Recalibrating the scope of scholarly publishing: A modest step in a vast decolonization process

https://publons.com/publon/10.1162/qss_a_00228

In 2018, Philip G. Altbach and Hans de Wit, two leading scholars of higher education at Boston College, published “Too Much Academic Research Is Being Published” in University World News. In making their case, Altbach and de Wit point out that although “no one knows how many scientific journals there are … several estimates point to around 30,000” (Altbach & de Wit, 2018)¹. Finding the number excessive, they declare “a crisis in academic publishing” involving “too much pressure on top journals” and “the rise of predatory journals and publishers that publish low or marginal quality research.” They recommend steps be taken to reduce the amount of research published². The analysis in this article not only challenges such journal estimates but calls for a recognition of the Global South’s research commitment and engagement. The source of this analysis is 25,671 journals that are using the open-source editorial management and publishing platform Open Journals Systems (OJS), first released by the Public Knowledge Project (PKP) in 2002³. Our association with PKP, as authors of this paper, is noted in the Competing Interests statement at the end of this paper.

Earlier studies of what we are calling JUOJS (journals using OJS) located 9,828 journals in 2015 (Alperin, Stranack, & Garnett, 2016). A survey of 2,114 of JUOJS staff members found 97.8% of them to be open access (OA), with 13.6% utilizing Article Processing Charges (APC) to provide OA (Alperin, Stranack, & Hanson, 2017; see also Edgar & Willinsky, 2010). These results suggest that 84.2% of the JUOJS are “OA diamond” (which charge neither authors nor readers for access). This result can be compared to a more recent study which placed JUOJS at 60% of OA diamond journals, among “at least 17,000, but likely up to 29,000, OA diamond journals” (Bosman, Frantsvåg et al., 2021, p. 93). Of relevance to this paper’s section on indexing, Bosman and coauthors conclude that “even though they are well embedded in academic structures, OA diamond journals struggle to be properly integrated into the ecosystem of scholarly publications” (p. 84).

This “struggle to be properly integrated” has a postcolonial history that Altbach (1981) addressed in the “University as center and periphery” some four decades ago. The “peripheral universities" of the “Third World,” he noted, are “basically distributors of knowledge … dependent on the central institutions for innovation and for direction.” Their “special problems with developing research capabilities” include “few outlets for scholarship, creative writing, and research reports by Third World intellectuals” as “publishers are largely uninterested in Third World authors.”⁴ Altbach saw this powerful “colonial educational heritage” as a form of “neocolonialism” not easily overcome: “The odds against a fully autonomous and effective publishing enterprise being developed in the [African] continent are high” (Altbach, 1976, p. 461). The universities there, he further warns, “not only have to confront the reality of the historically and economically based power of the industrialized nations but also … the widespread desire by the industrial nations to maintain their dominant positions” (Altbach, 1981)⁵.

The pervasiveness of this “intellectual imperialism”—as Syed Hussein Alatas at the University of Malaya has framed this center–periphery pattern, beginning in the late 1960s (Alatas, 2000)—has also given rise to an “academic dependency” theory of relevance to this paper⁶. Its leading scholar, Syed Farid Alatas, at the National University of Singapore and the University of Malaya, writes of multiple dependencies afflicting Global South researchers, including a “dependency on recognition” that leads to “many scholars [being] torn between satisfying the requirements of publishing in high ranked ‘international’ journals, particularly those listed in the [Web of Science], and publishing locally in their own languages” (Alatas, 2022, p. 20). The JUOJS attest to a certain break with the recognition dependency, just as these journals address what the Kenyan novelist and essayist Ngũgĩ wa Thiong’o identified as a “need to move the center from its assumed location in the West to a multiplicity of spheres in all cultures of the world,” which Thiong’o framed as part of a “vast decolonization process” (Thiong’o, 1993, pp. xiv, 3)⁷.

In the first decade of the 21st century, Ezra Ondari-Okemwa, an information science scholar at Machakos University, noted how the region’s shortfalls in “the technological capability to support electronic knowledge transfer and scholarly publishing” were matched by the regional publications’ “lack [of] visibility” (Ondari-Okemwa, 2007). What this means for researchers is summed up by PLOS publisher Alison Muddit: “The need to achieve credibility and visibility within the global research system” means researchers in the Global South choosing “between a desire to strengthen local platforms and outlets that better serve local needs and feeling pulled to ‘play the game’ in which norms have been set by the Global North” (Muddit, 2020; see also Alperin & Rozemblum, 2017).

Against this array of challenges, this paper seeks to bring to light how editors, publishers, and researchers around the world, and especially in the Global South, have taken advantage of an open-source publishing platform to create, against Altbach’s odds, “an effective scholarly publishing enterprise.” The scale of this enterprise will be mapped across the principal dimensions of scholarly publishing: country and region; language; discipline; and indexing. By counting those whose research has too often been discounted, this paper demonstrates the extent to which scholarly publishing in the Global South is shifting the assumed center to “a multiplicity of spheres.” This is enriching what is thought of as “the literature” across the disciplines, although its full generative potential for research is only beginning to be recognized and realized.

The JUOJS employ an optional PKP Beacon, released with the software in 2015, that enables PKP to notify OJS users of security patches and other software upgrades⁸. The beacon also transmits indexing information about the journal, including the journal’s title, ISSN, number of items published, titles, and abstracts. This study employed a standard for “active” journals of five items a year, set by the Directory of Open Access Journals (2020)⁹. The resulting publicly available data set for 2020 was 25,671 journals or 36.5% of the 70,214 OJS beacons operating at the time (Khanna, Raoni et al., 2021). The growth in journals using OJS over the previous decade has shown signs of increasing (Figure 1). The journals in the data set averaged 38.1 items in 2020, for 996,000 items, and these journals have published a total 5.4 million items since their inception. An OJS installation, from which multiple journals can be generated, is host to 2.62 active journals on average.

A journal’s country of origin goes through a three-step process. First, we used the ISSN (International Standard Serial Number) of a journal to query the LOC ISSN lookup API for location, which provides us with a MARC code. This MARC code can then be mapped to a country (Raoni, 2021). MARC codes that map to within-country regions or states are handled manually to reflect corresponding countries. Second, we extract the top-level domains from the journal URLs and map them to their corresponding ISO 3166-1 alpha-2 country codes. Third, we geolocate the IP address of the journal host using a weekly updated copy of the GeoLite2 Free Geolocation Data¹⁰.

The beacon data revealed JUOJS in 136 countries a country was not identified for 20 journals. Indonesia leads with 11,535 (45.0%) journals, followed by Brazil with 2,653 (10.3%)¹¹. The top 10 countries account for 74.6% of the total (Figure 2)¹². Yet the global distribution of the journals is widespread with pockets of more intense use (Figure 3).

The economic distribution of JUOJS was analyzed using the four World Bank income groups (Figure 4). By gross national income (GNI) per capita, 81.1% of the JUOJS are found in the World Bank’s middle-income countries, with the 27 low-income countries accounting for only 53 JUOJS (0.2%). This amounts to 81.6% of the journals coming from countries associated, by national income, with the Global South (Haug, 2021). From a geopolitical perspective, using the five United Nations Regional Groups (https://www.un.org/dgacm/en/content/regional-groups), 78.1% of the JUOJS are located in three UN regional groups associated with the Global South (Dados & Connell, 2012): Asia-Pacific states (54.7%), Latin American and Caribbean states (21%), and African states (2.4%) (Figure 5).

To determine the languages of publication for the JUOJS, we ran Google’s Compact Language Detector v3 (gcld3) (https://github.com/google/cld3) on the 100 most recent articles in the ISSN-verified subset of JUOJS. Gcld3 is a freely available pretrained neural classifier and has a built-in flag for language predictions that indicates when the predicted language classification exceeds an optimal probability threshold (https://github.com/google/cld3/blob/6b15e31d6d2840e5f34b1696f635cada9aff8f71/src/nnet_language_identifier.h#L111-L115). Only reliable language classifications were retained for each journal. Then, the most frequent language classification for each journal was designated as the journal’s primary language of publishing. If gcld3 classified a given journal’s article abstracts as being in multiple languages (i.e., with at least five articles per language), then the journal was designated as multilingual. Finally, a variety of heuristic checks (https://nbviewer.org/github/j-a-ball/ojs-global/blob/main/notebooks/JNB3_Languages.ipynb#Heuristic-approach-to-language-verification:) were applied to verify the primary language of publishing for each journal, for example, by checking predicted languages against top-level domains, which resulted in the semimanual correction of 478 journals. Two relatively underresourced languages, Balochi and Faroese, are currently unsupported by gcld3 and were instead discovered by searching top-level domains.

Language detection software indicated that research was being published in 60 languages by the JUOJS, for which an example journal was found by manually verifying each language (Table 1). The language classifications for all ISSN-verified JUOJS (n = 22,561), were shared with Google Scholar to extend the scope of its indexing of journals to additional languages¹³.

Close to half of the JUOJS published articles primarily in English (49.7%), with the top 10 languages accounting for 97.0% of the titles (Figure 6). By comparison, Scopus has publications in 40 languages, with the proportion of documents (rather than journals) in English at 92.6%, followed by 2.8% in Chinese and 1.3%% in Spanish (Vera-Baceta, Thelwall, & Kousha, 2019).

The high level of linguistic diversity among JUOJS calls into question the common linguistic assumption that “English is the language of science,” as a recent headline in Nature read (Elnathan, 2021). The reign of English in research and scholarship only began to take hold in the latter half of the 20th century, after sharing the spotlight earlier with French and German; and although Latin served as the language of learning in the medieval West, the influx of learning from the Islamic world in Arabic, through the 12th-century translation movement, made all the difference (Huttner-Koros, 2015; Willinsky, 2017, pp. 117–152).

Today, English's hegemony is facing challenges in the name of “bibliodiversity,” which Monica Berger, New York City College of Technology, places at the center of decolonizing the OA movement in research (Berger, 2021). A position statement on this question of research and language has been issued by the Chilean scholar Federico Navarro, La Universidad de O’Higgins, and 11 other faculty from around the globe, that “challenge(s) assumptions made about the use of English as a ‘lingua franca’ in scientific-academic contexts,” by identifying its deleterious effects on knowledge production, while arguing for “why we, as research communities in different fields and regions, should use multiple languages and varieties to promote transnational dialogue in scientific-academic contexts” (Navarro, Lillis et al., 2022). Similarly, Suresh Canagarajah, at Penn State, expresses his hope that bibliodiversity amounts to a “gradual chipping away at power, to decolonize writing and scholarship as diverse communicative spaces, which will complement the parallel activism for large scale institutional and policy changes” (Canagarajah, 2022, p. 120). The commitment to such change within scholarly communication is reflected in the international array of signatories to the 2019 Helsinki Initiative on Multilingualism (https://www.helsinki-initiative.org/en).

Although the research hegemony of English is present among the JUOJS, another linguistic feature of this collection is that 48.3% publish in more than one language (Figure 7). The relationship among the four leading languages can be tracked among mono- and multilingual journals (Figure 8). Although monolingual English language journals represent the largest segment at 6,651 titles (indicated by a single dot under the first column), the notable bibliodiversity among the JUOJS appears in the next column, with Indonesian–English bilingual journals (4,431 titles) and Portuguese–Spanish–English trilingual journals (550 titles). It is also worth noting that the use of English may be overestimated, as it is common for titles and abstracts, on which the language analysis relies, to be translated from other languages into English. As well, articles do, on occasion, employ multiple languages in their text, for example, Spanish and Castilian, or Bahasa Indonesia and Javanese.

To assess JUOJS’ research coverage, the title and abstract of the five most recent articles published in the ISSN-verified journals were first concatenated, then passed as inputs to a neural field of study classifier trained on English-language data (Weber, Kranzlmüller et al., 2020). The use of this classifier required the translation of all titles and abstracts into English, for which we used the Ubiquitous Knowledge Project’s EasyNMT (https://github.com/UKPLab/EasyNMT). The outputs of the classifier are distributions of likelihood over all classes, meaning that it supports multidisciplinary classification. But for ease of reporting, each journal was assigned the single most probable field of study label according to the Australian and New Zealand Standard Research Classification (ANZSRC; https://www.abs.gov.au/statistics/classifications/australian-and-new-zealand-standard-research-classification-anzsrc/latest-release). This schema was preferred over others because of the availability of the data set (https://zenodo.org/record/3490460) used to train the neural classifier, examples in which were labeled according to ANZSRC (Weber et al., 2020). ANZSRC fields map to the OECD’s Fields of Science and Technology (FOS), but differ from those embodied by the National Science Foundation’s seven research area directorates (https://www.nsf.gov/about/research_areas.jsp). Differences are mainly found in STEM fields, especially in the pure sciences. The classifier mitigates this problem of comparison by assigning basic labels such as Chemical and Physical Sciences, which the NSF classification then groups into broader research fields.

The disciplinary distribution of journals reveals that social science research is the leading area among journals, although the sciences, technologies, engineering, and mathematics (STEM) fields—led by the medical and health sciences, followed by engineering and technology—also play a major role by journal count (Figure 10). The STEM presence is notable because of the expectation is that research conducted beyond the center will likely be concerned with topics of cultural diversity (Figure 9). Where the social sciences and humanities are assumed to have a local significance that can be well served by researchers on site, the “universality” of the sciences frees them up for gravitating toward the center, with these results suggesting that there is a more globally distributed basis for research in the STEM fields than is commonly assumed.

In comparing the JUOJS disciplinary breakdown to the profile found in Scopus (with an overlap of less than 4% between them; see below), one finds a mixed picture. On the one hand, Scopus has a higher proportion of titles in mathematics and medicine and health sciences, and the set of JUOJS have a higher proportion in education and language, communication, and culture (Table 2). Yet in other disciplinary areas the journal coverage of the disciplines is roughly comparable between the two JUOJS and Scopus sets, such as economics, engineering, and computer science. One can imagine more advanced citational studies to determine the extent to which journals in these two sets draw on each other in their shared disciplinary interests.

The visibility and circulation of the JUOJS were assessed by determining the presence of these journals in 11 research indexes, directories, and lists. The analysis of the Web of Science Citation Reports and Cabell’s Predatory Reports involved drawing on Stanford University Library licensing agreements, with Cabell’s kindly determining the JUOJS in its database, based on its use of our data set. EBSCOHost, Scopus, OpenAlex, DOAJ, and Beall’s List (2021) make their respective journal lists publicly available. Bianca Kramer (2022) ascertained the number of JUOJS in ROAD, and SerpApi generously assisted in assessing Google Scholar’s indexing of JUOJS. With each of the indexes, we then measured how well JUOJS are represented in the index by assessing overlap based on matched ISSNs and/or matched journal URLs and domain names. The specific matching criteria used for each resource are also reported in Table 3.

As early as 1998, Ann María Cetto and Octavio Alonso-Gamboa, at the National Autonomous University of Mexico (UNAM), raised the marginalization of Global South research in the principal journal indexes (Cetto & Alonso-Gamboa, 1998). They found that the Web of Science’s coverage of scholarly publishing in Latin America and the Caribbean amounted to 0.5% of the journals indexed (Cetto & Alonso-Gamboa, 1998, p. 114). Cetto and Alonso-Gamboa were inspired to join forces with others in the region to form Latindex in 1997, which currently indexes 24,486 journals, as a way of countering what had otherwise been constructed, in Cetto’s words, as “the large periphery of the present world system” (Cetto, 1997, p. 40)¹⁴.

A quarter century after the Cetto and Alonso-Gamboa study, the Web of Science Citation Reports includes 1.2% of the JUOJS (Table 3)¹⁵. The coverage of JUOJS in EBSCO and Scopus is only marginally better. Our concern here is with the extent to which these three databases are commonly assumed to constitute the scientific literature, with researchers turning to them to write literature reviews and analyze characteristics of scholarly communication, for example, in the Global South (e.g., Cortés, Guix, & Carbonell, 2021).

On the other hand, a new generation of indexes, such as Dimensions and OpenAlex, does much better. They cover 54.3% and 63.8% of the JUOJS respectively, by relying on automated aggregation of open resources (such as Microsoft Academic Graph, Crossref, Unpaywall, DOAJ, ORCID, and PubMed). Finally, 88.3% of the JUOJS are indexed in Google Scholar. This is a result, in good part, of its lead engineer Anurag Acharya reaching out repeatedly to PKP, beginning in 2004, to improve its indexing of JUOJS through improvements on both sides. This suggests the deliberate effort needed to start to overcome the legacy of “the large periphery,” as Cetto put it.

Google Scholar data affords further insight into the JUOJS by offering article-level citation data (although by OJS installation domain rather than by journal). The 22,679 JUOJS in Google Scholar (June 2022) were hosted on 8,548 OJS domains, with 2.7 journals on average. Each identified domain was accompanied by the first page of search results, with citation counts for roughly the top 10 articles by citation (Figure 11). This revealed that 34 (0.4%) of the 8,548 journal domains had over 10,000 citations among their first 10 articles; 565 journal domains (6.6%) had over 1,000 citations, and 1,015 journal domains (11.9%) had over 500 citations. But then 547 domains (6.4%) had no citations on the first page¹⁶. The overall mean for the JUOJS’ first page of Google Scholar search results is 358.8 citations (median: 50.0).

As a further but still preliminary analysis of citation practices surrounding JUOJS, the most-cited articles in English, Indonesian, and Portuguese were examined for their citation by JUOJS and other publications. For the leading English-language article, which has 16,432 citations and is published in the Journal of Statistical Software, 95% of the articles citing it come from outside of JUOJS (Table 4). Judging by this one very successful article, the journal is contributing to the broader research literature. This is less the case with the top-cited article in Indonesian, published in Jurnal Ekonomi dan pendidikan, and the top-cited article in Portuguese, published in Outra travessia. These two articles are cited less frequently than the English leader, with 1,321 citations and 1,408 respectively, with those citations coming to a far greater extent from JUOJS. The leading Indonesian article had 73.2% of its citing articles in JUOJS, and for the leading Portuguese article the proportion was 77.7%. This likely reflects the prevalent role played by JUOJS in these two languages, although still roughly one-fifth of the citing articles come from sources other than JUOJS. Much more systematic work is needed on this question of citational dynamics in recalibrating scholarly communication, and this will be the subject of future studies.

As noted above, Latindex is a regional index, which draws on bibliographic data from 17 national resource centers across Latin America, the Caribbean, and the Iberian Peninsula. Of the 24,486 journals indexed in Latindex, 4,208 (17.2%) are JUOJS. On the other hand, fully two-thirds of the JUOJS that are published within this Iberoamerican region are indexed in Latindex.

The Directory of Open Access Journals (DOAJ) is an independent organization supported by libraries, publishers, and others, which accepts journals based on a thorough review of their adherence to scholarly standards. The Directory of Open Access Scholarly Resources is operated by the International Center for the registration of serial publications that manages ISSN (International Standard Serial Number) registration and accepts data from related sources, such as DOAJ, Latindex, and Scopus, which accounts for the greater number of journals it indexes and the corresponding increase in JUOJS that it indexes.

The University of Colorado Denver librarian Jeffrey Beall’s designation of “potential, possible, or probable predatory scholarly OA publisher[s],” for which he began to keep a list in 2008, has played an outsize role in the discounting of Global South journals. Although we have addressed this issue in some detail elsewhere (Khanna & Willinsky, in press), suffice it to state here that that there are two prominent lists of so-called “predatory” journals: Beall managed his publicly available list until 2017, at which time Cabell’s began to offer its subscription service Predatory Reports¹⁷. Beall’s 2017 list contains 366 JUOJS (1.4% of the JUOJS), and Cabell’s Predatory Reports has 237 (1.0), with the two lists sharing 82 JUOJS in common.

The relevance of this analysis for this study has to do with recent and repeated assertions that “predatory journals are a global threat” (Bhagat, 2021; Grudniewicz, Moher et al., 2019; Oviedo-García, 2021; Shrestha, Timsina, & Subedi, 2021; Strong, 2019), with other articles conflating predatory with OA publishing (Krawczyk & Kulczycki, 2021). We are no less opposed, to be clear, to deceptive and dishonest journals, wherever they are published¹⁸. Yet the way this is being framed as a global research threat plays into the peripheralizing world system, bemoaned by Cetto (1997) in the last century, while doing little to address the problem. What might be more effective, and less damaging to global science, is to set a publishing standard for the transparency of journal integrity, for which we’ve begun running initial tests (Willinsky, 2022).

This paper does not answer the question of how many journals there are in the world. It establishes, however, that the several estimates of 30,000, noted above, fall woefully short. Scholarly publishing is taking place on far more of a global scale than is commonly recognized in such estimates, as well as in studies that rely on the Web of Science Citation Index and Scopus. Dismissing this global scale out of hand, as too much research or predatory publishing, seems a less than scientific approach. Rather, there are grounds for beginning to recalibrate scholarly communication on a decidedly expanded, but no less rigorous, scale. This can only help the world take advantage of a yet to be fully grasped growth of the research enterprise. This growth is about more than journal counts. It concerns significant bibliodiversity in scholarly communication reflected, in this case, by the 60 languages in which research is published from 136 countries. It involves changes in scholarly communication economics. There are open-source alternatives to corporate domination of publishing services relied upon by profit and nonprofit publishers¹⁹. The open-source platform investigated in this study can be said, for example, to be driving an OA diamond model of equitable access for authors, users, and readers across the full range of academic disciplines²⁰.

Still, this study represents preliminary rescaling work in scholarly communication. It leaves many questions unanswered on how JUOJS operate within the larger world of research. How are the researchers they publish engaging and contributing to the work of others? The citation patterns need to be mapped to learn about how this work interacts within and beyond this set of journals. Is this work contributing, for example, to the improved efficiency that has been observed in research’s circulation and use²¹? Case studies are called for that investigate JUOJS’ influence on policy-making and professional practice. Such investigations will serve the 2030 goal set by Research4Life, in an encouraging accord among its 160 “publisher partners,” which is to raise the “profile … and support for the Global South research publishing industry” (Research4Life, 2022, p. 26).

Yet there is something more at issue with the global scale of this publishing activity that aligns with the point made by Achille Mbembe at the University of the Witwatersrand, in paying tribute to Ngũgĩ wa Thiong’o’s recentering project of the last century, that “the decolonizing project is back on the agenda worldwide” (Mbembe, 2015, p. 18). Mbembe envisions this agenda transforming universities into “platforms for the redistribution of different kinds of knowledges” (p. 7, original emphasis). This, too, is a part of what we have found here, linguistically and geopolitically. It suggests that another world of scholarly communication—more broadly global, diverse, and inclusive—is not just possible or on the agenda²². It is already under way for the benefit of all.

The authors wish to thank Anurag Acharya, Bianca Kramer, Ron Naoko, Kathy Kerns, and Emma Uhl, as well as SerpApi and Cabell’s, for their assistance in the preparation of this paper.

Saurabh Khanna: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Visualization, Writing—Original draft, Writing—Review & editing. Jon Ball: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Visualization, Writing—Original draft, Writing—Review & editing. Juan Pablo Alperin: Conceptualization, Investigation, Methodology, Resources, Writing—Review & editing. John Willinsky: Conceptualization, Funding acquisition, Methodology, Supervision, Writing—Original draft, Writing—Review & editing.

As acknowledged, the authors are associated with the Public Knowledge Project, developer of Open Journal Systems (JB and SK serve as research associates, and JPA and JW are coscientific directors).

This research was supported by the Graduate School of Education and the Khosla Family Professorship at Stanford University.

The study’s data set is available (Khanna et al., 2021), and Jupyter Notebooks (Ball, 2022; Khanna, 2022) containing the analysis code have been linked to the corresponding figures presenting the data.