Professional Documents
Culture Documents
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/4189184?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Taylor & Francis, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to Journal of
Advertising
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
Memory-Based Measures for Assessing Advertising Effects:
A Comparison of Explicit and Implicit Memory Effects
Stewart Shapiro and H. Shanker Krishnan
Prior marketing studies investigating memory for advertisements have relied almost exclusively on examining
effects contingent on explicit memory retrieval. This process involves a deliberate effort on the part of the
consumer to think back to an advertisement in an attempt to recall the ad information. Studies in this area
have shown that a lengthy delay between ad exposure and test, as well as divided attention during the ad
exposure episode, hinder or even eliminate successful explicit memory retrieval. The premise of this paper is
that an alternative retrieval process, implicit memory, may function differently. This form of memory retrieval
is automatic in nature and does not rely on consumers deliberately searching their memory for a previously
viewed advertisement. Comparisons with explicit memory retrieval suggest that implicit memory is preserved
even in conditions of delay and divided attention, whereas explicit memory is affected detrimentally by those
conditions. The two different forms of retrieval processes are validated with the use of a process dissociation
procedure. Implications and directions for future research are discussed.
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
The Journal of Advertising
Implicit memories have been shown to lead to a re- sure episode fails (explicit memory failure)
sponse bias in which there is a greater likelihood of (Richardson-Klavehn and Bjork 1988).
using the previously seen information to complete a Further evidence for implicit memory's importance
task without the awareness of doing so. One type of stems from the fact that it has been shown to corre-
response bias that has been found is an increased pref- late strongly with judgments, even when explicit
erence for previously seen information (Schacter 1987). memory does not (Kardes 1986; Shedler and Manis
One explanation for this effect is perceptual fluency, in 1986). For example, memory mediates the effects of
which a previously seen stimulus appears familiar, and vividness on judgment when implicit, but not explicit,
absent a successful search of memory to attribute this memory is used. The rationale for this effect is that
familiarity to the prior exposure episode, the familiar- implicit memory is closer to the behavioral predispo-
ity is attributed to a preference for the stimulus (Seamon sitions of a consumer and is a form of memory used in
et al. 1995). In our chewing gum example, implicit everyday situations (Krishnan and Chakravarti 1999).
memory would be revealed if exposure to the advertise- Therefore, focusing on implicit memories would re-
ment increased the likelihood that the consumer pur- move one weakness of focusing solely on explicit
chase the advertised brand without making a deliberate memory, that explicit memory retrieval does not cor-
attempt to recall the prior ad exposure episode. relate strongly with persuasion scores.
The exact nature of implicit memory has been a Gaining a greater understanding of the full impact of
subject of hot debate (see reviews by Richardson- an ad is also of great importance when trying to mea-
Klavehn and Bjork 1988; Schacter 1987). For example, sure ad effectiveness. Many companies, such as Procter
some take a two-systems approach that suggests im- & Gamble and Philip Morris, spend more than $2 bil-
plicit and explicit memory reside in different memory lion annually to advertise their products. Clearly, it is
systems. Others take an activation approach that in- important for these companies to assess the effective-
dicates a higher level of activation is needed for suc- ness of their advertising campaigns. Some firms rely
cessful explicit memory retrieval compared with on memory measures, such as recall and recognition, to
implicit memory retrieval. Although the exact nature gauge the effectiveness of an ad campaign. For ex-
of implicit memory may not be known, evidence sug- ample, in a day-after recall test, respondents are asked
gests that tests of explicit and implicit memory do not various questions about television commercials that
respond similarly to the same experimental manipu- aired the previous day. In a starch recognition test,
lations. This is particularly true in conditions in which respondents are asked to indicate whether they had
explicit memory retrieval has been found to be poor, seen specific print ads in a magazine. Responses to
such as after a long delay between ad exposure and these tests are aggregated and compared against norms
subsequent judgement or in the case of divided atten- for the category to indicate effectiveness. Decisions re-
tion, where the level of ad processing may not be garding ad campaigns and the management of the ad-
sufficient to yield successful explicit memory retrieval. vertised brand hinge on these judgments of effectiveness.
Because implicit memory does not rely on a directed It is the premise of this article, however, that these
search of memory, these same factors have not been measures are telling only part of the story and that
found to inhibit it to as great an extent (Richardson- implicit memory retrieval of ad information can affect
Klavehn and Bjork 1988; Schacter 1987). choice even when explicit memory retrieval fails.
In many purchase decisions, consumers do not make
a deliberate attempt to search their memory for pre- Hypotheses
viously encoded information. In these situations, im-
plicit (versus explicit) memory retrieval may be Dissociations Among Memory Measures
dominant; therefore, there is a need for marketers to
understand the effects of implicit memory. Consis- To examine the veridicality of our claim that research
tent with this, low retrieval motivation seems to un-based on explicit memory performance captures only a
derlie (Bogart and Tolley 1988; Park and Hastak 1994)portion of ad memory effects, we examine two factors:
findings that retail store consumers spend just five delay between exposure and test and divided attention
seconds on some brand choices. Even if consumers during exposure. In each case, we predict a dissociated
actively search their memory for prior ad informa- pattern of effects, with explicit memory performance
tion, the information may not be accessible.adversely Prior affected by the experimental manipulations
studies in psychology have shown that implicit and implicit memory performance unaffected.
memory retrieval may be successful even when delib- Although most studies investigating explicit
erate attempts to retrieve information from thememory expo- for ad information use a relatively short time
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
Fall 2001
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
The Journal of Advertising
rely on different retrieval processes. Specifically, ex- and automatic (implicit retrieval) influences may con-
plicit memory entails a deliberate, conscious search tribute to overall performance on a given task (Jacoby
of memory for the ad information, and implicit memory 1991), the PDP is used to parse out the effects due to
entails a response bias caused by the nondeliberate, conscious memory retrieval, yielding an unbiased es-
unconscious retrieval of ad information. Therefore, it timate of the amount of automatic influence caused
is imperative to provide convincing evidence that dif- by implicit memory retrieval. This procedure has been
ferent retrieval processes are being used in the two used widely in psychology to provide separate esti-
tests of memory. To provide convergent validity for mates for the level of contribution to an overall per-
this claim, two different methods are used. formance that results from conscious and automatic
Prior research has used a statistical test to demon-
processes (e.g., Anooshian and Seibert 1996; Debner
strate that study participants did not use explicit and Jacoby 1994; Hay and Jacoby 1996; LeCompte
retrieval processes when completing an implicit 1995; Nelson, Bennett, and Xu 1997; Reingold and
memory test (Perruchet and Baveux 1989). For Goshen-Gottstein
ex- 1996; Robinson-Riegler and Winton
ample, study participants would complete an implicit1996; Toth 1996) but has yet to be adopted in the
memory test without knowing that the test wasmarketing
re- literature (cf, Shapiro 1999). The proce
dure uses two tasks, an exclusion task and an inclu-
lated to a prior exposure episode. These same partici-
pants then would complete an explicit memory test sion task. The former provides an uncontaminated,
for the same information that asked them to think pure measure of automatic influence, whereas the
back and attempt to recognize whether several latter test provides a measure that contains both con-
items had appeared during the exposure episode. scious
If and automatic sources of influence. To under-
stand how these measures are developed, it is first
participants use explicit retrieval processes when com-
pleting the implicit memory test, their performanceimportant to understand the exclusion task.
on the implicit memory test should be correlated withIn a typical study using this procedure, words (e.g.,
FLOWER) are flashed on a computer screen for a
their performance on the explicit memory test subse-
quently administered; the use of similar retrieval very
pro- brief duration (e.g., 50 milliseconds). Subjects
cesses should yield correlated results. In contrast,
are then given a stem completion task that provides
them with the first three letters of half of the flashed
evidence that different retrieval processes were used
across the two types of memory tests would be pro- words (e.g., FLO_) and requires them to complete
each word stem. Conscious (C) and automatic (A)
vided if performance on both tests was uncorrelated.
This result is known as stochastic independenceinfluences
and are put in opposition with each other be-
cause subjects are asked not to complete the word
provides evidence that an explicit retrieval process
was not used to complete the implicit memory stems
test with the words that just had been flashed.
Because participants are told to exclude responding
(Perruchet and Baveux 1989). We would expect par-
ticipants' performance, whose original test waswith
ex- the originally flashed words, these instructions
are termed the "exclusion instructions.,, Successful
plicit in nature, to correlate with an explicit memory
test administered later because the same retrieval conscious retrieval of the flashed words would lead to
process would be used in both cases. We predict athat decreased likelihood of completing the word stem
H4: Performance on an implicit memory test with the test words. However, because automatic,
will be uncorrelated with performance on unconscious
a influences are by definition outside our
subsequently administered explicit memory volitional control, if conscious memory retrieval fails,
test, whereas performance on an explicit the automatic influences gained from exposure should
memory test will be correlated with perfor- increase the likelihood of completing the stems with
mance on another, subsequently adminis-the flashed words even though participants are try-
tered explicit memory test. ing to avoid doing so (Debner and Jacoby 1994). In
the exclusion task, an increased likelihood of com-
Although stochastic independence provides evidence
pleting
that explicit retrieval processes are not being used to the stems with the flashed words would occur
only if conscious memory retrieval failed (1?C) and
complete an implicit memory test, it does not demon-
strate that the retrieval process used to complete automatic
the memory retrieval led to a correct response
implicit memory test is unconscious. To add further(A). Formally,
credibility to our call for measures that capture im- Exclusion^ 1-OA. (1)
plicit memory retrieval, an additional procedure is
It is important to note
used, called the process dissociation procedure (PDP)
provides unambiguous e
(Jacoby 1991). Because conscious (explicit retrieval)
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
Fall 2001
Krishnan
enees, it actually underestimates the amount of such and Chakravarti 1993; Krishnan and
influence. Specifically, unless conscious memory re-1996; Shapiro 1999). We hypothesize that
Shapiro
trieval fails across all trials (for all flashed words),
H5:on
In the PDP, the extent of conscious influence
those trials it is successful, it will override automaticwill decrease as (a) the time delay between
influences and drive down the overall estimate of exposure and test increases and (b) attention
automatic influence (successful conscious memory to the ad decreases, whereas the extent of
retrieval would ensure that a word stem was not automatic influence will remain unaffected
completed to form a flashed word). The PDP corrects across these same conditions.
for this contamination by using inclusion instructions
in conjunction with the exclusion instructions. Method
In the inclusion instructions, word stems for the
remaining half of the test words are provided,An andoverview of the procedure we use is provided in
participants are asked to complete each of the Figure word 1. A lab experiment was designed to enable
stems with a word that appeared in the exposure performance pattern comparisons of memory tests as
episode. An increased likelihood of completing the
a function of two factors. The study used a 2 ? 2 plus 1
stem to form one of the flashed words would occur if between-subjects design with timing of the test (no
conscious memory retrieval was successful (C) or if delay/one-week delay) and attention (divided/full) as
conscious memory retrieval failed but automatic in- the manipulated factors, plus a control group that
fluences created a correct response (l-C)A. Formally,was not exposed to the target ads. In addition, three
Inclusion=C+(l-C)A. (2) dependent variables were used and collected between
subjects: one tapping explicit memory retrieval, an-
Using Equations 1 and 2, we can algebraically s
for C and A to obtain an uncontaminated estimate of
other implicit memory retrieval, and a third (the PDP)
to provide separate estimates of conscious and auto-
automatic influence, as well as an estimate of the
amount of conscious influence associated with com-
matic retrieval. A total of 369 participants (228 for
the explicit and implicit memory measures, and 141
pleting the task (in this example, the stem comple-
for the PDP) were recruited from a participation pool
tion task). Specifically,
at a large Midwestern university and received credit
Conclusion performance-Exclusion performance (3)for participation. All participants were assigned ran-
and domly to the treatment conditions, and all dependent vari-
ables were represented in each experimental session.
A=Exclusion performance/i 1-C), (4)
where performance is measured by the proportion Procedures of
correct responses.
Although these equations also could include All participants were exposed to several advert
a guess-
ing factor (baseline), in that a correct response ments could
for fictitious brands and asked to evaluate each
be given if both conscious and automatic memory Subsequently, they completed manipu-
advertisement.
failed and the participant simply guessedlation check measures for the attention manipulation
correctly,
baseline figures generally are disregarded because
(divided versus full), and then they undertook an un-
the level of analysis using the PDP typically related filler task that lasted 15 minutes. Partici-
involves
examining relative levels of influence across pantscondi-
in the no-delay group continued by answering
tions and showing that factors that affect questions the extent related to the dependent measures. Par-
of conscious influence have no effect on the extent of ticipants in the one-week delay group were asked to
automatic influences (Jacoby 1991). return at the same time the following week, at which
If the conscious and automatic components of thetime they completed the dependent measures (see
PDP yield a pattern of results identical to that of our Figure 1). Participants were assigned randomly to
measures of explicit and implicit memory, respec-one of the three dependent variable conditions: ex-
tively, we have more confidence that a different pat- plicit memory, implicit memory, or process dissocia-
tern of results across the explicit and implicit memorytion. Therefore, completion of the dependent variables
measures is due to different retrieval processes. This was also between-subjects to ensure that performance
would not only support the need to measure ad effects on one test did not contaminate performance on a
due to implicit retrieval, but also extend consumersubsequent test.
research on methods used to separate automatic and The attention manipulation was designed to vary
conscious memory processes (Duke and Carlson 1994;the amount of attention devoted to the ads during
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
6_ The Journal of Advertising
Measures
exposure. The manipulation involved the use o
audio program played for all participants (a s
story) while the advertisements were task
Explicit Memory. A recognition displaye
was used, ra
slides. The cover story than a free recall that
explained task, to sometimes
assess explicit mem
retrieval because pretests
sumers are exposed to advertising showed that recall
messages when was e
tremely
are listening to broadcast poor after
media a one-week
(e.g., delay. Furthermor
thumbing th
recognition task better
a magazine while listening to the radio or wat matches the implicit mem
task, because the test
television, watching the road while driving with stimuli remain relatively co
sistent between the recognition
radio on). The full attention participants were instr task and imp
to focus on the ads and memory
ignore task. The
thereal difference
audio is manifested in
program
actual instructions
divided attention participants were instructed given to participants. In to
the rf
on the ads but also to listen to the audio prog
ognition test, participants were given a list of tw
product categories
because they would be tested on the program lateassociated with the twelve ta
ads. Next to each product
the study. By varying the extent to which partici category were two b
focused on the radio ad, names,
thethe brand name listed in
amount ofthe ad and a
attention
distractor brand
sources available to process the name not shown
ads alsoin the exposure
varied.
episode. Subjects
pretest of this manipulation, thewere told that, for
former each product i
resulted
category
and the latter in divided listed, one of the
attention totwothealternatives
ads was a
duri
exposure. The desiredbrand name from the ads
attention they had seen,
level also and was
they
were asked to identify
firmed through manipulation checks. that brand name. Thus, this
Subjects in the control group
represents were
a two-alternative, not
forced choiceexpose
task that
requires subjects to
the target ads but completed think back
the to the exposure
implicit epi-
mem
measure. As will be discussed in the measurement sode and indicate which brand name appeared in the
section, this measure required participants toad
make
for each product category. Performance was mea-
brand choice decisions among various brands,sured
onlyas a proportion of correct responses with chance
levels of performance equal to 50%.
some of which appeared in the ads shown to subjects
in the experimental conditions. Therefore, this Implicit
con- Memory. In the implicit memory condition,
the twelve target test items were identical to those in
trol group provides a baseline level for the percentage
the recognition test; however, twelve additional
of times the target brands would be chosen indepen-
dent of any effects due to ad exposure. distractor product categories and brand names not
previously viewed were listed in an attempt to con-
ceal that this task referred back to the prior exposure
Advertising Stimuli
episode. Doing so decreases the possibility that par-
ticipants will become aware of the relationship be-
Because the primary dependent measure was memory
for brand names, advertisements were created to fea- this task and the exposure episode, which would
tween
lead
ture brands. All experimental participants were to the use of explicit retrieval processes when
ex-
completing the task (a possibility that is directly ad-
posed to and then tested on their memory for brand
names from 12 target advertisements. To simulate dressed
clut- by testing for stochastic independence be-
ter, these 12 ads were mixed with 12 distractor tween
ads on this task and a subsequent recognition task).
which the participants were not tested. For all 24 Participants
ads, in this condition were told that, for
each product listed, they should assume that they
print ad stimuli were created by modifying existing
were purchasing the product and that they should
color print ads from magazines. The advertisements
were selected to be representative of the types ofexamine
prod- the two brand names listed. Then, solely on
the basis of the brand names, they were to circle the
ucts students would use (e.g., watches, software, jeans).
brand they would be likely to choose. No reference
To control for experience effects, all references to actual
brand names were deleted. New brand names were was made to the advertisements, which made this an
implicit memory task (i.e., an active search of memory
chosen for these products and inserted in these adver-
tisements in clearly visible ad positions. The brand
for the previously seen brand names was not required).
names were pretested to ensure that they would Performance
be was measured by the proportion of cho-
sen brand names that had appeared in the prior ad
considered appropriate for the product category. Color
slides were created for each ad, and each ad was dis-
exposure episode. Implicit memory would be evident
if there was a greater likelihood of choosing the tar-
played for ten seconds with a slide projector. Partici-
get brands compared with a baseline group that was
pants were tested in small groups of eight to ten people.
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
Fall 2001
brand
not exposed to the target ads but completed the samenames to assess the correlations with the de-
dependent measure. pendent variables to test H4. In this test, partici-
Process Dissociation Procedure. Finally, in the PDP
pants were given 24 product categories, each of which
task, participants were asked to make two kindshad two
ofbrand names fisted next to it. Twelve of
choices. For six of the target brands, participants
these product categories were from the target adver-
tisements,
were given inclusion instructions that asked them to and the remaining 12 were distractor items
think back to the ads and choose those brands thatnot shown during the exposure episode. The target
were featured, because they had received superior brand name and distractor brand name fisted next to
the advertised products were the same as those used
ratings in Consumer Reports. As with the other mea-
in to
sures, each of the six target brands was listed next the dependent measures. Participants were told
that some of the brand names listed may have ap-
the appropriate product category and was paired with
a distractor brand that had not appeared during peared
ad in the ads they had seen, and they were to
indicate which they recognized. The test items thus
exposure, making this a two-alternative, forced choice
task. For the other six brands, participants were given
were similar to the original recognition measure, ex-
exclusion instructions that asked them not to choose
cept that additional distractor items had been added.
any brands they had seen in the ads, because these The test items were most similar to the implicit
brands received poor ratings in Consumer Reports. memory test, because the same number of distractor
Pretests indicated that these were credible manipu- items appeared. The purpose of this measure is to
lations of the inclusion/exclusion tasks. Each list of determine whether performance on the implicit
six brand names was counterbalanced across the in- memory test is stochastically independent from per-
struction type (inclusion/exclusion). From each formance on the explicit memory test, which would
participant's response, the automatic and conscious com- suggest different retrieval processes.
ponents were estimated. For example, if a subject cor-
rectly identified three of the six brands (50.0%) with the Results
inclusion instructions and two of the six brands (33.3%)
with the exclusion instructions, the extent of conscious ad Manipulation Check
influence, as given in Equation 3, would be ,50-.333=.167,
whereas the extent of automatic ad influence would be Results from the self-reported and objective mea-
[.333/(l-.167)]=.40, as given by Equation 4. sures of attention suggest that the attention manipu-
Attention Manipulation Check and Retrieval Pro- lation was successful. Two ANOVAs using all subjects
in the experimental conditions (see Figure 1) were
cess Check. Immediately after concurrent exposure to
the ads and radio program, all subjects completed run, with the self-reported measure as the dependent
manipulation check measures for the attention ma- variable for one of the ANOVAs and the objective
nipulation (full versus divided) (see Figure 1). These
measure as the dependent variable for the other. The
measures consisted of one self-reported measure of attention factor and time of test factor were included
as independent variables for both analyses. The time
attention and four objective questions to assess knowl-
edge about the radio program. The self-reported mea- of test factor was included simply to provide evidence
sure required subjects to indicate how much attention that there were no systematic differences in the level
they paid to the advertisements versus the radio pro- of attention paid to the ads across the no-delay and
gram on a 21-point scale from -10 (full attention one-week
to delay conditions.
the ads) to +10 (full attention to the radio program). Results of the self-reported measure show a main
The objective measure of attention involved four mul- effect of attention with greater attention paid to the
tiple-choice questions regarding the information pre-advertisements relative to the radio program in the
sented in the radio program. There were objectively full (x=-4.11) versus divided (x=+.88; F(l,363)=80.39,
correct answers to these questions, and the questionsp<.001) attention condition (negative numbers indi-
spanned information presented throughout the pro- cate greater attention allocated to the ads, and posi-
gram. One objective measure, ranging from zero cor-tive numbers indicate greater attention allocated to
rect to four correct answers, was developed for these the radio program). Neither the main effect of time of
multiple choice questions. Higher scores indicated testa (F(1,363)<1) nor the interaction term
greater amount of attention paid to the radio program(F(l,363)=1.59, p>.20) was significant. An ANOVA
and, hence, less attention to the advertisements. with the objective measure of attention indicate
greater amount of attention paid to the radio p
At the end of the study, all participants were ad-
ministered a forced choice recognition test for the
gram, and thus less attention to the advertisement
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
8 The Journal of Advertising
4*5
!9 ^
?. a?
ce
o ? ?o
Q-.C
? ?
F ?
?
"?
3
?s
f
?
?
S* c
u. F
e
k.
f
?.
?
LU
f ??
c F
]?
a:
o co
? = (0
?. _c
? ?
F ?
F to
? g>8
(?
CL W $
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
Fall 2001 9
Table 1
Mean Proportion of Target Brands Chosen
Notes: Conceptually raw scores should not be compared across the recognition and implicit memory measures. Each measure tests something
different. The recognition measure tests subjects' ability to recall ad information (the brand name) from the prior ad exposure episode,
whereas implicit memory tests for automatic ad influence that causes a response bias (increased likelihood of choosing the brand).
Therefore, patterns of effects for each measure across the manipulated factors (time of test and attention to ads) should be examined.
in the divided (x=2.52) versus the full (x=1.76; and divided attention condition (t(73)=2.0, p=.05),
whereas it was significantly worse than chance levels
F(l,365)=42.07,p<.001) attention condition. Again, nei-
ther the main effect of time of test (F( 1,365)=1.8, p>. in
18)the delay condition (t(41)= 3.24, p<.01). Although
nor the interaction term (F(1,363)<1) was significant.
recognition performance decreased significantly when
participants were required to allocate part of their
Dissociations Among Memory Measures attentional resources to a secondary task while at-
tending to the ads, ad processing in this condition
Means for the measures appear in Table 1. HI pre-was sufficient to facilitate explicit memory retrieval
dicts that explicit memory retrieval will decrease as (a) to a certain extent. Conversely, a one-week time de-
the time delay between exposure and test increaseslay between exposure and the test led to a complete
and (b) attention to the ad decreases during exposure. failure of explicit memory retrieval. Performance was
An ANOVA on recognition performance shows lower actually worse than chance levels in this condition,
memory performance in the delayed (x=.39) versus thewhich suggests that participants were merely guess-
immediate test condition (x=.70;F(l,lll)=47.44,p<.001) ing and that the distractor items were viewed as
and in conditions of divided (x=.56) versus full atten- equally or more likely to have appeared in the ads.
tion (x=.64; F(l,lll)=5.01,p<.05). The interaction be- The level of implicit memory retrieval was exam-
tween time of test and amount of attention was not ined to test H2, which predicts that implicit memory
significant (F(1,111)<1). These results support HI retrieval
and will be evidenced as a function of ad expo-
are consistent with the literature. sure. Consistent with H2, ad exposure led to a greater
T-tests were conducted on the recognition measure likelihood of choosing the advertised brands across
to determine whether recognition responses were sig- all conditions. A baseline level (x=.25) of brand choice
nificantly different than chance levels (x=.50). Re- was given by the 21 control group subjects who were
sults indicate that recognition was greater than chancenot exposed to the brands prior to completing the
levels in the immediate test condition (t(72)=6.97, choice measure. This baseline level was subtracted
p<.001), full attention condition (t(40)=3.20, p<.01),from the implicit memory performance, and t-tests
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
10 The Journal of Advertising
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
Fall 2001 11
Singh, Rothschild, and Churchill 1988). Because these The second research implication is that, in many
tests were expected to dissociate as a function of key cases, implicit, not explicit, memory measures are
variables, they should provide different perspectives appropriate. Specifically, in situations in which brand
on ad effectiveness (Krishnan and Chakravarti 1999). choice decisions do not involve an effortful search of
Here, the explicit/implicit memory distinctions were memory for previously seen ad information, memory
validated with the use of a PDP. Overall, the explicitmeasures based on noneffortful retrieval processes
test results parallel the conscious component of the (i.e., implicit memory) are relevant for assessing an
PDP, and the implicit test results parallel the auto- ad's effectiveness. For example, for many low involve-
matic component, which lends convergent validity to ment purchases, consumers do not try to retrieve ad
the memory retrieval distinction. Therefore, the re- information, and the only manner in which prior ad
sults of this study provide convincing evidence that exposure can have an effect on choice is through un-
researchers should take retrieval factors into account conscious means. As demonstrated in this research,
when using memory performance as a means of assess- prior exposure to advertising can affect brand choice
ing advertising effectiveness. Three specific implica- without the consumer being aware of the basis for
tions for researchers are discussed subsequently. that choice. Impulse buying also may be prone to
unconscious influences from advertising, because the
Research Implications choice is unplanned and thus made very quickly with-
out much thought (Rook 1987). In these situations,
The first research implication is that current mea-researchers should use implicit memory measures
sures of ad effectiveness based on explicit memory because consumers do not rely on effortful retrieval
processes. In addition, in these situations, an ad's
(e.g., recognition and recall) are valid only in brand
choice situations in which consumers engage in an effectiveness is not affected adversely by divided at-
effortful search of memory for prior ad information,tention or a one-week delay between exposure and
which may be the case in many purchase situations. brand choice decisions. Critical to the success of im-
For example, many product categories can be classi- plicit memory performance is simply exposure; thus,
fied as high in self-involvement because they are ex-the practical implication is that emphasis should be
pensive, risky, or simply important to the consumer. placed on an ad's reach rather than on its ability to
Furthermore, in situations in which a consumer is attract or sustain attention. Furthermore, firms may
new to the category and is actively evaluating brands, need to include implicit memory in their test battery
it is likely that an explicit retrieval process will be if they find that the product category is driven by low-
employed. In these situations, researchers should use involvement choice or high levels of impulse buying.
explicit memory measures because consumers are The third research implication is that consumers may
using effortful retrieval processes. If successful ex- use a mix of explicit and implicit memory retrieval to
plicit memory is a goal of an ad campaign, research- guide choice, and, therefore, the relative contribution
ers should be aware of factors that may inhibit explicit of these processes to choice should be understood. For
memory performance. For example, in this study, both example, even in cases in which an effortful search of
divided attention and a time delay between ad expo- memory for previously seen ad information is common,
sure and choice were found to decrease memory per- when such effort fails (as was found in the delay condi-
formance and, hence, the ad's effectiveness in tion), implicit memories may help drive choice. This
achieving successful explicit memory. What this im- result underscores the importance of using implicit
plies for practice is that attention to the ad shouldmemorybe performance in addition to explicit memory
increased through various techniques. This may mean measures as a means of assessing an ad's effectiveness,
that reach is sacrificed, because some media mayeven notin situations in which consumers use effortful
be conducive to achieving goals of higher attention. retrieval processes. Methodologies such as the PDP are
Furthermore, because initial exposures that leadparticularly
to relevant in parsing out the relative contri-
attention and elaboration may decay rapidly on explicit butions of automatic and conscious processes and pro-
memory measures, media forms that provide multiple viding an estimate of their relative contributions.
"touches" should be considered, not only to enable the
goals of elaboration and reinforcement, but also toDirections
in- for Future Research
crease the likelihood that a short time delay exists
between ad exposure and choice. Ad agencies shouldThis research compares explicit to implicit memory
retrieval and demonstrates how the PDP may be used
therefore adapt advertising campaigns to the retrieval
condition, such as explicit memory retrieval. to validate different types of retrieval processes.
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
12 The Journal of Advertising
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms
Fall 2001 13
This content downloaded from 144.173.6.94 on Thu, 25 Aug 2016 11:49:14 UTC
All use subject to http://about.jstor.org/terms