Design
4.2. Subjects
4.3. Materials and procedure
4.4. Dependent measures and Manipulation checks
4.5. Results
3.6. Discussion
Chapter 5
5.1. The mere exposure effect
5.2. Theoretical accounts of the mere exposure effect
5.2.1. Non-specific activation hypothesis
5.2.2. Perceptual tluencylAttributional model
5.2.3. Discrepancy-attribution account
5.3. Perceptual vs. conceptual fluency
Chapter 6: Experiment 3
6.1. Objectives of study
6.2. Subjects and Design
6.3. Stimuli
6.4. Dependent measures
6.5. Procedure
6.6. Results
6.6.1. Affect
6.6.2. Recognition
6.7. Discussion
Chapter 7: Conclusion
LIST OF FIGURES
Experiment L: Effect of attention and repetition on indirect memory test
Experiment L: Effect of attention and repetition on direct memory test
Experiment 2: Effect of attention and repetillon on consideration sets
Experiment 2: Effect of attention and repetition on choice
Experiment 2: Effrct ohttention and repletion on cued recall
Experiment 3: Possible relationships between conceptual tluency and perceptual tluency
Experiment 3: Effect of perceptual tluency and exposure level on affect when affect task preceded recognition task
Experiment 3: Effect of perceptual and conceptual fluency on affect when affect task preceded recognition task
Experiment 3: Effect of perceptual tluency and exposure level on recognition when affect task preceded recognition task
Experiment 3: Effect of perceptual fluency and conceptual tluency on recognition when affect task preceded recognition task
Experiment 3: Effect of perceptual fluency and exposure level on recognition when recognition task preceded affect task
Experiment 3: Effect of perceptual and conceptual fluency on recognition when recognition task preceded affect task
LIST OF TABLES
I . Experiment 3: Effect of perceptual fluency. conceptual fluency and exposure level on affect when affect task preceded recognition task
2. Experiment 3: Effect of perceptual fluency. conceptual fluency and exposure level on recognition when affect task preceded recognition task
3. Experiment 3: Effect of perceptual fluency. conceptual fluency and exposure level on recognition when recognition task preceded affect task
LIST OF APPENDICES
A. Experiment 1 : Critical brand names and slogans
B. Experiment 2: Critical brand names and slogans
C. Experiment 1: Visually-presented text passage
D. Experiment 2: Visually-presented text passage
E. Experiment 3: Homographs. study and test sentences
F. Experiment 3: Instructions to subjects at judgement phase
vli
Chapter One
1.1 Introduction
Consumers these days are bombarded by a barrage of advertising appeals. With the addition of the
internet to the advertiser's repertoire of media possibilities, consumers are greeted with more ads at every
turn than ever before. One consequence of this deluge of advertising messages is that they nave become so
much a pan of the landscape that a good portion of them are paid little attention or go unnoticed altogether.
One of the earliest papers to discuss the fallout dt' this advenising clutter is Krugman ( 1965).
Krugman made the observation that the psrvdsive and repetitive nature of ridvertislng has made the learning
d a d s akm tcl the learnme of tht: unimportant and nonsensical. More importantly, he noted that "much of the
Impact of trlevisian advertising is in the form of learnin2 wtthout involvement" and went on to suggest that
the mechanism by which persuasion occurs when invotvernent is low is significantly different from when
~nvolvement IS high. These comments proved to be visionary in a sense, for it foretold some of the
arguments that would be put forth by psychologists in accounting for the n w e exposure effect - the finding
that mere repetition of a stimulus enhances liking for it. Moreover, his comments underscored not only the
need to better understand the psychological processes that underlie low involvement learning and
pzrsuston. but to also properly distinguish between low and high involvement.
1.2 Low involvement vs. high involvement processing
How then does low involvement differ from high involvement? Krugman distinguishes bctween the
tho in terms of the degree of personal incohement or the number of thoughts attempting to relate the
stimuli to oneself. Houston and Rothschild ( 1977) defined tnvolvernent in terms of the extensiveness of
processing that occurs, while Petty, Cacioppo rind Schumann ( 1983) characterised involvement according to
which the target message or iswe is personally relevant. Mitchell ( 1983) conceptualized involvement as a
motivational state having two dimensions, intensky and direction. Intensity refers to the arncunt of attention
devoted tu prwesing the stimuli whereas direction refers to where that attention IS focused. which may be
determined by the processing goal of the individual. It was. however, Greenwald and Leavitt ( 1984) who
provided an overarching definition in a levels of involvement Framework that detailed four distinct leveIs of
involvement.
Greenwald and Leavitt ( 1584) named the four levels of invoIvement preanention, focal attentron.
conlprehertsion and ekzbornrion. in order of increasing involvement. The levels of involvement differ with
respect to the amount of anentianal capacity devoted to processing the stimuli and the cancornitant
dwibility of memo1 y lor the stimuIi, with higher levzls of involvement being associated with greater
atttntronal capaclty and durability. The preattention level involves minimal attentional capacity and
processing is consequently limited to the sensory or feature analysis level. At the focal attention level. a
modest amount of attentional capacity is allocated; enough to allow for identification of the stimulus at the
perceptual or semantic level. Yet greater attentional capacity is allocated at the comprehension level. making
i t possible to process at the syntactic level and to have propositional representation of the message in
memory. Finally. at the elaboration level. where the most amount of attentional capacity is required. not
only does one process syntauti;ally. but also attempts to relate message content to existing knowledge.
Consistent with these propositions is the expectation that memory for the stimulus. as measured on
traditional tests of recall or recognition, is better the higher the level of involvement.
1.3 Low involvement in consumer behaviour
Consumer researchers have a much better understanding of the processes that lead to learning and
persuasion when involvement is at least moderately high. With the exception of several papers that probe
cognitive processes that accompany incidental or peripheral processing (e.g. lanisrewski. 1993; Shapiro.
Maclnnis & Heckler. 1996; Obermiller. 1985). much research in marketing is focused on investigating the
effects of having processed stimuli in 3 delibrr~te and directed manner. as one might expert at higher levels
of inwlvement. In addition, the nature of experimental work is such that the experimental context itself
induces a m d r af processing charxter~st~c of ii higher lcvel of involvement than might be expected in a
more realistic setting. Hence. a subject brought into the experimental laboratory and presented with ads as
experimental stimuli may be reasonably expected to devote greater attentional resources to the ads and
process them more carefully thm if Jhe were to encounter the ads. say. while watching television.
Published studies that examine exposure effects at truly low involvement levels. such as at the preattention
or focal attention levels in Greenwald and Leavitt's ( L984) framework, remain few and far between in the
marketing literature.
The study by Petty. Cacioppo and Schumann ( 19831 often comes to mmd when one thinks of low
involvement persuasion, but even subjects in the low involvement condition of that study were permitted to
take as much time as they needed to peruse the advert~sing stimuli with their full attention. It differed From
the high involvement condition only in the amount of personal relevance or consequence the ads had for the
subjects and may as such be better understood as the lower involvement condition and not as one consuming
particularlv low amounts of attentional resources. (See Park & Hastlk ,1994; Sawyer & Howard. 199 1:
Miniard et 31.. 199 1; Maheswaran & Meyers-Levy, 1990: lohar, 1995; for similar manipulations of
involvement. \ involvement has also been manipulated by varying the processing goal given to subjects. Low
involvement subjects may be directed to process in such a way that minimizes the amount of elaborative
processing of the stimuli. often by calling their attention to some non-critical aspect of the stimuli (Mitchell.
1883; MacKenzir & Spreng, 1991: Hawkins & Huch. 1992). Others use a cambination of manipdating
personal relevance and varying the processing goal (Sengupta. G d s t e i n & Boninger, 1997; MacInnis &
Park, 199 1) These conditions are still fairly involving and if cast in terms of Greenwald and Leavitt's ( 1984)
framework. correspond most closely to the second highest involvement level. that of comprehension.
This dissertation is concerned with the lower levels of involvement, induced by limiting subjects'
attention to the stimuli. In the sections that follow, effects on memory and affect are discussed in the context
of such low levels of involvement.
1.4 Low involvement processing: Effects on memory
Given the low attention level associated with low involvement processing. what effects on memory
migh: one expect to find? It would not be surprising to find that memory for stimuli that are paid little
attention would be poor. This last stalement should. however, be qualified by the type of memory measure
~nvolved. A large literature has accumulated in cognitive psychology. providing compelIin evidence that
the 'absence of memory' as determined by one type of measure need not be corroborated by results found on
another type of measure. These two broad classes or' measures have come to be known as direct and itidirea
memory measures. tapping explicit and implicit memory respectively. Implicit memory. as opposed to
explicit memory. may be loosely defined as that aspect of our memory that a not associated with awareness
or intentionality. but which exerts an mfluence on our behavtour that is quite beyond our conscious control.
The one important feature that distinguishes indirect from direct memory measures is the lack of reference
to a prior learning episode or exposure phase. Subjects. rather than being told to recollect specific
information from an earlier phase. are merely instructed to complete the measure to the best of their ability.
Often. effort is made to ensure that subjects are unaware of the connection between the indirect measure and
the exposure phase. Hence. while direct measures encompass the traditional recall and recognition tests and
require one to think back to some point in time in order to form a response, indirect measures make no such
reference and incIude tests such as word-stem or word-fragment completion. perceptual identification, and
even general knowledge questions. Enhanced proficiency in completing an indirect measure following
exposure to tmpt stimuli is evidence for i t s cffecra memory. In comparing the two tvpes of measures.
interesting disscciatims have been fnllnd. with \ari:bles having an effect on one type ot measure having no
effect or the opposite eftect on the other (see Roediger & McDermott ( 1993) for crviev~. As well. an
apparent absence of memory for stimulus exposure u measured on a direct test is sometimes refuted by
better than control or above chance level performance on an indirect test. This points to the possibility that
in low involvement processing where attention levels are low and explicit memory poor. implicit memory
may nonetheless be evident on an indirect test.
1.5 Low involvement processing: Effects on affect
The previous section provided a brief discussion of the possible effects of stimulus exposure on
memory under low involvement processing conditions. We move on now to talk about effects on affect, a
discussion that is relevant to the third experiment of thrs dissertation where the dependent measure of
interest is affect.
Studies on the mere exposure effect provide the background Immure tbr assessing the effects of
repeated exposure on affect. of stimuli processed under low involvement conditions. a jonc pioneered
research in this area and the term 'mere exposure' first appeared in his 1968 seminal paper. In it, he
conducted a series of experi ments, presenting nonsense words (experiment 1 ), Chinese ideographs
(experiment 2) and faces (experiment 3) at various repetition levels and showed that liking increased with
the number of exposures. This represents the first evidence that stimuli encountered under low involvement
conditions. in particular at the level of focal attention according to the GreenwaId and Leavitt ( 1984)
framework, can have an effect on affect. An abundance of similar studies foltowed, providing corroborating
evidence for this effect that results from mere exposure to the stimulus. The effect has also been obtained at
the preattention level of involvement. The first among these studies was one conducted by Wilson ( 1979).
Using a dichotic listening procedure, Wilson ( 1979) presented subjects with melodies in one ear while
having them shadow a message played in the other ear. Despite an apparent inability to remember the
melodies on a later recognition test, subjects displayed a preference for these melodies over control ones.
Other studies using visual stimuli presented either subliminally or under incidental exposure conditions.
have found similar results (Kunst-Wilson & Zijonc. 1980; Mandler et al., 1987; laniszewski. 1993;
Burnstein & D'Agostino. 1999,). There IS therefore no tack of evidence of the impact of stimuli encountered
under low involvement conditions.
1.6 Synopses of experiments
In Lynch and SrulI (1982). it was noted that "One of the most imponant lessons of cognitive
psychology is that most of the memory and attentional factors that affect our judgement are simply
unavailable to consciousness . . . It would seem, then. that consumer research must not limit its focus to
conscious decision strategies, but also explore the unconscious and/or involuntary aspects of information
processing." This dissertation in a sense heeds that call by looking at the effects of stimulus exposure that
may not be amenable to conscious introspection or reporting.
The first two experiments of tnis dissertation look at what happens when ads are encountered whilst
one is engaged by some other task or stimuli, such that the ads are outside the focus of one's attention.
[According to Greenwaid and Leavitt's ( 1984) framework. a processing condition like this falls under the
preclttentmn level uf the involvment. I ?h= fwus of attention is manipulated by presenting subjects with two
sources of stimuli at the same time using a selective attention paradigm. One goup of subjects was asked to
perform an attention-demanding text-reading task while brand names were played to them through ear
phones. A second group of subjects was asked to do the reverse: attend to the brand names while ignoring
the text. In both of these groups, approximately one third of subjects received one exposure to the target
brand names, while another third received three exposures. The rest of the subjects were not exposed to the
brand names and represented the control group. Memory for the target brand names was assessed using both
a direct and an indirect measure. The most interesting results pertain to the condition where attention was
directed away from the brand names to the reading task. Despite an inability to report the brand names as
determined on the direct measure, there was nonetheless evidence that the exposure did, in fact, have an
effect on memory. Performance level on the indirect measure for subjects who were exposed to the brand
names three times was significantty higher than for control subjects who were not exposed to the target
brand names at all.
The third experiment reported in this dissenatron is a mere exposure study that examines the possible
causes of the phenomenon. Perceptual tluency has been proposed in the literature as a potential explanation
for the mere exposure effect. This experiment provides further evidence for the viabdity of that hypothesis
as we11 as tests conceptuat fluency as yet acother possible cause. Both perceptual and conceptual fluency
were manipulated in the same experiment, making it p~ssible to examine the relationship between the two in
producing the mere exposure effect. The critical stimuli consisted of homograph words presented in the
context of sentences. In the exposure phase, subjects were required to read these sentences aloud, each of
which were presented but brietly on a computer screen. [Given the short exposure time, the level of
involvement here corresponds most closely to the focal attention level in Greenwald and Leavitt's ( 1984)
framework.] In the judgement phase, subjects were presented with a new set of sentences, with a target word
embedded in each. Each target word was such that it was 1) identical in form and meaning, 2) identical in
form but different in meaning, 3) different in form but having the same meaning, or 3) different in form and
meaning, to a word seen earlier in the exposure phase. For each target word, subjects responded on a
recognition test and an affective preference measure. Results indicate that subjects were more likely to
express preference for those words that were elther identicai in form or meaning to words that they were
previously exposed to. This implicates both perceptual and conceptual fluency as factors driving the mere
exposure effect. Moreover. the results pointed to independent operation of the two sources of fluency.
2 Chapter Two
2.1 Implicit memory: A little bit of history
The term implicit memory was coined by Graf and Schacter ( 1985). with reference to the
phenomenon of remembering without awareness. Althcugh implicit memory research has been a hotbed of
research activity only in recent years. the awareness of such unawareness is not a discovery that modem day
scientists can lay proprietary claim to. The idea that memorial traces may manifest itself in words and action
w~thout one having conscious recollection of the acquisition event has been flirted with. speculated upon
and described in excruciating detail by philosophers and scientists before our time. The following provides a
historical perspective and insight into the precursors of implicit memory.
2.1.1 Korsakoff
In two papers published in 1889. Korsakoff provided rich accounts of the implicit memory
phenomenon. which still makes for an intriguing read today. The first of these was translated to English in
1955, while translation of the second Bas only accomplished very recently by Banks and Karam ( 1996). In
the latter, Korsakoff provided rather remarkable accounts of how his patients who. suffering from a
condition now known as Korsakoff s amnesia. exhibited memory disorders that proved fascinating. The
central feature of this syndrome is the inability to recollect new information. in particular events occuning
subsequent to the onset of the disease, which is in sharp contrast to the normal memory they have for events
pre-dating the condition. The extremity of the condition is reflected in the striking absence of memory for
actions that had taken place literally moments sgu. with the result that patients often engage in repetitive
behaviour simply because they have no recollection of ha\ ing done something they had just done. The more
relevant of the observations. in light of our present interest. is the finding that white there is no conscious
memory of recent cbents. memorial traces of them nonetheless persist to intluence behaviour in more subtle
ways. This is best illuminated by the example of one patlent who at each meeting with Korsakoff. guessed
that he was a physician. but stubbornly opposed the suggestion that they had met before. Then there is the
case nf the patient treated with electro-shock. who despite having no recollection of having undergone such
therapy, would guess merely from looking at the case containing the shock apparatus that Korsakoff was
there to administer shocks to him. Interestingly. Korsakoff reported a resurgence of these 'lost' memories as
his patients embarked on the road to recovery, as evidenced by accounts of patients gradually recounting
more and more events that they supposedly had no memory of during their amnesic period.
Korsakoff s thwretica1 account of the phenomenon which we now call implicit memory postulates
firstly that each event lays down a memorial mace and secondly that successful recovery of that tnce
requires the existence of associative connections leading to the target representations. A trace strength
exceeding threshold levels 1s a necessary cond~tion for conscious recollection. This should, however, not be
interpreted as a stmple trace strength theory of memory since conscious remembering is also contingent on
the presence of effective associative networks. such that a strong trace by itself is by no means sufficient for
awareness (Banks, 1996).
About two decades subsequent to Korsakoff s publications. Clap~ede ( 19 I 1, 195 1 ) provided
accounts of amnesic patients with conditions remarkably similar to those reported by Korsakoff. Among
these observations is the anecdotal account of the amnesic woman on whom Claparede tried a little
experiment. He hid a pin between his fingers and pricked her hand with it. As is typical of patients suffering
from Korsakoff s syndrome. the pain was all too quickly forgotten and no impression of it remained just
shortly afterward. Nevertheless. each time Clapartde placed his hand in the vicinity of hers. she would. as a
retlex, withdraw her hand. When the reason for the prompt action was pursued. she nonetheless was not able
to attribute it to s recollection of the preceding pricking occasion, but instead gave responses such as "Isn't
it allowed to w~thdrtrw one's hand'?" and "Perhaps there is a pin hidden in your hand". When questioned
about how she came to fathom such an idea. her replies were that "It is an idea that just came into my head"
and "Somrtimcs pins are hidden in hands" (Parkin, 1993).
Claparede's theoretical explanation of accounts such as the above, like Korsakoff s. appeals to the
existence of associative networks. Mowver. Claparede postulates two associative networks; one connects
representations to each other (peripheral associations). while the other connects representations to the self
(egocentric associations). Implicit expressions of memory are then a case of intact peripheral associations
but severed egocentric associations (Banks. 1996).
2.1.3 Janet and Freud
Janet and Fred were primarily interested in the form of implicit memory manifested in hysterical
amnesia. The condition refers to the absence of conscious memory for a traumatic event, although
impressions ot it nonetheless show through in thoughts and action. Janet concluded from his case studies of
such patients that t w factors were at play. Firstly. in hysterical amnesia. the patient relinquishes control of
voluntary and cimcious activation of memorm and sec.ondly, these same memories may still be evoked
automatically and in an untimely fashion. His proposed explanation appealed to the notion of an "inabiIity to
synthesize memories ir,tr, the personal consciousness" t Janet, 1904). Freud, like Janet. also believed that
hysterical symptoms were an expression of suppressed memories that do not enter the conscious state of
mind and further emphasized the potency of unconscious memories in influencing behaviour (cf. Schacter.
1987;.
2.1.4 Ebbinghaus
Ebbinghaus is well-known for his savings paradigm which neatly demonstrates economies in
learning when one relearns material for a second time (or n' time. for that matter). In the study phase of this
method, one learns material usually in the form of nonsense syllables and this is followed by a test phase
requiring the recall of the learned words. The number of trials. or alternatively. the time taken to master
perfect recall is noted for purpose of comparison when on subsequent occasions. the entire process is
repeated. Savings is measured by the difference in the number of trials or amount of time required to attain
equivalent levels of performance from one learniag occasion to another and the general finding is a positive
difference that occurs independent of whether one is capable of recollec~ing the original learning experience.
In present day terms, the savlnp method qualities 3s a test of ~mplicir memory because explicit recollection
of a prior learning episode is of no concern Ebbinghaus ( 1885. 1964) distinguished between types of
memory: voluntary recollection. which corresponds most closely to explicit memory in modern day
terminology; involuntary recollection, which refers to the spontaneous 'bringing to mind' of previous
experiences. More interestingly, Ebbinghaus conjectured that there may be a third type of memory -
memory that we're not aware of possessing but which nonetheless has an impact on our behaviour.
2.1.5 Warrington and Weiskrantz
Warrington and Weiskranu can rightly be credited for sparking the enthusiasm over implicit
memory among memory researchers by providing the first convincing empirical evidence for yet another
memory type in their work with amnesics. Warrington and Weiskrantz (1970) documented startling findings
of dissociation in performance on direct versus indirect tests. Normal and amnesic subjects were
administered two direct tests ( free recall, recognition) and two indirect tests (word fragment identification,
word-stem completion) to assess retention of words presented earlier in a to-be-remembered list. Normal
subjects outperformed amnesic subjects on the direct tests: a result that's hardly surprismg given the very
nature of amnesia. However. it is this outcome in combination with the tinding of statistically equivalent
performance by both groups on the indirect tests that proved to be enlightening. This virtually demolished
the then prevailing view of amnesia as arising from a failure to transfer information from short-term
memory to long-term memory. since the expenmental findings showed that there was indeed retention of the
learned material even if it was only apparent in the implicit memory tests (Roediger, 1990).
2.2 Implicit vs. Explicit Memory
Warrington and Weiskrantz's (19'70) landmark finding generated an impressive may of new
findings. theoretical perspectives. and procedures for investigating Ule relation between implicit and explicit
forms of memory. The distinction bet ween implicit and explicit memory rests on the intent~onality of the act
of remembering. Whereas explicit mcmory is manifested when subjects consider their own past experiences
in order to generate an appropriate response. implicit memory presents itself as a facilitation in performance
as 3 result of prior experience without the conscious intention to draw on past experience. The most often
used measures of explicit memory, referred to 3s direct measures, are that of rxal t and recognition. For
implicit memory, psychologists have developed a large set c~f mdirect measures ranging from word-fragment
completion tests, lexical decision tasks. possible/irnplssible object decision tasks, to general knowledge
questions (see Roediger & McDermort. 1993). The luge set of measures owes, in part, to the fact that
implicit memory, as psychologists have defined it, has become a blanket term for any facilitation in
performance by past experiences without intention to rely on those experiences. Hence, there are potentially
many measures that qualify as tapping implicit memory. In consumer research, generation of brand names in
response to category cues can be deemed an indirect memory measure so long as no reference is made to
any past episodes or events. Of greater practical relevance to marketers are choice and consideration set
measures (where subjects indicate the brands they would choose or consider for choice given a product
category) that can, in the absence of reference to prior encoding conditions, also be considered measures
representing implicit memory.
Many interesting dissociations between implicit and explicit memory performance have been
uncotered in the several decades of implicit memory research, For instance. levels of processing (Craik &
Tulving, 1975) msnipulations have beer, sh~wn to affect performance on duect tests but nct indirect tests
(Jawby & Dallas. 1981; Graf & Mandler, 1984). Changes in presentation format of stimuli. such ~s from
auditory at study to visual at test, h a w been shwn to adversely affect implicit memory performance. whde
having little or no impact on explicit memory performance (Jacoby & Dallas, 1981; Roediger & Blaxton,
1987). Age differences have also been found to dissociate the two classes of memory measures, with
younger subjects outperforming older subjects on direct tests but not on indirect tests (e.g. Light, Singh &
Capps, 1986; Light, 1987). FinalIy, more evidence of dissociations between implicit and explicit memory
has been reported in cognitive neuroscience, among patients with frontal lobe damage, with observations of
facilitation in performance on indirect tests in the absence of any performance improvements on direct tasks
(for recent reviews see Schacter, Chiu & Ochsner, 1993; Squire, Knowlton & Musen, 1993).
Whereas the evidence for dissociations is clear. the mechanism(s) underlying these dissociations has
been the subject of mtich debate. One tenable hypothesis is that these dissociations are the result of different
reuieval modes. with re&eval on direct measures being intentional while retrieval'on indirect measures
being incidental. Far this to be a plxusible explanation of dissociations. Schacter, Bowen and Booker (1989)
have noted the necessity to hotd all overt conditions (~ncluding test cues) at study and test constant, except
for retrieval instructions given at test. Only then can differences in implicit and explicit memory
performance be attributed to intentionaiity of retrieval. This necessary condition. called the retrieval
inrentionuiiry criteriort. has become a useful rule-of-thumb of sorts. for researchers relying on the
consciousness or intentionality o f retrieval argument as 3n exptanation for observed dissociations.
2.3 Attention and Memory
As discussed in the previous section, variables such as age, levels of processing and presentation
modality have been shown to have very dissimilar effects on direct and indirect memory measures. Yet
another one of these variables is that of attention. With attention. however, conflicting results have been
obtained and inconsistencies have yet to be resolved. Early studies investigating the role of attention in
memory performance demonstrated that dividing attention at study (wherein subjects are required to
perform two concurrent tasks) has no effect on implicit memory performance, while at the same time,
having large adverse effects on explicit memory performance (Parkin & Russo. 1990; Parkin, Reid & Russo.
1990). More recent studies (Woiters & Prinsen. 1997; Mulligan, 1998; Mulligan & Hanrnan, 1996;
Mulligan. 1997). however. have documented results that do not concur with these simple conclusions and
suggest the presence of moderating factors that were previously unaccounted for. One explanation invokes
the Transfer Appropriate Processing (TAP) framework (Morris, Branstbrd & Franks. 1977) and proposes
that the type of processing that takes place at test may reconcile inconsistencies between studies that found
an effect of attention on implicit memory and those that did not.
The principle espoused by the TAP framework. briefly stated, says that performance on a memory
test is enhanced to the extent that the cognitive processes engaged at study are re-engaged at test. The types
of cognitive processes. according to TAP, can be broadly classified into ones that are perceptually-driven
and ones that are conceptually-driven (e.g. Blaxton. 1989). For instance, if the type of processing that occurs
at study is predominantly perceptual (analysis of surface features) as opposed to conceptual (analysis of
meaning), performance on a later memory test will be facilitated insofar as the type of processing engaged in
is also primarily perceptual. The nature of processing that occurs at test is thought to be induced by the
demands of the memory test. Hence, in perceptually-driven memory tests such as word-fragment completion
or degraded stimulus identification. one has to rely on the perceptual attributes of the test cues to generate an
appropriate response. On conceptually-driven rests such ss category exemplar generation or general
knowledge questions. apprupnste responses are only conceptually related to the test cbes, thus cornpeiling
one to engage In more conceptual processing.
Hence. rhc TAP framew~rk ~ffers a poteli~d ex?!anation for the discrepancy in findings. Dividing
attention at study constlains che amount of cor~crptual ?rr)cessing that is possibk and performance on a later
itmceptually-dri ven memory test will be poor as a consequence of the mismatch in processes. Perceptual
processing. on the other hand, is presumed to be less resource-demanding, such that it is affected little, if at
all, by divisions of attention. According to this Iine of reasoning then, a reconciliation of the discrepant
findings reported in the literature may be obtained by appealing to the difference between these studies in
type of memory test administered. And indeed, the evidence that speaks to this issue appears to partially
support the contention: Studies in which perceptually-driven implicit tests are used have found little effect
of attention on memory pertommce (Muiligan & Hartman. 1996; Parkin & Russo, 1990; Szyrnanski &
MacLeod, 1996). The evidence, however. continues to be murky when it comes to conceptually-driven
implicit tests. Both support for (Wood & Cowan, 1995; Mulligan. 1997) and against (Koriat & Feuerstein,
1976; Isingrini. 1995) an effect of attention on memory performance on such tests have been found.
Further attempts at resolving this inconsistency have argued that TAP continues to be a useful
framework in predicting the effects of dividing attention. but that inconsistencies arise as a result of
differing operationalizations that lead to attention being divided to varying degrees in the different studies
(biulligan, 1997). When attention at study is but mildly divided (such as by imposing a concurrent task
that's not too resource-demanding), it is argued that conceptua1 processing is impaired, but not to the extent
that it is altogether absent. Hence in studies employing conceptually-driven impltcit tests and in which
attention is not strongly divided, it is unsurprising to find that implicit memory performance is not different
between full attention and divided attention conditions. Only when attention is strongly divided and
conceptual processing consequently c~mailed do d~fferences become apparent, as Wolters and Prinsen
( 1997) and Mulligan 11997) have shown.
This line of reasoning also implies that in studies of selective attention, wherein subjects are told to
focus on some task whde ignoring concurrently presented stimuli. there ought not to be implicit memory for
the ignored information on a conceptually-driven test. Assuming that the primary task is sufficiently
engaging such that subjects are not able to simuttaneously perform both tasks without performance on the
primary task suffering appreciably. there can be very little conceptual processing of the to-be-ignored
information. What processing goes on will likely be sontined to merely perceptual processing (see Craik.
1983: Cruk & Byrd. 1981). It follows then that the absence of conceptual processing or the little of it that
takes place means that there will be little evidence of memory on a conceptuctll y-driven indirect test.
To summarize, the TAP framework predicts that in divided attention studies where memory
performance is assessed on a conceptually-driven indirect test of memory. an effect of attention will be
observed only when attention is divided sufficiently strongly. In the case of selective attention studies, not
only will there be an effect of attention, but implicit memory for critical information will not be evident
when that information receives nonfocal attention.
A review of the marketing and psychology literatures reveals. however, that support for the second
prediction is equivocal. Of the few published studies that look at selective attention and memory
performance (both explicit and implicit), there are both studies that have found support for the presence of
implicit memuiy (Shapiro, MacInnis & Heckler, 1997; P h m & Vanhuelle, 1997; Eich, 1984) and ones that
provide evidence to the contrary (Wood & Cowan. 1995; Wood. Stadler & Cowan. 1997). Of particular
significance and relevance is the study by Eich ( L384), which presents one of the earliest pieces of evidence
of memory for information that is ignored and one that has generated some controversy as to the robustness
of the findings. I t s importance warrants a more detailed discussion of the experiment conducted and its
read ts.
Eich ( 1984) demonstrated that information presented in a dichotic listening procedure affected
implicit memory despite a c1ea.r absence of recognition for the same information. Subjects were given a task
of shadowing an audio version of an essay played to one ear. while word pairs were played to the other.
Some of these word pairs were such that the first would be compatible with a less Frequently occurring
homophone word rather than the more commonly used possibility. One example would be the word pair tari
-fare, in whichfare instead of fair would be the more likely interpretation given the word taxi. Subjects
were subsequently administered two tests, a recognition test (direct) and a spelling test (indirect). On the
latter of these. when subjects were asked to spell the homophone words. it was found that they were
significantly more likely than a baseline control group to write down the less frequently occurring word
encountered in the listening task than control words not presented. By contrast. recognition of these words
was not above chance level.
That the results reported by Eich 1 1981) were engendered by subjects' switching of attention between
the mended and to-be-ignored channels \vils raised as a posstbility by critics (Wood & Cowan. 1995; Wood,
S tadler & Cowan, 1997). In support of this contention, it has been pointed out that subjects shadowed prose
presented at a rather slow rate (85 words per minute) and that words presented in the to-be-ignored channel
were presented at the rate of one-second per word with a two-second interval between word-pairs. This
makes it likely that subjects had periodically shifted attention to the to-be-ignored channel. Moreover. as
mentioned earlier. to make a convincing argument that dissociations between direct and indirect tests are the
result of differences in the mode of retrieval and not merely to task differences. the retrieval intentionality
criterion (Schacter, Bowers & Bookers, 1989) holds that a11 cues in indirect and direct test conditions should
be held constant except for the instructions given n test. The Eich (1984) study, unfortunately, does not
satisfy this criterion and provides but tenuous testimony to the notion that the attentional resources can
dissociate performance on direct and indirect tests.
Wood and Cowan ( t 995. Experiment 3 & 4). in contrast, using a similar dichotic listening procedure
and conceptual1 y-driven memory tests that meet the retrieval intentionality criterion. found little evidence of
implicit memory. Subjects wore headphones through which two different prose passages were played in the
right and left charrnels. The task requmd the shadowing of input from thc right channel while ignoring input
trom the left. At the wnclusion of the listening task direct and indirect tests, each consisting of 14 phrases.
were administered. One phrase from each pair was extracted from the passage played in the ignored channel
and the other was ohtamed from unheard passages from the same book. On the direct test. participants were
asked to choose the phrase in each pair that they had hexd in the irrelevant channel. For the indirect test,
subjects were informed that half of the phrases came from a pubIished story by a well-known author and the
task was to identify whicn phrase of each pair was such. In comparing these results to those obtained from
control p u p s it was clear that both explicit and implicit memay for information presented ir, the to-be-
ignored channeI were absent. However. when the same stimuli were attended to, direct and indirect test
scores were significantly above those of control groups.
One may argue that the nature of the stimuli may have contributed to the failure to find implicit
memory in the ignore condition. It is conceivable that a certain minimal level of semantic processing has to
have occurred at study in order for implicit memory for phrases to be found, a condition that was not
achievable given the little attentional resources devoted to the information. Nevertheless. the finding of no
implicit memory under these conditions is not conclusive evidence that it cannot be found. It remains
possible that with smaller conceptual units. such as simple words or brand names instead of whole phrases.
implicit memory on a conceptually-driven test will be evident even when the critical information is but
minimally processed at exposure. Moreover. multiple repetitions of the information that receives only
nonfocal attention may reveal implicit memory effects.
The purpose of the present study then is to test the above proposition in a marketing relevant context.
We attempt to simulate. in an experimental context, situations of ad encounter commonly experienced by
consumers - situations in which ads are devoted little attention and are consequently minimally processed.
Specifically. we investigate whether advertising slogans that are heard while one is engaged in a resource-
intensive primary task may make the featured brands come to mind more readily on a later brand name
generation task. This allows us to see if there can be implicit memory on a conceptually-driven test when
smaller conceptual units are used and also offers the opportunity to put the predictions of the TAP model to
test. We 311~9 examine of the role of advertising repetition as a potential moderator of the relationship
between memory and attention.
3 Chapter Three: Experiment 1
3.1 Design
The experimental design was a 3 (number of exposures) x 2 (direction of attention) between-group
factorial. The first of these factors involved nor exposing subjects to the brand names (control group) to
establish baselines for comparison, or exposing subjects to one or three exposures (experimental groups).
These three [eveis of the first factor are also referred to elsewhere in this paper as control. no repetition and
repetition. The second factor entailed directing subjects to either focus attention on the brand names or to
Ignore them while attending to some other concurrently presented information.
3.2 Subjects
A total of 188 undergraduate students participated in the experiment for cred~t in an introductory
marketing course at a large northeastern university. The experiment was conducted on dn individual basis,
with each session lasting a half-hour. Participants were randomly assigned to one of six conditions on
a n ~ v s l .
3.3 Materials and Procedure
At the start of the experiment, individuals were seated in front of a computer and asked to read a set
of instructions on tl printed sheet. Those subjects assigned to the attend condition were told that their task
was to listen to audio messages played to them through earphones. while at the same time ignoring text
flashed on the computer screen. As part of the cover story. they were told that the aim was to assess how
comprehension of audio information may be affected by distracting visual stjmuli. For that reason. they
were to keep their eyes on the computer screen whiie paying full attention to what they were hearing
through the earphones. Ti, reinforce the need to focus attention on the audio message, they were told to
expect a later test of comprehension.
Subjects in the ignore condition. on the other hand, read instructions that told them to read the text
presented on the computer screen, as if they were reading a regular print article. except that they were to
read i t out loud. To be sure that subjects understood what was expected of them and as a w m up, they
were, at this point. required to read a short passage of 122 words flashed one word at a time at a pace that
corresponded to that used in the main task. After completing this practice trial. they were then told that they
were required to perform a similar task. but with concurrent audio messages played to them through
earphones. The cover story informed them that the purpose was to assess how comprehension of text
passages may be affected when the environment is a noisy one. As with the attend condition. it was
necessary to provide incentive to focus attention on the text passage and this was accomplished by telling
subjects to expect a later test of comprehension of the text.
After it was clear that subjects understood a11 instructions, the main task of the experiment began. The
experimenter then started a software program called Vortex, used for the purpose of displaying text on the
computer screen. The program parsed text passages and displayed them one word at a time at a
predetermined rate and in a manner that was reactiw to punctuation. This software allowed us precise
control over the rate of presentation and hence the speed of reading. In addition. its sensitivity to
punctuation enabled reading in a more natural manner. The text passage used was a concatenation of two
advertising messages. one for the product Ambien (a sleep medication) and the other tbr the Bose Wave
Radio. taken from print magazines. The passage was 605 words long, presented at the rate of 165 words per
minute. (Please see Appendix C for text passage.) This rate of presentation was chosen based on pretest
results that indicated that it was demanding enough to prevent subjects in the ignore condition from shifting
attention to the audio messages. Subjects in the pretest were presented with the text passage at various
reading speeds with a concurrent audio message played to them through earphones. They were subsequently
probed for awareness of the content of the audio message and explicitly asked if they had at any point
shifted attention to the audio message.
Immediately after clicking on the button that started the Vortex program running, the
experimenter began playing an audio tape containing the critical ad slogans. A total of 16 mock ad slogans
(featuring real brand names) were played. with a 3-second interval between slogans and a 40-second period
of silence at the start df the tape. This period of silence served the same purpose as the practice reading trial
- to ensure that subjects in the ignore condition had warmed up and were comfortably focused an the
reading task before they were presented with the critical ad slogans. Subjects in the no repetition condition
heard 4 critical ad slogans just once and each of the 4 filler slogans were repeated three times for a total of
16 slogans heard. Conversely. subjects in the repetitton condition heard each of the 4 critical ad slogans
three times and the tiller slogans once. again for a toul of 16 slogans heard. The brand names appearing in
the critical d slogans were seiected based on a pretest..' (See Appendix A for the critical brand names and
slogans ilsed. )
Upon completion of the reading/listening task, subjects were administered the dependent measures
contained in a questionnaire booklet. At the end of the session, they were debriefed. thanked for their
participation. and dismissed.
3.4 Dependent Measures and Manipulation Checks
In the ignore conditions. in which subjects were required to read the text on the computer screen,
whenever there was a break in continuity of the reading or when words were slupped, a record of it was
made. The experimenter had a printout of the text that was read before her and a highlighter ready, so that as
subjects read the text. errors were noted by a quick highllgnting of the words that were missed. This served
. as an indication of reading fluency with dysfluencies suggesting attentional shifts.
For greater assurance that subjects in the ignore (or nonfocal) and attend (or focal) conditions had
indeed focused attention respectively on reading the text and listening to the audio, subjects were given ten
bur-alternative multiple-choice questions based on the text passage.
One indirect and one direct memory test were contained in the questionnaire booklet. In the case of
the former. the associated instructions were to respond with the first brand name that came to mind for each
of ten product catepnes. Subjects were discouraged from deliberating for too long. For the latter. subjects
were required to recall the brand names that they may have heard in the audio messages and were prompted
rvith ten category cue:. different from those used in the indirect memory test. For each subject. memory for
two of the four critical brand names was assessed on the ind~rect test. and memory for the ather two on the
direct test. Thus the product categories were divided into two sets of two categories each. ~ u c h that each
subject would see one set on the indirect rest and the second set on the direct test. Use of each set on either
the indirect or direct test was counterbalanced across subjects, as was the order of the indirect and direct
measures. In keeping with the retrieval intentionality criterion. the tests were designed such that the indirect
and direct tests were distinguished only by ~nstructions.
The order in which the measures were collected was as follows:
i . Multiple-choice questions
2. Indirect [Direct] measure
3. Filler task
4. Direct [Indirect J measure
5 . Test fcr hyporhesis awareness
6 Demographic information
The filler task was s~mply a wurd search puzzle that took approximately 5 minutes to complete. It
served the purpose of clearing short-term memory as subjects moved from one task to the next. Hypothesu
awareness was tested fur with a series of questions placed near the end of the booklet, followed by some
demographic questions about each pan~cipant.
3.5 Results
Twenty percent of the 188 subjects run were deemed to be either aware of the hypothesis or of the
connection between the indirect task and the exposure phase and were excluded from further analyses. To be
sure that systematic biases were not introduced by doing this, two sets of analyses were run on the critical
dependent measures. one that included these subjects and another that did not. The results of both analyses
were not different on key hypothesis tests, but nonetheless we restrict discussion, in the following sections.
to data obtained from subjects who were not aware of the study hypothesis.
Given that the order of the memory tests was counterbalanced across subjects. we first checked for an
order effect on subjects* probability of responding with the brand names featured in the slogans by
conducting a logistic analysis on data from the experimental groups. The results reveal a null effect of test
order on ather the indlroct or direct memory tests ( X ' S c I. ps > 3). Thus. in subsequent analyses. data were
collapsed across test order.
Manipulation Checks. For assurance that subjects had indeed focused attention on the textlaudio as
instructed. their score (out of a possible ten) dn the multiple-choice questions assessing text comprehension
was subjected to an ANOVA test. There was only a significant main effect of direction of attention
(F( 1,144)= 195.3s. p<.0001) with subjects in the ignore conditions scoring higher (M=7.25) than subjects in
the attend conditions (M=3.25).
For subjects in the ignore conditions who were required to read text, we obtained a measure of their
reading fluency by computing the total number of words that each subject skipped over. This measure was
subjected to an ANOVA to assess if fluency, and hence the likelihood of attentional shifts, might have
differed across repetition levels. The results indicate that subjects in the different repetition conditions
cannot be distinguished on the basis of the number of words missed in the reading task (&, w,,=19.3.
&,Cl,UM1=27S. &onml=2 1.4; F(2.75)= 1.34, p>0.2)).
Indirect Memorv Mesure. For each of the two criticai brand names on the indirect memory test.
wbjects recaved a score of I ~f they generated the cc.rrec t brand name and 0 other wise. Since each subject
was rested fi>r implicit memory for two brand nzmcs. a cpeatcd measures logistic analysis. with time as the
repeated factor. was run on the data.
The analpis revealed a significant main effect of direction of attention (x2( 1)=3.99. pc.05'). with
subjects in the attend conditions being more likely to respond with the critical brand names than subjects in
the ignore conditions. Main effect of repetition was also significant indicating that the likelihood of
responding with those brand names was not equivalent (x2C)=20.08. p<.OI) across groups that received
zero. one or three exposues to the brand names. The time factor was not significant (~'(1)=1.19. p>.Z).
demonstrating that whether a product category was encountered early or later on had no significant effect on
subjects' responses. In addition. none of the interactions involving time was significant (X2s<l. pn.5 ) . Of
greater interest was the fact that the significant main effects are qualified by a significan~ direction of
differences between experimental and control groups across the repetition condition point to the presence of
explicit memory.
For subjects in the attend conditions, there was a steady increase in the proportion of subjects able to
recall the correct brand names as the number of exposures to the brand names increased. Pairwise
comparisons of proportions revealed a significance difference between the attend/control and attendno
repetition conditions (253.92. p=.001). as well as between the attendtcontrol and attendlrepetition
conditions ( 2 - 5 . 6 2 . p=.OOL). There was a marginally significant difference between the attendno repetition
and attendlrepetition conditions ( Z 5 - 1.65. p=. LO).
In the ignore conditions. on the other hand. none of the pairwiw comparisons yielded significant
differences (20--7. ps>.2). Hence. subjects in the different ignore conditions were not differentially able to
recdi the brand names they were exposed to.
3.6 Discussion
That implicit memory for the brand names increased with repetition when attention is focused on the
slogans is an unsurprising one. since the positive relation between priming and repetition in full attention
conditions has been established by previous studies (Greene. 1990; Roediger & Challis. 1992; Krishnan &
Shapiro. 1996). The more interesting results pertain to the ignore conditions. As the analysis above
indicates. there was no evidence of irnpliclt memory when only one exposure to the brand name was
afforded. a result that is consistent with the predictions of the TAP framework. Given that only perceptual
processing is possible under the conditions. there ought not to be evidence of memory for the information on
a later conceptually-driven test. Data yielded by the ignodrepetition condition. however. cast doubt on this
line of reasoning. Subjects in this condition were presumably similarly impaired in terms of abiIity to
process conceptually. but interestingly, displayed implicit memory for the brand names.
Could this owe to the hct that there had been attentional shifts to the audio presentation that went
undetected'? The manipulation checks lend little support to this argument. Subjects in the ignordno
repetition and ignorelrepetition conditions missed a statistically equivalent number of words. yet only the
latter showed implrcit memory for the brand names. This. however, rests on the assumption that attentional
shifts can be detected by breah in the tlow of reading. Glven that we had intentionatly made subjects read at
the fast pace of 165 words per minute, this does not seem to be an unreasonable assumption.
In summary. the results of this first experiment indicate that, contrary to the TAP model, there can be
implicit memory on a conceprualIy-driven task even when there is Iittle conceptual processing at time of
exposure. This dilutes the strength of the conceptual-perceptual distinction as an explanation for implicit-
explicit memory dissociations. The results yielded by this experiment also suggest that advertising
communications can have an effect on our behavior even in situations when they are not attended to or are
ignored. This occurs presumably by virtue of the information contained within these communications being
sufficiently activated (or made more accessible) to levels that facilitate its coming to mind when one is
prompted with the right cues. Two observations are noteworthy. First of all. in the context of the present
experiment, the results seem to indicate that the level of activation afforded to ignored brand names when
the number of exposures is limited to one is insufficient for ~mplicit memory to be observed. Each additional
exposure represents added opponuniry for activation. such that at three exposures. incrementaI activation is
ample for the emergence of implicit memory. Secondly. information has to be available in memory to be
activated or made more accessible. Given that the brand names employed were ones of at least moderate
familiarity to university students, activation remains a viable explanation for the phenomenon. This
experiment says nothing about situations in which coherent memory representations have to be formed for
novel or obscure brand names for which there is little pre-experimental familiarity. There is, however,
evidence in the psychology literature that attention is critical to the formation of memory representations of
novel items. Smith and Oscar-Bennan ( 1990) found priming on a speeded lexical decision task to be
equivalent in both full and divided attention conditions for word stimuli, but not when they were nonsense
words.
The present experiment suffers from several shortcomings, among which is the high proportion of
subjects classified as hypothesis-aware. By omitcmg the data of these subjects horn the lnalyses. one risks
having excluded subjects that may have been systemrtticalIy different from the remaining sample in respects
other than hypothesis awareness. Although our andyse using data from both larger and smaller samples
yield essential1 y the bame results. it docs no1 detract from the fact that ensuring low awareness levels would
have been preferable as well as prudent. Also. the ~rnplici t memory measure employed required subjects to
respond with the tirst brand name that came to mind. While first responses reveal top-of-mind brands. the
practical implications of this are somewhat limited. It remains to be seen if implicit memory for a brand
name would translate to a greater likelihood of tts inclusion in the consideration set and of being selected in
a memory-based choice situation. In addition, the products used in this experiment were of the consumer
packaged goods variety and it is not clear if the effects would generalize to other types of products. One
might expect that more durable products, associated in general with more high involvement type purchases,
would be somewhat immune to the effects found In the first experiment.
4 Chapter Four: Experiment 2
Experiment I demonstrated that brand names presented outside the focus of attention can nonetheless
be made more likely to come to mind with repeated exposures. From a marketing perspective, it would be
interesting to see if this greater accessibiIity of the brand names might actually extend to greater likelihood
of choosing or considering the brands. We therefore use the dependent measures of consideration set and
choice in experiment 2. in place of the fint response measure used in experiment I . We examine to see if the
same pattern of results found in experiment 1 would hold with these other measures. Attitude measures were
also administered to help ascertain if repeated exposures in this context might lead to enhanced attitude.
Experiment two was also conducted with an eye to avoiding the pitfalls discussed in the previous section.
One objective was to lower the level of hypothesis awareness with a modification in procedure. In addition.
brand names of products belonging to the inore durdble variety (watch, camera) were employed to allow a
stronger test of the effects found in Evperiment 1 Attitude measures were also collected in addition to the
dependent measures of choice. consideration set and cued recall.
4.1 Design and Subjects
The experimental design remains similar to that of Experiment I. Hence, thre e factors are involved
- two between subjects factors of direction of attention (attend vs. ignore) and number of exposures to
brand names (zero vs. one vs. three). A totaI of 203 participants were initially recruited by placing an ad in
the university newspaper. Eleven of the 203 recruitees turned out to be older non-university students and
were therefore eliminated from the study sample. All 192 subjects retained in the study were thus university
students who were paid $8 for their putic~pation. The session lasted about 30 minutes and as before. each
session was conducted on an individual basis.
4.2 Materials and Procedure
In an effort to reduce the likelihood of hypothes~s awareness, subjects were first informed that they
were about to participate in two unrelated research studies, Study A and Study B. each of which lasts about
15 minutes. The instructions for Study A were essentialIy identical to that for Experiment I. That is, subjects
were told to either focus all attention on reading text or listening to audio messages as both were
simultaneously delivered. Essentidly the same text for reading was used. with the exception that words that
Experiment I subjects had particular problem with were replaced or removed. to allow for smoother
reading. The resulting text was only slightly shorter with a total word count of 586 words. (Please see
Appendix D for text passage.) The audio stimuli. in similar fashion, contained mock slogans featuring actual
brand names, atthough an entirely different set of slogans and brand names were used in this second
experiment. There were. however, only two criticaI brand names instead of four, employed (see Appendix B
for critical brand names and slogans used).' In the repetition conditions, to avoid drawing particular
attention to the repeated brand names, thereby leading subjects, particularly those in the attend conditions, to
be hypothesis aware, we repeated the two other tiller brand names in addition ro the two critical brand
names. Upon completion of the reading/Iistening task. subjects proceeded to answer the same multiple-
choice questions that tested comprehension of the text passage. To reinforce efforts to minimize awareness.
subjects were given a 5-minute filler task cons~sting of open-ended questions of no consequence to the study
As soon as subjects completed Study A, they were presented with Study B containing six tasks.
Introductory comments informed subjects that the study was done in collaboration with a marketing research
firm, as part of an ongoing effort to find out about consumers' preferences in various product categories. In
Task 1, subjects were presented with twelve different product categories and asked that they list the brands
that they would consider purchasing as if they were actually about to make purchases in each category. In
Task 2. they were then asked to pick one brand from amongst the brands listed for each category. the one
they would actually buy. To see if there might be an effect of task order, a different version of the tasks
administered to half the subjects, required them to first name the brand they would buy before listing other
brands that were considered. For Task 3. subjects rated fifteen different brands on two attitude scales
mchored by very bad/very good and dislike very mrtcldfike very much. In Task 4, they were presented with
seven different advertising slogans that they were again to rate on similarly anchored scales. For Task 5.
subjects were asked to think back to Study A and to try to recall the brand names that they heard. Fifteen
different product category names, of which only one corresponded to the critical brand. were provided as
cues. The final task consisted of quest:ons that probed possible awareness of the hypothesis or af the relation
between the exposure phase and the considenticn set and chorce measures. It also checked for suspicions
that the cover story was redly a rust: t t ~ drsguise ;he relation between the two supposedly separate studies.
4.3 Dependent Measures and Manipulation Checks
Two critical brands belonging to two diKerent product categories were included in this experiment.
For any subject. one of these product categories would appear on the consideration set, choice and attitude
measures while the other would appear on the explicit memory measure, with counterbalancing done across
subjects.
On the consideration set and choice tasks, we were interested to see if there might be an increased
propensity tbr subjects to respond with the brand names that they were exposed to, even under nonfocal
attention conditims. We dso wanted to investigate if exposure to the brand names might have affected their
attltude towards the brands: hence the brand attitude measure. If subjects exposed to the critical brands are
indeed more Iibcly ct; include those brands in their consideratinn sets, there may be two possible underlying
reasons - that exposure had increased rrccessibi li ty of those brands and/or that subjects' attitudes towards
the brands had been enhanced.
Among the slogans that subjects rated in Task 3 was one that they were exposed to earlier. Given we
were interested in their evaluation of the slogan rather than the brands per se in this particular task, the brand
names that were originally associated with the slogan was replaced with fictitious names judged to be
neutral in terns of any connotations. The purpose of this measure was to assess the relation between
evaluation of the slogan (as opposed to the bnnd) and the exposure level.
The other manipulation checks used in this experiment were the same ones used in Experiment 1,
consisting of the measure of reading fluency and the multiple-choice test of comprehension of the text
passage.
4.4 Results
Of the 192 study participants. approximately 69 were classified as hypothesis aware. This level is
lower than that of the previous experiment, demonstrating some success in our attempt to minimize
awareness. The discussion that follows is based on data that exclude these subjects. As in Experiment 1. our
analyses revealed that including them did not change the substantive results.
Mani~ulation Checks. As before. subjects r e c e i d 3 score out of ten on the multiple-choice test for
comprehension of the text passage. If subjects had directed their attention to the audiohisual stimuli as
instructed, ignore subjects (told to ignore the acdic) ought to outperform attend subjects on th~s test. Indeed.
there was a significant main effect of direction of attention (F(1,174)=110.37, p<.000L, MUa=3.85,
Mi,,=6-74) on the test score.
The number of words that subjects in the ignore condition missed when reading the text passage was
taken as a measure of reading fluency. This was subjected to an ANOVA. the results of which indicated the
absence of a difference in fluency levels across the number of exposures conditions (F(2.95)= 1.84, p>. 17).
De-pendent Measures. The data were subjected to a logistic analysis, Recall that some subjects
received the choice task first while others received the consideration set task first. We therefore first
determined that task order did not have a significant effect on any of the dependent measures and collapsed
the data across task order for all subsequent analyses. Also, since two brands were used, we checked to see
if responses r n l g l ~ ~ hake been different for the two brands. Again, neither the main effect nor the interactions
were found to be .>ignificant.
Considerntion set. The main et'tects uf d~rect~on c.4 attention and number of exposures on the
li kzli hood ot mcluding critical brands in one's consideration sets were both bignificant at the .05 p-level
( ;('( I !=5.86. x'(z)= 14-16, respectively; x e Figure 3 ,L The direction of attention effect shows that subjects
who attended w the brand names were more likely to respond with the critical brand names than subjects
who ignored them. Also, subjects who had three exposures to the brind names were more likely to include
those brands in their consideration sets than the subjects who had one exposure; and the latter group were, in
turn, more likely to do so than the control group.
------*-*-*---------------.-----*---------
Insert Figure 3 about here.
Planned pairwise comparisons show that subjects In the attendno repetition and attendrepetition
conditions were significantly more likely to include critical brand names in their consideration sets than
atrend/control subjects (Z=-2.95, p=.002; Z=-2.00, p=.02; respectively), but the attend/repetition and
attend/no repetition groups were not signiticantty different from each other subjects (E-0.86. p=. 19). When
attention was directed away from the brand names, the proportion of subjects who had those brand names in
their consideration sets was significantly higher in the ignorehepetition than in the ignore/control conditions
(Z=- 1.78, p=.04). However, the comparison between the ignorefno repetition and ignorefcontrol conditions
yielded a marginally significant difference (Z=- 1.42, p=.O8). A final compilrison between the ignoreho
repetition and ignore/repetition conditions indicates that the two groups are not different from each other
(Z=-0.4 1. p.34).
Choice. The choice data were similarly submitted to a logistic analysis (see Figure 4 for results).
There was a significant main effect of repetition (~?2)=8.7?. pc.01). reflecting the higher proportion of
subjects who chose the critical brands as the number of exposures increased. The main effect of direction of
mention was marginally significant (x2( 1)=?.63. PC. LO).
Of greater interest. how ier. are the planned comparisons. Within the attend condition, the proportion
ot subjects responding w~th the critical brands hcreased with repetition. Pairwise campaiisons show rz
significant difference between attendkontrol and atteadho repetition (2-2.02, p=.02). as well as between
attendfconuol and attendJrepetihon (Z-2.29, p=.O! J conditions. Attendho repetition and attendrepetition
were, however, not significantly different from each other (Z-0.20, p=.42). A different pattern emerges for
the ignore conditrons. .4 comparison of the control and no repetition conditions shows that there is no
significant difference in choice probability (Z--0.10, p=.46), but a similar comparison between the no
repetition and repetition conditions revealed a significant difference (Z-1.79, p=.04). The difference
between the ignorekontrol and ignorehepetition conditions was sxgnificant as well (Z- 1.8 1. p=.04).
Cited Recall. As can be seen in Figure 5, on the explicit measure of category-cued recall. there were
significant main effects of attention (x2( 1 )=67.00. pc.01) and number of exposures (x2(2)=i 32.22, pc00 1 ).
Planned comparisons show a steady increase in the proportion of subjects who correctly recalled the
brand (see Figure 5) . when they attended to the brand names (Zsc-3.00. pse.005, for ali pairwise
comparisons). When they ignored the brand names, repetition led to recall rates that were significantly
greater than in the control (Z-3.04. p=.OO!) and no repetition (25-2.86, p = . W ) conditions. Recall rates in
the ignorekontrol and ignordno repetition conditions were equivalent at zero.
Other Measures. The two attitude subscales for brand (a=.87) and for slogan (e.90) were averaged
to give a single score for brand attitude and a single score for slogan attitude, ANOVAs were conducted on
the brmd attitude and slogan attitude data, to assess the effects, if any, of the number of exposures and
direction of attention on affective evaluation. Null effects were found for both attitude measures (Fscl).
An additional measure. that of consideration set size, obtained simply by counting the number of
dternativr brands subjects listed for the critical product categories. was similarly submitted to an ANOVA
test. The expertmental manipulations of number of exposures and direction of attention were again found to
have no effect on set stze (Fsc 1 ). The only s~gniticant effect that emerged was that of brand with the watch
category having, on average. a significantly larger consideration set size than the camera category (M =2.33
and M =3.02 . respectively; F= 19.53. p<.OOI ).
4.5 Discussion
To summarize. the objectives of Experiment 2 were to Iend credence to the findings of Experiment 1
and to see if similar effects may be detected with the more ecologicalIy valid measures of consideration set
and choice. The results in this experiment are consistent with findings in Experiment 1 and show that
subjects' pmbabihty of choosing a particular brand can be increased by simply ~ncreasing the accessibility
of the brand name through repetition priming. Although this increase is reliable with just one exposure when
the brand name is given focal attention. the effect of prior exposure is much weaker when the brand name is
minimally processed. In fact. in the ignore condition. no increase was observed with a single exposure but a
significant increase was found with three exposures. The pattern of results for choice is thus consistent with
those found for the indirect measure in Experiment I .
Slightly different results were obtained for the consideration set measure. The point of difference lies
in the ignordno repetirion condition in relation to the ignordcontrol and ignore/repctition conditions. As cm
be seen in Figure 4. there was an increase in probabitity of inclusim with just one exposure. beyond which
t h e was no appreciable increase. The consideration set measure IS, 3f course quite different from the
choice measure or the indirect memory test employed in Experiment 1. in that it allows for multiple
responses to a single product cue. We propose that even with a single exposure. the brand names were made
more accessible. This increase in accessibility may not be sufficient to be detected when one is given the
opportunity to generate but one brand name, by allowing the generation of multiple brand names. the chance
of detecting the effect of a single prior exposure is increased. Although we find in this experiment that the
probability of including the critical brand name in a multiple brand consideration set did in fact increase
with a single exposure. that iccrew attained but marginal significance. This is in contrast to what was
found by Shapiro. Maclnnis and Heckler (1997). In their study, a single exposure to a product presented in
peripheral vision increased the likelihood of its consideration for choice.
There are a number of signitkant differences between the two studies that may have contributed to
the different findings. Whereas target items were presented auditorially in this study. they were presented
v~sually in Shapiro et al.'s study. There may be fundamental differences between the two modalities that
allow preconscious processing of stimuli to occur more readily in the visual presentation mode than in the
auditory presentation mode. In addition. the dependent measure in Shapiro et al.'s study involved the
forming of consideration sets for ad-hoc categories ("food one would buy before p ing parachute-jumping")
whereas consideration sets were for established product categories in the present study. Subjects are likely to
have pre-existing preference and knowledge structures associated with these product categories, which may
dilute the effects of experimental manipulations. making it harder to detect potential effects.
Another finding of interest is the equivalence in brand attitudes across conditions. Had brand attitudes
been affected by the experimental manipulations. such as if brand attitudes had become more favorable with
the number of exposures. there would have existed a compelling alternative explanation for the effects found
in this experiment. In those conditions where repetition increased the probability of including the critical
brands in one's consideration set or of choosing those brands, the fact that attitudes were statistically
equivalent across conditions leaves increased accessi bi ti t y of the brand names. rather than favorabili ty of
attitudes toward those brands, as the primary underlying factor [see Nedungltdi (1990) for similar findings].
To further examlne the relation between brand attitude and choice as well as considerat~on set
~nclusion. we correlated the relevant measures in each of the SIX conditions. Specitically. subjects' affect
ratlngs were correlated with their choice and considerdtwn set responses (which were coded as I when they
responded with the target brand and G otherwise.) Interestingly. the only significant correlations were found
in the control conditions. For the attendfcontroi condition, brand attitude was significantly correlated with
choice (e.35. p . 0 3 ) and for the ignorekontrol condition. the correlations between brand attitude and
choice (e.30. p=.061 and between brand attitude and consideration set inclusion (e.30. p.06) were
marginally significant. These results suggest that without any exposure to brand-related information,
~ndividuals are likely to include a brand in their consrderauon set based primarily on attitudes. With minimal
Levels of processing of brands, the influence of any a priori act~tudes towards the brands does not appear to
increase. evidenced by the lack of correlation between brand attitude and consideration set inclusion or
choice in both the attend and ignore situations.
A frequent concern in implicit memory studies is the question of whether perform,ance on implicit
memory measures might have been contaminated by explicit memory strategies. That is. might subjects
have relied on intentional retrieval. and hence explicit memory. in responding on the implicit memory
measures'? There are three sources of evidence in Experiment 2 that suggest that this is not the case. First of
all, we correlated the recall measure with the choice and consideration set measures for the different
conditions in which expIicit memory was evident (i.e. attendno repetition. attendrepetition,
ignorelrepetition). None of the correlations turned out to be significant (ps>.3). Secondly. we had excluded
from the analyses subjects who saw through the cover story and expressed awareness of the relation between
the priming phase and the test phase. Hence. the remaining subjects were unlikely to have intentionally
recollected and responded with what was heard when completing the consideration set and choice tasks.
Finally. we note the following dissociation: Subjects in the ignoreho repetition condition. despite being
unable to recall the brand names that they heard (see Figure 5). were nonetheless as equally likely to
consider the brand for choice, as subjects in the ignorelrepetition condition. The latter subjects. on the other
hand, showed signiticant explicit memory for the brand names.
That explicit memory would be found in the ignoreIrapetition condition is somewhat unexpected. The
manipulation check on reading fluency certainly did not indicate attention shifting to be a potential
explanation. I t seems plausible. particularly given the use of familiar brand names. that repeated exposure
heightened xttvation to such levels that perm~tted conscious detection of their occurrence. Such an
explanation is in agreement with the atrenuator model of attention proposed by Treisman ( 1960, that will be
further elaborated on in the general discussion section.
Could subjects have momentarily switched artention to the unattended channel, heard the brand
names. but before coming around to responding on the explicit memory measures had forgotten what was
heard? This is an important question from a theoretical standpoint. but it is nevertheless still interesting to
know. particularly in an applied context, that previous encounters that have faded from conscrousness can
influence our behavior in ways that we are not aware of.
A final nbjrction may be made with repard to the assumption that shifts in attention can be detected
by breaks in the tlow of reading. The uggestion may be made to increase reading speed. so as to get a more
sensitive measure of attentional shifts As discussed earlier, subjects were made to read at the rather fast
pace of 165 wpm. a rue determined in n pretest to be comfortable enough not to undermine efficiency. It has
to be borne in mmd that excessive breaks in reading caused by reading speeds that are hard to keep up with,
may itself be the cause of attentional shifts and not merely the symptom. Hence, the issue of an optimal
reading speed remains a methodological challenge.
5 Chapter Five
The focus of experiments 1 and 2 was on the effects of repetition on memory for brand names when
these brand names were encountered under low involvement conditions. Memory for brand or product
information is, of course, but one factor influencing consumer buying behaviour. Affect or attitude toward a
brand is an important input to choice, and is a factor that's of immense interest to marketers. The focus of
this dissertation, beginning with this chapter. shifts to affective judgements of stimuli initially processed
under conditions of low involvement. As in the previous two experiments, the effect of repetition as a
critical moderating variable. is examined. The literature in the mere exposure effect, which is concerned
with the relat~onship between repetition and affect. provides the backdrop for discussing the third
experiment of thls dissertation. The following section provides an overview of the mere exposure effect.
5.1 The mere exposure effect
The term ntrre exposure effect was coined by Zajonc ( 1968). in reference to the phenomenon of
enhanced liking for stimuli to which one had been repeatedly exposed. That the phenomenon so intrigued
researchers 1s reflected in the more than 150 published studies investigating the mert exposure effect.
following Zajonc's initial
publication. The wide-ranging contexts within which it has been examined. fiom explaining food preference
(e.g Pliner. Crandall) to explaining liking for various stimufi in nonhuman mammals (Hill, 1978)- also attest
to its far-reaching impact.
The mere exposure effect has certainly received extensive empirical support. Within mainstream
social and cognitive psychology, the mere exposure effect has been found with viuious types of presentatron
stimuli - meaningful words (Grush. 1476. Whirtlesea. 1993). geometric shapes (Mandler. Nakarnura &
Van Zandt. 1987; Bumstein & D'Agustino. 1992). melodies (Wilson. 1979; Obemlller. 1985) -as well as
using various clic~tatiort measureb - forced-choice preference judgment (Seamon, McKenna & Binder.
1998; Mandter et al., 1987: Wilson. 1979: K~nst-Wilson & Zajonc. 1980). like-disiike rating scale
t Bornstem 4r D'Agost~no. 1992). multiple bipoh adjecile scales (Okrmiller. 1985; Jmiszewski, 1993).
WhiIe the robustness of the mere exposure effect is seldom in dispute, the reasons for its mxrrence
and theoretical accounts put forth have been the topic of fervent discussions and scrutiny. One of the most
Important aspects of the mere exposure effect that these theories have had to account for, is what's been
called the nonconscious mere exposure effect, the finding that repeated stimuli are better liked despite
subjects having no recollection of having encountered them. In Kunst-Wilson and Zrrjonc's (1980) study,
for instance. subjects were briefly exposed ( 1 ms) to target items at study. At test. they chose target items
signiticantly more often than distractor ones in a forced choice preference task, but picked out target items at
no better than chance level on a recognition task. An early theory proposed by Zajonc (1980, 1984) to
account for this tinding, one that has generated considerable controversy and debate. suggests that affective
and cognitive responses are the product of two separate and independent systems. Affective processing may
proceed independently of cognitive processing and may also precede it. thereby giving rise to the observed
independence between the two. Skeptics were quick to point out that the absence of recognition is not a
sufticisnt condition to establish the absence of cognition (Lazarus. 1982; Gordon & Holyoak. 1983). Rather,
affective responses may be mediated by cognitive activity in the form of perceptual registration (Lazarus.
1982). even when recognition is no better than chance.
The latter represents the cognitive mediation approach. an instance of which is the uncertainty
reduction model (Sawyer, 198 1). It posits that each exposure offers the opportunity for learning and a
reduction in uncertainty about the stimuIus is a direct consequence of this learning. Because a stimulus
about which there is less uncertainty is less likely to be associated with aversive tension. affective response
to the stimulus becomes more positive with repeated exposures. This explains the uncertainty reduction
model in general terms. More precise explications of the mechanism that Ieads from familiarity to liking are
contained in theories such as wo-factor theory (Berlyne, 1970). response cornperition (Harrison, 1968) and
optirrtul arousal (Crandall. 1970). the most prominent being the first of these. Two-factor theory asserts that
affect becomes more positive as one [ e m s about the stimulus and becomes more familiar with it with
repeated exposures, but that this enhancement occurs only up to a certain point. beyond which tedium or
boredum sets In. When this comes about. a decline in affect becomes evident with additional exposures.
Tu.0 factor theory hres fairly weli for the most pan. in accounting for tindings in the mere exposure
literature. Brxnstein ( 1989). in his rneta-analysis and review of the literature. noted that 'Qverall. a two-
factor learning-satiation model of exposure effects seems best supponed by this review." Nonetheless, the
rheury does have its difficulties in explaining some findings. the most significant of which is that stronger
mere exposure effects are obtained with sub1 iminal rather than supraliminal stimuli (Bornstein &
D'Agostino. 1992). Since the theory proposes that affect varies with subjective famdiarity, this finding
cannot be readily accommodated.
Today, the independence vs. cognitive-mediation debate has come to an impasse and newer theories
have in recent years gained favour at the expense or these earlier ones. They are the non-specific activation
hypothesis. the perceptual fluency/attributional model and the discrepancy-attribution account, each of
which is discussed in the following section. Although the experiment that will be described in the next
chapter docs not provide a critical test of the competing accounts, a discussion of these accounts is intended
to provide background understanding of the mere exposure effect in terms of recent theoretical
developments.
5.2 Theoretical accounts of the Mere Exposure Effect
5.2.1 Non-specific Activation Hypothesis
Mandler et al. (1987) argue that two things can happen when a stimulus is encountered - there can
be activation of the stimulus* representation in memory and there can be elaboration on the item involving
more conscious or effonful processes. The first of these results in the greater accessibility of the stimulus in
memory and is context-free. unlike elaboration. which involves integrating the stimulus with the context. In
studies demonstrating the mere exposure effect for stimuli that are not recognized. the target stimuli are
typically presented very briefly for but a fraction of a second (e.g. Kunst-Wilson & Zajonc. 1980). In such a
situation. Mandler et al. ( 1987) argue. there is activation of the target stimuli in memory with each
subsequent presentation even though elaboration is not passible. When later presented with a target. for
which there is an activated representation. and a distractor, for which there is none. the greater activation
biases subjects to select the target as the preferred stimulus. Because successful recognition is reliant on
having integrated. to some extent. the stimulus with the context at time of exposure. the very brief
presentations afforded at exposure ensured that recognition performance would be poor.
The more imponant pan of Mandler et al.'s ( 1987) account. however. is the claim that this bias in
response due to activation is not peculiar to affective or preference judgments. The account predicts that
when asked to make judgments on any other relevant dimension. subjects' responses will favour the target
in a similar fashion. This was indeed what was found in an experiment in which different groups of subjects
were asked to select from a pair of stimuli. the preferred stimulus. the brighter stimulus, the darker stimulus
or the stimulus that had been presented in an earlier exposure phase. Recognition performance was
unsurprisingly poor with target items selected 49.62 of the time; on the other hand. target items were
preferred over distractor ones at 59.2% of the time. were chosen as brighter 60% of the time and chosen as
darker 60.8% of the time.
That mere exposure effects can be found not just with affective judgments lessens the viability of
Zajonc's ( 1980.1984) independent hypothesis. Zajonc had accorded affect a special status. by arguing for a
reparate affective system that generates affective ruponses. unmediated by cognitive processes. If mere
exposure effectr can be found with other types of judgrcenls. then certainly. it %odd not be reasonable to
postulate that a scpante system is responsible for each type of response.
Mandler cr at ( 1987) refer to the sort of activatmn that occurs as nonspecific activxion because it h s
the potential to intluence judgment about any relevant Fmpeny or dimension of the stimulus. Hence. it is not
simply n response bias. but rather iorolves the subject attempting to relate the activation to a relevant
dimension of the stimulus. The authors point out that one would not expect to find similar effects with
dimensions such as odor or a t e because they are irrelevant. but this raises the question of what relevance
entails and if it is possible to specify a priori the dimensions that are considered relevant and ones that are
not. a question not dealt with in any detail by the nonspecific activation account. Moreover. this issue of
relevance. not being the focus of the paper. was not subject to testing in the experiment reported. The
dimensions of liking. brightness and darkness used in the experiment were all deemed to be relevant
dimensions. The nonspecific activation account also does not provide the conceptual link between the
simple activation of a stimulus* representation in memory and its effect on judgment. The reason for that
cause and effect is not clearly argued. Finally, the only attempt at replicating the experiment reported to
date. by Searnon, McKenna and Binder ( 1998). proved unsuccessful. Prior exposure was found to affect
responses an preference judgment. but not on judgments of brightness or darkness.
5.2.2 Perceptual Fluency/Attributional Model
The Perceptual FluencylAttributional Model of the mere exposure effect was proposed by Bornstein
and D'Agostino ( 1994) and motivated by the observation that subliminal stimuli tended to produce mere
exposure effects that were about twice as large as that for supraliminal stimuli (Bornstein, 1989). It therefore
attempts to explain two interesting aspects of the mere exposure effect. The first is what most currently
viable accounts have concerned themselves with, and that is that mere exposure effect can be found even for
stimuli that are not recognized. The second is the larger effect found for stimuli that are not recognized. in
comparison to clearly-recognised stimuli.
The model draws on Jacoby and Kelley's ( 1987) notion of perceptual tluency. which briefly defined.
is the ease of processing or encoding [he perceptual features of a stimulus. Perceptual fluency arises most
often as the result of repeated exposures to a stimulus - a previously encountered stimulus becomes easier to
process on subsequent encounters. This fluency in processing is experienced as a feeling state. which the
subject relies on in making an affective response to the stimulus. In fact. as Jaccoby and his colleagues have
argued (Jacoby. Kelley & Dywan. 1989). this misattribution of perceptual fluency is not peculiar ro affective
judgments. but has been found to occur on a variety of other judgment task that the subject is faced with.
For instance. in what has become known as the false fame experiment, in the absence of recognition for
previously presented names. subjects were signlticantly more likely to mistakenly call previously seen
names famous, compared to new names (Jacoby, Woloshyn & Keiley, 1989). In another study, subjects
were presented with sentences in a noisy background and were given the task of judging the loudness of the
noise (Jacoby. Allan. Collins & Larwill, 1988). Some of these sentences had been presented prior to the task
and were more fluently processed. As a cmsequence of a misattribution of this fluency, subjects judged the
noise to be less loud when the sentences were old rather than new.
Repeated exposure is not the only means by which perceptual fluency may be engendered.
Whittlesea. Jacoby and Girard (1990. Experiment I ) showed that by subtly manipulating the level of
masking of stimuli (20% vs. JOB) such that some words were easier to process than others unbeknownst to
subjects, the more fluently perceived words were more likely to be called 'old' in a recognition test. Similar
manipulations led subjects to judge with greater probability that fluently processed words were presented for
twice as iong (Whittlesea. 1993~.
Regardless of the manner in which perceptual tluency is created, the reason i t affects these various
judgmenrs has to do with subjects, being unaware of the source of the fluency, seelung the most
parsimonious explanation for the experienced feeling state. Since this perceptual fluency is experienced at
the same time the judgment task is presented, subjects misattribute the fluency to their evaluation of the
stimuli on the dimension concerned.
Bornstein and D'Agostino go a step beyond this explanation of the mere exposure effect by arguing
that a correcting mechanism is in place, that alIows one to discount the initial interpretation of the fluency.
This discounting occurs when subjects are aware of having been exposed to stimuli in an earlier phase that
might be at least partially responsible for the experienced tluency. They are therefore likely to discount the
fluency and report lower affective ratmgs than would otherwise be the case. This expiains the finding of
larger mere exposure effects for subIirninaI rather than supralrminal stimuli. No discounting process occurs
with the former because subjects are typically not cognizant of having been previously exposed to the
sr~muli. The key. however. is not in the subliminality of the stimuli, but rather the lack of awareness of the
relarronsh~p between the exposure phase and the affective judgment phase. and subliminal exposure is
merely one way of achieving it.
Comparing the nonspecific activation account and the perceptual fluency/attributional model. it is
apparent that the two are very similar and may be two sides of the same coin. The consequence of activation
in memory is greater accessibility and hence quicker processing (Collins & Loftus. 1975; Anderson. 1983).
Greater processing speed is also implied by greater ease of processing and perceptual fluency, in fact. is
often indexed by reaction time (e.g. Whittlesea. 1993). The perceptual fluencylattributional model, unlike
the nonspecific activation account . is more explicit about the process that leads from stimulus encounter to
affective judgment. Perceptuai fluency has also been the subject of numerous experiments (Reber,
Winkielman & Schwan. 1998; Whittlesea, 1993; Anand & Sternthal, 199 1 ) an3 has stood up well to
empirical testing. The discounting component of the perceptual fluency/attributional model. on the other
hand. awaits funher testing. There has not been much evidence either in support or against it, beyond the
experiments conducted by Bornstein and D' Agostino ( 1994). While the discounting argument is plausible, a
subtle distinction can possibly be made between an argument that says there is an initial misattribution of
fluency !o l i k i q . folIowed by a correction of that interpretation and one that says there is a smaller
likelihood of misattributing the tluency to begin with. when one is aware of other potentla1 sources of that
fluency. Future studies would have to make a distinction between the discounting argument and alternative
expianations like this one.
5.2.3 Discrepancy-Attribution Account
The Discrepancy- Attribution explanation was praposed by Whittfesea and Will lams i 1998) not so
much to account for the mere exposure effect. but rarher to explicate the circumstances under which feelings
of familiarity arise. Nevertheless, the account is based on the notion of processing fluency and as the authors
themselves have argued. the feeling of familiarity is but one of many possible feeling states (inciuding
feelings of pleasantness or liking) that may be aroused by processing fluency. As such the concepts
proposed in their arguments are equally relevant to an account of the mere exposure effect.
The feeling of familiarity is the subjective feeling of having encountered something before. The
prevailing view is that is arises because one experiences fluent processing and attributes the fluency to an
encounter in the past rwhittlesea. 1993; Jacoby & Girard. 1990; Lindsay & Kelley, L996). Whittlesea and
Will~arns argue that it is not processing tluency per sz that is responsible for these feelings of familiarity.
After 311. there are a multitude of stimuli that one can tluentty process. such as the familiar faces of one's
friends. the objects In one's home, ctc. Yet the feeling of familiarity does not always accompany all
occasions of tluent processing. Rather ir is experienced only when the fluency is unexpected or surprising
given a particular context.
The discrepancy-attribution model th~refore makes two important assertions. First, the feeling of
familiarity is the product of a compuison process between experienced fluency and expectation. It arises
when there is a mismatch between outcome and expectation. The second assertion follows from this and
says that. unlike that advocated in other theories, fluency need not be enhanced to produce the effect.
Rather. fluency need only be greater or less than the context leads one to expect. This in turn leads us to the
question of how these expectations come about. Whittlesea and Williams ( 1998) alluded to a general
background experience of an item. but provided no specifics as to how this background experience is
derived. or what it translates to in psychological!y meaningful terms. One conjecture might be that the
background experience of an item is the avenge fluency or activation level over the caurse of one's lifetime
experience with the item.
The discrepancy-attribution model is resonably applicable as an explanation of the mere exposure
rfkct. Under appropriate situational constraints provided by the stimuli. context and task. t1uenc.y may be
attributed to a feeling of pleasantness or liking; surprising tluency. that is. This. in fact. offers an
explanation. alternative to the perceptual tluency/atuibutional model. for why subliminal mere exposure
studies yield stronger effects compared to supralirninal ones. Subjects exposed to supraliminal stimuli are
less likely to be surprised when fluency is later experienced, since there is reason to suspect that the prior
exposures would exert an effect on fluency and expectations may be influenced accordingly. This
hypothesis that the mere exposure effect is the result af unexpected fluency, while a simple variant of
Whittlesea and William's ( 1998) hypothesis with respect to the feeling of familiarity, has as yet not been the
subject of any published study. Whether or not it will stand up to empirical testing remains to be seen.
5.3 Perceptual Fluency vs. Conceptual Fluency
The three accounts of the mere exposure effect discussed in the previous section are not too markedly
different from each other and, ;n fact. they all seem to share the theme of perceptual fluency The notion of
perceptual tluency is an intuitively appealing one and one that has received support in the literature. Much
less often ascribed to and examined is conceptual fluency. Conceptual fluency. like perceptual fluency,
refers to the case of processing a stimulus except with respect to the semantic properties or meaning of the
stimulus.
There has been but one experiment published to date that investigates the mere exposure effect in the
context of conceptual fluency. Whittlesea (1993) had subjects read a series of sentences, presented in such a
way that each sentence. except for its last word, was presented on one screen. followed by the last word that
completes it. on the next screen. The sentences were constructed such that the sentence stems were either
more (predictive context) or less (neutral context) predictive of the last word that would appear.
For example. in the predictive condition. the following might be presented on the fint screen
'The mother gently rocked her . . ." folIowed by the target word
"baby" on the next screen.
In the neutral conditlon, a sentence and a target word such as the following might have appeared
instead:
'The boy in the classroom could not . . ."
"draw"
This predicuveness of context represented [he ~~nceptuai tluency rnanipulat~on. In the predictive
context condition. the meaning of the sentence stems constrained the possible words that could appear next
and this presumably led to greater ease in processing the meaning of the word when it did appear. Subjects
were asked to indicate if target words. the last words of the sentences, were pleasant or neutral. Results
showed a significantly higher probability of judging target words to be pleasant when presented in a
predictive context. compared to a neutral context.
However. the mere exposure effect as has been defined, is the more favourable affective response to
stimuli as a result of unrehforced repeated exposures. One may argue that subjects in the predictive
condition were able to anticipate the target word and confirmation of that expectancy, on presentation of the
tarp: word. may be reinforcing in itself. thereby leading to greater lilung for the target word. Hence. it may
be the reinforcing effects of the predictive condition. rather than conceptual tluency per se. that may be
driving the mere exposure effect.
6 Chapter Six: Experiment 3
6.1 Objectives of Study
The objectives of the present study are four-fold:
1. We test the hypothesis that conceptual tluency can give rise to the mere exposure effect, in a way
that avoids the confounding effects of reinforcement discussed earlier.
2. We manipulate perceptual and conceptual tluency within the same experiment. thereby allowing
us to examine the relationship between the two. Four possible scenarios are depicted in the
graphs of Figure 6.
In the first instance, shown in Figure 6a, we find that perceptual fluency but not conceptual
fluency exerts an effect on affective judgments. Perceptual fluency effects have been effectively
demonstrated in the literature. whereas conceptual fluency has as yet, not received extensive
support, making this scenario a possible one. Alternatively. we might find that perceptual and
conceptual fluency operate independently to produce the mere exposure effect. This is shown in
the gaph of Figure 6b. where the effects are independent and additive. Figures 6c and 6d depict
scenarios where perceptual and conceptual fluency interact. In the former. the affect is at its
highest level when both perceptual and conceptual fluency are high. In the latter. only one of
conceptual or perceptual tluency needs to be high, in order for affect to be enhanced. There are
no additional gains from having both high conceptual and high perceptual fluency.
3. We also examine the effects ot'ctjnceptu~l and perceptual tluency on recognition. Jacoby and
dthers (Jacoby, Kelley & Dywan. 1989, Jacoby. Woloshyn & KeIley. 1989; Whittiesea, 1993;
Whittlesea & W~lliams. 1998) have argued that tluency when prccesslng a stimulus can give rise
to a feel~ng of familiarity and hence affect familiarity judgements. Since recognition performance
is widely thought of as being reliant on two component processes, famil~arity and recollection
(Mandler, L980), fluency may exert its effects on recognition performance by influencing
familiarity. In such a case, we may observe that both recognition and affect are affected by
conceptual and perceptual fluency in a paralie1 fashion.
4. We administer both affect and recognitron measures for each stimulus item. The use of multiple
measures allows us to observe if fluency can simultaneously influence affect and recognition or
if tluency attributed to one will not be available to influence the other. Moreoever, taking both
measures allows us to analyse the relationship between affect and recognition using measures of
association. If both affect and recognition are based on the same processes, a positive association
should be obtained.
6.2 Subjects and Design
Subjects were 51 undergraduate commerce students. who participated in the study for extra credit.
The design is a 1 (perceptual fluency: high vs. low) x 2 (conceptual fluency: high vs. low) x 1. (number of
exposures: 1 vs. 5) x 2 (order of dependent measures: affect first vs. recognition first). The first 3 of these
factors are within-subject factors, whereas the last is a between-subjects factor.
According to Bornstein's ( 1989) review. the effects of exposure on affect tend to peak between 10-20
presentations. In order to provide a clear assessment of the potential effects of fluency, it was necessary to
avoid this downturn in affect. Moreover. since natural words were used in this experiment. in contrast to the
meaningless stimuli (nonsense words. photographs. polygons, etc.) employed in the majority of the studies
reviewed by Bornstein. the likelihood that an earlier downturn might occur led to a decision to set the
number of exposures conservatively to one and five.
6.3 Stimuli
The stimuli were meaningful words sharing the common characteristic of being a homogr~ph.
Homographs are words that have two or more possible meanings. An example would be the word RIGHT,
which presented in the absence of a context. could mean the opposite of LEFT. or it could mean CORRECT.
48 homographs were selected according ro the following criteria: ( 1) they had to be between three and seven
letters long (2) hey had to have synonyms that were also between three and seven letters long. All 48
homographs were presented in the exposure or study phase, each presented either once or five times.
In order to constrain the meaning of the homographs, they were presented in the context of sentences.
The sentences were between six and ten words long and the homographs always appeared as the last word of
these sentences. For example. a subject might see the sentence. 'There are prizes for guesses that are right."
A subset of these sentences was drawn from Masson and MacLeocl(1992).
In the judgment phase. subjects again saw 48 sentences between six and ten words long with target
words at the end of sentences. The target words were either ( I ) one of the homograph words seen at
exposure (2) a synonym of one of the homograph words seen at exposure (3) one of the homograph words
seen at exposure bur with a different meaning :mplicd. c,r (4) a new word unrelated to any of the homqraph
words seen earlier.
Fur example. if the following sentence w s presented at study:
"His inheritance made him very rich."
One of the following four sentences was presented at judgment:
I. "He won the lottery and became rich."
2. "He won the lottery and became wealthy."
3 . 'The chocolate cake was very rich."
4. 'The chocolate cake was very tasty."
The first of these sentences represents the high conceptual fluencylhigh perceptual fluency condition.
since the target word has the same meaning and looks the same as the homograph word presented at study.
The second sentence represents the high conceptual fluencynow perceptual fluency case because the target
word is a synonym of the word seen at study. In the third sentence. the target word is the same word as that
at study. although a different meaning 1s implied by the context and represents the low conceptual
tlucncy/h~gh perceptual tluency condition. Finally. a target word unrelated to that seen at study is presented
in the lasr sentence and we therefore have a low conceptual tluency/low perceptual fluency. Which of the
four sentences was presented was randomly determined by the computer program. (Please see Appendix E
for the entire set of sentences used in the study.)
6.4 Dependent Measures
Aher each sentence is presented, subjects made two judgments about the target word. Approximately
half the subjects first responded to an affective judgment task. They were asked to indicate if they found the
word to be 'pleasant' or 'neutral'. This was followed by a recognition test that required them to respond
'old' or 'newb to each target word. The order of the two tasks was reversed for the rest of the subjects.
A variety of measures has been used with success to elicit the mere exposure effect (cf. Bornstein.
1989). The decision to use a pleasandneutral task here was in keeping with WhittIesea (1993). Also. a
choice between 'pleasant' and 'neutral' was deemed more appropriate than one between 'pleasant' and
'unpleasant' because. although the iattcr would present no problems in the case of words for which there is
high fluency ( we predict that subjects would misat?ri b ~ t e the tluency to pleasantness and hence respond
'pleasant'r, In the case of ~ o r d s for which there is low fluency or an absence of fluency. subjects may find
neither of -pleasantg or 'unpleajant' to be appropriate responses. A 'pleasant/ne~tral' judgement would
therefore be more congruent w~th high and low fluency. Moreover. an effort was made to avoid using words
that may be connotati-dy negative, which could overwhelm any positive effects of fluency and limit our
c h x e s of finding signiticant etfects. 'fhc use of such words also make a pleasantheutral judgement task
suitable.
6.5 Procedure
The entire study was administered on computer and each individually-run session lasted about thirty
minutes. On entering the lab. subjects were seated in front of a computer. They were told that the study
consisted of several tasks. that separate instructions precede each task and to follow instructions closely.
On the first task, which was the exposure phase of the experiment, subjects were informed that
sentences would appear on the computer screen, one at a time. Their task was to read each sentence aloud as
i t appeared. They were warned that the sentences would appear only briefly and they were to therefore focus
their attention on the centre of the screen where the sentences would appear. Each sentence was on-screen
for exactly 2.5 seconds, with a 1-second lapse, during which a '-' appeared on the screen, between
sentences. These sentences were the 48 sentences containing ~e homographs discussed in the previous
section. Half of these were presented just once, while the other haIf was repeated five times. The distribution
of the repeated sentences was randomly determined by the computer program.
The second task was a filler task intended to clear short-term memory. Subjects were given fairly
engaging math problems to solve for five minutes. These were presented on the computer screen. one
question at a time, and subjects input their answers using the keyboard. They were led to believe that their
math ability would be correlated with their performance in other parts of the study. Given that the critical
stimuli were words. math problems. involving only numbers. were chosen as the filler task to avoid possible
interference effects that might occur with a task that had words in it.
The third task was the critical phase where the dependent measures were administered. Subjects were
once again presented with sentences that appeared for 2.5 seconds each. The instructions given to them were
that they would once again be presented with sentences just as in the earlier part of the experiment. and they
were to read them aloud like before. This time around. however, after each sentence was read, the last word
of the sentence would remain on the screen. For the affective judgement task, the words 'pleasant' and
'neutral' appeared below the target word. Subjects were instructed to assess how pleasant they found the
word to be. If there was a pleasantness to the word, they were to indicate their choice by pressing the key
corresponding to the 'pleasant* option. Otherwise, they were to mdicate 'neutral'. The recognition task was
similar to the affect judgment task, except subjects chose between 'old* and 'new'. based on whether they
believed the same word [identical in meaning or form) had appeared earlier in any of the sentences
presented in the exposure phase. (See Appendix F for instructions.) The affect judgment and recognition
tasks were presented one after the other for each target word. Although both tasks were self-paced, with
each target word remaining on-screen until the two responses were made. subjects were advised not to
deliberate too long on each word before making their choice. Also. in order to ascertain that subjects had
understood the instructions correctly and to allow them to familiarise themselves with !he task. five
sentences that were not part of the set of 38 were presented for practice.
The final task of the experiment consisted of questions relating to the subject's demographic protile.
This concluded the experiment and subjects were debriefed and prompted for hypothesis awareness. They
were specifically asked to describe what they thought the experiment investigated. None of the subjects
related anything that hinted at hypothesis awareness.
6.6 Results
For each subject. the proportions of 'pleasant' responses for the affect task and 'old' responses for the
recognition task in each of the conditions were calculated and these proportions represent the data analysed.
Note that since we are looking at the probability of judging a word to be 'old', we are examining effects on
subjective recoption rather than recognrtion accuracy.
In terms of the results that might be expected, in view of scenarios depicted in Figure 6, we might
tind that perceptual fluency alone drives the mere exposure effect. In such a case. there should be a
significant morn effect of perceptual fluency or an interaction between perceptual tluency and exposure
Ievel. in the absence of effects involving conceptual fluency. Alternatively, should perceptual fluency and
conceptual tlurncy contribute independently to the phenomenon. the analyses should reveal effects
involving conceptual and perceptual fluency. but no interactions between the two. Two other scenarios
predict a significant perceptual fluency by conceptual fluency interaction - one in which the effect from
both sources of fluency combined is greater than can be expected from either alone. and the other in which
tluency from either source is sufficient to establish high affect levels, and no fwther gain is achieved with
tluency contributed by both sources. Post-hoc contrasts would help distinguish the latter czse tiom the
former.
Most studies in mere exposure use multiple exposures to generate the mere exposure effect, although
some have found effects with just a s~ngle presentation (e.2. Oberrniller, 1985; Whittlesea, 1993). We may
tind in the present case that multiple exposures are needed for effects of perceptual andfor conceptual
tluency to be observed. in which case tluency by exposure level interaction(s) can be expected. The
interactions also allow us to observe if the impact of perceptual and conceptua1 tluency on affect may differ
across the two exposure levels.
As 3 first step. the presence of an order effect is checked for. Results from a four-way ANOVA
performed on the affect data revealed a significant effect of conceptual tluency (F( 1 A 15)=5.09. p<.05). a
significant perceptual fluency x exposure level interaction effect (F(1,415)=4.35. pe.05). a significant order
effect (F( 1.4 15 ~=25.79. p<.0001) and a marginally significstnt effect of exposure level (F( 1.4 15)=3.54.
p<. 1). The rrcngnrtion d m were similarly analysed and yielded rather similar results. The following efTects
were significant: main effect of conceptual fluency (F( 1.4 l5)= 16.27, pc.000 1). main effect of perceptuai
tluency (FI 1.3L5)=202.57, p 4 0 0 1 ) , perceptual fluency x exposure level interaction effect (F(I.415)=16.01.
p<.0001) and main effect of exposure level (F(1.4 15)=41.32, p<.000 I). There was also a marginally
significant main effect for order of dependent measures (F( 1,4 15)=2.87, p<. 1).
Given the effects found for order, as well as for ease of interpretation of the results. both affect and
recognition data were further analysed by way of three-way ANOVAs for each level of order. There were
therefore four separate ANOVAs performed. the results for which are reported below.
6.6.1 Affect
When the affect measure preceded the recognition measure. analysis of the affect data yielded a
significant main effect of conceptual fluency (F( 1.3!5)=4.34. px.05). Subjects were more likely to call
target words 'pleasant* (&*=0.42. &ow,=0.36). when the target word had the same meaning as a word
presented in the exposure phase. This is therefore evidence that conceptual fluency can drive the mere
exposure effect. There was also a marginally significant effect of repetition (F(1.2 15)=3.08, p<. 1). with the
five exposure condition leading to a slightly higher likelihood of choosing 'pleasant' compared to the one
exposure condition (&,=0.42. &,,=0.36). The effect is. however. qualified by a significant perceptual
fluency by exposure level interaction (F( 1.215)=4.15, pc.05). The means for the various conditions are
presented In Table 1.
A closer look at the perceptual fluency by exposure level interaction with simple main effects tests
revealed a significant difference between high and low perceptual tluency within the five exposures
condition 1 & , @ ~ ~ ~ ~ = 0 . 4 7 . &0wpkxp=0.37. t=2.15. p<.OSj. but not within the one exposure condition (&,&,
l,xp=0.35. ~l,,w,1..p=0.36. t=0.63. p>.5). Hence high perceptual fluency had a greater impact on affect only
when the target words were repeated five times. Within the high perceptual fluency condition. five
exposures resulted in a significantly higher mean compared to one exposure (&ghpkp=0-47+ bP. 1,p=0.35. t=?.JO. p<.05). but within the low perceptual fluency condition. no such difference was found
~,,kxp=0.27. &1,p=0.38. 14 .2 1. pz.5). In other words. when perceptual fluency was low. repeated
exposure made little difference. (Please see Figure 7.)
Insert Figure 7 about here.
One ofthe ubjectives of this study was :o find out what the relationship between conceptual and
perceptual tluency might be in generating the mere exposure effect. The perceptud fluency x conceptual
fluency intermion was not significant. suggesting thatthe two have independent influences (Please see
Figure 8.) Further analyses in the form of separate two-way ANOVAs within each level of exposure were
conducted for a closer look at the relation. At five exposures. there were significant main effects of
conceptual fluency (F ( 1,107)=5.40, pc.05) and perceptual fluency (F( 1,107)=5.40, p<.05) and a non-
significant interaction effect. which again points to independence between perceptual and conceptual
fluency. At one exposure, however. none of the main or interaction effects was significant (Fsc0.5, ps>.4).
Insert Figure 8 about here. -------------"-------------------_------*-
Interestingly. when the affect measure was administered after the recognition measure, a three-way
ANOVA of the affect data did not result in any significant effects (Fx1.3. ps>.20). The same analyses
discussed above, including the simple main effects tests (tscl.4, ps>.l) and the two-way ANOVAs for each
level of exposure (Fsc1.7. ps>. 1). similarly failed to ttirn up any significant effects.
6.6.2 Recognition
When the affect measure preceded the recognit,on measure. a three-way ANOVA for recognition
yielded four significant effects. These included n significan~ main effect of perceptual fluency
(F( 1.2 L5)- 106.727. p 4 0 0 I ). with means in the high pcrxptual tluency condition higher than in the low
perceptual tluency condition (&,0p=0.50. &,,=0.19). a main effect of conceptual fluency (F( 1.2 Is)= 1 1.53.
pc.001), with the means in the high conceptual fluency condition again higher (&, , eO. JO. &,,=0.29).
There was also significant main effect of exposure level (F(1.215)=3?.22, p<.0001). Words presented five
times were more likely to be recognized than words presented just once (&,,=0.43, &,xp=0.26). The last
significant effect was for the perceptual fluency x exposure level interaction (F( 1.2 l5)=8.19. p 4 0 5 ) . Please
see Table 2 for the relevant means. With the exception of the perceptual fluency main effect. the effects
found here are the same as those found for affect.
Simple main eftkc ts tests were agaiil conducted on the perceptual tluency by exposure level
conditions and results indicate that when target words were presented five tlmes, the high atd low
perceptual fiuency conditions were significantly different h m each other (l(hlghpkxp=~.6% &owpkxp~.?3.
t=8. LO. pc.Wl).The same was true when target words were presented once (&,plcxp=0.37. sOwp ,,,=O. 14. t=5.67. pc.0001). There was also a significant difference between the five and one exposure
conditions when perceptual fluency was high (Gap Lxp=0.63. &,,,,~,~,~~=0-37, t=5.33. pc.000 1). When
perceptual fluency was low. the difference was smaller, but nonetheless still significant (Xl,,,zenp=0.23,
The hnal analysis is a two-way .ANOVA for each level of exposure. At one exposure, there was rt
main effect of conceptual fluency (F( 1.99)=4.15. Pe.05). a main effect of perceptual fluency
(F( 1.99)=23.78. p<.000 1 ) and a non-significant interaction effect. At five exposures. however. the main
cffec t of conceptual fluency was not significant (F( 1.99)= 1.68. p>. 1). although the main effect of perceptual
fluency was (F(l.99)=84.54, p<.0001). The perceptual by conceptual fluency interaction was again not
significant. (Please see Figure 12.) This absence of an interaction was consistently found in all of the
analyses performed.
Insert Figwe 12 about here.
The results thus far show that the independent variables exert very simiIu effects on both recognition
and affect. the exception being the case where the recognition measure came first and no effects of fluency
or repetition on affect were found. This suggests that both j~dgements may be based on similar processes.
We examine this by analyzing the relationship betweea affect and recognition. The phi coefficient was
computed for each order condition. based on frequencies in 3 2 x 2 [affeit(pleasanr/neutral) by recognition
(old/new)J conrmgency table. and the significance ~f the ~ssaciation was rested using a zL lest. It was found
that when the at'fect measure came t'irst. therz was a significant positive association between affect and
recognition (xL( 1 .N= l3U)=2O.? 1. pc.000:. i =. 1 3). When the recognition measure came tirst. there was o
smaller but still significant positive association found ( X 2 ( I.N=lX8)=8.18. p<.M)5, 4 =.O8). Correlation
analyses were also run for each order condition, using the proportion of 'old' responses and the proportion
of 'pleasant' responses for each subject. When the affect measure came first. there was it significant positive
correlation between affect and recognition (r=.22, p<.OO I). When the recognition measure came first. the
correlation was not signrficant (r=.09. p>.l).
While most of these results indicate a positive relationship between affect and recognition. the results
should be interpreted with caution since the estimate of phi is small but made significant by a large number
of observations. Hence, although the results from analyses of variance show similar effects on both affect
and recognition. measures of association were not as strong as might be expected if both judgements were
made on the basts of tluency.
6.7 Discussion
1 begin first with a discussion of the results for affect. With respect to perceptual fluency. the
experiment replicates previous studies in showing that the fluency experienced in processing the perceptual
features of stimuli as the result of previous encounter can lead to the mere exposure effect. This effect of
perceptual fluency. however, was found when the stimuli were presented five times but not when presented
just once. It therefore appears that it takes several exposures for fluency to be sufficiently enhanced to
produce the mere exposure exposure effect. at lest in the case of meaningful words. This is in contrast to
Obermiller ( 1985) where the mere exposure effect was obtained with a single exposure. The difference in
findings may owe in part to the different target stimuli used. Melodies were used in Obermiller ( 1985) and
perhaps because melodies rue defined by their perceptual properties. perceptual tluency becomes a more
potent factor in influencing affect. Also. as the simple main effects tests showed, when perceptual fluency
was low. liking was not enhanced through repeated exposures. Repetition only had an effect with high
perceptual tluency words.
The experiment also shows that conceptual fluency. like perceptual fluency, can underlie the mere
exposure effect. something that as yet has not been extensively demonstrated in the literature. Words that
had the same meaning as ones presented earlier were more likely to be found pleasant.
The data were also submitted to two-way ANOVAs within each level of exposure. The results
reinforce earlier conclusions in showing that effects of fluency were evident only at exposure level of Eve.
There were main effects of conceptual and perceptual fluency, but no interaction effect. A significant
interaction between the two was also absent when a three-way ANOVA was performed. Given these resulrs.
it sppem that conceptual and perceptual fluency represent independent bases for responding. The
independent and additive nature of the two sources of fluency conforms most closely to the relationship
depicted in Figure 6b.
Interestingly, the results discussed above hold only when the affect measure preceded the recognition
measure. No mere exposure effects were found in any of the conditions hen the recognition measure came
first. It thus appears that the order of dependent measures administered is crucially imponant or that fluency
effects are sens~tive co temporal delays. This finding is not puticulariy weIl-explained by the three accounts
of the mere exposure effect discussed. The oniy prec~ndition that the non-specific activation account
rpecifies for the finding of the mere expoaure effect is that the dimension being judged be a relevant one and
therefore makes no predictions with respect to temporal delays or intervening measures. The perceptual
tluencylmisatuibution model says that the effects of mere exposure are mitigated when the subject becomes
aware of the relationship between the exposure phase and the judgment task and consequently discounts the
experienced fluency. It may be argued that having the recognition measure inserted between presentation of
the stimuli and the affect task makes this relationship saiient to the subject- That is, since the very nature of
the recognition task makes explicit the role of the exposure phase, this awareness carries over to the affect
task. making it more likely that the fluency wili be discounted. However, this would make a more
compelling argument had the measures been administered in a blocked fashion, with the recognition task for
a11 of the words preceding the affect task for the same set of words. Recall that in the present experiment. the
measures were administered back to back for each word. Since the recognition and affect measures
alternated for the entire judgment phase. subjects in rhe recognition first condition should have been no
more likely to be privy to the potential effects of the exposures on their affect judgment than subjects in the
affect first condition. Finally, the element of surprising or unexpected fluency is central to the discrepancy-
attribution account. It may be argued perhaps that having the recognition task before the affect measure
leads subjects to attribute unexpected fluency to having seen the word earlier. Consequently. when it comes
to the affect measure, the fluency is already accounted for and no longer attributed to liking. This. however.
implies that an order effect ought to be found for recognition as well. The analyses found no evidence for
this.
We suggest that there may be a general unwillingness to attribute fluency to multiple causes. along
the lines of the notion of misguided parsinlony, or the unwillingness to ascribe the same consequence to
multiple causes (cf Nisbett & Ross, 1980: Whittlesea, 1993). According to this principle. a subject
experiencing fluency when processing a stimulus will unlikely think that the fluency is due to hidher liking
tbr the stimulus and to having encountered it before. at the same time. Instead, helshe might attribute it to
the aspect of the stimulus made salient by the first measure administered. Fluency will be attributed to any
measures that follow with a lesser likelihood. This explains why tluency had an effect on affect only when
the affect judpmrnt came before the recognition judgement. This brings us to the question of why tluency
had an effect on recognition irrespective of the order of measures. Recognition differs from affect. in that
there is an element of correctness to recognition performance. Subjects may be less willing to risk being
wrong by ignoring feelings of familiarity engendered by flilency and hence more consistently rely on
tluency in making judgements.
The results for subjective recognition were much more robust and consistently suong effects were
found regardIess of whether the recognition test preceded or folIowed the affect task. Except for the main
effect of perceptual fluency, the anaIysis of variance procedure yielded effects remarkably similar to that
found for affect. ~ k s u r e s of associatim rho* pos~tive hut weak dependence between affect and
recognition within trials. This. together with the analyses of variance results, suggests that affect and
recognition are dependent on only partly similar processes. Wile subjects may have relied on fluency to
some extent in making both affect and recognition judgements. alternative bases for responding may have
been avaihble to them. In the case of recognttion, for instance, subjects may have been reliant on
recollection or z directed search of memory rather than familiarity in making judgements (Mandler, 1980).
As an aside. it should be noted that hits are represented by the means in the h ~ g h conceptualhigh
perceptual tluency cells. The means in all other cells represent false alarm rates. bevine in mind that
subjects were required to respond 'yes' only when the word in question corresponded both in meaning and
form. to one seen in the exposure phase. Hit rates ranged between 31% and 67, while false a l m rates
ranged between 10% and 65%. (Please see Tables 1.2 and 3 for these numbers). It would seem that
recognition performance was rather poor. considering some words were repeated five times. This should not
be particularly surprising given the short exposure times for each sentence, the fairly large set of sentences
presented and the fact that subjects were asked to recognize words that could have had appeared in any part
of the sentences seen at exposure. Given that each sentence was six to nine words long, this makes for a
luge number of words to attempt to match target words to.
The fairly high rates of false alarm in the high perceptuaUlow conceptual and low perceptualhigh
conceptual tluency conditions are reminiscent of studies in false memory, or memory for items or events
that did not occur or occurred differently (Roediger & McDermott, 1995; Underwood, 1965; Anisfeld &
Knapp. 1968). Underwood ( 1965), for instance. found that on a recognition task where subjects were to
indicate i f a word had been previously presented in a list, they were likely to falsely recognize words that
wele asswiatixiy dated to prev~ously presented words. This result is often interpreted in terms of a
sprtxding activatton model of memory (Andersen & Bower, 1973; Collins & Loftus, 1975) where activation
of targets words at exposure spread to other related words in memory. Since words in the high
perceptuaVlow conceptual and low perceptuaVhiph conceptual fluency conditions were related to previously
presented target words. the high false alum rates may be similarly explained.
This study was conducted with the main objective of testing the proposition that both conceptual and
perceptual fluency can be antecedents of the mere exposure effect. The experiment was therefore not
designed to provide strong tests of competing accounts. but nonetheless does shed light on aspects of the
accounts that may or may not be supported. First of ali, it is clear that fluency effects are not limited to
perceptual fluency and the perceptual fluencytattributional model would therefore do well to include
conceptual fluency as another basis for the mere exposure effect. The same is true of the nonspecific
activation account. the key points of which are very much couched in terms of activation of the perceptual
features of the stimulus in memory. The discrepancy-attnbution account is the only one of the three that
explains the mere exposure effect in the braader sense of processing fluency. None of the accounts offers a
ready explanation fnr the order effect found. in pivticula~ !hat the mere exposure effect was evident onIy
when the affect measure preceded the recognition measure but not when it was preceded by the recognition
measure. The accounts, 3s they have been developed chcis far, are focused on relating fluency or activation to
affect and do not address moderating influences such as those that have to do with the measures themselves.
Hence. as mentioned earlier, the order effect found here isn't adequately explained and suggests the need to
extend the accounts to provide more definitive answers to the following questions: Can fluency experienced
when processing an item be attributed to multiple causes and hence detected on muItiple measures
completed by the same person? If not, wou1d fluency be attributed to the first measure administered, or
wuuld fluency be attributed to just one measure but with the first measure enjoying no particular advantage?
This leads to the next question of whether fluency is more likely to be attributed to certain properties of the
stimulus or experimental context than others. From a marketing perspective. these questions are important
because attempts at creating the son of fluency that might potentially lead to enhanced liking would be for
nought if that fluency is attributed to something else instead.
The number of exposures examined in this study was limited to one and five and there was sorne
indication that the effect af repetition on affect differs for perceptual fluency and conceptual. Over a larger
number of exposures. this difference may become more readily apparent. Numerous studies have shown that
affect takes a downturn or plateaus after sorne number of repetitions. the likely resuit of boredom setting
~ n t Stanp 8 O'ConneII. 1974: Zajonc et a1.. 1972; Crafidall. Montgomery & Rees. 1973; Kail & Freeman.
1973). Conceptual and perceptual fluency may perhaps differ ~4 th respect to when that apex or plateau is
reached. the increment in affect with each exposure. or perhaps the maximum affect attained. Results from
an investigation into these issues would have interesting implications for advertising execution. Most ad
repetitions within a given campaign are of the exact repetition. or high conceptual fluency-high perceptual
fluency. son. The additive nature of conceptual and perceptual fluency found here implies that the use of
exact repetition has the advantage of leading to the highest affect level, compared to repetitions that involve
conceptual or perceptual variations. Over a iarger range of exposure levels, we may be able to observe
perhaps that exact repetition generates an affect-repetitions curve that reaches its apex much sooner than say
repetitions that are perceptual variatlom around the same conceptual theme. Given the large number of
exposures typical of advertising campaigns. an execution strategy that avoids or delays that downrum in
affect would be of value.
7 Chapter Seven
What is common to a11 three experiments reported in this dissertation is the low involvement nature
of processing at time of initial exposure to stimulus. Experiments 1 and 2 presented target brand names to
subjects as they were engaged in some other task. such that it was possible to devote but minimal resources
to the processing of the brand names. Experiment 3 differs in that target words were presented in focal
attention, but nonetheless, these words were embedded in sentences that were presented fairly rapidly.
hampering attempts to extensively process or elaborate on the stimuli. The levels of involvement
investigated here. though lower than that typical of most consumer behaviour studies, does bear similarity in
reality to many ad encounters. and allows us to better understand behaviour in a wider range of contexts.
In experiments one and two. it was found that even with low attentional resources devoted to
processing the brand names, subjects who received multiple exposures were more likely to generate the
brand names on implicit memory tests. This despite subjects having little explicit memory for the brand
names. These results underscore the need to incorporate memory measures beyond the traditional
recognition and recall tests so often used in investigations of consumer memory. To the extent that there is
over-reliance on measures of the direct sort, we may be missing out on pieces of the puzzle in our attempts
to understand consumer memory. Moreover, it auggests a re-thinking of bays of evaluating admtising
effectiveness to include indirect measures. A case In point is the skepticism that has surfaced of late
regarding the effectiveness of web advertising. The point has been made that click-throughs are at a dismal
rate of 0.5% and that it is much too easy to ignore ads that one encounters while browsing the web. While
this may indeed be true, the conclusion need not necessruily be that such ads are ineffective if there is in fact
implicit memory for brand information that may have an influence on one's purchasing behaviour. Future
studies should perhaps also look to see if implicit and explicit memory may be differentially relied upon in
different types of purchase occasions. Implicit memory may play a bigger role in the case of a consumer
making a purchase of low monetary value and little import, who may make a choice from among brands that
readily come to mind. .4 consumer purchasing a bigger ticket item. on the other hand, may be more reliant
on retrieving from memory brands that hdshe had previouly seen an ad for, heard abu t from friends and
family. or perhaps purchased before.
In experiment three, it was tound rhat liking h r test words increased after multiple exposures to
words that were similar in meaning or form w the test words. This implicates perceptual and conceptual
tluency as causes of the enhanced affect. The prevailing view that tluency underlies the mere exposure
effect, unlike earlier theories that appealed to independent affective and cognitrve systems (Zajonc, 1980) or
to reduction in uncertainty with repeated exposures (Sawyer, 1981). has the advantage that it makes attempts
to increase liking more actionable, from a marketing standpoint. Beyoud simply using repetition,
practitioners can look to other means of creating the sort of tluency that leads to enhanced affect (e.g.
Whittlesea. 1993). In addition, since fluency may be attributed to dimensions other than affect, there may be
situations in which it would be advantageous to have fluency attributed to some aspect of the brand other
than affect. For instance. a brand that enjoys a high level of liking among consumers may benefit from
fluency that's attributed to the brand's prestige or other similar beliefs about the brand. At the same time. the
marketer may want to avoid having effects of fiuency accrue to competitive products that may have
packaging. advertising or brand names that are similar perceptually or conceptually to hidher managed
brand. Thus. there are certain benefits to understanding ( I ) the various means of generating fluency as well
as (2 ) the conditions that may lead one to attribute tluency to a particular dimension or aspect of a brand and
not another.
The results of the three experiments are consistent in showing that stimuli encountered under low
involvement conditions can affect us memorially and affectively. Nonetheless, one aspect of the findings
appears to be contradictory and that is that repetition had no effect on liking for target brands in experiment
two. while it influenced affect judgment in experiment 3. The stimuli and the dependent measures used were
very different, which may have contributed to the difference found. Rating scales were used in experiment
two whereas a forced choice pleasantness judgment task was used in experiment three. Reat brand names
were used in experiment two whereas words were used in experiment three. In the latter case. it is possible
that the use of real brand names that are familiar to subjects meant that they had well-formed. pre-existing
attitudes towards the brand names prior to participating in the study. The three additional exposures
provided in the lab would be but three among many other exposures encountered in the real world and are
hence unlikely to change liking towards the brand. Wtth words. on the other hand. people are less likely to
bring with them well-formed attitudes. allowing tluency to be more readily attributed to their affective
reaction to the words.
In terms of theory development. accountirtg fur implicit memory effects and for the mere exposure
effect has progressed dong rather independent paths. If as Seamon et al. ( 1995) and others have suggested.
the mere exposure effect is in fact just a mantfestation of implicit memory, then there is a need to further test
that proposition. If the results should support that contention. findings in both implicit memory and mere
exposure should be explicitly accounted for within a unified framework. Further findings that may need to
be accounted for within that framework come from studies investigating the truth effect (Hasher, Goldstein
& Toppino, 1977), which refers to the greater likelihood of j udging a statement to be true following repeated
exposures. A recent study by Reber and Schwan (1999) showed that statements were judged to be true with
greater likelihood when they were presented in a colour on a computer screen that made them highly visible,
implicating perceptual fluency as the likely cause of the tntth effect. There, therefore, appears to be a
confluence of evidence pointing to fluency as the antecedent for a variety of phenomena that need to be
accounted for by a generalized model of information processing.
BIBLIOGRAPHY
Anand. Punam and Bnan Stemthal ( 199 1). "Perceptual Fluency and Affect Without Recognit~on." Memory and Coenition. I9(May). 293-300.
Anderson, John Robert (1983). "A Spreading Activation Theory of Memory." Journal of Verbal Learnine: and Verbal Behavior. 22(June). 26 1-295
and Gordon H. Bower (1973), Human Associative Memory, Washington. D.C.: V.H. Winston.
Anisfled, Moshe and Margaret Knapp (1968). "Assoc~ation. Synonymity. and Directionality in False Recognition." Journal of Experimental Psvchology, 77(2). 17 1 - 179.
Banks. William P. (1996). "Korsakoff and Amnes~a" Consc[ousness and Comition: An International Journal, S(March). 22-26.
Berlyne. D. E. (19701, "Novelty. Complexity. and Hedonic Value." Perce~tion and Psycho~hvsics, 8(November). 279- 286.
. Moms B. Holbrook and Debra Stevens (1988). 'The Formation of Affective Judgments: The Cognitive- hffectlvc Model Venus the Independent Hypothesis." Journal of Consumer Research, 15(December). 386-391.
Blimton, Tcrcsa A. ( 1989). "Investigating Dissoc~arions Among Memory Measures: Support for a Tmsfer-Appropriate Proccss~ns Framework." Journal of Experimental PsvchoIoev: Learning. Mernorv and Comition, 15,657-668.
Bornstein. Robert F. ( 1989). "Exposure and Affect: Overvicw and Meta-Analysis of Research. 1968-1987," Psvcholoe~cal Bullet~n. 106(Septcmber~. 265-289.
and Paul R. D'Agostrno ( 1992). "Stimulus Rscognltion rtnd thc Mere Exposure Effect." Journal of Pcrsondiry and Socral Psvcholo~y. 63IOctobcr) 545-552.
and Paul R. D' Agosttno ( 1 994). 'Thc Attnbutron and Discountmg of Perceptual Fluency: Prelimnary Tests ofr! Perceptual Fluency/Attnbut~onal Model of the Mere Exposure Effect." Social Comtion, 12tSumer). 103- 128.
Collins. Allan M. and Elizabeth F. (1975). "A Spreading-Activation Theory of Semantic Processing," Psvcholaeical Review, 82(November). 407-128.
Cnik. Fergus I. M. (1983). "Selective Changes in Encoding as a Function of Reduced Processing Capacity." in C K , eds. F. Khx. J. Hoffman, & E. van der Mect. Amsterdam: North-Holland 152- 161.
and Mark Byrd ( 1982). "Aging and Cognitive Deficits: The Role of Attentional Resources." in Aaing and Conn~tive Processes, eds. F. I. M. Cnik and S. E. Trehub. New Work: Plenum 191-21 1.
and EndeL Tulving. ( 1975). "Depth of Processing and Lhe Retention of Words in Episodic Memory," Jaurn J of Exxrimcntal Psvcholoev: General, 125(2). 195-200.
Crandall, James E. ( 1970). "Preference and Expectancy Amusd: Further Evidence," w, 83(0ctober), 267-268.
. Victor E Montgomery and Willis W. Rces (1973). "Mere Exposure versus Familiarity wth Implications for Response Competition and Expccmcy Arousal Hypotheses." J-ofGenenl 88 ~lmuar).). 105-120.
Crandall. S. Chnstim ( 1985,. 'The Liking of Foo& as a Result of Exposure: Eating Doughnuts in Alaska" Journal of Soc~rtl Psvcholom:. 1 Z( Xpnl ), 187- 194.
E~ch. Enc [ 1984). "Memory h r Unattended Events: Re~em9enng Wtth and Without Awareness." Memow and Cogn~tron, 12. 105- 1 i I .
Gordon. Peter C. and Keith J. Holyoak ( 1983). "Irnpliclt Learning and Genenlimtion of the "Mere Exposure" Effect." Journal oi Personalitv and Soclal Psvchology, 35(Septernber). 492-500.
Graf. Peter and George Mmdler (1984). "Actlvacion Makes Words More Accessible. but not Necessarily More Remevhle." J O U ~ V ~ of Verbal Learnina and Verbal Behaviour, 23.553-568
and Daniel L. Schacter c 1985). "Implicit and Explicit Memory for New Associations in Normal and Amnesic Subjects." Journal of Experimental Psychology: Laming. Memory and Cognition, 1 l(Ju1y). 501-5 18..
Greene. Robert L ( 1998). "Spacing Effects on Implicit Memory Tests." Journd of Experimental Psvcholom: L r a m i n ~ Memorv. and Coenition. 16(November), 1004- 101 1.
Grecnwald. Anthony G. and Clark Leavitt 11984). "Auhence Involvement in Advertising: Four Levels." Journal of Consumer Research, 1 1 (June), 58 1-592.
Grush. Joseph E. (1976). "Attitude Formation and Mere Exposure Phenomena: A Nonartifactual Explanation of Ernp~ricd Findings." Journd of Persondit~ and Social Psvcholopru. 33(March). 28 1-290.
Hamson. Albert A. ( 1968). "Response Competition, Frequency. Exploratory Behavtor. and Liking." Journal of Pcrsonal~tv and Social Psvcholo~;41.9(4), 363-368.
Hasher. Lynn. David Goldstcin. Thomas Topplno ( 1977). "Frequency rind the Conference of Retercntial Vdidity." Journal of Verbal Learning and Verbal Behavtor. 16( Febru;~ry I . 107- 1 12.
Hawklns. Scurt md Stephcn Hoch 1 1992). "Low Invol~emcni Lzming: Memory Without Evaluation," Journal of Consumer Research. 19tScptember). 2 12-225.
Hilt. Winfred F. ( 1978). "Effects of Mere Exposure on Preferences in Nonhuman Mammals." Psycholoe~cal Bullet~n. S5( November). 1 177- 1 198.
Houston. Michael J. and Michael L. Rothschild ( 1977). "A Paradigm for Research on Consumer Involvement." unpublished working paper. Graduate School of Business. University of Wisconsin. Madison.
Jacoby. LYry L. and Mark Dallas ( 198 1). "On the Relationship Between Autobiographical Memory and Perceptual Learning," Journal of Exmimental Psvcholoev: General. 1 10.306-330.
and Colleen M. Kclley ( 19871, "Unconscious Influences of Memory for a Prior Event." Pcrsonalitv and Social Psvcholom~ Bulletin, 13(Septemberl, 3 14-336.
- . L o r m e G. Altan, lane C. Collins and Linda K. Lmvtll( 1988). "Memory Influences Subjective Expcriencc: Noisc Judgments." Joumal of Exwrimental Psvcholoev: Lcarnin~. Memorv and Comition. I4April). 230- 247.
, Colleen M. Kelley and Jane D&UI (1989). "Memory attributions." in Vxieties of Memorv and Consctousness: Essays in Honour of Endel Tulvlng. eds. Henry L. Roediger I11 and Fcrps I. M. Cnik. Hillsdalc, NJ: Erlbaurn. 39 1-422.
. Vcn Woloshyn and Colleen M. Kelley (1989). "Becomtng Famous Without Being Recognized: tinconscious Influences of Memory Produced by Dividing Attention." Journd of Ex~rimental Psvcholoev: General. 118(Junc). 1 15-13.
Juet. Pierre r 1901). "L'amneste et la dissoaa~~on des souvcnlrs par I'emotion [Amnes~o and the dissoctat~on of memoncs by emotion]," Journal de Psvcholoete Normale et Patholoe~aue. 1,417-453.
Jmszewsk~. Chrrs ( 1988 r. "Preconscious Processing EtTect: The Independence of Attitude Formarion and Conscious though^" Journal ol' Consumer Research. 15 (September). 199-209.
-, Chns ( 139Ob). 'The Intluence of Nonanended Material on the Processing of Advcnislng Claims." Journal of Markrrt~ng Research, 27(Au@), 263-27?.
. Chns ( 1993.1. ' Prcattentive Mcre Exposure Effects." Journal of Consumer Research, 20(December). 376392.
Johar. Gita V. (199%. Tonsunkr Involvement u d Decepuon from Implied Advertising CIaims." Journal of Marketing Research. 32(August:. 267-279.
Katl. Robert V and Harvey R. Fmman 1 1933). "Sequence Redundancy. Rating Dimensions and the Exposure Effect." Memorv and Comition, 1 (October). 554-458.
Korsakoff. S. S.. William P. Banks (Trans,, Sandra Jade Karam (Tms) (1996). "Medico-Psychoiopical Study of a Memory Disorder." Consciousness and Comition: An lntemationd Journal, S(March). 2-2 1.
Knshnan. H. Shaker and Siewm Shapiro (1996). "Comparing Implicit and Explicit Memory for Brand Names From Advertisements." Journal of Exwrimental Psvcholoev: Amlied, 2.147-163.
Krugman. Hrrben E. ( 1965 j. 'The Impact of Televis~on Xdvcrt~srng: Larnrng Without Involvemrnt." Public Ouinion Oumerlv. 29(Fdl). 349-356.
Kunst-Wilson. WiIliam R. and R. B. Zajonc ( 1980). "Affective Discnrninatron of Stimuli that Cannot be Recognized." Science. 207tFebruary). 557-558.
k a r u s . Richard S. 1982). ''Thoughts on the Relations Between Emotion and Cognition," Amencan Psvcholoeist, -
37(Septemkr). 10 19- 1024.
L~ght. Leah L. ( I 987), "Preserved Implicit Memory in Old Age," Practical Aspects of Memorv: Current Research and Issues, 2.90-95.
. Asha Singh and Janet L. Capps ( 1986). 'The Dissociation of Memory and Awareness in Young and Older Adults." J(, 8.62-73.
Lindsay. D. Stephen and Colleen M. Kelley (1996). "Creating llIusions of FarniIiarity in a Cued RecaII Remembcr/Know Paradigm." Journal of Memory and h ~ u a s , 35(April), 197-2 1 1.
Lynch. John G. and Thomas K. Smll(1982), "Memory and Attentional Factors in Consumer Choice: Concepts and Research Methods." Journal of Consumer Research. 9(June), 18-37.
Macinnis. Debonh I. and C. Whan Park (199 1). 'The Differential Role of Characteristics of Music on High and Low Involvement Consumers' Processing of Ads." Journal of Consumer Research. 18(Sepcember). 16 1 - 173.
Maheswarm. Duninj md Joan Meyers-Levy ( 1990). 'The Influence of Message Framing and Issue Involvement." Journal of Marketme Research, Z7(August). 361-367.
Mandler. George ( 1980). "Recognizrng: 'fir: Judgment of a Previous Occurrence." PsvchologrcaI Sc~encr.. 87(May). 252-27 1.
. Yoshro Nakamura rind Brllre fo Shebo Van Zmdt ( 1987). "Nonspecrfrc Effects of Exposure on Stlmuli That Cannot be Recog~~zcd." Journal of Exnenrncntd Psvcholorrv: Lcarntn~. Memow and Com~tron, 13(0ctobcr~. 64648.
Masson. Michael E. J. md Colin M. Mackod ( 1992). "Reenacting the Route to hterprctat~on: Enhanced Perceptual [dentitication Without Prior Pcrceprion," Journal of Ex~crnrncntal Psvcholo~v: Gcncml. 12 I ( 2 i . 145-176.
blrnliud. Paul W. Sun11 Bhatla Kenneth R. Lord. Peter R. Dtckson and H. Rao Unnava ( 199 I ) . "Picturc-based Persuasion Processes and thc Modemtlng Rolc of Involvement." Journal of Consumer Rcsearch, 18(Iune). 92- 107.
Mitchell. Andrcw A. ( 1983). "Cognitive Processes Initiated by Exposure to Advenising." in Information Processing Research in Advertising, ed. R. Harris. Hillsdde. NJ: Erlbaum. 13-42.
Moray. N. ( 1959). Attention in Dichotic Listening: Affective Cues and the Influence of Insuuctions. Quarterly Journal of E~perimental Psychology, 11.56-60.
Morris. C. Donald, John D. Bmsford and Jeffery I. Franks (1977). "Levels of Processing versus Transfer Appropriate Processing." Journal of Verbal Learntn~ and Verbal Behavior, 16.519-533.
Mulligan. Neil W. (1997). "Attention and Implicit Memory Tests: The Effects of Varying Attcntionai Load on Conceptual Pnrn~ng," Jvlemorv and Cormition. 25( 1 ), I 1-17.
(1998 I. 'The Role of Attention During Encoding in Implicit and Explicit Memory," Journal of Exwrimental Psvcholom: Learning. Memorv, and Comtion, 2 4 I) , 27-47.
Nedungadi. M a s h ( 1990). "Recall and Consumer Considemon Sets: influencing Choice without Atering Brand Evaluations." Journal of Consumer Research. l7t Decernberj. 263-276.
. Andrew A. Mitchell and Ida E. Berger (1993). "A Fmework for Understandmg the Effects of Advertising Exposure on Cho~ce." rn Advert~srne Enwsure. Memop and Choice. ed. Andrcw A. Mitchell. Hillsdde. NJ; Erlbaum. 89- 1 111.
Nisktt. Richard md Lee Ross ( 1980). Human Inference: Siratcnles md Shortcornin~s of Social Judamencs, Engiewood Clrffs. NJ. Prent~ce Hall.
Obcrrmller. Carl ( 1985). "Vxieues of Mere Exposure: The Effects of Processing Style md Repetrtion on Affective Response.'' Journal of Consumer Research. 12(J~ne!. 17-30.
Park, Jong-Won and Mmoj Hat& (l99J), "Memory-based Product Judgements: Effects of Involvement at Encoding and Retrieval." Journal of Consumer Research, 2 i(December), 534-517.
Parkin. Alan J.. Thomas K. Reid and Riccardo Russo (1990). "On the Differential Nature of Implicit and Explicit Memory," Memorv and Co~nition, 18.507-5 14.
(1993). Memorv: Phenomena. Exwnment and Theon, Cambridge, MA: Blackwell.
Petty, Richard E.. John T. Cacioppo and David S c h u m n ( 1983). "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement." Journal of Consumer Research, IO(September). 135-136.
Phm. Michel T. and Marc Vanhuete ( 1997). "Analyzing the Memory Impact of Advertising Fragments." Marketing Letters, S(4). 407417.
Plmr. Patnc~a i 1982). 'The Effects of Mere Exposure on Liking for U b l e Substances." Aowtite, 13(Septemberl. 283- 190.
Reber. Rolf and Sorben Schwart ( 1999). "Effects of Perceptual Flbency on Judgments of Truth." Consciousness and Coernitlon, 8tSeptember). 338-342.
. Piotr Winkielman and Norhn Schwm ( 1998). "Effects of Perceptual Fluency on Affect~ve Judgments." Psvch~logical Science. 9(Jmu;uy). 45-48.
Roediger. Henry L. and Teresa A. Blaxton ( 1987). "Effects of Varying Modality, Surface Fcxures. and Retention Intervd on hrning in Word-Fngment Completion." Memorv and Cognition. 15(5). 379-388.
and Bradford H. Chdlis ( 1992), "Effccts of Exact Repetition and Conceptual Repetition on Frec Recall and Pnmcd Word-Fngment Completion." Journal of Exwrimental Psvcholorrv: Leamine. Memorv and Comttion, ~ ~ ( J M u u ~ ) , 3- 14.
and Kathleen B. McDcmott (1993). "Implicit Memory in Nonnal Human Subjects." in Handbook of Ncuro~svcholow, eds. F. Boller and J. Gnfman. Amsterdam: Elsevier, 8.63- 13 1.
and Kathleen B. 'McDermott (1995). "Creating False Memories: Remembering Words Not frescntcd in bsts." Journal of Exwrimcnta! Psvcholorrv: Learning. Memorv and Comitioq. 2 11July). 803-8 14.
Sawyer. Alan G. & Daniel J. Howard ( 1991 1. "Effects of Omitting Conclus~ons in Advertisements to Involved and Uninvolved Audiences," Journal of Marketing Research, 28(Yovember), 467474.
(198 1 ). "Repetition. Cognitive Responses and Persuasion." in Coenitivc Reswnses in Persuasion. eds. Richard E. Petty. Tom M. Ostrorn. and Timothy C. Brock. Htllsdale. NJ: Erlbaum, 237-261.
Schacter. Daniel L.. C. Y. Peter Chiu and Kevin 9. Ochsner ( 1993h "Implicit Memory: A Selective Review." Annual Review of Ncurosc~ence. 16. 159- 182.
. Jeffrey Bowers and Jill Bcroker ( 1989). "Intention. Awareness and Implicit Memory: The Reuieval Intent~onai~ty Criterion." in ImDllCif%, ed. Stephim Lewandowsky, John C. Dunn and Kim Kmner. Hillsdale, KJ: Erlbaum. 47-65.
Scort B. MrtcKenzle and Richard A. Sprtnp c 1992). "How h e s Motivation Moderate the Impact of Central and Penpherd Prncessinp on Brand Attitudes and ~ntenuons." fuurnd of Consumer Researrh, I81M;lrch 1.5 19-529.
Seamon. John G.. Puicia A. McKenna and Neil Binder ( 19981. 'The Mere Exposure Effkc: is Differentidly Sensitive to Different Jtdprnent Tasks." ~onsciousness and Coanitlo~. 7(March). 85- 102
Seamon, John G.. Pepper C. Williams. Michael I. Crowley. Irene J. ffin Smthrt A. Lanper. Peter J. h e and Dana L Wishenpnd ( 1995). 'The Mere Exposure Effect Is Based on Implicit Memory: Effects of Stimulus Type, Encoding Conditiuns. and Number of Exposures on Recognition and Affect Judgments," Journal of Experimental Psvcholo~v: Learninp Memofi and Cognition. 21 (3 1.7 1 1-72 1.
Sengupta Jrudeep. Ronald C. Goodstein and David S. Boninger ( 1997). "All Cues Are Not Created Equal: Obtaining Attitude Persistence Under Low Involt cment Conditions." Journal of Consumer Research, 23(March). 35 1-361.
Shap~ro. Stewart. Deborah J. MacInnis and Susan E Heckler ( 1993. 'The Effects of Incidentat Ad Exposure on the Fonmtion of Considention Sets," Journai of Consumr Research, 24.94-104.
Sm~th. Michael E. md blarlene Oscar-Berman ( 1990). "Repetition Prirmng of Words and Pseudowords in Divided Attention and in Amnesia" Journal of Exwrimental Psvcholoev: Learnine, Memorv and Comition, 16. 1033- 1032.
Squire. L a q R., B. Knowiton and Gail Musen t 1993). '"The Stmcture and Organizatron of Mcmory." Annual Review of Psvcholoey. U. 453 495.
Stmg. Davrd I. and Edward J. O'Connell ( 1974). 'The Computer as Experimenter rn Social Psychology Research," Behaviour Research LMetfiods and Instrumentation, B(March). 223-23 1.
S z y m s k ~ . Katnn F. and Colin M. MacLeod t 1996). "Mmtpulation of Attention at Study Affects an Explicit but not an [mplicit Test of Memory." Consciousness and Comit~on, 5, 165- 175.
Trciismm. Anne 'M. (19#), "SeIective Attention in Man." British Medical Bulletin, 20. 12-16.
Tulving. Endcl. Daniel L Schacter md Heather A. Stark (1982). "Priming Effects in Word-Fragment Completion Are Independent of Recognition Memory." Journal of Ex~erimental Psvcholonv: Learning, Memorv. and Coenition, 8(4). 33G342.
Underwood, Benton J. ( 1965). "False Recognition Produced by Implicit Verbal Responses." JournaI of Emmimental m. 7O( I), 122-129
Wamngton. Elizabeth K. and L W e i z h t z ( 1970). "Amnesia Syndrome: Consolidation or Reuicval?" Nature, 228, 629-630.
Whittlesea. Bruce W. A. (1993). "Illusions of Fruniliruity." Journal of Exwrimental Psvcholoev: Learning. Memorv and Comition. 19(November), 1235-1 253.
. Lany L. Jacoby and Knsta Girard (19901, "Illusions of Immediate Memory: Evidence of an Attnbut~ond Bars for Feelings of Familiarity mr! Perceptual Quality," Iournd of Memory md Language. 29(Dcccrnber). 7 16-732.
rlnd L~sa D. iVill~rrms ( 1998). "Why Do Strangm Fcct Fam~liar. But Fnends Don't? A Dtscrepany- Attnbut~on Account o i Fecl~ngs of Fam~lrmty." .Act3 Psvcholoe~ca. 98(Apnl). 131-165.
Wilson. Wtll lm R. (19791, "Feeling More Than You Car Know: Exposure Effccts Without Lxming." Journid of Psrsonahtv and Soclal Psvch*, 37(lune,. 8 1 1-82 1
Woltrn. Ckzinus and Arno Pnnsen (1997). "Full versus Divrdcd Attent~on and Implrc~t Memory Performance." Memory md Comtion. 236). 7U-77 1 .
Wood. Noelle L. and Nelson Cowan (1995). 'The Cocktail Pany Phenomenon Revisited: Attention and ~Mernory in the Classrc Selective hstsnlng Procedure of Cherry ( 1953)," Journal of Experimental Ps~choloev: General. 1243). 243-262.
. Michael A. StadIer and Nelson Cowan ( 1997). "Is There Implicit iMernory Without Attention? A Reexamination of Task Demands in Eich's ( 1984) Procedure." Memorv and Comicion. 3 6 ) . 772-779.
Zajonc. R. B. 1 196&). "Attitudinaf Effects of Mere Exposure." Iournai of Person Jitv and Social Ps~chology Monom~hs. 9(2. Pt. 2). 1-27.
( 1980 i. "Feeling and Thmhng: Preferences Need No Inferences," American Psvchologist, 35(Febnrary). 15 1- 175.
(1984,. "On the Pnmcy of Affect." AmencanPsvcholonisf, 39(Febru;lry). 117-124.
and Hazel Markus ( 1982). "Affecthe and Cognitive Factors in Preferences." Iournd of Consumer Research. 9(Scptembcr). 123- 13 1.
, Ph~lip Shaker. Cuol Tams and David V ~ I Kreveld ( 1972). "Exposure, Satlation. a d Stimulus Discrirmnabilq." lournd of Personality and Soad Ps~ch~iogy, ?l(Mruch). 270-280
ENDNOTES
1. A variety of overlapping d~stinctions (e-g., unawarehware. unconscious/conscious. automaticlconuolled) have been used in the psychology literature to capture the distinction between implicit and explicit memory. Because of the prominence of their use in memory research. we repeatedly refer to implicit and explicit memory in drawing conceptual distinctions between them. The terms indirect and direct tests are used throughout our paper. to refer to measures that have been demonstrated in the literature to tap implicit vs. explicit memory, respectively.
2. The pretest that was conducted required subjects to respond with the Arst brand name that came to mind given each of 33 product category cues. To avoid potential floor and ceiling effects. brand names were selected from categories for which there were no dominant responses and for which the number uf brand names generated was not exceedingly large.
3. A separate pretest was conducted to identify product categories and brand names that would be suitable for inclusion in Expe~iment 2.
FIGURE 1
EXPERIMENT 1: EFFECT OF AlTENTlON AND REPETITION ON INDIRECT MEMORY TEST
V --
control no repetition repetition
+ attend + ignore
FIGURE 2
EXPERIMENT 1: EFFECT OF AlTENTlON AND REPETITION ON DIRECT MEMORY TEST
control no repetition repetition
+ ignore
EXPERIMENT 2: EFFECT OF ATTENTION AND REPETITION ON CONSIDERATION SETS
+attend -t- ignore '
control no repetition repetition - -- -----
FIGURE 4
EXPERIMENT 2: EFFECT OF ATTENTION AND REPETITION ON CHOICE
- attend - ignore
0 I control no repetition repetition
--- -
FIGURE 5
EXPERIMENT 2: EFFECT OF ATTENTION AND REPETITION ON CUED RECALL
attend
, -C- ignore
control no repetition repetition
FIGURE 6
EXPERIMENT 3: POSSIBLE RELATIONSHIPS BETWEEN CONCEPTUAL FLUENCY AND PERCEPTUAL FLUENCY
affect I
affect
low hhh conceptual fluency
high
-------------- low
low high conceptual fluency
low high conceptual fluency
low high conceptual fluency
----. low perceptual fluency - high perceptual fluency
FIGURE 7
EXPERIMENT 3: EFFECT OF PERCEPTUAL FLUENCY AND EXPOSURE LEVEL ON AFFECT WHEN AFFECT TASK PRECEDED RECOGNITION TASK
Low perceptual fluency High perceptual fluency
-+ 1 exposure . . -, - - 5 exposures
FIGURE 8
EXPERIMENT 3: EFFECT OF PERCEPTUAL AND CONCEPTUAL FLUENCY ON AFFECT WHEN AFFECT TASK PRECEDED RECOGNITION TASK
--- - -
Low perceptual fluency High perceptual fluency
- - -c - - High conceptual fluency + Low conceptual fluency
FIGURE 9
EXPERIMENT 3: EFFECT OF PERCEPTUAL FLUENCY AND EXPOSURE LEVEL ON RECOGNITION WHEN AFFECT TASK PRECEDED RECOGNITION TASK
Low perceptual fluency High perceptual fluency -----
1 exposure -. -s - - 5 exposures
FIGURE 10
EXPERIMENT 3: EFFECT OF PERCEPTUAL AND CONCEPTUAL FLUENCY ON RECOGNITION WHEN AFFECT TASK PRECEDED RECOGNITION TASK
Low perceptual fluency High perceptual fluency - -- * -- --- - - -+ . . High conceptual fluency +Low conceptual fluency
EXPERIMENT 3: EFFECT OF PERCEPTUAL FLUENCY AND EXPOSURE LEVEL ON RECOGNITION WHEN RECOGNITION TASK PRECEDED AFFECT TASK
I Low perceptual fluency High perceptual fluency -
+ 1 exposure - . -.- - . 5 exposures I
FIGURE 12
EXPERIMENT 3: EFFECT OF PERCEPTUAL AND CONCEPTUAL FLUENCY ON RECOGNITION WHEN RECOGNITION TASK PRECEDED AFFECT TASK
-6 0.58 / 0-52 t I
. * -
+-625 d-c' i i
Low peceptual fluency High perceptual fluency
- - - + - . High conceptual fluency -m- Low conceptual fI uency
TABLE 1
EXPERIMENT 3: EFFECT OF PERECEPTUAL FLUENCY, CONCEPTUAL FLUENCY AND EXPOSURE LEVEL ON AFFECT WHEN AFFECT TASK PRECEDED RECOGNITION TASK
Probability judged pleasant
High perceptual fluency Low perceptual fluency
No. of exposures High conceptual Low conceptual High conceptual Low conceptual fluency fluency fluency fluency
TABLE 2
EXPERIMENT 3: EFFECT OF PERECEPTUAL FLUENCY, CONCEPTUAL FLUENCY AND EXPOSURE LEVEL ON RECOGNITION WHEN AFFECT TASK PRECEDED RECOGNITION TASK
Probability judged old ---- ---- High perceptual Low perceptual
No. of exposures High conceptual Low conceptual High conceptual Low conceptual
1 0 . 4 1 0 . 3 3 0.18 0.10
TABLE 3
EXPERIMENT 3: EFFECT OF PERECEPTUAL FLUENCY, CONCEPTUAL FLUENCY AND EXPOSURE LEVEL ON RECOGNITION WHEN RECOGNITION TASK PRECEDED AFFECT TASK
-- Probabrlity judged old
--- High perceptuai Low perceptual
No. of exposures High conceptual Low ~onceptunl High conceptual Low conceptual
I 0 . 4 9 0.33 0 . 2 4 0 .13
EXPERIMENT 1: CRITICAL BRAND NAMES AND SLOGANS
"HEAD & SHOULDERS. The prescription for healthy hzir."
"V.ASELNE ~ T E N S I V E CARE. Give yoLr skin the drink ~t needs."
"DENTYNE. Because i t 's cool to be fresh."
"Wake up to the arornatlc wonder of NESCAFE."
APPENDIX B
EXPERIMENT 2: CRlTlCAL BRAND NAMES AND SLOGANS
'TIMEX watches. For function that does not compromise style "
'Take your best plctures with a CANON camera."
EXPERIMENT 1 : VISUALLY-PRESENTED TEXT PASSAGE
Advertising message 1 : Art: you one of the millions of people who are caught in the vicious cycle of not getting the sleep you so desperately need? And does it seem like no matter what you try, you can1 find reliet? It's time to wake up and smell the coffee. Because there is something you can do. See your doctor. You11 learn what things to stop doing. and what things to start doing, if you want to get back to sleeping more like the way you were used to. In some situations, your doctor may prescribe a sleep medication that you can use on a short-term basis when you have trouble sleeping. One of these sleep medications is called Ambien. It, along with appropriate lifestyle changes. may help you get back to your normal sleep pattern and routine. Ambien will usually put you to sleep quickly. You can rake it and not have to wait hours for it to work. Generally, in about twenty minutes, you11 fall asleep. Ambien is effective. It may help you get back to a natural way of sleeping. Also, Ambien leaves your system quickly, so you're less Iikely to be tired the next day and more likely to feel rested and refreshed in the morning. Your doctor will make you aware that you should only take Ambien when you can devote a full 7 to 8 hours of sleep. You should not engage in activities requiring complete alertness, such as driving a car. until you know how you react to the medicine. The doctor may also alert you to some side effects that may be associated with Ambien: drowsiness. dizziness. light- headsdness, and difficulty with coordination. Some people may also experience confusion. memory problems. and change in behaviour. Sleep medicines may cause dependence. especially when these medicines are used daily for longer than o few weeks or at high doses. So. if you're having a sleep problem. you may not have to live with it anymore. See your doctor, and find out whether Ambien may be appropriate for you. Isnl it time for you tci go to sleep? Sleep. It doesnl have :o be a dream.
-4dvrrtising message 2: Tabletop radios are popular for thelr convenience and small size. But their sound quality [eaves much to be desired. No one really expects high-fidelity sound from a radio. Until now. Bose presents the Wave radio. It's the one radio acclaimed by leading audio critics. Because it3 the one radio that delivers big, rich, lifelike stereo sound plus a small. convenient size. We think the Wave radio is the best-sounding radio you can buy. And audio cr~tics agree. Business Week named the Wave radio a "Best New Product of 1994". Popular Science gave it a prestigious "Best of What's New" award. The key is our patented acoustic waveguide speaker technology. Just as a tlute strengthens a breath of air to fill an entire concert hall. the waveguide produces room-tilling sound from a small enclosure. This technology and performance is available in no other radio. Youll touch a button and hear your favourite music come alive in rich stereo sound. Youll hear every note the way it's meant to be heard. The Wave radio fits almost anywhere. So you can listen in your bedroom, living room. kitchen, or any room. Operate the radio from across the room with the credit card- sized remote control. Set six AM and six FM stations, and switch between them at the touch of a button. The Wave radio is available for three hundred and forty-nine directly from Bose, the most respected ndme in socnd. Bose. Better sound through research.
APPENDIX D
EXPERIMENT 2: VISUALLY-PRESENTED TEXT PASSAGE
Advertising message I : Are you one of the millions of people who are caught in the vicious cycle of not getting the sleep you so desperately need? And does it seem like no matter what you try, you can\ find relief? It's time to wake up and smell the coffee. Because there is something you can do. See your doctor. Youll learn what things to stop doing. and what things to start doing. if you want to get back to sleeping more like the way you were used to. In some situations, your doctor may prescribe a sleep medication that you can use on a short-term basis when you have trouble sleeping. One of these sleep medications is called Ambien. It. along with appropriate lifestyle changes. may help you get back to your normal sleep pattern and routine. Ambien will usually put you to sleep quickly. You can take it and not have to wait hours for it to work. Generally, in about twenty minutes. youll hll asleep. Ambien is effec:ive It may help you get back tu a natural way of sleeping. Also. Ambien leaves your system quickiy. so you're less likely to be tired the next day and more ltkely to feel resred and refreshed in the mornlng. Your doctor will make you aware that you should only take Ambien when you can devote a full 7 to S hours of sleep. You should not engage in activities requiring complete alertness. such as drivmg a car. until you k n o ~ how you react to the medicine. The doctor may 3150 alert you to some side effects that may be associated with Ambien: drowsiness, dizziness. light- hcadedness, and difficulty with coordination. Some people may also experience confusion. memory problems. and change in behaviour. Sleep medicines may cause dependence, especially when these medicines are used daily for longer than a few weeks or at high doses. So, if you're having a sleep problem. you may not have to live with it anymore. See your doctor. and find out whether Ambien may be appropriate for you. Isnt it time for you to go to sleep? Sleep. It doesni have to be a dream.
Advertising message 2: Tabletop radios are popular for their convenience and small size. But their sound quality leaves much to be desired. No one realiy expects high-fidelity sound from a radio. Until now. Bose presents the Wave radio. It's the one radio acclaimed by leading audio critics. Because it's the one radio that delivers big. rich. stereo sound plus a small. convenient size. We think the Wave radio is the best-sounding radio you can buy. And audio critics agree. Business Week named the Wave radio a "Best New Product of 1994". Popular Science gave it a prestigious "Best of What's Ned' award. The key is our patented speaker technology. which produces room-filling sound from a small mclosure. This technology and performance is available in no other radio. Youll touch a button and hear yow favourte music come alive in rich stereo sound. Youll hear every note the way it's meant to be heard. The Wave radio fits almost anywhere. So you cm listen in your bedroom. living roam, kitchen. or any room. Operate the radio from across the room with the credit card- sized remoie control. Set six AM and six FM stations, and switch between them at the touch of a button. The Wave radio is avaihble for three hundred and forty-nine directly from Bose. the most respected name in sound. Bose. Better sound through research.
APPENDIX E
EXPERIMENT 3: HOMOGRAPHS, STUDY AND TEST SENTENCES
The second column contains sentences seen in the exposure phase. In the judgment phase. subjects saw one
of the four sentences in each cell of column 3. The b u r tluency conditions - ( 1) high conceptualhigh
perceptual. (2) high conceptuaVlow perceptual. (3) low concrptuaUhigh perceptud. (4) low conceptuaUlow
perceptual - are represented by the first. second. bird and fourth sentences in each cell. respectively.
Homograph I Study Sentence bear I The summer humidity can be hard to
beat The song had a lively beat.
bold
1 bright. 1
He behaved in a manner that was bold.
bright
I t -- I broad i The athlete had shoulders that were broad. I I
1
I I I
He was out to make a quick buck.
I
His excellent grades proved he was
correct / There are prizes for guesses that arc
1 I correct.
I ! 1 He was impatient and easy to cross.
Test Sentence 1. The pain was too much to bear. 2. The pain was too much to endure. 3. While camping he saw a bear. 3. While camping he saw a wolf. I . The audience moved to the music's beat. 2. The audience moved to the music's tempo. 3. The record he set was hard to beat. 4. The record he set was hard to break. 1. She spoke her mind and was bold. 2. She spoke her mind and was daring. 3. The text was printed in bold. 4. - The text was printed in italics 1. The gifted boy was extremely bright. 2. The gifted boy was extremely smart. 3. The sun's rays made the room bright. 4. The sun's rays made the room warm. I . The pathway to the house is very broad. 2. The pathway to the house is very wide. 3. The essay outline was much too broad. 4. The essay outline was much too vague. I . The panhandler was given a buck. 2. The panhandler was given a dollar. 3. The male of some animals is a buck. 4. The male of some animals is a bull. 1. To keep warm he wore a coat. 2. To keep warm he wore a jacket. 3. The wall was given a new paint coat. 4. The wall was given a new paint colour. 1. All answers to the questions were correct. 2. All answers to the questions were right. 3. The document had errors he had to correct. 3. The document had errors he had to delete. 1. His son's bad behaviour made him cross. 2. His son's bad behaviour made him angry. 3. The street was difficult to cross. 4. The street was difficult to ha te .
Loud noises caused the baby to cry.
cue The actor waited for the director's cue.
decline I Interest in the arts is on the decline.
dirty Walking in the mud makes your shoes dirty.
I
draw ] A re-match is necessary if there: a draw.
I
dull i The audience yawned because the talk 1 was dull.
i -- duty I Some goods are subject :o customs duty.
empty I All the shopping left her wallet empty.
fair The cheat was accused of not playing fair.
.4 race car goes very fast.
- project finally czme to a tinish.
i tbo t I Every month he hdd rent to foot.
goal / Becoming a good doctor was his life's
I
1. The touching m o w made her cry. 2. The touching movie made her weep. 3. The Tories adopted a new party cry. 4. The Tories adopted a new party slogan. 1. The coach gave the player a cue. 2. The coach gave the player a signal. 3. The hustler grabbed his cue. 4. The hustler grabbed his rum. 1. The town experienced a population decline. 2. The town experienced a population decrease. 3. The invitation was one she could not decline. 3. The invitation was one she could not refuse. 1. The homeless man's clothing was dirty. 2. The homeless man's clothing was filthy. 3. The con man was up to tricks that were dirty. 4. The con man was up to tricks that were illegal. 1. The competition ended in a draw. 2. The competition endea in a tie. 3. She took classes to learn to draw. 4. - She took classes to learn to sing. 1. The students skipped classes that were dull. * - . The students skipped classes that were boring. 3. The knife's useless because its edge is dull. 4. The knife's useless because its edge is chipped. 1. The purcnase price included import duty. 7 -. The purchase price included import tax. 3. To stand g~md all night was his duty. 4. To stand guard a11 night was his job. 1. After he moved out the apartment was empty. 2. After he moved out the apartment was vacant. 3. He often makes promises that are empty. 4. He often makes promises that are broken. L. A judge should strive to be fair. 2. .4 judge should strive to be just. 3. He took pony rides at the fair. 4. He took pony rides at the zoo. 1. The boxer's reflexes were fast. 2. The boxer's reflexes were quick. 3. His religion requires him to frequently fast. 4. His religion requires hin; to frequently pray. 1. The mokie was thrilling from start to finish. 7,. The movie was thrilhng from start to end. 3. The cabioets had a lacquered finish. 4. The cabinets had a lacouered surface. - -
L-- i
1. The spendthrift had many bills to foot. 9 -. The spendthrift had many bills to pay. 3- He yelped when someone stepped on his foot. 4. He yelped when someone stepped on his toe. 1. They worked hard to achieve the project's god. 2 - They worked hard to achieve the project's aim. 3. Their team finally scored its first goal. 4. Their team finally scored its first point.
kid t lobby
L-- 1 match
I modest
1 novel
pick I pile
The condition of his illness was grave. 1 1. The decision pondered was one that was grave. I 1 i: The decision pondered was one that was serious. He took his secret to the gave.
4. He took his secret to the police. She was very active as a kid. 1. In the stroller sat a cute kid.
In the stroller sat a cute child. He was not one with whom to kid. He was not one with whom to argue.
She eats only meat that's lean. I 1. Months of dieting made him look lean. ! 1. Months of dieting made him look thin. 1 1 3. He's a friend on whom one can lean. I
1 2. The hotel had an impressive foyer. I The reception desk is in the lobby.
I 3. The bill was one activists planned to lobby. I
4. He's a friend on whom one can count. 1. The hotel had an impressive lobby.
The burgh quietly picked the lock.
I 1. The man talking to himself could be crazy. 1
4 The bill was one activists planned to protest. 1. The door was secured with a lock. 2. The door was secured with a bolt.
The death of his family drove him mad.
1 3. The rude service he received got him mad. I
I 3. The cargo ships went through a lock. 4. The cargo ships wer~t through a tunnel. I 1. The man tdking to himself could be mad. 1
I 3- He earns an income that is modest. I
Fans cheered wildly at the soccer match.
He is well-liked for being modest.
4. The rude service he rece~ved pot him upset. 1. The sports channel sh~wed an exciting tennis match. 2. The sports channel showed an exciting tennis game. 3. The smoker lit his cigarette with a match. 4. The smoker lit his cigarette with a lighter. I . He is accomplished but also modest. 2. He is accomplished but also humble.
1 3. She was engrossed in reading the novel. I
She's in the habit of being neat.
i-ier design is one that's novei.
1 -- ! 4. She was engrossed in reading the book. He was rvotded because he behaved odd. ( 1. The man's habits were rather odd. 1
4. He earns an income that is meager. 1. He kept his desk cIean and neat. 7 - I He kept his desk dean and tidy. 3. His ideas are always rather neat. 4. His ideas are always raher nifty. 1. The company rewards ideas that are novel.
1 2. The mn's habits were rather strange. !
! 2. The company rew;uds ideas that are new.
1 3. One and three are numbers that ard odd. I She couldnl decide which dress to pick.
The book was hidden in the pile.
4. One and three are numbers that are prime. 1. There were many alternatives ftom which to pick. 7 b e There were many alternatives from which to choose. 3. He strummed the guitar with his pick. 4. He strummed the guitar with his finper. I The cards where put neatly in a pile.
The cards were put neatly in a stack. The carpet had a thick pile.
4. The carpet had a thick padding. I
i 1 I -. He served drinks from a jug. She bought an antique water pitcher. I !- He served drinks from a pitcher.
1 1 I f . He trained hard to be a baseball pitcher. I
/ 3: The secretary had many letters to mail. The colonel holds a high military post.
I Post The documents were sent by post.
pupil
raise.
4. He trained hard to be a baseball catcher. 1. The secretary had many letters to post.
I
, raise I The weights were ones he could easily His arms ached too much to lift. His boss gave him a raise.
He was cited as a model pupil.
4. Part of the eye is called the ins. 1. His arms ached too much to raise.
4. The colonel holds a high military rank. 1. He is his teacher's favourite pupil. 2. He is his teacher's favourite student. 3. Part of the eye is called the pupil.
4. His boss gave him a lecture.
1 2. He won the lottery and became wealthy. / 3. The chocolate cake was very rich.
1 rear I
I
I 1 rich
Some passengers prefer seats in the rear. His car was hit in the His car was hit in tht back.
spot
I ! Some fishes are difficult to rear.
His inheritance made him very rich.
j 1
1
stick
I
4. Some fishes are diff~ult to catch. 1. He won the lottery and became rich. I I
4. The chocolate cake was very tasty. I
/ store
t m e d 1 The cat in the burning house was saved. 1. The drowning woman was miraculously saved.
1 4. He beat the snake with a twig. To rent a movie he went to the video / 1. Sofas can be bought at a furniture store. I
' 2. The drowning womn was miraculously rescued. 3. He squandered the money he had saved. 4. He squandered the money he had won.
Mount Everest is a difficui t The wall took much stamina to scaIe. scale. The wall took much stamina to climb.
He choked on a fish scale. He choked on a fish bone.
Being tall he was easy to spot. The camouflaged lizard was hard to spot. The camouflaged lizard was hard to find. The Bahamas is a popular vacation spot. The Bahamas is a popular vacation place.
I I store. 1 2. Sofas can be bought at a furnitwe shop. 1
He used glue so the papers would stick. I. He licked the stamp to make it stick.
i 2. She went to the ai+n in a cab. 3 The plane waited its turn to m i .
- I 4. The plane waird its tun to land.
2. He licked the stamp to make it adhere. 1 3. He beat the snake with-a stick.
tail I He lost the man he was to tail.
3. The toys were bulky and hard to store. 4. The toys were bulky and hard to carry. k ! 1 2. 1. The The elusive elusive spy spy proved proved hard hard to to tail. follow.
i 1 I i 3. I
The Siamese cat has a slender tail. 1 i 4. The Siamese cat has a slecder body.
L
laxi He was late so he nailed a taxi. I 1. She went to the airpart in a taxi.
APPENDIX F
EXPERIMENT 3: INSTRUCTIONS TO SUBJECTS AT JUDGEMENT PHASE
Just before the judgement task began, subjects read the following instructions on-screen:
"Now for the next task. you will again be presented with sentences like the ones you just saw.
First, you are to read the sentence aloud and then indicate if you find the last word in the sentence
to be pleasmt or neutral. Two keys on the keyboard have been marked with a blue and a red
sticker. If the word is pleasant to you. then press the red key. If the word is neutral to you, then
press the blue key.
Atier you have classified the word as pleasant or neutral, you are to indicate if the word had been
presented to you in task 1 earlier in this experiment. If you remember having seen the word.
respond old by pressing the red key. If you do not think the word was presented earlier, respond
new by pressing the blue key ."
[The order of these ~nstructions were reversed for subjcc~s in the recognition-first condition.]
The experimenter then supplemented these on-screen ~nstructions by telling subjects the following:
1. that although the task was self-paced, they should respond quickly rather than deliberate for too
long before responding;
2 . that they should base their pleasantness judgement on their overall liking for or feeling toward the
word;
3. that they should respond 'old' on the recognition test only if the exact same word, with the same
meaning and spelling, had been seen earlier.