You are on page 1of 15

Situational Effects of Advertising Repetition: The Moderating Influence of Motivation,

Ability, and Opportunity to Respond


Author(s): Rajeev Batra and Michael L. Ray
Source: Journal of Consumer Research , Mar., 1986, Vol. 12, No. 4 (Mar., 1986), pp. 432-
445
Published by: Oxford University Press

Stable URL: https://www.jstor.org/stable/254303

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/254303?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
https://about.jstor.org/terms

Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Consumer Research

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
Situational Effects of Advertising
Repetition: The Moderating Influence
of Motivation, Ability, and Opportunity
to Respond

RAJEEV BATRA
MICHAEL L. RAY*

It is argued theoretically that the attitudinal gain from advertising repetition should
continue to increase rather than level off when consumers fail to generate cognitive
responses to message arguments in earlier exposures. An experiment shows that
repetition continues to increase brand attitudes and purchase intentions in conditions
where support and counter argument production is expected to be low, but that
these attitudinal gains level off under conditions in which a high level of such pro-
duction is expected.

The effects of message repetition on attitudes and ment with the message (the attitudinal outcome studied)
purchase intentions have been studied often (see increases, then decreases with higher exposure. Simi-
Sawyer 1974 and Belch 1982 for reviews). Most studies larly, Calder and Sternthal ( 1980) explain repetition re-
show that messages gain in impact for a few exposures sults in terms of the relative production of message-
but that further exposures begin to have a negative ef- related versus "own" thoughts on repeated exposure,
fect. Such inverted-U curves for repetition impact with "own" thoughts-assumed to be less positive than
emerge, for example, in the literatures on the attitudinal message-related thoughts-forming an increasing pro-
impact of mere exposure (Harrison 1977; Sawyer 1981; portion of total thoughts with increasing exposure.
Zajonc 1968) and on advertising wearout (Calder and These studies and theoretical perspectives would
Sternthal 1980). suggest that, in general, intermediate levels of message
Common to both the wearout and mere exposure exposure (two or three) should provide higher message
literature is the process view that while increasing ex- effects than either very low or very high levels. This
posure initially enhances learning and favorable atti- expectation finds support in studies of advertising rep-
tudinal affect, subsequent exposures-past the point of etition on eye movement responses (Krugman 1968),
"over-learning"-create tedium and negative affect attention (Grass and Wallace 1969), and brand attitudes
(Berlyne 1970; Stang 1975). Further, Cacioppo and (Gorn and Goldberg 1980; Naples 1979; Ray, Sawyer,
Petty (1979) use the cognitive response paradigm and Strong 1971; Winter 1973).
(Greenwald 1968; Wright 1973) to show that production In this paper, we extend the results of such prior re-
of support arguments increases, then decreases with ex- search by examining variations in attitudinal (and pur-
posure, while production of counter arguments de- chase intention) gains from advertising repetition across
creases, then increases with exposure; as a result, agree- situations that vary in the antecedent conditions re-
quired for advertisements to generate cognitive respon-
*Rajeev Batra is Assistant Professor, Graduate School of Business, ses. Such antecedent conditions include the message
512 Uris Hall, Columbia University, New York, NY 10027. Michael recipient's motivation, ability, and opportunity to gen-
L. Ray is Professor of Marketing and Communication, Graduate erate support arguments and counter arguments to the
School of Business, Stanford University, Stanford, CA 94305. This
research was funded by a grant from the Marketing Science Institute,
message. Evidence exists that the point at which ad-
and additional financial support was provided by the Marketing vertising repetition begins to create an attitudinal
Management Program of the Stanford Business School and by the downturn is moderated by various product, brand, user,
Faculty Research Fund of the Columbia Business School. The authors
execution, and media factors (Naples 1979; Ray et al.
would like to thank the many companies and advertising agencies
who provided commercials for use but who prefer anonymity. 1971; Silk and Vavra 1974). Though theoretical frame-
works exist to explain why the effects of repetition are,

432
?) JOURNAL OF CONSUMER RESEARCH * Vol. 12 0 March 1986

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
SITUATIONAL EFFECTS OF REPETITION 433

in general, curvilinear (e.g., Sawyer 1981), such situa- turn will occur earlier. In the latter situation, further
tional differences in where the downturn begins deserve repetition will yield no further attitudinal gain; attitudes
further research. may, in fact, decline with further repetition.'
In this article, a theoretical framework is developed
both to explain previous research results indicating such
Antecedents of Cognitive
situational differences and to make further predictions
about the differential gains from repetition for different Response Production
kinds of advertisements. In particular, hypotheses are
developed about the differential gain from repetition It is well known that the antecedent conditions for
for ads classified as being high or low on the levels of such support argument and counter argument produc-
support and counter argumentation expected on ex- tion are (1) the motivation, (2) the ability, and (3) the
posure. These hypotheses about differential repetition opportunity to respond to message arguments. The
effects are then tested experimentally. motivational antecedent-the feeling on the part of
message recipients that the content of the message is
important to them-is usually labelled '"involvement,"
THEORETICAL FRAMEWORK
and is the antecedent most commonly studied (Chaiken
Our "situational effects" framework builds on Ca- 1980; Roberts and Maccoby 1973; Wright 1973; see
cioppo and Petty's (1979) cognitive response explana- also Wright 1980). As shown by Petty and Cacioppo
tion of repetition effects. Cacioppo and Petty showed (1979), high levels of such motivation increase the pro-
that the typical pattern of attitudinal effects of message duction of support arguments and counter arguments.
repetition (an inverted-U curve) could be explained by The ability antecedent is discussed by Roberts and
the differential production of support arguments and Maccoby (1973), who show that increased respondent
counter arguments at different levels of message repe- knowledge about the issue leads to increased cognitive
tition. Several aspects of this process mechanism deserve response production to a message (cf., the inoculation
attention. As the authors point out, the mechanism in- research of McGuire 1964). Note that such ability, as
dicates that (1979, p. 105): usually defined in such research, depends on the issue
the repetition of the message arguments provided more
in question, and is not treated here as an individual
opportunities to elaborate cognitively upon them (sic) trait.2 In the advertising context, such response-enabling
and to realize their cogency and favorable implications. knowledge could come from generalized knowledge
Hence, counterargumentation declined at the moderate about the product category, the brand in the ad, or even
exposure frequency. At high exposure levels, however, from high prior familiarity with that specific advertising
tedium and/or reactance may have motivated the indi- execution. While the relationship is usually thought of
vidual to again attack the now offensive communication. as linear, it is possible that very high levels of prior
Thus, counterargumentation was renewed, and agree- knowledge might lead to reduced, rather than greater,
ment decreased at high exposure levels.
levels of response to message arguments (cf., Johnson
In other words, high exposure messages, having already and Russo 1984).
been thought about, become offensive. Such message- As pointed out by Petty, Cacioppo, and Goldman
related thinking would involve the generation of the (1981, p. 853), the motivation and ability antecedents,
support and counter arguments studied by Petty and though theoretically distinct, are usually highly corre-
Cacioppo. lated in practice, since those motivated to find out more
Two implications follow from this explanation. First, about an issue are likely to develop greater expertise
if the point at which the message is thought about is than are those who are less motivated (cf., Markus 1977,
delayed, the point at which it becomes offensive should who found that people have a greater ability to think
also be delayed, and the curve of dependent attitudes about those things relevant to their "self"). In an ad-
should continue to rise rather than fall to the familiar vertising context, Lutz, MacKenzie, and Belch (1983,
inverted-U shape. Second, the point at which the mes- p. 534) report that respondents high in motivation were
sage is thought about is a function of the antecedent also high in knowledge. Thus, while these two anteced-
conditions for support argument and counter argument ents should be manipulated independently where pos-
production in that situation. If the antecedent condi- sible (or motivation should be varied with ability kept
tions are at low levels, such message-related thinking
would not occur extensively in the first few exposures,
the message would not become offensive until later, and 'A similar model has recently been suggested by Cacioppo and
the point of the attitudinal downturn would be delayed. Petty (1985).
Repetition would thus continue to increase attitudes. 2It could be argued that individual trait-like differences in inherent
In contrast, when the antecedent conditions create high processing ability also ought to be considered. While the argument
levels of support argument and counter argument pro- has merit, such individual differences are usually ignored in the cog-
nitive response paradigm (for exceptions see Wright 1975 and Ca-
duction at low levels of exposure, the attitudinal down- cioppo and Petty 1982). For an exhaustive review, see Wright 1980.

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
434 THE JOURNAL OF CONSUMER RESEARCH

constant, as in Petty and Cacioppo 1983), such orthog- Situational Implications


onal manipulations are often difficult in some settings.
Combined with the process view on repetition pre-
The effects of the third antecedent-response oppor-
sented earlier, the literature on the three antecedents
tunity-are shown most clearly in the literature on dis-
of cognitive response production suggests that repetition
traction effects (cf., Festinger and Maccoby 1964; Os-
of a message will lead to increases in brand attitudes
terhouse and Brock 1970), where high levels of distrac-
only if these antecedents are at low levels. Such a sit-
tion (implying low opportunity) reduce the support and
counter argumentation evoked by a message.3 Addi-
uational framework is supported by and helps organize
results reported in much previous research on repeti-
tionally, Krugman (1967) and Wright (1981) have ar-
tion. Thus, Ray et al. (1971) reported that repetition
gued that television, unlike print, is a low "opportunity"
continued to increase intention to buy in ads for low-
medium since the message pace is higher and the re-
price "convenience" goods, but not for high-price
spondent control over exposure is lower. Wright (1981,
p. 276) has shown that, as a consequence, print messages
"shopping" goods, and for soft-sell "non-grabber" ads,
but not for hard-sell "grabber" ads-a result also re-
usually evoke more cognitive response production than
TV messages. From a different perspective, Calder, In- ported by Silk and Vavra (1974). Rothschild and Ray
(1974) found that while political advertising for a state
sko, and Yandell (1974), Chaiken (1980), and Petty and
Cacioppo (1984, p. 76) have shown that when messages
assembly race did lead to an increase in voting inten-
themselves contain more arguments, the number of
tions as a function of repetition, advertising for a pres-
idential race did not. These results support the situa-
support and counter arguments is higher than when the
tional framework suggested because more motivated
messages contain fewer arguments. The presence of
more arguments within the message thus provides more
message processing can be expected for shopping goods,
"opportunity" to the respondent for support and coun- grabber ads, and presidential elections. Similarly, the
ter argumentation: all other things being equal, a re- reason that Belch (1982) found no significant effects of
cipient should be able to generate a greater number of repetition using TV commercials may be that respon-
such responses to a message that contains more argu- dents were told before exposure that they would be
ments than to one that contains fewer. asked questions about the commercials, thus increasing
These antecedents are discussed further in Wright the motivation to generate message-related thoughts.
(1974, 1975, 1980, 1981) and Roberts and Maccoby Next, since high ability (from greater knowledge and
(1973). While these antecedents do interact, the nature prior familiarity) should enable more message-related
of the interactions is not fully understood. Wright initial processing,4 this should reduce the positive at-
shows, for instance, that while motivational differences titudinal effects of further repetition. Supporting this
impact on support argument production only if the op- expectation, Calder and Sternthal (1980) report that
portunity to do so is already high (as in the print me- higher exposure led to more total (both positive and
dium; 1974, 1981), counter argument production does negative) thoughts for an unfamiliar product, but to an
not show this interaction. Petty and Cacioppo (1984) increase (for high exposure levels) only in negative
show that the existence of more arguments within a thoughts for a well-known product.
message serves as an attitude-increasing "peripheral Finally, since Krugman (1967) and Wright (1974,
cue" under low motivation conditions, regardless of 1981) have argued that print, unlike TV, is a medium
their quality, but increases attitudes under high moti- that allows greater opportunity for attribute-based re-
vation conditions only if the arguments' "cogency" is sponse, the framework developed would suggest that
high. Clearly, therefore, three-way interactions are of the attitudinal downturn should occur earlier in print
interest, but no comprehensive theory of such inter- than in TV. Mitchell and Olson (1977) did find that
actions yet exists. repetition of two kinds of print ads did not have a pos-
itive effect on belief strength, attitudes, or purchase in-
tentions, while most studies using TV ads have found
3Some writers have combined the "ability" and "opportunity" positive repetition effects. Thus, there seems to be sub-
variables (e.g., Petty 198 1), including the effects of distraction under
stantial prior research suggesting that the antecedents
"ability." Some others, however (e.g., Wright 1980), use "opportu-
of such support argument and counter argument pro-
nity" as another antecedent, using it to understand the effects on
cognitive response production of low respondent control over ex-duction should moderate the attitudinal effects of mes-
sage
posure time, message length, the number of arguments in the message repetition, with continuing attitude gains in situ-
(as in Calder et al. 1974), distractor thoughts evoked by competing
ations wherein such message-related thinking is re-
messages (as in Wright 1981), etc. Cacioppo and Petty (1979) con- duced.
ceptualize repetition as providing "opportunities" for message re-
sponse. While "opportunity" could, therefore, be subsumed under
"ability," we prefer to keep them distinct, including under "oppor- 4Other studies, however, suggest that minimal familiarity is nec-
tunity" those response-enabling variables not under the respondent's
essary for any repetition impact: Ray et al. (1971) found that while
own control (e.g., the number of message arguments), and includingwell-known brands have positive repetition functions, brands that
under 'ability" those that are so controllable (e.g., generalized product
are not well known have flat repetition functions for at least eight
knowledge). exposures.

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
SITUATIONAL EFFECTS OF REPETITION 435

HYPOTHESES portunity) interactions, though the interactions of rep-


etition with the opportunity variable may occur through
Hi: Repetition will lead to continuing gains (in- different mechanisms in the two motivation/ability
creases over control) in purchase intentions conditions: cognitive response production in the high
motivation/ability level, and the just-described periph-
and attitudes for ads evoking a low number
of (claimed) brand thoughts (support argu-
eral route in the low motivation/ability level.
Finally, though data were also collected on measures
ments and counter arguments) in early (first
or second) exposures, but not for ads evoking of brand salience (recall and familiarity), repetition ef-
many such thoughts. fects on attitudes and purchase intentions via cognitive
response mechanisms are expected to be unrelated to
The following sections will describe how the ante- effects on measures of salience (cf., Cacioppo and Petty
cedent variables of support argument and counter ar- 1979, pp. 103-104, who found no interactions with fre-
gument production were operationalized in this study. quency on recall measures, and very low correlations
In terms of manipulations, the motivation and ability of recall with cognitive response production). Thus we
antecedents-which are expected to be highly correlated do not hypothesize any interactions of repetition with
(as described earlier)-were combined as one factor, these salience measures.
and opportunity was treated as the other variable of
interest. While it would have been preferable to keep
METHOD
them distinct, pilot testing (see Batra 1984) showed that
a manipulation of motivation through the product cat- In this study, 131 housewives were exposed to three
egory of the advertised brand also varied the ability an- 30-second TV test commercials each in a design that
tecedent, since respondents knew more about those varied product category and brand usage (manipulation
product categories they cared more about. Hypothesis
of antecedent motivation/ability), ad execution (ma-
1, which relates to the effects of support and counter
nipulation of antecedent opportunity), and frequency
argumentation on repetition responses, is thus extended level. The ads were embedded in program context with
to these antecedent factors individually. Specifically: an appropriate cover story, and the exposures were dis-
H2: Repetition will lead to continuing gains in tributed over two sessions, one week apart. The gains
brand purchase intentions and attitudes for over an internal, zero-exposure control group were an-
ads low in antecedent motivation/ability, but alyzed for brand purchase intentions, attitudes, and fa-
not for ads high in such antecedent motiva- miliarity; differences in ad recall were also studied.
tion/ability.
Design
H3: Repetition will lead to continuing gains in
The objective of the study was to study differences
brand purchase intentions and attitudes for
in repetition effects across different levels of the ante-
ads low in antecedent opportunity, but not
cedents of support/counter argument production: the
for ads high in such antecedent opportunity.
motivation, ability, and opportunity to respond to the
As will be explained later, the opportunity manipu- ad in such attribute fashion. However, as mentioned
lations were conducted in terms of the execution styles earlier, the motivation and ability antecedents are ex-
of the stimulus ads, with ads containing more attribute pected to be correlated, since high motivation to make
arguments providing greater opportunity for support careful brand choices would lead respondents to in-
and counter argumentation. Notice that no interactions crease their knowledge about how such better choices
are posited between the motivation/ability antecedent could be made. Thus, in this study both were treated
and the opportunity antecedent in their responses to as one antecedent variable. The design used was a 2
different exposure levels. This is because of recent re- X 2 X 3 factorial, with the first two factors (between
search suggesting that the number of message arguments subjects) being antecedent motivation/ability (high,
(the opportunity antecedent here) acts to raise attitudes low) and antecedent opportunity (high, low). The third
under conditions of both high and low motivation/ factor was a within-subjects frequency treatment, with
ability, although through different "routes." Petty and each respondent seeing each ad (a 30-second TV com-
Cacioppo have recently shown (1984, p. 78) that as long mercial) either one, two, or four times. (As a conse-
as the messages are "cogent" (strong rather than weak), quence of the scheduling patterns to be described later,
the presence of more arguments in the message increases the repetition treatment in this experiment is necessarily
attitudes either by increased support argument produc- confounded with some scheduling operationalizations
tion (in high motivation/ability situations) or by acting used; see section on scheduling.) The choice of a within-
as a peripheral cue facilitating attitude change (in low subjects treatment for frequency was made primarily
motivation/ability conditions). We thus do not expect on statistical grounds: it would control for individual
any three-way (repetition X motivation/ability X op- differences and thus increase the power of the test.

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
436 THE JOURNAL OF CONSUMER RESEARCH

Table 1 presents the design. To increase generality, TABLE 1

and to make possible the administration of ads at dif- EXPERIMENTAL DESIGN FOR REPETITION STUDY
ferent test frequencies to the same individual, the three
repetition levels within each between-subjects cell each Ad execution ("opportunity")a
used three different replicates of the ad type representing
the between-subjects treatment of that cell (e.g., high High (many arguments) Low (few arguments)
Motivation/
motivation/ability product category, low opportunity
Abilitya 1 b 2b 4b 1b 2b 4b
execution). A subject would thus see three test ads rep-
resenting her between-subjects treatment: the first High HM, c HM2 HM3 HF1 d HF2 HF3
would be seen once, the second twice, the third four HM2 HM3 HM, HF2 HF3 HF,
times. To avoid confounding a frequency level with a HM3 HM, HM2 HF3 HF1 HF2
particular ad (replicate), these ads were partially coun- Low LM1 LM2 LM3 LF1 LF2 LF3
terbalanced across frequency levels using a latin square LM2 LM3 LM, LF2 LF3 LF1
design (making the total design a split plot). Each be- LM3 LM1 LM2 LF3 LF1 LF2
tween-subjects cell thus had three replicate-frequency
Between-subjects: Antecedent motivation/ability (2 levels) X Antecedent opportunity (2
groups, with an average of 12 respondents per group levels).
(before no-shows and dropouts; actual n's are slightly b Within-subjects: Frequency of exposure.
c To be read: High motivation/ability, many arguments, replicate ad execution # 1
lower). Each of the three within-cell replicate-frequency
to be read: High motivation/ability, few arguments, replicate ad execution #1.
groups was run in two separate subgroups of six re- NOTE: The experimental design had two between-subjects factors at two levels each (mo-
spondents each to avoid intergroup history confounds tivation/ability and opportunity) and one within-subjects factor at three levels (frequency of
exposure). Each between-subjects oell in this design had three "replicate groups," using
(see also Footnote 6). The sessions were balanced across latin square counterbalancing. Data for each of these replicate groups was collected in two
day of week and time of day. sessions, from equal sub-groups.

In addition to providing data on the three test ads in


their treatment, respondents also provided data on the
three brands used in the experiment that did not appear (product categories) were used in all. Data from the
in their cell, thus providing a control (zero exposure) pilot study showed the selected stimuli differing signif-
condition for the analysis of treatment effects through icantly across both the motivation and ability measures.
gain scores. Since these were real ads for real brands, To further add to the potency of this "ability" manip-
with differing pre-exposure attitude and intentions ulation, the specific brands chosen for the high moti-
scores, a "pre-post" rather than "post-only" dependent vation/ability treatments were market leaders, implying
variable is clearly appropriate. However, a pre-exposure a high probability that the respondents would have used
measurement could be reactive, and so this "internal them before and thus have high usage-based knowledge
control" strategy was employed for the computation of about them; those brands chosen for the low levels were
gain scores. not market leaders. Pilot study data showed such dif-
ferences for prior usage for the selected brands to be
directionally as desired. These data also showed that
Stimuli
while higher usage brands had higher pre-exposure at-
Antecedent motivation/ability was first manipulated titudes (r = 0.50, p < 0.01), such higher usage did not
through the choice of the product category in the ad, appear to raise total support and counter argument
and selection was made on the basis of data collected production by itself (r = 0.03, n.s.). (The possibility of
from another study not described here (Batra and Ray usage differences confounding results is discussed later.)
1985) that measured the extent to which respondents Antecedent opportunity was manipulated through
rated the category as one in which brand choices mat- the choice of the execution of the ad, with a "many
tered to them (motivation) and the extent to which they attribute arguments" execution considered high in an-
considered themselves knowledgeable about that cate- tecedent opportunity, and "few attribute arguments"
gory through usage or otherwise (ability). As expected, executions considered low in such opportunity. The
such antecedent motivation and ability were signifi- presence of more attribute statements within the ad
cantly correlated in the earlier study (r = 0.49, p < 0.01, provides greater opportunity for the viewer to generate
n = 480). Three different brands, representing three dif- support arguments and counter arguments (cf., the re-
ferent product categories, were used as replicates within sults of Chaiken 1980 and Petty and Cacioppo 1984).
each of the two (high/low) motivation/ability treat- The ads were categorized based on pilot data (Batra
ments. For high motivation/ability the product cate- 1984, p. 383) on how much attribute information was
gories were photographic film, deodorants, and facial perceived to be in them: the mean informativeness rat-
moisturizers; for low motivation/ability, the categories ing was 2.79 for low opportunity and 3.88 for high op-
were instant coffee, instant chocolate drink mixes, and portunity (p < 0.01, n = 142). Results from this pilot
frozen pizzas. For reasons of confidentiality, brand study confirmed that ads rated as more informative
names cannot be disclosed. Thus, six different brands generated significantly more total support and counter

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
SITUATIONAL EFFECTS OF REPETITION 437

arguments (p < 0.01), though the absolute magnitude TABLE 2

of this correlation was low (r = 0.13). (Such low cor- SCHEDULING OPERATIONALIZATIONS
relations should be expected because for these real ads
on real brands, such cognitive responses are the out- Week I Week II
comes of message content interacting with prior atti- "Too Close for Comfort" "The Jeffersons"
tudes and knowledge structures, and are thus dependent (Daytime rerun break pattern) (Prime time break pattem)

on much more than just how many arguments there


Station ID Station ID
were in the ad.) The executions used across the two
Public service announcement Promotional spot
opportunity levels were pairs for the same brand, so Ad #4A (TEST)
that "number of arguments" execution differences were Filler ad

not confounded with differences in brands. However, PROGRAM PROGRAM


the pilot testing did reveal that real ads having few ar-
Ad #4A (TEST) Filler ad
guments also tended to be more "affective" in their ex-
Filler ad Filler ad
ecution, thus having a greater "emotional impact" Filler ad Filler ad
(mean emotional rating for low opportunity ads = 4.17
PROGRAM PROGRAM
and for high opportunity ads = 2.22, p < 0.01). To help
correct against such confounding, data were also col- Ad #1 (TEST) Ad #2 (TEST)
lected in the study on emotional impact and liking for Filler ad Filler ad
Filler ad Filler ad
the ad for possible use as covariates.
PROGRAM PROGRAM

Ad #4B (TEST) Ad #4B (TEST)


Scheduling Filler ad Filler ad
Filler ad Filler ad
As mentioned earlier, the test ads (embedded in pro-
PROGRAM
gram content) were distributed over two different ses-
sions, one week apart. The across-program and within- Ad #2 (TEST)
Filler ad
program scheduling operationalizations are presented
Filler ad
in Table 2. It should be noted that (1) within each rel-
evant commercial string, the test ads were given first End credits, etc. End credits, etc.

position to both equalize position string effects and to Total: 12 :30 ads Total: 11 :30 ads
create stronger manipulation effects for the test ads, (2) (4 test) (3 test)
the one-exposure ad was always shown in the first week, + ID + PSA + Promo + ID

while the two- and four-exposure conditions were evenly


divided between the two weeks, (3) the two exposures
within each weekly program for the four-exposure con-
Subjects and Procedure
dition are at least one commercial string apart, and (4)
the positions of the four-exposure and two-exposure ads One hundred and forty-four respondents were re-
across the last two strings are counterbalanced across cruited-36 for each of the four between-subjects cells;
the two sessions. after no-shows and dropouts, 131 provided usable data
The interval between last exposure and effect mea- for both sessions. These were housewives and working
surement is thus under 30 minutes for the two- and women aged 20 through 55 from the South San Fran-
four-exposure conditions and is one week for the one- cisco, California and nearby areas. They were paid $20
exposure condition. The result of the frequency/time- for their participation in the two one-hour sessions. Not
of-measurement confound built into this scheduling necessarily representative of American women gener-
choice is that the two- and four-exposure levels of de- ally, their assignment to different cells was randomized.
pendent measures could exceed the one-exposure levels The study was described as one in which university
either because of higher frequency or because they were researchers were studying the factors that influenced
closer to the time of measurement. This confound how consumers felt about the content of TV broadcasts
would not, however, explain away differences among (program content, network announcements, advertise-
the two- and four-exposure conditions themselves ments, etc.). Respondents were told that the videotaped
(which is where the downturn in dependent measures programs they would watch were actual TV programs
is expected) or differences across cells in the location already broadcast, and that since the researchers were
of the downturn, since the time-of-measurement con- interested in finding out how consumers felt about the
found would apply equally to all cells. Further, in the programs in conditions as similar as possible to watch-
analysis to follow, any interactions significant in the ing them at home, the programs would be shown to
total design were also tested for significance in the un- them complete with the sorts of announcements and
confounded two-to-four exposures part of the design. ads with which they would be seen at home.

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
438 THE JOURNAL OF CONSUMER RESEARCH

Respondents viewed the 30-minute videotape and dents in that treatment who recalled that test ad. Re-
then answered questions on the cover story (e.g., pro- spondents were asked to mention the names of brands
gram liking, viewing habits, the viewer's lifestyle), for which they remembered seeing ads in either the first
within which were embedded manipulation checks (see or second session. Brand familiarity was measured on
section on Measures) on the motivational and ability a seven-point "very unfamiliar" to "very familiar"
antecedents. Note that the procedure and measures were scale, with such familiarity coming not only from usage
identical for all frequency conditions and that no de- but also from magazines, ads, family and friends, and
pendent measures were collected in this first session in so on. The seven-point familiarity measure is expected
order to minimize possible reactivity. to be a somewhat more sensitive measure of brand sa-
In their second session, exactly one week later, an- lience than the dichotomous recall measure. Brand at-
other videotape with another program and embedded titudes were assessed through six semantic differential
commercials was played with identical viewing instruc- items: pleasant-unpleasant, useful-useless, cold-warm,
tions. Questions on prior episode viewing, episode rat- poorly made-well made, friendly-unfriendly, and the
ings, and episode likes and dislikes were then admin- best (in the category)-the worst.5 The mean for all six
istered as before. The next set of questions measured, items was used in the analysis; factor analyses revealed
in sequence, ad recall, brand familiarity, brand atti- one major factor, and the Cronbach alpha was 0.93.
tudes, purchase intentions, and brand usage. These were Brand purchase intentions were assessed on a seven-
embedded within cover story questions (e.g., program point scale anchored with "definitely would buy" and
content recall, liking, and videotaping intentions), and "definitely would not buy." (The use of a single inten-
included test, control, and filler brands. tions measure could pose potential problems of unre-
Finally, they were told that since "how you feel about liability, though the measure selected is one used fre-
the ads (in the programs) could influence your feelings quently.)
about the program" their ratings of certain ads were The test ads were rated on prior exposure (before the
desired. The three test ads in their cell were then shown study) as having been seen never, once or twice, quite
again, in sequence, and ratings were obtained for each often, or very frequently before the experiment. The
on (1) prior exposure to the ad (before the experiment), number of thoughts subjects had about the claims the
(2) the number of "thoughts about product claims" they ad was making about the product-either positive or
had while they watched the ad (manipulation check on negative (support arguments or counter arguments)-
the degree of support/counter argument production), were rated on a seven-point "many thoughts" to "few
(3) informativeness (manipulation check on the "op- thoughts" scale; this served as a manipulation check
portunity" variable), and (4) emotional impact and like- on the differences in the generation of such responses
dislike (possible covariates for an "affective-rational" across the four cells. It should be clearly understood
confound). Before leaving, they were told to write on that this is a measure of claimed-not independently
the back of the questionnaire any comments they had validated-support argument and counter argument
about either the purpose of the study or specific ques- production. Such a measure is not without precedent,
tions. These comments were used to assess the incidence however: Cacioppo and Petty ( 1985, p. 106) cite a study
of demand artifacts. by Schumann (1983) that also used a 10-point scale of
self-reported cognitive response production frequency
Measures in a study of repetition effects. Our pilot study (reported
in Batra 1984) found a significant (p < 0.0 1) correlation
The manipulation check on the motivational ante- between this self-reported measure and a count by a
cedent asked the respondent how important it was to judge (from a retrospective verbal protocol of the pro-
her, when buying any brand in each of the categories portion of ad-evoked thoughts that dealt with the
listed, that she bought exactly the brand that she did brand), although the absolute magnitude was low
(five-point scale, "most important" to "least impor- (r = 0.24).
tant"). The check on the ability antecedent (category The ads were also rated on the amount of product
knowledgeability) asked for a self-rating of her knowl- information (seven-point scale from "had no infor-
edge about which features one might look at, in choos-
ing among different brands, with such knowledge com-
ing not only from usage but also from magazines, ads,
5The choice of items used was based on the requirements of another
and friends (five-point scale, "very knowledgeable" to investigation into the dimensionality of brand attitudes. The pleas-
"quite unknowledgeable"). Prior brand usage was mea- ant-unpleasant, useful-useless, poorly made-well made, and best-
sured via classification of that brand as having been worst items are clearly evaluative and have been used extensively in
used most often, frequently, occasionally, at least once prior research. The cold-warm and friendly-unfriendly items were
used to see if they would respond as a distinctly more "hedonic"
before, or never before.
attitudinal factor; they did not in fact do so. For further details and
Ad recall was either reported correctly or not, and for reasons why the usual "good-bad," "like-dislike," and other items
was analyzed subsequently as the percentage of respon- were not used here, see Batra 1984.

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
SITUATIONAL EFFECTS OF REPETITION 439

TABLE 3
mation" to "had a lot of information"), (positive) emo-
tional impact (seven-point scale from "no emotional MANIPULATION CHECKS (OVERALL)
effect" to "strong emotional effect"), and liking or dis-
liking of the ad (seven-point scale). The first served as p-value
a manipulation check on the "opportunity" antecedent Measure Low High (d.f. = 385)
(ad execution style) variable, and the others served as
Motivation/Ability
possible covariates for the "affective-rational" con-
Motivation 2.64 3.62 <.001
founding noted earlier. Ability 2.77 3.47 <.001
Prior usage 2.81 3.42 <.001

ANALYSIS AND RESULTS Few Many


arguments arguments
Opportunity
Demand Artifacts Informativeness 4.02 4.73 <.001

Of the 131 respondents who attended both sessions, Execution covariates


two indicated that they may have suspected its stated Emotional impact 4.06 3.80 .096
intent. For this reason, analysis was restricted to the Liking of ad 4.81 4.29 .003

other 129 respondents. Claimed number of cognitive responses produced


(support/counter arguments)

Manipulation Checks Motivation/Ability


Low 3.33 (1)* 4.15 (3)
Manipulation checks (Table 3) indicated that the low High 4.53 (2) 4.23 (4)
motivation/ability cells were significantly lower than
the high motivation/ability cells on both the motivation NOTE: 'Scheffe mulple comparison tests indic
different were 1 and 2, 1 and 3, and 1 and 4.
measure (low = 2.64, high = 3.62, p < 0.001, 385 df.)
and the ability/knowledge measure (low = 2.77, high
= 3.47. p < 0.001, 385 d.f). Further, as intended, prior
indicated that this highest level was not significantly
usage was in fact higher for the high motivation/ability
different from the level for the high motivation/ability
level than for the low level (3.42 vs. 2.81, p < 0.001,
many-arguments cell (the one expected to be highest).
385 d.f).
The hypothesized differences in response between high
Analysis of the low and high "opportunity" marginals
and low "claimed cognitive response production"
showed that the "more argument" ads were seen as sig-
(CCRP) cells were therefore tested as planned, using
nificantly more informative ("few arguments" = 4.02,
cell contrasts between the high motivation/ability
"many arguments" = 4.73, p < 0.001, 385 d.f). Further,
many-arguments execution cell (high CCRP) and the
tests of the affective/rational confound showed that
low motivation/ability few-arguments cell (low CCRP).
while the "few argument" ads were significantly more
likeable ("few arguments" = 4.81, "many arguments"
= 4.29, p = 0.003, 385 df.), they were only directionally Tests of Hypotheses
stronger in emotional impact ("few arguments" = 4.06,
The repeated measures analysis of variance assump-
"many arguments" = 3.80, p = 0.097, 385 df).
tion of "compound symmetry" (Winer 1971, p. 596)
However, while the motivation/ability difference held
was tested through the hypothesis that the covariance
significantly not only for the marginals but also across
matrix of the dependent variable is a diagonal matrix
relevant individual cells, this was not true of the "num-
(Bock 1975, p. 459), using Box's M analog of Bartlett's
ber of arguments" difference. While the "informative-
test for sphericity. Since this homogeneity assumption
ness" rating was different (p < 0.001, 190 df.) across was upheld in every case to be reported here, only the
the "opportunity" cells within the low antecedent mo-
univariate results are given here. Further, since sample
tivation/ability treatment, it was not different within
sizes were slightly unequal across cells, unbalanced es-
the high antecedent motivation treatment. Probably as
timation was performed, using the unweighted means
a result, the four cells did not order as expected on the
regression approach (Winer 1971, p. 600).6 Tests for
claimed level of cognitive response production, as mea-
sured by the self-rating of whether the viewer had few/
many thoughts about the claims made by the ad about
6Note that the number of respondents rather than the number of
the brand. While the lowest cell mean was for the low experimental subgroups is used as the unit of observation, since in-
motivation/ability few-arguments execution cell, as ex- dividuals were allocated randomly to subgroups, which were allocated
pected, the highest cell mean was not for the high mo- randomly to treatments, with two subgroups per group to minimize
tivation/ability many-arguments execution cell but for inter-group history confounds. Further, experimental occasions were
randomized across time of day and day of week. (ANOVA runs using
the high motivation/ability few-arguments execution experimental sessions as covariates showed the covariates to be in-
cell (see Table 3). However, multiple comparison tests (Continued p. 440.)

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
440 THE JOURNAL OF CONSUMER RESEARCH

linear and quadratic trend were done using non-equal four. Frequency is thus not significant as a within-sub-
spacing intervals for the repetition levels, i.e., as 1, 2, jects factor (see Table 5), though the linear trend by
and 4 exposures. itself approaches significance at p < 0.10 (F = 3.09, d.f
For brand familiarity, brand attitudes, and brand 1,125).
purchase intentions, the dependent measure used was Of the interactions with frequency, only the inter-
the extent to which the treatment condition observa- action between frequency and motivation/ability is sig-
tions were higher than the mean for the control (zero nificant, at p < 0.05 (F = 4.30, d.f 2,250). While the
exposure) level, in absolute terms. Since three replicates intentions gain for high motivation/ability rises from
are involved in each cell, each with its own pre-exposure 0.289 for one exposure to 0.394 for two, it then drops
level, and with these replicates differing between the to 0.140 for four; for low motivation/ability, however,
motivation/ability levels, it is clearly more appropriate it shows almost no gain over control for one or two
to make between-cell comparisons using the mean of exposures but then rises dramatically to 0.778 for four.
each replicate's growth over its own control than to (It should be noted that here, as at other points discussed
compare absolute post-exposure cell means, which de- later, every reported mean that falls below the control
pend on the specific replicates used in each cell. These mean actually has a 95 percent confidence interval that
gain scores (averaged between replicates) were thus used includes zero gain, indicating that, for that level, the
in the analyses.7 Ad recall was analyzed as the per- mean growth was not significantly different from zero.
centage of respondents recalling the test ad. Several Obviously, this did not happen for all levels, for oth-
studies have shown that analysis of variance can be per- erwise there would be no significant results.9) Hypoth-
formed safely on nominal data if the number of obser- esis 2, therefore, is supported: the gain is greater for low
vations per cell is 30 or more (Cochran 1947; Mande- (than for high) motivation/ability. Further, this inter-
ville 1969), though the practice is controversial. action is significant even when analysis is restricted to
Only the within-subjects ANOVA results are reported the unconfounded two to four frequency range (F
below, since the between-subjects results are not of hy- = 7.97, p < 0.01, d.f = 2,250), and this level would
pothesized interest. Only two of the twelve between- stay significant even if the experimental subgroups were
subjects tests conducted were significant at p < 0.05; used as the denominator in the F-test (F to attain sig-
on both brand familiarity and recall, the differences be- nificance at p of 0.95 with 2, 20 d.f = 3.49). However,
tween the high and low motivation/ability treatments Hypothesis 3 (greater gain for low opportunity) is not
were significant. The mean square error terms in these supported (F = 1.75, d.f = 2,250), though the results
between-subjects tests, with 125 d.f, ranged from a low are directionally consistent. This interaction fails to
of 0.14 (for recall) to a high of 5.0 (for purchase inten- reach significance even when the emotional impact and
tions).8 liking covariates are used.
The contrast between high and low claimed cognitive
Purchase Intentions. As can be seen from Table 4, response production (CCRP) cells, as stated in Hy-
the mean gain over control in purchase intentions is pothesis 1 (greater purchase intentions gains for low
0.145 for one exposure, 0.128 for two, and 0.457 for than for high levels of claimed support argument and
counter argument production) is significant at p < 0.05
(F = 3.62, d.f = 2,250). (As mentioned earlier, the high
and low claimed cognitive response production cells
significant.) It could be argued that since subjects were run in such are, respectively, the high motivation/ability, many ar-
subgroups rather than individually, within-cell error variance was
guments cell and the low motivation/ability, few ar-
reduced artificially, and that the experimental sub-groups rather than
individuals should be the units of observation. In this study the effects
guments cell.) While intentions gains for high CCRP
of interest (interactions of frequency with the antecedents) are within falls from 0.35 for one to 0.08 for two exposures, before
subjects, so any reduced variance would not be correlated with across- falling further to -0.23 for four, intentions gains for
subjects variables. The effect of using individuals as data points would low CCRP starts at -0.25 for one exposure, rises to
thus be to use an inappropriately small error term. An adjustment
-0.02 for two, and keeps rising dramatically to 0.77
for this can be made by using individuals as the unit but increasing
the F required to achieve significance through adjusting the degrees for four. The interaction approaches significance at p
of freedom to reflect subgroup level analysis. While the use of
subgroups in the within-subjects statistical tests would reduce the
error degrees of freedom to 20 (there were six subgroups in each of
four cells), the test F statistic does not change markedly (for p < 0.05, 9For purchase intentions, the 95 percent confidence intervals for
from 3.04 to 3.49 and from 2,250 to 2,20 d.f), and the results reported all cell mean gains were as follows: at frequency level one, -0.19 to
in this article stay essentially unchanged. 0.48; at frequency level two, -0.19 to 0.45; and at frequency level
7An alternative to such "gain score" analysis would be to use ab- four, 0.12 to 0.80. For attitudes at the corresponding frequency levels,
solute post-exposure levels as the dependent variables, with pre-ex- the figures were -0.22 to 0.23, 0.00 to 0.43, and 0.04 to 0.53. For
posure levels serving as covariates. Such analyses were also conducted, brand familiarity at the same frequency levels, these were -0.36 to
but are not reported, since they gave results virtually identical to 0.23, -0.32 to 0.33, and 0.14 to 0.70. Figures for individual cells are
those reported here. omitted for brevity, but with sample sizes per cell ranging from 31
8Full results are available from the authors, and are omitted here to 33, standard deviations ranged from 1.08 to 2.1 1 for these measures
for the sake of brevity. and frequency levels.

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
SITUATIONAL EFFECTS OF REPETITION 441

TABLE 4

LEVELS OF REPETITION IMPACT ON "GAIN OVER CONTROL"

Low motivation/Ability High motivation/Ability Motivation/Ability Opportunity

Low High Low High Low High All


arguments' arguments arguments argumentSb Low High arguments arguments
Frequency (n =31) (n =33) (n =33) (n =32) (n =64) (n =65) (n =64) (n =65) (n =129)

Purchase Intentions

1 -.246 .231 .233 .348 -.000 .289 .001 .288 .145


2 -.023 -.254 .696 .084 -.142 .394 .348 -.088 .128
4 .768 .788 .497 -.228 .778 .140 .628 .287 .457

Attitudes

1 -.295 .195 .176 -.084 -.043 .048 -.052 .058 .003


2 -.009 .137 .546 .178 .066 .365 .277 .157 .217
4 .499 .507 .330 -.206 .503 .066 .412 .156 .283

Ad Recall (%)

1 00.0 15.2 12.1 3.1 7.8 7.7 6.3 9.2 7.8


2 32.3 33.3 21.2 18.8 32.8 20.0 26.6 26.2 26.4
4 96.8 90.9 84.9 81.3 93.8 83.1 90.6 86.2 88.4

Brand Familiarity

1 -.744 -.079 .299 .226 -.401 .263 -.206 .071 -.067


2 -.576 -.380 .748 .195 -.474 .476 .107 -.097 .004
4 .339 .342 .538 .470 .340 .504 .442 .405 .423

Lwclaimed cognitive response production (CCRP).


b High CCRP.

TABLE 5
= 0.065 when restricted to the unconfounded two- to
REPETITION IMPACT ON "GAIN OVER CONTROL": four-exposures portion of the design (F = 3.45).
WITHIN-SUBJECTS ANOVA

Brand Attitudes. The average attitude gain over


F-value
control was zero for one exposure, +0.22 for two ex-
posures, and +0.28 for four. The main effect for fre-
Purchase
Source df. intentions Attitudes Recall Familiarity quency is not significant (F = 2.08, d.f = 2,250), with
the linear trend alone significant at p < 0.05 (F = 3.78,
Frequency df = 1,125). The interactions of frequency with the
Total 2,250 1.55 2.08 201.32c 4.78c
within-subjects variables are significant only for moti-
Linear 1,125 3.Q9a 3.78a 544.33c 10.71c vation/ability, at p < 0.05 (F = 3.44, d.f = 2,250). Hy-
Quadratic 1,125 0.37 .77 3.79b .33 pothesis 2, predicting greater attitude gains for low than
Motivation/Ability with frequency
for high motivation/ability, finds support. While the
attitude gain for high motivation/ability rises rapidly
Total 2,250 4.30b 3.44b 1.35 2.63a
from 0.05 for one exposure to 0.37 for two, it subse-
Linear 1,125 7.46c 5.42b 1.57 3.68b
Quadratic 1,125 1.87 1.94 1.22 1.85 quently drops to 0.07 for four; for low motivation/abil-
ity, on the other hand, the attitude gain for one exposure
Opportunity with frequency
is -0.04 over control, rises somewhat to 0.07 for two,
Total 2,250 1.75 .86 .43 .98 but then continues to rise rapidly to 0.50 for four ex-
Linear 1,125 1.95 1.82 1.14 .63 posures. When analysis is restricted to the uncon-
Quadratic 1,125 1.59 .13 .02 1.24
founded two- to four-exposure levels only, this inter-
action stays
Motivation/Ability with opportunity withsignificant at p < 0.05
frequency (F = 6.19). Incon-
Total 2,250 .13 .10 1.35 .62
sistent with Hypothesis 3, however, high and low
opportunity ads do not respond significantly differently
Claimed cognitive response production with frequencyd
to frequency (F = 0.86, d.f 2,250), even when the "af-
Total 2,250 3.62b 3.15b 1.44 1.56 fective" ratings of the ads are used as covariates.
Hypothesis 1 predicted greater attitude gains for low
*p 0.10.
bp ?0.05.
CCRP levels. The contrast between the high and low
Cp ? 0.01. claimed cognitive response production (CCRP) cells is
d Cell 1-4 contrast.
significant at p < 0.05. Attitude gain for high CCRP

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
442 THE JOURNAL OF CONSUMER RESEARCH

rises from -0.08 for one to +0.18 for two exposures subject to those caveats. Some rival hypotheses, pur-
before falling to -0.21, while attitude gain for the low porting to explain these results, are now discussed.
CCRP cell keeps rising from -0.30 for one, through
-0.01 for two, to +0.50 for four. When tested for only
Rival Hypotheses
the two- to four-exposure levels, this interaction stays
significant at p < 0.05 (F = 4.41). First, it might be argued that the ads for the high
motivation/ability conditions were also different from
Ad Recall and Brand Familiarity. On average, across the ads in the low motivation/ability conditions in
all four cells, ad recall rose from 8 percent for one ex-
evoking more negative reactions to either message ar-
posure to 26 percent for two exposures and 88 percent
gumentation, execution style, or both, such that with
for four. The main effect of frequency is therefore highly
repeated exposures, such ads would be more likely to
significant (F = 201.3, d.f = 2,250, p < 0.001). The
show downturns rather than upturns in the dependent
linear trend is significant at p < 0.001 (F = 544.3, d.f measures of interest. This is not supported, however,
= 1,125) and the quadratic trend is significant at p =
by data from the pilot study (Batra and Ray 1985) in
0.054 (F = 3.79, d.f = 1,125). For familiarity, average
which the ads were pre-tested for the valence and quan-
gain over the control condition was -0.07 for one ex-
tity of various mediating responses they evoked (the
posure, 0.00 for two and 0.42 for four. The main effect
data are not repeated here, for the sake of brevity).
for repetition is thus significant at p < 0.01 (F = 4.78,
Second, it might be argued that ceiling effects pre-
d.fJ = 2,250). When partitioned into a linear and qua- vailed differentially between the high and low motiva-
dratic trend, only the linear trend is significant (p <
tion/ability conditions, such that ads in the high level
0.01; F = 10.71, df = 1,125). As expected, no signifi-
reached them first and thus failed to show further growth
cant interactions exist (at p < 0.05) between frequency
between the second and fourth exposures. Note that for
and either antecedent motivation/ability or antecedent
this to explain the observed results, the data must show
opportunity (ad execution style) for both recall and fa-
a pattern where the absolute levels of the dependent
miliarity. However, the motivation/ability antecedent
variables for the high motivation/ability level reached
interacts with frequency on brand familiarity at a mar-
their ceiling at the four exposure level but those for the
ginally significant level of p = 0.074 (F = 2.63, d.f
low motivation/ability level did not.
= 2,250).
Table 6 presents the absolute levels of post-exposure
attitudes and purchase intentions-as well as the gain
DISCUSSION
figures-for the two motivation/ability levels, by ex-
The results support the hypothesized interactions of posure frequency. The mean absolute levels for the
repetition effects (on attitudes and intentions) with the "high" product categories are significantly higher (p
levels of the motivation/ability antecedent of support < 0.05, d.f = 127) than the ones in the "low" level on
argument and counter argument production, and sup- the 1-7 scale of measurement for both attitudes and
port the interactions of repetition with the cells having purchase intentions for the one- and two-exposure lev-
the highest/lowest levels of such claimed production. els. This is so because more responses for the "high"
Although they are directionally consistent with the hy- level came from (presumably more satisfied) users, as
potheses, however, the results for the opportunity an- was intended in manipulating "ability." However, the
tecedent of such production (operationalized through means are not significantly different for the four-ex-
the number of arguments in the ad execution) are not posure level (p = 0.537 for attitudes, 0.868 for inten-
statistically significant. It is important to note that the tions, df = 127), so that if the dependent variables
motivation/ability results are significant even if analysis bumped into their ceilings at four repetitions, they must
is restricted to the two repetition levels (two and four) have done so for both the high and low motivation/
that do not suffer from the time-of-measurement con- ability cells, and a differential ceiling effects explanation
found mentioned earlier, as are the results for the con- would not hold. Further, the absolute values are at least
trasted levels (high, low) of claimed cognitive response a scale point away from the highest value (7). Finally,
production.10 Thus, while interpretation of the gain the high motivation/ability cell did show significant
from one to two exposures is confounded, the statisti- growth (p < 0.05 for both attitudes and intentions) over
cally significant crossover interaction of interest occurs control at the two-exposure level at the absolute ratings
between the two- and four-exposure levels and is not shown, so these two-exposure absolute levels are not
yet at the ceiling. Since the four-exposure absolute levels
are directionally lower than these two-exposure absolute
'0Post hoc contrasts showed the miotivation/ability treatment in- levels, the four-exposure absolute levels, too, must be
teracting with frequency for both intentions and attitude gains at p below the ceiling. Thus a ceiling effect explanation is
< 0.10 for the one-to-two exposures contrast and at p < 0.05 for the
unlikely, if still possible.
one-to-four exposures contrast. When just the high and low CCRP
cells were compared, the interactions with frequency were not sig- Third, it might be speculated that the ""high" moti-
nificant across frequency levels one and two, but were significant at vation/ability ads may have had higher exposure levels
p < 0.05 when frequency levels one and four were contrasted. prior to the experiment than the "low" ads, leading to

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
SITUATIONAL EFFECTS OF REPETITION 443

TABLE 6 higher in the low (r = 0.52) than in the high motivation/


CEILING EFFECT TESTS: ATTITUDES AND ability level (r = 0.35, z = 2.06, p < 0.05).
PURCHASE INTENTIONSa These differential relationships could be due to lower
variances for some of these dependent "gain" measures
Frequency in the high motivation/ability cell (F's for lower recall
and intentions gain variance differences are n.s. at 0.05,
Measure 1 2 4
but those for attitude and familiarity gain variance dif-
ferences reach significance, with 191, 194 df ). Still,
High motivation/Ability (n = 65)
they raise the possibility that even though recall and
Purchase intentions 'gain' 0.29 0.40 0.14 brand familiarity do not respond differentially to rep-
Attitudes "gain" 0.05 0.37 0.07
etition as a function of the level of antecedent moti-
Absolute post-exposure P.l. 5.14 5.28 4.89 vation/ability, their linkages to dependent measures of
Absolute post-exposure attitude 5.31 5.63 5.26 interest may be different, and the high impact of rep-
Low motivation/Ability (n = 64) etition on such "salience" measures may be transmitted
differentially to the dependent measures of interest,
Purchase intentions "gain" 0.00 ? -014b 0.78b
Attitudes "gain" -0.04 0.07 0.50b
leading to differential impact of repetition on those de-
pendent measures themselves. These data agree with
Absolute post-exposure P.l. 4.23c 4.08d 4.95
earlier results suggesting such differential impact of sa-
Absolute post-exposure attitude 4 59d 4.68d 5.10
lience in high and low motivational involvement con-
*Mean levels per cell, averaged across replicates. ditions, such as the "three hierarchies" research pre-
For tests across levels of motivation/ability, p s 0.10. sented in 1973 by Ray et al. (see also Beattie and Mitch-
c For tests across levels of motivation/ability, p s 0.05.
d For tests across levels of motivation/ability, p s 0.01.
ell 1985). This mechanism may operate in addition to
the cognitive response mechanism suggested by Ca-
cioppo and Petty (1979). Just as individuals appear to
lower measured growth within the experiment. Analysis determine their preferences on the basis of other pe-
that compared the self-reported prior (to the experi- ripheral cues only in conditions of low motivation and/
ment) exposure levels for the test ads between high and or ability, so also may they use brand salience as an
low motivation/ability levels showed, however, that no heuristic, or peripheral cue, only in such conditions.
significant differences in such prior exposure are de- This "familiarity heuristic" thus deserves further re-
tectable. search.
These three rival hypotheses, therefore, do not con-
vincingly explain the observed results. While the theo-
CONCLUSION
retical mechanism stated earlier for these results (the
differential production of support arguments and The motivation/ability interactions just discussed do,
counter arguments) was not tested in this study, some of course, require replication. This study combined
process evidence is available, though of a different kind. motivation and ability as one antecedent; as pointed
(Evidence on the differential production of support and out earlier, the two are conceptually distinct, and ought
counter arguments as a function of repetition has been to be studied as separate antecedents where possible.
provided by studies cited earlier, such as Cacioppo and Further, the operationalizations of these antecedents
Petty 1979.) ought to be different in future studies. High usage
Data from Table 4 showed that the greatest effects of brands, one of the ability operationalizations in this
repetition are quite obviously on ad (brand name) recall, study, might have added more than just ability; it might
which rises to 88 percent for four exposures, overall. also have increased support argumentation and reduced
Correlational analyses now show that such recall has a counter argumentation, and such a potential confound
significant (p = 0.01) relationship (r = 0.164) with should be avoided. Such studies could also then inves-
brand attitude gain only in the low motivation/ability tigate three-way interactions: how do the interactions
level, being low and insignificant in the high level (the of frequency with motivation depend on the levels of
difference between the two coefficients is significant at response opportunity prevailing within different levels
0.05 by Fisher's z statistic, z = 2.02). Since purchase of ability?
intention gain is related most of all to brand attitude Note also that this study did not collect mediating
gain (r = 0.84 overall), similar differential linkages are process information, in terms of the levels of support
found between recall and purchase intention gain (r and counter argumentation produced, relying instead
= 0.15 for low, r = -0.06 for high, z = 2.06, p < 0.05). on self-reports about such production which were not
Similarly, the relationship between brand familiarity independently verified. We are therefore not in a po-
gain and ad recall is much higher (r = 0.26) in the low sition to tell why the hypothesized opportunity inter-
than in the high motivation/ability level (r = -0.02, z actions with frequency did not achieve statistical sig-
= 2.79, p < 0.05), with the relationship between fa- nificance. It appears most likely that this occurred be-
miliarity gain and purchase intentions gain accordingly cause of a manipulation that was not strong enough, or

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
444 THE JOURNAL OF CONSUMER RESEARCH

that it was confounded with an "affective-rational" dif- Cochran, W.G. (1947), "Some Consequences When the As-
ference that could not be statistically controlled through sumptions for the Analysis of Variance Are Not Met,"
Biometrics, 3, 22-38.
the covariates used. Indeed, such differences in the "af-
Festinger, Leon and Nathan Maccoby (1964), "On Resistance
fective" nature of the executions could have influenced
to Persuasive Communications," Journal of Abnormal
the mediating response process in complicated and un-
and Social Psychology, 58, 359-366.
known ways: a highly affective execution might, through
Gorn, Gerald J. and Marvin E. Goldberg (1980), "Children's
raising the level of arousal, raise cognitive response Responses to Repetitive T.V. Commercials," Journal of
production even though the number of message argu- Consumer Research, 6 (March), 421-425.
ments is low (see Ray and Batra 1983 for a review), Grass, R.C. and Wallace H. Wallace (1969), "Satiation Effects
though this is obviously a speculative interpretation. of T.V. Commercials," Journal ofAdvertising Research,
Future studies in this area could thus usefully include 19, 47-57.
a check on actual cognitive response production to ex- Greenwald, Anthony G. (1968), "Cognitive Learning, Cog-
amine hypothesized mediating processes. nitive Response to Persuasion, and Attitude Change,"
in Psychological Foundations of Attitudes, eds. Anthony
G. Greenwald, Timothy C. Brock, and Timothy C. Os-
[Received December 1984. Revised July 1985.1 trom, New York: Academic Press.
Harrison, A.A. (1977), "Mere Exposure," in Advances in Ex-

REFERENCES perimental Social Psychology, Vol. 10, ed. L. Berkowitz,


New York: Academic Press.
Batra, Rajeev (1984), 'Low Involvement' Message Reception- Johnson, Eric J. and J. Edward Russo (1984), "Product Fa-
Processes and Advertising Implications, unpublished miliarity and Learning New Information," Journal of
doctoral dissertation, Stanford University. Consumer Research, 11 (June), 542-550.
and Michael L. Ray (1985), "How Advertising Works Krugman, Herbert E. (1967), "The Measurement of Adver-
at Contact," in Psychological Processes and Advertising tising Involvement," Public Opinion Quarterly, 30, 583-
Effects: Theory, Research, and Application, eds. L. Alwitt 596.
and Andrew A. Mitchell, Hillsdale, NJ: Lawrence Erl- (1968), "Processes Underlying Exposure to Advertis-
baum, 13-43. ing," American Psychologist, 23, 245-253.
Beattie, Ann and Andrew A. Mitchell (1985), "Advertising Lutz, Richard J., Scott B. MacKenzie, and George E. Belch
Recall and Persuasion," in Psychological Processes and (1983), "Attitude Toward the Ad as a Mediator of
Advertising Effects: Theory, Research, and Application, Advertising Effectiveness: Determinants and Conse-
eds. L. Alwitt and Andrew A. Mitchell, Hillsdale, NJ: quences," in Advances in Consumer Research, Vol. 10,
Lawrence Erlbaum, 129-155. eds. Richard P. Bagozzi and Alice M. Tybout, Ann Arbor,
Belch, George E. (1982), "The Effects of Television Com- MI: Association for Consumer Research, 532-539.
mercial Repetition on Cognitive Response and Message Mandeville, G.K. (1969), A Monte Carlo Investigation of the
Acceptance," Journal of Consumer Research, 9 (June), Adequacy of Standard Analysis of Variance Test Proce-
56-65. dures for Dependent Binary Variates, unpublished Ph.D.
Berlyne, D.E. (1970), "Novelty, Complexity and Hedonic thesis, University of Minnesota.
Value," Perception and Psychophysics, 8, 279-286. Markus, Hazel (1977), "Self-Schemata and Processing Infor-
Bock, R.D. (1975), Multivariate Statistical Methods in Be- mation about the Self," Journal of Personality and Social
havioral Research. New York: McGraw-Hill. Psychology, 35, 63-78.
Cacioppo, John T. and Richard E. Petty (1979), "Effects of McGuire, William J. (1964), "Inducing Resistance to Per-
Message Repetition and Position on Cognition Response, suasion: Some Contemporary Approaches," in Advances
Recall, and Persuasion," Journal of Personality and So- in Experimental Social Psychology, Vol. 1, ed. L. Ber-
cial Psychology, 37 (January), 97-109. kowitz, New York: Academic Press, 191-229.
and Richard E. Petty (1982), "The Need for Cogni- Mitchell, Andrew A. and Jerry C. Olson (1977), "Cognitive
tion," Journal of Personality and Social Psychology, 42, Effects of Advertising Repetition," in Advances in Con-
116-131. sumer Research, Vol. 4, ed. William D. Perreault, Ann
and Richard E. Petty (1985), "Central and Peripheral Arbor, MI: Association for Consumer Research, 213-
Routes to Persuasion: The Role of Message Repetition," 220.
in Psychological Processes and Advertising Effects, eds. Naples, Michael J. (1979), Effective Frequency: The Rela-
Linda F. Alwitt and Andrew A. Mitchell, Hillsdale, NJ: tionship Between Frequency and Advertising Effective-
Lawrence Erlbaum, 91 -111. ness, New York: Association of National Advertisers.
Calder, Bobby J., Chester A. Insko, and Barry Yandell (1974), Osterhouse, R.A. and Timothy C. Brock (1970), "Distraction
"The Relation of Cognitive and Memorial Processes to Increases Yielding to Propaganda by Inhibiting Coun-
Persuasion in a Simulated Jury Trial," Journal ofApplied terarguing," Journal of Personality and Social Psychol-
Social Psychology, 4 (Jan.-Mar.), 62-93. ogy, 15, 344-358.
and Brian Sternthal (1980), "Television Commercial Petty, Richard E. (1981), "The Role of Cognitive Responses
Wearout: An Information Processing View," Journal of in Attitude Change Processes," in Cognitive Responses
Marketing Research, 17 (May), 173-186. to Persuasion, eds. Richard E. Petty, Timothy M. Ostrom
Chaiken, Shelly (1980), "Heuristic versus Systematic Infor- and Timothy C. Brock, Hillsdale, NJ: Lawrence Erlbaum,
mation Processing and the Use of Source versus Message 135-139.
Cues in Persuasion," Journal of Personality and Social and John T. Cacioppo (1979), "Issue Involvement
Psychology, 39 (5), 752-766. Can Increase or Decrease Persuasion by Enhancing

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms
SITUATIONAL EFFECTS OF REPETITION 445

Message-Relevant Cognitive Responses," Journal of sions and Suggestions about Experimental Laboratory
Personality and Social Psychology, 37, 1915-1926. Research," in Buyer/Consumer Information Processing,
and John T. Cacioppo (1983), "Central and Peripheral eds. G.D. Hughes and Michael L. Ray, Chapel Hill, NC:
Routes to Persuasion: Application to Advertising," in University of North Carolina Press, 190-219.
Advertising and Consumer Psychology, eds. Larry Percy (1981), "Repetition, Cognitive Responses, and Per-
and Arch G. Woodside, Lexington, MA: Lexington suasion," in Cognitive Responses to Persuasion, eds.
Books, 3-23. Richard E. Petty, Timothy M. Ostrom, and Timothy C.
and John T. Cacioppo (1984), "The Effects of In- Brock, Hillsdale, NJ: Lawrence Erlbaum.
volvement on Responses to Argument Quality and Silk, Alvin J. and T.G. Vavra (1974), "The Influence of Ad-
Quantity: Central and Peripheral Routes to Persuasion," vertising's Affective Qualities on Consumer Response,"
Journal of Personality and Social Psychology, 46 (1), 69- in Buyer/Consumer Information Processing, eds. G.D.
81. Hughes and Michael L. Ray, Chapel Hill, NC: University
, John T. Cacioppo, and Rachel Goldman (1981), of North Carolina Press, 157-186.
"Personal Involvement as a Determinant of Argument- Stang, D.J. (1975), "The Effects of Mere Exposure on Learning
Based Persuasion," Journal of Personality and Social and Affect," Journal of Personality and Social Psychol-
Psychology, 41, 847-855. ogy, 31, 7-13.
Ray, Michael L. and Rajeev Batra (1983), "Emotion and Per- Winer, B.J. (1971), Statistical Principles in Experimental De-
suasion in Advertising: What We Do and Don't Know sign, New York: McGraw-Hill.
About Affect," in Advances in Consumer Research, Vol. Winter, Frederick W. (1973), "A Laboratory Experiment of
10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Individual Attitude Response to Advertising Exposure,"
Arbor, MI: Association for Consumer Research, 543- Journal of Marketing Research, 10 (May), 130-140.
548. Wright, Peter (1973), "The Cognitive Processes Mediating
-, Alan G. Sawyer, Michael L. Rothschild, Roger M. Acceptance of Advertising," Journal of Marketing Re-
Heeler, Edward C. Strong, and Jerome R. Reed (1973), search, 10 (February), 53-62.
"Marketing Communication and the Hierarchy of Ef- (1974), "Analyzing Media Effects on Advertising Re-
fects," in New Models for Communication Research, ed. sponses," Public Opinion Quarterly, 38, 192-204.
P. Clarke, Beverly Hills, CA: Sage, 147-176. (1975), "Factors Affecting Cognitive Resistance to
- Alan G. Sawyer, and Edward C. Strong (1971), "Fre- Advertising," Journal of Consumer Research, 2 (June),
1-9.
quency Effects Revisited," Journal of Advertising Re-
search, 11, 14-20. (1980), "Message-Evoked Thoughts: Persuasion Re-
search Using Thought Verbalizations," Journal of Con-
Roberts, Donald F. and Nathan Maccoby (1973), "Infor-
sumer Research, 7 (September), 151-175.
mation Processing and Persuasion: Counterarguing Be-
(1981), "Cognitive Responses to Mass Media Advo-
havior," in New Models for Communication Research,
cacy," in Cognitive Responses to Persuasion, eds. Richard
ed. P. Clarke, Beverly Hills, CA: Sage, 269-307.
E. Petty, Timothy M. Ostrom, and Timothy C. Brock,
Rothschild, Michael L. and Michael L. Ray (1974), "Involve- Hillsdale, NJ: Lawrence Erlbaum.
ment and Political Advertising Effect: An Exploratory Zajonc, Robert B. (1968), "Attitudinal Effects of Mere Ex-
Experiment," Communication Research, 1, 291-308.
posure," Journal of Personality and Social Psychology
Sawyer, Alan G. (1974), "The Effects of Repetition: Conclu- Monograph Supplement, 9 (2, Part 2), 1-28.

This content downloaded from


213.55.79.195 on Tue, 01 Mar 2022 07:09:39 UTC
All use subject to https://about.jstor.org/terms

You might also like