The Nuances of Brand Personality: A Corpus-assisted Linguistic Analysis of Web-based Communications of Fashion Brands

The ability to establish a particular brand personality (i.e. a set of human personality traits that consumers associate with a brand) is a key component of fashion brand management and communication. A given fashion brand may use language that communicates different personality traits (e.g. glamourous, exciting, youthful, exotic) as a way to define its own personality and distinguish itself from other fashion brands. Based on a corpus consisting of companyproduced texts collected from the websites and Facebook pages of over 100 fashion brands, this study aimed to determine which traits of brand personality emerge, which are the most frequent, and which nuances of meaning can be identified within them. This was accomplished by means of text analysis software that identifies statistically significant semantic domains to which conceptually-related lexical items are assigned. The analysis revealed 14 key semantic domains that were linked to various brand personality traits. Among the most prominent were Judgement of appearance: Positive, Time: New and young, Relationship: Intimacy and sex, and Unexpected, highlighting not only the traditional importance attributed to attractiveness, but also to sensuality and non-conventionality as desirable traits of fashion brand personality. Other distinctive traits that emerged as significant involved exclusivity (encoding the value of elitism) and iconicity (emphasizing high stature and uniqueness). The study offers insights into how fashion brands utilize web-based communications to convey brand personality. It also offers a useful methodology that fashion companies can adapt to ensure that they are effectively communicating the intended brand personality.


Introduction
In today's competitive business environment, companies are keen to communicate a strong brand as a way to distinguish themselves in an increasingly crowded global marketplace (Wheeler 2012). One way to achieve this goal is to establish a distinctive brand personality, defined by Aaker (1997: 347) as "the set of human characteristics associated with a brand". Defining and conveying to consumers a particular brand personality becomes an important strategy to encourage attachment and loyalty, as consumers seek brands that conform to their own self-image (Malär et al. 2011). The aim of the present study is to shed light on how fashion brands define and communicate brand personality through an in-depth linguistic analysis of their website and Facebook communications. Particular attention is paid to identifying specific traits of brand personality and nuances in their meanings, as well as possible variation across fashion market segments. In the following section, I provide an overview of brand personality as a theoretical construct, starting from related constructs discussed in the marketing literature. I then review previous research focusing on brand personality in the context of fashion brand communication.

Brand personality and its role in fashion brand communication
Over the years, marketing research has developed a number of fundamental and interconnected constructs that can be leveraged by companies to effectively promote their brands. Among these are brand identity, which consists of a unique set of meanings that a company defines for its brand and then aims to establish in the minds of consumers, and brand image, which refers to perceptions of a brand in terms of the attributes that consumers associate with it and retain in their memory (Keller et al. 2008). Brand distinctiveness, as perceived and valued by consumers, should also be "at the centre of brand strategy" (Romaniuk et al. 2007: 42).
Looking more closely at the attributes that consumers associate with brands, Keller et al. (2008) characterized them as either product-related or non-product-related. Product-related attributes refer to the physical features of the product itself (e.g. colour, size, materials), or to the purchase (e.g. price, point of sale) and consumption (e.g. packaging, informative literature) of the product. Non-product-related attributes are less tangible in nature and reflect imagery, experiences, and feelings that consumers have in relation to a product. Such attributes are expressed through the language consumers use to characterize a brand, e.g. elegant, modern, classic. Thus, a brand can be perceived "as a product, a personality, a set of values, and a position it occupies in people's minds." (Temporal 2011: 1). Building on this notion, we can identify brand personality (Aaker 1997) as another key construct discussed in marketing studies, which takes inspiration from pioneering studies on human personality research (Norman 1963, McCrae/John 1992. Aaker (1997) developed a highly influential theoretical framework for studying brand personality: the Brand Personality Scale 1 , based on five overarching personality dimensions of brands, including Sincerity, Excitement, Sophistication, Competence, and Ruggedness. These dimensions are then further articulated into different personality traits. For instance, the dimension Sincerity contains the traits of down-to-earth, honest, and friendly, while the dimension Excitement contains the traits daring, spirited, and unique. Aaker's (1997) large-scale empirical research based on consumer perceptions of various brands related to different types of products and services (e.g. clothing, appliances, insurance, hotels, drinks, newspapers) was able to identify not only the five overarching dimensions, but also 42 traits of personality across them. Following Aaker's work, other researchers have suggested similar models for investigating brand personality, although with a stricter psychological interpretation of the concept. Geuens et al. (2009) excluded the dimensions of Sophistication and Ruggedness that they considered to be unrelated to human personality. Azoulay/Kapferer (2003: 152) rejected the adjective feminine as a brand personality trait, which Aaker (1997) instead lists under the dimension Sophistication, arguing that "gender is absent from psychological scales of personality". Similarly, Azoulay/Kapferer (2003) interpreted the items "young" and "upperclass" found in Aaker's model as descriptions of age and social class, respectively, and not personality. Thus, there appears to be no clear consensus on a precise list of traits considered to be expressions of brand personality. Nonetheless, Aaker's more comprehensive scale continues to be extensively applied in studies of brand personality across different industries (e.g. Ekinci et al. 2006, Fetscherin/Toncar 2010 and cultures (e.g. Bosnjak et al. 2007, Sung/Tinkham 2005.
Brand personality is investigated in the present study in the context of fashion brand communication. From a theoretical perspective, fashion has been broadly defined as a "culturally endorsed form of expression" (King et al. 1974(King et al. /2011, which is widely accepted by a group of individuals and subject to change over time. Its expressive and social nature is also reflected in the language that fashion companies use to describe their brands, typically in terms of the socially desirable personality traits discussed above. Moreover, fashion brands are characterized by highly elaborate visual and tactile properties, as well as the iconic personalities of fashion designers (e.g. Valentino, Calvin Klein, Karl Lagerfeld) recognized within the fashion community, which may also emerge in the expression of brand personality. According to Müller/Chandon (2003), brand personality is affected by the nature of the product. Indeed, the pronounced sensorial and social nature of fashion products means that how fashion companies communicate brand personality takes on more importance compared with more utilitarian types of products.
Relatively few studies have focused specifically on the personality of fashion brands. These have mainly addressed luxury fashion brands from the perspective of consumer perceptions. Focus ing only on Aaker's (1997) Excitement dimension, Ismael/Spinelli (2012), Ismael/Melewar (2015, and Anggraeni (2015) showed that this dimension had a positive impact on wordof-mouth communications among young consumers of fashion brands in the United Kingdom, Switzerland, and Indonesia. Tong et al. (2017) applied Aaker's (1997) brand personality scale to analyse young American consumers' perceptions of luxury fashion brands. They identified six personality dimensions (Prestigious, Competent, Sociable, Creative, Snobby, and Romantic), thus building on the original scale and providing insights into the values of the survey respondents as luxury fashion consumers. Heine (2010) conducted interviews with luxury consumers seeking to develop a brand personality scale that specifically targets luxury fashion brands. He identified the five dimensions of Modernity, Eccentricity, Opulence, Elitism, and Strength, which show some overlapping with Aaker's (1997) dimensions, while also highlighting some traits that are distinctly linked to the world of luxury (i.e. Opulence and Elitism). With reference to challenges faced by luxury fashion brands, Kim/Hall (2014: 31) highlighted the need to contend with dynamic and fluid aspects of personality which can change in response to demand for fresh product offerings: "as even "classic" brands need to change their strategic direction and modernize their image, brands cycle through different personas through the years".
Even fewer studies have looked at fashion brand personality from the perspective of the company. While it is clearly important to understand how fashion consumers perceive brand personality, it is also vital for companies to ensure that they are communicating the desired personality traits of their brands in order manage and promote them in the most successful way possible. Ranfagni et al. (2016) performed a large-scale linguistic analysis of fashion brand personality as expressed by consumers and as defined by fashion companies. This was done by comparing personality-related adjectives found in fashion blogs (representing consumers) with those found on fashion company websites. The results revealed varying levels of alignment, suggesting that some fashion companies were more successful at communicating their brand personality than others. However, the main aim of this study was to identify alignment/disalignment of perceived vs. intended brand personality, without an in-depth analysis of the language used by fashion companies to communicate brand personality, which personality traits are emphasized, and which nuances of meaning emerge. To address these issues, the present study poses the following research questions: 1. Which traits of brand personality are conveyed through the language used by fashion companies on their websites and Facebook communications?

Web-based communications for promoting fashion brands
Web-based marketing communications represent an environment where consumers proactively seek out information about companies and their products, and therefore play a key role in shaping attitudes and perceptions in relation to brands (Madhavaram/Appan 2010). The corporate website constitutes an important tool that companies use to communicate about their brands in an effort to position themselves in the market and influence consumer choices (Salvi et al. 2007). In a study based on 160 corporate websites, Hwang et al. (2003) found that they tend to be multifunctional, both providing information and engaging in strategies to enhance brand image. Interestingly, the websites of clothing and footwear companies showed a significantly higher presence of "transformational" messages associated with the psychological characteristics of consumers who use the brand, as opposed to "informational" messages (Hwang et al. 2003: 12). In a study on the impact of visiting a brand's website on perceptions of brand personality, Müller/Chandon (2003) found that perceptions of youthfulness/modernity and sincerity/confidence increased significantly among participants who had been exposed to websites compared to those who had not. This underscores the importance of maintaining effective web-based communications to strengthen perceptions of desirable personality traits of brands. Research has shown that users respond favourably to fashion websites that provide both information about products and brands (Kim/Stoel 2004), as well as high quality visual images (Siddiqui et al. 2003), which likely work in synergy to trigger positive perceptions.
In recent years, the promotional role of the corporate website has been enhanced by social media platforms such as Facebook and Twitter, which are now an increasingly vital component of marketing communications, according to Kietzmann et al. (2011). These authors stressed the importance of companies "starting and manipulating a conversation" (Kietzmann et al. 2011: 245), if they hope to establish and maintain control of brand communication targeting today's digitally-savvy consumers. With particular reference to fashion brands, Kim/Ko (2012) noted that even some of the most traditional luxury fashion houses now have their own Facebook pages and Twitter accounts to engage with customers in an effort to build affection and boost desire for luxury. The same authors conducted empirical research that revealed a positive impact of luxury fashion brands' social media marketing activities on both customer relations and purchase intention. Thus, social media platforms provide opportunities for fashion brands to enhance and reinforce company-defined brand personality among consumers.
After having identified web-based communications of fashion companies as the research context for the present study, in the next section I provide a detailed description of the methodology implemented for the linguistic analysis of fashion brand personality.

Methodology
The present study is based on a corpus of texts collected from fashion brand web-based communications. It is thus grounded in corpus linguistics, a methodological approach that relies on special software programs to perform automated and empirical analyses of authentic language compiled into a body of electronically stored texts. The methodologies of corpus linguistics allow researchers to identify trends in language usage in a given communicative context which may escape detection in traditional manual discourse analysis, which is necessarily based on relatively limited amounts of text. In particular, this study follows the principles of corpus-assisted discourse analysis (see Baker et al. 2008, Partington 2010, which uses quantitative corpus methods to first retrieve linguistic items of interest within particular discourse types, and then integrates follow-up qualitative analysis within their context of usage to reveal any distinctive trends or patterns. This analytical process will be further described and articulated into phases in Section 4.2.

Data source and corpus collection
In an effort to identify a sufficiently representative sample of fashion brand communication to be included in the study, the starting point was the designer directory of Style.com 2 , a popular website dedicated to news about brands, people, and events in the international fashion community. At the time of consultation, the designer directory included 335 brands representing a wide range of nationalities and categories, e.g. luxury/retail, women's/men's clothing, well-established/emerging brands. For each of these brands, the Internet was searched to locate its website, or alternatively, its Facebook page from which to collect company-produced textual data to be copied and pasted into electronic files. For the purposes of this research, I limited data collection to these two sources because they contain paragraph-sized text material that would provide an adequate amount of linguistic data for quantitative elaboration with corpus tools (see section 4.2). Although many companies also utilize Twitter, the space restrictions of this medium (140 characters at the time of data collection) rendered it impractical for inclusion in this study which implemented a manual collection method to compile a relatively large amount of text suitable for corpus-assisted discourse analysis.
On the company websites, only texts with an underlying promotional purpose were collected, specifically those that contained information about the brand and/or designer, history of the company, brand-related press releases, descriptions of collections, and interviews with designers. On the Facebook pages, in line with the aims of the study, only company-produced texts were collected, for example, those describing promotional initiatives and upcoming events related to the brand. Both the website and Facebook texts were characterized by richly articulated descriptions, highlighting the positive attributes of brands, products, and associated people/entities, and thus steering consumer perceptions of them.
The data collection process revealed that a large number of brands -particularly the lesser known ones -did not have websites, Facebook pages, or any other source of company-produced textual data available on online sources. A number of sites had blocked text that did not permit copying and pasting. There were also many sites that contained essentially images with very little textual material beyond sporadic words or phrases, apparently opting to communicate their brand personality largely through the visual mode. At the end of the data collection process, there were 109 fashion brands whose online communications provided a sufficient amount of text to be included in the corpus. See the Appendix for an overview of the Fashion Brand Corpus (hereafter FBC), including the fashion brands, their market segments, the source(s) of textual data, and the number of words collected for each brand.
As can be seen, the FBC contains a total of 195,555 words. It therefore constitutes a small specialized corpus, lying within the range of 100,000-500,000 words considered suitable for corpus-assisted discourse analysis (Flowerdew 2004). The amount of textual data available on the fashion brand web sources varied considerably, ranging from a minimum of 127 words to a maximum of 9,209 words. The majority of the brands utilized both their websites and Facebook; others utilized websites or Facebook only.

Analysis
The analysis of the FBC was undertaken in three phases. In the first phase, the corpus was investigated with Wmatrix corpus analysis software (Rayson 2009). In addition to generating concordances of queried items, Wmatrix is capable of analysing texts from a semantic perspective by means of a semantic annotation tool which automatically assigns each lexical item to one of over 200 pre-established semantic domains. 3 It was then possible to extract key semantic domains, i.e. those that were statistically more frequent in the FBC compared to a reference corpus. Wmatrix calculates a keyness score with statistical significance at the 99% level of confidence (p < 0.01) based on the log likelihood test. To select an appropriate reference corpus between the options of American and British English, a preliminary analysis of orthographic variation (e.g. color vs. colour) showed considerably higher frequencies of American spelling across the corpus. Therefore, the FBC was compared to the AmE06 corpus of general written American English (966,609 words) incorporated within Wmatrix.
In the second phase, I carefully examined the lists of lexical items assigned to each key semantic domain to identify those that could be interpreted as personality traits associated with fashion brands, referring broadly to the five overarching personality dimensions and corresponding 42 traits of Aaker's (1997: 354) Brand Personality Scale, summarized in Table 1. To investigate potential differences in the expression of brand personality across the fashion market segments represented in the FBC, five corresponding sub-corpora were created. Each sub-corpus was then annotated semantically with Wmatrix and analysed using the same procedures described above. In the third phase, for qualitative insights, an in-depth contextual analysis was performed on selected items encoding fashion brand personality within the key semantic domains that emerged across the FBC. This entailed examining the individual items within concordances generated by Wmatrix in order to determine possible patterns and nuances in meaning in relation to how personality traits were used to qualify to the fashion brands and fashion-related entities. This section has provided a thorough description of the source of the linguistic data for the FBC, the collection process, and the analytical procedures. In the following section, I present the results of the analysis of brand personality across the FBC that integrated both quantitative and qualitative methods.

Key semantic domains in the FBC
The semantic analysis initially retrieved 94 domains that occur with statistically higher frequency in the FBC compared to the AmE06 reference corpus. More specifically, these are the domains with a keyness score of 6.63 or higher, representing the cut-off value of statistical significance at the 99% level of confidence (p < 0.01). After examining the lists of lexical items assigned to each of these domains, I identified 14 whose items could be interpreted as expressing various traits of brand personality. During this process, I broadly referred to Aaker's (1997) (51), youthful (12) latest (57), avant-garde (12), revolutionary (5) Evaluation: Good A5.1+++ A5.1+ 250.46

Brand personality traits in the FBC
A comparison of the keyness scores in Table 2 offers some insights into which traits of brand personality emerge with the highest frequencies across the corpus and are therefore the most distinctive. Across the 14 domains, the keyness scores ranged from relatively low (<100) to relatively high (100-400), and with one high outlier score of 1558.10, corresponding Judgement of appearance: Positive. Not surprisingly, fashion brands are keen to communicate traditional attributes related to attractive appearance (e.g. beautiful, stunning, lovely), but also to positive personality traits (e.g. chic, stylish, classy) that they want consumers to associate with their brands. Interestingly, Time: New and young had higher scores (381.15 and 71.21) than Evaluation: Good (250.46 and 67.13). On the whole, the fashion brands seemed to place more emphasis on modernity and youthfulness than on generic positive evaluation relating mainly to high quality. The strength of Judgement of appearance: Positive, Evaluation: Good, and Time: New and young across the FBC can be further explained by borrowing from marketing research the concept of brand authenticity, i.e. "a subjective evaluation of genuineness ascribed to a brand by consumers" (Napoli et al. 2014(Napoli et al. : 1091. According to Beverland (2009), companies build brand authenticity through the combination of tradition and innovation, which is reflected in these three most frequent domains.
The other relatively high scores of Relationship: Intimacy and sex, Unexpected, and Interested/excited/energetic highlight the importance attributed to the sensual aspect of brand personality (e.g. romantic, seduction, sexiness), as well as a desire to be perceived as atypical (e.g. surprising, astounding, unpredictable) and exciting (e.g. vibrant, dynamic, exhilarating). The domains Judgement of appearance: Positive, Time: New and young, Evaluation: Good and Interested/excited/energetic occurred in all or in the vast majority of the brands across the corpus, while Unexpected occurred in only 31 brands that opted to distinguish their personalities in this way. Most of the remaining domains with relatively low keyness scores were less well distributed across the corpus and tend to be associated with fewer brands, typically fewer than 50 out of the total of 109. For example, Ethical (G2.2+) occurred across the communications of only twelve brands. The exceptions were Important (A11.1+) that occurred across 74 brands and Comparing: Unusual (A6.2-) that occurred across 56 brands.
The comparative analysis based on fashion market segments is shown in Table 3. The wide variation in the word counts of each sub-corpus reflects the differences in availability of textual data during corpus compilation (see section 4.1), with the premium segment accounting for roughly two-thirds of the corpus. Although Wmatrix takes into account differing corpus sizes when calculating keyness, the disproportionately small sizes of the luxury, bridge, contemporary, and mass market sub-corpora resulted in insufficient quantities of lexical items within many domains to meet the 6.63 keyness score threshold. What appears most interesting is the presence/absence of key domains with respect to the FBC as a whole. Judgement of appearance: Positive and Time: New and young appeared across all segments, and Evaluation: good appears in 4/5 segments, suggesting that they constitute desirable personality traits of fashion brands in general. However, some segments distinguish themselves by also emphasizing other personality traits. Unexpected and Comparing: Unusual were key only in the premium and luxury segments, reinforcing the high value placed on uniqueness and exclusivity. Interested/excited/energetic and Happy emerged in premium, but also in contemporary, in line with its focus on young designers and modern-style clothing.  Table 3. Key brand personality domains across fashion market segments

Brand personality meanings in the FBC
A comparison of Table 1 and Table 2 shows that several semantic domains that emerged as key in the FBC overlap with some of Aaker's (1997) five overarching personality dimensions. The items found in Judgement of appearance: Positive, Evaluation: Good, and Expensive very largely align with Aaker's Sophistication, which similarly contains traits linked to beauty, high quality, and femininity. Time: New and young, Unexpected, Interested/excited/energetic, and Comparing: Unusual shared meanings of modernity, novelty, youth, and uniqueness with Aaker's Excitement. The items friendly, earthy, outgoing, happy, and joyful in the domains Informal/Friendly and Happy demonstrate conceptual convergence with some traits of Aaker's Sincerity. Able/intelligent included the same item intelligence listed under Aaker's Competence dimension. Tough/ strong contained tough and ruggedness also found in Aaker's Ruggedness. However, within these key semantic domains, some of the items seem to emphasize particular nuances of meaning within the personality dimensions, beyond those articulated by Aaker (1997). For example, in Judgement of appearance: Positive, the second most frequent item was exclusive, a meaning which does not emerge explicitly in Aaker's scale. In order to provide some context for how this personality trait was used, Figure 1 shows a sample of 15 lines from the concordance of exclusive generated by Wmatrix. Exclusive appears to have a rather multi-faceted usage, referring not only to fashion products as concrete entities, but also to abstract ideas, while reflecting the value of elitism. (1) Here, one of them: a stunning coat with a geometric pattern in sequins and stones. (Valentino/ O4.2+) (2) These products represent a joint effort between two companies whose philosophies complement each other with astonishing affinity and products of unsurpassed quality and precision. (Bottega Veneta/A5.1+)  The analysis also revealed three domains linked to certain personality traits that are not clearly accounted for in Aaker's (1997) scale, but that emerged in the FBC corpus as significantly frequent: Relationship: Intimacy and sex, Important, and Ethical. Follow-up analysis of Relationship: Intimacy and sex highlighted strong positive emotions attributed to both potential consumers (example 11), as well as the capacity of brands to enhance the sensual nature of whoever wears them (examples 12-13). Within the domain Important, the most frequent item was iconic, whose meaning goes well beyond mere importance to emphasize the unique stature achieved not only by the brand and wellknown designers (examples 14-15), but also by the people who wear the brands (example 16) The domain Tough/strong also contained the item ruggedness, but contextual analysis revealed some difference in its meaning. In example 17, ruggedness suggests practicality and resilience, while in example 18, tough invokes rebelliousness, further reinforced by the reference to studded leather. The domain Ethical revealed that some brands are keen to show their concern about ethical and social issues (example 19).
(19) As part of the textile and fashion industry we strive for a broad environmental and ethical approach to business (Cheap Monday/G2.2+) From the examples discussed above, we can also see that various items representing different personality traits may appear in combination within the same phrase. For instance, in example 2 unsurpassed (Evaluation: Good) is accompanied by astonishing (Unexpected). In example 7, the brand is described as alien and weird (Comparing: Unusual), but also strong (Tough/strong From a linguistic perspective, the brand personality traits discussed in this sub-section reflect the phenomenon of evaluation, i.e. how speakers and writers use language to express attitudes, opinions, and judgements. In the context of the company-produced communications, this actually amounts to "self-evaluation". The key personality traits that emerged show some convergence with Martin/White's (2005) appraisal model which articulates the linguistic expression of attitude into appreciation (evaluation of aesthetic quality), affect (expression of emotions), and judgment (moral evaluation of behaviours). 4 Most of the items encode appreciation by positively evaluating the fashion product/brand, for instance in the domain Judgement of Appearance (e.g. beautiful, elegance) or Evaluation: good (e.g. excellent, wonderful). Affect can be interpreted in the domains Happy (e.g. happy, joyful) and Relationship: Intimacy and sex (e.g. love, romantic) which invoke feelings and emotions associated with brands, while judgement is reflected in the domain Ethical referring to moral behaviour on the part of fashion companies. The overlapping of meanings within the key domains and the constructs of the appraisal model serve to further highlight the multi-faceted nature of fashion brand personality traits.
To summarize, in this section I have presented the quantitative analysis of brand personality in the FBC across 14 semantic domains from which distinctive personality traits emerged, including beauty, modernity, high quality, and sensuality (those that were the most frequent), but also non-conventionality, excitement, iconicity, and social ethics. Follow-up qualitative analysis revealed substantial alignment with Aaker's (1997) model of brand personality traits, as well as some additional nuanced meanings that appear to reflect particular attitudes and values associated with fashion brands. In the final section of this paper, I offer further interpretation of these trends, suggest ways to build on these findings in future research, and discuss possible applications in the area of fashion brand management.

Concluding remarks
Using corpus methods to systematically analyse the web-based communications of a large sample of fashion brands, this study has identified key traits of fashion brand personality. In terms of frequency, after positive judgements related to attractiveness, personality traits linked to newness and youthfulness emerged as the second most frequent domain. This is perhaps influenced by the nature of the fashion product which constantly renews itself from season to season, as well as the high value that modern society places on a youthful appearance. The relatively high frequencies of personality traits related to unexpectedness, atypicality, and excitement point to a desire to attract attention and 'stand out from the crowd' with respect to other fashion houses. Interestingly, these traits also show some conceptual overlapping with newness and youthfulness in that being new and young also implies being different and exciting. The comparative analysis across fashion market segments confirmed the key role of personality traits linked to attractive appearance and youthfulness in fashion brands, which were found across all segments. However, traits associated with uniqueness and exclusivity were key only in the premium and luxury segments, and excitement and happiness only in the contemporary segment, indicating some segment-related distinctiveness.
Across the corpus, many of the items aligned with Aaker's (1997) Brand Personality Scale. This instrument is based on empirical research involving multiple product categories and not limited to fashion, thus providing for a wide range of traits that may emerge during an analysis and, consequently, for more comprehensive results. However, other items encoded nuances of meaning that were not in alignment and appeared to be distinctive of the fashion brands. For example, a high premium was placed on sensuality, exclusivity, and the capacity to be different in pro-vocative and non-conventional ways. These results corroborate previous work focusing on luxury fashion brands, which similarly found personality meanings linked to elitism and eccentricity (Tong et al. 2017, Heine 2010. The present study also identified a strong emphasis on iconicity, highlighting the key role of fashion designers in company-defined fashion brand personality, which distinguishes it from other types of products (e.g. electronic goods, household appliances), where the individual who 'designs' them does not have the same status or visibility. In addition, the concern of some fashion brands to portray themselves as ethical could be an attempt to align themselves with the now widely accepted norm for companies to act in sustainable and socially responsible ways, an idea that was perhaps not yet so well consolidated within the social consciousness at the time of Aaker's (1997) study as it is today.
Thanks to the corpus-assisted analytical approach, this study has provided a comprehensive profile of the highly nuanced meanings of brand personality in the web-based communications of a large sample of fashion brands across different market segments, which could not have been achieved with traditional methods of marketing communication research. To build on these findings, it would be interesting to perform in-depth qualitative analyses of the web-based texts of the individual companies represented in the FBC. For example, a contrastive case-study approach could be implemented to determine which personality traits are used to characterize brands articulated at the level of product, collection, style, or design, while also taking into account other variables such as country origin or preferred channel(s) of communication. This would shed additional light on how companies may be similar or different in terms of the brand personality that they wish to communicate. This type of analysis could be complemented by field research within the fashion companies to elicit directly from brand managers which personality traits they intend for consumers to associate with their brands, which could then be compared to those that emerge from their web-based communications in order to determine possible discrepancies.
Given the fact that fashion brand websites also rely heavily on visual images to communicate brand personality (see section 4.1), it would be worthwhile to investigate the role of these features as semiotic resources that go beyond the verbal text. This would, of course, require a completely different analytical approach, for example multimodal discourse analysis (see Baldry/Thibault 2005), perhaps using data from image-driven social media such as Instagram or Pinterest. However, it would enable a better understanding of how fashion brands may use different modes of communication to convey a distinctive brand personality.
In terms of brand management, the methodology used in this study can offer a relatively user-friendly way for fashion companies to analyse their own web-based communications to ensure that the desired personality traits are actually emerging. They could also analyse the brand communications of their competitors to assess which personality traits might be shared, but also those that should render them distinctive. This would help fashion companies more effectively position themselves in the market and revitalize their brands when necessary, thus enabling them to have better control of their web-based communications. With particular reference to the social media landscape, as noted by Kietzmann et al. (2011), it is important for firms to utilize these platforms strategically by starting and then participating in ongoing conversations in order to steer perceptions of brand personality in the desired direction.