grammable towards all the specifications of the invited irritation. However, participation usually comes with provisions in place to minimise the risks.
The research system can easily decline recognition, and shut out certain attempts to participate.
One way of pushing the boundaries is to reflect on the thesis itself, within the thesis, just as in the previous paragraphs. While doing research, de
cisions made sometimes turn out to be dead ends. The return to start, and the learning process related to that, are usually made invisible, just like insecurities about methodological workarounds or conceptual inconsist
encies. I try to work against this trained reflex of obscuring insecurity in order to disclose the actual vulnerability of research as such. Even though the curriculum of my doctoral education asks for an authoritarian present
ation style, I hope to divest myself of as much authority as possible, being aware that this, inevitably, must be done with authority, in order to meet the requirements of the curriculum. This simple example shows that there is no way out of the system, once included, as long as selfexclusion and its serious consequences is not an option. The leeway for structural modifica
tions from within is very limited.
Together with the decision for an antiauthoritarian presentation style that, for instance, does not cover limitations, comes the refusal against any universalist standpoint that I noticed myself stepping into. The con
sequence is a radical and bold firstperson narration, since all the obser
vations I make for this research are mine alone, unless I let others speak within the framework I prepared. Of course, I hope that readers can com
prehend, feel stimulated and can relate their own observations to mine, thereby confirming them, and vice versa, but I never want to evoke the impression that I generalise beyond this, thereby replacing the striving for universality with the striving for solidarity. Research reception works like stepping into the author’s shoes, succeeding or failing in making them fit, and overcoming more or fewer obstacles while using them. I see my task as an author in offering a design that fits my research problem, and I am aware that there cannot be a design that appeals to everyone equally well.
The most important consequence moves beyond modes of presentation and representativeness disclaimers; since my research problem and my over
arching research questions have a huge scope, it would be pretentious to try to do right by that by collecting an enormous corpus of data with a global scope, or by uncritically adopting an existing large corpus. In contrast, the empirical corpora that are weaved into my research range from replies to a smallscale survey with openended questions, to UNEScO demographic data. Including numerous small analyses of data excerpts from a variety of sources is a way of making more obvious that, first and foremost, my con
ceptualisations, influenced by uncountable events that I picked up more or less consciously, shape my results, while data play a subordinate role.
When reading research reports, I often get the impression that data is seen as external to the research results, which are only incorporated via ana
lysis. I argue that incorporation starts with the first decision that is made concerning the data. For instance, even though my influence on how UN
EScO demographic data were compiled is zero, I take responsibility for their limitations because I use them in support of my argumentation. With ease, data can be selected because their shortcomings work in favour of the ar
gumentation, or neglected because they do not, justifying the neglect with their shortcomings. Because of this shaky role that data play in research, I do not ‘let them speak for themselves’, because they are unfit to do so. The reflection on the nature and provenance of data might take up more space in this thesis than their analysis. Therefore, the thesis is interspersed with small empirical analyses only, rather than being based on them.
According to my argumentation so far, comparing the use of a local pub
lishing infrastructure by local researchers to their use of the ‘international’
infrastructure could provide valuable insights into the impact of coloniality on the research system. Bibliometrics is usually a popular tool of choice to such an end. Bibliometrics, in general, serves as the principal method
ology for collecting and analysing bibliographical data, including citation data. This subfield of LIS relies heavily on data derived from WOS, or, de
pending on the type of research questions, on institutional or statewide research information systems (CRIS, see Sīle, Pölönen et al. 2018). Both Scopus and Google Scholar (GS) appeared on the market of sources for cita
tion data in 2004, but have not been able to shake the position of WOS so far. It is well known that for large regions in this world, a representative
amount of research results produced in the SSH is discoverable neither in WOS, in Scopus nor in CRIS, because only a few ‘Global South’ institutions maintain them. A good amount might be covered by subject databases which potentially are used as a GS source. GS, while providing a broader coverage for most disciplines, especially for the SSH (Prins, Costas et al.
2016), is contested because of its questionable data quality (Jacsó 2010).
Although this seems to be improving (Harzing and Alakangas 2016), its usability for scientometric research is limited.
The only subject database that has indexed a substantial amount of litera
ture published in Africa was discontinued in 2011, and older bibliographic data is basic and not comprehensive (see Section 4.4). If required data are not readily available, this does not mean that research relying on such data is impossible. What I would call nonmainstream scientometrics¹⁸—that is, scientometrics based on handpicked data—might be tedious. However, while providing insights into otherwise unsolvable research problems, non
mainstream scientometrics also point out the shortcomings of databases and tools. Databases always rely on inclusion criteria, and while it can be optimistically assumed that scientometricians are aware of the respective criteria, they might be critical about them, or about the result of the selec
tion.¹⁹ Usually, this is acknowledged in a footnote, but does not impact the actual research. Further, even if the scientometrician agrees with the
18 ‘Mainstream scientometrics’ is rarely mentioned in the literature. This is probably because the vast majority of papers published in the pertinent journals, like Scientometrics, refer to the data found in the mentioned databases. However, Thevall 2008 makes use of the term ‘mainstream bibliometrics’ without explicit definition, but seems to equal it to methods based on ‘core citationbased impact measures’, facilitated by WOS, now also by Scopus and GS, and further supplemented by researcherbased metrics (hindex), CRIS
based methods, knowledge domain visualisation, and analysis of usage data from digital libraries. Thelwall contrasts mainstream bibliometrics with webometrics, which is, today, in this context, often referred to as altmetrics. Altmetrics is based on social media mentions, a specific form of link analysis (and, I would add, also on counts of views and downloads of digital publications).
19 See Chavarro, Ràfols et al. 2018 for a recent position. The current inclusion criteria for WOS are found on Web of Science Group, Web of Science Journal Evaluation Process and Selection Criteria, https://clarivate.com/webofsciencegroup/journalevaluationproce ssandselectioncriteria, visited on 27 April 2020; discussed in Sections 2.2.4 and 3.4.
criteria, this does not mean that every publication which complies with the criteria is really included in the database. Apparently, mainstream sciento
metrics helps to find answers to a certain type of questions, but to others, as demonstrated in this thesis, answers typically found are insufficient. This is often due to the design of highly specialised commercial tools, and it can be worked around by not limiting the database to preselected data.
Intentionally or unintentionally, mainstream scientometrics—just like any other indexing or selection process related to academic literature—
often excludes research from poorly funded research environments, from formerly colonised countries, from the ‘peripheries’ (see Chapter 3). As I mentioned before, if journal or book publishers do not receive recogni
tion, they will also lose potential authors since those are encouraged to compete with their publications globally. If journals and book publish
ers do not play by the rules set by the streamlined industry in the ‘Global North’, they will most likely have no chance of surviving. It is not deciding whether these are cases of exclusion or nonselection that is of central im
portance, but rather how both variants are deeply inscribed in the research system—and equivalent in effect.
Nonmainstream scientometrics are especially required for studying re
search communication taking place where registration mechanisms work imperfectly. Even though institutional registers like CRIS can provide vis
ibility to the publications they contain, they are also used to foster compet
ition between researchers, which is based on quantity, not quality of their work. With my thesis, I do not aim to point at the scantiness of African research documentation. I see good reasons to deliberately not participate in this competition, even though the actual reasons that few African univer
sities participate might instead be related to a lack of resources. However, I actually observe a certain scantiness of effort undertaken by information providers and libraries in the ‘Global North’ to make African scholarly ma
terials available, because from the official versions of those organisations’
missions and professional codes, it follows that a closeto full worldwide coverage of relevant research literature can be expected (see Chapter 5).
Relevance is rarely explicitly defined, and even the relevanceranking al
gorithms of commercial academic search engines are a black box. Deco
lonial thinking can help to get to those latent definitions. It sets out at the border between coloniality and modernity, making visible how global designs erase local attainments. With my laborious approach, I set up a sample of ‘local’ Southeast African journals in order to arrive at a sample of publications by Southeast African authors that were not necessarily pub
lished by WOS or Scopusindexed journals. This way, I work around the straightforward path set by globalised information systems. Counteracting these standard tools of bibliometrics, produced in the ‘Global North’, the approach is denoted by ‘decolonial scientometrics’ (see Chapter 4).
In this study, quantitative results solely serve to establish a cursory over
view of the accessibility and visibility of Southeast African SSH content, specifically in Europe, and only because of that, an idea about the total quantity of this content is needed as well. There is no way around the use of databases that are also used for audit purposes, and each has its own set of limitations. The complex sampling strategies described in Chapter 4, combining several databases, and going down to the individual researcher level, is an attempt to overcome the problem of observing something huge and complex with very limited vision, which itself becomes very visible through this approach.
There are points where I deem it necessary to include a comparison in or
der to make inequalities—that I do not intend to deny—palpable. Those inequalities delineated in Section 3.6 refer only to WOS and Scopus, and to the reality created by them, and not to the reality of academic publish
ing as such. A bibliometric study is part of a rather conceptual chapter, because, just like the other considerations of this chapter, it is part of the attempt to think globally. At the core of Chapter 3 is a quest for under
standing why, if society is a global system, does it then play out differently in different local contexts. It would be impossible to understand this if the focus were confined to only one certain location. Even though the most sense might be made from a familiar context, a system of communication that has global reach must be analysed as such. The second reason why the small comparative study on global scale, making use of WOS and Scopus databases, is located in a conceptual chapter is its meta level of analysis;
those quantitative results can only be interpreted on the level of fourth
order communication (or observation: communication always relies on observation). Based on Luhmann’s theory of observation²⁰ that I touched upon earlier, omnipresent in his mature œvre, the levels can be described as follows:
1. communication takes place: e. g. in the form of thinking, writing, and distribution,
2. communication is structured to increase its likelihood of success (see Section 2.1.3): primary registration as academic knowledge and scholarly communication, however limited in its scope,
3. structured communication is observed: indexing by, in this case, two service providers who themselves are observing scholarly com
munication, and
4. observations of structured communication are observed: my study observes the indexing of those services.
Basically, the level of meta observation in Section 3.6 is the same as in Chapter 4, where I focus on Southeast African authors, but in the latter, I tried hard to get the big picture, combining all sources of indexing that I could find. The difference therefore is that, in Chapter 4, although still re
maining on the fourth level, I aim at getting closer to the observations from level three in order to form a more precise picture. There are uncountable observations happening on this level, so this picture is only achievable as an excerpt. Therefore, my sampling method includes as many addresses of registration and indexing as were feasible, avoiding copying the limitations of single observers on level three into this study. Unfurling limitations of single observers with the aim of reducing or relocating limitations is, I sug
gest, the general aim of any conceptual chapter.