Exploring the themes of the territory: a topic modelling approach to 40 years of publications in International Journal of Lifelong Education (1982–2021)

ABSTRACT

The question of what the research of lifelong education is all about needs to be revisited from time to time. Not only is this line of research approached from a multitude of academic disciplines – such as sociology, psychology or philosophy – the very concepts that are used to denote the field also undergo important changes over time, e.g. from lifelong education to lifelong learning. In this contribution, we will explore this rather elusive research territory using a large-scale text analysis of a specific scientific journal, International Journal of Lifelong Education, based on metadata (abstracts, keywords and titles) from 1,185 articles published between 1982 to 2021. Based on topic modelling techniques, we identify the main themes that have been prevalent within the journal, and how the journal’s content has changed character over time. We end the paper with a more critical examination of what kind of political and scientific currents might help explain what has led research practices to be more descriptive, micro-oriented and work-related over time.

KEYWORDS:

• Sociology of science
• lifelong education
• bibliometrics
• research

Introduction: what is the research on lifelong education all about?

The question of what the research on lifelong education is all about needs to be revisited from time to time. Not only is this line of research approached from a multitude of academic disciplines – such as sociology, psychology, history or philosophy – the very concepts that are used to denote the territory and its legitimate objects also undergo important changes over time. The 40th anniversary of the International Journal of Lifelong Education (IJLE) provides a good occasion to make use of the ample research contributions that have been produced in this journal to explore how the content of research in lifelong education has changed thematic orientation and to understand the current ‘state-of-the-art’ as well as potential blind spots. In this contribution, we will treat lifelong education as synonymous with the education and learning of adults in all forms they may take when used in the language of researchers.

The fact that research on lifelong education is characterised by heterogeneity and elusiveness is by no means news to the devoted readers of IJLE. Throughout the years we have become well-acquainted with research from various countries and continents, from various scientific vantage-points and based on a multitude of methodological procedures. Yet the pluralism of the research field and this journal also makes it challenging to sum up the developments from a birds-eye point of view. Over such a long-time span as 40 years, the formative academic traditions of the past easily fall into oblivion and all the epistemological positions and discursive camps that play a part in it are hard to sum up neatly. The ideas that one has with regards to the shape and developments of the past contributions might stem as much from one’s own position within it, as it does from empirically grounded observations of fundamental changes in research and discourse over time. Is the map of the research territory perhaps bound to be provincial and partial?

Mindful of these pitfalls or potential problems, we have set up to explore all peer-reviewed articles that had metadata from abstracts, titles and keywords that were published in IJLE between 1982 to 2021. By performing a topic modelling analysis on all the words from abstracts, keywords and titles, we provide an empirically grounded semantic map of the research territory throughout the full lifespan of the journal. In what is to follow, we first specify the rationale for studying the knowledge production of these journal entries using a large-scale textual analysis. Next, we section out the main thematic clusters that have been identified by the topic model as the best representation of the linguistic variation published within the journal. By providing examples of contributions of each of the distinct topics that have been identified as pertinent by the topic model, we compare these various lines of thought and research practices with one another. Finally, we critically discuss which thematic clusters have come to dominate and which have fallen out of favour over time.

Background and previous research

Ever since the start of IJLE in 1982, there has been a strong wave of transnational collaboration within humanities and social science, with the English language increasingly establishing itself as the lingua franca of academic knowledge production around the globe (Heilbron & Gingras, 2018; Fejes & Nylander, 2019; Fejes et al., 2020). In the Western industrialised world, the systems of adult education grew rapidly from the 1970s and onwards, creating a need for knowledge on the teaching and learning of adults. In the preface to the inaugural issue of IJLE, Husén (1982) identified how more and more individuals were engaged in education and learning beyond the age of childhood and youth traditionally associated with educational research. He also pointed out that this ‘spectacular growth’ was related to changing economic demands, higher life expectancy as well as the political vision of ’meritocracy’.

The birth of IJLE and its initial aim was in no way exclusively oriented to educational processes within the Western and industrialised centres of the capitalist world order. Importantly, its early scope focused on learning processes in former colonies and the Global South, a foci that have remained somewhat of a subtheme (cf. Leach, 1982; Maruatona, 1999; Chuks, 2004). Looking back at the first few numbers of IJLE, we can identify a strong ambition to transgress the established national, linguistic and institutional ramifications of educational research in various ways. Apart from opting for ‘International’ in its title, the second editorial of the journal invited contributions ‘from any part of the world’, and as a means to reach a wider audience and circulation, the accepted articles were initially provided with abstracts in French and German (Editorial, 1982). Although the ambitions towards a more global coverage were clearly present from IJLE’s early stage, its international commitment has proven hard to materialise and the preponderance of authors of anglophone origins and from the Global North has remained stark. How the research field of the education and learning of adults takes shape, and the geopolitical and international ramifications of who partakes in the research and receives scholarly recognition, have been highlighted and problematised in previous research and scholarly debates (cf. Fejes & Nylander, 2019; Field et al., 2019; Larsson et al., 2019; Milana, 2019).

In more recent times, and within the academic environments where knowledge on lifelong education is produced, the importance of peer review English articles has generally increased further. This might partly be due to the career strategies and publication priorities of researchers themselves, who, within relatively small research niches like lifelong education, can benefit from engaging in international exchanges and plugging themselves into broader research networks. Importantly, these strategies and adaptations among fellow scholars can also be seen as a response to standardised benchmarks of evaluating what counts as worthy and good research from a management perspective (Gingras, 2014, 2016; Nylander et al., 2013). It is a development that also reinforces power relations between people of different languages and with a given distance or proximity to ‘the centre’ of the globalised system of scientific knowledge production (Gingras, 2014; Heilbron & Gingras, 2018).

The pretence of peer-reviewed English articles of providing an unbiased and universal representation of the state-of-the-art research in the field has thus been hard to sustain. As for the research field of lifelong education, numerous studies have pointed out how extra-academic factors such as gender, geography, editorial power, etc. seem to skew the dynamics of scholarly recognition, which might indicate that it is a weak scientific field (Larsson et al., 2019; Rubenson & Elfert, 2019). Previous bibliometric research on leading journals has found, among other things, that research in this field is dominated by authors from anglophone countries, and that the contributions cite primarily male authors and draw rather eclectically from other research traditions (sociology, psychology, philosophy, political science, etc.) (Fejes & Nylander, 2019). In terms of methodological approaches, the field can be characterised as rather uniform, as there is near-total dominance of qualitative studies (Boeren, 2018; Daley et al., 2018; Fejes & Nylander, 2015). Theoretically, three perspectives have been found to gain prominence and traction in recent years: socio-cultural perspectives, critical pedagogy and post-structuralism (Fejes & Nylander, 2015).

Although these studies might be illuminating for understanding institutional, social and geographic power relations that underpin research practices and the endorsement of certain scholarly contributions, they have provided little in-depth knowledge to the question of scholarly content. Yet given that the databases that index peer-reviewed English journals have gathered much of the relevant information about the content of these articles – such as their abstracts, keywords and titles – the possibility of generating cartographic meta-descriptions of this elusive research territory is clearly within reach. Topic modelling the content of the IJLE is a pertinent case to start these explorations, as it has been an important outlet for the research on the learning and education of adults for more than 40 years and contains a wide array of scholarly contributions.

The study: a topic modelling approach to the research territory

In 1982, the inaugural year of IJLE, Rubenson (1982) published an article entitled ‘Adult Education Research: in Quest of a Map of the Territory’. Rubenson’s quest was directed to the pertinent question of what governs the knowledge production of adult education research and how this might be objectified into a ‘territory’.

Building on the conceptualisation of adult education research as a territory composed of rivalling discursive strategies, this contribution analyzes metadata from all peer-reviewed articles published in IJLE during its 40 years of publication. Drawing on large-scale text analysis known as topic modelling, we identify which latent topics have been most prevalent within this journal, and how the journal’s content has changed in character over time. The following two research questions are addressed:

What kind of research on lifelong education has been published in IJLE during its 40 years of history when taking into consideration the wordings used in abstracts, titles and keywords?

What kinds of overarching thematic changes have characterized the production of knowledge published in IJLE, and what perspectives have come to dominate this academic territory over time?

The first question relates to the content of IJLE in terms of what themes and research problems have been used and thus been considered legitimate among its authors, reviewers, editors and readers. The second question relates to the scholarly traditions and ideals that lie behind knowledge production and that are also made visible through the analysis. The two problems address the thematic orientation in an empirical and explorative way using metadata assembled about all the journal entries on an equal and symmetrical basis.

As compared to prevalent theories in the sociology of science, our approach bears some resemblance to Bourdieu’s relational approach in that we map out a specific area or space, interrogate the prevailing structure of knowledge as a social practice and try to provide room for collegial reflexivity by objectifying the conventional forms of objectification (Bourdieu, 1988; Bourdieu & Wacquant, 1992). However, as compared to Bourdieu’s programme of mapping out ‘academic fields’, we are not exploring how different discursive strategies might be related to institutional and dispositional power relations, nor are we conducting a fully-fledged field analysis, as we study a single journal. Zooming in on a single journal does have the added benefit of providing room for reflecting on the agentic dimensions of artefacts, in this case through a scientific journal, aspects that have been given more relevance in post-bourdieusian research (cf. Boltanski & Chiapello, 2005; Chiapello & Gilbert, 2019).

Methodological considerations

Within the social sciences, more broadly, topic modelling approaches are effective in revealing the ‘hidden thematic structure in large collections of documents’ (DiMaggio et al., 2013, p. 577). Topic modelling is a probabilistic text analysis based on LDA, which stands for Latent Dirichlet Allocation.¹ The topic model produces a set of interpretable ‘topics’, which are groups of words that belong together under a single thematic structure. Regardless of their syntax and word location, the topic model uses statistical inferences to capture these words in patterns of co-occurrences (DiMaggio, 2015). For the LDA to infer these hidden topics in each text corpus, it imagines a generative process in which the documents are created by randomly mixing latent topics. In this ‘bag of words’, each topic is characterised by a distribution over all the words, where each word is given a distribution of which topic it belongs to and each document is given a distribution in the overall map of topics (Blei et al., 2003).

Compared to traditional forms of text analysis of ‘manifested content’, which require deep readings and hand-coding procedures to derive themes, latent text mining techniques such as topic modelling can extend the volume and provide more precision to the analysis (cf. Berelson, 1952; Blei et al., 2003; Franzosi, 2010). As topic modelling is an ‘unsupervised’ and probabilistic form of text analysis, it does not require a predefined selection of search terms, nor does it require any manual coding or classification. In previous research on the knowledge production of adult education and lifelong learning, multiple forms of manual content analysis have been performed on single journals (cf. St.Clair, 2011; Taylor, 2001) as well as on a collection of journals (cf. Fejes & Nylander, 2015; Hayes & Smith, 1992; Larsson, 2010). A slightly more data-driven approach is deployed by Boeren (2018), who based her review of Adult Education Quarterly, Studies in Continuing Education and International Journal of Lifelong Learning on a predefined group of search terms that are meant to represent different research approaches and ways to conduct research. However, none of these studies using bibliometric data as their starting point has made much use of possibilities, which novel forms of text mining techniques offer when it comes to comprehensively identifying latent content and thematic orientations over time.

Even if topic modelling is inductively deriving patterns and co-occurrences of words, interpretative work has to be conducted to evaluate and describe which model makes sense and finalise the analytical process. For this reason, it has been important to apply the method to a line of research that we are already acquainted with and where our interpretations are likely to fall back on familiarity with the affiliated authors, themes, modes of inquiry and research conventions.² Once these underlying thematic clusters had been identified in the form of linguistic clusters from the full corpus of text, we looked closely at the relative and absolute frequency of words in each of the topics in an interactive map provided by the LDA visualisation tool (Chuang et al., 2012). By comparing the interpretation of the discursive elements that were aggregated under each topic, we could build an empirically anchored understanding of those topics that were generated and their relationship to one another. Once that topic model had been developed and we settled on the preferred model, it was also possible to derive useful supplementary bibliographic data such as statistics of citations or institutional affiliations. For easing the presentation of our results and deepening our understanding of each line of research, we also compiled lists of the articles with the ‘best-fit’ in each of the topics and read these articles more closely. Utilising the advancements from data science and text mining, while maintaining our interest in the qualitative dimension of the underlying material, can be seen as one way to overcome the unfortunate divide between qualitative and quantitative methods of inquiry haunting modern social science (Ragin, 1987). In the final stages of our analysis, we tried to reduce the complexity of our model to its common denominators, focusing on the thematic structure that shapes the research territory as a whole.

Data

The study was conducted by downloading available texts from IJLE in Scopus. We integrated the text corpus to the software Python, within which we deployed the instruments and topic model analysis. For the topic model analysis to be at its best the full body of text had to undergo certain clean-up procedures. In this case, we only took non-pronouns that consisted of alphabetic characters and reduced them to their simplified form, and took out all stop words and the words that were too frequent or too rarely used. Words that were frequently mentioned in a sequence of two or three were treated as one entity instead of individually, to better capture their full meaning. For example, lifelong learning was kept intact as a unit, ‘lifelong_learning’, and not separated as if it contained two different words.

This underlying database first consisted of 1,276 peer-reviewed articles, i.e. excluding the editorials, book reviews and other commentaries. However, in 91 cases these articles lacked abstracts and were thus deleted from the sample. For the rest of the entries (1,185) abstracts, titles and keywords were assembled into an integrated text corpus. Thus, the missing data compose less than 10 per cent of the full publication’s record of the journal over time. It is of course very important to bear in mind that the sample is not evenly distributed across the lifespan of the journal, as there is more text emerging in the latter half of the period than in the first decade. For instance, in the period 1982–1992, the articles were below 20 annually, whereas the period 2010–2020 consisted of roughly 40 articles a year (see Figure 1). That means that for the global map of topics most contributions come from the last two decades of the journal.

Figure 1. The number of articles included in the topic model analysis over time. Based on metadata from IJLE, 1982–2021.

Another potential limitation is that the study is based on metadata derived from abstracts, keywords and titles, not the wordings of the full peer-reviewed papers as such. If there are certain conventions of writing an abstract, choosing index words and titles that differ from the actual research found in the articles, our research approach does not pick this up. A more modest way to frame this contribution would be to say that it is a study of wordings used in abstracts, titles and keywords. But as we have reasons to assume that there is a relationship between the summary of the article, its title and keywords, and its actual content, we think that our results can be taken to represent the full knowledge production of IJLE over time.

Next, we will present the results of this study based on a model of ten topics that maps out the research terrain of lifelong education based on IJLE publications to date. When we present these results, we will refer to territorial ‘regions’ (East, South, West, North), which are thematic subdivisions derived from the structure of the overarching intertopic map (see Figure 2). Each of these regions thus aggregates topics that bear ‘family resemblance’ to each other to a varying degree. The closer these topics are located to one another in the map the more semantic similarities they share, and the further away they are from one another the more they differ in terms of wordings and language use. We will also highlight some of the words that have given shape to each of these semantic clusters in both absolute and relative terms. References to absolute frequency stand for the frequency of words in a given topic, whereas when we talk about distinguishing words or their relative frequency, it is referring to the words that are featured more in this topic as compared to the other prevalent topics. Figure 2 (below) thus depicts the overall semantic map of the content from IJLE. For the interpretation of each of these topics, we refer to the description found throughout the findings section, where each of the overarching regions is presented in greater detail based on examples from article entries in IJLE.

Figure 2. Intertopic map of the IJLE from 1982 to 2021 based on ten topics and their marginal distribution.

The results

THE EASTERN REGION: critical research on lifelong learning

In the Eastern region of the map, we find topics 7, 3 and 6. Here, research is oriented to adult learning as a political phenomenon. There is a strong focus on citizenship and social movements in this sphere, especially in topic 7, which also displays a strong normative orientation. The frequency of words – such as social, society, adult_education, political, public, educational, immigrant, new, movement, state, argue and democratic – illustrates that research within topic 7 engages in argumentation around social and political issues and largely sees research as a tool for bringing about change.

One example of an article that has a high matching value to this thematic orientation is Clover’s (2015) entry ‘Adult education for social and environmental change in contemporary public art galleries and museums in Canada, Scotland and England’. In this article, Clover uses selective examples from public arts and cultural institutions in these three anglophone countries to show how these can be understood as pedagogical spaces that play an important role in the struggle for social and environmental change. Another example is Wilson and Cervero’s (1997) article ‘The song remains the same: the selective tradition of technical rationality in adult education programme planning theory’, in which they draw inspiration from urban planning theory and discourse analysis to argue that adult education planning represents ‘a selective tradition’ within a dominant discourse that privileges ‘technical rationality’.

Over time topic 7 has remained fairly constant in IJLE, often drawing on various forms of critical theory formulated outside the territory and translated to become suitable for the context of adult learning. Names such as Peter Boshier, Jack Mezirow and Stephen Brookfield are recurring references within this topic. Popular education, informal learning and social movements play a crucial role in many of these entries. Stressing civic engagement as an important facet of adult education and learning and conceptualising these activities as underpinned with political axiologies unite the entries. By drawing from various critical perspectives (post-Marxism, feminism, postcolonial theory), social justice is advocated, and normative claims meant to lead to equality and fairness are often verbalised.

Closely related to these critical tropes in the extreme Eastern territory of the map, we find topic 3. The most used words here include adult_education, lifelong_education, paper, educational, curriculum, development, theory, analysis, and concern. In relation to the other topics of IJLE, this is characterised by a high number of studies directed towards curriculum and drawing from philosophy, whereas research endeavours are (similarly to topics 6 and 7) often directed to critique dominant ideologies and the status quo based on various social and political theories.

One example from topic 3 is an article by Holst (1999) titled ‘The affinities of Lenin and Gramsci: implications for radical adult education theory and practice’. In this article, the focus is on clarifying the relationship between Leninism and the work of Antonio Gramsci, and the relevance of such work for radical adult education. Another example from this topic is an article by Connelly (1992) titled ‘A critical overview of the sociology of adult education’. Here, Connelly explores critically and conceptually if there is a sociological tradition within adult education attempting to provide a theoretical foundation for the discipline.

Topic 3, thus represents critical research of a slightly different kind, although closely associated with topic 7. A common denominator among those articles that fit topic 3 is that they are conceptual rather than empirical. On the one hand, there is more political and normative research on adult education, and on the other hand, there is critical research on the discipline of adult education itself. Following topic 3 over time, we can see how it was a very common theme in the 1980s, peaking in 1984/1985 when it represented 35 per cent of all tokenised words. Since then, however, there has been a strong decline of this line of research and, in 2020, the topic only represented 5 per cent of all the tokenised words. This amounts to a major thematic decline, which warrants further discussion later.

The Eastern region also includes topic 6. The overall most used words within this topic are: lifelong learning, lifelong learner, concept, and policy. Similar to the other topics in the East, it retains a strong theoretical orientation combined with an interest in informal learning. One example from topic 6 is an article by Watkins and Marsick (1992) titled ‘Towards a theory of informal and incidental learning in organizations’. In this well-cited article, the authors develop a theoretical framework for understanding incidental learning in organisations. Their framework includes, among others, ideas regarding action, proactivity, critical reflexivity, and creativity. Another example is an article by Bagnall (2010) titled ‘Citizenship and belonging as a moral imperative for lifelong learning’. Here, the author critiques the conception of citizenship and belonging as defined by a policy organisation – and then proposes a definition that is more focused on the moral imperative for citizenship and belonging.

THE SOUTHERN REGION: qualitative research of teaching and learning

Down in the south of the intertopical map, we find topics 9 and 5. As you can see by Figure 4(below), both topics gain traction throughout IJLE’s lifespan. However, topic 9 increased its share of the total content very rapidly between 2003 to 2012, but since this peak has somewhat faded out in popularity relative to the other topics. Topic 5 has had a more sustained and steady growth throughout the lifespan of the journal, but it remains smaller than topic 9 looking at the full frequency of the tokenised words represented in Figure 2.

The semantic characteristics of topic 9 clearly illustrate the increasing interest in workplace learning after the turn of the century. The experiences of learners at work, especially specific groups such as women, workers, migrants and elderly, is a common research object within this cluster. In absolute terms, the most frequently used words are: learning, woman, practice, work, learner, process, community, and workplace. In relative terms, and compared to the other topics, words such as identity, worker and transformation are formulated more often here.

One example that aggregates in topic 9 is ‘Personal epistemologies and older workers’ by Billett and Woerkom (2008), in which the authors evaluate older workers’ needs and strategies to develop and deploy effective and critical personal epistemologies. The focus on personal epistemologies and critical reflection are discussed as a means for older workers to respond to the changing requirements of work and the labour market. Another article that typifies this line of research, according to the results derived from the topic model, is ‘Filtering informal learning in everyday life: invoking ordinariness and moving to civic engagement’ by Grummell (2010) focuses on informal learning from television watching. Grummell explores the role of informal learning within the ordinariness of daily life rather than in a work-based setting, and shares some interest with the scholarship in the Eastern region as she problematises television for reproducing established social order, rather than providing room for critical reflection on the power structures of everyday life.

A topic that increases steadily over time is topic 5. By the start of IJLE, it gathered some 5 per cent of all the tokenised words, whereas it grew to compose 15 per cent of the full-text corpus in the final years of 2020. The most frequent words for this topic are: experience, student, research, study, and learning. Compared to the other topics it is more focused on art and culture, as well as teachers and teaching.

One example of an article that scores high on matching the overall content in topic 5 is Ng et al. (2002) study ‘The teacher’s role in supporting a learner-centred learning environment: voices from a group of part-time postgraduate students in Hong Kong’. The study presents how 29 part-time adult learners perceive teaching and how this can best support their learning. The results indicate that there are important differences in how the group of students perceive teachers and tutors, but that a framework of learner-centred education is foundational to support students’ conceptions of learning. Another example of an article belonging to this topic is McLean and Vermeylen’s (2014) study ‘Transitions and pathways: self-help reading and informal adult learning’. In this article, they outline the results of interviews conducted with 134 readers of self-help books. Zooming in on the statistical distributions of how these books were read unravelled how the readers understood and narrated transitions and pathways in their lives as well as how these conceptualisations were gendered.

While exploring different contexts of formal and informal education, topic 5 focuses on students and teachers’ engagement in specific learning activities and outlines a specific outcome related to the focus group under scrutiny. It is similar to topic 9 in that a large part of the research is based on interviews centred on people’s experiences. But, while topic 9 is directing attention to workplaces and learning in non-formalised settings, topic 5 is composed of research with a stronger orientation to schools, teachers and students.

THE WESTERN REGION: surveying lifelong education

The western side of the territory is shaped by research on formal educational institutions, of student participation, motivation, assessments and so on. It is composed of one big and one smaller topic. The big one is topic 8, where the most frequent words are: study, student, participation, adult, school, and level. In more relational and relative terms it assembles most of the research dealing with participation, employment, as well as labour market outcomes . Verbs such as increase and motivate are popular, while institutional contexts of secondary and higher education, and the target groups of students, go to show that it is an educational research cluster with a clear emphasis on empirically mapping out different facets and factors of more formal and institutional environments. Contrary to most other topics, quantitative methods make a clear mark within this topic, with words such as factor, level and survey scoring high. This topic increases substantially in IJLE over time. Making up approximately 10 per cent of all tokenised words in the early years of 1982–1984, while increasing to around 20 per cent of all tokenised words in the last period under scrutiny.

An example of research that lies close to the inertia within this topic is Lee and Desjardins (2019) study ‘Inequality in adult learning and education participation: the effects of social origins and social inequality’, in which they used data from the Programme for International Assessment of Adult Competences (PIAAC) to compare inequality of participation in two different kinds of job-related adult learning. They relate these participation measures to social origins at the micro-level and three categories of social inequality at the macro-level (economic, education, and skill inequality), and use statistical methods to derive effects. Another example of an article that scores high in topic 8 is Støren and Børing’s (2018) ‘Training of various durations: do we find the same social predictors as for training participation rates’. This study examines numerous hypotheses about participation in non-formal training by frequency, drawing on PIAAC data for eight different European countries. PIAAC data allows the authors to compare countries as well as to break down participation rates between social groups within a given population.

Next to topic 8 in the Western region is topic 10, a smaller and more centrally located topic. The most frequent words here are higher education, policy, university, access, academic, assessment, student, and knowledge. In more relational and relative terms topic 10 assembles most of the research dealing with prior learning, higher education, access, assessment, academic, and prior learning. In sum, this topic focuses on higher education, but specifically on issues regarding participation in higher education.

One example of a study that has a high degree of match to topic 10 is Harris and Wihak’s (2017) ‘To what extent do discipline, knowledge domain and curriculum affect the feasibility of the Recognition of Prior Learning (RPL) in higher education?’. In the article, the authors focus on the relationship between academic knowledge and experiential knowledge in academia. They argue that RPL is more feasible in some professional academic programmes than others. Another example is D’Agostini and Titton’s (2020) ‘Youth in the context of chronic unemployment in Spain and Brazil’. Here, the authors analyse youth policies concerning the structural crisis of the capitalist system in Spain and Brazil. They argue that, despite fashionable policies, the structural crisis threatens the future of the young generation in these countries – countries where unemployment rates among youths are high.

NORTHERN REGION: organisational perspectives

In the Northern region three related clusters are aggregated: topics 1, 2 and 4. Topic 1 is located more towards the northwest, and topics 2 and topic 4 are placed in the northeast. Many of the titles in the North specify a geographical locality and societal need, in relation to which the studies are conducted. Research often deals with system-oriented problems and the provision of educational opportunities in a given context. Examples of research objects include community education, trade union education, human resource development (HRD) and literacy programmes. Policy-oriented and needs-driven forms of analysis are important facets of the northern part of the map.

Common words used within topic 1 are: programme, literacy, development, teacher, and training, where the first two (programme and literacy) are highly distinctive words for topic 1 in relation to the global map. One example is an article by Chuks (2004) titled ‘Literacy/numeracy and vocational training among rural women in Nigeria for a good livelihood and empowerment’, in which the author discusses the importance of women for economic productivity in Nigeria. With such importance of women, the author argues for the need for further knowledge of women’s literacy, numeracy, and vocational skills in rural Nigeria. Other articles scoring high on matching the content in this topic are Rachal’s ‘Measuring English and American historical literacy: a review of methodological approaches’ and Duke’s ‘Adult education, poverty and development’, both published in 1987. What unites these different contributions is an interest in adult learning and adult education as a means for poverty reduction and societal development. As the Western region was primarily characterised by an interest in school settings, participation and the formation of skills and competencies, it is reasonable that research on literacy is significant in this thematic cluster.

Still in the Northern region, but more oriented to the Eastern one, characterised by research on citizenship and political movements, we find topics 2 and 4. The overall term frequency of topic 4 generates the following set of words: policy, lifelong_learning, community, continuing_education, change, development, country, strategy, local, government, and system. In more relational terms it is characterised by a strong emphasis on policies and political reforms as well as continuing education and distance education. Examples that can be seen as typifying this discursive approach to the research territory include Boyadjieva and Ilieva-Trichkova (2018) study ‘Adult education as a common good: conceptualization and measurement’. This research was funded by a European Horizon 2020 scheme and built on cross-country data from 24 European countries in order to develop an index intended to measure the extent to which adult education fulfils the aims of the common good. Another example is Duke’s (1991)article ‘Revision and adaptation in postcompulsory education in Australia,’ in which the Australian educational system is discussed, focusing on policy documents and the transformation over time.

Topic 2 is related to the previous cluster in terms of its semantic content. The most common words here are: health, college, self, university, group, and professional. It gathers much of the research on health and professions but has an emphasis on college and schools, which might help to explain why the content is recognised as different from the articles in the Southern region. Common among the articles in topic 2 is that the empirical material foremost includes texts of different kinds as opposed to interviews and experiences that were prevalent in the South. One example is Evans (1983) article ‘Further education pressure groups: the Campaign for adult education in 1944ʹ, which focuses on the clauses of the British Education Act of 1944 that concerned adult education. Another example is Maruatona’s (1999) article ‘Adult education and the empowerment of civil society: the case of trade unions in Botswana’ that traces the history of trade unions in Botswana to discuss how adult education could address the workers’ problems.

Discussion: a text mining approach to the research territory

In the result section, we have presented an intertopical map based on regional representations of content derived from the large-scale textual analysis. As illustrated by this topic model, there are common denominators among the topics that lie close together, at the same time as there are differences among the latent thematic orientations in the distinct sections of the map. In this section, we turn to a meta-reflection about the results from a birds-eye point of view. First, we are looking at the semantic map of the territory in its multiplicity on an aggregated level. What are the main discursive strategies and scholarly traditions that have been unravelled by this latent topic analysis? Secondly, we discuss the kind of thematic changes that have occurred in the publications in IJLE over time. What research problems, epistemologies and perspectives have come to dominate, and what might help explain the changing contours of this research landscape?

A territory of correlated topics: pointing out the multiple ‘points of view’

When looking at the map on an aggregated level, we can see how topics in the Northern region are more oriented towards system-level analysis of policy, organisations and educational programmes. Analyses based on historical, nationwide and regional analysis are commonplace here, as are policy-oriented evaluations initiated by external political bodies like the UNESCO or the European Union (EU). By contrast, in the Southern region, we find topics focused on teaching and learning in a microcosmos, often located within workplaces or among a specific group of learners denoted as migrants, women, non-traditional students, etc. Contrary to the research in the North, research in the south of the territory tends to be formulated in close proximity to these groups and paying interest to their experiences and perceptions. To a large extent, this research in teaching and learning is drawing on qualitative approaches, especially interviews.

In the Western region, more topics of empirical and descriptive orientation are aggregated, whereas topics in the Eastern region are critical and normative in their outlook. The explorative and descriptive endeavours in the West tend to be in a quest for provincial and partial answers to specific research problems, whereas the problems addressed in the East are more built into the social order as a whole. The role of the critical research endeavour is theory-driven but contains both post-structuralist and normative approaches. In the descriptive-empirical region, certain methods make their mark, for example, by surveying patterns of participation and motivation (quant) or seeking to understand acts of recognition, perceptions, or experiences (qual). The critical paradigm is often weaker in its empirical grounding but has a strong affiliation towards theoretical and conceptual models derived from auxiliary disciplines such as philosophy or sociology.

In Figure 3 we have provided an overview of some of the distinguishing words in the form of a relational semantic map.

Figure 3. Relational map over the research territory based on some of the distinguishing words in the topics derived from articles published in IJLE from 1982 to 2021.

Figure 4. Frequency of the ten topics published in IJLE over the period 1982–2020 by year and per cent over time.

The rough illustration of the pertinent research clusters (Figure 3) that have surfaced in the correlated map of topics might be good for cultivating reflexivity. The variation of the map is the product of scientific divisions of labour within academia, and researchers would probably feel more or less affiliated to these various discursive camps depending on their place in IJLE as a pluralistic ‘thought collective’ (Fleck, 1935/1979). By elaborating in greater detail how the research territory of lifelong education, as represented in this specific journal, is shaped, we can hopefully trigger collegial discussion, debates and reflection regardless of where in the map one might place oneself and what camp one might identify with (cf. Bourdieu, 1988; Schön, 1987). Whether the analytic frame is based on individuals, interactions, organisations or social structures, the correlated semantic map reveals issues that tend to be addressed in conjunction within the dominating scholarly traditions that have shaped the journal.

In these reflexive deliberations, one would have to bear in mind that the contributions to IJLE are part of a bigger universe of possible journals whose status, circulation, orientation and thematic focus indirectly helped shape these results. For example, the kind of content that was published in IJLE and subsequently visualised in our topic model of 40 years, might look different in the later period due to the introduction of new journals such as Vocations and Learning (2008–) and The European Journal for Research on the Education and Learning of Adults (RELA) (2009–) as alternative publication venues for specific kinds of research. It is also worth remembering that most publications originate from the Global North and the anglophone universe, thereby reflecting the interest of research communities that themselves are situated both geographically and in social space.

One might think that the static and asynchronous depictions above provide an unfulfilling representation of the journal’s contributions, as it collapses all the words and semantics into a frozen time horizon. Research practices are far from static, especially research in lifelong learning that is rather faddish and loose in orientation. During the four decades of IJLE publications, important changes have occurred in the development and management of higher education as well as the very ‘ethos of capitalism’ (cf. Boltanski & Chiapello, 2005; Desjardins & Ioannidou, 2020). In the next section, we are going to discuss the changing thematic orientations based on which topics have increased and decreased, connecting it both to policy and their socio-historical condition of possibility.

Changing themes of the territory

Looking at the cartographic changes over time (Figure 4), we can see that the general pattern is a move of semantic contributions from the Northern to the Southern region, and from the eastern to the western end of the territory, i.e. an increase in the frequency of words among topics that are surveying rather than critiquing ‘the state-of-affairs’, and an increase in topics focused on teaching and learning as opposed to system-level issues such as educational provision, evaluation and organisation.

Looking closer at these thematic changes we note that topics 1 and 3 have decreased substantially over the 40 years under scrutiny. As shown above, these topics differ from one another in important aspects, but together they represent research aimed at political- and societal change. While topic 1 is based on terms of societal improvement through political reform, topic 3 is constituted by a more radical normative approach critiquing the state-of-affairs and the status quo through critical theory. In contrast to the general decline of these topics, we note that topics 5, 8, 9 and 10 have become more pronounced. What these four topics have in common is their empirical orientation, and a joint focus on participation, learners and students. The centre of attention conveys different objects – higher education (topics 5 and 10), the workplace (topic 9) or a wide range of educational institutions (topic 8) – but it is the grammar of applied and empirical research that surfaces here.

So how can we explain these thematic changes within the IJLE against the broader structural changes that have occurred within this time frame and in the conduct of lifelong education research? IJLE was established in connection to the expansion of adult education institutions in Western societies that took place in the 1960s and 1970s. Adult education was commonly framed within a humanistic discourse as something that would support the formation of humankind as well as societal change (Faure, 1972). Concerning such discourse, IJLE emerged as a forum for publications on ‘lifelong education’ – taking up the name of the policy discourse of the time. With a strong international ambition, as outlined in its early editorials (Editorial, 1982; Husén, 1982), it became an important outlet for ambitions aimed at changing society through developmental projects in less economically developed countries, as well as to formulate critique towards the state-of-affairs in more affluent increasingly post-industrialised societies.

The shift in the research territory from the strong humanistic ethos of the 1970s and early 1980s to empirical surveying of specific facets of lifelong education and learning, might also be connected to the changing landscape of education in the West during this time. In the 1980s and 1990s, we witnessed further development of adult education institutions. Not least the growth of higher education institutions, and demographic changes that affected participation in tertiary education (cf. Bourdieu, 1988; Husén, 1982). With a more diverse student population, issues regarding non-traditional students, motivations and access to education became common topics within lifelong education research in the Global North.

We can also benefit from situating the broad structural changes in thematic orientation, as published in the IJLE, within the wider landscape of policy changes taking place in the 1980s and 1990s. The shifts from lifelong education to lifelong learning marked a changing political perception of what education and learning of adults were good for (cf. Edwards, 2002; Fejes & Dahlstedt, 2014; Jarvis, 1987). Boltanski and Chiapello (2005) argue that the governing rationales in Western societies have left the bureaucratic and standardised form of Fordist mass production and increasingly been shaped by a management discourse of flexibility, future-oriented action, upskilling and project-based tasks. In this ‘new spirit of capitalism,’ companies organised themselves differently and the demands of workers to be constant learners, open to upskilling and reskilling became heavily emphasised.

Although these changes are far from new at this point and have arguably not been very abrupt in materialising, their significance for the research field of lifelong education shall not be underestimated. By introducing the notion of lifelong learning, a stronger focus was put on adults needing to continuously take part in learning, all the time, in different contexts, to stay up to date and be perceived as ‘employable’. Whether or not this should be seen as an expression of the new or ancient mode of governing can be debated, but it seems reasonable to see the rise and extension of ‘the learning imperative’ in virtually all facets of life as connected to these changes occurring in the dominant ideology and the transformation of the capitalistic ethos. Within the more modest claims that are visualised by the results from this study, we see a general trend from civic to work-related research themes, as well as a turn towards learning in a wider array of contexts. The most identifiable trend here is perhaps that topic 9, with its focus on qualitative descriptions of different facets of workplace learning environments, has increased significantly from the middle of the 1980s and has become one of the most common topics in the journal.³

The policy interest in the arrangements and politics of adult learning systems has also grown throughout the period we studied (cf. Desjardins & Ioannidou, 2020; Desjardins et al., 2006). Parallel to the developments described above, we have witnessed a change in academia as well as in the wider field of lifelong education. Public spending on higher education and adult education has increasingly come with certain ‘strings attached’. The discourse of evidenced-based and effective education has picked up steam, which for the terrain of adult education might mean that policy-oriented surveys on lifelong learning such as PIAAC warrant more legitimacy than classical issues of exploring, or advocating for, democracy, social rights and justice (see e.g. Fejes & Salling Olesen, 2010; Rubenson & Elfert, 2019).

The wealth of data created since the late 1990s, and especially in the last decade, for steering and managing policies and educational activities, also allows for comparisons to be made across countries and demographic groups, i.e. directly and indirectly facilitating empirical comparisons based on certain standards and benchmarks which are publicly sanctioned. While the wealth of studies contributing to topic 8 cannot be easily summoned, the expansion and growth of this topic in IJLE over time can probably be attributed both to the strength of quantitative research in surveying different aspects of participation and motivation within adult and higher education, as well as the ever-growing interest in evaluating and governing populations employing different management tools and policy instruments (Boltanski & Chiapello, 2005; Chiapello & Gilbert, 2019). The weakness of such research for policy is that it easily ends up reifying a rather limited analytical framework sanctioned by agents operating within the field of power (EU, nation-states, educational management), thereby lending itself to problematisations and easily accessible instruments which themselves should be critically scrutinised.⁴

Alongside these broader developments in the way lifelong education has been conceptualised, and the shifts in the political landscape, we also need to be attentive to the changing politics of scientific research in social science and humanities (SSH) as such (cf. Fleck et al., 2019). The observed moves from critiquing to surveying, and from researching system-level issues to the microcosmos of teaching and learning, might be related to the way research in lifelong education is financed and valued, and possibilities for career progression within universities and other research institutions. For example, over the last few decades, some countries have introduced or revised national professional evaluation systems for researchers to progress and be appraised in their academic career trajectories (e.g. the National Scientific Professional Qualification in Italy). An increasing number of countries have also revised national research evaluation systems for the distribution of public funds across higher education institutions (e.g. the Research Excellence Framework in the UK). Both qualification frameworks draw in part on measures based on citation and impact analyses to assess the quality and impact of research produced by individuals as well as institutions. Moreover, there has been an increase in competitive grants (public and private) to support SSH to be available from the national to the international levels, for example, at EU level through the creation in 2000 of the European Research Area. This development, coupled with a growing institutional demand on researchers to seek external funding to support their own research, might propel more empirical work within institutional settings of schooling, as it is more easily legitimised within dominating perceptions of ‘the common good’.

We also witnessed a growing trust in evidence-based claims to justify investment of public resources in research and development that emphasises the scientific, social and political impact of research, and the ways these can be evaluated. It is reasonable to assume that these processes may produce effects on the type of research problems, epistemologies and perspectives that lie behind the observed thematic changes within lifelong learning over time. Small-scale qualitative studies that explore the experiences and perceptions of teachers and learners in single institutional settings remain a viable option regardless, even if the researcher does not receive grants from the established research councils. The trends we have identified in producing small-scale empirical research might therefore be conceived as an accessible low-cost opportunity for many researchers to persevere and advance in their work and careers, in the wake of the heightened scholarly demands and the absence of other long-term economical and institutional affordances.

A final way to read these results is to ponder what phenomena and lines of research have remained unexplored, unthought and unpublished. Given the results emerging from our analysis: what have been the avenues of the unthinkable and underexplored? Which areas have remained underutilised and might take us to new and interesting horizons? One such area might be social scientific research on lifelong education mobilising the novel technological developments that occurred in data science, such as data mining of large-scale texts. In our own effort to bring new perspectives to the scientific territory that IJLE has been the outlet for, we have pursued one such metapragmatic approach based on large-scale topic modelling. In doing so, however, we should not lose track of the humanitarian ethos handed down to us by previous generations of scholars within lifelong education, as well as the inevitable quest to create something better tomorrow by critiquing the current.

Acknowledgments

We are grateful to the research assistant Erik Anders at the Department of Computer and Information Science at Linköping University who, in collaboration with Erik Nylander, helped integrate and analyse the bibliometric data in Python. Preliminary findings from this research were presented to a meeting of editors and editorial advisory board members of IJLE in June 2021; the authors wish to thank those who contributed to that discussion. We also received useful comments on early drafts of this article from the research seminar at the Division of Education and Adult Learning (Linköping University), and the two anonymous reviewers of IJLE.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. More on the mathematical basis of the model can be found in Blei et al. (2003).

2. In fact, based on these iterative processes between modelling and interpretation, topic modelling construction has been compared to ‘grounded theory’ (Baumer et al., 2017).

3. Hypothetically, the decline of this workplace learning topic in the journal after 2010 might be connected to the introduction of Vocations and Learning (2008–), which is a specific journal oriented to issues of workplace learning.

4. This scepticism towards sanctioned instruments and tools of governmentality should, of course, extend to the use and misuse of bibliometrics itself (cf. Gingras, 2014; Nylander et al., 2013).