Product Samsung

The current issue and full text archive of this journal is available at www. emeraldinsight. com/0309-0566. htm EJM 44,7/8 Consumer responses to brand extensions: a comprehensive model ? ? Eva Mart? nez and Jose M. Pina ? Facultad de Ciencias Economicas y Empresariales, The University of Zaragoza, Zaragoza, Spain Abstract Purpose – This paper aims to understand the reciprocal spill-over effects of brand extensions by testing a comprehensive model that gathers both the brand extension evaluation process and the later in? uence on brand image. Design/methodology/approach – Data were obtained from 699 face-to-face interviews conducted in Spain.Structural equation modelling was used to test the proposed hypotheses. Findings – The results indicate that brand extensions have feedback effects on brand image depending on the attitude toward the new product and perceived image ? t. Consumer attitude depends, in turn, on initial brand associations, perceived category ? t, perceived image ? t and consumer innovativeness. Brand familiarity also shows indirect effects. Research limitations/implications – The model should be tested with extensions of the same (line extensions) or different categories.It is also necessary to analyse non-? ctitious products, and to take different moderating effects into account. Practical implications – The results suggest how to protect the brand image from unsuitable extension strategies. The paper shows what kind of perceived ? t is more important for consumers as well as the direct and indirect role of several variables. Originality/value – The paper extends previous research by proposing a complete framework that considers the factors that in? uence either the attitude to the extension or the attitude to the extended brand.Samsung Distribution ChannelKeywords Brand extensions, Brand image, Brand equity, Consumer behaviour, Spain Paper type Research paper 1182 Received January 2008 Revised October 2008 January 2009 Accepted February 2009 Introduction Brand extension is a strategy that many companies follow with the aim of bene? ting from the brand knowledge achieved in the current markets (Aaker and Keller, 1990; Milberg et al. , 1997). When a new product is marketed under a well-known brand name, failure rates and marketing costs are reduced (Milewicz and Herbig, 1994; Keller, 2003). Keller (2003) states that more than 80 per cent of ? ms resort to brand extensions as a way of marketing goods and services. The support that the brand gives to the new product often leads to a change in the brand image associations. Both the affection and the speci? c knowledge associated with the brand and the new product are interchanged in the consumers’ mind (Czellar, 2003). European Journal of Marketing Vol. 44 No. 7/8, 2010 pp. 1182-1205 q Emerald Group Publishing Limited 0309-0566 DOI 10. 1108/03090561011047580 The authors would like to thank the following sources for their ? nanc ial help: CICYT (Ref: ?SEJ2005-02315) and Government of Aragon (â€Å"GENERES†, Ref. S-09; â€Å"PM0262/2006†). They also gratefully acknowledge the constructive comments of the three anonymous EJM reviewers. This feedback process can increase the memory and strength of brand associations (Morrin, 1999; Aaker, 2002) and, thus, improve the positioning of the brand (Park et al. , 1986). Nevertheless, several authors indicate that the dilution of current beliefs is more likely (Tauber, 1988; Ries and Trout, 1993; John et al. , 1998). This dilution effect can take place even though the extension is not related to negative information (Morrin, 1999; ?Ahluwalia and Gurhan-Canli, 2000; Mart? nez and Pina, 2003). Virgin, for instance, is a company that has grown through extensions into the audiovisual sector, retailing, alcoholic drinks, passenger transport (by railway and air) and space tourism, among others. However, market research studies suggest that customers’ per ceptions of the Virgin brand mainly depend on the performance of the airline, which implies a constant threat of image dilution (Hughes, 2007). The in? uence of brand extension on brand image is explained by several theories, most of them coming from Psychology.According to the â€Å"associative network theory†, brand image may be understood as a mental scheme formed by a network of concepts (nodes) interconnected by linkages or associations (Anderson, 1983; Morrin, 1999). Park et al. (1993) explain that extensions which are coherent with the brand schema will not lead to image dilution (assimilation process). On the other hand, the brand schema will be modi? ed to accommodate examples that are far from current brand attitudes and beliefs (accommodation process). Following Weber and Crocker’s (1983) ? ork, Gurhan-Canli and Maheswaran (1998) suggest that the image modi? cation could be re? ected in the formation of a mental subcategory inside the brand scheme (sub-typin g model) or in a complete modi? cation of brand associations (conversion model). The sub-typing or conversion processes may occur when perceived ? t or typicality between the extension category and the brand is low. However, it is just possible that brand attitudes and beliefs would always change because of the new information, which is called the bookkeeping model (Weber and Crocker, 1983; Loken and John, ? 993; Gurhan-Canli and Maheswaran, 1998). Consumers could react according to the bookkeeping model when the information on the new product is highly accessible. Regardless of perceived ? t, higher accessibility gives rise to an image enhancement, whereas lower accessibility has a negative effect on brand evaluations (Ahluwalia and ? Gurhan-Canli, 2000). The brand extension literature shows that brand extensions can affect both the ? general brand associations (Mart? nez and de Chernatony, 2004) and the beliefs in speci? attributes (Keller and Aaker, 1992; Loken and John, 1993). T he beliefs related to the most representative product of the brand, or ? agship product, are more resistant to dilution ( John et al. , 1998; Chang, 2002), as well as the perceptions linked to the brand personality (Diamantopoulos et al. , 2005). Most previous research on brand extensions develops experimental designs, focusing on a reduced number of variables (e. g. Loken and John, 1993; John et al. , 1998; Alexander and Colgate, 2005). Some authors have tested models through structural ? equation modelling (e. g.Bhat and Reddy, 2001; Volckner and Sattler, 2006) although they concentrate on consumer attitude toward brand extensions and not on reciprocal spillover effects. According to literature, brand extensions may give rise to both a â€Å"forward† effect from the parent brand to the new product and a â€Å"feedback† or â€Å"backward† effect from the new product to the parent brand (Milberg et al. , 1997; Responses to brand extensions 1183 EJM 44,7/8 1184 B alachander and Ghose, 2003). Neglecting this potential backward effect affords a limited view of consumer behaviour and may lead to inappropriate marketing actions.With the goal of better understanding the way that extensions in? uence brand image, our work proposes and validates a theoretical model that, according to the previous literature, integrates the most relevant variables. With the exception of the ? contribution of Volckner and Sattler (2006), previous models only focus on a few variables, which makes it dif? cult to determine how the consumers’ responses to brand extensions are generated. Furthermore, the proposed model considers both the brand image before the extension and the image variation, which is a step forward in literature.As well as brand image, we will analyse the effects of brand familiarity, attitude to the extension, extension-brand ? t (category and image ? t), perceived dif? culty in manufacturing the extension product and consumer innovativeness. Hence, the study expands previous research by testing a comprehensive model that gathers both the brand extension evaluation process and the later in? uence on brand image. This model can help brand managers to protect their brands from unsuitable brand extensions by showing the main determinants of spillover effects and the direct and indirect effects of the speci? variables. Relationships that have been individually supported in previous works could be rejected when considering complex models with several dependent and independent variables. The study is structured in four sections. The next section contains a brief review of the literature to justify the theoretical model and the relations established in the hypotheses. The third section describes the methodology used to validate the model, and the results are reported in the fourth section. Finally, we address the conclusions and managerial recommendations.Proposed model and hypotheses The proposed model helps us to understand t he in? uence of brand extensions on brand image. For this reason, the model includes the variables with the greatest impact on extension attitude (Aaker and Keller, 1990; Hem et al. , 2003). This attitude will determine the development of the brand image (Lane and Jacobson, 1997), affecting the current associations. The model stems from the initial brand image and attempts to identify the main relations and interactions that follow the launching of the brand extension and its potential effects on the established associations.Generally, consumer attitudes toward brand extensions can depend on factors related to brand associations, extended category, perceived ? t, and consumer characteristics (Czellar, 2003; Reast, 2005; ? Volckner and Sattler, 2006). Hence, two brand knowledge factors, brand familiarity and initial brand image, are considered. In relation to the new product and its ? t with the parent brand, we consider perceived dif? culty in manufacturing, perceived category ? t a nd perceived brand image ? t. Extension attitude and consumer innovativeness are also taken into consideration. Whereas brand associations and ? have been examined in nearly every study on brand extensions, perceived dif? culty and consumer innovativeness have received lesser attention. Since Aaker and Keller’s (1990) fundamental study and all subsequent replications (Barrett et al. , 1999) analysed perceived dif? culty with inconclusive results, it seems necessary to study this variable more in depth. On the other hand, the whole literature on brand extensions relies on the assumption that a known brand reduces the risk associated with buying new products (Smith and Park, 1992), and consumer innovativeness re? ects the consumer’s risk aversion.The proposed effects of these variables and the remaining ones are depicted in Figure 1. The ? rst variable included in our model is brand familiarity. This variable is closely related to the dimension of brand equity labelled a s awareness by Aaker (1996), since familiar brand names usually present high awareness. Moreover, it is also akin to the brand image construct, which refers to the different â€Å"perceptions about a brand re? ected as associations existing in the memory of the consumer† (Keller, 1993). Direct effects on extension attitude are expected for brand familiarity as well as indirect ones through brand image.First, individuals will have a better initial image of the brands they are familiar with (Low and Lamb, 2000; Lemmink et al. , 2003). By means of a â€Å"halo effect†, the impressions of familiar attributes are used to form precise opinions on brands (Reynolds, 1965) and develop more complete knowledge structures (Alba and Hutchinson, 1987; Grime et al. , 2002). Furthermore, familiarity indirectly re? ects the experience with a brand (Alba and Hutchinson, 1987), presenting a clear relationship between experience and brand image (Hoek et al. , 2000).Familiarity can also ha ve a direct effect on brand extension evaluations. Consumers are more inclined to buy products of brands they have previously consumed (Swaminathan, 2003) and know better, unless the experience has been unsatisfactory (Swaminathan et al. , 2001). Although some works have failed to prove that familiarity affects consumer attitude to an extension (Glynn and Brodie, 1998) and to the extended brand (Diamantopoulus et al. , 2005), we hypothesise: H1. The greater the familiarity of the core brand, the more positive the initial brand image. H2.The greater the familiarity of the core brand, the more favourable the attitude to the extension. Brand image is an essential factor for understanding consumer attitude toward brand extensions, since the credibility of the new product increases when brand perceptions become more favourable (de Ruyter and Wetzels, 2000). If the brand image consists of Responses to brand extensions 1185 Figure 1. Proposed model to analyse the effect of brand extension strategy on brand image EJM 44,7/8 1186 associations such as a high-perceived quality, the extension attitude will be better (van ? Riel et al. 2001; Volckner and Sattler, 2006). In the same vein, the extension attitude is positively related to the perceptions of reputation (Hem et al. , 2003), prestige (Park et al. , 1991) and the consumers’ affection for the brand (Sheinin and Schmitt, 1994). In the case of corporate and service brands, a positive image also clearly generates favourable perceptions of the new products (Brown and Dacin, 1997; de Ruyter and Wetzels, 2000). Given that the extension leverages the current brand associations, the better the initial brand image the more positive will be the consumers’ response.Therefore: H3. The more positive the initial brand image, the more favourable the attitude to the extension. If consumers perceive a high ? t between the brand and the new product, the brand leveraging increases and the potential negative effects are less likely (Czellar, 2003). Some authors state that consumers can consider a category ? t or an image ? t (Bhat and Reddy, 2001; Grime et al. , 2002; Czellar, 2003). Thus, individuals can believe that the new product is physically similar to the other products of the brand (category ? t) or coherent with the general brand associations (image ? ) (Grime et al. , 2002; Czellar, 2003). Whatever the case, the consistency between cognitive elements and the similarity among various stimuli ease and improve consumers’ evaluations (Aaker and Keller, 1990; Eagly and Chaiken, 1993). Brand image-perceived ? t interaction effects are revealed in the literature (Boush et al. , 1987; Aaker and Keller, 1990) as well as ? direct effects (Volckner and Sattler, 2006). The next hypotheses deal with the direct effects of perceived ? t dimensions on extension evaluation. As commented above, perceived category and image ? will directly affect the consumer attitude to the extension. Generally, the assessment of an extension will be more positive as perceived closeness with the brand grows (Aaker and Keller, 1990; ? Volckner and Sattler, 2006), even in the case of non-prestige brands (Park et al. ,1991). However, consumers believe that extensions to non-related categories are not very reliable and offer low quality, which causes a negative assessment (Kirmani et al. , 1999). According to the literature, a high-perceived category or image ? t makes success more likely (Boush et al. 1987; Boush and Loken, 1991; Park et al. , 1991). The important thing is to get the consumers to relate the new product to the brand, independently of the kind of closeness. This discussion leads to the following hypotheses: H4. The greater the perceived category ? t between the extension and the core brand, the more favourable the attitude to the extension. H5. The greater the perceived image ? t between the extension and the core brand, the more favourable the attitude to the extension. Another va riable included in our model is perceived dif? ulty in manufacturing or offering a new good or service. This variable has been analysed in numerous works, although it is not clear whether it in? uences consumer behaviour or not (Barrett et al. , 1999; van Riel et al. , 2001). Moreover, present research does not clarify whether this in? uence is positive (Aaker and Keller, 1990; van Riel and Ouwersloot, 2005) or negative (Semeijn et al. , 2004). This diversity of results re? ects that the in? uence of dif? culty in manufacturing might depend on the study settings and the variables interacting with such dif? culty.Generally, consumers who think that the new product category requires little manufacturing effort may question its advisability (Aaker and Keller, 1990). They could even think that high-quality brands are trying to make fast money by overpricing trivial products (Aaker and Keller, 1990; van Riel et al. , 2001). In a sense, easy-to-make extensions could resemble downscale ext ensions, where the brand stretches down by offering lower price-quality products (Kirmani et al. , 1999). Consequently, we posit: H6. The greater the perceived dif? culty in manufacturing the new product, the more favourable the attitude to the extension.The last variable of our model to explain attitude to the extension is consumer innovativeness, a concept that represents the consumers’ propensity to buy new products and consider new ideas (Roehrich, 2004). Since innovative people are more risk-prone (Klink and Smith, 2001; Hem et al. , 2003), they show a better attitude toward brand extensions, whatever their perceived ? t (Klink and Smith, 2001). In this sense, some authors have found that higher consumer innovativeness increases perceived quality and purchase intention of new services (Hem et al. , 2003; Siu et al. , 2004) and ? tangible products (Volckner and Sattler, 2006).Rogers (1983) claims that one of the most salient traits of consumer innovators is the comfort th ey gain from taking risk. Unlike later adopters, highly-innovative individuals ? nd far extensions appealing (Xie, 2008) and, consequently, do not mind trying products that get away from the company’s core business. As a matter of fact, they should be more prone to try new products regardless of the degree of brand knowledge or perceived ? t. Consequently, we posit: H7. The greater consumer innovativeness, the more favourable the attitude to the extension. The following hypotheses relate to the feedback effect on brand image.Because of the new information, the brand schema could vary its structure of nodes and links (Morrin, 1999). There is no doubt that most brand associations will remain stable after stretching to new categories, being the ? nal perceptions mainly determined by the ? initial ones (Lee and Ulgado, 1993; Mart? nez and Pina, 2003). However, product introductions in the marketplace involve providing consumers with information, which not always ? ts with the ini tial beliefs and feelings about the brand. As elucidated by previous research, the attitude to the extension is a major driver of spillover effects from the extension to the parent brand.Low quality or negatively ? assessed extensions will entail a detriment of brand image (Chang, 2002; Mart? nez and ? Pina, 2003), diluting both general and speci? c beliefs (Mart? nez and de Chernatony, 2004). Diamantopoulos et al. (2005) found that brand personality is more dilution-resistant, although any brand association is exposed to the risk of dilution. A way of reducing this risk is to strengthen the attitude to the extension, given that consumers who are satis? ed with the extension are usually satis? ed with the brand (Alexander and Colgate, 2005). The following hypothesis is based on these arguments. H8.The better the attitude to the extension, the more favourable the feedback effect on the extended brand. Responses to brand extensions 1187 EJM 44,7/8 1188 The literature reveals that the attitude to an extended brand directly depends on the degree of ? t with the extension (Grime et al. , 2002). The introduction of extensions far from the core business will involve losing brand differentiation and credibility, whereas extensions to related markets will avoid potential damage (Aaker, 2002). Some authors like Milberg et al. (1997) have proved that low-? t extensions generate negative feedback in terms of attributes or image.Similarly, Lee and Ulgado (1993) ? veri? ed that ? t has a positive effect on the image of service ? rms, whereas Mart? nez and de Chernatony (2004) veri? ed the same for tangible product extensions. Other works equally suggest that the impact of brand extensions on the parent brand is ? directly related to similarity (Mart? nez and Pina, 2003) or image ? t (Loken and John, 1993; John et al. , 1998). All in all, we expect a more positive feedback effect provided the brand stretches coherently with either its image or current products. H9. The great er the perceived category ? between the extension and the core brand, the more favourable the feedback effect on the extended brand. H10. The greater the perceived image ? t between the extension and the core brand, the more favourable the feedback effect on the extended brand. Methodology An empirical study was conducted to contrast the hypotheses and validate the model displayed in Figure 1. Following the usual procedures, we utilised real brands and realistic hypothetical extensions (Aaker and Keller, 1990; van Riel et al. , 2001; van Riel and Ouwersloot, 2005) that were previously selected through three pre-tests.Below, we explain these and other aspects related to the methodology applied. Pre-tests In line with previous research, a sample of undergraduates was employed in the pre-tests (Sheinin and Schmitt, 1994; Kim, 2003). The speci? c brands and extensions were selected by means of Wilcoxon tests, which were necessary due to the lack of normality in the data. The aim of the ? rst pre-test, conducted with 91 students, was to choose brands in three sectors (fast moving consumer goods, durable consumer goods and services) that were familiar (F) to individuals and had a different image perception (I).Familiarity is an essential requisite to guarantee that consumers have a clear image to evaluate (Low and Lamb, 2000). Two questions were thus formulated to assess those concepts in seven-point Likert scales (1 ? Totally unfamiliar/7 ? Very familiar; 1 ? Bad image/7 ? Excellent image) for a total of 11 brands. According to the results, Colgate ? and Signal (FC ? 6. 38; FS ? 5. 50), Nike and Puma (FN ? 6. 56; FP ? 5. 64), Telefonica Movistar and Amena (FT ? 6. 64; FA ? 6. 27) were chosen as familiar brands. The image is signi? cantly different in toothpaste brands (IC ? 5. 74; IS ? 4. 96; Z ? 2 4. 618; p , 0. 0001), sports brands (IN ? 6. 21; IP ? 5. 10; Z ? 2 5. 449; p , 0. 00001) and mobile phones (IT ? 5. 67; IA ? 4. 88; Z ? 2 4. 001; p , 0. 00001). The seco nd and third pre-tests, where 98 and 81 students, respectively, participated, were aimed at ? nding two extensions –one for each sector– with differences in perceived ? t. Both perceived category ? t (CF) and brand image ? t (IF) were considered (Bhat and Reddy, 2001) in two Likert scales (1 ? Not at all similar/7 ? Very similar; 1 ? Non-coherent/7 ? Very coherent). For the toothpaste brands, â€Å"sugar-free whitening tooth decay-preventing sweets† and â€Å"sunglasses† were selected.The ? rst showed a higher perceived ? t than the second for Colgate (CF1 ? 5. 36; CF2 ? 1. 31; Z ? 2 5. 341; p , 0. 00001) (IF1 ? 5. 69; IF2 ? 1. 54; Z ? 2 5. 339; p , 0. 00001) and Signal (CF1 ? 4. 86; CF2 ? 1. 19; Z ? 2 5. 120; p , 0. 00001) (IF1 ? 5. 19; IF2 ? 1. 25; Z ? 2 5. 019; p , 0. 00001). On the other hand, for the sports brands, we chose â€Å"skis† as a close extension and â€Å"DVD players† as a far extension, both from the perspective of product c ategory of Nike (CF1 ? 3. 33; CF2 ? 1. 28; Z ? 2 5. 120; p , 0. 00001) and Puma (CF1 ? 3. 32; CF2 ? 1. 14; Z ? 2 4. 910; p , 0. 00001).Similarly, there were statistical differences between the image ? t of the extensions for Nike (IF1 ? 4. 23; IF2 ? 1. 36; Z ? 2 5. 561; p , 0. 00001) and Puma (IF1 ? 3. 89; IF2 ? 1. 14; Z ? 2 5. 113; p , 0. 00001). Finally, â€Å"telecommunication on-line courses† and â€Å"insurance† were the service extensions selected. Speci? cally, the perceived category ? and image ? t were statistically different for Telefonica Movistar (CF1 ? 4. 67; CF2 ? 1. 84; Z ? 2 5. 475; p , 0. 00001) (IF1 ? 4. 72; IF2 ? 1. 72; Z ? 2 5. 543; p , 0. 00001) and Amena (CF1 ? 3. 73; CF2 ? 1. 76; Z ? 2 4. 283; p , 0. 00001) (IF1 ? 4. 27; IF2 ? 1. 84; Z ? 2 4. 61; p , 0. 00001). Sample and procedure Subsequent to the pre-tests, we elaborated 12 questionnaires with a different brand-extension combination. On the ? rst page, individuals had to indicate their consume r innovativeness and answer some questions about the corresponding brand (familiarity and image) and product category (perceived dif? culty). Then, on the second page of the questionnaire, respondents were required to imagine that the speci? c brand launched the extension. Questions then assessed the ? t, the respondents’ attitudes towards the extension and the brand image, supposing the existence of the new product category.No additional information about the products’ attributes was provided in order to avoid bias that could defeat the objective of the study (Bhat and Reddy, 2001). The surveys were answered by a total sample of 720 individuals (699 valid cases) in a Spanish city, which is sometimes considered as a test market for products aimed at Spain. The respondents were approached by a team of interviewers in different parts of the city, on different days and at different times during May 2005. By following a quota sampling procedure, the sample was required to match the population structure by sex (50. 9 per cent women and 49. per cent men) and age (46. 5 per cent 26-45 years, 33. 3 per cent 16-25 years, 20. 2 per cent 46-64 years). These demographical variables may be strong predictors of changes in attitudes and behavior (Hansman and Schutjens, 1993) and, therefore, should be controlled to get adequate variance in the data. Table I shows the type of questionnaires used in our research and the speci? c number of individuals who satisfactorily responded to each. No individual answered more than one questionnaire. Measures Variables were measured through seven-point Likert scales by requesting individuals either to state their level of agreement with the speci? statement (1 ? Totally disagree, 7 ? Totally agree) or directly assess the variable (e. g. 1 ? Not at all familiar, 7 ? Very familiar). In all cases, items were extracted or based on the literature. In order to avoid potential order effects (Klink and Smith, 2001), perceived Respons es to brand extensions 1189 EJM 44,7/8 N8 Brand 49 Colgate Extension (high ? t) Sugar-free whitening tooth decaypreventing sweets Sugar-free whitening tooth decay†¦ Skis Skis Telecommunication online courses Telecommunication online courses N8 Brand 50 Colgate 48 49 49 80 Signal Nike Puma ? Telefonica Movistar 75 AmenaExtension (low ? t) Sunglasses Sunglasses DVD players DVD players Insurance Insurance 1190 Table I. Type and number of questionnaires Signal Nike Puma ? Telefonica Movistar 79 Amena 49 48 49 74 dif? culty was assessed prior to brand characteristics and ? t. For the same reason, ? nal image was measured once the individuals had formed an opinion about the brand extension. Table II shows the scales used for each factor. First, consumer innovativeness was measured with the items proposed by Roehrich (1994), who considers a dual perspective, â€Å"hedonistic† and â€Å"social†. Perceived dif? ulty was assessed through an item used by Aaker and Keller (19 90) and two additional items coherent with the concept. For brand familiarity, we used Dawar’s scale (Dawar, 1996), whereas the scale validated by Martinez et al. (2004) was employed to assess initial and ? nal brand image. This scale utilises items from several works (Martin and Brown, 1990; Weiss et al. , 1999) which attempt to assess tangible (functional image) and intangible (affective image) attributes and bene? ts, as well as the global attitude to the brand (reputation). The distinction made by several authors between category ? t or similarity and image ? or consistency with brand image (Park et al. , 1991; Bhat and Reddy, 2001; Grime et al. , 2002) was used to measure perceived ? t. Thus, a series of items that assess ? t from both perspectives (Aaker and Keller, 1990; Taylor and Bearden, 2002) were chosen. Finally, extension attitude items were suggested by authors like Aaker and Keller (1990) or Pryor and Brodie (1998) considering both the general assessment of the new product and purchase intentions. Results The collected data were analysed by means of structural equations methodology, assessing both the measurement and the structural model (Kline, 2005).The structural model allows us to know whether there is evidence to reject the proposed hypotheses, although previously the measurement model has to evaluate the psychometric properties of the scales in terms of unidimensionality, reliability and validity. Furthermore, some ? t indicators show whether the measurement and structural models explain the collected data with relative precision (Hair et al. , 1998). Scale validation Prior to analysing all the variables as a whole, we studied whether initial brand image, ? nal brand image, consumer innovativeness and perceived ? should be considered as multidimensional or unidimensional factors, since the distinction between the Scale Consumer innovativeness. Roehrich (1994) Measured concept Hedonist innovativeness (HINN) HINN1: I am more intereste d in buying new than known products HINN2: I like to buy new and different products HINN3: New products excite me Social innovativeness (SINN) SINN1: I am usually among the ? rst to try new products SINN2: I try new products before my friends and neighbours SINN3: I know more than others about the latest new products DIFF1: Dif? ulty in designing and making the product DIFF2: Complex techniques or knowledge are needed DIFF3: Specialised resources are needed (personnel, facilities. . . ) FAMI1: Familiarity with the brand’s products FAMI2: Purchase frequency of the brand’s products FAMI3: Knowledge of the brand’s products Functional image (FUIM) (initial/? nal) FUIM1i/FUIM1f: The products have a high quality FUIM2i/FUIM2f: The products have better characteristics than competitors’ FUIM3i/FUIM3f: The products of the competitors are usually cheaper Affective image (AFIM) (initial/? al) AFIM1i/AFIM1f: The brand is nice AFIM2i/AFIM2f: The brand has a personalit y that distinguishes it from competitors AFIM3i/AFIM3f: It is a brand that does not disappoint its customers Reputation (REIM) (initial/? nal) REIM1i/REIM1f: It is one of the best brands in the sector REIM2i/REIM2f: The brand is very consolidated in the market Category ? t (CAFI) CAFI1: The extension is similar to the brand’s products CAFI2: The ? rm’s resources are helpful to make the product extension Image ? t (IMFI) IMFI1: The product extension ? s with the brand image IMFI2: Launching the extension is logical for the company IMFI3: Launching the extension is appropriate for the company EXAT1: Favourable attitude towards the extension EXAT2: Perceived quality of the extension EXAT3: Likelihood of trying the extension Responses to brand extensions 1191 Perceived dif? culty (DIFF). Aaker and Keller (1990) Brand familiarity (FAMI). Dawar (1996) ? Brand image. Mart? nez et al. (2004). Based on: Martin and Brown (1990) Aaker (1996); Weiss et al. (1999); Villarejo (2002) Perceived ? t. Aaker and Keller (1990); Taylor and Bearden (2002) Extension attitude (EXAT).Aaker and Keller (1990); Pryor and Brodie (1998) Table II. Scales used in the questionnaires EJM 44,7/8 1192 proposed dimensions (e. g. hedonistic and social innovativeness) could be statistically non-advisable. Through a previous analysis with SPSS 13. 0, we detected a weak item-total correlation of FUIM3i (corr. ? 0. 281) and FUIM3f (corr. ? 0. 296) with the respective dimensions of functional image. After eliminating them, we conducted an explanatory factor analysis for the unidimensional and multidimensional models using the EQS 5. b and ERLS (elliptical re-weighted least squares) estimation method.The initial image, ? ?nal image and perceived ? t scales proved to be reliable in both models (Joreskog and ? Sorbom, 1993), although it was advisable to eliminate HINN1 related to consumer innovativeness. Although the factor loadings exceeded the cut point lU ? 0:540; lM ? 0:673? ; the R 2 co ef? cients ? R 2 ? 0:292; R 2 ? 0:453? were below those recommended in the literature (Hair et al. , 1998). Once the scales had been properly re? ned, we proceeded to compare the unidimensional and multidimensional models through several indicators (Hair et al. , 1998; Kline, 2005). Tables III and IV display the coef? ients obtained, which clearly favour the consideration of independent dimensions for all the factors analysed. The only indexes in which the unidimensional model surpasses the multidimensional one are PNFI and PGFI for the factors of initial brand image (PNFI ? 0. 511 , 0. 638; Comparative indicators Initial image Unidimen. Multidimen. 126. 181 0. 047 0. 221 112. 181 0. 160 0. 638a 0. 466a 154. 181 72. 177a 0. 034a 0. 152a 61. 177a 0. 088a 0. 511 0. 377 106. 177a Final image Unidimen. Multidimen. 211. 559 0. 053 0. 343 197. 559 0. 283 0. 628a 0. 449a 239. 559 51. 082a 0. 027a 0. 122a 40. 082a 0. 057a 0. 516 0. 382 85. 082a x2RMSR (Root mean square residual) ECVI (Expec ted cross-validation index) NCP (Noncentrality parameter) SNCP (Scaled noncentrality parameter) PNFI (Parsimonious normed ? t index) PGFI (Parsimonious goodness of ? t index) AIC (Akaike information criterion) Table III. Indicators of the alternative models of brand image (initial and ? nal) Note: aCoef? cients that are favourable to the speci? ed model Comparative indicators Consumer innov. Unidimen. Multidimen. 195. 411 0. 079 0. 309 190. 411 0. 272 0. 453a 0. 292a 215. 411 31. 088a 0. 022a 0. 076a 27. 088a 0. 039a 0. 394 0. 261 53. 088a Perceived ? t Unidimen. Multidimen. 77. 634 0. 34 0. 140 72. 634 0. 104 0. 483a 0. 314a 97. 634 50. 164a 0. 025a 0. 103a 46. 164a 0. 066a 0. 391 0. 256 72. 164a x2 RMSR (Root mean square residual) ECVI (Expected cross-validation index) NCP (Noncentrality parameter) SNCP (Scaled noncentrality parameter) PNFI (Parsimonious normed ? t index) PGFI (Parsimonious goodness of ? t index) AIC (Akaike information criterion) Table IV. Indicators of the alter native models of consumer innovativeness and ? t Note: aCoef? cients that are favourable to the speci? ed model PGFI ? 0. 377 , 0. 466), ? nal image (PNFI ? 0. 516 , 0. 628; PGFI ? 0. 382 , 0. 449), consumer innovativeness (PNFI ? . 394 , 0. 453; PGFI ? 0. 261 , 0. 292) and perceived ? t (PNFI ? 0. 391 , 0. 483; PGFI ? 0. 256 , 0. 314). Nevertheless, the parsimony indicator, AIC, which allows us to choose between models with a different number of latent variables, as in our case, presents better values in the multidimensional structure: initial image (AIC ? 106. 177 , 154. 181), ? nal image (AIC ? 85. 082 , 239. 559), consumer innovativeness (AIC ? 53. 088 , 215. 411) and perceived ? t (AIC ? 72. 164 , 97. 634). After verifying the multidimensional character of initial brand image, ? nal brand image, consumer innovativeness and perceived ? , our next step was to conduct a factor analysis of all the scales. Again, we used EQS and ERLS, obtaining the results shown in Table V. We can i nfer from these results that the scales present good statistical properties. As can be seen in Table V, all the proposed items unidimensionally ? t the respective 13 factors or latent variables. The values obtained in composite reliability coef? cients and extracted variance analysis (EVA) are above 0. 6 and 0. 5, respectively, which guarantees the internal consistency of the scales. Moreover, the validity criterion was satis? ed from both convergent and discriminant viewpoints.Thus, all lambda coef? cients for the observed variables are signi? cant (t . 1. 96) and they load on the corresponding factors with standard loadings above 0. 5. The con? dence intervals of between-factor correlations were calculated to analyse discriminant validity. No intervals included value 1, which indicates the differentiated character of the factors. The main goodness-of-? t indicators for the measurement model are shown at the bottom of Table V, distinguishing between global and incremental ? t index es. On the whole, the indicators are positive and above the minimum established by researchers (Hair et al. 1998; Kline, 2005). With regard to global ? t, GFI is above 0. 8 (GFI ? 0. 884), whereas RMSEA and SRMR error statistics were below the maximum values of 0. 06 (RMSEA ? 0. 053) and 0. 08 (SRMR ? 0. 040) recommended by Hu and Bentler (1999). The only unsuitable indicator is the Chi-square test (x 2(417) ? 1224. 142; p , 0. 001), which often occurs in samples of over 400 observations. On the other hand, all the incremental ? t measures were above the required 0. 8 (AGFI ? 0. 844) and 0. 9 (CFI ? 0. 973; IFI ? 0. 973; NFI ? 0. 960; NNFI ? 0. 966) levels, which proves the statistical convenience of the proposed model.The validation process concluded with the estimation of three second-order models for the dimensions of brand image (initial and ? nal) and consumer innovativeness. These models presented favourable ? t indicators for initial image (GFI ? 0. 958; SRMR ? 0. 035; NFI ? 0. 975; IFI ? 0. 979), ? nal image (GFI ? 0. 972; SRMR ? 0. 028; NFI ? 0. 985; IFI ? 0. 989) and consumer innovativeness (GFI ? 0. 978; SRMR ? 0. 022; NFI ? 0. 985; IFI ? 0. 987). Model and hypotheses contrasting After analysing the psychometric properties of the scales, we proceeded to the estimation of the structural model, which corresponds to the structure shown in Figure 1.Previously, the global effect of extensions on brand image was analysed, comparing the values of initial and ? nal image in each scenario. Responses to brand extensions 1193 EJM 44,7/8 Factor HINN SINN Items HINN2 HINN3 SINN1 SINN2 SINN3 FUIM1i FUIM2i FUim1f FUIM2f AFIM1i AFIM2i AFIM3i AFIM1f AFIM2f AFIM3f REIM1i REIM2i REIM1f REIM2f FAMI1 FAMI2 FAMI3 DIFF1 DIFF2 DIFF3 EXAT1 EXAT2 EXAT3 CAFI1 CAFI2 IMFI1 IMFI2 IMFI3 Reliability t (. 1. 96) l(. 0. 5) 22. 230 20. 993 26. 547 25. 862 19. 829 22. 534 20. 543 24. 779 24. 208 21. 076 19. 473 17. 864 21. 545 21. 680 17. 880 23. 342 18. 125 25. 834 19. 868 22. 112 19 . 930 20. 822 18. 05 24. 402 18. 291 22. 956 18. 606 21. 579 22. 312 18. 837 26. 733 26. 683 24. 607 0. 861 0. 820 0. 915 0. 899 0. 744 0. 835 0. 776 0. 873 0. 859 0. 787 0. 741 0. 693 0. 798 0. 802 0. 694 0. 871 0. 706 0. 919 0. 751 0. 838 0. 771 0. 799 0. 729 0. 926 0. 725 0. 831 0. 712 0. 795 0. 839 0. 730 0. 906 0. 905 0. 859 Convergent validity * CRC (. 0. 6) EVA (. 0. 5) 0. 828 0. 891 0. 787 0. 857 0. 785 0. 810 0. 770 0. 825 0. 845 0. 839 0. 824 0. 763 0. 920 0. 707 0. 733 0. 650 0. 750 0. 550 0. 587 0. 629 0. 704 0. 645 0. 638 0. 610 0. 618 0. 793 1194 FUIM (i) FUIM (f) AFIM (i) AFIM (f) REIM (i) REIM (f) FAMI DIFF EXAT CAFI IMFITable V. Reliability, convergent validity and ? t of the measurement model Notes: Fit indices: Global ? t: x 2 ? 1224. 142 (417) p , 0. 001; GFI ? 0. 884; RMSEA ? 0. 053; SRMR ? 0. 040. Incremental ? t: AGFI ? 0. 844; CFI ? 0. 973; IFI ? 0. 973; NFI ? 0. 960; NNFI ? 0. 966; CRC: Composite reliability coef? cient; EVA: Extracted variance analysis, GFI : Goodness of ? t index; RMSEA: Root mean square error of approximation; SRMR: Standardised root mean square residual; AGFI: Adjusted goodness of ? t index; CFI: Comparative ? t index; IFI: Incremental ? t index; NFI: Normed ? t index; NNFI: Non-normed ? t indexGiven that the Cronbach alphas exceeded 0. 7, a single measure of initial and ? nal image, obtained as the mean of all the underlying items, was considered. Figures 2-4 gather the results according to the sector. For a better understanding of the effect on image, a single initial image (IMAG * (i)), calculated as the mean of initial brand images for close and far extensions, was taken into consideration. A new ? nal brand image (IMAG * (f)), resulting from adding IMAG * (i) to the difference obtained between the ? nal and the initial image in each scenario, was also considered. In general, these Responses to brand extensions 195 Figure 2. Brand image variation (toothpaste brands) Figure 3. Brand image variation (sport brands) Figure 4. Brand image variation (mobile phones brands) graphics suggest that ? rms should avoid entering markets far from their sector, since such extensions clearly entail brand image dilution. Once the global effect of extensions was analysed, the model hypotheses were tested. To test hypotheses related to feedback effects we created new variables based on unstandardised residuals. These residuals represent the brand image variation in such a way that higher values indicate more favourable feedback effects.They were obtained by regressing the post-test scores against the corresponding post-test scores, and the psychometrical properties of the resulting construct were similar to those of brand image factors (Cronbach’s alpha ? 0. 795). EJM 44,7/8 1196 Table VI contains the results of the model estimation and goodness of ? t measurements, which are acceptable and above the thresholds established in literature. Again, reasonable values were obtained for the error statistics ( RMSEA ? 0. 044; SRMR ? 0. 077) and the global ? t GFI (0. 892). The incremental ? t indexes also met the statistical requirements (AGFI ? 0. 74; CFI ? 0. 972; IFI ? 0. 972; NFI ? 0. 952; NNFI ? 0. 969). Next, the speci? c results concerning the hypotheses are commented. First, familiarity has a direct and signi? cant in? uence on initial brand image (best ? 0. 485; t-value ? 10. 419), as proposed in H1. However, contrary to H2, familiarity seems to have no signi? cant effect on extension attitude (best ? 2 0. 052; t-value ? 2 1. 443). Consequently, the most familiar brands will lead to more favourable brand associations, although not necessarily to a better assessment of the extension. The effect of initial brand image on extension attitude is signi? ant and positive (best ? 0. 232; t-value ? 6. 351), as proposed in H3. Therefore, consumers will prefer the brand extensions of companies that have managed to build and communicate positive brand associations. Since brand image depends on brand familiarity, consumer attitude toward brand extensions seems to be the result of a cognitive-affective sequence (Fishbein and Ajzen, 1975). Supporting H4, category ? t seems to be a clear determinant of extension attitude (best ? 0. 299; t-value ? 2. 439). In the same way, extension attitude is signi? cantly dependant on image ? t (best ? 0. 587; t-value ? 4. 76), which con? rms H5. Consequently, consumers will prefer those extensions marketed in a category that ? ts the brand portfolio, especially in terms of general brand associations. The effect of perceived dif? culty on extension attitude is positive (best ? 0. 035), as expected. Nevertheless, the coef? cient relating both factors fails to reach statistical signi? cance (t-value ? 1. 186), which implies rejecting H6. This lack of statistical signi? cance reveals that consumers do not consider dif? culty of manufacturing as a heuristic of the perceived quality of the new product. Hypotheses H1: FAMI !IMAG (i) H2: FAMI ! EXAT H3: IMAG (i) ! EXAT H4: CAFI ! EXAT H5: IMFI ! EXAT H6: DIFF ! EXAT H7: INNV ! EXAT H8: EXAT ! IMAG variation H9: CAFI ! IMAG variation H10: IMFI ! IMAG variation Standardised b (t) 0. 485 * 2 0. 052 0. 232 * 0. 299 * 0. 587 * 0. 035 0. 093 * 0. 631 * 2 0. 050 0. 159 (10. 419) (2 1. 443) (6. 351) (2. 439) (4. 876) (1. 186) (2. 924) (5. 846) (2 0. 313) (1. 004) Hypotheses validation Yes No Yes Yes Yes No Yes Yes No No Table VI. Results of the structural model Notes: *Signi? cant at p # 0. 05; Fit indices: Global ? t: x 2 ? 1131. 700 (481); p , 0. 001; GFI ? 0. 892; RMSEA ? . 044; SRMR ? 0. 077. Incremental ? t: AGFI ? 0. 874; CFI ? 0. 972; IFI ? 0. 972; NFI ? 0. 952; NNFI ? 0. 969; CRC: Composite reliability coef? cient; EVA: Extracted variance analysis, GFI: Goodness of ? t index; RMSEA: Root mean square error of approximation; SRMR: Standardised root mean square residual; AGFI: Adjusted goodness of ? t index; CFI: Comparative ? t index; IFI: Incremental ? t index; NFI: Normed ? t index; NNFI: Non-normed ? t index Regarding H7, consumer innovativeness appears to have a clear, though reduced, effect on extension attitude (best ? 0. 093; t-value ? 2. 924).All in all, attitude towards extensions will be fundamentally explained by the initial brand image (H3), perceived ? t (H4 and H5) and, to a lesser extent, by other factors such as consumer innovativeness (H7). H8 to H10 indicate the factors that explain the potential feedback effects of brand extensions on brand image. With respect to H8, extension attitude has a positive and signi? cant effect on brand image variation (best ? 0. 631; t ? 5. 846). Hence, the more favourable the attitude to the extension is, the more favourable the attitude toward the extended brand will be. Because of the high coef? ient obtained, companies launching brand extensions will have to avoid damaging their brands with low quality products. Contrary to our expectations, perceived category ? t has no direct effect on brand image variation, which rejects H9 (best ? 2 0. 050; t ? 2 0. 313). Despite showing a relatively high and positive coef? cient, the effect of image ? t proposed in H10 is not signi? cant either (best ? 0. 159; t ? 1. 004). The lack of signi? cance in both coef? cients suggests that the in? uence of ? t on brand image variation is only indirect through extension attitude (H4 and H5).To sum up, then, while perceived image and category ? t are essential factors for the success of a brand extension, it is signi? cant that extension attitude synthesises their effects. The centralising role of extension attitude was also corroborated by checking through the estimation of competitive models that neither brand familiarity nor consumer innovativeness nor perceived dif? culty have direct effects on brand image variation. Given the importance that literature attaches to perceived ? t to explain feedback effect (e. g. Loken and John, 1993; John et al. , 1998) and the lack of signi? ant effects in our model, we took a new step in the analysis. According to Czellar (2003), perceived ? t may moderate the in? uence of the attitude to the extension on the attitude to the extended brand. In the same way that high-perceived ? t increases the transference of brand associations to the new product (Aaker and Keller, 1990; Czellar, 2003), we think that the opposite effect could take place. This possibility was explored by means of two multi-sample analyses for each of the ? t dimensions, category ? t and image ? t. Speci? cally, the sample was split into high ? t (mean . 4) and low ? (mean , 4) and the structural model were replicated without considering direct effects of ? t. The Lagrange Multiplier (LM) Test and the maximum likelihood estimation method determined whether the model coef? cients are signi? cantly different (Iglesias and ? Vazquez, 2001). The comparison between the considered sub-samples yields interesting results. Although the effect of extension attitude on image variati on was similar for category ? t (x2dif ? 0. 182; p . 0. 1), the results lend support to the existence of moderating effects for image ? t at 90 per cent (x2dif ? 2. 868; p ? 0. 090). In the expected direction, the in? ence of extension attitude was higher in the high ? t condition (best ? 0. 810; t ? 12. 740) than in the low ? t one (best ? 0. 666; t ? 11. 203). In consequence, spillover effects between the brand and the extension (forward and backward) will depend on image ? t perceptions rather than on category ? t. Responses to brand extensions 1197 EJM 44,7/8 1198 Discussion A brand is one of the most important assets for ? rms and, therefore, marketing managers must be on the alert for inadequate strategies that erode brand assets. One of this potentially risky strategies involves the launching of unsuitable brand extensions ? hat erode extended brand bene? ts and associations (Mart? nez and de Chernatony, 2004; Diamantopoulos et al. , 2005). However, so far there is no clear u nderstanding of the main variables leading to spillover effects between brand extensions and parent brands and their relative in? uence. The present work proposes a model to ? nd out how extension strategies affect brand image, one of the major dimensions of brand equity. Unlike most previous research, this paper focuses on extension evaluation and feedback effects on the core brand as interrelated rather than independent phenomena.Moreover, it incorporates a few key variables into an operative model instead of considering most of the potential variables that might divert the attention of researchers and practitioners alike. The estimation of this model showed positive goodness-of-? t indexes and, without considering non-validated relationships, it sheds some light on the main factors and processes explaining consumer attitude. According to the literature, core parent brand experience positively in? uences probability of extension trial (Swaminathan et al. , 2001; Swaminathan, 2003) .However, our results reveal an indirect effect of brand experience or brand familiarity on consumer attitude to brand extensions. This variable has a distinctive in? uence on brand image, which, in turn, affects the assessment of the new category. These results are coherent with the behaviour models de? ned by some authors who maintain that the individual’s beliefs determine attitude and this, in turn, determines purchase behaviour (Fishbein and Ajzen, 1975). From this perspective, brand image, rather than brand familiarity, would explain consumer attitude to the extension. Our ? dings validate previous results in the literature concerning the positive effects of perceived ? t, either category or image ? t, on consumer attitude. In the same way, it was con? rmed that consumer innovativeness increases likelihood of consumer ? acceptance, although to a lesser extent than perceived ? t (Volckner and Sattler, 2006). Nevertheless, we could not verify the proposed relationship bet ween the attitude to the extension and dif? culty in manufacturing the new category. Due to the clear inconsistency of results along studies, the relevance of this variable proposed by Aaker and Keller (1990) should be questioned.In relation to feedback effects, our results suggest that perceived ? t (category and image) has no direct effect on the extended brand image, though an indirect effect occurs through attitude to the extension. Previous works focusing on the in? uence of perceived ? t on parent brand associations have mostly resorted to experimental settings (e. g. Loken and John, 1993; Milberg et al. , 1997; John et al. , 1998) rather than SEM models. Therefore, this relationship cannot be taken for granted in complex models where several constructs are interrelated. The estimation of the model also revealed that image ? moderates the effect of extension attitude on image variation. In the light of the results, consumers that perceive the extension as coherent with the bra nd image will modify their brand associations mainly on the basis of their resulting attitude. A high ? t perception usually entails a categorisation process where the extension is associated to the brand category and leverages the current beliefs and attitudes (Monga and Houston, 2002). According to our results, this process occurs in the opposite direction in such a way that a high ? t will involve the leveraging of the attitude to the extension.The results obtained are thus in line with those works that indicate that consumer attitude toward brand extensions mainly depends on perceived ? t (Aaker and ? Keller, 1990; van Riel et al. , 2001; Volckner and Sattler, 2006). Moreover, it contributes to the body of knowledge by showing that the effect of perceived category and image ? t on the extended brand image is not direct. On the contrary, it occurs an indirect effect through extension attitude and, in the case of image ? t, a further moderating effect on the relationship between e xtension attitude and image variation. To sum up, the coef? ients obtained indicate that extension attitude is especially determined by perceived category ? t, image ? t and initial brand image, which, in turn depends on familiarity. Consumer innovativeness is also a factor that explains consumer response to brand extensions. Furthermore, the results reveal that the existence of positive feedback effects will be an immediate consequence of the attitude to the extension. These results clearly support the basic argument of our model: the consumer will assess the product according to a series of variables and, as a result, the consumers will modify the initial brand schema.Implications Considering all the results obtained as a whole, we can make some recommendations for ? rms launching brand extensions. There is no doubt that the most important aspect for the success of an extension is coherence with the image of the extended brand. Though positive, it is not essential that the new pro duct or service belongs to a new category, but the ? rm has to be able to communicate the brand essence to the different markets (Kim, 2003). Once the new product is ? rmly associated to the current brand image, consumers will perceive a high quality of the new product and the risk associated to purchasing it will be lowered.Although innovative consumers are expected to prefer low-? t products (Xie, 2008), consumer innovativeness is a factor with a weak effect on the attitude to the extension. In comparison to introducing a new brand name, brand extensions will increase consumer trust and reduce the weight of consumer innovativeness as a risk reliever. Since consumer behaviour will be relatively similar regardless of consumer predisposition to new products, this factor should not be used for potential market segmentation. In consequence, companies must identify other consumer characteristics able to alter perceptions of quality and purchase ntentions of speci? c product categories. A favourable initial image will also be positive for consumer acceptance increasing the appeal of the new product. This image is hard to obtain in the short term, although our model suggests that increasing familiarity through communication or brand trials is an effective way of building brand associations. Since brand familiarity does not directly in? uence extension attitude, companies do not have to worry when their brands are not familiar enough or the current market share is scarce.Whenever they are capable of transmitting a positive brand image and ? t is high, success should be easy to obtain. Moreover, launching products perceived as trivial or very easy to make will not prevent consumers from trying the new product, a concern highlighted by Aaker and Keller (1990). Responses to brand extensions 1199 EJM 44,7/8 1200 Once consumers have developed a favourable attitude toward the new product, the brand associations might not be diluted but even strengthened. Provided perceived ? between the extension and the core brand is high, especially on the basis of image ? t, the attitude to the extension will be the main driver of feedback effects. Consequently, increasing the success of brand extensions and protecting the leveraged image are not con? icting but complementary goals. Companies should thus address their efforts towards the success of the extension by building a bundle of coherent and strong brand associations. This is the best way to avoid the risk of image dilution. Future research Our ? ndings raise several issues for future research. The ? st issue refers to the lack of time between the extension stimulus and the subsequent measurement of brand image, which is the common procedure in most studies. The fact of the matter is that higher experience reduces the likelihood of negative feedback effects (Sheinin, 2000; Swaminathan, 2003), since the mere exposure to the new product affords consumers to establish links with the brand that, otherwise, woul d not exist (Klink and Smith, 2001). However, experiments requiring the cooperation of respondents over time are likely to suffer from a â€Å"history problem† caused by the in? ence of external events (Campbell ? and Stanley, 1963). By analysing FMCG through a longitudinal study, Volckner and Sattler (2008) show that feedback effects diminish over time, although they also admit the possibility of confounding effects. Taking into account the advantages and disadvantages of the different procedures, the present study opted to exclude extraneous variables by minimising the time between pre and post-test scores. Since we aimed to test the interrelationships between factors, the setting of the study was designed to reinforce internal validity as much as possible.Consequently, it must be observed that the paper generates a picture of feedback effects in the short-term and these effects should be checked through a long period of time. It would be also advisable to verify whether th e validated relationships are consistent when consumers are exposed to all the market signals (competitors action, distribution support, etc. ) by using real extensions. Another issue to consider is whether the model can be applied to extensions of the same category or line extensions. Since line extensions are products with a higher perceived degree of ? t (Grime et al. 2002), there is a possibility that the relationships are sustained. It might be even more interesting to study whether service companies can successfully extend to the goods markets and vice versa. Indeed, it would be worthwhile to examine the brand and extension conditions that lead to higher effects of perceived ? t dimensions on the extension attitude toward the brand. Given that the in? uence of consumer innovativeness on extension attitude was less than expected, further research could also explore whether consumer innovativeness has moderating effects rather than mediating ones.Klink and Smith (2001) proved th at the in? uence of perceived ? t on extension attitude is lower among innovative consumers, who are more receptive to new products. The in-depth study of other variables related to personality, such as sensation-seeking or impulsive decision-making, also deserves attention Finally, it would be convenient to revise other measurement scales for brand image, which include a higher number of items. Brand image is a complex construct that sums up every association linked to the brand and may involve attributes, bene? ts, and attitudes (Keller, 1993).Although the proposed scale can be a suitable proxy, further research should deal with the limitations derived from the items used for measuring brand image and the remaining factors as well. References Aaker, D. A. (1996), â€Å"Measuring brand equity across products and markets†, California Management Review, Vol. 38 No. 3, pp. 102-20. Aaker, D. A. (2002), Brand Portfolio Strategy, The Free Press, New York, NY. Aaker, D. A. and Kell er, K. L. (1990), â€Å"Consumer evaluations of brand extensions†, Journal of Marketing, Vol. 54 No. 1, pp. 27-41. Ahluwalia, R. and Gurhan-Canli, Z. 2000), â€Å"The effects of extensions on the family brand name: an accessibility-diagnosticity perspective†, Journal of Consumer Research, Vol. 27 No. 3, pp. 371-81. Alba, J. W. and Hutchinson, J. W. (1987), â€Å"Dimensions of consumer expertise†, Journal of Consumer Research, Vol. 13 No. 4, pp. 411-54. Alexander, N. and Colgate, M. (2005), â€Å"Customers’ responses to retail brand extensions†, Journal of Marketing Management, Vol. 21 Nos 3/4, pp. 393-419. Anderson, J. R. (1983), â€Å"A spreading activation theory of memory†, Journal of Verbal Learning and Verbal Behavior, Vol. 22 No. 3, pp. 61-95. Balachander, S. and Ghose, S. (2003), â€Å"Reciprocal spillover effects: a strategic bene? t of brand extensions†, Journal of Marketing, Vol. 76, January, pp. 4-13. Barrett, J. , Lye, A. and Venkateswarlu, P. (1999), â€Å"Consumer perceptions of brand extensions: generalising Aaker and Keller’s model†, Journal of Empirical Generalisations in Marketing Science, Vol. 4, pp. 1-21. Bhat, S. and Reddy, S. K. (2001), â€Å"The impact of parental brand attribute associations and affect on brand extension evaluation†, Journal of Business Research, Vol. 53 No. 3, pp. 111-22. Boush, D. M. nd Loken, B. (1991), â€Å"A process-tracing study of brand extension evaluation†, Journal of Marketing Research, Vol. 28 No. 1, pp. 16-28. Boush, D. M. , Shipp, S. , Loken, B. , Genturck, E. , Crockett, S. , Kennedy, E. , Minshall, B. , Misurell, D. , Rochford, L. and Strobel, J. (1987), â€Å"Affect generalization to similar and dissimilar brand extensions†, Psychology and Marketing, Vol. 4 No. 3, pp. 225-37. Brown, T. J. and Dacin, P. A. (1997), â€Å"The company and the product: corporate associations and consumer product responses†, Journal of Marketing, Vol. 61 No. 1, pp. 68-84.Campbell, D. and Stanley, J. (1963), Experimental and Quasi-experimental Designs for Research, Rand McNally, Chicago, IL. Chang, J. W. (2002), â€Å"Will a family brand image be diluted by an unfavorable brand extension? A brand trial-based approach†, Advances in Consumer Research, Vol. 29, pp. 299-304. Czellar, S. (2003), â€Å"Consumer attitude toward brand extensions: an integrative model and research propositions†, International Journal of Research in Marketing, Vol. 20 No. 1, pp. 97-115. Dawar, N. (1996), â€Å"Extensions of broad brands: the role of retrieval in evaluations of ? †, Journal of Consumer Psychology, Vol. 5 No. 2, pp. 189-207. de Ruyter, K. and Wetzels, M. (2000), â€Å"The role of corporate image and extension similarity in service brand extensions†, Journal of Economic Psychology, Vol. 21 No. 6, pp. 639-59. Responses to brand extensions 1201 EJM 44,7/8 Diamantopoulos, A. , Smith, G. and Grime, I. (2 005), â€Å"The impact of brand extensions on brand personality: experimental evidence†, European Journal of Marketing, Vol. 39 Nos 1/2, pp. 129-49. Eagly, A. H. and Chaiken, S. (1993), The Psychology of Attitudes, Harcourt Brace Jovanovich, Fort Worth, TX. 202 Fishbein, M. and Ajzen, I. (1975), Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research, Addison-Wesley Publishing, Reading, MA. Glynn, M. S. and Brodie, R. J. (1998), â€Å"The importance of brand-speci? c associations in brand extension: further empirical results†, Journal of Product & Brand Management, Vol. 7 No. 6, pp. 509-18. Grime, I. , Diamantopoulos, A. and Smith, G. (2002), â€Å"Consumer evaluations of extensions and their effects on the core brand: key issues and research propositions†, European Journal of Marketing, Vol. 36 Nos 11/12, pp. 415-38. Gurhan-Canli, Z. and Maheswaran, D. (1998), â€Å"The effects of extensions on brand name dilution and enhancement†, Journal of Marketing Research, Vol. 35 No. 4, pp. 464-73. Hair, J. F. , Anderson, R. E. , Tatham, R. L. and Black, W. C. (1998),