Abstract
Contemporary research on the social side of cultural taste highlights a notable trend of cultural omnivorous taste among socially privileged groups. However, social stratification of taste encompasses different cultural domains, such as music, film or literature. But are omnivores in music also more likely to be omnivores in film and literary taste, and otherwise? Motivated to contribute to the debates on comparability in omnivore studies, this study tests the assumption of overlap of the omnivorous taste, using the KuBiPaD I survey data from Germany. Employing latent profile analysis, this study offers empirical insights revealing a partial overlap. The results show that it is common for omnivores in one domain to be omnivores as well as paucivores in other domains. The co-occurrence of omnivorousness in one domain and univorousness in another one is rare. When comparing the social stratification of omnivorous classes, three domains demonstrate dissimilarities. In the end, the overlap hypothesis is only partially supported, which encourages future research to select cultural domains to study omnivorousness more thoughtfully.
Introduction
Contemporary studies have provided compelling evidence of the social stratification of cultural taste along the omnivore-univore coordinates (Johnston et al., 2019). The departure point of this theoretical thinking refers to Peterson’s (1992) thesis. Analyzing data from the United States, a move from highbrow to omnivorous taste patterns among individuals in high social positions became evident, whereas those in lower positions tended to embrace univore taste (Peterson and Kern, 1996). These results have stimulated a substantial body of empirical research in the realm of cultural omnivorousness characterized by conceptual flexibility, methodological diversity and, as a consequence, contradictory findings (de Vries and Reeves, 2022).
Studies in this field are criticized for applying ‘consistently inconsistent’ theoretical and methodological approaches (Hazır and Warde, 2015; Peterson, 2005). One of the most prominent examples of inconsistency touches upon the lack of consensus regarding the conceptualization, operationalization and measurement of cultural omnivorousness. Frequently, omnivore–univore studies are conducted in different cultural domains – music (Daenekindt and Roose, 2014; Kunißen et al., 2018; Rossman and Peterson, 2015), literature (Purhonen et al., 2010; Voronin, 2024), and film (Hanchard et al., 2019; Weingartner, 2021). Scholars also tend to combine data from multiple fields (de Vries and Reeves, 2022; Warde and Gayo-Cal, 2009) or focus on participation in a range of diverse cultural activities (Katz-Gerro and Jæger, 2013; Sintas and Álvarez, 2002).
Given the lack of consensus, comparative studies of omnivorousness are frequently based on the measurement of cultural stratification through different cultural forms (Peterson, 2005). Following this, Hazir and Warde (2015: 80) raise several substantive questions that should be addressed to understand the incommensurable nature of omnivore studies, one of which is framed as: ‘Given the profuseness of cultural practices, which cultural domains and items are important?’ The variety of approaches raises a conceptual question: in which domains of cultural consumption or taste should we study omnivorousness, and what role does it play? Are patterns of omnivorousness overlapping across different domains, meaning that those who exhibit omnivorous taste in one domain (e.g. music) are also likely to do so in other domains (e.g. literature and film)? This study aims to investigate the measurement of cultural omnivorousness across cultural domains and assess whether the measurement of it through preferences in music, literature, and film plays a significant role. The selection of music, film and literature as domains is motivated by their wide accessibility and ubiquitous representation in prior studies of cultural stratification.
Building upon the recent survey ‘Cultural Education and Cultural Participation in Germany (KuBiPaD I)’ (Otte et al., 2022), I analyze taste within these three domains using preferences for a rich set of genres. To model audience segments and identify omnivore patterns, I employ latent class/profile analysis (LCA/LPA). Subsequently, I compare the composition of omnivore classes across the domains of music, film, and literature. For visualizing the findings, Sankey diagrams are presented. Finally, I evaluate whether the composition of these classes differs in terms of socioeconomic characteristics. After that, the main findings are discussed outlining potential limitations and further perspectives.
The concept of cultural omnivorousness and its ambiguity
Social patterning of taste, as described by Bourdieu (1984), refers to the long-established homology concept: the hierarchical nature of social positions is reflected in the tendency to like certain genres or cultural objects in an exclusive way. Focusing on cultural capital and habitus, Bourdieu is concerned with individuals’ educational attainment and early experiences connected to cultural and general education that contribute to the development of certain taste patterns (Bourdieu, 1984: 13) – e.g., elite or highbrow taste for holders of high educational qualification (Bourdieu, 1984: 20) and popular or lowbrow taste for lower educated. An appreciation of highbrow products requires certain aesthetic competencies and intellectual resources, while lowbrow categories are seen as simple and unpretentious. Individuals in different socio-economic positions tend to draw symbolic boundaries in taste to perform the social distinction (Bourdieu, 1984) as well as judge the social positions of others relying on their cultural repertoire. As shown by prior studies, cultural activities and tastes still carry different status recognitions, with both lower and upper segments recognizing and adhering to these distinctions (Jæger and Larsen, 2024; Jæger et al., 2023).
Relying on Bourdieu’s idea, Peterson and Kern (1996: 904) revised the social mapping of taste and documented ‘strong support for the shift from snobbishness to omnivorousness’, signifying an openness of highbrow consumers to appreciate a variety of cultural repertoire, traditionally assigned to lowbrow or middlebrow categories. Similarly, Peterson and Simkus (1992) found no evidence of the exclusive inclination to the highbrow culture among high-status individuals. These findings led to the development of the omnivore-univore thesis. Since then, the term ‘omnivore’ has been used to describe the openness to appreciate diverse cultural objects and taste patterns that cross symbolic boundaries, while the term ‘univore’ indicates exclusive, predominantly popular or lowbrow, taste. In-between patterns can be referred to as ‘paucivores’ (neither omnivore nor univore).
The original findings have been challenged by reanalyzing the same data by Brisson (2019). He pointed out biased selection of cultural items for analysis, unreliable criteria for capturing the crossing of symbolic boundaries and weak correlation between emotional (liking) and behavioral (doing) dimensions. Consequently, a tendency toward omnivorousness may be a methodological artifact (Brisson, 2019; Rossman and Peterson, 2015).
The actual sense of omnivorousness, its conceptualization and operationalization are far from the consensus in contemporary studies (Hazır and Warde, 2015; Johnston et al., 2019). Studies have revealed different audience structures and omnivorous profiles. Disentangling the omnivorous pattern in music, Savage and Gayo (2011) defined such listeners as rather musical ‘experts’ who took leading positions in the musical field, where the rethinking of the boundaries of musical genres took place. Although omnivorousness is related to openness and tolerance, it is only one potential manifestation of them (Roose, van Eijck, and Lievens 2012: 508). Omnivorousness can take different meanings and unfold in different taste patterns, as universal logic of differentiation hardly exists (Villarroya and Llopis-Goig, 2021).
The lack of consensus has led to mixed results in comparative research on omnivorousness (Peterson, 2005). To provide conceptual clarity, de Vries and Reeves (2022) proposed a division into two versions of the omnivore-univore thesis: weak and strong interpretation. This division reflects the ground of the separation between omnivorousness by volume and composition, as suggested by Warde and Gayo-Cal (2009). The weak interpretation suggests that upper social groups are more ‘culturally engaged’ than lower ones (de Vries and Reeves, 2022: 293). The strong interpretation suggests that this engagement includes crossing the boundaries between upper- and lower-level cultures and avoiding class-based exclusion (de Vries and Reeves, 2022: 293–295). This distinction adds conceptual clarity to the omnivore-univore thesis and addresses previous concerns about its conceptual flexibility.
Another pitfall inherent in comparative studies lies in the implication of diverse potential cultural indexes of status, for example, in music, films, books, visual arts, food, and so on (Peterson, 2005: 266–268). The assumption behind this is that the inclination toward tolerance of cultural taste operates on the level of dispositions (habitus), structuring tastes in a similar way in across domains and producing consistent cultural practices within those domains (Bourdieu, 1990). Assuming a close link between social positions and cultural taste, omnivorous dispositions are visible using different cultural forms to an approximately similar extent, so that omnivores in one cultural domain tend to show extensive taste in other domains. For example, Purhonen et al. (2010) showed a strong correlation between omnivorousness in music and literature by volume, suggesting partial overlap and homology (Purhonen et al., 2010: 280). However, the analysis of the overlap was not the central point of the study and relied on additive indices to measure omnivorousness that do not show evidence of the distinctiveness of the omnivorous patterns in each domain. Systematic comparisons of omnivorous taste across domains remain scarce.
A lack of time may challenge the homology assumption between cultural domains, as extensive engagement in multiple fields requires substantial time resources. Simultaneously, different cultural domains vary in terms of time demands. Consequently, consumers may be forced to be selective in time-demanding domains, thereby limiting the diversity of their cultural repertoires. For example, listening to music usually takes less time than watching films, which in turn is less time-demanding than reading books. This suggests that developing omnivorous dispositions in music may be less challenging in terms of time constraints compared to literature or film. However, the time constraint is mainly relevant to consumption acts rather than preferences, as individuals may express preferences even with limited consumption.
Notably, different cultural domains entail distinct levels of access that extend beyond time demands. Feder (2023) suggests conceptualizing access through four layers: rights (formal barriers), opportunity (physical possibility), participation (individual barriers), and reception (ability to appreciate, often shaped by cultural capital). For instance, listening to music online can be considered one of the most accessible cultural activities in terms of opportunity and participation. However, it can involve constraints at the level of reception, as illustrated by the cultural stratification of music genres. In contrast, engagement with literature might require higher levels of intellectual skills, making the participation barrier more prominent, compared to the music domain.
The literature on omnivorousness covers studies that measure taste through both practices and preferences in multiple domains (de Vries and Reeves, 2022; Roose et al., 2012; Warde et al., 2008), capturing a more detailed pattern that goes beyond one cultural form. However, general population surveys rarely measure cultural taste across multiple domains, leading researchers to assume homology and rely on single-domain measures to study omnivorousness (Kunißen et al., 2018; Leguina, Widdop, and Tampubolon, 2016; Voronin, 2022). This study addresses the potential differences across three selected domains used in previous studies and characterized by different degrees of time devotion – music, film, and literature – and tests whether omnivorous patterns are universal and consistently visible.
Data, operationalization, and methodology
Data
This article is built upon the secondary data analysis of the population-representative survey ‘Cultural Education and Cultural Participation in Germany (KuBiPaD I)’ (Otte et al., 2022). The survey investigates patterns of cultural participation and taste in Germany. In particular, preferences for a set of diverse music, film, and literary genres are measured. The fieldwork lasted from July to December 2018. The survey was conducted in the form of computer-assisted face-to-face interviews (CAPI) and employed a multistage probability sampling procedure (Otte et al., 2022). The final sample consists of 2592 respondents. Basic descriptive statistics of employed socio-demographic variables is provided in Appendix 2. Details about the survey’s methodology as well as extensive descriptive statistics are available in the methodology report (Prussog-Wagner and Sandbrink, 2019).
Operationalization
Taste is measured via preferences for a list of genres (see Appendix 1) using a 5-point scale variable (Otte et al., 2022). The list was provided to respondents on a separate sheet of paper. Respondents select whether they do not know, strongly like, like, neither like nor dislike, dislike or strongly dislike each genre. While questions about music and film tastes were provided to all respondents, those pertaining to book preferences were exclusively presented to individuals who had engaged in reading books or listening to audiobooks within the past year. Although the measurement of taste is limited to genre categories, the rich number of different genres allows identifying a variety of taste patterns.
Methodology
The current studies on cultural stratification and differentiation involve modeling audience segments that share distinct characteristics. One of the most widespread approaches is to use latent class analysis (LCA) – a finite mixture modeling driven by data to identify groups with the most similar item response patterns (Masyn, 2013). If indicators have a continuous rather than categorical nature, latent profile analysis (LPA) is applied.
LCA/LPA is a statistical method aimed at dealing with unobserved heterogeneity by allocating respondents into latent groups (classes) based on their responses to observed variables (Hagenaars and McCutcheon, 2002; Weller et al., 2020). The goal is to identify a model with the fewest latent classes that adequately fit the data (Muthén and Muthén, 2000: 883). The number of classes is determined by the researcher using a stepwise technique (class enumeration) to find the best-fit model (Muthén and Muthén, 2000; Nylund-Gibson and Choi, 2018). After specifying the number of classes, the results include individual posterior probabilities of belonging to each class, the most likely class membership and size of classes. Since Lazarsfeld (1950) introduced this method to the academic community, it has been evolving and developing in different fields.
LCA/LPA is widely used in contemporary studies of cultural audience segments. For example, Alvarez Sintas & Álvarez (2002) analyzed cultural lifestyles in Spain and distinguished four classes: the one with a lack of cultural activity, popular, highbrow, and omnivore classes. Olivos and Wang (2022) relied on the indicators of cultural participation in Chile and classified audience into omnivore, middlebrow, cinephile and passive profiles. Coulangeon (2015) modeled music consumption patterns, proposing six classes, including french pop univore, pop, selective omnivorous, rock, youth pop and extensive omnivores classes. Music preferences were also analyzed in Germany by Kunißen et al. (2018). Scholars used survey data to build six audience segments called omnivore, shallow entertainment (paucivore), excitement entertainment (univore), integration (paucivore), self-actualization (paucivore) and harmony (univore) classes. These examples illustrate the extensive use of LCA/LPA in research on cultural consumption and tastes.
In order to investigate the structure of taste, I begin the analysis by running models of latent profiles for up to eight classes separately in each cultural domain – music, film and literature, — using non-weighted data. 1 Preferences for genres are used as indicators of class belonging. These variables are considered to be continuous variables (5-point scale).
To minimize misclassification errors, multicollinear taste measures within each domain are addressed by eliminating one variable from strongly correlated pairs (> 0.5), unless removal would compromise conceptually significant dimensions (Sinha, Calfee, and Delucchi, 2021). In the domain of literature, preferences for thrillers (highly correlated with crime), children’s and youth books (with fairy tales and legends), classical and modern literature to 1945 (with sophisticated contemporary literature since 1945) and fantasy (with science-fiction) are excluded. However, drama and poems, as distinct literary forms, are retained. In films, fantasy and action are omitted due to their high correlation with science-fiction. In addition, preference for film d’auteur is not considered due to high rate of missing values. Literary adaptations are also excluded due to interpretive ambiguity, as literary adaptations can encompass various styles. In music, preferences for indie rock, alternative, punk (highly correlated with rock and heavy metal, hard rock) and traditional German folk (with German schlager) are removed. Another closely linked pair pertains to classical music and opera; however, both variables are retained due to their well-established highbrow status (Peterson and Kern, 1996) and different social determinants (Ho et al., 2021).
To address the skewed distribution of variables, a maximum likelihood estimator with robust standard errors (MLR) is applied. A numerical integration algorithm is selected. The results are discussed in terms of similarities and differences across domains. Cases with missings on all taste variables in one domain are not considered. The analysis is conducted in Mplus software (Muthén and Muthén, 2017).
To identify the minimum sufficient number of classes, I rely on diverse statistical criteria: Akaike information criteria (AIC), Bayesian information criteria (BIC), sample-size adjusted BIC (SABIC), Lo-Mendell-Rubin adjusted likelihood-ratio test (LMR), and parametric bootstrapped likelihood ratio test (BLRT) (Nylund et al., 2007; Sinha et al., 2021). The decision is complimented with qualitative criteria such as an interpretative value, the appropriateness of class size, and consistency with previous research. BLRT, BIC and SABIC are found to outperform other indicators in big samples (Nylund et al., 2007; Sinha et al., 2021; Yang, 2006). In models with a very large degree of separation, LMR can also be an accurate criteria (Tein et al., 2013).
After determining the optimal number of latent classes, I interpret audience divisions based on observed patterns. To identify omnivorous patterns, I use two main criteria: (1) the presence of the largest volume of taste compared to other classes and (2) the absence of a consistent exclusion of other taste patterns. Key indicators include the average conditional means/median, the number of genres with a conditional mean > 3, the number of genres with a conditional mean < 2, and the comparison of conditional means across classes. Respondents are assigned to the most likely class based on the posterior probabilities (means) of liking music, film and literary genres. I then compare class compositions and test for overlapping omnivorous characteristics using Sankey diagrams (Naqvi, 2023 [2022]), followed by a description of the socio-demographic structure of omnivore classes.
Results
Descriptive statistics
The tastes in music, film, and literature vary in the population (as illustrated in Figure 1). While the absolute majority show preferences for pop and rock music, less than one-quarter of the population is fond of heavy metal, indie rock, traditional music or opera. Classical music and musicals are well regarded as more than 40% show such preferences. Other genres are located in between.
Tastes in music, books and films, weighted, KuBiPaD I.
Regarding film taste, documentaries and comedies are the most preferred genres. They are followed by other highly liked genres: action, crime and thrillers, historical films and classics of film history, which are preferred by more than one-half of the population. The popularity of horrors and homeland-films is the lowest. The film genres are in general more favorable compared to the music or literature genres. For example, eight out of 17 genres are rated by more than three scale points on average.
The majority of readers have a fondness for crime novels, non-fiction and technical literature. Such genres as humorous literature, thrillers and historical novels are also characterized by good popularity. The least preferred are horror novels and comics.
The detailed summary statistics on tastes are presented in Appendix 2, Tables 8–10.
Deciding on the number of classes
In the first step, I run LPA separately for two to eight classes in each domain. Fit indicators are provided in Appendix 3. Comparing models that differ by only one latent class, BLRT always supports a better fit of the model with more classes. Adding one class to each model results in decreasing AIC, BIC and BIC-adjusted values which signifies that a model with more classes has a better fit. However, the LMR test identifies the five-class solution as optimal for music, film, and literature.
LMR tests also support an eight-class solution for literature and a seven-class solution for film. However, it provides a too-nuanced separation of diverse paucivore taste patterns. Consequently, the five-class solutions are the most parsimonious. An additional disadvantage of models with more classes is that one of the classes becomes significantly small in size. The entropy coefficients of the five-class models exceed 0.8, signifying good distinctiveness.
Interpreting the latent classes
The domain of music
When comparing the posterior mean preferences for music genres across various classes, distinct differences in taste patterns become evident (Table 1, Model M). I proceed to interpret the meaning of these classes and assign labels to them. The interpretation relies on identifying similar characteristics among the most (dis)liked genres within each class. Where applicable, I draw upon the four-factor (Rentfrow and Gosling, 2003) and the revised five-factor models (Rentfrow et al., 2011) to categorize musical preferences. The analysis reveals one omnivore, three paucivore and two univore classes.
Conditional means to like music genres for the members of the five classes (Model M), N = 2590, KuBiPaD I, non-weighted.
Entropy = 0.827.
The first class (‘Traditional’) displays a distinct inclination toward conventional genres, particularly German schlager and country, folk. Notably, the conditional predicted mean for liking those genres is higher than the average preferences in the sample. Traditional folk music from other cultures and musicals also obtain a relatively high scores. Members of this class show univore traditional taste, rejecting most of the highbrow (e.g. opera), intense (heavy metal) or contemporary (electronic music) genres. The first class stands out as the smallest subgroup, representing 8.8% of the sample.
The second class (‘Intense’) signals a clear orientation on intense music, namely rock and heavy metal, hard rock. Interestingly, they still hold moderate preferences for pop and electronic music. However, their strongest aversion is directed to the sophisticated (e.g. opera) and church music. Around 18.4% of the respondents belong to this class.
Members of the next class ( ‘Traditional & Sophisticated’) combine sophisticated genres such as opera, classical, musical and jazz with traditional or local music. They express strong objections to heavy metal (marginalized intense genre) as well as electronic music and hip hop (contemporary genres). Other genres are rated somewhere in between. Although members moderately like both so-called highbrow and unpretentious traditional genres, they are selective in their choices. This class accounts for 13.3% of the sample.
A strong taste for pop music and moderate preferences for hip hop, electronic, R’n’B, soul and rock are prominent indicators of the second class (‘Popular & Contemporary’). Developing a sightly narrow taste for popular/contemporary music, members unfavor some sophisticated genres (e.g. opera) and marginalized music (e.g. heavy metal). This class is the second largest, comprising 26.1% of the sample.
Omnivore dispositions are evident in the fifth class (‘Omnivore’). For nine out of 14 genres, the expected scale value exceeds three, with no genre rated below two, indicating tolerance for diversity. The mean preference score is the highest one compared to other classes. Hence, the participants of these classes prefer a variety of music and are not strongly discouraged by lowbrow or unpopular genres, displaying openness toward diversity. It is the largest class, representing 33.4% of respondents.
The domain of film
A distinctive characteristic of the film domain is the audience’s greater openness to a much larger number of genres than in the music domain. For example, members of all classes share high posterior probabilities to like historical and period pieces, classics of film history and documentaries. Nonetheless, the classes in this domain display a diverse range of volume and composition in their taste patterns (Table 2, Model F).
Conditional means to like film genres for the members of the five classes (Model F), N = 2589, KuBiPaD I, non-weighted.
Entropy = 0.850.
The members of the first class (‘True comedy-crime’) accumulate a relatively narrow palette of taste. They display moderate preferences for crime and comedy genres – both commonly blended in modern cinematography (IMDb, n.d.). This crossing of genres is appealing, as comedies often lacks intrinsic subject and frequently need to rely on other genres to incorporate humorous elements in their storylines (Leitch, 2002: 266). Other moderately preferred genres are sophisticated products. However, members reject other genres, such as horror, homeland and romance. This class is the second largest, making up 23.4% of the sample.
The second class (‘Emotional & Entertaining’) demonstrates a selective omnivorous pattern. Members prioritize genres with entertaining and emotion-evoking elements as well as unpretentious plots (e.g. comedy, romantic, drama, dance films) but also show a strong preference for historical and period pieces, classics of film history as well as documentaries. However, the latter one is also found to be the most preferred genre in Germany, so it is likely that members of many classes share positive feelings about it. Moreover, documentaries can be classified as a highbrow genre due to their popularity among the highly educated groups (De Rosa and Burgess, 2014), making the fondness for it potentially indicative of social nature. The members are still averse to horror and sci-fi genres, but are characterized by relatively high medial conditional probability across genres. This class makes up 37.5% of the sample.
The third class (‘Imaginary’) is characterized by high conditional probabilities of liking adventure and imaginative genres, e.g. science fiction, horror, crime and thriller. Since comedic elements can be integrated into any genre-specific stories, they are also highly preferred in this class. Genres featuring whiny stories (e.g. dramas and romantic movies) are highly excluded. This class accounts for 10.5% of the sample.
The next class (‘Omnivore’) is the smallest in size, reaching only 6.3%. However, members show true omnivore characteristics, combining popular and sophisticated genres, including highly disliked genres in the population. For 11 genres out of 13, members have considerably high conditional probabilities. The lowest conditional mean refers to the preferences for homeland films and westerns, two genres that are not common in Germany.
The prominent characteristic of the fifth class (‘Sophisticated & Popular’) is the combination of sophisticated, imaginative and popular genres in their repertoire. However, they are less inclusive and more selective than the omnivore class, performing status-based exclusions. For instance, they tend to dislike emotion-evoking (e.g. romantic, drama) and marginalized (e.g. horror) genres. This class makes up 22.4% of the sample.
The domain of literature
Taste in literature displays a division into one univore, two paucivore and two omnivore classes (Table 3, Model B). Notably, nearly all classes show a tendency to like comedy and non-fiction, two of the most popular genres in this domain. However, the taste patterns are more exclusive compared to the domain of film.
Conditional means to like book genres for the members of five classes (Model B), N = 1929, KuBiPaD I, non-weighted.
Entropy = 0.805.
The first class (‘Imaginary & Detective’), encompassing 10.2% of the sample, develops taste for detective (e.g. crime) as well as fantastic and mysterious genres, e.g. science fiction and horror. The members of this class display a profound distaste for highbrow categories (e.g. sophisticated contemporary literature since 1945, drama, poems).
The second class (‘Non-fiction’), which accounts for 12.5% of the sample, is the most univorous one. Conditional means are getting relatively high for a single genre – non-fiction and technical literature, guidebooks. All other genres, particularly the fiction ones, are excluded from the cultural repertoire of the individuals in this class.
The next class (‘True comedy-crime’), which accounts for the second largest segment of the sample (46.2%), develops preferences for crime, historical and comedy products. Imaginary genres, such as science fiction, as well as horror and comics, are highly rejected. This taste pattern in literature mirrors the preferences of the first class in the domain of film.
The fourth class (‘Omnivore’), which is the smallest one (10.0%), shows substantially high conditional means for most of the genres. Only two genres receive a conditional mean lower than the scale middle point. Members of this class tend to like different genres, combining contrasting repertoire.
The fifth class (‘Sophisticated & Popular’) is marked by exceptionally high conditional means for highbrow cultural products, e.g. sophisticated contemporary literature, biographies and memoirs, historical novels, poems and dramas, indicating strong aesthetic competencies. In contrast, marginalized genres such as horrors or science-fiction are usually rejected. Nevertheless, members selectively expand their taste to genres from popular culture (e.g. crime), showing a tendency toward omnivorousness. This class is the second largest one, comprising 21.0% of the sample.
Comparing omnivorousness across domains
Relying on the assigned classes based on posterior probabilities, omnivorous taste across domains exhibit strong partial overlap (see Appendix 2, Table 11 for direct comparison). To visualize multiple flows between omnivore, paucivore and univore classes, I employ Sankey diagrams.
Music and film
Among the members of the omnivore class in music, approximately 49% and 6% end up in selective and true omnivore classes in the domain of film, respectively (Figure 2). Depicting a flow from music to other classes, it becomes evident that a significant share of those classified as omnivores in music belong to paucivore film classes. It is quite rare for omnivores in music to be univore in the domain of film, taking into account the size of the univore class. A big share of omnivores in the film finds themself in the paucivore music classes. Among omnivores of any kind in the field of films, 41% is assigned to the class of omnivores in music. Similarly, the flow from the univore film class to the omnivore music class is infrequent.
Sankey diagram, the domains of music and film, KuBiPaD I.
Music and literature
Among selective omnivores in literature, around 56% display omnivorousness in music, whereas in the group of true omnivores, 37% are assigned to omnivore music class (Figure 3). The rest of omnivores in literature mainly belong to paucivore classes in music. Among the members of the omnivore class in music, approximately 41% are assigned to either one omnivore class in the domain literature. Notably, the combination of univorousness in one domain and omnivorousness in another is a rare occurrence.
Sankey diagram, the domains of music and literature, KuBiPaD I.
Film and literature
As illustrated in Figure 4, among true omnivores in literature, 39% are assigned to any omnivore class in film. Among selective omnivores in literature, 64% are also classified as selective or true omnivores in film. Among true omnivores in film, 57% are in omnivore classes in literature. Among selective omnivores, only 36% are classified as either selective or true omnivores in literature. It is common for omnivores in film to be paucivores in literature.
Sankey diagram, the domains of film and literature, KuBiPaD I.
The socio-demographic characteristics of the most omnivorous classes
The main assumption of the omnivore-univore thesis refers to the cultural stratification of taste (Peterson and Kern, 1996). As a final step, I estimate the social characteristics of omnivore classes. The corresponding descriptive analysis is available in Appendix 4. I consider several indicators of socio-economic positions: higher education, father’s higher education, age, gender. For a detailed comparison, I run fractional logistic regressions to predict the posterior probability of assignment to each omnivore class (Figure 5, full results in Appendix 5). Respondents under 25 are filtered out to avoid bias related to incomplete higher education.
Fractional logistic regressions to predict the probability of becoming a member of omnivore classes, average marginal effects, KuBiPaD I, non-weighted.
The results indicate notable similarities between omnivores in music and selective omnivores in literature and film, alongside clear contrasts with other classes. Higher education significantly increases the probability of being an omnivore in music and a selective omnivore in literature and film, but decreases the one for true omnivores in these latter domains. The father’s higher education only slightly increases the probability of being an omnivore in music and a selective omnivore in literature.
Next, women have higher probabilities of being omnivores in music or selective omnivores in literature and film, whereas no significant gender differences are found for true omnivores in film and literature. Moving to the age patterns, a higher probability of being omnivores in music is among middle-aged segments, true omnivores in film and literature – among younger adults and selective omnivores in these domains – among older adults. Generally, the results suggest distinctive social meanings attached to different forms of omnivorousness across domains.
Conclusion and discussion
This study has investigated the assumption of overlap of the omnivorous taste at the level of genres between cultural domains, meaning that if individuals are omnivores in one domain (e.g. music), they are also omnivores in another cultural domain (e.g. literature or film). Relying on the KuBiPaD I survey data from Germany, one of the key findings is only partial support for the assumption of universal omnivorous dispositions across the domains.
Certainly, achieving complete overlap is highly improbable due to the differing number of stratified genres across the three domains. In the music domain, such genres as classical, opera, jazz or musicals are regarded to be sophisticated and highbrow genres (Rentfrow et al., 2011). Conversely, in the film domain, such cultural stratification between genres are less common, with several exceptions (e.g. documentaries, see De Rosa and Burgess, 2014). Thus, broadening one’s taste in the music domain frequently involves embracing genres with varying social recognition, while in film, this pertains to the increased volume within the same level of cultural stratification. Consequently, taste in film is generally more likely to be extensive than taste in music. In the domain of literature, there is already a social difference between readers and non-readers (Bennett et al., 2009: 110) that may weaken the genre hierarchy.
Furthermore, consumers can face time constraints even on the level of preferences, limiting their ability to engage equally in all cultural forms. Consequently, being omnivore in one domain leads to partial selectivity in other domains, since omnivorousness in one (especially time-demanding) domain already requires sufficient resources. Thereby, it is common for omnivores in one domain to be paucivores in other domains. Following this result, I advocate for the alternative formulation of the assumption, so that omnivorousness in one domain and univorousness in another is rarely, if ever, combined. However, it is common to observe individuals who are omnivores in one domain and paucivores in another. This result can serve as an additional explanation for incommensurable findings by previous omnivore-univore studies.
It is noteworthy that the omnivorous taste pattern and its differences from the paucivore classes are much more distinctive in the domain of music compared to film or literature. The music domain, therefore, offers a particularly appealing environment for omnivore studies. In terms of socio-demographics, music omnivores share significant similarities in education and gender with only selective omnivores in the (less accessible) domains of film and literature, while the most omnivorous classes in these latter domains differ greatly. This intriguing discrepancy can be attributed to the different logic of cultural stratification in these domains. Current omnivore–univore studies that strictly assume homology and direct transferability of findings between cultural domains might be misleading.
Future research should carefully select domains to study cultural stratification of taste at the level of genres since the results also show that it is advantageous to go beyond one domain. While omnivore studies in music are well-established, studying more than one domain can bring new and unique information in terms of the social stratification of omnivorous taste patterns. The prominence of this pattern in the domain of literature should not be overlooked by future research. Previous studies have also supported the importance of social stratification of cultural taste in music and literature (Bennett et al., 2009; Purhonen et al., 2010).
The current study is not without limitations. First, the measures used in this article are limited to the self-assessment across a set of various genres. However, genres can represent misleading characteristics for measuring taste (Brisson, 2023; Brisson and Bianchi, 2022). For example, there is a lack of evidence for a common general identification of genres (Brisson, 2023). Also, modern cultural products can go beyond a single genre, combining characteristics from multiple styles (e.g. crime-comedy films or folk-rock music). Second, audience segmentation can take place within genres rather than between them (Nault et al., 2021). The logic of cultural legitimacy can function distinctively at different levels of culture – the level of genres and objects in those genres (Childress et al., 2021). Third, this study focuses on omnivorousness within cultural domains and does not consider the one that spans various cultural forms. Fourth, numerous studies employed cultural practices rather than preferences to measure cultural omnivorousness (e.g. Katz-Gerro and Jæger, 2013; Roose et al., 2012; Sintas and Álvarez, 2002). This study does not address this perspective.
The methodology bears extra limitations. Relying on the LPA, the analysis treated variables to be independent within classes. The violation of this assumption may lead to the misspecification of models. As a second critical point, to evaluate the size of the overlap between domains, individuals were assigned to one class following posterior probabilities. However, this approach accounts neither for probabilities of belonging to other classes nor for the error terms.
Despite limitations, this study contributes to the debates on comparability in omnivore studies by examining the overlap of cultural omnivorousness across different domains. The extensive coverage of objective and subjective genres in the questionnaire allowed for the consistent identification of distinct audience segments, including the omnivore ones. The findings suggest partial support for the overlap of omnivorousness across domains. However, the complementarity of paucivorousness and omnivorousness between domains is notable. Studying cultural taste at the level of genres in multiple domains brings new information that remains unobserved when solely working in a single domain.
Footnotes
Appendix 1
Appendix 2
Appendix 3
Appendix 4
Appendix 5
Author’s Note
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Open Access funding enabled and organized by Projekt DEAL.