Literature review

Even before the institutionalization of social sciences at the university level, several authors carried out extensive analyses of the bibliographic output of what is now called Brazilian social and political thought (^{Ramos, 1995}; ^{Santos, 2002}; ^{Fernandes, 1958}). However, this bibliography was more interested in identifying dominant approaches than in categorizing recurrent themes, although these are occasionally mentioned. Moreover, they looked at literature that predated the institutionalization of postgraduate programs in the country and the consolidation of institutional forms of academic research and writing. It was not until the late 1980s that more systematic typologies of the sociological research agenda emerged, accompanied by the growth, geographic dissemination, and thematic diversification of output in this discipline.

Roughly speaking, these typological efforts were animated by two kinds of concern. The first, common to a considerable part of the studies developed until the end of the 1990s, was interested in the processes of autonomization of social sciences and the types of connection built or destroyed between sociological and political research agendas. From this perspective, it is the responsibility of scholars to assess the degree of public commitment of local social scientists and the impacts of such commitment on the type of research conducted (^{Villas Bôas, 2007}; ^{Vianna et al., 1998}; ^{Melo, 1999}). The second type of concern, which has gained importance since the 2000s, assumes the autonomy of the academic field of social sciences as a premise, and focuses on its growth and state regulation, as well as on the effects of these processes on the disciplines’ research agenda (^{Leite, 2015}; ^{Leite & Codato, 2013}; ^{Maia, 2016}; ^{Simões, 2018}; ^{Marenco, 2016}).

The study of Glaucia Villas Bôas is a good example of the first type. Based on an analysis of the books published by humanities⁸ scientists between 1945 and 1966, Villas Bôas argues that that generation’s political concern with national development was the main determinant of the most studied themes and their transformations during the period. In addition to an increased production in book format, the bibliography produced in those years was divided into two fundamental approaches: one focused on the conceptual and theoretical problems of the disciplines included, and the other focused on broader problems of Brazilian society, the latter being more common in the analyzed corpus (^{Villas Bôas, 2007}, pp. 51-52). Another characteristic of this period is the marked preponderance of historiographical books, although it is also marked by the growing importance of disciplines such as political economy, sociology, and political science, which are more connected to the political issues emerging at the time (^{Villas Bôas, 2007}, p. 57). In the author’s terms:

It was thus perceived that the expansion of the social sciences corresponded to a renewal of texts that largely translated the specific problems of that time into their themes and analyses. Researchers questioned the country’s economic development, the chances of improving the lives of the population, the precarious living conditions in rural areas, and socioeconomic inequalities. These issues were discussed in literary and artistic circles, as well as in student circles. The renewal of social sciences was thus linked to a knowledge demand arising from the problems faced in those years. (^{Villas Bôas, 2007}, p. 213, own translation).⁹

Also, according to the author, the large increase in publishing companies, universities and their postgraduate programs, and in the possibility of cultural production in social sciences, led to an expansion of their intellectual circuit beyond São Paulo and Rio de Janeiro elites, towards more diverse social problems of the time. From this emerges a record of demands that would influence the area in the second half of the 20^th century in Brazil (^{Villas Bôas, 2007}, p. 186). While her more central conclusions concern the political content of the motivations underlying the themes of Brazilian social sciences from the 1940s to the 1960s, Villas Bôas also highlighted the centrality of an editorial and university infrastructure in the autonomization of research agendas in the field.

The independence of the problems addressed by social sciences in relation to public debate remains a fundamental concern of studies conducted in the 1990s by the group of researchers of the project “Social Sciences in Brazil”, led by Luiz Werneck Vianna. In one of the pioneering studies on the profile of social science doctoral students, ^{Vianna et al. (1998)} analyzed, among many variables, the general characteristics of the theses defended between 1990 and 1997 in seven postgraduate programs¹⁰ in sociology, anthropology, political science, and social sciences. These theses were manually categorized according to their methodology and objects of study. In this regard, a thematic distribution into 32 distinct object types was found, as described below:

Thus, of the 32 thesis objects listed in this table, nine of them concentrate 50% of the theses, namely: “culture” (27 theses), “studies of religion and churches” (26), “indigenous studies” (24), “agrarian studies” (24), “trade unions and workers” (24), “other societies” (21) . . . , “political attitudes, movements, and ideologies” (21), “public policies” (20), “gender studies” (19), totaling 206 theses out of 411 consulted. If we add the theses included under the headings “Brazilian thought” and “social science”, which in fact deal with related themes, i.e., how and with what instruments Brazil is studied, we would then have another thematic area of concentration of the theses produced in the period, comprising 22 of them. In this case, there would be ten thematic areas of the authors’ preference, with 228 theses included in those areas, making up 55.5% of the total research. Also, from a thematic point of view, the theses that make up this 55.5% collection can be considered from the following angle: 84 of them focus on objects unmistakably present in the modern Brazilian agenda, including “trade unions and workers”, “political attitudes, movements, and ideologies”, “public policies”, and “gender studies”. Moreover, by examining the keywords of the theses on “studies of religion and churches”, “agrarian studies”, and “culture”, we can see that 48 of these theses are also part of the modern agenda - eleven of them on “studies of religion and churches”, fifteen on “agrarian studies” and 22 on “culture” - totaling 132 theses. (^{Vianna et al., 1998}, own translation).¹¹

From the interface between this thematic composition and its relationship with the three researched disciplines, the authors deduce that there is a close relationship between Brazilian social sciences and the dilemmas of our modernization. Sociology and political science are responsible for studies in knowledge areas connected to the “drama of Brazilian democratic construction in the beings of its modernity”,¹² while anthropology was responsible for studying what “is backward in our society, especially the characters of rural life”¹³ (^{Vianna et al., 1998}, own translation). Thus, the specialization of this field did not occur at the expense of its connection with public debate, but rather the opposite. The agenda of social sciences therefore combined autonomy in relation to the waves of public opinion, yet without completely neglecting public opinion’s pressing issues.

However, the same was not true about the relationship between postgraduate studies and society’s administrative agencies: “a science that has managed to become so exposed to public opinion has, however, shown fragile relations with public administration agencies and market organizations” (^{Vianna et al., 1998}, own translation).¹⁴ Moreover, the number of researchers and scientific outputs in sociology has grown in a way that makes it complex to analyze it from a single perspective. The institutionalization of the social scientist career has become the phenomenon with the greatest effect on its scientific and organizational field. This scenario is important because it highlights the constitutive elements of the history of national social sciences, clarifying how scientific and organizational discourses have influenced their transformation over time and space. The relationship between society, the state and the scientific world also undergoes fundamental changes, showing the processes in which their knowledge interacts and reflects Brazil’s national development. Currently, there are 4,640 curricula registered on the Lattes Platform as related to the field of sociology,¹⁵ originating from doctorates in one of the three social sciences. This number has also grown since 2010, as shown in Figure 1:

Source: Authors’ elaboration, based on data from the Conselho Nacional de Desenvolvimento Científico e Tecnológico [National Council for Scientific and Technological Development] (CNPq).

Figure 1 Number of PhDs related to the area of sociology on the Lattes Platform 

From the study of Werneck Vianna and his team emerges the image of a highly institutionalized field, but one that seems to respond, at least in an impressionistic way, to the pressing concerns of national politics. However, although their declared motivation refers to the relationship (or lack thereof) between social sciences and public life, the collected data reveals little about this interface, indicating the emphasis of subsequent studies, restricted to the internal dynamics of studied disciplines. Perhaps for this reason, an important part of the thematic analyses of agendas was motivated by the initiative and sponsorship of professional associations, especially the Anpocs, SBS, ABCP, and ABA.

In 2018, the Revista Brasileira de Sociologia, SBS’ scientific journal, produced a dossier called “The Brazilian postgraduate system and the expansion of the field of sociology”.¹⁶ Among other texts that looked at aspects of the development of the discipline, the article by Marina Melo, Ana Cláudia Bernardo and Selefe Gomes (2018) was the only one to focus more on the content of sociological research, though mainly considering the identification of methodological characteristics and implementing a classification, for which they used SPSS software. The authors’ database included 282 theses and considered variables such as author gender, geographic origin, and program rating. One of the study’s main contributions was to demonstrate sociologists’ low attention to describing research methodologies, with an even worse result among women. From a thematic perspective, however, it is worth noting that, like ^{Vianna et al. (1998)}, they delimited criteria that were subjective and prone to dissent, without detailing them. Of the 17 categorized areas, the most frequent in the studied cases were “Intellectuals/Social thought/Theoretical theses”.¹⁷

More recently, Mariana ^{Nascimento (2021)} discussed the agendas of Brazilian sociology and its funding sources based on the research lines of ten of the country’s best-rated postgraduate programs. ^{Lima and Cortes (2013)} used a similar approach to discuss issues of interdisciplinarity in the field. The former author is one of the exceptions in the studies about the discipline at the national level, because she sought a more objective way of determining research themes, making use of resources such as word count in abstracts and project titles. Nascimento also points out that these new forms of subfield classification are essential, since the criteria used by CNPq are not sufficient to cover the diversification of sociological output. The study concludes that most research projects do not receive funding and are supported by volunteers. Guided by methodological techniques like those of ^{Nascimento (2021)}, the thesis by Tatiana ^{Maranhão (2010)} covers the scope and includes in the same type of analysis books and academic articles, with a focus on understanding whether there is autonomy in the perspectives of intellectual work with regard to the field’s internal and external constraints, such as state interests. According to this author, there is independent reflection among Brazilians, who nevertheless coexist with a certain provincialism by not seeking the internationalization of agendas or dialogues with foreign sociologists.

As for anthropology, Paula ^{Montero (2004)}, at the invitation of the ABA, sought to outline the thematic panorama of the field by comparing the research lines declared by postgraduate programs in reports sent to the CNPq and congresses and research groups in the area. Her basic intention was to identify “the main problems that organize the field of the discipline in the country, and how they have evolved over the last ten years” (^{Montero, 2004}, p. 117). Resuming a distinction made by Roberto Cardoso de Oliveira, she also questions whether national anthropological research remains internally distinguished by the researched object (traditionally grouped into the poles “indigenous” vs. “white”), or whether the different modes of theoretical construction of these objects have become a fundamental criterion for dividing its subfields. According to the author, on the one hand, the relative absence of references to empirical objects of anthropology in the labels of the compiled research lines leads us to contest Oliveira’s interpretation; on the other, however, the objects remain in the labels of the thematic groups at the main events of the discipline, indicating the existence of two modes of construction of anthropological problems (^{Montero, 2004}, p. 119). Thus, the consensus in the area is still “the emphasis on classical training and the need for fieldwork” (^{Montero, 2004}, p. 121, own translation).¹⁸

As for the observation of the recorded research groups, the author says it indicates three prominent themes: urban anthropology, anthropology of religion, and anthropology of family and gender relations (^{Montero, 2004}, p. 124). For her, this discontinuity between the themes of research lines, congresses and research groups thus evidences a certain disjunction between the organization of postgraduate education in anthropology and the topics that congregate research. Such a scenario, she says, results from a discipline expansion not accompanied by the consolidation of specific problems, which arises from an uneven and incomplete internationalization of academic dialogue (^{Montero, 2004}, p. 128).

Following Montero’s effort, Júlio Assis ^{Simões (2018)} resumes the analysis of research lines to investigate changes and continuities in Brazilian anthropology’s themes. Based on a list of over twenty categories, he concludes that research lines are widely dispersed across time and location, with only a few recurring in the more than 21 programs recorded in 2012. The exceptions are the programs comprising research in lines such as Identity, territory, and interethnic relations (15), Anthropology of politics (9), and Communication, art, and culture (9) (^{Simões, 2018}, p. 64). In general, most of the lines highlighted by Montero continued to appear in the program reports in 2012, though with the addition that “the cleavage between ‘indigenous ethnology’ and ‘anthropology of national society’, while still echoing in some intra-discipline tensions and quarrels, is not representative of the transformations that these major areas have undergone” (^{Simões, 2018}, p. 79, own translation).¹⁹

However, Both Montero and Simões recognize the problems of taking research lines as evidence of a research agenda, given the widespread practice of including into it professors, publications, and research related to distinct subfields (^{Montero, 2004}, p. 118). Moreover, such lines seem to reflect weak commonalities among the various faculty members of a program rather than its specific vocation. To assess this correlation, Velcimiro ^{Maia (2016)} contrasts the themes of the discipline’s bibliographic output with the research lines of postgraduate programs in social sciences, as filed with Capes. In all, he analyzed 583 documents on research lines and 547 qualified articles referring to the 2007-2009 and 2010-2012 triennial evaluations by the same agency. The author used natural language processing techniques such as lemmatization and clustering to join semantically similar documents and check whether there was an interface between research lines and published works. One of his results indicated that the subjects studied by the professors differ substantially from the descriptions of research lines from each program.

Another finding explored by ^{Maia (2016}, p. 123) is the stratification of sociology’s organizational field into a clearly defined center and periphery. This stratification is coercively reinforced through the state’s evaluation model via Capes, which leads peripheral groups to imitate its practices, thus mirroring the central groups in the field. In adopting the term isomorphism from ^{Dimaggio and Powell (1983)}, the author explains that, based on the institutionalization of an organizational field, constraint processes impose certain practices around the same legitimacy conditions. Peripheral institutions thus tend to participate in mimetic processes with regard to their research practices, even in opposition to the consolidation of their own research identity, in order to gain some legitimacy in the scientific field. ^{Maia (2016)} demonstrates that, in a study on scientific collaboration networks, (i) the endogeny of article publication strategies from the best-rated programs was maintained regarding the preference for journals from their own institution and (ii) peripheral programs, in their legitimacy techniques, copied actions from the leading programs, preferring to publish in their best-rated journals, even guiding the topics covered in their own outputs.

In this respect, Velcimiro Maia’s research addresses the problem of competition or, to use his term, “coopetition” - a mixture of collaboration and scientific competition - to think about the obstacles experienced by the area of social sciences in the country in its growth in recent years. Although its specialization in different thematic domains is well-known, processes of hierarchy and confrontation for resources also affect the organization of its research field, affecting the relationship between postgraduate programs and researchers themselves. In addition, the design of Brazilian postgraduate programs is reflected in discipline arrangements, making it more difficult to visualize which themes are becoming the focus of excellence in the field.

Unlike sociology and anthropology, Brazilian political science’s institutionalization occurred later, but was more intense and in line with the criteria proposed by the Capes evaluation system (^{Leite, 2015}). For this reason, its analysts have favored the investigation of scientific articles that are more identified with the area than its theses, books, and lines of research. This is the case of the article published by Fernando Leite and Adriano Codato (2013), focusing on the role of the Qualis-Capes system in the discipline’s growing methodological and disciplinary autonomy in relation to its counterparts. Though much smaller than sociology in number of programs, political scientists published more qualified articles than sociologists in 2009 (^{Leite & Codato, 2013}). Based on 364 articles from the six journals classified in the upper strata of the Qualis-Capes system, the authors identified six thematic areas: performance of political institutions; values, attitudes, participation and politics; political theory, analysis of concepts, and history of ideas; state, society and government policies; political communication, democracy, and electoral processes; and international relations (^{Leite & Codato, 2013}, p. 16). These insights are further developed in Leite’s (2015) doctoral thesis, which investigates the multiple dimensions of political science in its post-institutionalization and growth stage, focusing on its dominant approaches and thematic areas. Despite the various maps of the field drawn by the author, the research theme is almost always taken as an auxiliary independent variable for understanding other dimensions of the discipline, such as predominant methodological and epistemic approaches.

Under the auspices of the ABCP, André Marenco also analyzed the research agenda of political science through its production in academic articles, however, he focused only on one of the oldest publications in the discipline, the journal Dados (^{Marenco, 2016}). In the same book, organized by ^{Avritzer, Milani, and Braga (2016}), several other chapters use bibliometric techniques to examine the evolution of research subfields, usually categorized manually.

Also, in the field of political science, in an issue of the Brazilian Political Science Review, a journal published by the ABCP, Candido, Campos, and Feres Júnior (2021) used a programming package in R programming language to identify the most recurrent subjects in the discipline’s national journals, and whether there was a gender division in their authorship according to the most or least published themes. While the data showed that men and women proportionally tended to be linked to different research agendas, there was no sign of discrimination in the editorial processes. Feminized subfields were published as much as those with a male predominance.

Despite the profound differences, it is possible to identify some methodological gaps in parts of this bibliography. The most obvious of these is the rare use of computational techniques for thematic attribution to the units of analyzed material. Theses, articles, papers, and projects are almost always categorized as being in one thematic area by the completely discretionary decision of the research authors. It is therefore not surprising that thematic typologies of the same discipline vary so much from one study to another. More than a methodological detail, this absence has an impact on the construction of the typologies themselves and the conclusions therefrom. To overcome this, we use the Topic Modelling technique, like Candido, Campos and Feres Júnior (2021) did in political science. While it does not eliminate some discretion in the imputation of topics, it allows some technical control over the process, as we will explain in the next section.

Another important gap concerns the little attention given to discerning between consolidated and emerging themes. The predominance of some themes in each corpus or discipline often hides their variation over time. Historically consolidated themes may be undergoing marginalization, just as seemingly marginal topics may experience some growth. Therefore, the following sections take into account the chronological distribution of the most recurrent themes in sociology. In this way, we intend to identify not only consolidated and underestimated themes, but also emerging and marginalized topics.

Methodology

To identify the themes studied by Brazilian theses in sociology in the last eleven years, we used a resource for processing large volumes of text, namely Topic Modelling, and a corpus mining technique, in order to identify semantic patterns. This modelling is based on natural language processing, and consists in identifying the semantic similarity between texts from the division of groups of the terms that co-occur most frequently. As defined by Scarpa:

In this representation, a document is seen as a bag of words; that is, the positions of the words in the documents are not considered, only the number of times each word appears. . . . The sparse matrix derived from these texts allows using storage techniques that require considerably less memory, such as a keyword dictionary, with only non-null terms. . . . Latent semantic indexing (LSI) is a set of automated statistical procedures to quantitatively measure the similarity of meaning between two words or groups of words. (^{Scarpa, 2017}, pp. 9-11, own translation).²⁰

Topic Modelling operates in five steps. In the first stage, the analyzed documents are “cleaned out” in order to isolate only the terms with some meaning. In this sense, numerals, pronouns, terms with less than two letters, etc., are eliminated. In the second step, very frequent and very rare terms are excluded. This is necessary because neither of these types of words contributes to the definition of semantic patterns. In the present corpus, the term “thesis”, for example, is so ubiquitous that it would hardly help us characterize a corpus, the same applying to a rare term like “Amerindian”. A third step consists in isolating word radicals, and eliminating semantically irrelevant suffixes and prefixes.

Topic Modelling was performed using the Latent Dirichlet Allocation (LDA) package in R language. All abstracts were translated from Portuguese to English with the help of the Google Translate function in the Google Sheets application. Although Topic Modelling can be performed with texts in Portuguese, there are no good radicalization and lemmatization dictionaries for the language, unlike those available for English. Even though we know that the automatic translation of abstracts may raise semantic problems, given the chance that it does not respect the original meaning of some terms, we have good reasons choose this procedure. Firstly, virtual mass translation tools have evolved substantially in recent decades. Second, the most important thing for Topic Modelling is individual words, with little regard for possible nuances and synonyms.

To estimate the number of topics that would best divide the analyzed corpus, we used the density-based method via the LDA package, as proposed by ^{Cao et al. (2009)}, run in R programming language. In this method, an algorithm divides the corpus into distinct numbers of topics predefined by the user. The model assumes that the number of topics is excessive when most of them are characterized by many identical terms, and that the number of topics is insufficient when most of them are characterized by very specific terms. Thus, the optimal number of topics is defined when there are intersections between most of the topics, but they are not large.

Source: Authors’ elaboration, based on the model proposed by ^{Cao et al. (2009)}.

Figure 2 Adjusting the number of topics (X) to the analyzed corpus according to topic density (from 0 to 1) 

Originally, the database had 3,256 theses, but 66 of them were eliminated because they did not contain abstracts with the predefined minimum of 500 characters. The remaining 3,190 abstracts have, on average, 1,852 characters. In the present analysis, we set up a test with the number of themes oscillating between 5 and 50, and according to Figure 2, the optimal number can be said to be between 15 and 24, since these points are closer to the ideal adjustment (1.00). Strictly speaking, the number of topics closest to this was 18, and, therefore, we chose this division of the corpus.

General characteristics of the theses

As expected, most of the 3,190 theses analyzed are concentrated in the Southeast region of Brazil. Despite the increase in the percentage of theses defended in programs in other regions, especially in the South, it was not enough to drastically reduce the concentration in the Southeast, as shown in Figure 3. It is worth noting that this result should be contextually considered, as it was in the Rio-São Paulo axis that the first postgraduate sociology programs in the country were founded, and most of them are still operating in these cities.

Source: Authors’ elaboration, based on data from the CGEE.

Figure 3 Number of theses by program region and year of defense (2006-2016) 

The regional disparity is equally noticeable when we look at the educational institutions in isolation, as shown in Table 1. Almost all of the universities at the top of the ranking, which exceeds one hundred theses defended, are concentrated in the Southeast, with some also in the Northeast. There are a few exceptions, such as the Postgraduate Program in Sociology at the Universidade de Brasília (UnB), which is old in the country, the Universidade Federal do Paraná (UFPR) or the Universidade Federal do Rio Grande do Sul (UFRGS).

Table 1 Number of theses by postgraduate program (2006-2016) 

Postgraduate program	Quantity
Social Sciences - PUC-SP	447
Social Sciences - Unicamp	205
Sociology - USP	188
Social Sciences in Development, Agriculture and Society - Universidade Federal Rural do Rio de Janeiro (UFRRJ)	148
Sociology - UnB	139
Social Sciences - Universidade Federal do Rio Grande do Norte (UFRN)	134
Sociology - Universidade Federal do Ceará (UFC)	129
Sociology and Anthropology - Universidade Federal do Rio de Janeiro (UFRJ)	129
Sociology - Universidade Federal de Pernambuco (UFPE)	120
Sociology - Universidade Federal da Paraíba (UFPB)	120
Social Sciences - Universidade do Estado do Rio de Janeiro (UERJ)	111
Sociology - UFRGS	100
Sociology - Unicamp	97
Social Sciences - Universidade Federal da Bahia (UFBA)	95
Political Sociology - Universidade Federal de Santa Catarina (UFSC)	86
Sociology - Institute for Social and Political Studies (IESP)/UERJ/Iuperj	85
Sociology - UFPR	84
Sociology - Universidade Estadual Paulista “Júlio de Mesquita Filho” (Unesp)/Araraquara	80
Social Sciences - Universidade Federal de Campina Grande (UFCG)	60
Sociology - Universidade Federal de São Carlos (UFSCar)	60
Social Sciences - Unesp/Marília	58
Social Sciences - Universidade do Vale do Rio dos Sinos (Unisinos)	55
Sociology and Anthropology - Universidade Federal do Pará (UFPA)	53
Sociology - Universidade Federal de Minas Gerais (UFMG)	47
Social Sciences - UFPA	44
Social Sciences - Unesp/Araraquara	44
Social Sciences - Universidade Federal de Juiz de Fora (UFJF)	39
Social Sciences - PUC-MG	37
Political Sociology - Universidade Estadual do Norte Fluminense Darcy Ribeiro (UENF)	34
Sociology - Universidade Federal de Goiás (UFG)	32
Sociology and Politics - UFMG	29
Social Sciences - PUC-Rio	24
Sociology - Universidade Federal de Sergipe (UFS)	23
Social Sciences - UFSCar	15
Social Sciences - PUC-RS	12
Latin America Integration - USP	11
Sociology - Universidade Candido Mendes (UCAM)	8
Social Sciences - Universidade Federal do Maranhão (UFMA)	8

Source: Authors’ elaboration, based on data from the CGEE.

Dominant topics in the theses

As indicated by the estimation of the density of topics (see Figure 1), we chose to divide the corpus into 18 distinct categories. Table 2 lists the ten most recurrent terms in each topic, as well as the labels we assigned them based on those terms. Figure 4 presents the same list of topics according to their frequency in the analyzed content.

Table 2 Topics detected in the theses 

Topic	Label	Terms
1	Theory	theori; structur; concept; knowledg; scienc; chapter; part; scientif; food; assumpt
2	Culture	cultur; narrat; symbol; imag; music; popular; art; memori; artist; produc
3	Health	health; servic; data; inequ; care; access; factor; medic; incom; person
4	Work and economy	worker; market; capit; labor; econom; industri; compani; economi; union; cooper
5	Environment	develop; region; environment; econom; natur; local; actor; sustain; environ; innov
6	Methods	interview; interact; daili; agent; qualit; methodolog; particip; method; data; dynam
7	Crime and violence	right; law; violenc; polic; state; legal; control; crime; crimin; secur
8	Education	educ; school; profession; train; cours; teacher; student; higher; perform; academ
9	Race	movement; ident; nation; black; fight; recognit; action; mobil; indigen; racial
10	Generations	valu; age; old; good; home; class; materi; percept; consum; children
11	Urban	citi; space; urban; paulo; rio; resid; são; hous; place; center
12	Rural	famili; communiti; rural; region; land; territori; farmer; agricultur; state; migrat
13	Religion	religi; youth; young; religion; church; cathol; symbol; ritual; death; name
14	Gender	women; speech; gender; media; communic; sexual; men; categori; address; produc
15	Intellectuals	concept; critic; histori; literatur; intellectu; thought; theori; think; dialogu; read
16	Historical sociology	time; histori; part; open; transit; logic; american; half; evid; foreign
17	Democracy	polit; polici; parti; state; govern; class; elect; democraci; ideolog; disput
18	Public Policies	polici; public; state; manag; program; particip; implement; action; govern; municip

Source: Authors' elaboration, based on data from the CGEE.

Source: Authors’ elaboration, based on data from the CGEE.

Figure 4 Dominant topics most recurrent in the theses 

As can be seen in Figure 4, there is a preponderance of topics related to labor and economy (301 theses; 9.4% of the corpus), culture (274; 8.6%), and crime and violence (261; 8.2%). Among the less recurrent topics, we can highlight studies of historical sociology (71; 12.2%), methods (86; 2.7%), and generations (104; 3.3%). Despite the differences in methodology and focus, this is very similar to the data mentioned by ^{Vianna et al. (1998)} when they detected the preponderance of theses in the 1990s on what they called culture, trade unions and workers, and public policies. But it is the distinctions that are most striking, given the apparent emergence and growing centrality of themes such as crime and violence, absent from the that survey, or the apparent decline of studies on religion. At the same time, these results are also different from those of other studies we discussed above (^{Maranhão, 2010}; ^{Melo et al., 2018}). But despite these similarities and distinctions, it is worth noting the historical displacement of themes related to national development, which, as ^{Villas Bôas (2007)} indicated, dominated the agendas of sociological research between the 1940s and 1960s. Table 3 shows the chronological oscillation of each theme.

Table 3 Frequency of each topic over time (2006-2016) 

Topic	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016
Historical sociology	5%	1%	1%	2%	2%	3%	3%	3%	2%	1%	2%
Methods	1%	2%	2%	4%	2%	3%	2%	3%	3%	3%	4%
Generations	2%	4%	3%	2%	3%	4%	4%	3%	3%	4%	3%
Theory	3%	4%	8%	4%	3%	3%	5%	3%	4%	3%	5%
Health	4%	4%	5%	6%	4%	3%	2%	6%	6%	6%	4%
Race	5%	4%	5%	5%	6%	3%	5%	4%	6%	4%	5%
Environment	6%	5%	7%	2%	7%	5%	6%	7%	4%	5%	5%
Intellectuals	5%	5%	7%	7%	6%	6%	5%	5%	5%	5%	4%
Urban	7%	4%	4%	5%	5%	6%	6%	6%	5%	4%	6%
Education	6%	4%	4%	2%	6%	5%	4%	6%	7%	7%	7%
Religion	8%	6%	7%	8%	4%	8%	6%	5%	4%	5%	4%
Democracy	7%	5%	7%	8%	6%	5%	5%	3%	5%	6%	7%
Gender	4%	7%	7%	5%	5%	7%	5%	5%	7%	6%	6%
Rural	5%	6%	7%	8%	6%	4%	6%	7%	6%	8%	7%
Public policies	9%	6%	7%	4%	5%	6%	11%	9%	7%	5%	6%
Crime and violence	6%	9%	7%	7%	9%	7%	7%	9%	9%	12%	8%
Culture	5%	11%	4%	10%	10%	11%	9%	8%	9%	7%	10%
Work and economy	13%	14%	8%	12%	11%	12%	8%	8%	8%	8%	7%

Source: Authors’ elaboration, based on data from the CGEE.

Source: Authors’ elaboration, based on data from the CGEE.

Figure 5 Analysis of correspondences between postgraduate programs and topics 

Thus, the greater or lesser preponderance of a topic does not imply its consolidation in the discipline. As Table 3 indicates, the number of theses on sociology of labor and economics fell between 2006 and 2012, stabilizing thereafter. A similar trend, albeit less marked, is noted with religion. On the other hand, theses on education have experienced some growth in the analyzed period, as have those on crime and violence. However, in general, the temporal distribution of theses across topics is fairly stable. Some exceptions are the theses on public policies, which grow sharply between 2009 and 2012, decreasing thereafter until 2014. Topics such as gender and race are surprisingly stable in terms of theses defended per year, at around 6% and 5%, respectively. As we will see below, this trend may reflect the indirect incentives, equally continuous over time, for postgraduate programs to diversify thematically, which Velcimiro Maia called “institutional isomorphism” (^{Maia, 2016}).

The correspondence analysis, illustrated in Figure 5, shows the permanence of certain principles of thematic organization of doctoral programs in sociology and social sciences. On the horizontal axis, the opposition between urban and rural, although quite modified, continues to structure the thematic distinctions of the discipline. As shown in the map, the specialized nature of the Social Sciences Postgraduate Program in Development, Agriculture, and Society (CPDA) at the UFRRJ makes it the main producer of theses on the environment and ruralities. At the opposite extreme, but less specialized, are programs such as UERJ’s Social Sciences Postgraduate Program (PPCIS) and the sociology program at UFS, which are more focused on urban sociology. The vertical axis, in turn, opposes the programs more focused on the study of intellectuals, among which the sociology program at Unicamp is the most prominent, and the programs more dedicated to the study of religion, among which UERJ’s PPCIS appears again.

Apart from these key topics, there seems to be a convergence and thematic plurality of doctoral programs, as most of them are in the middle of the map and close to multiple themes. This seems to confirm Velcimiro Maia’s hypothesis of an institutional isomorphism of research lines and bibliographic output. According to him, Brazilian scientific policies, especially the Qualis-Capes system, has encouraged a double movement of convergence and uniform diversification of postgraduate programs’ research lines (^{Maia, 2016}, p. 138). In his view, the hierarchization of journals based on institutional exogeny criteria and the consequent financial and symbolic rewards arising from this system encourage programs to diversify the topics studied by their faculty in order to broaden their scope of publications, thus reducing the number of thematically specialized programs. It seems that this isomorphism is not restricted to the articles by the professors in Maia’s study, but also migrates to the supervised theses.

Conclusions

The aim of this article was to analyze the development of sociological research agendas in Brazil, focusing on more recent generations of academics, i.e., doctoral theses defended in specialized postgraduate programs in the country. The advantage of selecting these materials is that they are individually authored outputs and long-term investments. Articles in scientific journals, in contrast, may result from shorter research projects and dispersed collective collaborations. Based on doctoral theses, we were able to examine how the dimensions of teaching, research, institutional belonging, and scientific policies can be articulated in the definition of themes. The data we discussed reinforces the diagnosis that there is a greater diversification of subjects being explored in specialized postgraduate courses, which seems to come from institutional stimuli from evaluation and development agencies.

In addition, one of the contributions that we postulate in dialogue with the literature concerns the research design. We adopted the Topic Modelling technique, considered the oscillations of the studied themes over time, and performed correspondence analyses to check the weight of the institutional variable on the choices and perspectives defended in the theses. This set of options allowed us to render a more objective and neutral portrait of what is studied in Brazilian sociology, capturing its contemporary transformations. Over the course of a decade, some research agendas lost importance while others gained relevance. At first, the great dispersion in approaches stands out. Culture is the theme that gains the most attention in the latest period (10% of theses in 2016), with historical sociology occupying the opposite extreme (2%). Traditional topics such as labor and economics have lost ground. Gender and race, pressing issues in public debate and expanding in the social sciences, have consolidated in the agenda.

The data indicates that Brazilian sociology has acquired characteristics of what Craig ^{Calhoun (2007)} calls a “professionalized science” with high specialization and thematic diversification. But unlike the American example analyzed by the author, the scenario here seems to reflect rather the effects of scientific evaluation and promotion agencies (^{Maia, 2016}), which have recently taken on the task of guiding national scientific development.

While this rapprochement between state regulatory agencies and academia seems to have significantly reshaped the field, the dialogue between the two seems problematic. ^{Nascimento (2021)} and ^{Lima and Cortes (2013)}, among others, show that the formalization of areas in bodies such as Capes and CNPq reflects little of what the discipline produces, which impacts the strategic planning of the area and the still fragmentary and poorly structured character of its funding.

The morphology of postgraduate programs seems to contradict the thematic pluralization of their theses. While these are increasingly structured around diverse themes, the programs are undergoing a process of strong standardization (^{Maia, 2016}). In addition, important methodological deficiencies remain, highlighted by several of the studies discussed here. Among the topics we have listed, the low recurrence of approaches to research methods is notorious. This gap among sociological theses defended in the country corroborates the results of the study of ^{Melo et al. (2018)}, which highlights the low overall concern of the Brazilian academic community with a more accurate description of the methodologies of their investigations in the discipline. This problem, however, is not restricted to this area of social sciences and has also been frequently pointed out as a major challenge in political science (^{Soares, 2005}; ^{Neiva, 2015}).

The doctorate is the highest degree of training in higher education, and in this study we cover a whole decade of outputs completed by this stratum of researchers in Brazilian sociology from 2006 to 2016. After the respective period, the country and the rest of the world went through circumstances that would radically change the interactions and living conditions of several populations, with the outbreak of the covid-19 pandemic and the occurrence of climate disasters that are increasingly common in the face of global warming. As hypotheses of research agendas that may grow in sociology due to the most recent situation, we emphasize, for example, health and the environment.

Sociology has grown academically over the last century in Brazil, having approached and adapted strongly to the agencies of regulation and professionalization of science. The multiplication of studies about the discipline, including its history and morphology, attests to its high degree of institutionalization, but not only that. The strengthening of this field of knowledge can be observed through different methods and variables, which help us situate the area within broader political and social contexts. Since it became a specialized sector, sociological analyses have coexisted with left and right-wing governments, military coups, and the electoral rise of an authoritarian administration in the executive branch.

Despite these changes, which often had direct negative consequences for scientists’ work routines (^{Chaguri et al., 2023}), the structuring of the national postgraduate system and its evaluation policies seem to be contributing to a certain stability of thematic pluralization in research agendas. On the other hand, however, the regulatory and funding agencies themselves do not seem to recognize the thematic form that the discipline has taken. Another limit to this specialization process, as we have pointed out, is the relative underdevelopment of methodological discussions, which has been pointed out by other authors, and which we have shown to be evident in the small space that this theme has occupied in the list of categories presented here. Thus, there are achievements, but also future challenges for the discipline.