how prevalent is the negativity effect in consumer environments?

Journal of Consumer Research, Inc.

How Prevalent Is the Negativity Effect in Consumer Environments?Author(s): Rohini AhluwaliaSource: Journal of Consumer Research, Vol. 29, No. 2 (September 2002), pp. 270-279Published by: The University of Chicago PressStable URL: http://www.jstor.org/stable/10.1086/341576 .

Accessed: 18/09/2013 10:52

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR todigitize, preserve and extend access to Journal of Consumer Research.

http://www.jstor.org

This content downloaded from 130.60.206.43 on Wed, 18 Sep 2013 10:52:19 AMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=ucpress

http://www.jstor.org/stable/10.1086/341576?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


270

� 2002 by JOURNAL OF CONSUMER RESEARCH, Inc. ● Vol. 29 ● September 2002All rights reserved. 0093-5301/2003/2902-0009$10.00

Re-Inquiries

How Prevalent Is the Negativity Effect inConsumer Environments?

ROHINI AHLUWALIA*

The negativity effect, or the greater weighing of negative as compared with equallyextreme positive information in the formation of overall evaluations, is widely be-lieved by media planners and appears to be a well-proven phenomenon in con-sumer psychology. Although this effect has been extensively documented underconditions of moderate to high processing involvement in the literature, its ro-bustness in consumer environments may be overstated. Specifically, there areimportant differences between the experimental settings in which this effect hastypically been obtained and marketplace conditions. For instance, subjects in paststudies have typically evaluated unknown or hypothetical targets with the goal offorming an accurate impression. In the marketplace, consumers may be familiarwith brands and likely to process brand-related information with a variety of otherprocessing goals, such as impression and defense motivation. Using two experi-ments, this re-inquiry delineates conditions under which the negativity effect islikely to emerge and those under which it may be less likely to occur.

Media gurus often give negative news quadruple weightas compared with positive news, in accordance with

the Merriam formula used to compute the persuasiveness ofmedia (Kroloff 1988). In other words, the negativity effect,or the greater weighing of negative as compared with equallyextreme positive information in the formation of judgments,is a well-accepted assumption by both managers and aca-demics (e.g., Bunker 1996; Herr, Kardes, and Kim 1991).The extensive use of negative advertising in the most recentas well as past presidential campaigns attests to the strongbelief of the media managers in this assumption.

Although the negativity effect is a robust finding in thisliterature, some exceptions have emerged in recent years(e.g., Block and Anand-Keller 1995; Skowronski and Carls-ton 1987). Specifically, it has been absent when the negativeinformation is not considered to be diagnostic or informative

*Rohini Ahluwalia is associate professor and Charles W. Oswald FacultyFellow at the School of Business, University of Kansas, Summerfield Hall,Lawrence, KS 66045 ([email protected]). The author acknowledges the help-ful input of the editor, associate editor, and reviewers. In addition, theauthor thanks Bob Burnkrant and Zeynep Gurhan-Canli for their insightfulcomments on a previous version of this manuscript. This research wassupported by a grant from the General Research Fund at the University ofKansas.

(e.g., Skowronski and Carlston 1987). A related stream ofresearch on message-framing effects in public service cam-paigns has uncovered additional exceptions. In this research,information is framed as either a benefit of performing (pos-itive) or a negative consequence of not performing (nega-tive) the advocated behavior, with a negativity effect im-plying greater persuasion with negative as compared withpositive framing. The persuasive advantage of negativeframing, however, has not been observed under conditionsof low information elaboration, for example, low issue in-volvement (Maheswaran and Meyers-Levy 1990) and highmessage efficacy (Block and Anand-Keller 1995), both ofwhich reduce the need to carefully scrutinize the message(see Shiv, Edell, and Payne [1997] for an exception).

The current re-inquiry examines two additional factorslikely to moderate the well-known negativity effect underconditions previously known to be conducive to it (moderateto high involvement). Importantly, these factors focus onthe differences between the settings of past research studiesand conditions found in naturalistic environments and raisequestions relating to the generalizability of this phenomenonto consumer settings. For instance, in research settings inwhich the negativity effect has been obtained, subjects eval-uate targets with the goal of forming an accurate impression.



NEGATIVITY EFFECT IN CONSUMER ENVIRONMENTS 271

In contrast, consumers tend to process information aboutbrands with a variety of personal and situational processinggoals, such as defending their prior beliefs and managingtheir impression with others (Chaiken, Giner-Sorolla, andChen 1996). Also, the negativity effect typically has beenexamined in the context of unknown targets, while consum-ers are more likely to be familiar with brands they use. Thesedifferences are the focus of this re-inquiry.

Two experiments were conducted to examine these issues.Experiment 1 examines the role of brand familiarity, whileexperiment 2 manipulates the processing motivation of theconsumer. The results suggest that pervasiveness of a neg-ativity effect may be overstated on the basis of past liter-ature. Conditions under which this effect is likely to emerge,be attenuated, and even be reversed in the marketplace arediscussed.

CONCEPTUAL FOUNDATION

The Negativity Effect and the Role ofDiagnosticity

Consistent with past research (e.g., Herr et al. 1991; Ma-heswaran and Meyers-Levy 1990), the negativity effect isdefined here as the greater weighing of negative as comparedwith equally extreme positive information in the formationof evaluative judgments. It is assessed via the relative weightgiven to negative versus positive information in the for-mation of evaluations.

It has its roots in the impression formation research inthe 1970s and 1980s, which focused on evaluations of un-known targets under conditions of accuracy motivation (e.g.,Anderson 1981; Wright and Weitz 1977). The most acceptedexplanation for the negativity effect is that negative infor-mation is perceived as more useful or diagnostic than pos-itive information for categorizing targets into evaluative cat-egories (Herr et al. 1991). Therefore, negative informationreceives greater weight in evaluations. This explanation im-plies that the negativity effect should be attenuated if theperceived diagnosticity of negative information is lowered.Consistent with this rationale, Skowronski and Carlson(1987) demonstrate a reversal of the negativity effect in thecontext of ability-related human behaviors, a domain inwhich the diagnosticity of negative information tends to below. This is because an individual’s successes are moreinformative about his abilities than his failures.

Although negative information typically has been per-ceived as highly diagnostic in the context of product judg-ments, it is important to note that diagnosticity judgmentsare a subjective assessment. In addition to valence, they aredependent on factors that influence the evaluative directionof information processing (e.g., goals of the perceiver).Goal-based directional processing often leads to inferentialbiases. These biases are typically manifested as goal-con-sistent overestimation or underestimation of the diagnosticvalue of a piece of information (Herr et al. 1991).

In other words, negative product-related information hasbeen perceived as more diagnostic than positive information

in settings in which subjects are expected to engage in rel-atively open-minded processing and are accuracy motivated.The question is, would the negativity effect still emergewhen factors that influence the nature of information pro-cessing and, therefore, diagnosticity assessments exist in theenvironment? Two such factors are brand familiarity andgoals of consumers.

Brand Familiarity

When consumers are familiar with a brand and thereforehave prior beliefs and attitudes, although they may not nec-essarily have strong feelings or past experiences related toit, their preexisting cognitive structure is expected to guidethe interpretation and integration of any new information(Petty and Cacioppo 1986). As such, existence of a priorattitude, even if weak, may lead to consistency-based pres-sures in information processing (Russo, Meloy, and Medvec1998). These pressures result in a biased memory searchsupporting attitude-consistent external information, accept-ing it at face value (e.g., Ditto et al. 1998) and therebyenhancing its perceived diagnosticity.

For instance, consumers are more likely to support pos-itive information relating to a liked familiar brand as com-pared with an unfamiliar brand. In doing so, they are ex-pected to enhance the perceived diagnosticity of a new pieceof attitude-consistent positive information, perceiving it asmore diagnostic than would consumers not familiar withthis brand. However, this pattern of biased processing is notexpected with negative brand-related information since aweak positive attitude may not be sufficient to invoke morecounterarguments in response to a new piece of negativeinformation (Ahluwalia, Burnkrant, and Unnava 2000). Assuch, even consumers who like the brand are likely to rec-ognize its diagnostic value. In other words, these consumersare expected to perceive the diagnosticity of both positiveand negative information as high, weighing them bothequally and thereby exhibiting an attenuation of the nega-tivity effect. Consumers who dislike the brand, however, arelikely to demonstrate a negativity effect since negative in-formation is attitude consistent for them.

In the current research, I will focus on consumers wholike the brand since this scenario is more representative ofthe marketplace; even though all consumers dislike a fewbrands, their attitudes toward most brands tend to be mod-erately positive (Mizerski 1982).

H1: When exposed to positive information relating toa brand, subjects who are familiar (vs. unfamiliar)with it are likely to (a) generate more supportarguments, (b) perceive it to be more diagnostic,and (c) give it more weight.

H2: Subjects exposed to new information relating to afamiliar brand are likely to (a) elicit an equivalentnumber of support arguments in the positive andnegative conditions, (b) perceive positive infor-mation to be as diagnostic as negative information,



272 JOURNAL OF CONSUMER RESEARCH

and (c) weigh positive information as much as neg-ative information in forming their brand evaluation.

Information-Processing Goals of the Consumer

Much of information processing in the naturalistic en-vironments is goal driven (Chaiken et al. 1996). Importantly,most models of persuasion have traditionally assumed thatthe primary information-processing goal of individuals isaccuracy (e.g., Petty and Cacioppo 1986), or the desire tohold attitudes and beliefs that are objectively valid. Recenttheorizing (e.g., Chaiken et al. 1996; Hilton and Darley1991), however, has concluded that even though it is rela-tively common for student subjects to be accuracy moti-vated, alternative processing goals are likely to dominate innaturalistic environments. Two such goals are defense mo-tivation and impression concerns. Defense motivation, gen-erally an outcome of personal factors such as prior beliefsand attitudes, is defined as the desire to confirm the validityof a preferred position and to disconfirm the validity ofcontrary positions. Impression motivation, or the desire toexpress attitudes that address the specific interpersonal goalsarising from the social context, is associated with situationalconcerns. It biases evaluative reactions in the direction con-sistent with demands of the social situation.

The presence of goals is known to result in selectiveinformation processing, with perceivers focusing on goal-consistent information and rejecting goal-inconsistent in-formation. Such processing may affect negative and positiveinformation asymmetrically, depending on its goal consis-tency. For instance, Tykocinski, Higgins, and Chaiken(1994) found that individuals whose psychological self-dis-crepancy (mismatch between “actual” and “ought” self-be-liefs) elicited anxiety in the presence of negative outcomeswere motivated to counterargue a negatively framed mes-sage but yielded to its positively framed counterpart. Incontrast, individuals whose self-discrepancy (mismatch be-tween “actual” and “ideal” self-beliefs) caused anxiety inthe presence of positive outcomes were more motivated toyield to a negatively framed message, as compared with apositively framed message.

In our research, processing goals were operationalized viatypes of consumer involvement. This operationalization waschosen because of its direct relevance for consumer envi-ronments as well as its widespread acceptance in the liter-ature (Chaiken et al. 1996; Petty and Cacioppo 1986).

Outcome Involvement. The most well-known andwell-researched form of involvement in the literature cor-responds to accuracy motivation (Chaiken et al. 1996; Chen,Shechter, and Chaiken 1996; Johnson and Eagly 1989;Leippe and Elkin 1987; Petty and Cacioppo 1986). It isexpected to lead to a relatively impartial, open-minded treat-ment of information with the intent of forming an accurateimpression. It is typically manipulated via making a decisionimminent or using an issue that may have personal relevancefor the subject.

Under such conditions, consumers may be expected toperceive negative information related to familiar and likedbrands as more diagnostic than positive information becauseof two processes. First, a concern to be accurate is likelyto attenuate the biasing effects of brand familiarity discussedearlier, thereby lowering the perceived diagnosticity of pos-itive information. Second, as outcome involvement increases,people tend to become more risk averse (Wright and Weitz1977), focusing their attention on the negative information(Maheswaran and Meyers-Levy 1990) and thereby perceivingit as more diagnostic. Thus,

H3: Outcome-involved subjects exposed to new in-formation about a familiar brand are likely to (a)perceive the negative information to be more di-agnostic than the positive information and (b) givemore weight to the negative as compared with thepositive information.

Impression-Relevant Involvement. In contrast, a sec-ond form of involvement, often termed impression-relevantinvolvement (Johnson and Eagly 1989; Leippe and Elkin1987), is marked by a selective bias in processing aimed atsatisfying immediate social goals and, therefore, corre-sponds to impression motivation. Specifically, the subject isinvolved in processing the message because his or her re-sponse is likely to be scrutinized in some way and possiblyto be instrumental in obtaining social approval. For instance,consumer decisions may need to be justified to family mem-bers or friends. Similarly, an organizational buyer may beaccountable to others. This type of involvement is typicallymanipulated by informing subjects that they will be askedto justify and/or explain their responses to consequentialothers whose opinions may or may not be known (Chaikenet al. 1996).

In fact, whether these opinions are known or not is ex-pected to moderate the effects of this involvement (Chaikenet al. 1996). Specifically, when these views are known, im-pression-involved subjects are likely to express judgmentsthat mirror these views in order to win approval. However,when the views are not known, their processing correspondsto their desire to avoid appearing foolish in front of anaudience. That is, they want to demonstrate that they havecarefully considered all the information available. This de-sire results in a heightened use of all pieces of information,even nondiagnostic ones. A consequence is an overinter-pretation of the relevance of nondiagnostic information (Tet-lock and Boettger 1989).

It is important to note that even though outcome-involvedand impression-involved respondents have both been knownto carefully scrutinize the message and engage in elaborativemessage processing, these involvement states are likely tolead to different types of processing (Chen et al. 1996). Forinstance, Leippe and Elkin (1987) found that outcome-in-volved subjects were sensitive to variations in message-ar-gument quality, whereas impression-involved subjects at-tenuated the differences between strong and weak messages.

Therefore, if impression-involved consumers are aware




of the views of their anticipated audiences, their judgmentsof perceived diagnosticity are likely to depend on theseviews. When these views are not known, they are likely tooverestimate the relevance of positive information, perceiv-ing both positive and negative information as highly diag-nostic and weighing them equally in their decision making.The current research will focus on the latter scenario, thatis, when the views of the audience are not known. Thus,

H4: Impression-involved subjects exposed to new infor-mation about a familiar brand are likely to (a) per-ceive the negative information to be as diagnostic asthe positive information and (b) give equivalentweight to the negative and positive information.

Position Involvement. The third kind of involvementcorresponds to the goal of defense motivation. It has typi-cally been operationalized as commitment to a particularstance or brand (Chaiken et al. 1996) as opposed to beinginvolved with the issue in an unbiased, open-minded fash-ion. It occurs either because the person is publicly attachedto this stance (Ahluwalia et al. 2000; Kiesler 1971) or be-cause the issue relates to important values (Johnson andEagly 1989).

This type of message recipient is likely to counterarguenew information that contradicts his or her vested stancebut support information that validates it (Chaiken et al.1996). Thus, while position involvement, like the otherforms of involvement, is likely to lead to increased messagescrutiny, it is expected to differ in the nature of processinginvoked and the level of persuasion obtained. The selectivecognitive processing invoked by it is expected to lower theperceived diagnosticity of attitude-inconsistent negative in-formation (Ahluwalia et al. 2000) while enhancing the diag-nosticity of attitude-consistent positive information. That is,a reversal of the negativity effect, or a positivity effect, isexpected for these individuals.

H5: Position-involved subjects exposed to new infor-mation about a familiar brand are likely to (a)perceive the negative information to be less di-agnostic than the positive information and (b)weight the negative information less than the pos-itive information.

EXPERIMENT 1

Methodology

Design and Procedure. Hypotheses 1 and 2 weretested using a 2 (information valence: positive, negative) #2 (brand: unknown, known) between subjects design. Sev-enty-seven undergraduate students participated in the study.Subjects were run individually. They were informed thatthey were participating in a consumer survey being con-ducted by a market research company and the businessschool to collect consumer opinions of two products plannedfor introduction in the U.S. market. They were told they

would see at least one piece of information about each prod-uct. The subject was handed an information folder that ma-nipulated brand familiarity. In the unfamiliar-brand condi-tion, all the materials in the folder related to a filler product(Sonar cameras). In the familiar-brand condition, three pagesrelated to the target product (Avanti athletic shoes) and oneto the filler (camera). Of these, one Avanti page describedthe company as an established foreign manufacturer of ath-letic shoes planning an entry in the U.S. market; the othertwo were ads touting its durability and technology. That is,in the unfamiliar condition, the folder did not contain anyinformation relating to Avanti shoes, while in the familiarcondition, it contained two ads and a brief companydescription.

The manipulation of information valence was adminis-tered after the subject read the folder. At this time, the ex-perimenter went to his desk to get the questionnaire. Heacted surprised to find a loose-leaf page on the desk andinquired whether the subject had read it. After receiving theexpected negative response, the experimenter apologized,told the subject that the page had apparently slipped out ofthe folder unnoticed, and requested the subject to read itbefore filling out the questionnaire. The binder holes in theslipped-out page were deliberately torn to explain the mis-hap. The slipped-out page was either a positive or a negativenewspaper article (described later) relating to Avanti shoes.After reading it, subjects completed the dependent-measuresquestionnaire.

In a pretest of the familiarity manipulation, 29 subjectsfrom the subject pool that was used for the main experimentread one of the two information folders (without the slipped-out page) and reported their evaluation of (good/bad, like/dislike, favorable/unfavorable) and familiarity with (not atall familiar/very familiar) Avanti shoes (seven-point scales).Those in the familiar (vs. Sonar folder) condition reportedmore positive attitudes ( vs. 3.79,M’s p 4.81 F(1, 27) p

, ) and greater familiarity with the brand14.15 p ! .01( vs. 1.89, , ).M’s p 4.73 F(1, 26) p 35.88 p ! .01

Target Messages. The positive and negative messagesdeveloped in the format of newspaper articles containedeither positive or negative statements about the shock ab-sorption ability of the target brand of shoes. The positive(negative) message reported the findings of a recent studythat found Avanti shoes to be superior (inadequate) in theirlevel of shock absorption. The messages were adopted fromAhluwalia et al. (2000), who reported their equivalence onextremity, believability, and strength.

Dependent Variables. Subjects were first asked to listall their thoughts while reading the target article. Twojudges, blind to the conditions and hypotheses, codedthoughts into three categories: counterarguments, supportarguments, and other thoughts (Petty and Cacioppo 1986).Since support (counter) arguments consist of reasons forbelieving in (arguing against) a message, in the negativearticle condition they included negative (positive) thoughtsabout Avanti, while in the positive condition they included




TABLE 1

EXPERIMENT 1: CELL MEANS AND STANDARD DEVIATIONS

Variable

Unknown brand Familiar brand

Positive Negative Positive Negative

Support arguments .45a (.59) 1.10b (.85) 1.50a (.73) 1.47a (.96)Counterarguments 1.05a (.99) 1.00a (.86) .71a (.92) .53a (.70)Total thoughts 3.00a (1.39) 3.52a (1.34) 3.75a (1.38) 3.68a (1.27)Perceived diagnosticity

(seven-point scale) 3.38a (1.17) 5.03b (1.31) 4.83a (1.37) 5.28a (1.32)Weight (10-point scale) 4.20a (2.10) 7.25b (2.11) 6.13a (2.13) 7.26a (1.97)Mean attention (seven-

point scale) 4.50a (1.19) 4.55a (1.26) 5.16a (1.25) 5.18a (.95)

NOTE.—All contrasts are within each familiarity condition. For each familiarity condition, means in the same row that have different superscriptsdiffersignificantlyat . Standard deviations are in parentheses.p ! .05

positive (negative) thoughts about it. There was 90% agree-ment between the judges. Disagreements were resolved viadiscussion.

Next, weight given to the newspaper article in evaluatingAvanti shoes was assessed via two 10-point scales (noweight at all/a lot of weight, not at all important/very im-portant; ). Subjects assessed the diagnosticity of ther p .86information in the newspaper article on three seven-pointscales (extremely relevant/extremely irrelevant, not at alluseful/of very great use, not at all indicative/very indicative;

) on the basis of Klar (1990). Attention to the news-a p .90paper article was measured via two seven-point scales (con-centrating very little/concentrating very hard, paying verylittle attention/paying a lot of attention; ). This mea-r p .77sure was used to assess the subject’s level of involvementin message processing.

Results

Subjects unfamiliar with the brand were expected to sup-port negative information more than positive brand-relatedinformation, perceiving it as more diagnostic and giving itmore weight. However, subjects familiar with the brand wereexpected to support positive brand-related information morethan the unfamiliar subjects, perceiving it to be more di-agnostic and giving it more weight than their unfamiliarcounterparts (hypotheses 1a, 1b, and 1c). Consequently, fa-miliar subjects were likely to support both positive and neg-ative information equally (hypothesis 2a), perceiving themto be equally diagnostic (hypothesis 2b) and therefore ex-hibiting an attenuation of the negativity effect (hypothesis2c). Please refer to table 1 for the cell means.

Cognitive Responses. The 2 # 2 ANOVA on supportarguments demonstrated a significant interaction betweenbrand familiarity and information valence (F(1, 73) p

, , , ), indicating differences in3.90 p ! .05 r p .23 b p .50support arguments generated in response to positive andnegative information by the subject’s brand familiarity. Asexpected, subjects who were familiar with the brand gen-erated significantly more support arguments when exposedto positive information than subjects in the unfamiliar-brand

condition. Consequently, subjects who were familiar withthe brand tended to support the positive brand-related mes-sage as much as its negative counterpart ( , effect sizep 1 .90

), although the subjects who were unfamiliar withd p .03the brand generated significantly more support argumentsin the negative as compared with the positive condition( , effect size ).p ! .01 d p .97

Perceived Diagnosticity. The ANOVA on mean diag-nosticity scores revealed a significant interaction betweenbrand familiarity and information valence (F(1, 73) p

, , , ). Specifically, as expected,4.14 p ! .05 r p .23 b p .52subjects in the familiar-brand condition perceived positiveinformation as more diagnostic than subjects in the unfa-miliar-brand condition. Additionally, a statistically signifi-cant difference emerged between the valence conditions forthe unfamiliar subjects ( ). Consistent with predic-p ! .01tions, this difference was not statistically significant (p 1

) for the familiar subjects. More important, its effect size.30( ) was considerably smaller than the one obtainedd p .34with the unfamiliar group ( ). Additionally, maind p 1.36effects of valence ( , , ) andF(1, 73) p 12.57 p ! .01 r p .38familiarity ( , , ) were also sig-F(1, 73) p 8.23 p ! .05 r p .32nificant, indicating that negative information was perceivedas more diagnostic than positive information ( vs.M’s p 5.153.99), and information about familiar brands was consideredas more diagnostic than information about unfamiliar brands( vs. 4.16).M’s p 5.07

Weight Given to Information. A significant interac-tion emerged between brand familiarity and information va-lence ( , , , ) for theF(1, 73) p 4.00 p ! .05 r p .23 b p .51measure of mean weight. The pattern of the interaction fol-lowed the diagnosticity results. Consistent with expecta-tions, subjects in the familiar-brand condition weighed thepositive information more than subjects in the unfamiliar-brand condition, and a statistically significant negativity ef-fect emerged with the unfamiliar brand ( ). Hy-p’s ! .01pothesis 2b relating to an absence of the negativity effectin the familiar-brand condition also received support. Al-though the means in this condition were in the direction ofa negativity effect, it is important to note that this difference




was not statistically significant ( , ). Morep p .11 b p .53important, its effect size was significantly smaller (d p

) than the one obtained with unfamiliar subjects (.58 d p, , ). In other words, even though the1.48 Z p 95.00 p ! .05

data do not unequivocally support the elimination of thenegativity effect, they clearly indicate that this effect is sig-nificantly attenuated in the familiar-brand condition (as com-pared with the unfamiliar-brand condition). Additionally, themain effects of valence ( , ,F(1, 73) p 19.23 p ! .01 r p

) and familiarity ( , , ).45 F(1, 73) p 4.10 p ! .05 r p .23were also significant; the pattern was consistent with thediagnosticity results.

Message Attention. The ANOVA revealed only amain effect of brand familiarity ( , ,F(1, 73) p 5.84 p ! .05

) such that subjects paid significantly more attentionr p .27to the newspaper article in the familiar as compared withthe unfamiliar-brand condition. These data suggest that theattenuation of the negativity effect in the familiar-brand con-dition is more likely to be an outcome of the nature ofprocessing than the amount of elaboration. More important,the range of the mean attention scores (4.50–5.16 on a seven-point scale) indicates that subjects were moderately involvedin the message-processing task.

Discussion

Experiment 1 demonstrates that brand familiarity is likelyto attenuate the negativity effect. Specifically, consumerswho were familiar with the brand and liked it tended tosupport the attitude-consistent positive information signifi-cantly more than unfamiliar consumers, perceiving it asmore diagnostic than did the unfamiliar group. However,their prior positive attitude did not seem to impact theirprocessing of negative information. Both groups engaged inrelatively similar processing of this information, perceivingit to be equally diagnostic. That is, the attenuation of thenegativity effect in the familiar-brand condition was drivenby the enhanced diagnosticity of positive information in thatcondition.

Additionally, consistent with other research in marketing(e.g., Pechmann and Stewart 1990), this study finds thatsignificantly more attention is paid to information aboutfamiliar as compared with unfamiliar brands. This suggeststhat consumers may be more likely to notice news and ad-vertising relating to familiar brands. That is, it may be mostmeaningful to examine negative information effects in thecontext of familiar brands. As such, experiment 1 demon-strated that the existence of a weak positive attitude for afamiliar brand, when no processing goals dominate, signif-icantly attenuated the negativity effect. Experiment 2 wasconducted to examine the likelihood of a negativity effectwith familiar brands, when various goals dominate infor-mation processing.

EXPERIMENT 2

Method

Design and Subjects. Hypotheses 3, 4, and 5 weretested using a 3 (involvement: outcome, impression, posi-tion) # 2 (valence: negative, positive) between subjectsfactorial design. One hundred four undergraduate studentsparticipated in the experiment. Subjects were recruited overthe phone to participate in this study.

Procedure. Subjects were run individually. They wereinformed that the study was being conducted on behalf ofthe university athletic department, and it was explained thatrecent policy initiatives from the chancellor’s office requiredthe athletic department to consider student input before mak-ing major purchases. At present, the athletic department wasobtaining student input on three relatively new brands ofathletic shoes that were under consideration. In the interestof time, each student would evaluate only one of thesebrands. All subjects were given a folder containing infor-mation about Avanti shoes from experiment 1. After readingit, they were asked to provide their thoughts about Avantishoes on an audiotape. They were specifically asked to pointout the positive qualities of the brand that appealed to themso that the athletic department could gain a better under-standing of the features that were likely to appeal to students.This tape recording was conducted to accomplish the ma-nipulation of position involvement described later.

The manipulation of involvement was administered afterthe subjects thoughts were tape recorded. They were givendirections (described later) that assigned them to the dif-ferent involvement conditions. After these instructions, theexperimenter went to his desk to get a questionnaire for thesubject and went through the slipped-out-page incident de-scribed in experiment 1. Depending on the valence condi-tion, the slipped-out page was either the positive or thenegative newspaper article used in experiment 1. That is,subjects who were equated on their familiarity with the targetbrand but differed in their processing motivation were ex-posed to either a positive or a negative article about it.

After reading the article, subjects filled out a questionnairethat included key dependent measures such as cognitiveresponses, perceived diagnosticity, and weighing. Finally,subjects were debriefed and quizzed for potential hypothesesguessing. Two subjects (one position-involvement negative,one issue-involvement negative) who reported being sus-picious about the article slipping out were dropped from theanalysis.

Involvement Manipulations. It is important to notethat the involvement instructions were provided to all sub-jects at the same point in the procedure: after formation ofan attitude toward the target brand (reading the informationfolder and thought elicitation) but before exposure to thenewspaper article. In the outcome-involvement condition,the subjects were asked to carefully evaluate the brand. Theywere informed that their evaluations would play a very im-




portant role in the committee’s choice because they wereone of few students who had been asked to provide inputon these shoes (e.g., Maheswaran and Chaiken 1991).

In the impression-involvement condition, they were toldthat the purchasing committee was interested in obtainingstudent evaluations as well as understanding the reasonsunderlying them. They would have to provide a writtenjustification of their evaluation, and a representative fromthe committee would interview them later for further ques-tions and clarifications relating to their rationale (e.g., Leippeand Elkin 1987).

In the position-involvement condition, after tape record-ing of thoughts, the subject was told that the Avanti cor-poration was looking for individuals to include in their pro-motional campaign and was asked if the company could usetheir taped thoughts and photograph in their campaign. Toestablish the face validity of this cover story, it was im-portant that subjects verbalize their thoughts so the exper-imenter could place his request after listening to them. Thesubject was photographed and asked to sign a release state-ment to this effect. Importantly, since the thought-elicitationinstructions were likely to lead to favorable brand-relatedthoughts, this procedure was expected to attach the subjectto the brand and commit him/her to a favorable brand eval-uation (e.g., Ahluwalia et al. 2000). It is important to notethat subjects in all conditions underwent the same thought-elicitation procedure. The only difference in the position-involvement condition was that they signed the release andwere photographed.

In order to alleviate any self-presentation and impression-management concerns in the position- and outcome-involve-ment conditions, subjects in these conditions were told thattheir responses to the questionnaire were anonymous andwere specifically instructed not to put their name on it and,upon completion, to drop it in a collection box near the exit.These instructions were provided after administration of theinvolvement manipulation but before exposure to the targetmessage (newspaper article) and were expected to removeany impression-management pressures (Lindskold and Propst1981). A pretest, described next, was conducted to furthertest for the possibility of impression-management concernsin the position-involvement condition.

Two conditions were run ( ): position and outcomen p 43involvement with negative information. The former was po-tentially most susceptible, while the latter was least suscep-tible to impression concerns. The procedure mirrored themain experiment. A measure of impression-managementconcerns that asked subjects to indicate the degree to whichtheir answers were influenced by what others may think ofthem (seven-point scale; not at all/to a very great extent)was also included.

Thoughts elicited after exposure to the negative articlewere coded into two categories: impression-management re-lated and non-impression-management related. The formercategory included any thoughts dealing with the conse-quences of endorsing the brand publicly or managing theirimpression with others (friends, the experimenter, etc.). The

results revealed no differences in the number of impression-management-oriented thoughts ( ,F(1, 42) p .95 M’s p

vs. .00, ) as well as the level of concern with.04 p 1 .30self-presentation ( , vs. 2.00,F(1, 42) p .60 M’s p 2.38

) in the two conditions. More important, impression-p 1 .40management thoughts as well as reported concern were verylow in both conditions.

A second pretest ( ) was conducted to test the va-n p 55lidity of the involvement manipulations. In the interest ofefficiency, only the positive-message condition was run.Specifically, the cost of doubling the sample (subjects runindividually) far outweighed the extra insight provided byincluding the negative-message condition. Since the manip-ulations were administered before exposure to the targetmessage, there was no reason to expect a simple change inthat message to influence their effectiveness. More impor-tant, cognitive responses, which are an indicator of involve-ment type, were also elicited in the main experiment. Thesedata (reported in “Results”) support the validity of the in-volvement manipulations in both valence conditions.

The procedure of the pretest mirrored the main experi-ment. The dependent measures related to the processing ofthe slipped-out article. To assess impression involvement(e.g., Siegel-Jacobs and Yates 1996), subjects were asked(seven-point scales) if they thought they would have to jus-tify their judgments to others (definitely did not believe Iwould have to justify/definitely believed I would have tojustify) and how often they thought about ways of justifyingtheir judgments to others while reading the slipped-out ar-ticle (never/frequently; ). Position involvement wasr p .72assessed via two items: I feel committed to Avanti shoes;and I feel a sense of loyalty to Avanti shoes (seven-pointscale: strongly disagree/strongly agree; ). Outcomer p .90involvement has typically been assessed via effort and at-tention measures/scales. However, past research shows thatall types of involvement enhance attention and effort and,therefore, are not distinguishable on these measures (e.g.,Chen et al. 1996; Leippe and Elkin 1987). Instead, outcomeinvolvement is better assessed via criticality of thoughts; itleads to more critical thoughts than impression (Leippe andElkin 1987) and position involvement (Ahluwalia et al.2000). Thus, cognitive responses elicited in response to theslipped-out page were used to assess this variable. A mea-sure of mood was included to assess if the endorsementopportunity in the position-involvement condition inadver-tently influenced the subjects’ mood, thereby leading to amood-management effect.1

There was a significant effect of processing instructions onthe measures of impression ( ) and positionF(2, 52) p 14.37involvement ( ) as well as counterargumentsF(2, 52) p 10.57( ) and support arguments ( ),F(2, 52) p 7.16 F(2, 52) p 6.82all . Subjects in the impression-involvement con-p’s ! .05

1On the basis of Peterson and Sauber (1983), subjects indicated theiragreement on a five-point scale (strongly agree/strongly disagree): Cur-rently, I am in a good mood; As I answer these questions, I feel verycheerful; For some reason, I am not very comfortable right now; and Atthis moment, I feel edgy or irritable.




TABLE 2

EXPERIMENT 2: CELL MEANS, STANDARD DEVIATIONS, AND EFFECT SIZEOF THE CONTRASTS FOR EACH INVOLVEMENT CONDITION

Variable

Outcome involvement Impression involvement Position involvement

Positive NegativeEffect

size (d) Positive NegativeEffect

size (d) Positive NegativeEffect

size (d)

Support arguments 1.33a (.77) 2.00b (.84) .85 1.94a (1.25) 1.71a (1.21) .20 2.06a (1.07) .81b (.75) 1.40Counterarguments 1.17a (1.04) .61b (.78) .69 .22a (.55) .53a (.94) .41 .44a (.63) 1.31b (1.13) .98Total thoughts 4.22a (1.06) 4.27a (1.36) .04 4.19a (1.07) 4.00a (1.32) .16 4.06a (1.28) 4.18a (1.59) .09Diagnosticity 4.30a (1.25) 5.67b (1.30) 1.10 5.24a (1.14) 5.43a (1.27) .16 5.63a (.88) 4.56b (.92) 1.22Weight 5.58a (2.16) 8.39b (2.21) 1.32 7.56a (1.78) 8.00a (1.90) .21 7.74a (1.24) 6.09b (1.60) 1.19

NOTE.—All contrasts are within each involvement condition. For each involvement condition, means in the same row that have different superscripts differsignificantly at . Standard deviations are in parentheses.p ! .05

dition were more concerned about justifying their responsesthan the other two groups ( vs. 3.33 and 3.00,M’s p 5.60

). Subjects in the position-involvement conditionp’s ! .01reported significantly higher commitment to the brand thanthe two other groups ( vs. 2.35 and 2.18,M’s p 3.72

). The three groups did not significantly differ onp’s ! .01the mood measure ( ). Analysis of thoughtsF(2, 52) p 1.59revealed that outcome-involved subjects were more criticalin their information processing than subjects in any othercondition. They had an equivalent number of counterargu-ments and support arguments ( vs. 1.57), whileM’s p 1.28both impression- ( vs. .65) and position-involvedM’s p 2.11( vs. .36) subjects elicited more support thanM’s p 2.53counterarguments. Therefore, the validity of the involvementmanipulations was supported.

Dependent Variables. The thought-listing task, diag-nosticity, and information weight measures followed thoseused in experiment 1.

Results

Outcome-involved subjects who were likely to supportnegative information were expected to perceive it as morediagnostic than positive information and thereby demon-strate a negativity effect (hypothesis 3). Impression-involvedsubjects were likely to support both messages and, therefore,were expected to perceive both types of information asequally diagnostic, weighing them equally (hypothesis 4).Position-involved subjects who were likely to support pos-itive information more were expected to perceive it as morediagnostic than negative information, thereby exhibiting apositivity effect (hypothesis 5). Please refer to table 2 forthe cell means.

Cognitive Responses. Cognitive responses were en-listed to help assess the validity of the involvement manip-ulations. As mentioned earlier, the different types of in-volvement can be distinguished by the direction ofprocessing they elicit. As expected, outcome-involved sub-jects who were critical as well as risk averse elicited moresupport arguments in the negative (vs. positive) conditionand more counterarguments in the positive (vs. negative)

condition. Subjects who were impression involved and there-fore motivated to support the presented information did notexhibit any differences in the number of either support ar-guments or counterarguments in the two message conditions.Subjects who were position involved exhibited defensive pro-cessing, eliciting more support arguments in the positive- (vs.negative) message condition and more counterarguments inthe negative- (vs. positive) message condition.

Perceived Diagnosticity. Consistent with expecta-tions, outcome-involved subjects perceived the negativemessage as more diagnostic than its positive counterpart.Impression-involved subjects did not perceive a differencein the diagnosticity of the two messages, while position-involved subjects perceived the positive message to be morediagnostic than the negative one.

Weight Given to Information. Consistent with thediagnosticity results, outcome-involved subjects exhibited anegativity effect, giving more weight to negative as com-pared with positive information. The impression-involvedsubjects gave equivalent weight to negative and positiveinformation. Position-involved subjects exhibited a reversalof the negativity effect, giving more weight to positive ascompared with negative information.

Discussion

These results indicate that people who are familiar witha brand but differ in their type of involvement or processinggoal are likely to assess the diagnosticity of a new piece ofbrand-related positive or negative information very differ-ently. This difference in diagnosticities is likely to influencethe weight allocated to this information.

The data reveal that outcome-involved individuals whostrive for accuracy not only support a new piece of negativeinformation more than its positive counterpart but alsocounter it less. Consequently, they perceive negative infor-mation as more diagnostic than its positive counterpart, giv-ing it more weight. Impression-involved subjects, in contrast,motivated by their concern to appear thorough, supported bothpieces of information. In doing so, they tended to enhancethe perceived diagnosticity of positive information. Con-




sequently, they weighed both pieces of information equally,exhibiting an elimination of the negativity effect.

The position-involved subjects who were motivated todefend their existing attitudes attempted to counterargue theattitude-inconsistent negative information, rating it as rela-tively weak and specious while supporting the attitude-con-sistent positive information. Consistent with defensive pro-cessing, they perceived the negative information to be lessdiagnostic than positive information and gave it less weight,exhibiting a positivity effect or a reversal of the negativityeffect.

GENERAL DISCUSSIONThe negativity effect is widely believed by media planners

and appears to be a well-proven phenomenon in consumerpsychology. However, there are some remarkable differ-ences between the marketplace conditions and settings inwhich the past negativity research has been conducted. Spe-cifically, consumers in the marketplace are likely to (a) bemore familiar with brands about which they receive infor-mation, (b) pay greater attention to messages about familiar(vs. unfamiliar) brands, and (c) process brand-related in-formation with a variety of motivations including defenseand impression concerns. The current re-inquiry attempts toexamine the robustness of the negativity effect under theseconditions.

Experiment 1 reveals that when consumers are familiarwith a brand and like it, the negativity effect is attenuated(as compared with the scenario of the unfamiliar brand).The results suggest that under these conditions, if the con-sumer pays attention to new information about the brand,s/he is likely to demonstrate only a weak negativity effect.This is because even a weak positive attitude is likely toinvoke consistency motivation, enhancing the weight givento attitude-consistent positive information.

Importantly, as the consumer’s involvement and famil-iarity with the brand increases, s/he is likely to pay moreattention to new information about it (Pechmann and Stewart1990). Whether this increased attention translates intostrengthening, eliminating, or reversing of the negativityeffect is dependent on the subject’s processing goal. Spe-cifically, when the perceiver is motivated by accuracy con-cerns (e.g., outcome involved), a negativity effect is likelyto be observed. However, this effect is eliminated with im-pression-involved individuals and reversed into a positivityeffect by defense-motivated (position-involved) consumerswho may be committed to the brand.

As such, this article suggests that although all types ofinvolvement may enhance the subjects’ message elaboration,whether this enhanced elaboration leads to a negativity effectis dependent on the nature of processing invoked. Specifi-cally, the nature of processing influences the perceived diag-nosticity of information, which in turn determines the weightgiven to it. Only when the subject’s involvement motivatescritical and risk-averse processing is negative informationlikely to be perceived as more diagnostic than positive in-formation and, therefore, weighed more. This situation oc-

curs when the subject is outcome involved. However, theother types of involvement do not invoke processing thatenhances the perceived diagnosticity of negative informa-tion. They lead to either similar perceived diagnosticitiesfor negative and positive information, implying an elimi-nation of the negativity effect (impression involvement), oran enhanced diagnosticity of positive information, implyinga reversal of the negativity effect (position involvement).Therefore, a negativity effect is likely in the marketplaceonly to the extent to which consumers are accuracy drivenand risk averse in their message processing. Past researchindicates that in naturalistic environments, this type of pro-cessing is less likely to occur than that motivated by situ-ational and personal concerns (corresponding to impressionand position involvement; e.g., Hilton and Darley 1991).

It is important to note that defense motivation elicited byposition involvement is a stronger form of the consistencymotivation induced by brand familiarity. That is, brand fa-miliarity and position involvement represent different pointson the continuum of brand attitude strength. As the attitudestrength increases, so does the extent of processing bias,weakening the impact of negative information. This canrange from an attenuation of the negativity effect (with brandfamiliarity) to its reversal into a positivity effect (with brandcommitment).

Although the reversal has been examined before (Ahlu-walia et al. 2000), this article extends our understanding ofthe defensive processing goal. More important, it examinesadditional goals (e.g., impression motivation) and proposesan integrative framework for understanding the effect ofgoals on processing of negative information.

It is also very important to note that past research hasobtained a robust negativity effect typically under conditionsof moderate to high involvement (Block and Anand-Keller1995; Maheswaran and Meyers-Levy 1990). Although thelevel of involvement was not explicitly manipulated in thecurrent experiments, it was expected to correspond to mod-erate levels in experiment 1 and high levels in experiment2. Specifically, subjects in experiment 1 were explicitlyasked to evaluate the brand but were not given a strongmotivation for engaging in extensive processing. Data fromthe attention scale indicate that, as expected, these instruc-tions led to moderate involvement in the information-pro-cessing task. Experiment 2 instructions, in contrast, werespecifically tailored to increase the subject’s processing in-volvement and, therefore, correspond to high levels of thisvariable. In other words, this re-inquiry examines limitationsto the negativity effect under conditions previously knownto be conducive for it.

In this regard, it is important to note that since my researchdid not examine the negativity effect under conditions oflow involvement, the findings presented in this article maynot generalize to such conditions. Future research is neededto fully address the likelihood of a negativity effect underlow involvement situations.

In sum, the research reported here suggests that a strongnegativity effect is only likely to emerge when consumers




are highly involved in a decision or product category butnot attached to the brand in any way or prefer it, and norare they motivated by any social concerns in processing thenew information. That is, a negativity effect may be morelimited in the marketplace than is currently believed. Thefindings of this re-inquiry call for a reassessment of thenegativity assumption in the marketing literature and an up-dating of media impact formulas that assume a robust neg-ativity effect.

[Received May 2001. Revised March 2002. David GlenMick served as editor and Kent B. Monroe served as

associate editor for this article.]

REFERENCES

Ahluwalia, Rohini, Robert E. Burnkrant, and H. Rao Unnava(2000), “Consumer Response to Negative Publicity: TheModerating Role of Commitment,” Journal of Marketing Re-search, 37 (May), 203–214.

Anderson, Norman H. (1981), Foundations of Information Inte-gration Theory, New York: Academic Press.

Block, Lauren G. and Punam Anand-Keller (1995), “When to Ac-centuate the Negative: The Effects of Perceived Efficacy andMessage Framing on Intentions to Perform a Health-RelatedBehavior,” Journal of Marketing Research, 32 (May), 192–203.

Bunker, Ted (1996), “Right on the Money,” Brandweek, 37 (8),30–32.

Chaiken, Shelly, Roger Giner-Sorolla, and Serena Chen (1996),“Beyond Accuracy: Defense and Impression Motives in Heu-ristic and Systematic Information Processing,” in The Psy-chology of Action: Linking Cognition and Motivation to Be-havior, eds. Peter M. Gollwitzer and John A. Bargh, NewYork: Guilford, 553–578.

Chen, Serena, David Shechter, and Shelly Chaiken (1996), “Gettingat the Truth or Getting Along: Accuracy- versus Impression-Motivated Heuristic and Systematic Processing,” Journal ofPersonality and Social Psychology, 71 (2), 262–275.

Ditto, Peter H., James A. Scepansky, Geoffrey D. Munro, Anne-Marie Apanovitch, and Lisa K. Lockhart (1998), “MotivatedSensitivity to Preference-Inconsistent Information,” Journalof Personality and Social Psychology, 75 (1), 53–69.

Herr, Paul M., Frank R. Kardes, and John Kim (1991), “Effectsof Word-of-Mouth and Product-Attribute Information on Per-suasion: An Accessibility-Diagnosticity Perspective,” Journalof Consumer Research, 17 (March), 454–462.

Hilton, James L. and John M. Darley (1991), “The Effects ofInteraction Goals on Person Perception,” in Advances in Ex-perimental Social Psychology, Vol. 24, ed. Mark P. Zanna,San Diego, CA: Academic Press, 235–267.

Johnson, Blair T. and Alice H. Eagly (1989), “The Effects of In-volvement on Persuasion: A Meta-analysis,” PsychologicalBulletin, 106 (2), 290–314.

Kiesler, Charles A. (1971), The Psychology of Commitment: Ex-periments Linking Behavior to Belief, San Diego, CA: Aca-demic Press.

Klar, Yechiel (1990), “Linking Structures and Sensitivity to Judg-ment-Relevant Information in Statistical and Logical Rea-

soning Tasks,” Journal of Personality and Social Psychology,59 (5), 841–858.

Kroloff, George (1988), “At Home and Abroad: Weighing In,”Public Relations Journal (October), 8.

Leippe, Michael R. and Roger A. Elkin (1987), “When MotivesClash: Issue Involvement and Response Involvement as De-terminants of Persuasion,” Journal of Personality and SocialPsychology, 52 (2), 269–278.

Lindskold, Svenn and L. Rebecca Propst (1981), “Deindividuation,Self-Awareness, and Impression Management,” in ImpressionManagement Theory and Social Psychological Research, ed.James T. Tedeschi, New York: Academic Press, 201–222.

Maheswaran, Durairaj and Shelly Chaiken (1991), “PromotingSystematic Processing in Low-Motivation Settings: Effect ofIncongruent Information on Processing and Judgment,” Jour-nal of Personality and Social Psychology, 61 (1), 13–25.

Maheswaran, Durairaj and Joan Meyers-Levy (1990), “The Influ-ence of Message Framing and Issue Involvement,” Journalof Marketing Research, 27 (August), 361–367.

Mizerski, Richard W. (1982), “An Attribution Explanation of theDisproportionate Influence of Unfavorable Information,”Journal of Consumer Research, 9 (December), 301–310.

Pechmann, Cornelia and David W. Stewart (1990), “The Effectsof Comparative Advertising on Attention, Memory, and Pur-chase Intentions,” Journal of Consumer Research, 17 (Sep-tember), 180–191.

Peterson, Robert A. and Matthew Sauber (1983), “A Mood Scalefor Survey Research,” in Proceedings of the American Mar-keting Association’s Educators’ Conference, Chicago: Amer-ican Marketing Association, 409–414.

Petty, Richard E. and John T. Cacioppo (1986), Communicationand Persuasion: Central and Peripheral Routes to AttitudeChange, New York: Verlag.

Russo, J. Edward, Margaret G. Meloy, and Victoria H. Medvec(1998), “Predecisional Distortion of Product Information,”Journal of Marketing Research, 35 (4), 438–452.

Shiv, Baba, Julie A. Edell, and John W. Payne (1997), “FactorsAffecting the Impact of Negatively and Positively Framed AdMessages,” Journal of Consumer Research, 24 (3) (Decem-ber), 285–294.

Siegel-Jacobs, Karen and J. Frank Yates (1996), “Effects of Pro-cedural and Outcome Accountability on Judgment Quality,”Organizational Behavior and Human Decision Processes, 65(1), 1–17.

Skowronski, John J. and Donal E. Carlston (1987), “Social Judg-ment and Social Memory: The Role of Cue Diagnosticity inNegativity, Positivity, and Extremity Biases,” Journal of Per-sonality and Social Psychology, 105 (January), 131–142.

Tetlock, Philip E. and Richard Boettger (1989), “Accountability:A Social Magnifier of the Dilution Effect,” Journal of Per-sonality and Social Psychology, 57 (3), 388–398.

Tykocinski, Orit, Tory E. Higgins, and Shelly Chaiken (1994),“Message Framing, Self-Discrepancies, and Yielding to Per-suasive Messages: The Motivational Significance of Psycho-logical Situations,” Personality and Social Psychology Bul-letin, 20 (February), 107–115.

Wright, Peter and Barton Weitz (1977), “Time Horizon Effects onProduct Evaluation Strategies,” Journal of Marketing Re-search, 14 (4), 429–443.



how prevalent is the negativity effect in consumer environments?

Documents