OUP user menu


Darren Dahl (Editor in Chief)Eileen FischerGita JoharVicki Morwitz

16 out of 115

When Good Brands Do Bad

Jennifer Aaker, Susan Fournier, S. Adam Brasel
DOI: http://dx.doi.org/10.1086/383419 1-16 First published online: 1 June 2004


This article reports results from a longitudinal field experiment examining the evolution of consumer-brand relationships. Development patterns differed, whereby relationships with sincere brands deepened over time in line with friendship templates, and relationships with exciting brands evinced a trajectory characteristic of short-lived flings. These patterns held only when the relationship proceeded without a transgression. Relationships with sincere brands suffered in the wake of transgressions, whereas relationships with exciting brands surprisingly showed signs of reinvigoration after such transgressions. Inferences concerning the brand's partner quality mediated the results. Findings suggest a dynamic construal of brand personality, greater attention to interrupt events, and consideration of the relationship contracts formed at the hands of different brands.

  • Brand Equity/Extensions
  • Brand Loyalty
  • E-commerce/computing/Internet
  • Satisfaction
  • Experimental Design and Analysis (ANOVA)

Because of its relevance and potential for insight generation, the relationship paradigm has enjoyed much resonance among marketing academics and practitioners. To date, however, research that examines relationships within the evolutionary context that defines them has been limited. Longitudinal field experiments have been particularly sparse, leaving unanswered many foundational questions regarding the factors that make relationships lasting and strong. Empirical investigations have also favored application domains where relationships are actively constructed by human partners, thereby especially limiting our understanding of the influences that operate in the context of consumers' relationships with brands. One factor affecting relationship strength that has received much attention concerns the transgressions that befall long-term relationships. Studied primarily within the services field, this research operates on the assumption that the response to the transgression, and not the transgression itself, is of critical importance to relationship quality and course (Hart, Heskett, and Sasser 1990). Questions thus remain as to the effects of relationship breaches independent of the recovery efforts that balance them and the conditions under which these effects may be more or less detrimental to the relationship at hand. The effects of the personalities committing transgression acts become especially interesting in this regard, due to their potential to influence relationship strength both directly and indirectly.

Attempting to address these gaps, we embarked on a longitudinal field experiment in which relationships were formed between consumers and an online photographic products and services brand. Evolving relationship strength profiles were monitored over a 2-mo. period, in response to brand personality and transgression manipulations. To shed light on underlying process mechanisms, the mediational role of character inferences concerning the quality of the brand as relationship partner was also explored. Below we draw on research from both the consumer and interpersonal relationship fields to develop our conceptual model and hypotheses.

Building Strong Relationships

Relationship strength, broadly construed in the spirit of durability and impactfulness (Petty and Krosnick 1995, p. 3), is the most frequently studied relationship characteristic in the interpersonal relationships domain (Fincham and Bradbury 1987) and the one quality most centrally linked to relationship stability both directly and indirectly (Price-Bonham and Balswick 1980; Rusbult et al. 1991). Further, relationship strength predominates in marketing, where it constitutes the top goal of managers (Gummesson 2002) and a priority for academic research (Marketing Science Institute 2002). The study of relationship strength is inherently concerned with the specification of factors that systematically influence development trajectories and, hence, the depth of resulting relationship bonds. Two such factors, the personality of the relationship partner and the commission of transgression acts, merit particular attention in light of their controllability through marketer action and the significance of their relationship effects.

Personality Effects on Relationships

Research has shown that relationships are influenced by the personalities of the partners involved (Robins, Caspi, and Moffitt 2000). Additive and joint effects have been observed for traits such as extroversion (Gifford 1991), traditionalism (Robins et al. 2000), warmth (Hill 1991), and flexibility (Shoda, Mischel, and Wright 1993). Fletcher et al. (1999) specify three trait clusters of particular note: status, warmth, and vitality. These traits underlie peoples' conceptions of ideal partners in intimate relationships and thus exert particular influence on relationship strength potential. The effects of personality on the relationship are both direct and indirect, as partner personality systematically influences the behaviors displayed in a relationship and biases the character inferences that are derived from the observation of these behaviors over time (Auhagen and Hinde 1997). Indeed, it is suggested that middle- and later-stage development is centrally concerned with character inferences regarding the relationship partner, as these shape maintenance processes (Hinde 1979; Holmes and Rempel 1989).

Partner Quality Inferences

One notable subclass of character inferences affecting relationship evolution concerns evaluations of partner capabilities and efforts in managing the relationship along implicit and explicit contract lines (Altman and Taylor 1973). In a marketing context, such inferences include whether the partner is likely to behave in such a manner that promises are kept (Iacobucci, Ostrom, and Grayson 1995), relationship failures are avoided (Smith, Bolton, and Wagner 1999), problems are resolved (Sirdeshmukh et al. 2002), and long-term consumer interests are served (Braun and Zaltman 2000). General perceptions regarding the partner's dependability and reliability (Boon and Holmes 1999), trustworthiness (Holmes and Rempel 1989), supportiveness (Moorman, Deshpande, and Zaltman 1993), and accountability (Altman and Taylor 1973) also appear significant in gauging partner capabilities, intentions, and motives in meeting relationship obligations. In line with act-frequency theories of impression formation (Buss and Craik 1983), these character inferences cohere into a generalized perception of the quality of the relationship partner through a dynamic process that considers a string of partner behaviors over time. Research has shown that partner quality inferences are used to calibrate general beliefs about the relationship (Fletcher and Kininmonth 1992), gauge the significance of and formulate responses to partner transgressions (Holmes and Rempel 1989), and ascertain overall satisfaction and loyalty levels (Sirdeshmukh et al. 2002), thereby influencing relationship strength and course.

Sincere and Exciting Brand Personalities

Two brand personality templates merit attention in light of their prominence in the marketing landscape. Further, these two personalities are fundamental in that they compose two of the three partner ideals in intimate personal relations (Fletcher et al. 1999) and capture the majority of variance in personality ratings for brands (Aaker 1997; Caprara, Barbaranelli, and Guido 2001), a finding that is robust across individuals, product categories, and cultural contexts (Aaker, Benet-Martínez, and Garolera 2001). First are “sincere” personalities that dominate the world of classic brands such as Hallmark, Ford, and Coca-Cola (Smith 2001). The sincere personality has been pursued by both smaller companies seeking to establish themselves as warmer and more caring and considerate than larger, unwelcoming rivals (e.g., Gateway Cow campaign) and by larger companies seeking a more down-to-earth face in consumer-brand interactions (e.g., MetLife's use of Snoopy). Research suggests that sincere brands will garner relationship advantages. Traits of nurturance, warmth, family-orientation, and traditionalism, which have been positively related to relationship strength (Buss 1991; Robins et al. 2000), are characteristic of sincere personalities (Aaker 1997). Sincerity can also spark inferences of partner trustworthiness and dependability (Aaker 1999), which temper feelings of vulnerability and support relationship growth (Moorman et al. 1993).

A second personality type that has received increased marketing attention is that of the “exciting” brand built around qualities of energy and youthfulness (Aaker 1997). Exciting brands, including such exemplars as YAHOO!, Virgin, and MTV, attempt differentiation through unique and irreverent advertising, atypical brand logos, and hip language. Brands have pursued exciting personalities when chasing younger demographics (e.g., Mountain Dew's “Do the Dew” campaign), repositioning for increased cultural vitality (e.g., BMW's “Driving Excitement” campaign, circa 1993), and seeking differentiation against incumbent market leaders (e.g., Dr. Pepper vs. Pepsi and Coca-Cola). Branding critics charge that, although exciting brands are attractive and attention-getting and thus highly capable of generating interest and trial, they are seen as somehow less legitimate long-term partners (Altschiller 2000). Thus, although the exciting trait is held as an ideal in intimate relations, this personality may harbor inherent disadvantages relative to the sincerity template in fostering perceptions of partner quality and encouraging long-run relationship strength.

Acts of Transgression

A second factor often singled out for its determinant effects on relationship strength is the commission of a transgression, which refers to a violation of the implicit or explicit rules guiding relationship performance and evaluation (Metts 1994). Some argue that how people cope with negative threats to the relationship has greater impact on relationship strength than positive relationship features (Rusbult et al. 1991) and that the true status of a relationship is evident only under conditions of risk and peril that activate the attachment system (Reis and Knee 1996). The significance of transgression acts derives at least in part from the high levels of salience and diagnosticity of negative events (Fiske 1980). Building on the literature above, transgressions provide opportunities for learning about the qualities of the relationship partner, which guides subsequent development paths (Altman and Taylor 1973). Accordingly, although transgressions will vary in their severity and cause and differ in their ultimate negotiations, all are significant in their ability to affect relationship progress. In this sense the transgression stands as the hallmark of the relationship, representing perhaps the most significant event in the relationship history.

The seeming inevitability of transgressions in long-term relationships contributes to their significance as well. As interdependence increases and partners interact across more domains or with increased frequency, the likelihood of a transgression augments in kind (Grayson and Ambler 1999). Interestingly, as is true with personal relationships (Reis and Knee 1996), consumers' expectations regarding brand transgressions are antithetical to this relational reality. Smith et al. (1999), for example, suggest that customers do not expect failures in their service interactions and adopt a no-transgression scenario as their operative reference point.

Research is equivocal regarding the likely effects of the transgression interrupt event. The most commonly held view is that transgressions are inherently damaging as they precipitate a string of negative inferences that threaten the relationship core (Buysse et al. 2000). In essence, the transgression reveals disconfirming evidence of the partner's intentions to act according to the terms of the relationship contract and thus exposes vulnerabilities, doubts, and uncertainties that alter and undermine partner quality perceptions (Boon and Holmes 1999). Research has shown that, once these perceptions begin to erode, it can be difficult to slow relationship decline, despite recovery efforts that may appear successful in the short run (Maxham and Netemeyer 2002).

Some researchers are less fatalistic in their views and propose contingency theories that govern the destructive influences of transgression acts. Of particular note is the relationship context in which the transgression is committed, such that relationship-serving biases dilute the negative effects of transgressions in strong unions and past positives cancel them in long-standing relations (Wiseman 1986). Partner traits have also been offered as potential moderators of transgression effects, as with competitive personalities that exacerbate negative effects and cooperative personalities that assuage them (Cupach 2000). In marketing, research has focused on the moderating effects of the recovery effort, or interactions between the recovery and characteristics of the transgression (Bitner, Booms and Tetreault 1990; Smith et al. 1999). For inspiration, this research draws on causal attribution theory, which highlights judgments of culpability and seriousness of the transgression, and distributive justice theory, which considers the fairness and equity-balance of reparations and costs (Bolton and Lemon 1999; Tax et al. 1998). Findings in this research stream demonstrate how marketer-initiated recovery attempts can dilute what is regarded as the inevitable negative fallout from failures, sometimes driving the relationship to satisfaction levels beyond pre-event marks (Smith and Bolton 1998).

Conceptual Model

The literature reviewed above provides the basis for understanding how the strength of consumer-brand relationships may be affected directly and indirectly by different brand personalities, particularly as they commit transgression acts (see fig. 1). The proposed model rests on the premise that consumers make inferences regarding a brand's character based on the observation of brand behaviors over time and that these inferences cohere into a generalized assessment of the brand in its role as a relationship partner (Blackston 1993; Fournier 1998). These partner quality inferences allow development processes to occur, thereby governing reactions to interrupt events such as transgressions. Hypotheses derived from the model are developed below, building first from anticipated effects of brand personality within the two transgression conditions to the mediating effects of partner quality predicted to govern results overall.

We hypothesize that stronger relationships will accrue for brands with sincere relative to exciting personalities. Specifically, compared to the young and trendy characteristics of the exciting brand, the sincere brand should (a) harbor inherent advantages in fostering strong relationships and (b) encourage more positive perceptions of partner quality, which, in turn, deliver strength advantages. We suggest that the sincere personality advantage will hold only in relationships that persist without the commission of a brand transgression:

  • H1: In conditions of no transgression, stronger relationships will accrue for sincere brands relative to exciting brands.

Two opposing predictions concerning the interactive effects of brand personality and transgression acts are offered in light of previously cited research. One literature stream suggests that a transgression should disproportionately harm relationships with sincere brands, where the disconfirming evidence of the transgression threatens existing partner quality perceptions. The transgression may have a different meaning and thus consequence in relationships with exciting brands, where partner quality foundations may not have been established to the same degree. Put differently, a decline in relationship strength is expected when a transgression occurs (vs. does not occur) for sincere brands, whereas such a result should not hold for exciting brands.

  • H2a: Relationship strength will be weakened for sincere brands when a transgression is present versus absent; this result will not hold for exciting brands.

A second stream of research suggests that the partner quality foundations underlying strong relationships in fact allow the negative effects of a transgression to be overcome, thus helping to maintain relationship strength levels. This implies an inherent advantage for the sincere brand committing a transgression. In other words, a decline in relationship strength should result when a transgression occurs (vs. does not occur) at the hands of the exciting brand, whereas such a result should not hold in the case of sincere brands.

  • H2b: Relationship strength will be weakened for exciting brands when a transgression is present versus absent; this result will not hold for sincere brands.

The above hypotheses are predicated on the premise that sincere and exciting personalities harbor differential abilities to garner the partner quality foundations that, in turn, affect relationship strength levels, thus governing transgression effects. Specifically, an overriding mediation effect is predicted:

  • H3: The interactive effect of brand personality and transgressions on relationship strength will be mediated by perceptions of partner quality.

The Study

A longitudinal field experiment involving a 2 (brand personality) × 2 (transgression) × 3 (time) mixed-factorial design was conducted in spring of 2000. Participants were recruited under the guise of a prelaunch beta test for a new online film processing and digitizing company named Captura Photography Services. The beta test cover story helped enhance external validity and justify the high level of interaction and monitoring the study required. The choice of the online photographic service was relevant in light of technology innovation of the time, and it allowed a setting for multiple consumer contacts required of this relational study context.

To recruit participants, 100 invitations were sent from the Captura e-mail address to students, administrators, and broader community members who had volunteered for participation in research sponsored by a West Coast business school. The invitation announced the upcoming brand launch and inquired about beta test participation. Invitees were informed that they would interact with the Captura brand via e-mail and Web site visits for a period of 2 mo., during which time they would be asked to provide ongoing opinions and reactions. For compensation, participants were paid $20, received free gifts, and were entered into a cash prize drawing.

The 69 participants who agreed to take part in the study were sent a follow-up e-mail directing them at random to one of two Captura Web sites (described below) for completion of a background questionnaire on photography habits, demographics, and online behaviors. Forty-eight participants completed this questionnaire (mean age = 21.09, range 18–50, 65% female) and followed the study to its completion. The participant sample was photography involved (90% camera ownership, average film purchase 7.72 rolls per year, high self-reported interest in photography). However, participants also indicated low levels of familiarity with eight online photographic service brands launched in CY2000 (aggregate m = 1.57, SD = 1.22, where 1 = unfamiliar, 7 = familiar). Profiles of invitees and final participants were compared to examine the possibility of response bias: mean age, gender, and category involvement did not significantly differ.

Study participants interfaced with the brand one to three times each week for a total of 12 interactions over a 2-mo. period. Participants were sent notification e-mails asking them to return to the Captura Web site for these relationship exchanges. Table 1 details the sequence of 12 interactions making up the staged relationship development process. Although not explicitly operationalized as such, the interactions were designed with general relationship development goals in mind. Some provided knowledge toward increased intimacy, for example, whereas others encouraged deepened affect toward the brand or more habitual behavioral interactions.

View this table:
Table 1

Captura Relationship Interaction Time Line

Time lineContent and goals of consumer interactions
Day 1: Interaction 1Relationship initiation (personalized e-mail). Formal beta-test study invitation mailed to participants.
Day 12: Interaction 2 (background questionnaire)Knowledge exchange (Web site visit). Participants randomly assigned to personality conditions. Participants explore content and register on Web site.

Day 15: Interaction 3Affect development (Web site visit). Participants learn of gift of free camera, to be returned within 3 weeks for film processing. Cameras express mailed to participants with prepaid return mailers.
Day 22: Interaction 4 (questionnaire 1)Behavior encouragement (Web site visit). Magazine-like article posted on “Home” page noting innovative uses for disposable cameras.
Day 25: Interaction 5Habit reinforcement (Web site visit). “Home” page article posted containing top 10 tips for improved picture-taking and better photographic expression.
Day 34: Interaction 6Contract salience (personalized e-mail). Reminder to return camera for NetAlbum creation. Participants sent a “Your camera has been received” confirmation e-mail upon receipt of camera.
Day 42: Interaction 7Affect and behavior reinforcement (Web site visit). Participants invited to visit Web site to view their newly created online photo albums (NetAlbums).
Day 45: Interaction 8Transgression manipulation (personalized e-mail). Random selection of participants notified that NetAlbum was accidentally deleted; half reminded that NetAlbum is ready to be viewed.
Day 47: Interaction 9 (questionnaire 2)Emotion trigger (Web site visit). Contest announced in which photograph judged by experts to be of the highest quality would earn cash prize of $250, with three runners-up receiving $50 each.
Day 48: Interaction 10Apology and recovery (personalized e-mail and Web site visit). Participants in transgression condition receive apology and notification of NetAlbum restoration, tailored according to brand personality. All participants notified that contest is underway. Participants return to site where entry form is posted.
Day 54: Interaction 11, questionnaire 3Relationship decline (Web site visit). Participants informed of upcoming study termination. Participants return to site, photo contest winners announced.a
Day 58: Interaction 12Relationship termination (personalized mailing). Participants mailed their processed photos, compensated, and debriefed.
  • a Cash prizes were in fact awarded randomly to four participants at the end of the study, one in each condition. Analyses at time 3 were rerun without contest winners in the sample. Results remained the same.

Three augmentations to basic study procedures merit highlighting. First, on interactions 4, 9, and 11 (days 22, 47, and 54, respectively), participants not only visited the Web site for new content but also were asked to complete questionnaires including partner quality and relationship strength measures, as well as other brand diagnostics. These are referred to as time 1, time 2, and time 3 data throughout the article. Second, after interaction 6, participants were asked to mail their disposable cameras (which were earlier provided as a gift) to Captura via prepaid FedEx. Cameras were processed and digitized at a local studio. Each participant's photos were then organized into secured, personalized, online “NetAlbums” designed in a style consistent with the personality condition (see fig. 2), which participants were invited to view in a subsequent interaction. Third, the transgression manipulation occurred on day 45 at interaction 8; the apology and recovery occurred on day 48 at interaction 10.

Figure 2

Examples of Sincere versus Exciting Brand Web Pages

Independent Variables

Brand Personality

Two Web sites, each with a distinct URL to prevent contamination across conditions, were created: one conveyed a sincere Captura brand personality, and the other an exciting personality. Participants were randomly assigned to brand personality conditions. To enhance external validity, professionals were retained for brand personality execution, including graphic design of the Web site and logo and copy writing for the text of all consumer-brand interactions. Personality was manipulated through four venues: (1) overall tonality, as conveyed through vocabulary choice and phrasing (e.g., “Hello” for the sincere brand vs. “Hey!” for the exciting brand); (2) brand identity elements consistent with intended personalities, as based on pretest results (i.e., sitting St. Bernard dog vs. jumping Dalmatian puppy logos); (3) Web site visuals, including colors (soft browns, oranges, yellows vs. bright reds, greens, purples) and font (Comic Sans vs. Jester); and (4) content, as contained in Web site postings (e.g., family picnic vs. rock-climbing photographic references), page links (Disney, Kodak, and Life magazine vs. MTV, Polaroid, and Spin magazine), and tag lines (“Because Life Is Too Meaningful to Let You Pass It By” vs. “Because Life is Too Exciting to Let You Pass It By!).

Captura Web site content was organized into five sections, each accessible throughout the study via navigation buttons on the left-hand sidebar frame. “Home,” the first page accessed when entering the site, contained the content of the particular interaction at hand. “About Us” described the services provided by Captura. The “Stories” section contained photographs from (fictional) Captura users, with descriptions of where, when, and why the pictures had been taken. “Links” offered connections to other photography and lifestyle Web sites that were pretested to have personality associations consistent with the sincere or exciting Captura brand. “Contact Us” launched the e-mail address that participants could use to send comments to Captura: a preformatted response acknowledged comment submissions. Figure 2 provides illustrative content in the “About Us” section for the two personality conditions.

A pretest (n = 32) ensured that the two personality conditions did not differ in personal relevance (e.g., brand image relevant to me, makes sense to me; seven-point scale, r = .78; Msincerity = 5.03; Mexcitement = 4.83, F < 1) or category relevance (e.g., relevant in photography services category, makes sense in photography category; r = .87, Msincerity = 4.40; Mexcitement = 5.00, F < 1). Further, when participants rated the Captura name on sincere and exciting traits, no significant differences were found (Msincerity = 3.60; Mexcitement = 4.03, F < 1).


A pretest was conducted to identify a transgression that was (1) sufficiently under Captura's control as to be recognized as a preventable relationship violation and (2) severe enough to be noticed but not so severe as to be debilitating. The accidental erasure of digital prints by a Captura employee was selected for the manipulation versus three other pretested scenarios. Participants were randomly assigned to the two transgression conditions at interaction 8, where half received a personalized e-mail reminding them to view their online photo albums if they had not done so already (transgression absent) and half received notice that their online photo albums had accidentally been erased (transgression present). Tonality announcing the transgression was consistent with each brand personality manipulation: “We are very sorry, but we are unable to locate your NetAlbum. … Some of our employees have yet to master the system. … We deeply apologize for this unfortunate occurrence and hope to make amends” (Sincere) versus “Sorry, but we can't find your NetAlbum! … Some of our employees are still a little green. … Sorry about that, hope we can make it up to you!” (Exciting). Relationship strength and partner quality were measured 2 days following the transgression to allow diagnosis of transgression effects.

At interaction 10, 3 days after the service failure, the apology-recovery component of the transgression event sequence occurred. Recovery efforts were crafted to fit the nature of the transgression, such that the magnitude of recovery was similar to that of the transgression (Bolton and Lemon 1999). Specifically, those in the transgression condition received notice that their NetAlbums had been restored, as announced through e-mail using personality-consistent language: “Hello. We are contacting you with some news that we are sure you will welcome. We are happy to report that your on-line photos have been restored. … We apologize for the worry we put you through. Our greatest hope is that you continue to remain a part of the Captura Family, and that you accept our sincere regret at the inconvenience we caused” (Sincere); “Hey! We have some very cool news that we think you'll be excited to hear! We've restored your pics! … Sorry about the mix-up again! We hope you keep on being part of the Captura Experience; and again, sorry about the problem we caused!” (Exciting). Relationship strength and partner quality were measured 6 days following the recovery to allow full diagnosis of transgression dynamics.

Dependent and Mediating Variables

Four relationship strength indicators, each capturing a noteworthy conceptualization of the construct in the interpersonal or marketing literatures, made up the dependent variable set. Each construct was measured via multiple items at three points in time (time 1 on day 22 of the relationship, time 2 on day 47, 2 days after the transgression, and time 3 on day 54, post–recovery attempt). The four indicators—commitment, intimacy, satisfaction, and self-connection—were treated independently in the analyses to allow exploration of different relationship effects.

Commitment, defined as an enduring desire to continue the relationship combined with a willingness to make efforts toward that end (Morgan and Hunt 1994), was operationalized via items that tapped the instrumental investments underlying commitment, the time horizon implicit in the construct, and more general behavioral indicators of loyalty (Gundlach, Achrol, and Mentzer 1995). Intimacy was defined as a deep understanding about the relationship partners as created through information disclosure (Altman and Taylor 1973) in light of research demonstrating that the reduction of uncertainty accounts for the greatest percentage of variance in friendship closeness ratings (Hays 1985). Intimacy items assessed the perceived depth of consumer understanding exhibited by the brand, consumers' understanding of the brand, and consumers' willingness to share informational details toward the goal of more intimate relational ties. The Satisfaction measure included items indicating appraisals of satisfaction with and happiness in the relationship (Lewis and Spanier 1979), as well as comparisons of relationship performance versus expectations (Oliver 1997). Self-connection indicated strength through activation of the person's identity system (Aron et al. 2000) and contained items capturing the degree to which the relationship delivered on centrally held identity themes (Fournier 1998), or helped express real and collective selves (Belk 1988).

The mediating variable, Partner Quality, concerned character inferences regarding the brand's performance in its partner role. Items tapped selected aspects of trust and trustworthiness (Braun and Zaltman 2000; Moorman et al. 1993; Sheppard and Sherman 1998) and included behavioral indicators of benevolence (i.e., whether brand acts with consumer interests in mind), problem-solving prowess (i.e., brand's responsiveness to problems), and perceived reliability and dependability. Two customized brand character items relating specifically to transgression commission (Smith et al. 1999) were also developed in light of the behavioral orientation of the manipulation. Table 2 lists the specific items used to measure these constructs of interest.

View this table:
Table 2

Construct Indicators, Measurement Items, and Scale Reliabilities

Time 2
Time 3
Relationship strength indicators:
  CommitmentI am very loyal to Captura.
I am willing to make small sacrifices in order to keep using Captura.
I would be willing to postpone my purchase if the Captura site was temporarily unavailable.
I would stick with Captura even if it let me down once or twice.
I am so happy with Captura that I no longer feel the need to watch out for other photography alternatives.
I am likely to be using Captura one year from now.
  IntimacyI would feel comfortable sharing detailed personal info about myself with Captura.
Captura really understands my needs in the photographic services category.
I'd feel comfortable describing Captura to someone who was not familiar with it.
I am familiar with the range of products and services Captura offers.
I have become very knowledgeable about Captura.



  SatisfactionI am completely satisfied with Captura.
I am completely pleased with Captura.
Captura is turning out better than I expected.
  Self-ConnectionThe Captura brand connects with the part of me that really makes me tick.
The Captura brand fits well with my current stage of life.
The Captura brand says a lot about the kind of person I would like to be.
Using Captura lets me be a part of a shared community of like-minded consumers.
The Captura brand makes a statement about what is important to me in life.
  Partner qualityI can always count on Captura to do what's best.
If Captura makes a mistake, it will try its best to make up for it.
I know I can hold Captura accountable for its actions.
Captura is reliable.
Given my image of Captura, letting me down would surprise me.
A brand failure would be inconsistent with my expectations.

Note.—All Likert items are anchored by 1 = strongly disagree, 7 = strongly agree. In line with the multifaceted conceptualization that the alternate measures were intended to indicate, intercorrelations among the four relationship strength variables were high, and they varied across the study horizon. Time 3 correlations ranged from .73 (Satisfaction and Commitment) to .98 (Self-Connection and Commitment); time 2 correlations from .63 (Satisfaction and Commitment) to .92 for Self-Connection and Commitment; and time 1 correlations from .58 (Satisfaction and Intimacy) to .86 (Self-Connection and Commitment).


    All analyses were run with the full set of 48 participants as well as with the subset that completed all three questionnaires (n = 40). Although the latter analyses yielded slightly weaker statistical effects on a few isolated measures, a nearly identical pattern of results was obtained for the two groups. Demographic profiles of the two subject pools were also similar, and no evidence for response bias was found. Results using the more conservative smaller sample are reported.

    Manipulation Checks

    To assess the effectiveness of the brand personality manipulation, participants rated the degree to which Captura could be described by sincerity traits (sincere, wholesome, sentimental, family-oriented; α = .87) as well as exciting traits (exciting, unique, young, trendy; α = .90; 1 = strongly disagree, 7 = strongly agree) (Aaker 1997). To minimize demand effects, these checks were assessed only at time 3. A 2 (brand personality) × 2 (transgression) ANOVA run on the Sincerity index yielded only a main effect of brand personality, where higher ratings on the Sincerity index were found in the sincere condition versus the exciting condition (M = 5.74 vs. M = 3.66; F(1,36) = 12.81, p < .01). A 2 × 2 ANOVA on the Exciting index also yielded a brand personality main effect, where higher ratings on the Exciting index were found for the exciting versus sincere condition (M = 5.50 vs. M = 4.30; F(1,36) = 3.65, p < .05). No other significant effects were found in either analysis.

    To assess the transgression manipulation, participants at time 3 were asked the degree to which they agreed with the following: “Captura makes mistakes,” “There are times when Captura lets me down,” and “Captura can let me down” (averaged to create a three-item Transgression index; α = .84). A 2 × 2 ANOVA run on the Transgression index yielded only a transgression main effect wherein those in the transgression present condition had higher ratings on the index than those in the transgression absent condition (M = 5.87 vs. 4.83; F(1,36) = 3.97, p < .05).


    Four analytic routes shed light on the relationship issues of interest in this study. First, trend analyses were conducted to reveal general relationship development patterns within the four experimental conditions. Second, a set of 2 (brand personality) × 2 (transgression) × 3 (time) repeated-measures ANCOVAs was run on each of the strength indicators to test hypotheses using methods appropriate to longitudinal studies of relationship development in the interpersonal literature (Hays 1985). Third, for insight into the theoretical relationships suggested in our model, mediation analyses were undertaken. Finally, ancillary analyses were conducted to provide diagnostic information regarding the more provocative findings.

    Trend Analyses

    Trend analyses revealed the developmental trajectories for the two brand personalities in the transgression present and absent conditions, and, in particular, helped determine whether transgression effects shifted the development patterns for the two brands. Dummy variables for Time2 and Time3 were created whereby the coefficient of the Time2 dummy denoted the change in the dependent variable from time 1 to time 2, and the Time3 dummy reflected changes from time 1 to time 3. The difference between these two coefficients indicated time 2 to time 3 strength changes (with p values obtained through a separate regression using Time3 as the reference dummy). The four dependent measures were regressed against these dummies as well as age and gender: demographics yielded no consistent effects.

    We first consider development trajectories for the two brands in relationships without a transgression. Relationships with the sincere brand displayed a trend of increased strengthening over time across all four indicators (p's < .05). In contrast, relationships with the exciting brand progressively weakened from time 1 levels, as per commitment, satisfaction, and self-connection indicators (p's < .05). The trajectory for the sincere brand thus evokes the human relationship analogy of a close partnership or friendship (Price and Arnould 1999), whereas the exciting brand trajectory parallels more closely the trend of a short-lived fling (Aron et al. 2000).

    A very different and interesting story unfolded in relationships struck by transgressions. At the hand of sincere brands, transgressions precipitated significant weakening of the relationship across all indicators (p's < .05).Of importance, the sincere brand relationship showed no signs of recovery despite reparative attempts (p's > .10).Surprisingly, progressive relationship deterioration did not result when the exciting brand committed a transgression (p's > .10).Whereas the natural trajectory for the exciting brand relationship was one of progressive decline, this decline was abated at time 2 with the transgression event, and it sustained through the recovery. Figure 3 illustrates typical development trajectories across the four experimental conditions using the self-connection indicator.

    Figure 3

    Self-Connection Trajectories across Four Experimental Conditions


    To shed more light on the trend analysis results and to test the hypotheses put forth in the study, a set of 2 (personality) × 2 (transgression) × 3 (time) ANCOVAs was conducted on each of the relationship strength indicators. Brand personality and transgression served as between-subjects variables, time as a repeated measure, and age and gender as covariates. Each analysis revealed a significant three-way interaction (p's < .05), which was driven by a significant two-way brand personality × transgression interaction (p's < .05; with one exception: satisfaction yielded only a significant two-way interaction). Simple effects that examined the nature of each interaction at single points in time were run; two-tailed tests were used. Means are provided in table 3; table 4 presents the simple effects tests.

    View this table:
    Table 3

    Relationship Strength and Partner Quality as a Function of Brand Personality and Transgression Manipulations (Means amd Standard Deviations)

    Sincere brand personalityExciting brand personality
    Strength measuresAbsence of transgression
    (n = 10)
    Presence of transgression
    (n = 10)
    Absence of transgression
    (n = 10)
    Presence of transgression
    (n = 10)
      Time 13.21 (.70)3.01(1.17)3.26(1.99)3.18 (.96)
      Time 24.62(2.43)1.5  (.83)2.16(1.36)3.17(1.90)
      Time 34.35(1.79)2.22(1.19)1.58 (.62)3.00(2.12)
      Time 14.07 (.85)4.27(1.31)3.82(2.17)4.41(1.11)
      Time 25.41(1.37)2.86 (.62)3.53(1.60)4.72(2.04)
      Time 35.44(1.38)2.67 (.98)2.83(1.28)4.49(2.44)
      Time 15.38 (.89)4.96 (.87)5.61(1.88)5.42(1.08)
      Time 26.15 (.93)3.32(1.94)4.40(1.99)4.37(2.32)
      Time 35.69(1.38)3.31(1.96)3.58(1.58)4.36(2.56)
      Time 13.46 (.91)3.28(1.09)3.52(2.04)3.68(1.19)
      Time 24.65(1.71)1.83 (.71)2.36(1.34)3.24(1.85)
      Time 34.58(1.67)1.95 (.60)2.15(1.12)3.93(2.65)
    Partner quality:
      Time 13.76 (.58)3.94 (.87)3.70(2.08)4.66(1.25)
      Time 25.68(1.63)2.35(1.16)3.90(1.46)4.43(2.59)
      Time 35.32(1.72)3.00(1.14)3.40(1.35)5.12(2.40)

    Note.—The transgression preceded time 2 measurement; the apology and recovery preceded time 3 measurement. Higher means indicate greater agreement with the measure and are associated with stronger relationships. The means used in our significance tests relied on repeated-measures of a general linear model (GLM) and include controls for serial correlation, age, and gender.

      View this table:
      Table 4

      Multiple Analysis of Variance Using a General Linear Model (with Repeated Measures) Procedure: Simple Effects Analysis

      Strength measuresSincere brand
      (presence vs. absence
      of transgression)
      Exciting brand
      (presence vs. absence
      of transgression)
      Absence of transgression
      (exciting vs. sincere
      brand personality)
      Presence of transgression
      (exciting vs. sincere
      brand personality)
        T2, post-transgression−3.16**1.09−2.87**1.62
        T3, post-recovery−2.15*1.96*−3.70** .60
        T2, post-transgression−3.36**1.23−2.42*2.42*
        T3, post-recovery−3.18**2.12*−3.34**2.03*
        T2, post-transgression−2.96**−.26−1.90+1.01
        T3, post-recovery−2.38*1.03−2.22**1.36
        T2, post-transgression−3.78**1.08−3.27**1.88+
        T3, post-recovery−2.69**2.21*−2.90**2.19*

      Note.—Values in the table are the t-statistics for tests of significance of the effects between conditions. Degrees of freedom are 2, 33.

      For insight into hypothesis 1, which suggests inherent strength advantages for sincere versus exciting brands, the pattern of effects at times 1, 2, and 3 within the no transgression condition was examined. Sincere and exciting brand relationships progressed similarly in the early stages of the engagement: no significant differences in strength patterns between the two brand personalities were found at time 1, day 22 of the relationship (p's > .10 for all indicators). This null effect is also consistent with the premise that sincere and exciting brands were equally valued in the current context, as highlighted earlier regarding the study design. Relationship strength was considerably greater for the sincere versus exciting brand at both time 2 and time 3, across all strength indicators (p's < .05).Thus, in line with hypothesis 1, stronger relationships eventually accrued for brands associated with sincere relative to exciting traits. As discussed below, this strength advantage held only in relationships without transgressions.

      To explore the viability of hypothesis 2a versus 2b, we next examined the differential impact of the transgression on relationships with the two personality types. At time 2, after the transgression act, the relationship with the sincere brand exhibited lower levels of intimacy and self-connection relative to the exciting brand (p's < .05).This result is consistent with hypothesis 2a, and it casts doubt on hypothesis 2b. The same sincere brand disadvantage persisted after recovery efforts were engaged (Time3 p's < .05). That is, even after the apology and recovery, participants felt a stronger bond with the exciting versus the sincere brand. This pattern suggests that the meaning of the transgression and the ensuing recovery efforts differed fundamentally for the two personality types, with the information being destructive and disconfirming in the case of the sincere brand and somehow constructive in the case of the exciting brand.

      Support for hypothesis 2a over hypothesis 2b was also found by comparing relationship strength patterns for the sincere brand when a transgression did versus did not occur. Contrasts showed that strength was significantly greater for the sincere brand when no transgression took place relative to when it did take place, an effect observed across all strength indicators (Time2 p's < .01). Again, these differences remained after recovery attempts (Time3 p's < .05). Such consistent and persistent effects support the contention that transgressions have a particularly damaging effect on relationships with sincere brands, where the simple occurrence of the transgression harms strength on all dimensions and creates difficulties that remedial efforts did not assuage.

      Finally, we examined the effect of the transgression on the exciting brand. These findings also do not support hypothesis 2b. No differences in strength were found at time 2 for the exciting brand in the transgression absent versus present condition (p's > .10). In other words, the transgression did not produce negative effects beyond those reflected in the naturally deteriorating trajectory of the exciting brand. It is interesting that there were significant differences at time 3, where an increase in strength was observed for the exciting brand when a transgression had taken place relative to when it had not. That is, consumers interacting with the exciting brand were more committed, indicated stronger feelings of intimacy, and reported higher levels of self-connection (p's < .05) when the brand made a mistake and pursued subsequent recovery attempts.

      Collectively, these results suggest that the transgression and subsequent recovery helped to somehow invigorate the relationship with the exciting brand, perhaps by injecting new meanings and salience into the relationship and thereby reversing the natural decline that would otherwise accrue. This is consistent with the premise that a transgression may act as an inflection point in the relationship: in this case, one that allowed the exciting brand an opportunity to (re)-activate the attachment system underlying the relationship with the brand (Reis and Knee 1996), thus prompting consumer reconsideration of levels of connection, intimacy, and commitment. The transgression had a different meaning and hence consequence for the sincere brand, where injected meanings involved disconfirming evidence of brand capabilities and intentions to act according to contract terms—a fundamental breach that harmed the relationship at its core.

      Inside the Invigoration Effect

      Post hoc analyses using ancillary relationship diagnostics were conducted to explore the viability of the (re)invigoration proposition for the exciting transgressing brand and, more generally, decipher what exactly was happening to the exciting brand relationship in the context of the transgression. First, to shed light on whether the transgression in fact made the exciting relationship more salient and engaging, a set of 2 (personality) × 2 (transgression) ANOVAs were run on time 2 measures of relationship dependency (Blankfield 1987) and future intention. Follow-up contrasts conducted in light of a significant 2 × 2 interaction (p's < .05) revealed that relationships with exciting brands in the transgression present versus absent condition were rated higher on two of three dependency indicators: “I feel like something is missing when I haven't used Captura in awhile” (Mpresent = 2.59, Mabsent = 0.90, p < .05) and “I am addicted to Captura in some ways” (Mpresent = 2.46, Mabsent = 0.98, p < .05). Heightened relationship intentions were also expressed at the time of the transgression: “I'd like to see my relationship with Captura get stronger” (Mpresent = 3.37, Mabsent = 1.70, p < .05) and “I will very likely be using Captura one year from now” (Mpresent = 3.62, Mabsent = 1.56, p < .05). Such differences in dependency and future intention were not found in contrasts between sincere transgression present versus absent conditions. Thus, it appears that the transgression act itself served to make salient consumers' otherwise unacknowledged relationships with the exciting brand. An analogous effect is perhaps observed in the context of power outages, whereby consumers' true feelings about their appliances and electricity service providers are revealed only through the denial of the consumption experience, which brings below-awareness attachments to the fore.

      Perceptions of relationship interdependence (Hinde 1979) at time 3 provided further insight into saliency changes in the exciting brand relationship precipitated by the transgression. A series of 2 (personality) × 2 (transgression) ANOVAs were run on eight semantic differential items describing relationship interdependence taken only at time 3, post-recovery. Follow-up contrasts conducted in light of two-way interactions (p's < .05) showed that relationships with exciting brands in the transgression present versus absent condition were more likely to be rated as: permanent (Mpresent = 4.89, Mabsent = 1.80; 1 = fleeting and 7 = permanent, p < .05); habitual (Mpresent = 4.44, Mabsent = 2.16; 1= sporadic and 7 = habitual, p < .05); and characterized by frequent interactions (Mpresent) = 3.89, Mabsent = 1.97; 1 = occasional and 7 = frequent interactions, p < .05). Again, such differences were not found in contrasts between transgression present versus absent conditions for sincere brands. These findings also support the contention that the transgression/recovery sequence made the exciting brand relationship more salient and engaging and increased interdependence between the consumer and brand.

      A final analysis of partner quality inferences shed further light on the reinvigoration proposition by revealing whether new meanings specific to the partner were precipitated by the transgression and recovery acts. A 2 (personality) × 2 (transgression) × 3 (time) mixed-factorial ANOVA run on the partner quality index yielded significant three-way and two-way personality × transgression interactions (p's < .05). Follow-up contrasts at time 3 showed that the exciting brand was rated higher in terms of partner quality after the transgression and recovery took place (Mpresent = 5.12, Mabsent = 3.40, p < .05). It is interesting that these partner quality foundations were not developed for the exciting brand in the no transgression scenario, where levels of partner quality remained flat over time (MTime1 = 3.70, MTime2 = 3.90, MTime3 = 3.40, p's for all contrasts > .10). This pattern of results suggests that the transgression and ensuing recovery helped establish positive perceptions of Captura's trustworthiness and accountability that were otherwise not available. A very different result held for sincere brands, where the transgression eroded established partner quality foundations (Mabsent = 5.32, Mpresent = 3.00, p < .05). These findings support the contention that the meaning of the transgression differed for the two brand personalities, providing disconfirming evidence of established partner capabilities in one regard and constructive evidence allowing their development in the other.

      Mediation Analyses

      Hypothesis 3 suggests that the interactive effect of the two brand personality types and the act of a transgression leads to different levels of partner quality, which, in turn, affects relationship strength. To test this premise, four sets of regressions (a–d below) were conducted whereby brand personality, presence of transgression, and brand personality × transgression were the independent variables in a mediated moderation analysis (Baron and Kenny 1986). The four relationship strength indicators served as dependent variables. Results indicated that (a) when each of the four strength indicators were regressed on the main and interactive effects of brand personality and transgression, the interactive effect was significant (p's < .01 for all four relationship indicators). Further, (b) when partner quality was regressed on the main and interactive effects of brand personality and transgression, the interactive effect was significant (p < .01). In addition, (c) each of the four strength indicators was associated with higher levels of partner quality (p's < .0001). Importantly, (d) when the strength indicators were regressed on partner quality, brand personality, transgression, and the brand personality × transgression interaction, partner quality remained a significant predictor of relationship strength (p's < .0001), but the interaction effect became insignificant (p's > .15). The only exception to this pattern involved intimacy, where the interaction decreased to marginal significance (p = .08) when partner quality was taken into account. These results support a mediating role of partner quality judgments in consumers' reactions to transgressions, consistent with hypotheses 3.

      A final analysis examined whether the effects above would remain robust with prior judgments taken into account (Smith and Bolton 1998). The four sets of regressions outlined above were rerun including two dummy variables for temporal effects (Time2 vs. Time1; Time3 vs. Time2) involving each of the core dependent measures. Two findings resulted. First, in each of the regressions described in d, partner quality remained significant (p's < .0001), while the brand personality × transgression interactions did not (p's > .10 for all strength indicators). Thus, the results above did not change when time trends were taken into account. Second, the dummy variables were insignificant in all analyses, except those involving satisfaction (p's < .05). Thus, although prior satisfaction levels influenced subsequently felt satisfaction, brand personalities and transgression acts affected relationship strength profiles above and beyond the updating of satisfaction judgments over time.

      General Discussion

      Results from the 2-mo. longitudinal field experiment support hypotheses regarding the interactive effects of brand personality and transgression acts on the evolving strength of consumer-brand relationships and the mediating role of partner quality inferences in governing these dynamics. Specifically, brands characterized by sincere personality traits encouraged progressively stronger relationships analogous to close friendships in the interpersonal space, but this was the case only when the relationship proceeded without a transgression. Transgressions were particularly damaging to relationships with sincere brands, which showed no signs of recovery despite subsequent reparation attempts. Result patterns suggest that the transgression provided contrary evidence that disconfirmed expectations of the partner, as per the deterioration of partner quality perceptions (e.g., “This brand is not as concerned about me as I thought”) and violation of assumed intimacy levels with the brand (e.g., “This is not the brand I thought it was”). Of importance, the transgression also appeared to damage the fundamental meanings on which this seemingly close partnership was based, per weakening of bases of self-connection and diminishing satisfaction and commitment overall. Development patterns were markedly different for exciting brands, which displayed more of a spike-decay pattern characteristic of flinglike engagements when transgressions did not take place and, surprisingly, strength improvements in the wake of transgressions. At the hands of exciting brands, the transgression-recovery sequence appeared to (1) provide useful knowledge about the brand and relationship (per intimacy and partner quality development), (2) reactivate attachment systems (per self-connection gains), and (3) increase interdependency levels and hence commitment in the relationship. In this sense, transgressions operated at least in part as a means of (re)invigorating the exciting brand relationship, a function not engaged in relationships with sincere brands.

      Contribution and New Insights

      These findings are consistent with consumer research suggesting that transgressions can serve as defining moments that distance the relationship in some instances but propel it forward in others (Fournier and Deighton 1999) and extend this work through the specification of conditions that govern alternative reactions. Interaction results also shed light on conflicting findings in the services literature, where higher satisfaction is found with failure-recovery in some instances (Bitner et al. 1990; Smith and Bolton 1998), though dampened levels are observed in others (Bolton and Drew 1991). Findings also resonate with discussions of the dark side of long-term marketing relationships (Grayson and Ambler 1999) and highlight the risks involved in the invariant pursuit of deep consumer relationships grounded in foundations of trust.

      These results also extend theories regarding the role of expectations and prior experiences in consumer judgments. Findings corroborate the view that objective evidence, such as that revealed by a transgression, may be interpreted differently depending on prior experiences and relationships (Hoch and Deighton 1989). Although it is generally assumed that consumers assign greater weight to evidence consistent with experiences, these results suggest that consumers in strong relationships with sincere brands did not discount the transgression, which was likely inconsistent with relationship expectations. This finding encourages a rethinking of conventional wisdom regarding the safety cushion of tolerance that is assumed through strong, trusting bonds. Moreover, our research sheds light on the mechanisms guiding expectation setting, an issue on which extant theories have been silent. The current research illuminates this issue by specifying brand personality as one potential source of relationship expectations, and it supports a process explanation that focuses on an important subclass of expectations concerning partner quality as influential to relationship progress over time.

      It is also useful to reflect on additional process explanations that may underlie observed effects, particularly in light of the surprising result concerning the beneficial role of exciting brand transgressions. One potential mechanism concerns the differential application of uncertainty reduction processes typically assumed constant across relationship types (Boon and Holmes 1999). It is possible that ambiguity regarding appropriate actions on the part of the brand partner was higher in the case of exciting versus sincere brands and that these perceptions served to diminish negative repercussions of the transgression act. That is, by their very nature, exciting brands might have encouraged consumers to “expect the unexpected” through their more flexible and lively spirit, thereby reducing feelings of vulnerability and risk and diminishing the severity of apparent violations. This explanation, though compelling, does not explain the stronger relationships found at time 3 for the exciting brand in the transgression condition. Nonetheless, there remains a need for a more complete understanding of the traits embedded in various personality templates and for research that explores the relationship biases inherent therein.

      Another promising explanatory direction hinges on the distinct trust processes potentially encouraged by the different brand personalities, raising the possibility that either the transgression event was a violation of only one of these trust forms, or the recovery assuaged only one type of infringement. Rousseau et al. (1999) posit two dominant trust forms: calculative trust, based on the weighing of specific gains and losses for exchange relationships, and emotional trust, based on identification and attachment for communal relationships. The conjecture here is that the relationships encouraged by the exciting personality may have been more exchange-oriented in spirit and thereby characterized by calculative trust. Violations of calculative trust would have been satisfied by reparative actions designed to balance losses with reciprocated gains, as was the case with our manipulation. Sincere brands, on the other hand, may have encouraged communal relations with emotional trust forms that were debased through what consumers considered careless brand actions. Here, recovery efforts designed simply to recoup losses would be perceived as ineffectual in repairing the relationship breach at hand. Accordingly, in designing what was purportedly a neutral recovery event, we may have inadvertently crafted a reparation in line only with calculative trust violations. Although this theoretical direction does not fully explain results concerning reinvigoration signals, it does nonetheless point to value in a contingency explanation specifying the types of relationships in which strength accrues from fairness and equity judgments relative to socioemotional rewards.

      Caveats and Calls for Future Research

      This research was inspired in part by calls for more longitudinal field-based experiments in consumer research, particularly those involving relationship phenomena (Bitner et al. 1990; Mick 1999; Morgan and Hunt 1994). Our aim was to create a study setting that would blend some of the advantages of an experimental approach, which involves a controlled context in which to manipulate constructs of interest, with some of the advantages of a naturalistic approach, which include the capture of a broad base of behaviors in a real-world setting and induction of illuminating patterns from this data. It was the marriage of these two philosophical approaches that produced a relatively novel study design pursued here and allowed insight into the ways in which relationships grew and were diminished at the hands of different brands.

      This method, however, is not without its limitations, which include a restrictive sample size, noise stemming from the study environment, and limits to generalizability associated with the exploration of two personality types within a single product category. Further, although our beta test cover story allowed ecological validity advantages, relationships were nonetheless artificially construed and came to an end at the 2-mo. mark. Indeed, based on the magnitude of means obtained on selected relationship strength measures, some may debate the degree to which relationships were created at all. The constrained timespan of the study raises questions regarding the sustainability of observed effects as well. A study design that more explicitly recognizes the staged, complex, and cumulative process of relationship evolution across an extended time horizon is suggested, particularly in light of the diagnostic value of self-connection and intimacy measures. Internal validity can also be strengthened through attempts to disentangle the communication of the transgression from the transgression act itself and by direct consideration of transgression versus recovery components of the service failure.

      On more substantive levels, our research emphasizes the determinant role of brand personality in establishing consumer relationship bonds, both in terms of the direct effects of different personality templates and the partner quality inferences that each entails. Evidence of actionable links between brand personality and consumer behaviors supports rejuvenation of what has been an underemphasized domain in consumer research (Holt, forthcoming). Focused work is needed to understand what exactly is manipulated through certain personality templates and whether it is more appropriate to consider “trait constellations” when characterizing the personalities of brands. Of importance, the present work supports a dynamic view of personality that extends beyond trait snapshots to consider the actions committed by the personality and the various character inferences that these spark. It may be that the enlivened brand personality is best conceptualized in terms of relationship roles rather than ascribed traits, which opens up new avenues for conceptualization. Questions regarding when and how consumer personalities influence the relationship also merit attention, including research into links between brand and consumer personalities and identification of relationship-relevant styles that may affect consumers' interactions with brands.

      The present research also suggests empirical inquiry dedicated explicitly to the transgression event itself. Foundational work is needed to clarify our conceptualization of transgressions, beyond and including the miscarried deeds of brand partners (Vinokur and Van Ryn 1993). Research delineating the factors that enhance or dilute transgression significance is warranted as well. To date, transgression severity has rarely been examined outside a causal attribution frame, nor has the content of transgression breaches been explicitly considered. A broadened reconceptualization of transgressions in terms of the class of “interrupt events” that they represent may in fact prove relevant and promising, particularly in light of the information-provision functions highlighted in the current research.

      Finally, any research on transgressions or inquiries regarding role-based construals of the personality of the brand must be conducted with sensitivity toward the types of relationships at hand. Two distinct relationship classes are implicated through the longitudinal patterns revealed here: close, increasingly intimate, long-term-oriented friendships and the initially enthused, but subsequently declining, flings. Research is needed to ascertain the degree to which these different relationships are indeed nurtured by distinct brand personalities, and more broadly, to identify the various relationship types operating in the consumer-brand world. Such research would ideally specify the contract terms that govern each relationship type including, for example, relationship goals, behavioral norms, and rules for satisfaction assessment (Argyle 1986; Fitzsimmons and Bargh 2003). An empirical exploration such as this would further serve to sharpen the conceptualization of transgression events themselves, and provide a framework for understanding transgression severity. Research on relationship contracts would also allow researchers to move beyond expectation levels to explicitly recognize the content of expectations and the relationship rules that create them (Fournier and Mick 1999), particularly as these vary along temporal and cultural lines.


      View Abstract