Abstract
This study explored trends of diversity in Communication and Media Studies (CMS) between 1974 and 2023 through an automated content analysis of journal publications. An increasing proportion of first-authors from non-WEIRD countries and non-U.S. WEIRD countries was found alongside a declining proportion of U.S.-based scholars (WEIRD refers to “Western, Educated, Industrialized, Rich, and Democratic”). The visibility of women first-authors grew markedly to 54% in the 2020s, although still a bit below the benchmark. A growing prevalence of qualitative methods was found relative to quantitative methods and conceptual work. Interestingly, these trends were intertwined. Qualitative studies were more likely to be authored by women, by scholars from non-U.S. WEIRD countries, and based on non-WEIRD populations. Women authorship was less likely among the increasingly visible non-WEIRD and non-U.S. WEIRD first-authors. Furthermore, less diversity—in particular, geographic and methodological diversity (but not gender diversity)—was found in journals with higher impact-factor.
Keywords
Introduction
A comprehensive scientific understanding of societal issues requires a diversity of perspectives: Without diversity, fields risk biased knowledge production (Matthes, 2024; Trepte & Loths, 2020), reinforcing the status quo, and creating gaps in the literature (Waisbord, 2022, 2025). Inclusive scientific disciplines are better positioned to reduce existing inequalities, by ensuring that minority communities receive similar opportunities as majority groups (Dhanani & Jones, 2017). The field of Communication and Media Studies (abbreviated as CMS, see Demeter, 2019) has increasingly become aware of the need to break down existing geographic, gender, and epistemological hierarchies (Bowman, 2026; Chakravartty et al., 2018; Jansen et al., 2026; Matthes, 2024; Ng et al., 2020; Vaccari, 2022). Even when editors strived for diversity and inclusivity, challenges remained such that they cannot always bring diversity into practice even when trying hard (e.g., Ewoldsen et al., 2023; Stromer-Galley et al., 2023).
The current study investigated the shifting diversity in CMS journal article publications in the 50-year time span between 1974 and 2023. Thereby, we aimed to contribute to the literature with a comprehensive understanding of diversity in CMS by analyzing it along four indicators: (a) geographic representations in terms of where authors’ institutional affiliations were located and (b) where study populations were geographically located; (c) the first-author’s gender and the resulting representation of women versus men authorships; and (d) the empirical research methodologies in CMS by exploring the representation of qualitative versus quantitative methods. Representation of perspectives refers to the extent to which academic publications proportionately mirror the diversity of the global CMS community—taking ICA memberships as a benchmark. Additionally, we explored how diversity was associated with journal impact factor and how intertwined these diversity indicators were; investigating whether diversity on one indicator was associated with diversity on another indicator.
Beyond understanding where the field came from and currently stands, our study provides the necessary points-of-comparison for future analyses to see how CMS succeeded in its mission to advance diversity. Contributing to existing insights, we investigated diversity with a manually validated, automated content analysis of publications from a wider selection of journals across the impact-factor spectrum (including lower and higher impact journals), a longer time frame, and a combination of four diversity indicators. Furthermore, despite recent attention to WEIRD scholarship (Henrich et al., 2010), the geographic diversity of study populations in CMS was rarely analyzed before. Moreover, new insights were yielded by linking methodological diversity with geographic and gender diversity. We also share our data for further analysis and to allow country-level insights (see section:
We focus on published journal articles as the unit-of-analysis, defined as research articles published in peer-reviewed scientific journals; widely regarded as the most visible and institutionally-valued forms of scholarly output, often serving as a measure of academic success and recognition. Journal publications represent knowledge products endorsed by disciplinary gatekeepers, such as editors and reviewers, and reflect what a field deems to be legitimate and high-quality output (Vanderstraeten, 2010). Publications are a tangible indicator of whose voices—whether marked by geographical, gender, or methodological differences—are emphasized or marginalized within the CMS discipline.
The Value-Laden Production of Knowledge in CMS
The notion of value-free knowledge, encapsulated in Max Weber’s concept of Wertfreiheit, posits that knowledge should remain neutral and free of personal or societal values (Weber, 1904). In contrast, the value-laden perspective holds that knowledge, as a human endeavor, is inherently shaped by values, serving as both a product and a medium of power relationships (Kincaid et al., 2007). The value-laden perspective resonates with Foucault’s (1995) power-knowledge framework, where those who define what knowledge is and how it is or should be produced wield significant influence over others; thus, shaping the boundaries of what is considered normative or legitimate knowledge within a field (Foucault, 1980). In the sociology of knowledge, discussions around knowledge production are thus fundamentally linked to social relations through which knowledge is created, legitimized, and disseminated. In other words, knowledge production is not a neutral, disembodied or value-free enterprise, but a value-laden endeavor (Bosch, 2025; Kincaid et al., 2007).
CMS’s evolution also reflects the relations and dependencies that shaped this field. World War II boosted the development of CMS in the U.S.: Substantial U.S. government funding and institutional support brought sociologists and political scientists together to conduct empirical investigations into government communication and propaganda in wartime (Glander, 1999; Simpson, 1994; Wahl-Jorgensen, 2004). Economically, the shift toward knowledge-based economies after WWII gradually redefined knowledge as a commodity (Powell & Snellman, 2004). Along with this, U.S. universities took on a pivotal role as incubators of “brain power” and producers of knowledge products (Bridgman & Willmott, 2006). The rise of knowledge-based economies—driven by U.S. institutions, and fueled by technological advancements and neoliberal reforms—also drove the global expansion of CMS (Waisbord, 2019).
Epistemological Hegemonies and Decolonization
While optimists view globalization as an opportunity to participate in shared spaces of idea exchange, others caution that globalization risks dissolving valuable cultural distinctions: It may cause U.S. or Western-centric norms and epistemologies to increasingly dominate, thereby sidelining the perspectives of other regions unless they fit—or are made to fit—the Western frameworks or standards (Gondwe, 2025; Stein & Andreotti, 2017; Waisbord, 2019).
Such “standardization” has raised concerns around cultural imperialism (Schiller, 1976), neo-colonialism (Nkrumah, 1965), and epistemological hegemonies (Pascale, 2016). The dominance of the quantitative empirical paradigm, often assumed to be a value-free scientific approach (Lang, 2013; Perloff, 2013), is a prime example of such an epistemological hegemony. In the critiques of epistemological hegemonies, many scholars emphasized the importance of “decolonizing the mind” (Wa Thiong’o, 1986), embracing cosmopolitan imagination (Waisbord, 2025), and advocating for diversity in academia (De Sousa Santos, 2007). Diversity—as a concept that can be operationalized, measured and observed—provides valuable insights into the extent of (de)colonization within academia (Albayrak, 2018), beyond being just a branding exercise or hollow metaphor (Bosch, 2025).
Diversity of Perspectives in CMS
Scholars conceptualized and measured diversity in various ways to assess its presence in certain systems, including academia. For instance, Weitzman (1992, p. 64) addressed that it is challenging to define “a meaningful value-of-diversity objective function.” Following up on this, Stirling (2007) proposed a general diversity heuristic applicable to scientific fields. Stirling’s approach captured the combination of three necessary, but distinct basic attributes of diversity: variety, disparity, and balance (Stirling, 2007). Variety refers to the number of categories into which elements of any type of system can be allocated (Stirling, 2007). Disparity reflects the distinguishability of these different types of “things” in a system. And, balance reflects the pattern of allocation across the different distinguishable categories. Balance is, thus, about how much of each distinct category is represented in a system and whether this is proportional to its relative prevalence in the population.
It is important to notice that each attribute requires the other two attributes to be meaningful: Without variety or disparity, balance is irrelevant, and vice-versa. Specifically, diversity can be captured with a higher versus lower variety of categories; for example, regarding geographic diversity, the number of separate countries (according to the UN: 195) versus country categories (in this study three categories: non-WEIRD; non-U.S. WEIRD; the U.S.). Although exploring separate countries allows more granular analyses (i.e., higher variety), it overlooks that many countries are relatively similar (i.e., lower disparity) and that real diversity can only be achieved if distinguishable country-categories are represented proportionally (i.e., in balance); moreover, working with distinguishable categories allows for more generalizable conclusions.
Among the three diversity attributes, balance stands out as the critical lens to empirically examine power dynamics (Stirling, 2007). Balance moves beyond mere presence (variety) and reflects the proportional representation of disparate voices and subjects. Whether balance reflects proportional representation requires a benchmark, for which this study relies on ICA memberships. In an unbalanced situation, certain authors or perspectives are represented more prominently in journal publications than they are present among the members of CMS’s largest international association of scholars; this suggests an uneven dominance by certain groups at the cost of the relative absence of other groups. More balance, thus, indicates that academic output is more proportionately distributed across geographies, gender, and methodological approaches.
Geography, gender, and methodology are this study’s central barometers to operationalize diversity, because of their role as structural indicators of epistemic hegemony within a research field as well as their alignment with decolonization theory’s call to dismantle hierarchies in knowledge production through the geo-body politics of knowledge framework (Mignolo & Escobar, 2010). The geo-politics dimension underscores how knowledge is shaped by geo-historical locations, calling for a delinking from Western hegemonic narratives (Gondwe, 2025; Mignolo & Escobar, 2010). The bio-politics dimension explores the interplay between gender identity and epistemology; emphasizing the need to liberate historically marginalized groups, such as women’s voices, to participate in knowledge production (Mignolo & Escobar, 2010). Additionally, decolonial scholars increasingly critiqued the biases embedded in methods, especially the dominance of quantitative methodologies (Smith, 2021); urging to improve the balance between qualitative and quantitative methods.
Geographic Representation of Authors and Study Populations in CMS
Attention for diversity in CMS has grown in the past decade, amongst others sparked by the #CommunicationSoWhite study of Chakravartty et al. (2018). Indeed, authors from a few countries appear to publish a major share of CMS articles (Trepte & Loths, 2020). This imbalance in favor of Western countries seems prevailing, with a strong influence from the U.S. (De Albuquerque et al., 2020; Demeter, 2019; Trepte & Loths, 2020; Walter et al., 2018; Wiedemann & Meyen, 2016). The dominance of, especially, the Anglo-Saxon countries among authors in CMS is even stronger than in other fields (Demeter, 2019; Guenther & Joubert, 2017; Lauf, 2005). As language is central within CMS, scholars based in an English surrounding—and especially native-English speakers—have a built-in advantage (Dimitrova & Palmer, 2025): They do not have to translate the main elements of their research (Demeter, 2019) nor their own lines of reasoning, which is a barrier for non-native English speakers (Suzina, 2021). Some even speak of linguistic imperialism, as colonialism and education have always been linked (Dimitrova & Palmer, 2025). The business of academic publishing has even further strengthened the push toward one common language as to maximize profit (e.g., through scale and efficiency).
The existing dominance of U.S.-based authors in the literature (Walter et al., 2018) and editorial boards (Asuman et al., 2025; Goyanes & Demeter, 2020) might induce non-U.S. authors to provide more substantive contextualization for their study contexts or to make explicit comparisons to the U.S. context (Chan et al., 2021); thus, reinforcing the (perceived) centrality of the U.S. in CMS. Since many CMS theories were formulated from a Western democratic perspective by Western authors (Miike, 2007), an application to Western contexts—even with research conducted in other geographic areas—is a convenient choice and often a necessity.
Nonetheless, both Trepte and Loths (2020) and Goyanes et al. (2020) found a significantly improved balance in terms of authors’ geographic affiliations. In that regard, especially Asian communication scholarship has increasingly been published (So, 2010). More specifically, Ekdale et al. (2022) found a proportional decrease in the dominance of North American authors between 1990 and 2019, but they still published significantly more articles than scholars residing in any other part of the world. Additionally, Western bias is kept in place through various mechanisms: Articles written by Northern American and European authors were more likely published by higher-ranked journals (Ekdale et al., 2022; Lauf, 2005), whereas scholars from other regions published more regularly in lower-ranked journals (Xu et al., 2024). Even the articles about the Global South or de-Westernization are often authored by scholars from Western institutions (Marques et al., 2025).
Hence, CMS is still insufficiently geographically diverse (Goyanes et al., 2020). Walter et al. (2018) found that by 2016 still 67.4% of work in the Journal of Communication originates from North America; comparable to the 65.7% that Ekdale et al. (2022) found for the Top-10 CMS journals. Thus, while geographic diversity overall increased, the U.S. (and to a lesser extent: Northwestern-Europe) still dominates in flagship journals. Scholars residing in the U.S. seemingly maintained their dominant role in the discipline: Being central actors within ICA (Wiedemann & Meyen, 2016), in ICA fellowships (Park & Barnett, 2024), and taking most places on editorial boards (Asuman et al., 2025; De Albuquerque et al., 2020; Goyanes & Demeter, 2020; Trepte & Loths, 2020). Hence, we treated authors affiliated with U.S. academic institutions as a distinguishable category for geographic diversity.
Next to the U.S., we distinguished non-U.S. WEIRD countries and non-WEIRD countries, where WEIRD refers to “Western, Educated, Industrialized, Rich, and Democratic” (Henrich et al., 2010). Previous studies that looked at geographic distinctions found that classifications based on location alone (e.g., Global North and Global South) were insufficient. For example, European regions outside of Northwestern-Europe—but categorized as Global North—were not among the most represented countries in terms of publication rates (Demeter, 2019): Eastern Europe is more comparable to Global South regions, such as South America or Africa (Demeter, 2018). The WEIRD categorization, instead, foregrounds structural imbalances more comprehensively: Not only where people are geographically located, but also how they are situated in a global system of inequality determines academic power dynamics (Bosch, 2025) and shapes the validation or spread of scholarship. The WEIRD framework captures economic and political power disparities, but also social, educational, and cultural dimensions—and, thereby, allows a comprehensive analysis of the structural forces that influence knowledge production globally. This is important, because differences in resources partly determine the possibility of scholars to receive feedback at conferences, the ability to use specialized software/hardware, research time versus teaching load, or even up-to-date library access (Dimitrova & Palmer, 2025; Gondwe, 2025). The following research question was explored to investigate whether the geographic publication gap has persisted or whether non-U.S. WEIRD and non-WEIRD countries increased their representation:
Geographic Diversity of Study Populations
Developments of geographic diversity regarding study populations have been mostly overlooked, while this is crucial for the generalizability of research findings. Henrich et al. (2010) claimed that most psychology theory is based on research with WEIRD people: 96% of research subjects in the top Psychology journals were drawn from Western industrialized countries (Arnett, 2016), which represent only 12% of the world’s population; underscoring the disproportionate influence of WEIRD contexts. Whether the same holds for CMS and how this developed over time is unknown.
Even scholars in non-WEIRD contexts may opt to study communication phenomena in a WEIRD context. A recent special issue of Journal of Communication, for example, explicitly strived to be more geographically balanced, which mostly succeeded in terms of the authors’ affiliation, but not in terms of study context (Stromer-Galley et al., 2023): Several of the non-U.S. authors still studied the U.S. context. One reason for non-WEIRD scholars to collect data in a WEIRD-context is that reviewers or editors tend to (unconsciously) prioritize studies from their own country (e.g., U.S. reviewers being more positive about studies with data collected in the U.S.; see Gondwe, 2025; Goyanes & Demeter, 2020). With authorship potentially becoming more geographically diverse (Bucy & Evans, 2022; Goyanes et al., 2020; Trepte & Loths, 2020), we also explored the still unknown evolution of study populations’ geographic diversity:
Gender Diversity of Authors in CMS
The widely discussed Matilda Effect explicates the systematic under-recognition, denial of credit, or erasure of contributions made by women in intellectual fields (Rossiter, 1993)—often in favor of attributing achievements to men colleagues. We investigated the gender imbalance between men and women first-authors in CMS. Representation of women scholars is also important as it provides inspiring role models for aspiring academics (Braun et al., 2023).
Previous work found that women scholars experienced structural barriers in CMS (García-Jiménez & Herrero, 2022): Their work received fewer citations than articles by men (Knobloch-Westerwick & Glynn, 2013; Wang et al., 2021), causing a persistent publication-citation gap (Jansen et al., 2026), and in the long-run a lower H-index (Goyanes et al., 2024). Moreover, women’s work is perceived to be of lower quality (Knobloch-Westerwick et al., 2013), even when it was read more often (Rajkó et al., 2023). The underrepresentation of women scholars and their work is intensified by the imbalances inherent in the construction of gender (Eckert & Bachmann, 2021).
Increased representation of women scholars was found in ICA presentations (Braun et al., 2023), across 85 communication journals (Jansen et al., 2026), and in published works of CMS sub-disciplines (Guenther & Joubert, 2017; Taipale & Fortunati, 2014). An investigation of 11 European countries and the U.S. found no improvement of the gender imbalance across countries in the last decade with still significantly more men authors in each country (Rajkó et al., 2023). Given the limited time frame and geographic focus of the existing studies, Rajkó et al. (2023) called for research that extends beyond the Euro-U.S. context and to consider larger time frames. Hence, we explored the following research question:
Arguably, the indicators of diversity (i.e., geographic, gender, methodology) may be mutually dependent. Taking an intersectional perspective (Bachmann & Eckert, 2022), it is important to examine the intersections of geographic and gender balance, because gender as a construct does not operate in isolation. Social identity theory would predict that academic roles (e.g., professors or researchers) are perceived as more masculine in certain regions/cultures; therefore, gender causes more perceived role (in)congruity in specific contexts (Braun et al., 2023). Increasing geographic balance could, thus, potentially go at the cost of gender balance: If women’s scholarship mainly originates from a limited set of countries, the perspective of women could become less prominent with an increasing geographic balance of the field.
Trepte and Loths (2020), for example, found more women authors among U.S. scholars, whereas this was less likely for European and Asian scholars. Braun et al. (2023) confirmed this finding for Asia, whereas Guenther and Joubert (2017) demonstrated that woman-authored studies were less likely from specific European countries (see also Rajkó et al., 2023). To gain more insight into the relationship between geographic and gender balance, we asked:
Diversity of Methods in CMS
Evident from a decades-long discussion about the (dis)advantages of quantitative versus qualitative methods in the social sciences, also in CMS (Lang, 2013; Perloff, 2013), methodological choices are highly consequential in the scientific process. They mirror different ideological and philosophical approaches to the sociology of knowledge—and whether one perceives knowledge to be value-free or value-laden (Lee & Wallerstein, 2004). While quantitative methods, especially experiments, are often regarded as the “holy grail” when it comes to predicting communication phenomena, they can be criticized for their claims of external validity. Qualitative research, instead, helps researchers understand human experiences in-depth. Qualitative and quantitative methods can be mutually beneficial, offering richer understanding of the same phenomenon (Kreps, 2011; Young & Gray, 2013). Yet, the epistemological divide seemingly persists as quantitative and qualitative CMS journals remained distinguishable with limited overlap in editorial boards (De Albuquerque et al., 2020). Also, mixed-methods studies are rather rare in CMS (Kamhawi & Weaver, 2003; Trumbo, 2004).
In the early days of CMS, both empirical paradigms were fairly balanced; yet, quantitative methods have gained greater prominence (from 42% in 1965 to 64% in 1989, see Cooper et al., 1993); although this gap was found to not further grow in later periods (Trumbo, 2004). Thus, quantitative methods seem to lead (Kamhawi & Weaver, 2003; Kim et al., 2017), which may even be stronger in certain outlets (Freelon et al., 2023). For instance, almost 95% of the publications in the Journal of Communication in the 1950s to 83% in 2019 used quantitative methods (Walter et al., 2018). To understand how the representation of empirical paradigms within CMS journals developed, the following research question was examined:
Geographic and methodological diversity may be associated, because U.S.-based scholars are a majority in editorial boards of quantitatively oriented journals (67.5%), but not for qualitatively oriented journals (41.2%, De Albuquerque et al., 2020; see also Freelon et al., 2023). Moreover, quantitative methods may require more resources to obtain sufficient statistical power (i.e., larger samples) or to purchase costly statistical software and hardware (e.g., for computational methods). Goyanes and Demeter (2020), however, did not find a relationship between editorial board diversity and empirical diversity of publications. Therefore, we asked:
Which empirical approaches are more employed by men or women scholars is an open question (Dow & Condit, 2006), because “there is no distinctive feminist method of research” (Bachmann & Eckert, 2022, p. 7). The relationship between methodology and gender was explored through the following research question:
The Impact of Impact-Factor
Previous studies used impact-factor as a selection criterion to decide which journals to include in their research (e.g., Freelon et al., 2023; Günther & Domahidi, 2017; Trepte & Loths, 2020). Hence, studies on academic diversity often concentrated on publications in journals with highest impact-factors. Rajkó et al. (2023) suggested, therefore, that future research should also use the impact-factor as a predictor of diversity in CMS publications. Yet, Jansen et al. (2026) found no difference in the proportion of women first-authorships across levels of journal prestige.
A positive relationship has been found between the proportion of U.S. authors and impact-factor (Ekdale et al., 2022; Lauf, 2005). Goyanes and Demeter (2020) found, in contrast, no relationship between impact-factor and geographic diversity of authors nor data; yet, they demonstrated a negative relationship between impact-factor and diversity overall (Goyanes & Demeter, 2020). Although they have more geographically diverse editorial boards (Asuman et al., 2025; Bowman, 2026), these findings altogether suggest that higher-impact journals may limit, rather than enhance, diversity in the field. To explore this suggestion, we asked:
Benchmarks
There are no conclusive, non-arbitrary benchmarks; this makes it hard to claim when “enough” balance would be achieved. Trepte and Loths (2020) made good steps by comparing publication rates to population size and the estimated number of CMS scholars in a country. Additionally, Freelon et al. (2023) compared references in CMS to ICA memberships. Both studies concluded that it is neither population size nor composition of the academic community, which explain the overrepresentation of certain groups in CMS publications.
In this study, representation regarding geographic diversity of authors and gender of authors was assessed by the extent to which academic publications proportionately mirrored the diversity of the global CMS scholarly community in terms of ICA members. Inspired by earlier work (e.g., Braun et al., 2023; Freelon et al., 2023), we used ICA membership details as the benchmark of how the international academic CMS population would look like. Likely, this is a conservative estimate of the population’s diversity. After all, one could challenge how representative the ICA is of the global academic CMS community; arguably, it already represents a pre-selection of the most successful and affluent academics. While ICA has taken sincere actions to internationalize (Bowman, 2026), the association has also been said to still be a force of Americanization that conserves power structures in the field (Wiedemann & Meyen, 2016). Table 1 presents the benchmarks applied in this study; no such standards are available for methodological diversity (i.e., proportion of quantitative and qualitative studies or scholars).
Benchmarks for Geographic and Gender Diversity Based on ICA Memberships.
Method
Data and Sampling
We concentrated on CMS journals with a generic focus, coined as “general-field journals” by Ewoldsen et al. (2023, p. 535). These are journals that do not solely focus on specific sub-domains; thereby excluding journals, such as Health Communication, Political Communication and Public Relations Review, in which authors from other fields also frequently publish—allowing a cleaner assessment of the CMS field. We included all English-written general-field SSCI-ranked journals (N = 33, see Table 2) in the “Communication” category of Journal Citation Reports of which the published article’s metadata could be downloaded from EBSCO’s “Communication & Mass Media Complete” database by August, 2023. 1 The metadata included the information about journal title, article publication date, article title, article DOI, article abstract, article subject, article language, names of authors, affiliations of authors, and article type. To focus on research publications, we only included articles classified as “Article” in “Academic Journal” with “English language” in EBSCO’s database; thereby, purposefully excluding alternative publication types; that is, book reviews (N = 8,027), editorials (N = 304), essays (N = 261), and non-English texts (N = 857). Duplicates were identified and removed from the dataset by detecting cases that had the same DOI, journal, year, (preprocessed) title and/or first page. Our analysis was restricted to begin in 1974, when a growing number of journals entered the field and became available in the EBSCO database; thereby, avoiding a dataset that was skewed toward the few longer-existing journals, while still allowing an analysis of a 50-year time frame. In total, our sample included N = 32,300 journal publications.
Overview of Journals With Number of Article Publications Included in the Analyses.
To examine the validity of the automated measurements, a manual content analysis was conducted on a subsample of publications that could be matched with the publications of the automated analysis based on DOI (N = 1,093). For this manual content analysis, we selected nine general-field journals with the highest 5-year journal impact-factor in the Journal Citation Reports-ranking 2020. Coding work for the manual content analysis was carried out by three paid research assistants (master-level communication science students from the University of Amsterdam), experienced in reading and analyzing scientific CMS articles. They were rigorously trained in seven sessions to achieve optimal intercoder reliability. A pilot test was conducted on a subsample of research papers randomly selected from the original sample to test for the intercoder reliability (ICR) of measured items. All variables in the manual content analysis scored above Krippendorff’s α > .80 and are thus considered reliable (Riffe et al., 2019, p. 129). The data of the manual content analysis were also used to give additional insights into the geographic representation of study populations and methodologies, because the automated measurement of these variables did not achieve optimal reliability for all categories.
Measurements
Geographic Diversity: First-Authors’ Country of Affiliation
Because an author’s own nationality (or race) is difficult to obtain from a content analysis, we focused on the country of the first-author’s affiliation (i.e., most often university). The manual coding of this variable showed no disagreements between coders (K-α = 1.00). For the automated measurement, we relied on the article’s metadata from the EBSCOhost database. In case the affiliation of a first-author was missing, we employed the CrossRef REST API with Python package habanero to retrieve and supplement authors’ affiliation where possible. Next, we extracted the last elements of the affiliation, as they commonly indicated the associated country (e.g., University of Copenhagen, Denmark). When affiliations were solely mentioned as names of universities or academic institutions, we employed a “country-university” dictionary to identify the corresponding country. In total, 118 countries were identified.
For analytical purposes, this country variable was recoded into one variable with three categories: (0) U.S., (1) non-U.S. WEIRD, and (2) non-WEIRD country of residence. We followed the theoretical categorization of Henrich et al. (2010) to decide which countries were WEIRD: countries with a long democratic tradition of British-descent societies (i.e., Australia, Canada, Ireland, New Zealand, United Kingdom, United States) and those clustered in the northwest of Europe with highest GDP per capita (respectively, Liechtenstein, Luxembourg, Monaco, Norway, Switzerland, Iceland, Denmark, the Netherlands, Sweden, Austria, Finland, Belgium, Germany). All other countries were categorized as non-WEIRD. The automated measurement of authors’ geographic diversity strongly overlapped with manual coding (97.5% agreement; Krippendorff’s α = .95), which demonstrated its reliability.
Geographic Diversity of Study Populations
To retrieve the country of data collection, we followed the procedure set out by Chan et al. (2021): Papers’ abstracts and titles were searched for country keywords that would indicate where a study was conducted. The keywords-list was shared by Chan, which then consisted of 21 countries. We extended the list on two aspects: (a) we included all 249 countries on the International Organization for Standardization ISO-3166 list; (b) we included a wider range of keywords, by including the country name and conjugations of it, synonyms for the country name, the country’s language (when that was unique), and country’s capital and/or biggest city (keyword list is available in the corresponding country dataset on OSF, see Data Availability Statement). Challenging for this measurement was that authors of studies with U.S. data often do not describe this in the abstract or title (Chan et al., 2021; Walter et al., 2018). For a majority of cases, therefore, no country could be detected (53.48%) and only 15.0% of the studies could be traced back to data being collected in the U.S, which obviously was an underestimation.
For the manual validation, the country of the study populations was coded as the country in which the data collection occurred as described in the paper (often in the Abstract, Title, or Method section). To allow for cross-country studies, coders could select multiple countries.
The lack of context description in specific titles or abstracts caused an unacceptably low reliability for the automatic measurement of geographic origin of study populations in general (K-α = 0.19), particularly caused by the problematic automated measurement of U.S. study populations (66.7% agreement; K-α = 0.11). Yet, the automated measurement of data collection in non-WEIRD countries was acceptable (91.5% agreement; K-α = 0.58) just as the detection of data collection in non-U.S. WEIRD countries (90.7% agreement; K-α = 0.52). This confirms that non-U.S. authors are more likely to prominently report geographic details (Chan et al., 2021). The statistical analysis, therefore, concentrated on the trends in the presence of study populations from non-WEIRD and non-U.S. WEIRD countries.
Gender
Gender can be understood as a social construct of practices influenced by culture rather than biological differences alone (Butler, 1990). As authors typically do not explicitly disclose their gender identity in publications—only names are available—our analysis was restricted to a binary gender classification (men/women); we could not capture a more diverse spectrum of gender identities (Jansen et al., 2026, described the same limitation).
We employed CrossRef REST API to retrieve authors’ name when this information was absent in the EBSCO database. The manual annotation of the author’s gender was based on the masculinity or femininity shown in the name (K-α = 0.82). If not entirely clear, coders searched the Internet to find author’s gender (e.g., faculty page, social media profiles, etc.). If it was not possible to find the author’s gender or if an author indicated to identify themselves outside of the man-woman binary, coders chose “other/uncertain” (coded as 2; 2.4% of cases).
For the full dataset, we detected authors’ gender using an automated approach similar to Sebo (2024), who used ChatGPT. However, dealing with copyrighted material this would violate copyright laws and the rights of the original creator. Therefore, we used the locally-run Meta-LlAMA-3-8B-Instruct architecture (Grattafiori et al., 2024), which appeared to even achieve slightly better measurement performance. We selected this model as it enables high-precision analysis without extensive computing resources. Additionally, its specialized instruction-tuning ensures strict compliance with our prompt’s requirements. Prior to full-scale deployment, we conducted an iterative prompt engineering phase on the manually coded subsample to test which prompt achieved the best performance (Supplemental Appendix B1 presents the prompt with optimal performance).
We implemented a probability-based filtering protocol to identify how (un)certain the model was about its prediction. Specifically, we evaluated a range of probability thresholds from .60 to .99; predictions with probabilities below each threshold were coded as “Unknown.” For each of these, the model’s performance was compared with the manual annotation using correlation metrics (see Supplemental Appendix B1). Higher probability thresholds (i.e., more certainty), logically, improved correlations with human annotations, but substantially reduced data coverage (i.e., high precision, low recall). Based on this trade-off, a cutoff threshold of 65% was selected, which provided an optimal balance between data retention of 97% and classification reliability (95.7% agreement; K-α = 0.91 when unknown category was excluded; and 91.8% agreement; α = .84 with unknown category included).
Methodology
Methodology was manually annotated as a nominal variable by coding the abstract, categorizing the primary research method into four groups: 0 = conceptual work without method; 1 = quantitative, 2 = qualitative, 3 = mixed methods (K-α = 0.89). If the article was a conceptual paper or a literature review, it was coded as 0. If quantitative and qualitative research methods were both combined in one paper, we coded it as “mixed” (3).
Using the automated approach analogous to the gender detection procedure, method was automatically classified using the Meta-LlAMA-3-8B-Instruct architecture. Multiple prompts were tested on a manually annotated subsample to refine the model’s classification logic. The final prompt (see Supplemental Appendix B2) used the same 65% cutoff point as for gender, which optimized the balance between reliability and data retention (see Table B1 in Supplemental Appendix for comparison): 81.3% agreement; K-α = 0.70. The mixed-methods category was excluded, because it was too rarely present (1% of cases) to be detected reliably; increasing reliability to K-α = 0.72.
Year of Publishing
Time was operationalized as the year in which an article was published. This study included articles from the year 1974 to the year 2023. To give insight into the long-term dynamics, we ran alternative models with “decade” as an independent variable (0 = 1970s; 1 = 1980s; . . . ; until 5 = 2020s).
Journal Impact-Factor
The 5-Year Impact Factor of journals was obtained from Clarivate’s Journal Citation Reports 2023. We chose the 5-Year Journal Impact Factor, because this provided a more comprehensive and stable measure of journals’ influence over time and smoothened out year-to-year variations (more robust against annual outliers). Since it was not possible to retrieve year-by-year impact-factor scores dating back to 1974, we used the 2023 impact-factor as a reference point that indicated which journals were considered lower- or higher-impact at the time of data collection.
Analyses
Our analyses dealt with dependent variables that were dichotomous or categorical. Therefore, binary and multinomial logistic regression analyses were run to answer the research questions. Since observations were not independent of each other (i.e., different journals have varying diversity scores, see Trepte & Loths, 2020), we ran multi-level versions of these models in StataNow/SE 19.5 (i.e., melogit and xtmlogit): Publications were nested within journals. Moreover, when analyzing one dependent variable (e.g., geographic diversity), we controlled for the influence of the other dependent variables (e.g., gender and method) to capture possible intricacies between the diversity variables and, thereby, also answer RQ3, RQ5, and RQ6. Supplemental Appendix C provides the full regression tables of the analyses. The findings of the multinomial logistic regression analyses, moreover, were replicated by Bayesian multinomial logistic regression models using journal-level random intercepts (see Supplemental Appendix D).
Results
Table 3 presents a decade-by-decade overview of the presence of the four diversity indicators between the 1970s and 2020s; highlighting their long-term developments.
Overview of Diversity Indicators (%) per Decade.
Note. Separate percentages of both the automated and manual analysis are presented for (a) geographic diversity of study populations and (b) methodological diversity, because these measurements were not optimal in the automated content analysis. Different percentages are yielded there, because they inherently dealt with different samples; the automated content analysis with 33 journals of diverse impact factors, whereas the manual content analysis included 9 journals of relatively high impact. For the manual analysis, the 2010s and 2020s are merged, because only publications from the year 2020 could be included in that analysis. The percentage for study populations do not add up to 100%, because of multi-country studies where multiple countries can be present in one publication.
Geographic Diversity of Authors’ Country of Residence (RQ1a)
On average and over the full 50-year period, authors residing in the U.S. published 41.5% of all CMS articles. Additionally, non-U.S. WEIRD authors were responsible for 38.1%; totaling up to 79.5% WEIRD authorship. First-authors from non-WEIRD countries represented 20.5%.
Figure 1 depicts the proportion of authorship per context per year: Both non-WEIRD and non-U.S. WEIRD authors steadily increased at the cost of U.S.-based authors representation—particularly with strong increases around the mid-1980s and the 1990s. When analyzing the trend over time (i.e., year) and controlling for the other diversity indicators, the multilevel multinomial logit model demonstrated that the likelihood of having either a non-WEIRD (R.R.R. factor of 1.05, p < .001) or a non-U.S. WEIRD (R.R.R. = 1.03, p < .001) author increased by a factor 1.05 or respectively factor 1.03 with every year compared to the prominence of U.S.-residing authors (see Table C1 in Supplemental Appendix C).
Predicted geographic representation of first-authors’ country of residence (categorized as U.S. vs. non-U.S. WEIRD vs. non-WEIRD) in CMS publications over time.
In terms of decades, this means that with every decade (i.e., 10-year) the odds of having a non-U.S. WEIRD author increased with factor 1.31, relative to U.S. authors. The presence of non-WEIRD first-authors even increased with a factor 1.57 per decade. Concretely, U.S.-based authorship still represented 90.4% of the publications in the 1970s and this went down to 35.2% in the 2020s. Instead, non-U.S. WEIRD authorship increased from 6.6% in the 1970s to 39.6% in the 2020s, and non-WEIRD-based authorship from 3.0% to 25.3% in the 2020s.
Looking at the most recent years (2020–2023), U.S. authors were no longer overrepresented compared to the benchmark of ICA memberships (35.2% in the 2020s compared to a benchmark of 45.3%), but actually underrepresented. By now, it is non-U.S. WEIRD authors that seemingly are overrepresented; with 39.6% of all CMS publications in the 2020s compared to 24.6% of ICA members. The publications by non-WEIRD authors increased to 25.3% in the 2020s; still not reaching the benchmark of 30.2% in ICA membership.
Zooming in on the specific countries where first-authors resided (see Figure 2 and Supplemental Appendix E, Figures E1), it was clear that the U.S. still held a dominant position with the largest share of publications; although its proportional dominance noticeably declined over time. The other most represented countries were the U.K. and Australia—both Anglo-Saxon and WEIRD. Other non-U.S. WEIRD countries, including the Netherlands, Germany, and Canada, also ranked highly. Within the non-WEIRD regions, the strongest increase was found of scholars residing in China, Spain, Israel, and also Hong Kong, South Korea, and Singapore.
First authors’ country of affiliation over time (proportion).
Impact-Factor and Authors’ Geographic Diversity (RQ7a)
A negative impact was found of impact-factor on authors’ geographic diversity (Supplemental Appendix Table C1). Compared to authors from the U.S., the odds of having a non-U.S. WEIRD author decreased with a non-significant factor 0.84 (p = .125) for every one-point increase in impact-factor (vis-à-vis U.S. authors). Slightly more negatively, and significant, the likelihood of a first-author from a non-WEIRD country decreased with factor 0.80 (p = .015) for every one-point increase in impact-factor.
Thus, overall representation of non-U.S. authors considerably increased; yet, journals with high impact-factor were still more likely to publish work from U.S.-based authors. Figure 3 shows that the predicted likelihood of U.S. authorship is larger in journals with a higher impact-factor. In contrast, authors from non-U.S. WEIRD countries were especially more likely to publish in journals with lower impact-factor compared to the U.S. authors. Authors from non-WEIRD countries were relatively unlikely to publish in all journals, irrespective of impact factor.
Representation of first-authors’ country of residence in journals with comparatively lower (0–2.3), medium (2.4–4.5), or higher (4.6 and above) 5-year Journal Impact-Factor. Cut-off points decided based on terciles (each group of journals represented 33% of publications).
Geographic Diversity of Study Populations (RQ1b)
A multi-level binary logistic regression model (see Supplemental Appendix Table C2) showed that the likelihood of study populations originating from non-WEIRD countries increased every year, O.R. = 1.02, p < .001: This steadily increased from 4.5% in the 1970s to 32.3% in the 2020s. Because the automated measurement was not very precise (e.g., false negatives when authors did not mention country of data collection in the title or abstract; 53.5% of the data), we also inspected the manual content analysis results. The manually annotated data (n.b., fewer journals, all relatively high-impact) replicated a similar pattern for data collected in non-WEIRD countries: Excluding the studies without empirical data, this increased from 6.3% in the 1970s to 20.5% of the data collections in the 2010s. Thus, compared to the benchmark of non-WEIRD ICA memberships (i.e., 30.2%), non-WEIRD study populations were relatively well represented, although less in the smaller pool of manually-annotated high-impact journal articles (20.5%) than in the larger set of automatically-annotated journal articles (32.3%).
The proportion of studies with data from the non-U.S. WEIRD countries was not found to change over time (automated analysis: O.R. = 1.00, p = .119). Since we analyzed the full population of publications, an inspection of the decade-by-decade percentage of publications with non-U.S. WEIRD data is still insightful. It revealed an upward trend: from 4.4% in the 1970s to 15% from the 2000s onwards (stable from then, which explained why no significant linear effect of “year” was detected). The overall lower percentage of non-U.S. WEIRD data (14.6%) than non-WEIRD data (25.4%) also flags that this measurement was probably affected by more false negatives in studies from the WEIRD contexts (i.e., WEIRD authors less likely to prominently report where their data originated from). Exploring the data of the manual analyses, we found that the proportion of non-U.S. WEIRD study populations increased from 3.1% in the 1970s to 23.3% in the 2010s, O.R. = 1.03, p = .003 (significant, with the same set of controls), rather close to the benchmark of 24.6%.
Study populations from the U.S. accounted for 90.6% of the research in the 1970s and this decreased to 65.9% by the 2010s (O.R. = 0.98, p = .020; insight from manual content analysis only, because the automated analysis was unable to detect U.S.-based studies as it is too often left unmentioned when data were sourced from the U.S. context; see also Chan et al., 2021; Walter et al., 2018). Altogether, the prominence of publications with non-WEIRD study populations gradually increased over time and those with non-U.S. WEIRD study populations until the 2000s; this inherently came at the cost of U.S.-based data collections.
Impact-factor and study population’s origin (RQ7b)
The likelihood of having non-WEIRD data decreased with higher impact-factor (O.R. = 0.78, p = .002). A negative association was also found between impact-factor and the likelihood of collecting data in non-U.S. WEIRD countries (O.R. = 0.86, p = .034). Thus, and similar to the authorships, higher impact-factor journals were less likely to feature studies with data collected outside of the U.S.
Gender Diversity (RQ2)
The proportion of women first-authors has grown with every year (O.R. = 1.03, p < .001; see Supplemental Table C3): In terms of decades, women’s authorship increased every 10 years with factor 1.32. Women authored 18.2% of publications in the 1970s, and this increased to 53.9% in the 2020s. Thus, an improved gender balance was found, though still below the 58% to 61% benchmark. Yet, the growth curve of woman first-authorship seemingly continues to rise (Figure 4).
Representation of women versus men first-authors in CMS publications over time.
Impact-Factor and Gender (RQ7c)
Replicating the finding of Jansen et al. (2026), woman first-authorship and journal impact-factor were not significantly associated (O.R. = 0.96, p = .133), after controlling for geographic origin, study method, and time.
Methodological Diversity (RQ4)
When predicting the likelihood of publications with a qualitative method versus a quantitative method (i.e., the reference category), an upward trend was found over time (R.R.R. = 1.01, p = .011; see Supplemental Table C4). The percentage of qualitative studies increased from 6.6% in the 1970s to 52.4% by the 2020s (Table 3). The proportion of quantitative methods also increased over time compared to the “no method” category (year: O.R. = 0.96, p < .001) in the model that controls for geographic origin, gender, impact factor, and journal; yet, the percentage of quantitative studies showed no upward trend in Table 3. This volatility may be attributed to the changing composition of CMS’ publishing field; specifically, the recent entry of journals oriented at qualitative research compared to the historical dominance of quantitative outlets.
Due to the false-negatives (i.e., when authors did not mention method in title or abstract) and, consequently, the limited reliability of these measurements, it may be that the percentage estimates were underestimated. The manually annotated data confirmed the two upward trends vis-à-vis publications without a method. It showed that 6.6% of the studies in the 1970s used a qualitative approach and that this increased to 13.8% in the 2020s. Also for quantitative methods an increase was found: from 33.0% in the 1970s to 52.6% in the 2010s and 2020s; but one must note that this were mainly journals with high-impact factor.
Impact-Factor and Methodology (RQ7d)
The representation of qualitative methods compared to quantitative methods decreased considerably with a higher impact-factor, O.R. = 0.61, p < .001. High-impact factor journals publish more quantitative research.
Geographic Diversity and Gender Diversity of Authors (RQ3)
Negative associations were yielded between geographic diversity and women’s first-authorship: The odds of a non-U.S. WEIRD first-author vis-à-vis a U.S.-based first-author was smaller when the first-author was a woman than a man (R.R.R. = 0.89, p < .001; Supplemental Table C1). Similarly, but even more strongly, non-WEIRD first-authorship was a factor 0.82 less likely when the first-author was a woman compared to work first-authored by a man (R.R.R. = 0.82, p < .001).
Testing the relationship in the opposite direction (i.e., gender as dependent variable; Supplemental Table C3), the odds of having a woman first-author was factor 0.90 less likely (p = .002) when they resided in a non-U.S. WEIRD country compared to first-authors residing in the U.S. Even more strongly again, the odds of a woman first-author were 0.84 lower (p < .001) when the author resided in a non-WEIRD country than U.S.-based authors.
These findings demonstrated that a decreasing representation of U.S.-residing first-authors was negatively associated with the rise of women’s authorship, and vice-versa: More women first-authorship went hand-in-hand with a dampened geographic diversification of non-U.S. authorship. In other words, the rise of woman first-authorship as found regarding RQ2 seems not to be a general trend, but particularly driven by scholarship from the United States. Similarly, the increase of non-U.S.-based first-authors especially pertained to men scholars.
Geographic and Methodological Diversity (RQ5ab)
Studies with a qualitative method (compared to quantitative methodology; independent variable) positively predicted the likelihood of non-U.S. WEIRD authorship (O.R. = 1.28, p < .001) compared to U.S.-residing authors. Whether a study used a quantitative or qualitative method did not significantly predict the likelihood of it being authored by a scholar from a non-WEIRD country (O.R. = 0.91, p = .095) versus a U.S.-based first-author. Authorship from non-U.S. WEIRD countries was, therefore, also more likely to be associated with the use of a qualitative method than non-WEIRD authorship (O.R. = 1.45, p = .002).
Testing this in the opposite direction (method as DV), we confirmed that scholars from non-U.S. WEIRD countries were a factor 1.37 (p < .001; Supplemental Table C4) more likely to employ qualitative methods versus quantitative methods than U.S.-based authors. Yet, qualitative methodology was factor 0.88 (p = .034) less likely among non-WEIRD first-authors than for U.S.-based authors. Thus, non-U.S. WEIRD first-authors were also more likely to employ qualitative methods than authors from non-WEIRD countries (O.R. = 1.55, p < .001). In sum, qualitative methods were most likely authored by non-U.S. WEIRD scholars and least by non-WEIRD scholars.
A somewhat different pattern was found when predicting the geographic origin of study populations (Supplemental Table C2). We found no significant differences in the likelihood of data collection in non-U.S. WEIRD countries between qualitative and quantitative methodology (p = .572). Yet, the use of non-WEIRD data was more likely in qualitative relative to quantitative studies, O.R. = 1.48, p < .001. In short, research with data collected in non-WEIRD countries was more likely to be conducted with qualitative methods, whereas qualitative methods were most likely authored by non-U.S. WEIRD scholars.
Gender and Methodological Diversity (RQ6)
Method (independent variable) and first-author’ gender (dependent variable, Supplemental Table C3) were significantly associated: Studies with a qualitative method (compared to quantitative method) were a factor 1.46 (p < .001) more likely to be first-authored by a woman. This was replicated when predicting the method of a publication (Supplemental Table C4): Qualitative methods (vs. quantitative) were more likely if the first-author was a woman than a man, O.R. = 1.48, p < .001.
Discussion
Summary of Research Findings
In line with a general trend of internationalization of science (Goyanes et al., 2020), CMS has also become more geographically diverse. The dominance of U.S.-based first-authors decreased over the decades. This proportion dropped from 90% of CMS articles to roughly one-third of all articles in the 2020s, resulting even in an underrepresentation of U.S.-based scholars compared to ICA-membership (45%). Currently, U.S.-based authors are on par with authors from non-U.S. WEIRD institutions, who appear to be overrepresented (benchmarked by ICA-membership). Thereby, Western knowledge-based economies remained dominant (Powell & Snellman, 2004) with almost 75% of first-authors from WEIRD countries (adding up non-U.S. WEIRD and U.S. authorship). This study confirms a persisting WEIRDness in the CMS literature, manifesting existing patterns of geo-politics in knowledge production (Mignolo & Escobar, 2010). In other words, the decrease of U.S.-based first-authors has not considerably enriched the field with disparate cultural perspectives; especially scholars based in countries relatively similar to the U.S. (foremost the UK, Australia, Netherlands, Germany, or Canada) saw their publication rates increase. Nevertheless, and contributing to earlier insights on geographic imbalance within CMS (Chan et al., 2021; Demeter, 2019; Freelon et al., 2023; Lauf, 2005; Trepte & Loths, 2020), our findings demonstrated increasing geographic diversity, where the representation of non-WEIRD first-authors considerably increased over the decades: Non-WEIRD authors are now represented closer to the benchmark of ICA memberships, suggesting their increased representation is more than just a sign of tokenism (see Marques et al., 2025).
Study populations also diversified geographically; especially, studies were more frequently conducted with data collection in non-WEIRD contexts. Yet, data are still more often sourced from the U.S. than from any other region (60.9%, and 65.9% including multi-country studies; manual content analysis data), which further manifests geo-political patterns (Mignolo & Escobar, 2010). It can be concluded that also non-U.S. based scholars work with U.S.-based study populations, which reinforces the U.S. as the dominant context for CMS research (Goyanes & Demeter, 2020). Due to peer-review practices, non-U.S. WEIRD and non-WEIRD scholars may feel pressured to study the U.S. context if they want to publish in high-impact journals, because this is where many future reviewers and editorial board members come from (Asuman et al., 2025; De Albuquerque et al., 2020; Marques et al., 2026; Trepte & Loths, 2020).
Gender representation has become more balanced, with a steep increase in women’s authorship amounting to 54% in the 2020s, continuing the upward trend observed by Jansen et al. (2026) and in specialized CMS subfields (Brosius & Haas, 2009; Taipale & Fortunati, 2014), but found to come to a halt by others (Rajkó et al., 2023; Trepte & Loths, 2020). Although the benchmark of ICA memberships (58%–61%) has not been reached yet, our data show a growth curve that predicts women scholarship to still gain more prominence in the years ahead. Compared to neighboring fields, women’s first-authors are already better represented in CMS than, for instance, in political science (see also Goyanes et al., 2020; Teele & Thelen, 2017).
The proportion of qualitative studies has clearly increased over time. This trend likely reflects changes in the journal landscape, with several newer journals—such as Communication, Culture & Critique, Critical Discourse Studies, Journal of African Media Studies—having a strong qualitative orientation. In contrast, longer-established journals like Journal of Communication, Communication Research, and Human Communication Research maintain a more quantitative profile. This difference also helps explain why the large-scale automated analysis yielded different patterns than the smaller-scale manual analysis, because the latter focused primarily on these high-impact, quantitatively oriented journals. Even within that subset, however, there is a noticeable—though more modest—increase in qualitative research. Overall, whether quantitative research remains the dominant methodology in CMS (Kamhawi & Weaver, 2003; Kim et al., 2017) depends largely on which journals are included in the analysis.
Intertwined Trends of Diversity
Although we found an upward trend for all indicators of diversity, our results also showed that these trends were intertwined. Specifically, geographic and gender diversity were negatively associated. The increase of scholarship from especially non-WEIRD and also non-U.S. WEIRD authors occurred particularly among men first-authors. In other words, the spectacular growth of woman first-authors appeared to be mainly driven by publications from U.S. authors, which aligns with Trepte and Loths (2020) findings.
Whereas these two trends seemed to dampen each other, methodological diversity was found to flourish more among studies with more diverse features. Regarding gender, women first-authors were more likely to publish qualitative studies. Moreover, a publication was more likely to employ qualitative methods when authored by non-U.S. WEIRD scholars. Research with non-WEIRD based study populations was also more likely to be conducted with qualitative methods. This provides evidence for the claim that de-Westernization may promote alternative epistemologies (Marques et al., 2025). Beyond separate epistemologies, it may be (financial) circumstances that steer non-WEIRD scholarship to qualitative methods; quantitative methods often require large(r) budgets to obtain sufficiently-powered sample sizes or to purchase expensive software.
Impact-Factor and Diversity
Interestingly, no specific gender barrier was found in higher-impact journals, as woman first-authorship and journal impact-factor were unassociated, corresponding with Jansen et al.’s (2026) finding. However, impact-factor was negatively associated with the likelihood of, especially, non-WEIRD authorship. High-impact journals, instead, were more likely to publish research by U.S.-residing first-authors (see also Ekdale et al., 2022; Walter et al., 2018) and less likely to feature studies with data collection in non-WEIRD and non-U.S. WEIRD countries. Furthermore, publications in higher impact-factor journals were less likely to employ qualitative than quantitative methods. These findings can be interpreted as signs of manifesting epistemological and institutional hegemonies in CMS via the flagship outlets (Glander, 1999; Wahl-Jorgensen, 2004). Altogether, lower impact-factor journals offer the space and possibilities for authors and research that are typically less well represented in CMS, which gives these outlets a crucial role in the field to further promote diversity.
Limitations and Next Steps
Geographic diversity was measured by the country of first-authors’ affiliation. This does, however, not reflect the full identity of scholars in terms of personal, ethnic, or cultural citizenship: There are many non-native English scholars with a non-WEIRD background that bring their experiences to departments in the WEIRD context. They, inherently, will also be confronting challenges that have not been recognized in this analysis. Thereby, the current approach may actually provide a conservative estimate of the geographic diversity of the field; WEIRD institutions will be more heterogeneous (i.e., internal diversity) than presented here.
Our analyses of geographic and gender diversity focused on first-authors. Although this differs by field, first-authors in CMS are generally assumed to have contributed the most to a publication and are most prominent, especially since APA referencing of three authors or more is shortened to “et al.” (Braun et al., 2023; Freelon et al., 2023). Hence, the first-author position is most crucial when exploring the visibility and diversity trends in the field. Yet, future studies could extend our findings by moving beyond first-authors to consider entire author teams and their geographic or gender composition. Such work could open up questions, such as how author teams in CMS have diversified or become more homogenous (see Torre et al., 2025), and how such diversification influences the geographic representation of study populations or methodological approaches. For example, the co-authors of non-WEIRD first-authors might still reside in the U.S., through which the U.S. dominance could still be kept in place.
We have consciously opted for the WEIRD categorization of countries in our study (Henrich et al., 2010)—it extends beyond where countries are located and classifies countries on whether they are relatively similar on multiple indicators: cultural tradition, educational levels, industrialization, affluency, and democratic standards. Yet, the WEIRD classification may evoke the impression that very diverse societies (e.g., the Nordic-European ones and the Anglo-Saxon countries) can be easily lumped together. After all, “WEIRDness” is used as a fixed category rather than a continuum, which it arguably is. Besides, obscuring this between-country diversity—necessary to draw generalizable conclusions—the WEIRD concept inherently has normative assumptions and, ironically, comes from (and is often used in) Western academia, which may subtly reproduce the same ethnocentric perspective that it critiques.
Gender diversity was operationalized by coding women and men first-authors, where this could be automatedly recognized based on first name. This approach could not include identities outside the gender binary (e.g., non-binary or transgender people) and, thereby, inevitably overlooked important nuances in this respect (i.e., 2% of all ICA members identified as non-cisgender; Freelon et al., 2023). Future research ideally deals with this more inclusively.
Conclusion
So, has CMS become more diverse over the last five decades? When understanding diversity as a multidimensional and interrelated concept (Stirling, 2007), the answer to the above question is a tentative yes; but the devil is in the details, or rather, in the intertwined observations (Bachmann & Eckert, 2022). In terms of variety, we observed a clear increase in the number of countries where publications originated from as well as the methods that were employed. Regarding disparity, we found distinguishable differences in CMS regarding geographic authorship diversity (i.e., non-WEIRD vs. non-U.S. WEIRD vs. the U.S. affiliations), gender, and a divide between quantitative and qualitative research designs. These translate into an increasingly balanced pattern regarding the first-author’s gender, a less persistent dominance of positivist approaches, but also an enduring imbalance regarding the dominance of WEIRD regions. Given that these dimensions of diversity constitute one another (Stirling, 2007), we conclude that CMS has become more diverse over time—and, arguably will continue to do so.
A more diverse CMS field will help to develop theory that better describes, explains, and predicts real-world phenomena beyond the WEIRD context. The feminist movement in CMS has made major steps to diversify the field (Dow & Condit, 2006), and a greater openness and welcoming attitude to a more diverse palette of perspectives will prepare the community to study contemporary phenomena and deal with grand societal challenges. For example, phenomena such as social media trolls or the role of fake news in elections (Rossini, 2023) had already been studied in less represented contexts, such as Brazil, the Philippines, and India, before they challenged the Western democracies (and became dominant topics in our field). Had this non-WEIRD work reached a wider international audience earlier, then the WEIRD world could have been better prepared for the misinformation crises that it is now confronted with.
Supplemental Material
sj-docx-1-crx-10.1177_00936502261454813 – Supplemental material for Fifty Years of Diversity in Communication and Media Studies (1974–2023): The Intertwined Trends in Geographic, Gender, and Methodological Representation
Supplemental material, sj-docx-1-crx-10.1177_00936502261454813 for Fifty Years of Diversity in Communication and Media Studies (1974–2023): The Intertwined Trends in Geographic, Gender, and Methodological Representation by Mark Boukes, Irina Lock and Shujun Liu in Communication Research
Footnotes
Acknowledgements
We thank Nikki P. Dekker and Soohyun Bae for their valuable contributions to the first version of this manuscript, as well as for their efforts in the manual annotation process that they conducted together with Artemis Tsoulos and Nathalie Koubayová.
Ethical Considerations
The study received ethical approval from the University of Amsterdam.
Consent to Participate
This was a study without collecting human data, so no informed consent applied.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Dutch Research Council (NWO) with a Vidi grant (project number: VI.Vidi.211.101) awarded to Dr Mark Boukes.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Open Practices
On OSF (
), we share the article-level dataset (with copyrighted materials removed) and a spreadsheet with country-level indicators of diversity for every decade (1970s to 2020s) to allow further analysis and/or to gain specific country-level insights.
There, we also share an appendix with (a) details on the manual content analysis; (b) measurements of gender, methods, and country classification; (c) full multilevel logistic regression tables; (d) Bayesian multinomial logistic regression models; and (e) visualizations of first-authors’ country of affiliation. See: .
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
