Tempted or Not?

The Effect of Recent

Purchase History on Responses to
Affective Advertising
ANIRBAN MUKHOPADHYAY
GITA VENKATARAMANI JOHAR*

Three experiments investigate the emotions that arise from buying or not buying
at an unintended purchase opportunity and how they color evaluations of affective
advertising appeals that are viewed subsequently. We demonstrate that buying
can cause happiness tempered with guilt, while not buying causes pride. Consistent
with the felt affect, respondents who had bought at time 1 subsequently prefer
happiness appeals to pride appeals, while those who had refrained prefer pride
appeals. Drawing attention to the initial purchase decision and varying the affect
by manipulating the discount both moderate this effect. These results contribute
to the literatures on self-regulation, emotions, and persuasion.

C onsumers are constantly being exposed to marketing

stimuli and go from moment to moment making strings
of decisions. There is therefore a distinct possibility that a
to buy. We focus on how these emotions then influence
responses to affective appeals that are viewed subsequently.
We examine situations of unintended purchase because
seemingly irrelevant prior decision may exert a continuing these occur frequently but have received scant research at-
influence. Yet, much of consumer research analyzes each tention. As Stern (1962) suggests, often a need is activated
decision independent of preceding decisions. It has recently only on exposure to a product. Research suggests that a
been demonstrated that emotions such as sadness and disgust substantial proportion of purchases may be in categories that
induced by irrelevant episodes may carry over and affect consumers had not intended to buy. For instance, Bellenger,
financial decisions (Lerner, Small, and Loewenstein 2004). Robertson, and Hirschman (1978) found that up to 62% of
This research investigates one such case of intertemporal purchases at a department store were for products not on
affective dynamics, where a decision at one point gives rise the shopping list that the consumer entered the store with,
to specific emotions that then drive subsequent judgments. and Heilman, Nakamoto, and Rao (2002) found that the
Specifically, we study the emotions that arise from choosing mere presence of a promotion can prompt purchases in cat-
to buy or not buy at an occasion where one had not intended egories that consumers had not even had in mind. After
consumers have had to decide to buy or not buy a chanced-
upon item, they may often be exposed to advertising or
*Anirban Mukhopadhyay (Anirban.Mukhopadhyay@ust.hk) is assistant promotional materials for completely unrelated products.
professor of marketing and a Wei Lun Fellow at the Hong Kong University
of Science and Technology, Clear Water Bay, Kowloon, Hong Kong; Gita This scenario is extremely likely in today’s marketing en-
V. Johar (gvj1@columbia.edu) is Meyer Feldberg Professor of Business at vironment and may occur in a store, online, while viewing
the Graduate School of Business, Columbia University, 3022 Broadway, catalogs, or simply when walking in a mall. Our question
New York, NY 10027. Correspondence: Anirban Mukhopadhyay. This
article is based on the first author’s doctoral dissertation at Columbia Uni-
concerns the consumer’s reactions to advertisements viewed
versity, supervised by the second author. The research was supported by subsequently for products that may be unrelated to the initial
grants from the Graduate School of Business at Columbia University and unintended purchase decision.
the Hong Kong Research Grants Council. The authors thank Agnes Chan The results described in this article contribute to three
and Martin Zemborain for invaluable research assistance, and the editor,
three reviewers, committee members Andy Gershoff, Ran Kivetz, Don
streams of research. First, they add to the small but growing
Lehmann, Eldar Shafir, and Klaus Wertenbroch, as well as Gerry Gorn, body of research on the interdependence of behaviors and
Michel Pham, and Bob Wyer, for their happy, proud, and mixed comments judgments over time. Here, endogenous to a sequence of
at various stages of this research. consumption occasions, the affect experienced on buying or
Dawn Iacobucci served as editor for this article. not buying at time T1 moderates the evaluation of adver-
tising seen at time T2. Our demonstrations of the specific
Electronically published December 12, 2006
and differentiated nature of the observed positive emotions,
445

䉷 2007 by JOURNAL OF CONSUMER RESEARCH, Inc. ● Vol. 33 ● March 2007

All rights reserved. 0093-5301/2007/3304-0003$10.00
446 JOURNAL OF CONSUMER RESEARCH

and the corresponding congruence effects over time, also of one goal at the cost of the other. Progress toward a goal
contribute to the emerging literature on discrete emotions. can generate positive affect, while lack of progress toward
Moreover, the specific emotions observed are the first ex- or giving up on a goal can generate negative affect (Carver
perimental confirmation of those reported in Gardner and and Scheier 1990). As Carver and Scheier (1998) speculate,
Rook’s (1988) interviews. The following section develops mixed emotions may therefore arise from situations involv-
the theoretical bases for predicting differentiated emotional ing trading off multiple goals. Therefore, specific to unin-
responses to decisions made at unintended purchase oppor- tended purchase, both buying and not buying should gen-
tunities and the possible consequent effects over time. erate positive affect from one source (acquisition or not
spending) but also negative affect from the other source.
THEORETICAL FRAMEWORK Goal-Based Appraisals and Specific Emotions. What
Several researchers have discussed various aspects of specific types of emotions are likely to result from this de-
the interaction between affect and self-regulation (Carver cision to buy or not buy? A key proposition in the area of
and Scheier 1990; Fishbach, Shah, and Kruglanski 2004; research on goal-based appraisals is that specific emotions
Gardner and Rook 1988; Shiv and Fedorikhin 1999). Gi- are caused by distinct cognitive appraisals (Ellsworth and
ner-Sorolla (2001) demonstrated that delayed-cost and de- Smith 1988; Ortony, Clore, and Collins 1988), which are
layed-benefit dilemmas correlated differentially with self- related to the specific “emotivational” goals and action ten-
reports of various types of affect (studies 1 and 2) and that dencies involved (Roseman, Wiest, and Swartz 1994; see
priming such affect led to differential amounts of snack also Lerner and Keltner 2000). For example, Roseman
food consumption (study 3). However, there remains rel- (1991) proposes a taxonomy of discrete emotions based on
atively little experimental evidence of how various self- various combinations of five outcome appraisals: motiva-
regulation-related situations may systematically generate tional state, situational state, probability, legitimacy, and
different emotions or how these emotions may influence causal agency. Different emotions are predicted to derive
judgments in subsequent seemingly unrelated situations. from specific combinations of these appraisals. For instance,
The following analysis first develops hypotheses about the “pride” is reported for punishment-absent outcomes that are
emotions felt upon buying or not buying at an unintended self-caused, where one deserves a positive outcome. Simi-
purchase opportunity and then discusses how these emo- larly, “joy” is reported in the presence of a certain reward
tions may influence evaluations of advertisements en- where one deserves a positive outcome.
countered subsequently. The idea that different goal-directed appraisals may be
related to differentiated emotions provides a theoretical basis
for survey findings where respondents report mixed feelings
Affect Felt at Time 1 subsequent to an impulse purchase. For instance, Gardner
Goal Conflict and Mixed Emotions. Stern’s (1962) and Rook (1993) found that happiness (“pleasure and ex-
taxonomy of impulse purchases refers to “suggestion im- citement”) was frequently reported after an impulse pur-
pulse buying” or the fact that merely viewing a product may chase, but this positive affect was equally tinged with guilt.
activate a need for it. This need is necessarily accompanied This may be explained in the appraisal terms described
by the requirement to spend money where one had not above, as happiness derives from the presence of a certain
planned to. Such temptations therefore challenge the overall reward (the purchased item), while guilt conversely derives
goal of wealth maintenance, that is, not spending money from deserved punishment caused by actions taken by the
unnecessarily (Wärneryd 1999). Fishbach, Friedman, and self (the unplanned spending). This indicates that both buy-
Kruglanski (2003) show that exposure to a temptation to ing and not buying at an unintended purchase opportunity
violate a given goal (e.g., an attractive price) may auto- should evoke specific combinations of positive and negative
matically activate the goal itself (e.g., the goal of not spend- emotions. Specifically, we hypothesize that affective conse-
ing money unnecessarily). We therefore propose that an un- quences of buying are likely to be happiness at having ac-
intended purchase opportunity may simultaneously trigger quired something inherently pleasurable and guilt for having
both—the goal of not spending and the goal served by ac- spent money unplanned, while consequences of not buying
quiring and using the product.1 These goals will be in con- are likely to be pride at protecting the goal of not spending
flict if they are activated beyond a threshold, which may along with regret at passing up a tempting object (Carmon
then cause complex affective responses, since either deci- and Ariely 2000).
sion, to buy or not buy, involves the protection or furthering However, not all unintended purchase opportunities are
the same. There may be situations where consumers expe-
1
A pilot study not reported here supports this claim. Given a purchase rience multiple positive emotions. For example, buying at
opportunity, respondents not having preexisting purchase intentions (i.e., an extremely attractive purchase opportunity, such as when
unintended purchase) reported that they thought equally about the product a tempting product is offered at a large discount, may en-
as well as the money they would have to spend when making the purchase gender both happiness from acquisition and pride at saving
decision (M’s p 4.89 vs. 4.82 on seven-point scales anchored at “not at
all”–“a lot”; F ! 1), while those with prior intent focused on the products money. In this case the consumer who buys the product may
to be acquired rather than the money to be spent (M’s p 5.41 vs. 4.32; feel that both goals—acquisition and savings—are furthered.
F(1, 131) p 26.24, p ! .0001). Consequently, both affective states are likely to result.
PURCHASE HISTORY AND AFFECTIVE ADVERTISING 447

Responses to Advertising Viewed Subsequently not buying at an unintended purchase opportunity. We then
describe two experiments that test the basic hypothesis in
How might these emotions influence responses to an ad- different ways and test important boundary conditions.
vertisement viewed subsequently? A simple mood congru- Study 2 shows that the pattern of ad evaluations differs
ence prediction would say that those experiencing positive across regular and heavy discount conditions in a manner
affect prefer appeals featuring positive affect, while those consistent with the affect felt at time 1. Study 3 then provides
experiencing negative affect prefer negative affective ap- support for the proposed affect misattribution mechanism
peals. This is not informative in this situation because all by demonstrating that the effect can be eliminated if atten-
respondents may be experiencing both positive and negative tion is drawn to the source of the felt emotion—the initial
affect. This prediction also does not distinguish between the purchase decision. To avoid demand effects, the consumption
specific emotions being felt or between different types of opportunities at time 1 and time 2 are not only intrinsically
positive or negative affective appeals. The feelings-as-in- different (scenario vs. advertisement) but also feature different
formation framework provides a relevant basis for deriving product categories (books and software). Note that endoge-
hypotheses in this regard. nous to the consumption sequence, the integral affect contin-
A vast literature has established that affective responses gent on the decision at time 1 would act as incidental affect
to targets are often used as sources of information in eval- at time 2.
uations of the target (Schwarz and Clore 1983, 1996). Under
this view, consumers in a good mood may misattribute the
source of their feelings to a target advertisement and evaluate
the ad favorably (Pham 1998). This reasoning may be ex- STUDY 1—AFFECT FELT ON BUYING OR
tended from generalized moods to specific emotions. When NOT BUYING
consumers are faced with advertising that features happiness
or pride appeals, they are likely to use their feelings as input The aim of the first study was to investigate the emotional
to their ad judgment. However, they are unlikely to spon- responses to buying or not buying at an unintended purchase
taneously reflect on the source of these emotions (i.e., the opportunity. We expected that those who had bought would
previous purchase decision). Hence, they are likely to mis- feel guilty for having spent money but happy at having
attribute their emotions to the advertising appeal and to eval- acquired something, while those who had not bought would
uate the ad more favorably when the ad is a likely source feel proud of their restraint but also regretful at the missed
of their felt emotion. When the ad features a pride appeal opportunity.
and consumers are feeling proud as a result of not giving
in to a tempting purchase, they are likely to attribute their
feelings of pride to the ad and to rate the ad favorably. A
feeling of pride cannot be easily attributed to a happiness Method
appeal; hence, the positive felt emotion is unlikely to carry
over to ad evaluations in this case. An analogous argument Three hundred twenty-four students at a large North-
can be made in the case of consumers who buy at an un- eastern university were presented a scenario in which they
intended purchase opportunity and are feeling happy at the were strolling through a mall waiting for a friend and came
time of exposure to an ad appeal. Hence, we predict that upon an attractive sale on software products that they had
happiness appeals should be preferred over pride appeals by not really intended to buy. In order to disguise the task,
those who had previously bought, while pride appeals should allow for heterogeneity in preferences, and guard against
be preferred to happiness appeals by those who had not the lone alternative bias, the temptation was presented as a
previously bought. However, if the purchase was made under choice between two equally attractive offers (see Dhar
a heavy discount, we predicted that consumers are likely to 1997). These offers had been pretested to be equally at-
feel happy as well as proud. In this case, the above reasoning tractive, and the prices and deal amounts ($45.00 discounted
suggests that both happiness as well as pride appeals will to $19.95) had been equated as well as calibrated such that
be favorably evaluated. These are our key propositions. approximately half the sample would choose to buy one of
These predictions are supported by Lee and Aaker’s the two offers while the other half would not buy either
(2004) demonstration that perceptual fluency may cause offer. Given these offers, respondents indicated whether they
messages congruent with regulatory states to be processed would buy either one or not buy at all. They were then
more effectively. Our premise is also consistent with re- presented with an alphabetized 25-item affect scale, based
search showing that the value of “feeling right” from reg- largely on Richins (1997) with relevant items added from
ulatory fit transfers onto the persuasiveness of messages Izard (1977) and Simonson (1992). These items consisted
(Cesario, Grant, and Higgins 2004). However, Cesario et al. of affective terms such as “happy,” “proud,” “guilty,” and
(2004) studied messages framed in terms of eagerness or “remorseful.” Respondents indicated the extent to which
vigilance, which are conceptually different from the affect- each item reflected how they felt when they thought of
related focus of the current study. having spent or not spent money at the sale. All responses
In the following sections, we first describe a study that were on a four-point scale anchored at 1 p not at all, 2 p
investigates the emotions that arise from either buying or a little, 3 p moderately, and 4 p strongly.
448 JOURNAL OF CONSUMER RESEARCH

Results and Discussion FIGURE 1

STUDY 1 RESULTS
The 25-item scale used to measure affect was factor an-
alyzed, and a four-factor solution explaining 63% of the
variance was obtained. The extracted factors mapped onto
the predicted constructs—guilt (eigenvalue 7.51), happiness
(5.28), pride (1.80), and remorse (1.08). Fifty-four percent
of respondents chose to buy at the unintended purchase
opportunity. In order to investigate the predicted different
affective responses to buying versus not buying, a 2 # 4
mixed ANOVA was run, with purchase decision (buy vs.
not buy) as the between-subjects factor and affect (guilty
vs. happy vs. proud vs. remorseful) as the within-subjects
factor. Scores on each affective dimension were calculated
by averaging all items that loaded higher than .5 on only
that factor.2 The between-subjects factor was significant such
that not buying caused less affect overall than buying STUDY 2—THE EFFECT OF A PRIOR
(M’s p 1.45 vs. 1.64, F(1, 323) p 20.38, p ! .0001), and PURCHASE DECISION ON SUBSEQUENT
this was qualified by a significant interaction (F(1, 323) p
AD EVALUATIONS
45.37, p ! .0001; see fig. 1). Follow-up contrasts indicated
that respondents were likely to feel more guilt if they had Study 1 demonstrated how responses to unintended pur-
bought than not bought (M’s p 1.57 vs. 1.28, F(1, 323) p chase opportunities can have complex affective conse-
19.08, p ! .0001) and also more happiness (M’s p 2.20 quences such as happiness and pride. This experiment tests
vs. 1.66, F(1, 323) p 56.74, p ! .0001), less pride (M’s p our hypotheses that happiness appeals will be evaluated
1.56 vs. 1.73, F(1, 323) p 4.40, p ! .05), and, contrary to more favorably if one has previously bought, while pride
predictions, more remorse (M’s p 1.21 vs. 1.12, F(1, 323) p appeals will be preferred if one has not bought. If the offer
is extremely attractive—for example, available at a large
4.44, p ! .05). Additional analyses conducted using factor
discount—those who buy will feel proud as well as happy
scores from the factor analysis replicated these results. because they furthered their goals of controlled spending as
These results support the hypothesis that both buying and well as acquisition. Hence, both types of appeals should be
not buying at an unintended purchase opportunity give rise effective here. Neither the pride nor the happiness appeals
to specific combinations of emotions. Specifically, buying should be as effective in the case of those who do not buy
seems to give rise to happiness tempered with guilt and a at a large discount.
little remorse, while not buying causes pride. A follow-up
study, not reported here for reasons of brevity, investigated
whether the same pattern of results would be obtained if Method
product-focused affect (“the product that you acquired or Participants and Procedure. Two hundred thirty-
did not acquire”) were solicited rather than spending-focused seven students were recruited at a large Northeastern uni-
affect (“the fact that you spent or did not spend money”). versity, namely, Columbia University; this was relevant to
It is possible that respondents who did not buy might report the stimulus design. On arrival at the laboratory, they were
more regret when asked to focus on the product as opposed randomly assigned to experimental conditions and presented
to the spending. The nature of the product was also manip- with a packet containing stimuli for several unrelated stud-
ulated, since different types of products may induce different ies, including this study. The materials for this study were
reactions. We found that the results replicated those of study at the top of the packet. The experiment was presented in
1. The same affective factors were observed, and focusing two parts, with the time 1 decision being presented as a
respondents on the product as opposed to the spending did “Decision-Making Study,” followed by a separate and ap-
not create a noticeable difference in the patterns of reported parently unrelated “Advertisement Appraisal Study.” Par-
emotions.3 Studies 2 and 3 investigate the conditions under ticipants worked through the packet at their own pace and
which these emotions have an influence on advertising that upon completion were paid, thanked, and debriefed.
is encountered subsequently.
Stimuli and Design. This experiment was in the form
2
The items used were guilt (blameworthy, guilty, upset with myself); of a 2 (level of T1 discount: regular vs. heavy) # 2 (T1
happiness (eager, enthusiastic, excited, fulfilled, happy, joyful, thrilled); decision: buy vs. not buy) # 2 (advertising appeal: pride
pride (proud, relieved, satisfied); remorse (remorseful, sad, miserable).
3
Similar patterns were observed when the time 1 decision was manip-
vs. happiness) between-subjects design. All respondents
ulated by random assignment (cf. Arkes, Kung, and Hutzel 2002) rather were presented with an unintended purchase scenario similar
than measured. to that used in study 1, except that the chanced-upon sale
PURCHASE HISTORY AND AFFECTIVE ADVERTISING 449

was for books and not software. In the regular discount in the heavy discount at T1 condition revealed the expected
condition, the list price of the books was $55.00 discounted reversal among those exposed to the pride ads, such that
to $29.95. The products and price points had been pretested those who had bought at T1 found the pride appeal signif-
to be reasonably attractive, yet not overly so, and within the icantly more attractive at T2 compared to those who had
discretionary spending budget of the subject population so not bought at T1 (M’s p 5.79 vs. 3.57, F(1, 229) p 6.72,
as to maximize goal conflict. The reduced price in the heavy p ! .01). Those who had bought under high discount also
discount condition was $19.95. Given this scenario, re- found the happiness appeal more attractive than those who
spondents indicated a decision to buy or not as well as their had not bought at T1 (M’s p 5.59 vs. 3.63, F(1, 229) p
propensity to buy on a 100-point scale. 5.68, p ! .05). Similar patterns, not reported here for reasons
All participants then moved on to the ostensibly unrelated of brevity, were observed on the Aad (a p .97) as well as
ad appraisal study, introduced as a pretest for an advertise- purchase intent (a p .92) measures.
ment that the campus bookstore was planning to run in the
university newspaper. This ad featured a sale on a set of Affect Felt at T1. Responses to the 25-item affect scale
software products, all pretested to be equally attractive to were factor analyzed, and a four-factor solution explained
this population. The ad in all conditions featured the same 62% of the total variance. While the factors obtained in
photograph of smiling students, the same body copy with study 1 were happiness, guilt, remorse, and pride, here
brief product descriptions, and the same message stating, guilt and remorse appeared collapsed into one factor, and
“As a special summer offer, Columbia University teams up anger stood out as a separate factor. The emergence of this
with leading software providers to offer you any of the top new factor provides preliminary indication that the intro-
games or personal hobby software of the year 2003.” The duction of the heavy discount might have influenced af-
nature of the appeal was manipulated by altering the head- fective responses to buying or not buying at T1. Happiness,
line, caption, and tag line, as follows. The pride appeal pride, remorse/guilt, and anger scores were computed by
featured the headline “The Pride of a Columbia Lion” and averaging items that loaded uniquely and heavily onto each
the caption “Acing the last exam. Winning the big race. individual factor. The four affect scores were then entered
Receiving deserved recognition. Proud to be a Columbia into a 2 # 2 # 4 mixed design, with discount at T1 and
Lion.” (adapted from Aaker and Williams 1998). The tag decision at T1 as the between-subjects factors, and the four
line read “Be Proud.” The headline for the happiness appeal types of affect as the repeated measure. Results showed a
was “Happy to be at Columbia!” The caption was identical significant main effect for decision at T1, such that those
except for exclamation marks instead of periods in the punc- who bought reported more intense affect than those who
tuation. The tag line read “Be Happy!” The ad appeal in all did not buy (M’s p 1.66 vs. 1.41, F(1, 233) p 26.04,
conditions was designed to be relevant to the advertised p ! .0001). This was qualified by interactions with overall
products without cuing the T1 decision in any explicit way. reported affect (F(1, 233) p 14.05, p ! .0001), discount
After viewing the ad, participants responded to an overall at T1 (F(1, 233) p 3.81, p ! .06), and a three-way inter-
ad attractiveness measure, a five-item Aad scale (Aaker and action (F(1, 233) p 6.26, p ! .05; see fig. 2, bottom pan-
Williams 1998), and a purchase intention measure, all on els). Planned contrasts revealed that under a regular dis-
13-point scales, and, finally, the same 25-item affect scale count, buying caused significantly more happiness than
as in study 1. did not buying (M’s p 2.11 vs. 1.62, F(1, 233) p 16.98,
p ! .0001), while not buying caused somewhat more pride
than did buying (M’s p 1.97 vs. 1.75, F(1, 233) p 3.14,
Results p ! .08). These results are identical to those obtained
in study 1. However, a different pattern emerges under
Ad Attractiveness at T2. A 2 (discount at T1) # 2 a heavy discount. Here, while buying is still seen to
(decision at T1) # 2 (appeal) between-subjects ANOVA cause more happiness than not buying (M’s p 2.26 vs.
on the ad attractiveness measure found a significant main 1.44, F(1, 233) p 46.31, p ! .0001), it also causes mar-
effect for decision at T1 (F(1, 229) p 5.85, p ! .05) qual- ginally more pride than not buying (M’s p 1.98 vs. 1.75,
ified by two significant two-way interactions, between deci- F(1, 233) p 3.42, p ! .07). Hence, the presence of a heavy
sion at T1 and appeal (F(1, 229) p 6.48, p ! .05) and be- discount alters the pattern of affective responses and dif-
tween decision at T1 and discount (F(1, 229) p 6.51, p ! ferentially influences the evaluations of advertising appeals
.05), and a significant three-way interaction (F(1, 229) p viewed at time 2.
8.19, p ! .005; see fig. 2, top panels). Planned contrasts were
run, comparing across T1 decision within the two types of Discussion
appeal and within the two levels of discount. As predicted,
respondents who saw a regular discount at T1 found the The pattern of results on affect reported after a T1 pur-
happiness appeal at T2 more attractive if they had bought chase decision was almost identical to the results on the T2
(vs. not bought) at T1 (M’s p 6.71 vs. 4.50, F(1, 229) p ad attractiveness, attitude, and purchase intent measures. We
7.39, p ! .01). Conversely, they found the pride appeal more suggest that consumers who buy are happy; if they are then
attractive if they had not bought (vs. bought) at T1 exposed to an ad featuring a happy appeal, they attribute
(M’s p 5.97 vs. 3.65, F(1, 229) p 7.21, p ! .01). Results their happiness to the happy ad and hence evaluate it more
450 JOURNAL OF CONSUMER RESEARCH

FIGURE 2

STUDY 2 RESULTS

favorably than consumers who do not buy and are not feeling on the sale with much less emotion. The lower amount of
happy. Respondents who choose not to buy at a regular pride reported at not buying under such conditions is pos-
discount feel a certain amount of pride; they attribute their sibly a symptom of less conflict felt. Consistent with re-
pride to the pride ad appeal and hence evaluate it more ported emotions, happy as well as pride appeals are eval-
favorably than those who do not buy. It is noteworthy that uated more favorably under a buy decision under heavy
respondents feeling proud (happy) do not respond favorably discount compared to a no buy decision.
to happiness (pride) appeals; we argue that this is because
the attribution of their emotional state to the ad is not an
easy one to make in these situations. These results argue STUDY 3—DRAWING ATTENTION TO
against simple valence-based mood congruence predictions THE INITIAL PURCHASE DECISION
and are consistent with the proposed affect misattribution
mechanism. Study 2 suggests that happiness experienced on buying
The reversal on reported pride from regular to heavy dis- at T1 colors evaluations of advertising viewed at T2. Our
count levels is noteworthy. Study 1 and the regular discount explanation of this effect relies on the affect misattribution
condition in this experiment show that not buying at an view that the feeling of happiness is misattributed to the ad
unintended purchase opportunity can cause pride. However, resulting in increased ad evaluations. If this is the case,
faced with a heavy discount, those who choose to buy report directing attention to the emotion and its source before the
feeling both happiness and pride, while those who do not ad is viewed should cause a correction in the misattribution
buy report less of both. Evidently a heavy discount causes of the emotion to the ad and hence eliminate the effect
some consumers to seize the deal, generating happiness from (Schwarz and Clore 1983). Study 3 tested this proposition.
the acquisition as well as pride, presumably from the large We restricted the investigation to happy ads viewed after a
amount of money saved. Other consumers, who may not heavy discount at T1 because this combination evoked the
have been as favorably disposed toward the product, pass most happiness on buying, as seen in study 2.
PURCHASE HISTORY AND AFFECTIVE ADVERTISING 451

FIGURE 3

STUDY 3 RESULTS

Method subjects ANOVA. Only the interaction term was significant

(F(1, 66) p 4.60, p ! .05). Planned contrasts revealed that
Seventy students were recruited at Columbia University in the no recall condition, those who had bought at T1 were
as before. The procedure mirrored that of experiment 2. significantly more favorable toward the happiness appeal
This experiment was in the form of a 2 (T1 decision: buy than those who had not bought (M’s p 6.62 vs. 4.65,
vs. not buy) # 2 (recall of T1 decision vs. no recall) be- F(1, 66) p 5.58, p ! .05), thereby replicating study 2. In
tween-subjects design. All respondents were first presented contrast, in the recall condition, there was no significant
with the heavy discount scenario used in study 2, after which difference in ad evaluations regardless of whether they had
they indicated their decision to buy or not as well as their chosen to buy or not (M’s p 5.12 vs. 5.72, F ! 1; see fig.
propensity to buy on a 100-point scale. The decision to buy 3). Calling attention to the source of the affect made re-
was measured as before. Respondents in the no recall con- spondents correctly attribute their happiness to their buying
dition then moved on to the ad appraisal study as in study decision. Because their happiness was no longer attributed
2, featuring only the happy ad. In contrast, before being to the ad, attitudes to the ad were attenuated (M’s p 6.62
presented with the ad appraisal study, respondents in the vs. 5.12, F(1, 66) p 3.43, p ! .07). This pattern of correc-
recall condition were asked to recall the book-buying sce- tion is consistent with the proposed affect misattribution
nario that they had just gone through by writing down what mechanism.
their decision had been and how it had made them feel. This
manipulation was intended to direct their attention to any Alternative Explanations. The two items measuring
affect that they may be experiencing, as well as its source. fluency were averaged to form a composite measure (a p
Then, after viewing the ad, all participants responded to the .95) that was submitted to a 2 (decision at T1) # 2 (recall)
five-item Aad scale used in study 2, a two-item fluency mea- between-subjects ANOVA. Only the main effect of recall
sure (“How easy was the ad to understand/process?”; Lee was significant, such that those who had been made to recall
and Aaker 2004), and a three-item ambivalence measure their initial decision reported greater fluency than those who
(“When you saw this ad, how uncomfortable/confused/con- had not (M’s p 8.62 vs. 6.59, F(1, 66) p 5.28, p ! .05).
flicted did it make you feel?”; Williams and Aaker 2002). There was no main effect of decision at T1 (F ! 1) and no
All measures were taken on 13-point scales. interaction (F ! 1). The finding that felt emotion as induced
by prior purchase decisions did not influence ease of ad pro-
Results cessing rules out the fluency explanation for the pattern of
results.
Attitude to the Ad. The five items assessing Aad were The three items measuring ambivalence were also aver-
averaged to form a composite measure (a p .97), which aged to form a composite measure (a p .76). A 2 (decision
was submitted to a 2 (decision at T1) # 2 (recall) between- at T1) # 2 (recall) between-subjects ANOVA on this mea-
452 JOURNAL OF CONSUMER RESEARCH

sure again revealed only a significant main effect of recall, not necessarily due to grim necessity. The goal of not spend-
in this case such that recalling the initial decision led to ing, with an eye toward wealth maintenance, is conceptually
less reported ambivalence than not recalling the decision different from the goal of saving, as visualized in extreme
(M’s p 4.14 vs. 5.82, F(1, 66) p 6.13, p ! .05). Again, the cases as belt-tightened survival. For instance, restraining
lack of a time 1 decision effect (F ! 1 ) and any interaction overexpenditure, thereby putting money in a bank account,
(F ! 1) help to rule out a differential elaboration account can by itself generate intrinsic motivation and provide plea-
for the results. sure (Wärneryd 1999, 266). Giner-Sorolla (2001, experi-
Taken together, these results support the proposed affect ment 3) demonstrates interesting effects of existing affect
misattribution account for the influence of unintended pur- on subsequent behavior, and it would be a natural extension
chases on evaluations of advertisements seen subsequently. of that research to investigate analogous effects of mixed
Our results indicate that fluency and differential elaboration emotions. However, the effect of the interaction of specific
cannot account for the observed variations in Aad across negative and positive emotions is beyond the scope of this
conditions. The finding that recalling the T1 decision leads article.
to less ambivalence toward the ad suggests that recalling Our third main contribution is that we add to the feelings-
the decision, with its attendant goal conflict, leads respon- as-information literature by demonstrating that specific emo-
dents to attribute any felt ambivalence to the T1 decision tions (as opposed to generalized moods) may be used as
rather than the ad at T2. Future research is needed to pursue input when evaluating a target object, as long as the felt
this line of inquiry and further triangulate on the process emotion is easy to attribute to the target. Feelings of pride,
driving the results observed here. for example, can be attributed to an ad that features pride
in being a member of the respondent group (i.e., student at
GENERAL DISCUSSION the university). This results in enhanced evaluations of the
advertisement. Feelings of happiness, however, are not easy
The studies described above demonstrate that complex to attribute to an ad that features a pride appeal; hence, ad
emotions can arise from decisions to either buy or not buy evaluations are not affected by felt emotion even though the
at an unintended purchase opportunity. These emotions can emotion is positively valenced.
then carry over and color evaluations of advertising that is Some important caveats apply to the results described
viewed subsequently. This research makes three key con- above. Studies 2 and 3 provided a first look into how the
tributions. First, it adds to the growing body of work study-
emotions generated at an unintended purchase opportunity
ing behavior across consumption events. The innovation
may influence subsequent judgments. The time 1 decision,
here is that affect generated at one point in time, endogenous
which generates the mixed emotions, was in both cases op-
to the consumption sequence, influences responses to an
erationalized as a scenario with the time 2 opportunity op-
unrelated advertisement at a later point in time. Second, the
erationalized as an ad response. Our results were obtained
research adds to the relatively new literature on mixed emo-
tions. Rather than combinations of generalized positive and using this specific setup, and hence we do not make any
negative affect, specific mixtures of differentiated emotions claims about generalizability. Second, specific emotional ap-
are predicted theoretically and observed empirically. We be- peals were chosen to test the theory because they were rel-
lieve that this is the first experimental demonstration that evant to the affect generated at T1 and also to control for
unintended purchase decisions may be antecedent to specific valence. Clearly, these do not represent the universe of
combinations of emotions. The specific emotions caused by affective appeals. As mentioned above, it would be inter-
buying or not buying on these occasions are predicted to esting to look at other emotions, such as guilt and remorse
arise based on the view of emotions as arising from goal- as observed in study 1, and combinations of emotions in
based appraisals. Our research can be viewed as an empirical specificity as well as valence. A third question concerns
test of the appraisals underlying pride and happiness. the bracketing of the two episodes. As Read, Loewenstein,
It is important to note that while the theoretical framework and Rabin (1999) discuss, choices may be bracketed in sev-
we use might appear similar to that developed by Giner- eral ways. The bracketing here was temporal, with the T2
Sorolla (2001), there are some key differences. Conflict gen- appeals following closely on the T1 scenario. We cannot
erated at a purchase decision is conceptually different from tell whether the affect is strong enough to linger across other
the guilty pleasure–grim necessity distinction, because it is such situations or is too ephemeral to transcend all but the
not strictly true that people have an immediate goal to ac- strongest bracketing. One instance in which our operation-
quire (one that leads to short-term pleasure) and a long-term alization seems particularly relevant is shopping on the In-
goal to save money (that leads to guilt or regret). People ternet—presumably making a purchase decision and putting
may plan consumption well in advance, and the goal served an item into the shopping basket should enhance feelings
by acquisition and usage may also be realized independent of happiness and hence responses to happiness-based ap-
of hedonic value either immediately or in the long term, peals seen on the next Web page. This has implications for
depending on the type of product considered. In this context, the design of Internet shopping sites.
we observe the same pattern of emotional responses across Future research can build on these demonstrations in sev-
both virtues (long-term benefit) and vices (short-term ben- eral ways. For instance, it would be interesting to study
efit). Moreover, we believe that restraint from buying is also responses to mixed emotional appeals, since it is possible
PURCHASE HISTORY AND AFFECTIVE ADVERTISING 453

that a T1 decision that generates mixed emotions may also ——— (1993), “In the Mood: Impulse Buying’s Affective An-
increase tolerance for mixed emotions at T2 (Williams and tecedents,” Research in Consumer Behavior, 6, 1–28.
Aaker 2002). This would imply that a mixed pride and Giner-Sorolla, Roger (2001), “Guilty Pleasures and Grim Ne-
happiness appeal would work best after a decision to buy cessities: Affective Attitudes in Dilemmas of Self-Control,”
Journal of Personality and Social Psychology, 80 (Febru-
at a heavy discount. There may also be differences in emo-
ary), 206–21.
tions experienced at the time of making the decision as Heilman, Carrie M., Kent Nakamoto, and Ambar Rao (2002),
opposed to the point of payment, which also may influence “Pleasant Surprises: Consumer Response to Unexpected In-
responses to types of affective ad appeal. Indeed, the inten- Store Promotions,” Journal of Marketing Research, 39 (May),
sity of affect felt on exposure to an ad may itself be de- 242–51.
pendent on prior state. Further, Fishbach et al. (2003) dem- Izard, Carroll E. (1977), Human Emotions, New York: Plenum.
onstrate individual differences in counteractive control Lee, Angela Y. and Jennifer L. Aaker (2004), “Bringing the Frame
effects, which may lead to variations in the extent of ex- into Focus: The Influence of Regulatory Fit on Processing
perienced conflict and thereby moderate our observed ef- Fluency and Persuasion,” Journal of Personality and Social
fects. As such, the nature of the products viewed at T1 and Psychology, 86 (February), 205–18.
T2, the joint effects of affect and cognition, and various Lerner, Jennifer S. and Dacher Keltner (2000), “Beyond Valence:
different types of bracketing are all variables whose effects Toward a Model of Emotion-Specific Influences on Judgment
and Choice,” Cognition and Emotion, 14 (July), 473–93.
may only be guessed at. All of these suggest that the effect
Lerner, Jennifer S., Deborah A. Small, and George Loewenstein
of emotions across unintended purchase opportunities is an (2004), “Heart Strings and Purse Strings: Carryover Effects
area with much potential for future research. of Emotions on Economic Decisions,” Psychological Science,
15 (May), 337–41.
Ortony, Andrew, Gerald L. Clore, and Allan Collins (1988), The
Cognitive Structure of Emotions, Cambridge: Cambridge Uni-
REFERENCES versity Press.
Pham, Michel T. (1998), “Representativeness, Relevance, and the
Aaker, Jennifer L. and Patti Williams (1998), “Empathy versus Use of Feelings in Decision Making,” Journal of Consumer
Pride: The Influence of Emotional Appeals across Cultures,” Research, 25 (September), 144–59.
Journal of Consumer Research, 25 (December), 241–61. Read, Daniel, George Loewenstein, and Matthew Rabin (1999),
Arkes, Hal R., Yi-Han Kung, and Laura Hutzel (2002), “Regret, “Choice Bracketing,” Journal of Risk and Uncertainty, 19
Valuation, and Inaction Inertia,” Organizational Behavior and (December), 171–97.
Human Decision Processes, 87 (March), 371–85. Richins, Marsha L. (1997), “Measuring Emotions in the Con-
Bellenger, Danny N., Dan H. Robertson, and Elizabeth C. Hirsch- sumption Experience,” Journal of Consumer Research, 24
man (1978), “Impulse Buying Varies by Product,” Journal of (September), 127–46.
Advertising Research, 18 (December), 15–18. Roseman, Ira J. (1991), “Appraisal Determinants of Discrete Emo-
Carmon, Ziv and Dan Ariely (2000), “Focusing on the Forgone:
tions,” Cognition and Emotion, 5 (May), 161–200.
How Value Can Appear So Different to Buyers and Sellers,”
Roseman, Ira J., Cynthia Wiest, and Tamara S. Swartz (1994),
Journal of Consumer Research, 27 (December), 360–70.
“Phenomenology, Behaviors, and Goals Differentiate Discrete
Carver, Charles S. and Michael F. Scheier (1990), “Origins and
Emotions,” Journal of Personality and Social Psychology, 67
Functions of Positive and Negative Affect: A Control-Process
(August), 206–21.
View,” Psychological Review, 97 (January), 19–35.
Schwarz, Norbert and Gerald L. Clore (1983), “Mood, Misattri-
——— (1998), On the Self-Regulation of Behavior, Cambridge:
bution, and Judgments of Well-Being: Informative and Di-
Cambridge University Press.
rective Functions of Affective States,” Journal of Personality
Cesario, Joseph, Heidi Grant, and E. Tory Higgins (2004), “Reg-
ulatory Fit and Persuasion: Transfer from ‘Feeling Right,’” and Social Psychology, 45 (September), 513–23.
Journal of Personality and Social Psychology, 86 (March), ——— (1996), “Feelings and Phenomenal Experiences,” in Social
388–404. Psychology: Handbook of Basic Principles, ed. E. Tory Hig-
Dhar, Ravi (1997), “Consumer Preference for a No-Choice Option,” gins and Arie W. Kruglanski, New York: Guilford, 433–65.
Journal of Consumer Research, 24 (September), 215–31. Shiv, Baba and Alexander Fedorikhin (1999), “Heart and Mind in
Ellsworth, Phoebe C. and Craig A. Smith (1988), “Shades of Joy: Conflict: The Interplay of Affect and Cognition in Consumer
Patterns of Appraisal Differentiating Pleasant Emotions,” Decision Making,” Journal of Consumer Research, 26 (De-
Cognition and Emotion, 2 (December), 301–31. cember), 278–92.
Fishbach, Ayelet, Ronald S. Friedman, and Arie W. Kruglanski Simonson, Itamar (1992), “The Influence of Anticipating Regret
(2003), “Leading Us Not Unto Temptation: Momentary Al- and Responsibility on Purchase Decisions,” Journal of Con-
lurements Elicit Overriding Goal Activation,” Journal of Per- sumer Research, 19 (June), 105–18.
sonality and Social Psychology, 84 (February), 296–309. Stern, Hawkins (1962), “The Significance of Impulse Buying To-
Fishbach, Ayelet, James Y. Shah, and Arie W. Kruglanski (2004), day,” Journal of Marketing, 26 (April), 59–62.
“Emotional Transfer in Goal Systems,” Journal of Experi- Wärneryd, Karl-Erik (1999), The Psychology of Saving: A Study
mental Social Psychology, 40 (November), 723–38. on Economic Psychology, Cheltenham: Elgar.
Gardner, Meryl P. and Dennis W. Rook (1988), “Effects of Impulse Williams, Patti and Jennifer L. Aaker (2002), “Can Mixed Emo-
Purchases on Consumers’ Affective States,” Advances in Con- tions Peacefully Coexist?” Journal of Consumer Research,
sumer Research, 15, 127–30. 28 (March), 636–49.