What is local research? Towards a multidimensional framework linking theory and methods

This is a preprint. It is undergoing peer review. The final published version may differ. 1 What is local research? Towards a 1 multidimensional framework linking 2 theory and methods 3 4 Victoria Di Césare1* and Nicolas Robinson-Garcia1 5 6 1Unit for Computational Humanities and Social Sciences (U-CHASS), 7 Department of Information and Communication, University of Granada (Spain) 8 *Corresponding author. Email: [email protected] 9 10 Abstract 11 In this research article, we propose a framework of local research, a concept of growing importance due 12 to its far-reaching implications for public policy. Despite being increasingly mentioned in scientific 13 publications and policy documents, there seems to be little agreement on what the concept means and 14 on which phenomena it covers. A clear understanding of it is crucial for informed decision-making 15 when setting research agendas, allocating funds, and evaluating and rewarding scientists. Our goal is to 16 determine how the concept of local research relates to its measurement, highlighting the nuances 17 involved in its theoretical depiction and the risks of applying different empirical methods 18 indiscriminately across regional and disciplinary contexts. We first explore various ideas and 19 approaches to local research found in the literature. Then, we synthesize them into three dimensions – 20 locally situated, informed, and relevant research – and test six approaches – Toponyms, Languages, 21 Authors, Databases, References, and Citations – in 10 million publications from the Dimensions 22 database. Our findings suggest divergences across regional and disciplinary settings. The operational 23 approaches to identify local research operate differently across both the disciplinary landscape and the 24 Global North-South divide. In addition, we observe no complete alignment between theory and practice. 25 This article highlights the complex, multifaceted nature of local research. We propose a framework to 26 facilitate the study of its different dimensions and their intersections, to help mitigate many of the 27 barriers current research evaluation systems face when assessing and supporting local research. 28 29 Keywords: local research; scientific knowledge; scientific policy; research evaluation; scientometrics. 30 This is a preprint. It is undergoing peer review. The final published version may differ. 2 Introduction 31 Local knowledge is central for the development and implementation of efficient policies, which 32 adapt and consider the conditions and needs of different social groups (Stiglitz 1999). This is especially 33 relevant in peripheral communities, where mainstream knowledge, structured around Global North 34 norms and values, may inadequately adapt to local contexts (Cancino et al. 2024; Marginson & Xu 35 2021). Hence, scientific policy awareness is most needed in order to foster and prioritize local research, 36 ensuring a better distribution of resources (Miguel et al. 2015). However, despite the great interest 37 surrounding it, the linkage between how local research is defined and measured across fields of science 38 is unclear. It is therefore essential to grasp the concept thoroughly in order to make informed decisions 39 when setting research agendas, allocating funds, evaluating scientists, and rewarding them for their 40 work. Conversely, a lack of systematization in defining local research may leave a vacuum that can lead 41 to the imposition of colonizing standards, universalistic criteria, and ranking regimes (Ishikawa 2014; 42 López Piñeiro & Hicks 2015). This can erode the work of local researchers, who provide “the best 43 prospects for deriving policies which are both effective and engender broad-based support” (Stiglitz 44 1999 p. 3). 45 The importance of linking theory and practice when defining local research lies in its far-46 reaching implications for public policy and research evaluation. Although recurrently used (e.g. 47 Cancino et al. 2024; Chavarro et al. 2014; Hicks et al. 2015), researchers seem to disagree both on what 48 the term means —its defining features— and on the phenomena it covers (Hampton 2020). Concrete, 49 unified definitions are scarce, and only recently have partial methods for its measuring been developed, 50 although the approaches’ exemplars do not necessarily reflect those defining features. The most typical 51 method passed on over time uses toponyms for measurement (for instance, in Chavarro et al. 2014; 52 Ordóñez-Matamoros et al. 2010; Zhuang et al. 2020). Other approaches use national languages, journal 53 location, or absence from international databases as proxies for the local (e.g. Gupta & Dhawan 2009; 54 Kulczycki et al. 2020; Tijssen 2007). Besides scientists, other stakeholders like governments and 55 funders have also expressed growing interest in local research, but they avoid clearly stating the 56 concept’s scope (Cancino et al. 2024; Consejo Latinoamericano de Ciencias Sociales 2020; Delgado-57 López-Cózar et al. 2021; Hicks et al. 2015; United Nations Educational, Scientific and Cultural 58 Organization 2021; Zhang & Sivertsen 2020). 59 In this article, we explore various definitions and approaches to identify local research found in 60 the literature. Our aim is to elucidate how the concept of local research relates to its measurement, 61 highlighting the nuances involved in its theoretical depiction and the risks of applying different 62 empirical methods indiscriminately across regional and disciplinary contexts. To do so, we first turn to 63 some post-World War II discussions about science’s social goals and universalist framing, which 64 emerged in Global North countries but were counteracted by the Global South’s concern for their own 65 This is a preprint. It is undergoing peer review. The final published version may differ. 3 local challenges. We synthesize these diverse views into three recurring foci, where research is tied to 66 place, knowledge producers, and audiences. Then, we identify the most used methodological 67 approaches to local research, which we compute to more than 10 million Dimensions publications at 68 the disciplinary, country, and world region levels. Considering all theoretical and empirical 69 contributions to the matter, as well as their limitations, we propose a framework of local research that 70 articulates the dimensions we call locally situated, informed, and relevant research. The measurement 71 results suggest, on the one hand, that the approaches operate differently across the Global North-South 72 divide and, on the other, that there is no complete alignment between theory and practice when it comes 73 to local research. 74 Literature review 75 Defining local research 76 Concern for the local in relation to science and technology can be traced back to the aftermath 77 of World War II, when science policy was established as a distinct policy domain under the so-called 78 ‘social contract’ of science (Elzinga & Jamison 1995). In the United States, wartime capacities were 79 redirected towards strategic social goals (Bush 1945), while in Europe, reconstruction and the emerging 80 Cold War fostered new forms of scientific nationalism (Krige 2006; Krige & Barth 2006). Yet, science 81 presented itself as universal and self-governing (Merton 1968; Polanyi 1969), masking its national 82 interests. This universalist framing was exported to the Global South, where it shaped priorities, failing 83 to address local challenges. It is in this context that the notion of local research began to be articulated, 84 in explicit opposition to the ‘global’ agendas promoted from the North (Sábato & Botana 1968; Vessuri 85 1987). 86 The contrast between Global North and South is particularly visible in how evaluation cultures 87 have shaped legitimacy of local research. In the North, under Merton’s ideal of universalism, research 88 rooted in specific contexts was rarely identified as local but presented as general knowledge (Sarewitz 89 1996). Scientometric indicators such as Impact Factors have reinforced this by treating quality as 90 universally comparable, rendering locality invisible (López-Piñeiro & Hicks 2015; Ràfols et al. 2016). 91 This has become an issue, both in the North as well as the South, calling for more socially robust and 92 participatory knowledge production (Sarewitz 2016; Gibbons et al. 1994) or by explicitly denouncing 93 the clash between imported evaluation systems and developmental needs (Singer et al. 1970; Sutz 94 2000). 95 These contrasting trajectories reveal why the concept of local research has remained elusive. In 96 the North, locality was masked under claims of universality; in the South, it was explicitly foregrounded 97 but often delegitimised through evaluation practices. As a result, the term has been used in multiple, 98 This is a preprint. It is undergoing peer review. The final published version may differ. 4 sometimes inconsistent ways. Taken together, these debates point to three recurring foci: where research 99 is produced, by whom it is conducted, and for whom it is intended. Each emphasises a different aspect 100 of how research is tied to place, knowledge producers, and audiences, and each has been supported by 101 different strands of literature. 102 One of the recurrent ways in which local research has been defined is through its geographical 103 boundedness. Here, the object of study is tied to a specific place, population, or ecosystem, and the 104 locality of the research derives from this explicit geographical anchoring. For instance, Duncan (1989) 105 argues that empirical research should be sensitive to spatial differences. In the same sense, Sommer 106 (1990 p. 205) describes research as local when it is “site-specific, in that the researcher seeks to 107 understand or change conditions in a particular location at a particular time”. In sum, here we see that 108 local research can refer to specific locations with their territorial particularities at different geographical 109 levels. 110 Another trend of literature considers it especially important to conduct research that 111 acknowledges the unique features of the local, in order to diagnose and develop solutions in line with 112 the expertise of systematically marginalized communities (Moore et al. 2019; Sommer 1990). Here, 113 local research is conceptualised by reference to the individuals, communities, or sources upon which 114 research is produced. Local context and knowledge are seen as key characteristics of this type of 115 research because they help design and direct research priorities in alignment with societal needs and 116 demands. This is reflected in policy actions directed at promoting co-creation and productive 117 interactions, by stressing the need for alliances with local communities to generate knowledge that is 118 both scientifically robust and socially relevant (Spaapen & van Drooge 2011; Trencher et al. 2014). 119 Other examples are those drawing on indigenous or traditional knowledge systems as a foundation for 120 research (Akena 2012; Beigel 2013; Datta 2018). 121 At the same time, there are limitations to equating provenance with locality. The fact that 122 research is conducted by geographically concentrated authors or references does not guarantee that the 123 work addresses local issues. Conversely, external researchers can contribute to studies that are highly 124 contextualised and responsive to local needs. As several scholars have argued, the orientation and use 125 of knowledge cannot be assumed simply from the institutional or geographic origins of its producers 126 (Beigel 2013; Sutz 2000). This is reflected in the third stream of literature, which focuses on 127 communities, regions, or audiences for whom knowledge is most relevant. From such a view, research 128 can be labelled as local when oriented to address the specific needs or challenges of a bounded 129 community. The literature on the societal impact of research relates to this, as relevance is considered 130 a key characteristic (Bornmann 2013; Vanclay 2002). 131 The interest in knowledge production, scientific impact, and the localization of research 132 activities is problematized as well by spatial scientometrics, in order to draw attention to the 133 geographically concentrated attributes of research outputs that are normally circumscribed as global 134 (Frenken et al. 2009). The scientometric community generally agrees with a notion of local research 135 This is a preprint. It is undergoing peer review. The final published version may differ. 5 that implies “research related to either local, regional or national contexts, conditions or topics, as 136 opposed to research that is universalistic or decontextualized” (Chavarro et al. 2014 p. 2). As a result 137 of this type of investigation, “relevant knowledge is derived, contributing therefore to the local stock of 138 information necessary to increase local understanding and to produce new knowledge valuable to solve 139 local intellectual, technical, or social issues” (Ordóñez-Matamoros et al. 2010 p. 421). As we will see 140 below, all the ideas about local research discussed so far permeate attempts to measure the phenomenon. 141 Identifying local research 142 From an empirical perspective, different bibliometric approaches have been suggested in the 143 literature to explore and analyse ‘underrepresented’ or local research in bibliographic databases. Hence, 144 we find studies focusing on visibility (Tijssen et al. 2006), authorship and collaboration (Ordóñez-145 Matamoros et al. 2010; Tijssen 2007), interdisciplinarity (Chavarro et al. 2014), publishing dynamics 146 (Ishikawa & Sun 2016; Kulczycki et al. 2020; Mironescu et al. 2023), and topics (Miguel et al. 2015; 147 Zhuang et al. 2020), among others. These studies tend to focus on countries from the Global South. 148 However, the methods used to identify local research tend to vary across studies, with only a few 149 framing their methodological design from a theoretically driven perspective. In most cases, these 150 definitions are pragmatic and context-driven, more the result of technical solutions than of reflection on 151 the proxy used and its capacity to identify research inherently local. Next, we will review the most 152 common operational approaches to identifying local research in the scientific literature, which can be 153 grouped as follows (see Table 1 for a summary): 154 1) Toponym-based methods, which rely on the presence of place names to indicate a geographical focus 155 in research outputs. 156 2) Language-based methods, where the use of non-English languages is taken as an indicator of a 157 domestic orientation in research. 158 3) Journal-based methods, which assume that research published in national journals by local authors 159 and/or indexed in non-mainstream databases inherently reflects domestic concerns or topics. 160 Toponym-based methods 161 The starting premise for studies centred on toponyms is that research is locally situated in a 162 particular place or context (Chavarro et al. 2014; Miguel et al. 2015; Ordóñez-Matamoros et al. 2010; 163 Sommer 1990). One of the most commonly used methods for this approach is to identify toponyms and 164 demonyms in titles, abstracts, keywords, and various other sections of publications (Orduña-Malea et 165 al. 2024). It has been used to analyse multiple aspects of local research, like the share of contributions 166 on local issues authored by national and foreign researchers, and the effects of international 167 collaboration on team performance and agenda setting in Global South countries (Ordóñez-Matamoros 168 et al. 2010). Additionally, this method was employed to investigate the relationship between the degree 169 This is a preprint. It is undergoing peer review. The final published version may differ. 6 of interdisciplinarity and the local orientation of publications in Colombia (Chavarro et al. 2014). Other 170 studies focused on the growth rate in different disciplines within Argentina’s thematic domain (Miguel 171 et al. 2015) and the interactions of global and local geography topics in China and the United States, 172 especially regarding local responses to global issues and the global performance of local matters 173 (Zhuang et al. 2020). 174 These studies have contributed to gaining insights into the nature and scope of research 175 addressing local issues (Chavarro et al. 2014; Ordóñez-Matamoros et al. 2010). Still, toponym-based 176 methods pose a series of methodological challenges that can undermine their capacity to identify local 177 research. First, there is an unequal use of toponyms by region in titles and abstracts that emerges from 178 epistemological hegemony rather than numeric dominance in scientific production (Castro Torres & 179 Alburez-Gutierrez 2022; Castro Torres 2024). Authors from the Global North are less likely to include 180 concrete geographical references, particularly country names, in their articles than authors from the 181 Global South (Castro Torres 2024), who end up being more exposed to the evaluation and citation 182 disadvantages of following naming conventions (Kahalon et al. 2022; Miguel et al. 2023; Mongeon et 183 al. 2022). This leads to the reinforcement of the belief that “evidence produced in and about the Global 184 North is assumed to be more ‘universal’, whereas evidence from or produced in the Global South is 185 considered valid only for specific contexts (i.e. ‘localized’)” (Castro Torres & Alburez-Gutierrez 2022 186 p. 1). Second, certain disciplines and even subdisciplines, such as demography or international relations 187 within the Social Sciences, are more geographically bounded; thus containing more explicit geolocation 188 references than others, like general or educational psychology (Castro Torres & Alburez-Gutierrez 189 2022). Behavioural sciences provide an illustrative case where findings are frequently generalized from 190 predominantly Western, Educated, Industrialized, Rich, and Democratic (WEIRD) samples, 191 highlighting the existence of local research that remains unlabelled as such (Henrich et al. 2010; 192 Kahalon et al. 2022; Miguel et al. 2015). All these differences in localization habits, disciplines, and 193 data sources translate into fewer toponyms’ mentions, that is, false negative results. 194 Furthermore, working with toponyms is complex on a technical level. While their isolation 195 from natural language text might be straightforward when it comes to unique proper nouns, it can 196 produce false positive results when dealing with common place names, anthroponyms, and ordinary 197 nouns. A close example to us is Granada, which in Spanish refers to several cities in Spain, Peru, 198 Nicaragua, and the United States, a province in Spain, a department in Nicaragua, a country in the 199 Caribbean, a fruit (pomegranate), and a weapon (hand-grenade). Similar issues can be found in the 200 coincidence between streets and people’s names (Jerônimo et al. 2018) or in the close linkages between 201 toponyms and dermatological conditions (Radhika et al. 2021), just to mention a few examples. As the 202 level of territorial aggregation decreases, these types of scenarios multiply, which could explain the 203 strategy of most studies working with toponyms at the country level only, and hence looking more into 204 national research per se. While novel methodologies using Natural Language Processing (NLP) 205 techniques have the potential to overcome these limitations (Jerônimo et al. 2018; Mongeon et al. 2022), 206 This is a preprint. It is undergoing peer review. The final published version may differ. 7 there is space for improvement, especially due to the risk of false positives and false negatives in the 207 obtained results. 208 Language-based methods 209 Studies using language to identify local research usually take English as a proxy for global 210 research. This is because English has, for many decades now, been broadly considered the lingua franca 211 of science (Garfield 1967), with mainstream databases primarily covering publications in this language 212 (van Leeuwen et al. 2001). Added to this is the notion that publishing in the English language allows 213 research to potentially reach broader audiences, achieve greater impact and, therefore, gain wider 214 recognition and reward (Buela-Casal & Zych 2012; López-Navarro et al. 2015). This has led many non-215 Anglophone countries to implement national policies that promote publishing in English as a strategy 216 to boost internationalization and enhance their scientific impact (Robinson-Garcia & Ràfols 2020). 217 Fields that are especially affected by this linkage between language and internationalization are 218 those of the Social Sciences and the Humanities. Researchers will attempt to address different 219 audiences: global or national. Depending on their goal, they will publish in their own language in order 220 to reach local communities (Nederhof 2006). This is evidenced by the resistance researchers in these 221 areas pose to policies pushing them towards publishing in the English language (Ishikawa & Sun 2016), 222 highlighting the importance of their national language as a means to reach local audiences (Kulczycki 223 et al. 2020; Mironescu et al. 2023). 224 Another example of the use of language as a proxy for identifying local or global research is 225 found in the CWTS Leiden Ranking (https://www.leidenranking.com/), which aims at analysing the 226 scientific performance of over 1,500 universities worldwide. This ranking is based on bibliometric data 227 derived from Web of Science1. However, they exclude publications in non-English languages (CWTS 228 2024) and do not consider them as core publications suitable for analysis, thus reinforcing a double 229 bias: one related to the predominance of Global North coverage within WOS, and another to the 230 preference for English language publications. Given that languages other than English also contribute 231 to knowledge advancement, the main problem with language-based methods is that they overshadow 232 and underrate non-English scientific outputs (van Leeuwen et al. 2001). 233 From a technical perspective, language can be a problematic variable to identify local research 234 on a large scale. Many of them, like English, Arabic, or Spanish, are official or co-official in several 235 countries, which means they cannot be indicative of a single place. Even if this approach is controlled 236 for geographical affiliation, working at lower levels of territorial aggregation would generally not be 237 possible, since most regions and cities share a national language. Especially difficult is to extrapolate 238 the method to Anglophone countries, as it would not be possible to distinguish between local and global 239 scientific outputs based solely on their language of publication. 240 1 Although since 2024 it also includes an open version based on OpenAlex data. This is a preprint. It is undergoing peer review. The final published version may differ. 8 Journal-based methods 241 Studies on the origin of journals often assume that research published in nationally oriented 242 sources inherently reflects local concerns (e.g. López Piñeiro & Hicks 2015; Miguel et al. 2015; Navas-243 Fernández et al. 2018; Tijssen 2007). In contrast, international journals are typically defined by criteria 244 such as being produced by large editorial conglomerates, affiliated with renowned institutions, edited 245 by internationally diverse committees, selected by authors worldwide, published in English, or indexed 246 in mainstream databases (Buela-Casal et al. 2006). In this sense, one common approach focuses on the 247 geographical affiliations of key actors involved in the journals’ production, including publishers, 248 editors, reviewers, and authors. This method has been applied in studies on African journals, where 249 their relevance and performance are investigated within the global research landscape but are still 250 considered local (Tijssen 2007; Tijssen et al. 2006). However, these methods face challenges, as the 251 mere geographical location of a journal or its contributors does not necessarily indicate that the 252 knowledge produced is related to the context in which it is generated. Mainstream journals, often owned 253 by large academic publishing companies in the United Kingdom, the Netherlands, or the United States, 254 publish research internationally. This understanding of locality reflects the assumption that Western, 255 Anglophone journals are largely deemed global, while those edited in peripheral countries must meet 256 several extra criteria to gain the same recognition (Navas-Fernández et al. 2018). 257 Although many different citing behaviours have been identified (Borgman & Furner 2002), 258 several studies also suggest looking into the geographical location of citations to define international or 259 national journals, thus adding another perspective into locality: that which is used or consumed by 260 individuals geographically located in a single place (Grančay et al. 2017; Hladchenko & Moed 2021; 261 Moed et al. 2021; Pajić 2015). This citation approach is further developed in combination with journal 262 indexing in order to investigate two sets of highly cited papers within Spanish sociology according to 263 two databases: an international one (Web of Science) and a national one (In-RECS). As a result, citation 264 differences by language are revealed, pointing towards a relation between journal, citation impact, 265 language, and topics (López Piñeiro & Hicks 2015). 266 The use of database indexing as a criterion to identify local journals is based on the notion that 267 WOS (and Scopus by extension) index only those journals that constitute the core of each specialty. 268 This premise dates back to Bradford’s Law of Scattering and its application by Garfield to abstracting 269 services (Bensman 2012). Garfield based the inclusion of journals in WOS on universalistic criteria, 270 such as editorial standards and scientific impact. But over time, the selection procedure2 has been 271 modified as other products and citation indexes have been added, revealing a disciplinary and 272 geographical bias in its coverage (Chavarro et al. 2018; van Leeuwen et al. 2001). Furthermore, an 273 additional source of bias in what is considered to be the mainstream lies in the referencing practices of 274 2 More information on the journal selection process available at https://clarivate.com/products/scientific-andacademic-research/research-discovery-and-workflow-solutions/webofscience-platform/web-of-science-corecollection/editorial-selection-process/editorial-selection-process/. This is a preprint. It is undergoing peer review. The final published version may differ. 9 elite journals which, by disproportionately citing research from a few dominant countries, reproduce 275 existing global hierarchies (Bornmann et al. 2018). 276 Journal indexing is used in many countries as an evaluation criterion, enforcing scholars to 277 publish in journals included in mainstream databases as a strategy to internationalize their research. We 278 find examples of such an approach in countries such as Spain, Colombia, and Brazil (Chavarro et al. 279 2018). In the case of the former, we even find categorizations of journals based on a combination of 280 indexing and journal impact metrics (Torres-Salinas et al. 2010). By presenting mainstream database 281 content as universally representative, the presence of local research in the Global North is obscured. 282 More broadly, the dangers of using these biased and incomplete sources as proxies for global science 283 are significant. They provide a distorted view of the volume and nature of research agendas worldwide, 284 reinforce the notion that only science from the Global North constitutes international mainstream 285 research, and risk informing policy decisions that undervalue local contributions. This is particularly 286 problematic when assessing researchers’ performance, as it can lead to the marginalization of local 287 research efforts (Torres-Salinas et al. 2010). 288 Table 1. Summary of operational approaches to identify local research. 289 Operational approaches Proxies of local research Units of analysis References Toponym-based Toponyms Toponyms and demonyms mentioned in scientific publications Articles E.g. Castro Torres & Alburez-Gutierrez 2022; Chavarro et al. 2014; Miguel et al. 2015; Mongeon et al. 2022; Ordóñez-Matamoros et al. 2010 Conference papers E.g. Zhuang et al. 2020 Language-based Languages Non-English languages of scientific publications Articles E.g. CWTS 2024; Ishikawa & Sun 2016; Kulczycki et al. 2020; Miguel et al. 2023 Journals E.g. Chavarro et al. 2018; Mironescu et al. 2023; Navas-Fernández et al. 2018 Journal-based Authors Geographical affiliation of publishing authors Articles E.g. López Piñeiro & Hicks 2015; Miguel et al. 2015 Journals E.g. Hladchenko & Moed 2021; Moed et al. 2021; Navas-Fernández et al. 2018; Pajić 2015; Tijssen 2007 Databases Non-mainstream indexing of scientific publications Articles E.g. López Piñeiro & Hicks 2015 Journals E.g. Navas-Fernández et al. 2018; Tijssen et al. 2006; Torres-Salinas et al. 2010 References Articles E.g. Bornmann et al. 2018 This is a preprint. It is undergoing peer review. The final published version may differ. 16 Figure 2. Intersection of local journal sets across operational approaches. 434 Figure 3 delves into field and category differences within the subsets of local journals according 435 to each operational approach. The colour scale represents the proportion of local journals that belong to 436 each disciplinary category, grouped by Dimensions fields. In general, most shares are low, and any 437 pronounced differences are hard to observe. The Social and Health Sciences are the only fields with 438 categories that exceed 10% of journals identified as local by the various operational approaches. In the 439 first case, a few highlights can be seen in the Toponyms and References approaches for Human Society 440 (0.18 and 0.13, respectively), and in the Languages and Citations approaches for Education (0.13 and 441 0.12). In the second case, Health Sciences informs some of the highest shares, especially in Authors 442 and Databases for the Biomedical and Clinical Sciences category (0.22 and 0.20). As for the 443 Humanities, Life and Physical Sciences, the proportions remain low across all categories and 444 approaches. This distribution becomes more informative when looking at the values by country and 445 world region in the next Figures. 446 447 This is a preprint. It is undergoing peer review. The final published version may differ. 17 Figure 3. Share of local journals according to each operational approach by field and disciplinary 448 category. 449 Local research at the country and world region levels 450 In Figure 4, we show six different world maps, each one of them corresponding to an 451 operational approach. The colour scale represents each country’s proportion of papers published in local 452 journals with respect to their total number of publications during the period of study. The darker hues 453 indicate a low share of publications in local journals, whereas the lighter shades reveal higher 454 proportions. Overall, Toponyms stands out because it presents the highest proportions. Approximately 455 50% to 80% of the papers produced in parts of Africa, Southeast Asia, and Latin America and the 456 Caribbean are published in local journals according to this approach. Databases also presents medium 457 to high proportions, particularly in Southeast Asia, Eastern Europe, and Latin America and the 458 Caribbean. Indonesia stands out because 85% of its papers are not indexed in mainstream databases. 459 Most parts of the Citations map show low shares of local research except for countries like Indonesia 460 (0.52), Ukraine (0.50), Russia (0.40), and Brazil (0.30). From an audience point of view, these countries 461 publish around 30% to 50% of their papers in local journals. These same countries stand out in Authors, 462 with 30% to 45% of their papers being published in sources that mostly disseminate articles by 463 researchers declaring the same geographical affiliation. The References approach, in contrast, highlights 464 cases like the United States (0.30) and Canada (0.18). From an information source perspective, around 465 20% to 30% of their output is published in journals referencing papers that are geographically 466 constrained. Finally, Languages reports that almost all countries publish in journals operating in 467 English, although most of Latin America and the Caribbean exhibit slightly lighter hues than the rest of 468 the world. 469 This is a preprint. It is undergoing peer review. The final published version may differ. 18 470 Figure 4. Countries’ publication share in local journals by operational approach. 471 Figure 5 delves into category differences per operational approach at the regional level. Again, 472 the colour scale represents the proportion of papers published in local journals with respect to each 473 world region’s total number of publications during the period of study. The lighter the shade, the higher 474 the proportion, which allows us to observe that Latin America and the Caribbean have the most locally 475 oriented scientific production of all regions, across all fields and approaches. Africa and Oceania report 476 their higher proportional values in Toponyms, whereas the United States and Canada’s most noticeable 477 column is References. Asia and Europe display quite similar patterns of high to moderate values, 478 especially in Toponyms and Databases. Overall, Figure 5 informs that the highest publication shares in 479 local journals tend to concentrate towards Global South regions, and that one or several combined 480 approaches successfully detect the local in those more peripheral parts of the world. Still, despite their 481 differences per approach, all regions share two lighter-coloured sections that spread across the Social 482 Sciences and Humanities, including a couple of disciplinary categories from the Life Sciences. 483 This is a preprint. It is undergoing peer review. The final published version may differ. 19 Focusing on the operational approaches, Toponyms shows medium to high proportions in most 484 categories from the Humanities, Social, and Life Sciences. They are particularly evident in Africa 485 (Human Society = 0.88) and Oceania (History, Heritage and Archaeology = 0.84). On the contrary, 486 Languages is barely present in Latin America and the Caribbean (Education = 0.34) and Asia (Creative 487 Arts and Writing = 0.24). Authors appears quite muted in most regions but Latin America and the 488 Caribbean (Language, Communication and Culture = 0.58) and Asia (Education = 0.56). Next to it, 489 Databases has some of the brightest areas in Latin America and the Caribbean around Social Sciences 490 (Education = 0.85) and Humanities (Creative Arts and Writing = 0.89), with several categories also 491 standing out for Asia (Philosophy and Religious Studies = 0.77). The References approach informs 492 considerable shares in the United States and Canada across the Social Sciences (Education = 0.79), 493 Humanities (Philosophy and Religious Studies = 0.72), and even part of the Health Sciences. A few 494 Humanities categories are also quite noticeable from this approach’s perspective in Oceania, Africa, 495 and Europe. Lastly, Citations is almost only highlighted in Latin America and the Caribbean’s Social 496 Sciences (Law and Legal Studies = 0.72) and Humanities (Language, Communication and Culture = 497 0.73), along with a few light-colored categories in Asia (Philosophy and Religious Studies = 0.71). 498 This is a preprint. It is undergoing peer review. The final published version may differ. 20 499 Figure 5. World regions’ publication share in local journals according to each operational approach by 500 field and disciplinary category. 501 Finally, Figure 6 displays the results of a PCA conducted on the operational approaches at the 502 country and world region levels. The analysis includes countries with at least 3,000 articles published 503 within the 3-year period of study. Each dot represents a country marked with its corresponding two-504 letter ISO code although, for readability reasons, the least productive ones are left unlabelled. The 505 colours group countries by world region. The arrows show that, on the one hand, the Toponyms and the 506 References approaches are inversely related. This means that the countries where toponyms work best 507 for signalling local research tend to exhibit less geographically concentrated references, and vice versa. 508 On the other hand, Databases, Authors, Languages, and Citations point in similar directions and are 509 This is a preprint. It is undergoing peer review. The final published version may differ. 21 more closely correlated, as supported by Figure A1 (Appendix). The countries located on the right-hand 510 side of the biplot are therefore positively associated with higher coverage in non-mainstream databases, 511 more diverse linguistic practices, and more geographically concentrated authorship and readership. 512 The colour pattern reveals that the United States, Canada (in dark blue), and Oceania countries 513 (in purple) cluster to the bottom-left of References. Close to them, and extending towards the centre of 514 the biplot, appears most of Europe (dark magenta) together with several Asian countries (bright orange), 515 like Singapore (SG), the United Arab Emirates (AE), and South Korea (KR). Asia also spreads around 516 Toponyms, where a stronger presence of Africa (light orange) can be observed. Latin America and the 517 Caribbean (yellow) lean towards the four approaches shaping PC1, especially Databases. Further along 518 the x-axis, Russia (RU), Brazil (BR), Ukraine (UA), and Indonesia (ID) seem to behave like outliers 519 within this main structure of operational approaches. Overall, Global North countries cluster more 520 densely towards the bottom-left quadrant of Figure 6, suggesting that almost all approaches to 521 measuring local research have little explanatory power over them. In contrast, Global South countries 522 appear to be clearly characterized by most approaches, although regional differences persist within this 523 group. 524 525 Figure 6. PCA of local research operational approaches at the country and world region levels. 526 This is a preprint. It is undergoing peer review. The final published version may differ. 22 Discussion 527 In this article, we propose a framework of local research. Although the term is increasingly 528 mentioned in the scientific literature, it is often used indistinctively to denote different forms of locality: 529 the territorial scope of a study, the production and use of local knowledge, or the targeting of specific 530 audiences. Our goal is to determine how the concept of local research relates to its measurement, 531 highlighting the nuances involved in its theoretical depiction and the risks of applying different 532 empirical methods indiscriminately across regional and disciplinary contexts. To address this, we first 533 review how the concept of local research has been addressed, both theoretically as well as empirically. 534 Based on our literature review, we identify three dimensions – locally situated, informed, and relevant 535 research – and six scientometric approaches to analyse it – Toponyms, Languages, Authors, Databases, 536 References, and Citations. We then compare these approaches and examine how they link to the three 537 theoretical dimensions identified. 538 Our main findings suggest divergences across regional and disciplinary settings by operational 539 approach. Except References, where the United States and Canada stand out, most other approaches are 540 more prevalent in Global South regions alongside a few Eastern European countries. Toponyms and 541 Databases stand out, although they are based on problematic assumptions that give a heightened 542 impression of locality in certain places while hiding this quality in others. The Authors and Citations 543 approaches display similar world patterns, while Languages yields almost no information at the country 544 level other than emphasizing English as the lingua franca of science. 545 When combining this geographic data with fields and disciplinary categories, we see that the 546 Social Sciences and Humanities from the Global South have a stronger tendency towards the local. 547 Latin America and the Caribbean are most successfully captured by approaches like Databases, 548 Citations, and Authors. Whereas for African countries, the Toponyms approach is more prevalent in 549 those areas, alongside the Health Sciences. These results align with those of previous studies (e.g. 550 Castro-Torres & Alburez-Gutiérrez 2022; Hladchenko & Moed 2021; Navas-Fernández et al. 2018; 551 Tijssen 2007; Torres-Salinas et al. 2010). On the contrary, Life Sciences remains constant across 552 approaches and regions, and Physical Sciences are barely captured by any method. In sum, this initial 553 testing of the operational approaches to identify local research indicates that they operate quite 554 differently across both the disciplinary landscape and the Global North-South divide. 555 While the PCA supports our findings on regional differences, it also suggests that the 556 approaches’ behaviour does not fully match the theoretical views on local research. The approaches tied 557 to territory, such as Toponyms and Languages, do not align with a unified locally situated dimension, 558 although they help capture the local for many African, Asian, and Latin American countries. The same 559 happens with the approaches linked to knowledge producers, like Authors and References, which do 560 not converge in the locally relevant dimension. The most present countries in the former approach are 561 This is a preprint. It is undergoing peer review. The final published version may differ. 23 Latin American and Asian, whereas the latter displays a closer correspondence to the Global North. 562 This last group is formed by many of the largest scientific producers worldwide, so it is difficult to 563 conclude whether the method is capturing local research from the Global North or global research led 564 by these world scientific powers. On the contrary, the approaches tied to knowledge recipients, such as 565 Databases and Citations, seem to better explain the locally relevant dimension. This is especially visible 566 for most of Latin America and the Caribbean, as well as for a few Asian and Eastern European countries. 567 On the whole, these findings reveal that there is no complete alignment between theory and practice 568 when it comes to local research. 569 We acknowledge that our study is not free from limitations. One of them is the source used for 570 our analyses, as observed when working with Languages (Figure 4), where the coverage of Dimensions 571 beyond English seems to be relatively poor compared to more comprehensive databases such as 572 OpenAlex (Céspedes et al., 2025). Furthermore, some methodological decisions may be adopted 573 differently. For instance, local journals could be defined based on inequality indices such as Gini or 574 Herfindahl, instead of relying on the share of publications of the most contributing country. It would be 575 interesting to further explore how these methodological decisions may impact the findings. Finally, the 576 approach based on Toponyms could be more exhaustive by adding author keywords or searching in the 577 full text of publications, which was not feasible in our case due to the unavailability of such information 578 in the chosen database. 579 Concluding remarks 580 Our findings suggest that no individual approach entirely captures a dimension of local research 581 in every context. Nor are these existing measurements free from technical limitations and geographical 582 biases, as we have pointed out. This means that studies assessing the impact or production of local 583 research need to adopt nuanced methodological designs that adapt to the geographic region and 584 disciplines being analysed. Still, knowing to what extent what we capture with scientometric methods 585 is the phenomenon we aim to analyse remains contested. Therefore, we see great potential in their 586 combination to help seize such a complex, multidimensional phenomenon. Complementarity between 587 approaches and dimensions may be key to harnessing all their potential while minimizing their 588 limitations when addressing the local. This framework could prove useful to study not only the different 589 dimensions that comprise local research, but also their intersections. Possible instances of analysis could 590 be, for instance, the role of non-mainstream journals as communicating vessels between global and 591 local scientific circles (Chavarro et al. 2017; Di Césare et al., 2025), or the evolution of research fronts 592 that emerge to respond to localized problems but then expand when the issue acquires a global scale 593 (i.e. the spread of vector-borne diseases, as in Simon et al. 2008). Despite the challenges described at 594 length, we believe this contribution can serve as a starting point for integrating theoretically informed 595 This is a preprint. It is undergoing peer review. The final published version may differ. 24 conceptions with empirical approaches of local research, potentially helping to mitigate many of the 596 barriers current research evaluation systems face when assessing and supporting this type of scientific 597 research. 598 Acknowledgments 599 We would like to sincerely thank the anonymous reviewers for their careful reading and 600 constructive feedback, which helped us to significantly improve this work. We are also grateful for the 601 valuable comments received from colleagues when presenting earlier versions of this research at the 602 Conference on Scientific Mobility and Talent Attraction: Challenges and Opportunities (University of 603 Granada, 2023), the Seminar on Diversity and Recognition in Research Teams (University of Bristol, 604 2024), and the RESPECT Fellows Workshop (Georgia Institute of Technology, 2024). Our special 605 thanks to Dr. Cassidy Sugimoto (School of Public Policy, Georgia Institute of Technology), Dr. Rodrigo 606 Costas (Centre for Science and Technology Studies, Leiden University), Dr. Thema Monroe-White 607 (School of Policy and Government, George Mason University), and Dr. Diego Kozlowski (School of 608 Library and Information Sciences, University of Montreal) for many insightful discussions. Also, many 609 thanks to PhD candidate Carolina Pradier (School of Library and Information Sciences, University of 610 Montreal) for her comments on the code of our first manuscript. 611 Funding information 612 This work is part of the COMPARE project (Ref: PID2020-117007RA-I00) funded by the 613 Spanish Ministry of Science (Ref: MCIN/AEI/10.13039/501100011033 FSE invierte en tu futuro). 614 Victoria Di Césare is currently supported by a FPI grant from the Spanish Ministry of Science (Ref: 615 PRE2021-097022). Nicolas Robinson-Garcia is currently supported by a Ramón y Cajal grant from the 616 Spanish Ministry of Science (Ref: RYC2019-027886-I). 617 References 618 Akena, F. A. (2012). ‘Critical analysis of the production of western knowledge and its implications for 619 indigenous knowledge and decolonization’, Journal of Black Studies, 43/6: 599–619. DOI: 620 10.1177/0021934712440448 621 Beigel, F. (2013). ‘Introduction: the politics of academic autonomy in Latin America’. The Politics of 622 Academic Autonomy in Latin America, Public Intellectuals and the Sociology of Knowledge, 623 pp. 1–26. Ashgate Publishing: Farnham Burlington. 624 Bensman, S. J. (2012). ‘The impact factor: its place in Garfield’s thought, in science evaluation, and in 625 library collection management’, Scientometrics, 92/2: 263–75. DOI: 10.1007/s11192-011-626 0601-9 627 This is a preprint. It is undergoing peer review. The final published version may differ. 25 Borgman, C. L., & Furner, J. (2002). ‘Scholarly communication and bibliometrics’, Annual Review of 628 Information Science and Technology, 36/1: 2–72. DOI: 10.1002/aris.1440360102 629 Bornmann, L. (2013). ‘What is societal impact of research and how can it be assessed?: a literature 630 survey’, Journal of the American Society for Information Science and Technology, 64/2, 217–631 233. DOI: 10.1002/asi.22803 632 Bornmann, L., Wagner, C., & Leydesdorff, L. (2018). ‘The geography of references in elite articles: 633 Which countries contribute to the archives of knowledge?’, PLOS ONE, 13/3: e0194805. DOI: 634 10.1371/journal.pone.0194805 635 Buela-Casal, G., Perakakis, P., Taylor, M., & Checa, P. (2006). ‘Measuring internationality: reflections 636 and perspectives on academic journals’, Scientometrics, 67/1: 45–65. DOI: 10.1007/s11192-637 006-0050-z 638 Buela-Casal, G., & Zych, I. (2012). ‘How to measure the internationality of scientific publications’, 639 Psicothema, 24/3: 435–41. 640 Bush, V. (1945). Science, the Endless Frontier: A Report to the President. Washington, DC: National 641 Science Foundation. 642 Cancino, R., Albis Salas, N., Villaroel Valenzuela, J., Robles Belmont, E., Oliveira, T., Ràfols, I., 643 Palacios Núñez, G., et al. (2024). ‘Manifiesto por las métricas socioterritoriales de ciencia, 644 tecnología e innovación’. Presented at the XV Jornada Latinoamericana de Estudios Sociales 645 de la Ciencia y la Tecnología, Campinas, Brasil: Asociación Latinoamericana de Estudios 646 Sociales de la Ciencia y la Tecnología. 647 Castro Torres, A. F., & Alburez-Gutierrez, D. (2022). ‘North and South: naming practices and the 648 hidden dimension of global disparities in knowledge production’, Proceedings of the National 649 Academy of Sciences, 119/10: e2119373119. DOI: 10.1073/pnas.2119373119 650 Castro Torres, A. F. (2024). ‘Revisiting North-South disparities in naming practices: an extension and 651 update’, Social Currents, 12/2:103–110. DOI: 10.1177/23294965241297932 652 Céspedes, L., Kozlowski, D., Pradier, C., Holmberg, M., Shokida, N. S., Benz, P., Poitras, C., et al. 653 (2025). ‘Evaluating the linguistic coverage of OpenAlex: an assessment of metadata accuracy 654 and completeness’, Journal of the Association for Information Science and Technology, 655 76/6:884–895. DOI: 10.1002/asi.24979 656 Chavarro, D., Tang, P., & Ràfols, I. (2014). ‘Interdisciplinarity and research on local issues: evidence 657 from a developing country’, Research Evaluation, 23/3: 195–209. DOI: 658 10.1093/reseval/rvu012 659 Chavarro, D., Ràfols, I., & Tang, P. (2018). ‘To what extent is inclusion in the Web of Science an 660 indicator of journal “quality”?’, Research Evaluation, 27/2: 106–118. DOI: 661 10.1093/reseval/rvy001 662 Chavarro, D., Tang, P., & Ràfols, I. (2017). ‘Why researchers publish in non-mainstream journals: 663 Training, knowledge bridging, and gap filling’, Research Policy, 46/9: 1666–1680. DOI: 664 10.1016/j.respol.2017.08.002 665 Consejo Latinoamericano de Ciencias Sociales (CLACSO) (2020). Evaluando la evaluación de la 666 producción científica. Para una transformación de la evaluación de la ciencia en América Latina 667 y el Caribe, p. 25. Buenos Aires: Consejo Latinoamericano de Ciencias Sociales (CLACSO). 668 CWTS. (2024). ‘CWTS Leiden Ranking’. CWTS Leiden Ranking. Centre for Science and Technology 669 Studies (CWTS). Retrieved from <https://www.leidenranking.com/> 670 Datta, R. (2018). ‘Decolonizing both researcher and research and its effectiveness in Indigenous 671 research’, Research Ethics, 14/2: 1–24. DOI: 10.1177/1747016117733296 672 Delgado-López-Cózar, E., Ràfols, I., & Abadal, E. (2021). ‘Letter: a call for a radical change in research 673 evaluation in Spain’, El Profesional de la Información, e300309. DOI: 674 10.3145/epi.2021.may.09 675 Di Césare, V. (2024). Local research code (Version 1.0.0) [R language]. 676 <https://github.com/vdicesare/Local.Research> 677 Di Césare, V., & Robinson-Garcia, N. (2025). ‘Dataset of the manuscript “What is local research? 678 Towards a multidimensional framework linking theory and methods”’. Zenodo. 679 <https://zenodo.org/records/17278297> 680 Di Césare, V., Costas, R., & Robinson-Garcia, N. (2025). ‘Testing the knowledge bridging function of 681 non-mainstream journals in scholarly communication’. Presented at the 29th Annual 682