Internet slang, consisting of distinct pronunciation, word, morphology, and syntax derived from online context, is a variant of the standard language (Liu et al., 2019). The emergence of Internet slang depends on two factors: its users and the context. On the one hand, netizens, active online for a long time, create and speak Internet slang instead of the standard language to express their unique identity. On the other, compared to real communication context, online communication is more accessible, random, and secretive, constructing a different environment for the continuous evolution of Internet slang as an independent variety.

As such, we draw from schema theory to clarify the conceptual nature of Internet slang style. Schema theory describes how people recognize and understand the world by using cognitive structures to organize prior knowledge (Fiske, 1982). In this vein, schemata refer to the knowledge unit in the human mind, and an intermediary between objects and language. People organize schemata as a psychological structure network that represents shared meaning among many individuals. In the marketing context, this network allows individuals to form mental representations of ads, brands, or products, process, retrieve, and categorize information related to them, and finally facilitate decision-making (Sujan and Bettman, 1989; Halkias, 2015).

Similarly, people with direct or indirect experiences with online communication will develop a cognitive representation of Internet slang. In other words, they establish prototypical knowledge and consensual meaning of this particular language variety, forming the “style” perception of Internet slang (Jeffries and McIntyre, 2010, p. 127). As people see similarities between events or experiences during processing, and weave them into different schema categories (Schank, 1982), the “style” of Internet slang can also be further understood and measured based on the “types” of thematic meaning delivered.

To the best of our knowledge, the literature contains no conceptually useful definition of Internet slang style and its dimensions that can be further extended to a scale and used in business practices. Therefore, the definition of Internet slang style should be first established before scale development. To achieve this purpose, we searched extensively to identify articles on “linguistic style,” “language style,” “communication style,” and “advertising style” published in leading journals that encompass communication or linguistic topics. By carefully understanding these articles, we tried to establish an appropriate framework to introduce the conceptualization of the Internet slang style.

Recent general conceptualizations in prior literature seem to coalesce around two streams of frameworks. On the one hand, the construct “communication style” emphasizes how people communicate with others. In this vein, communication style can be “conceived to mean the way one verbally and paraverbally interacts to signal how literal meaning should be taken, interpreted, filtered, or understood (Norton, 1978, p.99).” Prior research often uses this conceptualization to evaluate different ways that service providers use to interact with customers (e.g., Webster and Sundaram, 2009; Hwang and Park, 2018). On the other hand, “linguistic style,” or “language style,” reflects the linguistic nature of a word within a sentence structure, … and the meaning of a word provided by the semantics of the word and the rest of the sentence (Hung and Guan, 2020, p. 597). Operationalization of linguistic style usually relies on LIWC, a coding dictionary that categorizes nearly 6,400 words into 89 themes (Pennebaker et al., 2015; Kovacs and Kleinbaum, 2020). Eighteen out of these 89 themes directly capture linguistic style, and can be used to predict intentions or personality traits (e.g., Lee et al., 2019; Koh et al., 2020).

General conceptualizations outlined in this section underlie divergent theoretical natures. Although linguistic styles are indeed psychologically-derived, their operationalization is based on specific linguistic features. For example, a social linguistic style is associated with the number of social intercourse-related words, such as family, employee, neighbor, and personal pronouns (Lee et al., 2019). Therefore, the linguistic style framework cannot directly describe consumers’ perceptions of Internet slang as a unique language variety. Internet slang may contain very distinct elements or rules from the standard language and generate different consumer mindsets perceptions. By contrast, the framework of “communication styles” seems more in line with the essence of our intended definition of Internet slang style (i.e., the schema theory), as it emphasizes more on people’s schematic perceptions.

Responding to the argument that language features should be examined with consideration of social meanings embedded in the applied context (Coupland, 2007; Moore, 2012), we aim to provide a conceptualization that encompasses the essence of Internet slang style specific to the marketing domain while still offering consistency with prior communication style literature. Therefore, in our research, we formally define Internet slang style as consumers’ schematic perception of the characteristics conveyed by the Internet slang expressions adopted in marketing-related contexts (e.g., in an advertisement, on the product package).

As communication style plays as the conceptual basis for Internet slang style, we reviewed all the extracted articles again, focusing on possible dimensions of this scale. Generally, communication style consists of nine dimensions: dominant, dramatic, contentious, animated, impression-leaving, relaxed, attentive, open, and friendly (Norton, 1978). However, in line with our aim of developing a parsimonious scale specific to the marketing domain, we determined whether some dimensions needed to be summarized or eliminated and whether additional components should be considered. Embedding Internet slang in the marketing context is not the same as communicating with others in everyday interactions. We, therefore, examined the unique features of Internet slang (especially those in the marketing context) to decide on possible dimensions for the conceptualization of Internet slang style.

The item generation procedure followed Churchill’s (1979) scale development. To generate the initial pool of 156 statements, three sources were used: (1) the literature on linguistic style, geographical dialects, and accent characteristics (Cheyne, 1970; Edwards and Jacobsen, 1987; Tsalikis et al., 1992), including Chinese classics; (2) in-depth interviews with consumers to elicit descriptions regarding ISS; and (3) additional search for online resources and Chinese dictionaries.

To eliminate redundancy, five marketing or economics graduate students were recruited to judge the statements after presenting the ISS concept and example. The appropriateness in the marketing context was explicitly listed as a critical evaluative standard. This process reduced the initial 156 statements to only 55 items. Afterward, the authors invited three marketing professors and one linguistic professor to omit to ensure the authority and rationality of the results and further reduced the statements to a manageable set of 35 items.

To identify a comprehensive and representative set of Internet slang, the authors used the ‘‘Network Buzzwords’’ database, a Chinese official website¹ collecting Internet slang sentences voted on for popularity by Internet users. The collection period was from January 12, 2014, to December 24, 2014. The authors gathered 1,241 sentences, to which 280 additional sentences were supplemented after thoroughly reviewing a professional Chinese language journal [“Yao Wen Jiao Zi (咬文嚼字)”] and exploring rankings published by a Chinese online search engine (“Baidu”).

A set of 350 Internet slang sentences was randomly selected from the sentence pool, which was constructed according to the definitive standards in the pilot study and Study 1. Then, these sentences were distributed into 44 groups², and 44 versions of the questionnaire were designed to limit fatigue and boredom. Specifically, each participant was randomly assigned to one version of the questionnaire and indicated their agreement with how each Internet slang sentence reflected the original 35 items generated from Study 1. All 35 items were rated on a five-point Likert scale anchored by 1 (Strongly disagree) and 5 (Strongly agree).

Participants were approached at airports, railway stations, or a university in a major city in southern China. A final set of 443 complete questionnaires was returned and reserved (valid rate = 76.3%, 47% male, 528 adults participated). Of the valid responses, 50.8% of participants were aged between 18 and 22 years, and 54.6% spent 2–4 h online daily.

Because our objective was to identify the dimensionality of ISS from consumers’ perspective and not individual variations in evaluations of each sentence, the average scores of each sentence for each item were computed across participants (each sentence evaluated by an average of 10 participants). The resulting data consisted of 350 cases, each reflecting an Internet slang sentence with average scores on 35 different items.

Iterated-principal-component exploratory factor analyses (EFAs) with direct-oblimin-rotation were conducted. First, item-to-total correlations less than 0.4 were eliminated (resulting in one item being deleted). Then, 22 items were removed individually based on the rotated component matrix and the expected interpretable meanings brought to the structure, as their loading coefficients exceeded 0.4 in more than one factor. The remaining items entered the final factor analysis, presenting a four-factor solution. The Kaiser–Meyer–Olkin value of 0.81 (Bartlett’s test of sphericity: χ² = 1789.55, p < 0.001) and the eigenvalues greater than one both indicated that the analysis was appropriate for the data. The final set of 15 items accounted for 62.37% of the variance, and the Cronbach’s α was 0.89, within the guidelines for scale development (Nunnally, 1978). The scree plot showed a significant dip, confirming the rationality of this four-factor solution. The four factors were labeled amiability (five items, α = 0.84), overtness (five items, α = 0.78), candor (three items, α = 0.63), and harshness (two items, α = 0.66). A summary of the factor analysis is shown in Table 2.

To confirm whether the four-dimension solution was the general structure for ISS, additional research was conducted to collect data from a second independent sample of subjects to examine the structure via a series of confirmatory factor analyses. Similar to the procedure from Study 2, a total of 415 questionnaires were sent to participants with a similar demographic profile.

Finally, 350 useable responses (a response rate of 84.3%) formed the basis for the iterated confirmatory factor analysis models (Table 3): (a) the null model; (b) a one-factor model (all items were loaded on a single factor); (c) a two-factor correlated model (amiability and overtness items were loaded on the same factor, while the remaining items loaded on the other; (d) a four-factor uncorrelated model with items loaded on their respective hypothesized factors; (e) a four-factor correlated model with the same structure as model (d); and (f) a four-factor model with one second-order factor.

Table 3 shows that, compared with the fit results of models (a)–(d), models (e) and (f) presented superior fit results (Kelloway, 1998). Model (e) was constructed as a correlated four-factor model (χ²/df = 2.93, CFI = 0.90, GFI = 0.92, IFI = 0.90, RMSEA = 0.07). Model (f) was to confirm these four factors constituted the higher-order construct ISS (χ²/df = 2.87, CFI = 0.90, GFI = 0.92, IFI = 0.90, RMSEA = 0.07). Closer examination revealed identical results from these two models. First, nearly every indicator’s t value was statistically significant (p < 0.01), and almost all the coefficients exceeded 0.50 (Table 4). Furthermore, the estimates of Cronbach’s α for the ISS four factors based on the second independent sample were similar to those of Study 2: amiability α = 0.83, overtness α = 0.78, candor α = 0.64, and harshness α = 0.74. Finally, composite reliability and AVE indices were computed to justify the scale’s convergent and discriminant validity (Fornell and Larcker, 1981; Table 5). Composite reliability was higher than 0.60, and AVE did not fall below the corresponding pairwise squared correlation coefficients. In general, results supported the stability of the four-factor structure for the scale and suggested good psychometric properties.

Study 4 was an experiment to test how different ISS dimensions induced by slogans help build corresponding brand personality dimensions in advertisements with the same product. Besides, we also wanted to provide further evidence to differentiate it from brand personality in this study.

ISS theoretically differs from brand personality, whose formal definition is “the set of human characteristics associated with a brand” (Aaker, 1997). Brand personality reflects consumers’ generalizable impressions of the brands (Aaker et al., 2004) and is helpful for companies to establish deeper consumer-brand relationships and favorable brand attitudes (Fournier, 1998). Prior research reveals that even subtle marketing cues can influence consumers’ brand personality perception, including visual symmetry (Bajaj and Bond, 2018), haptic product attributes (Ranaweera et al., 2021), and disclosure of the brand’s corporate social responsibility activities (Tarabashkina et al., 2020). More importantly, extending prior findings of linguistic reflexes of personality (Mairesse and Walker, 2011; Kovacs and Kleinbaum, 2020), we propose that brand personality can also be shaped by the specific language features adopted in the advertising. For example, using metaphors in marketing communications leads consumers to perceive the products as more innovative and less socially responsible (Luffarelli et al., 2021).

To examine the effectiveness of ISS, we adopted brand personality as the key criterion variable. For operationalization, we used the brand personality scale developed by Aaker (1997), broadly validated and generalized cross-culturally and widely applied to academic and practical settings, despite slight changes in some countries (Aaker et al., 2001). Precisely, the scale consists of five trait dimensions: sincerity, competence, excitement, ruggedness, and sophistication. Our theoretical model highlighted the trait dimensions of sincerity and competence, which constantly emerge as parts of brand personality in both eastern and western cultures (Aaker et al., 2001). Accordingly, to further understand how brand personality would dictate the kind of language used in advertisements, two ISS dimensions with relatively orthogonal meanings, amiability, and harshness, were selected in advance.

The trait of sincerity captures the extent to which consumers characterize a brand with adjectives such as “warmth,” “cheerful,” and “genuineness,” while the competence dimension is composed of efficient, successful, and confident impressions (Aaker, 1997; Aaker et al., 2001). On the one hand, consumers often form the perception of brand sincerity when a proximal psychological distance is elicited (Hu and Shi, 2020) or when a sense of social belongingness is boosted (Chang et al., 2019). In this vein, an amiable Internet slang slogan generates an atmosphere with original, pleasing, and adorable properties, allowing consumers to feel psychologically close to the brand to establish a sincere brand perception. On the other hand, a harsh slogan would convey a sense of straightforwardness and roughness, which easily leads consumers to induce associations of confidence and capability from the brand to embed in the advertisement, thereby building perceptions of brand competence (Chen, 2021). Therefore, we formally hypothesized that:

Although Internet slang is recognized as a novel approach for corporations to conduct marketing practices, some gaps still exist in understanding its characteristics distinct from those of the standard language. In closing these gaps, this research attempts to link three literature streams–the communication literature, psychology literature, and marketing literature–by introducing a new construct that evaluates the unique characteristic associated with Internet slang as an emerging language variant. In this vein, we make three significant theoretical contributions.

First, some previous studies on Internet slang have focused only on one particular feature (e.g., humor or novelty) (e.g., Liu et al., 2019). We develop a theoretically well-grounded and comprehensive conceptualization of the Internet slang style as a multi-dimensional construct based on the schema theory and the communication style literature. Specifically, Internet slang style comprises the four dimensions of amiability, overtness, candor, and harshness. Accordingly, a scale of Internet slang style is constructed to provide researchers with an instrument for contextualized Internet slang research in the marketing domain. Investigators can select particular dimensions or comprehensively use our scale in their study. As such, the understanding of Internet slang and its influence in the marketing context can be deepened, and the results can be comparable.

Second, in addition to empirical results that consistently show high convergent and discriminant validity of the ISS scale (studies 1–3), we demonstrate that our scale differs from brand personality and brand attitudes under experimental conditions (Study 4). Specifically, we empirically show that Internet slang style dimensions impact consumers’ perception of brand personalities, but exert no influence on their brand attitudes. These results provide support for recent conceptual propositions that brand personality can be constructed through the linguistic identity of a brand (Carnevale et al., 2017) and extend the classic concept of “linguistic styles as the individual difference (Pennebaker and King, 1999)” to a marketing context. Meanwhile, these empirical findings also respond to the call of emphasis on exploring whether social media context (i.e., Internet slang) would infuse a collectively-derived meaning into brands (i.e., brand personality) (Carnevale et al., 2017).

Third, this research explores an interdisciplinary topic by examining Internet slang and its influence on marketing communications. In such a field that bridges linguistics and marketing theories, most research focuses on the code-switching effect (e.g., Lin and Wang, 2016; Ahn et al., 2017), or impacts of concrete linguistic elements (e.g., pronouns, phonetics, rhyme) (e.g., Hung and Guan, 2020; Liebrecht et al., 2021). Few studies have examined Internet slang individually as an essential phenomenon. This paper extends this research field by establishing a typology of the prominent perceptual characteristics manifested in Internet slang as a unique language variety.

Marketing practitioners already acknowledge the importance of Internet slang by extensively integrating popular slags into their advertisements. Therefore, our research helps managers to gain insights into applying Internet slang in four ways. First, a refined definition of Internet slang style provides practitioners an objective recognition and shared understanding of the unique but common perceptual characteristics that Internet slang would generate among consumers. This would help establish an essential foundation for managers who consider employing Internet slang in marketing communications, and avoid possible risks of subjective judgment of Internet slang.

Second, we conceptualize and operationalize ISS as comprising four dimensions (i.e., amiability, overtness, candor, and harshness). This expanded conceptualization of the Internet slang style shows that solely associating Internet slang with separated features such as novelty, youth, or interestingness is inadequate to marketing practitioners (e.g., Crystal, 2006; Liu et al., 2019). Study 2 reveals that although amiability contributed to the highest variance, overtness, candor, and harshness each accounted for approximately 10% of the variance in the EFA. Therefore, companies should consider all four dimensions and their possible influences in their decisions to use Internet slang in their marketing communications. For instance, when adopting concise or efficient Internet slang in an advertising context, companies should realize that the slang statement might convey sharpness and even imply non-deference.

Third, we provide an easy-to-apply scale that consists of only 15 clear items for companies to measure Internet slang styles. As such, this instrument enables practitioners to predict consumers’ conception of particular Internet slang statements before launching an advertisement or promotional activity. In addition, with a large body of Internet slang available for marketing use, practitioners can establish a corpus in which all the Internet slang phrases or sentences are categorized by specific ISS dimensions that are evaluated and determined in large-scale consumer surveys beforehand. Such a corpus would allow companies to choose appropriate slang embedded in the advertisements to activate particular associations among consumers.

Finally, we find evidence that different Internet slang style dimensions correspond to differential brand personalities (Study 4). As such, Internet slang could serve as a means for building a particular brand personality. Therefore, companies should align their use of Internet slang statements with branding decisions. Strategic promotional activities that include Internet slang should not violate the brand’s predetermined personality. In turn, when launching a new brand or new product in the market, marketing managers can also take advantage of the readily available associations of certain popular Internet slang statements to construct a particular brand personality in a much easier manner.

This research still contains some limitations, which also suggest avenues for future research. First, the external validity of the ISS scale and related findings might be limited because all the studies were conducted in China. Although our conceptualization encompasses general theories of psychology, communication, and marketing that involve no country-specific factors, replication studies in other countries are required in the future. Second, Internet slang is context-specific (which is ignored in the current paper). Therefore, it would be helpful to account for the interactional roles of product categories, launching platforms, and target audiences when examining the downstream effect of ISS. Third, the relationships between ISS dimensions and brand personality dimensions require more theoretical development and clarification. For example, what would the impact of overtness and candor be on brand personality dimensions? Would amiability and harshness still influence consumers’ perception of the other two brand personality dimensions? Finally, the nomological validity of the ISS scale should be further tested. This research only identifies brand personality as a possible outcome, but finds no relationship between ISS and brand attitude. Future research should reveal relationships between ISS dimensions and other consequential variables (e.g., word-of-mouth intentions) and try to identify possible antecedents that help build different ISS dimensions.