You are on page 1of 20

Computers & Education 137 (2019) 12–31

Contents lists available at ScienceDirect

Computers & Education


journal homepage: www.elsevier.com/locate/compedu

Exploring two decades of research on classroom dialogue by using


T
bibliometric analysis
Yu Songa, Xieling Chenb, Tianyong Haoc,∗, Zhinan Liua, Zixin Lana
a
School of Education, South China Normal University, No.55 Zhongshan Road, Tianhe District, Guangzhou, Guangdong Province, 510631, China
b
College of Economics, Jinan University, No. 601 Huangpu Avenue West, Tianhe District, Guangzhou, Guangdong Province, 510632, China
c
School of Computer Science, South China Normal University, No.55 Zhongshan Road, Tianhe District, Guangzhou, Guangdong Province, 510631,
China

A R T IC LE I N F O ABS TRA CT

Keywords: Classroom dialogue is a commonly used method for teaching and learning, and a close study of
Classroom dialogue dialogue has increasingly become an active field of research. In order to have a comprehensive
Bibliometric analysis overview of the field, a bibliometric analysis was conducted on 3914 papers published from 1999
Visualization to 2018 retrieved from WoS database in relation to classroom dialogue. Specifically, we analyzed
Research hotspots
trends in publications and citations, recognized prolific authors, institutions and journals,
identified geographical publication distributions, visualized the characteristics of collaboration
among authors, institutions, and countries/regions, as well as revealing the evolution of themes
over the past 20 years. Findings include, firstly, the fact that publications and citations in relation
to classroom dialogue have grown consistently over the past 20 years. Secondly, the USA has
contributed dramatically more publications, especially since the year 2012. Thirdly, scientific
collaborations in perspectives of country/region, institution and author can be explored by ac-
cessing the dynamic social networks. Fourthly, thematic features in relation to research on
classroom dialogue were revealed by analyzing keywords, with several recurring keywords being
identified throughout the period (e.g. ‘classroom’, ‘discourse’, ‘student’) and at the same time,
new keywords have emerged (e.g. ‘technology’, ‘computer-mediated communication’), which
reflect the shifting trends in the field. This work is useful in terms of indicating the current status
of research to scholars as well as practitioners, enabling them to be more aware of the research
hotspots when making decisions about which topic to address.

1. Introduction

Classroom dialogue, defined as communication in classroom settings in which “one individual addresses another individual or
individuals and at least one addressed individual replies”, is a commonly used method for learning and teaching (Howe & Abedin,
2013). The close study of classroom dialogue has increasingly become an active field of research since the 1970s (Howe & Mercer,
2017). Interest in understanding and improving the quality of dialogue is higher than it has ever been, not only within academia but
also amongst teachers in schools, and potentially in governments (Howe, 2017). There are reviews of the topic (e.g. Gillies, 2016;
Henderson, McNeill, González-Howard, Close, & Evans, 2018; Howe & Abedin, 2013), yet, to our knowledge, none of them has been
based on quantitative methods. Bibliometric analysis is a set of quantitative methods used for the study or measurement of texts,


Corresponding author.
E-mail addresses: sungyuepku@foxmail.com (Y. Song), shaylyn_chen@163.com (X. Chen), haoty@m.scnu.edu.cn (T. Hao),
273989225@qq.com (Z. Liu), 1837451683@qq.com (Z. Lan).

https://doi.org/10.1016/j.compedu.2019.04.002
Received 26 January 2019; Received in revised form 30 March 2019; Accepted 1 April 2019
Available online 11 April 2019
0360-1315/ © 2019 Elsevier Ltd. All rights reserved.
Y. Song, et al. Computers & Education 137 (2019) 12–31

especially in the form of big datasets (Zupic & Čater, 2015). It is useful to generate/construct literature structure, assess the im-
portance of publications and identify a research topic's trends and evolution (Gimenez, Salinas, & Manzano-Agugliaro, 2018). By
using bibliometric analysis, this paper aims to present a comprehensive investigation of the scientific literature across the globe that
relates to classroom dialogue.

1.1. Research on classroom dialogue

Classroom dialogue is often regarded as crucial for pedagogy to be effective (van der Veen & van Oers, 2017), and through talking,
students are expected to gain access to diverse ideas, gain practice in critical thinking and deepen understanding (Mercer & Dawes,
2014). This may consequently be beneficial in terms of achieving satisfactory learning outcomes (Dallimore, Hertenstein, & Platt,
2010; Resnick, Asterhan & Clark, 2015). In practice, challenges exist to establishing dialogic pedagogy and to bringing its role into
full play, which results in generally encouraging as many students as possible to talk in class and improving teachers’ skills in relation
to conducting dialogic teaching effectively (Howe & Mercer, 2017).
Works on classroom dialogue are usually warranted with reference to the work of Vygotsky (1978), who proposed a socio-cultural
theory bridging the relationship between thought, action, communicate on and culture. There are several scholars who have con-
ducted reviews on classroom dialogue and relevant topics (e.g. classroom talk), integrating what is currently known and describing
the development of this field. These reviews have normally been conducted by taking a meta-analysis approach. For example, Howe
and Abedin (2013) systematically reviewed research conducted over a period of 40 years on classroom dialogue amongst primary and
secondary students. Mercer and Dawes (2014) reviewed studies on classroom talk between teachers and students, from the 1970s
until the 2010s. These reviews present a clear picture of the main areas of research in the field, as follows.
Patterns of classroom dialogue. Verbal exchanges are the focus of the research in relation to classroom dialogue, nevertheless the
role of listening has started to receive attention (O'Connor, Michaels, Chapin, & Harbaugh, 2017). Teacher-student interaction and
student-student interaction are the two main types of classroom dialogue, and they have both been studied extensively (Howe &
Abedin, 2013). A three-step pattern, initiation-response-feedback (IRF), as illustrated by Sinclair and Coulthard (1975), is taken to be
the dominant pattern within classroom dialogue (Howe & Abedin, 2013; O’Connor & Michaels, 2007). Initiation means that one
individual, usually the teacher, starts a dialogue by posing a question or raising a topic. Then responses are given, usually by students.
Finally, initiators provide feedback on these responses, which may be further divided into short judgments (e.g. ‘good’, ‘interesting’)
or detailed evaluations (Wiebe Berry & Kim, 2008; Schwab, 2011). Another main contribution to sorting out dialogic patterns has
been made by Mortimer and Scott (2003), who generated four types of communicate approaches: ‘authoritative interaction’, ‘dialogic
interaction’, ‘authoritative non-interaction’ and ‘dialogic non-interaction’. As indicated by Howe and Abedin (2013), the first three
approaches are consistent with the IRF pattern.
Predictions of variability in dialogue. A sizeable body of research has focused on charting factors that predict variations in classroom
dialogue (e.g. Burns & Myhill, 2004). A variety of factors have been found to be related to classroom dialogue, including student
factors (e.g. attainment, age, ethnicity, gender, personality), teacher factors (e.g. age, expectation, experience, gender, pedagogy,
personality) and environmental factors (e.g. class size, group size, resources, curriculum, school type, seating, culture) (see Howe &
Abedin, 2013). Of these factors, gender, ethnicity and attainment are the most extensively studied (Howe & Abedin, 2013).
Methodology issues. The discussion of whether using a quantitative approach, a qualitative approach, or both, has aroused much
attention. The approaches have their strengths and weaknesses, and tensions between them continue to influence developments in
the field (Howe & Mercer, 2017). Quantitative methods are those using coding schemes to capture the countable features of talk and
behaviors, among which systematic observation is most frequently used; nevertheless, computer-based text analysis is popular be-
cause of its ability to measure patterns of language use (Mercer, 2010). Using quantitative findings, researchers can look for asso-
ciations between the relative incidence of dialogic features and other variables. The qualitative approach usually attaches to eth-
nographic and sociolinguistic traditions, and researchers analyze by recording, transcribing and coding with careful examination
(Mercer, 2010). By using the method, it is possible to reveal the nature and pattern of exchanges, and examine how the structure and
content of dialogue develops (Mercer & Dawes, 2014). The adoption of detailed coding has been quite popular during the past five or
six years, especially in Britain. Nevertheless, there are some reflections that indicate the difficulty involved in using it in large-scale
studies. Hennessy et al. (2016) developed a Scheme for Education Dialogue Analysis (SEDA) which clustered 33 codes to describe
dialogic sequences, and Howe and Hennessy (in press) proceed to refine the codes into 10 key categories.
Assessment of classroom dialogue. Model-based assessment of classroom dialogue has actually become more popular recently
(Howe & Mercer, 2017). Many scholars have put forward persuasive arguments for its use and there is evidence from small-scale
studies that suggests the importance of classroom dialogue when it comes to developing students’ reasoning and thinking, and
understanding curriculum subjects, especially mathematics and science (e.g. Alexander, 2001; Resnick, Asterhan, & Clarke, 2015). It
is explained that through dialogue, students not only embrace diverse ideas, but also gain practice in thinking through problems and
organizing concepts, formulating arguments and counterarguments, and responding thoughtfully and critically to diverse points of
view (Mercer & Dawes, 2014). Nevertheless, only a small number of studies have attempted to evaluate dialogic practices in a target-
based fashion, and most of these have focused upon small-group interaction amongst students (Howe, 2017). There is scant evidence
to show how classroom dialogue affects the performance scores that student actually obtain (Howe et al., 2015). A reuse of quan-
titative methods is called upon to assess whether some modes of dialogic organization are more beneficial than others, whether the
verbal exchanges viewed as having high value are indeed significantly associated with positive outcomes (Mercer & Dawes, 2014).
Recognizing the significance of the previous reviews, it should be noted that there are important issues uncovered, or at least not
elaborated sufficiently. For example, what are the development trends in the field? Which scholars and institutions are active in, and

13
Y. Song, et al. Computers & Education 137 (2019) 12–31

contribute more to this field? What are the focuses and hotspots in relation to classroom dialogue? To solve these questions, re-
searchers must be more open to new methods, for example, those used in computer science.

1.2. Bibliometric analysis and its application in the field of education

Bibliometric analysis is a useful tool to map the literature concerning a specific research area (Falagas, Karavasiou, & Bliziotis,
2006). There are many definitions of bibliometrics. It is defined as a set of methods used for the study or measurement of texts and
information, especially in the form of big datasets (Cobo, López-Herrera, Herrera-Viedma, & Herrera, 2011). Zupic and Čater (2015)
affirm that bibliometrics is a tool for analyzing the evolution of disciplines on the basis of the intellectual, social, and conceptual
structure. It utilizes different methodologies in order to study the evolution of, and trends in, a research topic as well as assessing the
relative importance of publications in a specific research field (Gimenez et al., 2018). As indicated by Lewison and Devey (1999),
bibliometrics is to scientific papers what epidemiology is to patients. This can be helpful for understanding the discipline of bib-
liometrics as well as its significance.
Not only can bibliometrics facilitate research retrospectives, it can also help in the exploration of research hotspots and devel-
opment trends in disciplines, both quantitatively and objectively. The results from bibliometric analysis contribute to progress in and
advancement of, a certain research area in many different ways. On the one hand, it allows progress made to be assessed, identifies
the most reliable and popular sources of scientific publication, recognizes major scientific actors, such as author and institutions, lays
the academic foundation for the evaluation of new developments, identifies emerging research interests and predicts future research
success (Geng et al., 2017; Mazloumian, 2012). On the other hand, it helps researchers pinpoint their potential research topics,
recognize the appropriate research institutions with which to conduct joint work, and identify potential academic collaborators
(Martínez, Cobo, Herrera, & Herrera-Viedma, 2015).
Many studies have applied bibliometric analysis in order to present a general overview of a research field, especially an inter-
disciplinary research field, and has been shown to be an objective, reliable, and cost-effective measurement of academic outputs
(Campbell et al., 2010). Among these bibliometric studies, it is worth mentioning some key areas, such as fuzzy research (Merigó, Gil-
Lafuente, & Yager, 2015a), residential energy consumption and greenhouse gas emissions (Geng et al., 2017), applying artificial
intelligence to electronic health records (Chen, Liu, Wei, Yan, Hao & Ding, 2018a), event detection in social media (Chen, Wang,
Tang, & Hao, 2019), natural language processing in medical research (Chen, Xie, Wang, Liu, Xu & Hao, 2018b), and text-mining in
medical research (Hao, Chen, Li, & Yan, 2018).
Focusing on education research, a few bibliometric studies are available, among which some representative works are as follows.
They normally center on bibliometric analysis of education in certain disciplines, such as nursing education (Hunt, Jackson, Watson,
& Cleary, 2013), medical education (Sampson, Horsley, & Doja, 2013), engineering education (Xian & Madhavan, 2014), and STEM
(i.e. Science, Technology, Engineering, and Mathematics education) (Assefa & Rorissa, 2013). Bibliometric approaches have also been
adopted in a range of settings, for example, higher education (Budd, 1988; Kosmützky & Krücken, 2014), adult education (Fejes &
Nylander, 2014), open educational resources (Zancanaro, Todesco, & Ramos, 2015), and federal spending on education research
(Milesi, Brown, Hawkley, Dropkin, & Schneider, 2014). One representative work involved the applying application of science
mapping and performance analysis methods to the analysis of literature on virtual and remote laboratories in education (Heradio
et al., 2016). They identified the most influential publications, the most researched topics, as well as the evolution of topic interest in
the field.

1.3. Research aims

For the purpose of this study, namely to present a general and comprehensive overview of classroom dialogue research, bib-
liometrics is a suitable approach. Currently, no bibliometric study in relation to literature on classroom dialogue is available. To that
end, this paper presents a bibliometric analysis of scientific literature on classroom dialogue at the first time with specific objectives
as follows: first, to identify the development trends of publication and citation across time in the field; second, to recognize the
prolific journals, authors, and institutions; third, to identify the geographical distribution of the prolific countries/regions; fourth, to
reveal the characteristics of scientific collaboration among countries/regions, institutions, and authors; finally, to discover the
evolution of the most frequently used keywords across time.
This study contributes to the research field from several aspects. First, it provides education scholars with a comprehensive
understanding of the research status and development of classroom dialogue. Second, it helps education scholars identify the authors,
institutions and countries/regions with the most potential to develop and share classroom dialogue research. Third, it enables
education scholars to be more aware of the research hotspots when making decisions about which topics to investigate. Fourth,
researchers and students are able to identify the most influential journals dealing with classroom dialogue and with an international
scope. Furthermore, this research helps companies and governments identify which R&D centers to be financed in order to conduct
scientific and technological activities.

2. Dataset and method

2.1. Data retrieval and preprocessing

In conducting data retrieval, one critical procedure or challenge was to identify search terms. Popular as classroom dialogue is,

14
Y. Song, et al. Computers & Education 137 (2019) 12–31

people have not reached a consensus over its terminology and conceptualization (Howe & Mercer, 2017). Besides ‘classroom dia-
logue’, terms such as ‘accountable talk’, ‘exploratory talk’, ‘dialogic teaching’, and ‘classroom discourses’ also appear in the literature
(Hennessy et al., 2016). The choice of terminology is sometimes arbitrary, and there is a lack of distinction; different terms are used
even though many of them are intended to describe similar phenomena (Howe & Mercer, 2017). One challenge was to maximize the
identification of studies that addressed dialogue in class as opposed to other settings. Classes can be conducted either on campus
within in traditional classrooms, or off campus such as through online classes. There are two types of dialogue, those that take place
between teacher and students in whole-class settings, or those that occur between students themselves in small groups. In the end,
‘classroom’ and its two associated terms ‘whole-class’ and ‘small-group’ were taken as one set of keywords in the course of the
literature search. Another challenge was to search the literature in a fashion that would maximize the identification of studies that
were consistent with our definition of dialogue. Howe and Abedin (2013) used 21 associated terms (i.e. answer, argumentation,
communication, conversation, dialogic, dialogue, discourse, discussion, feedback, ground rules, interaction, interactive, IRE, IRF,
language, oracy, question, reciprocal, recitation, speaking and listening, talk, turn taking) to conduct the search in their review. The
terms were identified by examining reference lists of items and iterating the process until no new ones emerged. We intended to use
these terms in literature retrieval, nevertheless tens of thousands of articles resulted and many of them were not relevant to dialogue
research, but happened to contain one of the terms in titles or abstracts. In this case, we decided to employ the Delphi method to cue
terms associated with dialogue. Four relevant domain experts were invited to list any associated terms of which they were aware, and
then they were encouraged to refine their earlier answers in light of the replies of other members. In the end, ten terms were
identified as the other set of search keywords: dialogue, dialogic, discourse, conversation, discussion, language, interaction, talk,
communication and speaking.
Although the study of classroom dialogue started in the 1970s, it is from around the year 2000 that the field has experienced rapid
growth (Howe & Abedin, 2013). Thus, it seemed reasonable to examine two decades of publications, namely, the years 1999–2018.
As Web Of Science (WoS) is the most important bibliometric database (Cobo, Martínez, Gutiérrez-Salcedo, Fujita, & Herrera-Viedma,
2015), raw bibliographic data concerning the area of research on the classroom dialogue were retrieved from it with the following
advanced query, in which “TS” (Topics), referring to the title, abstract or keywords of a publication, was used as a search field. The
specification was for any publication with which at least one keyword from each of the two sets was associated. The retrieval was
conducted on January 1, 2019.
TS=((‘classroom’ OR ‘small-group’ OR ‘whole-class’) AND (‘dialogue’ OR ‘dialogic’ OR ‘communication’ OR ‘conversation’ OR
‘discussion’ OR ‘discourse’ OR ‘interaction’ OR ‘talk’ OR ‘speaking’ OR ‘language’))
Publications retrieved from the query were considered if they were: 1) articles, since research articles usually provide more
original research findings and give more specific information on authors and their affiliations (Geng et al., 2017). 2) in English, but
did not have to be conducted in English-speaking countries. 3) published between 1999 and 2018; 4) in a Web of Science Category
with ‘education’ or ‘educational’ in the name; and 5) indexed in Science Citation Index Expanded (SCIE) and Social Sciences Citation
Index (SSCI). Based on the above restrictions, 9880 publications with full bibliographic information as well as citations published
between 1999 and the end of 2018 were downloaded in plain text. A Python program was used to extract key elements, for example,
title, abstract, journal, year published, and author address of each publication from the downloaded raw data which were entered
into Excel for further processing.
In order to ensure the efficiency and reliability of the analysis, data filtering and preprocessing work were conducted. First,
replicated data were deleted by checking the title, journal, author address, and year of publications. Second, to guarantee the high
relevance of the data to the research target, a filtering exercise was carried out by two researchers based on the criteria listed in
Table 1. The two researchers piloted filtering by assessing 200 articles and the inter-rater reliability was 94%, and the inconsistencies
were fully discussed before reaching agreement. They then assessed and filtered the rest of the articles in division. According to the
criteria, 3914 publications were selected for the analysis of the final data set. Further, bibliographic information concerning each of
the 3914 publications was confirmed and recorded according to the original articles, especially the author address information.
Finally, the names of authors, institutions, and countries/regions extracted from the author address information were recorded
manually to ensure they were expressed in a consistent fashion since inconsistent expressions are common. After the data filtering
and preprocessing, we finalized the 3914 publication records for analysis.

Table 1
The inclusion and exclusion criteria for manually verifying the retrieved publications.
Inclusion criteria I1 Talk
I2 Textual verbal exchanges
I2 Non-verbal behaviour (including paralinguistic and prosodic features) was only of interest when it supported language
I3 Computer-mediated communication
I4 Dialogue carried out in distance or online learning
Exclusion criteria E1 Not focused on classroom dialogue or associated terms
E2 listed above
E3 One individual's commentary on contrasting ideas
E4 One individual's assertion of a single, received idea
E5 One individual's speech
E6 Dialogue happened outside class

15
Y. Song, et al. Computers & Education 137 (2019) 12–31

2.2. Methods

2.2.1. Bibliometric indicators


Several bibliometric indicators were used to assess the influence of the publications. First, publication and citation counts were
considered since this is the most popular bibliometric method (Ding, Rousseau, & Dietmar, 2014). They measure productivity and
influence, respectively (Svensson, 2010). Second, the Hirsch index, namely H-index was used to evaluate the academic level from
both quality (citation count) and quantity (publication count) perspectives (Peng et al., 2018). Initially, the H-index was an author-
level metric attempting to measure both the productivity and citation impact of the publications of a scientist or scholar1. It is now
also widely applied to the productivity and impact of a scholarly journal, country/region, institution, and author. This is defined as H,
meaning that an individual's publications have at least H citations each (Hirsch & Buela-Casal, 2014), whereas a higher H-index value
reflects a higher level of scientific achievement. Third, three citation thresholds (citation count ≥ 300, > =100, and ≥ 50) were
utilized to identify the publication count with a certain degree of influence by achieving a specific threshold (Merigó, Mas-Tur, Roig-
Tierno, & Ribeiro-Soriano, 2015b).
Furthermore, with regard to the measurement of journal influence, indicators such as journal impact factor, JCR quartile in
category (Q), five-year impact factor (5-year IF), SCImago journal rank (SJR), impact per publication (IPP), as well as source nor-
malized impact per paper (SNIP) were also applied. The impact factor (IF) of an academic journal is a measure reflecting the yearly
average number of citations of recent articles published in that journal2. For a specific year, IF was assessed by dividing the total
citations of publications from a journal published during the preceding two years by the total publications published in that journal
during the same period (Geng et al., 2017). In this study, IF derived from the Journal Citation Reports 2017 was applied. A five-year
impact factor, calculated by dividing the citation count for a journal in a given year by the total publications published in that journal
in the previous five years, was also included. Q measures a journal's relative position in a certain JCR category (Chen et al., 2019).
Since a journal belonging to more than one JCR category may have different Q values, in this study, the highest Q value was
considered. As a Scopus-based indicator, rather than one that regards all citations as equal, SJR measures weighted citations received
by a journal, where the weighting of citations depends on the subject fields, quality, and public reputation of the cited journal. IPP,
which also refers CiteScore, is the average number of citations in the previous three years dividing all publications published in those
years3. As the normalized version of IPP, SNIP measures a journal's contextual citation impact by weighting citations on the basis of
the total citations in a subject field, helping to make direct comparisons between journals in different subject fields4.
Finally, with regard to measuring the influence of institutions, three of the most widely read university rankings in the world,
namely the 2018 Academic Ranking of World Universities (ARWU), the QS World University Rankings 2018 (QS), and the World
University Rankings 2018 (THE) are also used.

2.2.2. Geographic visualization method


Geographic visualization is a set of techniques which support geospatial or spatial data analysis. With the combination of various
technologies, such as image processing and virtual reality, computers can help present information in a way that better reveals
patterns. Location is usually used as the key index variable, and through visualization, the unseen information can be got/displayed/
presented more effectively in a visual environment than when using textual or numerical description.
In this study, publications from England, Scotland, and Wales were treated as publications from the UK. However, publications
from Hong Kong, Macau, and Taiwan were not treated as Chinese publications and were independently counted. Geographic vi-
sualization analysis was conducted using GeoDa and ArcGIS on the publication count statistics for each country/region by drawing a
global research heat map.

2.2.3. Social network analysis


Social networks are composed of actors who are tied to one another through social relations (Prell, Hubacek, & Reed, 2009).
Social actors, such as groups or individuals, are represented by nodes, while links connecting the nodes indicate relations or inter-
actions. Social network analysis aims to understand the actors and investigate their social relations in a specific social context through
the use of networks and graph theory (Tobler, 1970). It has been increasingly important in the social sciences and has been applied in
a variety of areas, such as psychology, health, and business organization (Serrat, 2017). The benefits of social network analysis are
obvious, for example, discerning actors playing central roles, identifying structural holes and isolated actors, leveraging peer support,
and strengthening the efficiency and effectiveness of the existing relations (Serrat, 2017).
In this study, the exploration of collaborative relations between authors, institutions and countries/regions was conducted using
social network analysis with an R package named networkD3. In the generated networks, authors, institutions and countries/regions
were represented by nodes with lines indicating collaborations. The node size reflected publication counts. The node colors indicated
specific countries/regions of institutions or authors, or the specific continents of countries/regions. The line width indicated fre-
quency of collaboration.

1
https://en.wikipedia.org/wiki/H-index.
2
https://en.wikipedia.org/wiki/Impact_factor.
3
https://journalmetrics.scopus.com/index.php/Faqs.
4
https://service.elsevier.com/app/answers/detail/a_id/14884/c/10547/supporthub/scopus/related/1/.

16
Y. Song, et al. Computers & Education 137 (2019) 12–31

2.2.4. Topic analysis


Since publication keywords usually represent the major research focus of a publication, and can help to identify the research topic
of the publication quickly (Zhong, Geng, Liu, Gao, & Chen, 2016), author-defined keywords as well as ISI KeyWords Plus were used
for topic analysis in this study. To ensure analysis effectiveness, keywords preprocessing was applied. First, for the 171 publications
without author-defined keywords or ISI KeyWords Plus, keywords were manually supplemented according to title and abstract
content. Specifically, keywords were mainly extracted from titles, as titles normally reveal the nature of a work. For example,
keywords ‘problem orientation’ ‘small-group’ ‘discussion’ and ‘biochemistry teaching’ were extracted from the following title ‘Pro-
blem-oriented small-group discussion in the teaching of biochemistry laboratory practical’. Keywords were then revisited by referring to
abstracts. Second, a de-duplication process was applied (Cobo, Martinez, Gutierrez-Salcedo, Fujita & Herrera-Viedma, 2015). Spe-
cifically, abbreviations were replaced by corresponding full names by reviewing the corresponding publication content, for example,
L2 was replaced by ‘second language’, L1 was replaced by ‘first language’, FL was replaced by ‘foreign language’, SLA was replaced by
‘Second Language Acquisition’. Third, all the keywords were converted to lower case and were used in the singular form. Further-
more, keywords representing the same concepts were grouped, for example, ‘color’ and ‘color’, ‘behaviour’ and ‘behaviour’.
Finally, the focuses of research on classroom dialogue can be identified by word cloud techniques given that keyword frequency in
scientific publication has been commonly used to indicate the research trends in and focus of, academic literature (Mathews et al.,
2013; Gimenez et al., 2018). An R package named wordcloud2 was used to generate word clouds by visualizing the importance of
keywords through keyword frequency analysis. In the word cloud, the size of font indicated frequency of keyword occurrence in
literature. To explore the emergence and evolution of research interests through time, subsets of data in four periods were obtained:
1999–2003 (361 publications), 2004–2008 (496 publications), 2009–2013 (1215 publications), and 2014–2018 (1842 publications).
By comparing the word clouds of the four periods, the evolution of the research focus could be determined.

3. Results

The result of bibliometric analysis of publication in the field of classroom dialogue are presented in this section, including the
analysis of trends in publications and citations, analysis of prolific journals, authors, institutions and countries/regions analysis,
scientific collaboration analysis, together with the analysis of research area distributions and evolution of the thematic research
hotspots.

3.1. Analysis of trends in publications and citations

Statistical computing was applied to the retrieved publications, and the distributions of total publications and citations by year are
presented in Fig. 1. We conducted polynomial regression analysis with year as the independent variable x to fit the trends of total
publication and citation counts. In the regression modelling, the period 1999–2018 was used for regression fitting. Two polynomial
regression curves with the R2 were integrated in the figure.
From the results, it can be seen that there is a trend towards an increase in the quantity of publications, reflecting a growing
enthusiasm for research in the field. This is also demonstrated by the positive coefficient of x2 in the estimated regression model
(R2 = 0.9676). With the regression model, the predictive value for future years can be estimated, for example, the predictive value
for the year 2019 is calculated as: 1.048416*20192–4190.712*2019 + 4187817 = 494.5912. The citation counts were also recorded

Fig. 1. The trend analysis of publication and citation counts.

17
Y. Song, et al.

Table 2
Top journals with the most publications and citations.
R Journal Classroom dialogue publications in the journal(1999–2018) All publications of the journal(1999–2018)

TP TC H %TP > =300 > =100 > =50 1999–2008 2009–2018 IF (Q) 2017 5-Year IF TP TC H

TP TC TP TC

1 Computers & Education 175 4630 36 5.73 1 6 27 30 155 145 4475 4.538(Q1) 5.568 3053 93375 130
2 International Journal of Science Education 113 2124 22 5.47 0 3 8 40 242 73 1882 1.325(Q3) 1.936 2065 39188 81
3 Modern Language Journal 105 3064 34 4.10 0 5 22 42 352 63 2712 2.789(Q1) 2.578 2564 19361 67
4 System 100 777 15 8.73 0 0 0 0 0 100 777 1.547(Q2) 1.826 1145 6936 36
5 Foreign Language Annals 88 750 16 8.75 0 0 1 40 111 48 639 0.802(Q4) 1.286 1006 7382 85
6 Teaching and Teacher Education 86 1480 20 3.61 0 1 6 27 76 59 1404 2.473(Q1) 3.335 2381 54104 92
7 Science Education 80 3688 29 4.00 1 11 18 34 349 46 3339 3.035(Q1) 4.367 2002 35560 83
8 Language and Education 76 684 14 15.54 0 0 2 5 1 71 683 1.262(Q3) 1.487 489 2350 22
9 Journal of Research in Science Teaching 71 3051 26 6.24 1 9 19 27 352 44 2699 3.21(Q1) 4.132 1138 43806 97
10 Linguistics and Education 65 149 6 22.34 0 0 0 0 0 65 149 0.892(Q3) NA 291 873 14
11 Tesol Quarterly 64 1602 22 5.22 0 3 9 24 281 40 1321 2.256(Q1) 2.828 1227 19249 68
12 Language Teaching Research 61 686 15 12.68 0 0 1 8 4 53 682 2.086(Q1) 2.536 481 4521 33
13 Early Childhood Research Quarterly 45 972 20 4.59 0 0 4 8 57 37 915 2.364(Q1) 3.415 981 23888 72

18
14 Computer Assisted Language Learning 43 415 12 11.44 0 0 1 2 0 41 415 1.928(Q1) 2.366 376 3893 31
15 Language Learning & Technology 42 923 16 7.82 0 1 8 8 65 34 858 2.113(Q1) 3.008 537 9119 54
16 Learning and Instruction 42 1292 19 4.38 0 3 8 14 171 28 1121 3.967(Q1) 5.695 960 33751 87
17 Research in Science Education 42 465 10 5.95 0 0 2 6 7 36 458 1.568(Q2) 1.914 706 8216 37
18 Teachers College Record 42 466 12 1.78 0 0 2 9 22 33 444 1.072(Q3) 1.688 2357 24614 66
19 Journal of Computer Assisted Learning 41 1057 17 4.24 1 1 2 16 74 25 983 1.859(Q2) 3.055 967 20847 68
20 Educational Technology & Society 40 584 14 2.61 0 0 2 11 26 29 558 1.767(Q2) 2.326 1534 20099 60
21 Reading Teacher 40 217 8 2.11 0 0 0 6 14 34 203 0.782(Q4) 1.297 1892 9734 36
22 Educational Studies in Mathematics 38 344 9 5.58 0 1 1 0 0 38 344 1.1(Q3) 1.438 681 5184 28
23 Journal of the Learning Sciences 38 1961 23 8.62 1 5 10 20 155 18 1806 3.0(Q1) 4.471 441 16763 69
24 Learning Culture and Social Interaction 38 303 9 17.19 0 1 1 0 0 38 303 1.125(Q3) 1.568 221 913 12
25 Journal of Literacy Research 37 427 12 8.43 0 0 1 19 84 18 343 1.71(Q2) 2.438 439 4664 33
26 Research in the Teaching of English 36 564 15 8.22 0 0 1 18 75 18 489 1.976(Q1) 2.056 438 3808 29
27 Asia-Pacific Education Researcher 34 118 6 5.72 0 0 0 0 0 34 118 0.633(Q4) 0.753 594 2042 17
28 Instructional Science 34 860 14 5.36 0 2 5 11 102 23 758 1.922(Q1) 2.62 634 12843 54
29 Language Learning 34 1363 18 4.68 0 3 13 12 191 22 1172 1.655(Q2) 3.297 727 19438 72
30 English Teaching-Practice and Critique 33 167 7 8.92 0 0 0 1 0 32 167 0.442(Q3) 0.678 370 1399 17

Abbreviations: TP: total publications; TC: total citations; H: H-index; %TP: percentage of classroom dialogue publications in the journal of all publications of the journal during years
1999–2018; > =300, > =100, > =50: publication counts of publications with more than 300, 100, or 50 citations; IF (Q): impact factor and JCR quartile in category for year 2017.
Computers & Education 137 (2019) 12–31
Y. Song, et al. Computers & Education 137 (2019) 12–31

Fig. 2. The comparison of IF, SJR, IPP, and SNIP for the top 10 most prolific journals.

and analyzed. There has been a dramatic increase in the number of citations, particularly from 2006. Similarly, we also employed a
regression model to predict the trend relating to citation change as z = 31.40926*x2-125662.9* x +12568800.

3.2. Prolific journal analysis

Through the processing of all the publication sources, all the relevant journals were identified, along with their journal impact
factors (IF), JCR quartile in category (Q), and H-index. Table 2 presents the top 30 journals containing the most publications. The top
30 most prolific journals account for 45.55% of the total publications. The top five most prolific journals are Computers & Education,
International Journal of Science Education, Modern Language Journal, System, and Foreign Language Annals. Ranking by TC or H index,
the ranks of the top five journals changes slightly. Science Education has a relatively higher TC and H, thus it is ranked at no.2 (by TC)
and no.3 (by H index), respectively. Computers & Education has the highest rank among all the three metrics, reflecting the high
quality of its publications. By employing more evaluation metrics, for example, SJR, IPP, SNIP, the comparison of the top 10 journals
is visualized in Fig. 2. Computers & Education and Journal of Research in Science Teaching are the two most prestigious journals, even
though the latter has a relatively low TP count.

3.3. Prolific countries/regions analysis

All the institution countries/regions, institutions, and authors participating in each publication were used for the analysis of the
most predominant ones. The publications originate from 81 countries/regions. Fig. 3 shows the geographical distribution of pub-
lication counts for the countries/regions. Table 3 presents the statistics of the top 30 countries/regions with regard to the total
publications in this field. The USA is the most prolific country with 1844 publications (47.11%), followed by the UK with 444
publications (11.34%). The two countries together account for 58 percent of the total publications, indicating their prominent
position in the research field. Furthermore, the USA possesses the highest H-index of 82, showing that it not only dominates in the
publication count, but also has a high academic influence in the research field. The topic five countries/regions are the same when
ranked by TP, TC, or H.
The annual number of publications from the top five countries/regions exhibits almost an upward trend, as shown in Fig. 4. The
USA has had a leading position throughout nearly the entire period. The publication counts for the USA have experienced sharp
increases during the period 2011–2018. In particular, the publication count for the USA has had a dramatic increase in the year
2017–2018. The publication counts for the other four countries/regions have experienced relatively small increases compared with
that of the USA, and especially for Australia, Canada, and Taiwan.

3.4. Prolific institution analysis

There are 1782 institutions from all over the world with relevant research work among the publications. Table 4 presents the top
30 most prolific institutions for the years 2009–2018. Of the top 30, 22 originate from the USA, which again indicates the USA's
dominant position in the research field. The University of Cambridge and University of Virginia have the highest TP (51) during the

19
Y. Song, et al. Computers & Education 137 (2019) 12–31

Fig. 3. Geographical distribution of publication counts.

Table 3
Top most prolific countries/regions ranked by total number of publications.
Rank Name TP (%) TC H > =300 > =100 > =50 1999–2008 2009–2018

TP (%) TC TP (%) TC

1 USA 1844(47.11) 33328 82 3 63 164 493(57.53) 2886 1351(44.19) 30442


2 UK 444(11.34) 8929 45 3 10 38 108(12.60) 752 336(10.99) 8177
3 Australia 234(5.98) 3459 30 0 5 14 56(6.53) 331 178(5.82) 3128
4 Canada 167(4.27) 2443 28 0 1 9 33(3.85) 210 134(4.38) 2233
5 Taiwan 160(4.09) 1959 24 0 0 5 20(2.33) 55 140(4.58) 1904
6 Spain 131(3.35) 1367 20 0 1 5 11(1.28) 47 120(3.93) 1320
7 Netherlands 104(2.66) 1522 20 0 3 7 18(2.10) 95 86(2.81) 1427
8 Hong Kong 92(2.35) 944 15 0 1 5 11(1.28) 33 81(2.65) 911
9 Turkey 90(2.30) 508 12 0 0 1 5(0.58) 0 85(2.78) 508
10 Sweden 89(2.27) 766 13 0 1 3 6(0.70) 31 83(2.72) 735
11 China 81(2.07) 775 13 0 2 4 4(0.47) 0 77(2.52) 775
12 Singapore 79(2.02) 1177 17 0 3 4 16(1.87) 56 63(2.06) 1121
13 Finland 78(1.99) 906 16 0 0 5 12(1.40) 87 66(2.16) 819
14 Germany 72(1.84) 1898 22 1 4 8 16(1.87) 181 56(1.83) 1717
15 South Korea 71(1.81) 542 14 0 0 1 5(0.58) 19 66(2.16) 523
16 New Zealand 64(1.64) 740 14 0 0 4 12(1.40) 43 52(1.70) 697
17 Israel 63(1.61) 987 16 0 0 7 18(2.10) 66 45(1.47) 921
18 South Africa 54(1.38) 463 12 0 0 2 8(0.93) 33 46(1.50) 430
19 Norway 44(1.12) 318 8 0 0 2 8(0.93) 20 36(1.18) 298
20 Japan 40(1.02) 447 13 0 0 1 3(0.35) 4 37(1.21) 443
21 Belgium 28(0.72) 543 10 0 2 2 1(0.12) 7 27(0.88) 536
22 Italy 27(0.69) 326 10 0 0 3 9(1.05) 35 18(0.59) 291
23 Chile 26(0.66) 261 6 0 0 2 3(0.35) 8 23(0.75) 253
24 France 24(0.61) 384 10 0 1 2 6(0.70) 34 18(0.59) 350
25 Malaysia 24(0.61) 170 7 0 0 0 0(0.00) 0 24(0.79) 170
26 Ireland 23(0.59) 207 6 0 0 1 3(0.35) 5 20(0.65) 202
27 Brazil 22(0.56) 427 7 0 1 2 7(0.82) 25 15(0.49) 402
28 Iran 22(0.56) 87 6 0 0 0 0(0.00) 0 22(0.72) 87
29 Denmark 19(0.49) 126 5 0 0 0 2(0.23) 6 17(0.56) 120
30 Greece 19(0.49) 163 7 0 0 0 5(0.58) 6 14(0.46) 157

Abbreviations: TP (%): publication count and percentage; TC: total citations; H: H-index of a country/region; > =300, > =100, > =50: pub-
lication counts of publications with more than 300, 100, or 50 citations.

years 2009–2018. The King's College London has the highest TC (2,300) though it has 20 publications only, indicating the high
quality and influence of its publications.
The annual publication distributions of the top five institutions are presented in Fig. 5. Nanyang Technological University has had

20
Y. Song, et al.

21
Fig. 4. Annual publication distributions of the top 5 countries.
Computers & Education 137 (2019) 12–31
Y. Song, et al. Computers & Education 137 (2019) 12–31

Table 4
Top institutions ranked by total number of publications.
Institution Country 2009–2018 1999–2008 1999–2018 ARWU 2018 QS 2019

TP TC TP TC TP TC H > =300 > =100 > =50

University of Cambridge UK 51 1256 7 58 58 1314 20 0 2 10 3 6


University of Virginia USA 51 1447 5 36 56 1483 21 0 5 6 151–200 192
Nanyang Technological University Singapore 49 1002 14 55 63 1057 16 0 3 4 96 12
Michigan State University USA 42 975 15 51 57 1026 18 0 2 5 101–150 141
Arizona State University USA 41 780 9 37 50 817 15 0 1 5 101–150 212
Pennsylvania State University USA 41 1407 13 90 54 1497 21 1 3 7 74 95
The University of Hong Kong Hong Kong 41 449 6 27 47 476 11 0 0 3 101–150 25
University of Texas at Austin USA 39 1132 12 118 51 1250 18 0 3 8 40 63
The Ohio State University USA 37 779 8 20 45 799 16 0 2 3 94 89
Columbia University USA 36 715 4 13 40 728 14 0 0 4 8 16
University of Michigan, Ann Arbor USA 33 1251 17 214 50 1465 20 0 5 8 27 20
University of Washington USA 31 1000 9 52 40 1052 15 0 3 5 14 66
The University of Auckland New Zealand 29 319 3 26 32 345 9 0 0 2 201–300 85
University of Helsinki Finland 29 364 5 26 34 390 9 0 0 2 57 110
University of Jyvaskyla Finland 29 289 2 9 31 298 11 0 0 1 601–700 336
University of Pittsburgh USA 29 1591 18 105 47 1696 19 1 6 8 90 136
University of Illinois, Champaign USA 27 610 10 160 37 770 13 0 3 4 41 71
Urbana
The University of Queensland Australia 27 502 7 33 34 535 12 0 0 3 55 48
University of Wisconsin–Madison USA 27 531 14 63 41 594 16 0 0 2 28 53
University of North Carolina at USA 25 521 4 40 29 561 11 0 1 4 30 83
Chapel Hill
Utrecht University Netherlands 25 425 6 46 31 471 11 0 1 2 51 124
Purdue University USA 24 488 7 72 31 560 12 0 0 5 70 100
University of Colorado at Boulder USA 24 460 4 8 28 468 14 0 1 2 38 190
University of Maryland, College USA 24 398 6 45 30 443 12 0 0 2 51 20
Park
Florida State University USA 23 756 9 28 32 784 15 0 1 5 201–300 472
University of California, Los USA 23 1066 15 110 38 1176 19 0 2 9 11 32
Angeles
University of Georgia USA 23 678 15 116 38 794 13 0 1 7 201–300 431
Boston College USA 22 733 3 35 25 768 12 0 4 5 301–400 376
University of California, Davis USA 22 404 6 19 28 423 12 0 0 2 96 100
Georgia State University USA 21 276 2 17 23 293 8 0 0 2 NA NA

Abbreviations: TP: publication count; TC: total citations; H: H-index of a country/region; > =300, > =100, > =50: publication counts of pub-
lications with more than 300, 100, or 50 citations; ARWU: The 2018 Academic Ranking of World Universities; QS Rank: QS World University
Rankings 2019.

a relative stable increase in publications over time, while University of Virginia has changed significantly. The University of Cam-
bridge has shown a dramatic increase in the last two years, indicating its recent rapid development recently in the research field.

3.5. Prolific author analysis

The most prolific authors were identified. Table 5 shows the top 20 authors ranked by publication count. Twelve of them come
from the USA, and four are from the UK. Specifically, four of the top five authors are from the USA, demonstrating its active role in
the research field. The most prolific one is Neil Mercer with the highest publication count and Robert C. Pianta with the highest H
index. Jonathan Osborne from King's College London has the highest citation count (1,596), even though his publication count is not
among the top 20.

3.6. Collaboration analysis

The scientific collaboration in perspectives of country/region, institution and author were reported. The collaboration among
countries/regions is visualized as Fig. 6, with 81 nodes and 238 links as well as seven sprase nodes using network analysis, which can
be accessed via http://www.zhukun.org/haoty/resources.asp?id=JCE_country. Each country/region is presented as a node, with the
node size representing its proportion of publications and the node color denoting the continent to which it belongs. The thickness of
each line indicates collaboration strength between two countries/regions. According to Fig. 6, the USA (the largest node in blue) has
the most collaborations with other countries/regions. In addition, the collaboration network between European countries (the nodes
in yellow) is very dense. The collaboration among institutions shown as Fig. 7 can be accessed via http://www.zhukun.org/haoty/
resources.asp?id=JCE_affiliation. The institutions with publications ≥ 15 are involved in publication collaboration. In the network,
the node count and edge count are 80 and 314. Most of the institutions are in the USA and UK. Similarly, the collaboration among

22
Y. Song, et al.

23
Fig. 5. Annual publication distributions of the top five institutions.
Computers & Education 137 (2019) 12–31
Y. Song, et al. Computers & Education 137 (2019) 12–31

Table 5
Top most prolific authors ranked by total number of publications.
Rank Author Institution Country TP TC H > =300 > =100 > =50 1999–2008 2009–2018

TP TC TP TC

1 Neil Mercer Univ Cambridge UK 21 642 12 0 2 5 2 1 19 641


2 Robert C. Pianta Univ Virginia USA 18 905 13 0 5 5 2 26 16 879
3 Katherine L. McNeill Boston Coll USA 17 538 10 0 3 4 2 1 15 537
4 Richard C. Anderson Univ Illinois Champaign Urbana USA 17 430 9 0 2 2 4 59 13 371
5 Kim Nguyen-Jahiel Univ Illinois Champaign Urbana USA 14 397 8 0 2 2 3 59 11 338
6 Robyn M. Gillies Univ Queensland Australia 14 333 9 0 0 2 5 19 9 314
7 Christine Howe Univ Cambridge UK 11 343 6 0 1 4 1 2 10 341
8 Diane L. Schallert Univ Texas Austin USA 11 127 5 0 0 1 1 9 10 118
9 Jason T. Downer Univ Virginia USA 11 501 7 0 3 4 1 13 10 488
10 Laura M. Justice Ohio State Univ USA 11 332 8 0 1 2 1 1 10 331
11 Hansun Zhang Waring Columbia Univ USA 10 207 8 0 0 2 1 0 9 207
12 Sara Hennessy Univ Cambridge UK 10 232 8 0 0 1 2 14 8 218
13 Bridget K. Hamre Univ Virginia USA 9 399 7 0 2 3 0 0 9 399
14 Carollee Howes Univ Calif Los Angeles USA 9 315 7 0 1 2 1 10 8 305
15 Jennifer LoCasale-Crouch Univ Virginia USA 9 193 4 0 1 1 0 0 9 193
16 Paul Warwick Univ Cambridge UK 9 227 8 0 0 2 0 0 9 227
17 Sara E. Rimm-Kaufman Univ Virginia USA 9 248 7 0 1 1 2 12 7 236
18 Baruch B. Schwarz Hebrew Univ Jerusalem Israel 8 190 5 0 0 1 1 1 7 189
19 Chee-Kit Looi Nanyang Technol Univ Singapore 8 155 7 0 0 0 1 1 7 154
20 Eduardo F. Mortimer Univ Fed Minas Gerais Brazil 8 356 6 0 1 2 4 23 4 333

Abbreviations: TP: publication count; TC: total citations; H: H-index of a country/region; > =300, > =100, > =50: publication counts of more
than 300, 100, or 50 citations.

Fig. 6. Collaboration network among countries/regions (The orange nodes represent countries/regions from Europe, red for Asia, purple for Africa,
blue for North America, brown for South America, and green for Oceania). (For interpretation of the references to color in this figure legend, the
reader is referred to the Web version of this article.)

24
Y. Song, et al. Computers & Education 137 (2019) 12–31

Fig. 7. Collaborative network of institutions with publications ≥ 15 (Different colors represent institutions from different countries/regions, for
example, the red nodes represent institutions from Chile, yellow for Taiwan). (For interpretation of the references to color in this figure legend, the
reader is referred to the Web version of this article.)

authors with publications amounting to ≥ 5 shown as Fig. 8, can be accessed via http://www.zhukun.org/haoty/resources.asp?id=
JCE_author. In the network, the node count and edge count are 76 and 68, with 30 sprase nodes. By accessing the dynamic networks,
users can explore the collaborative relationships with respect to specific countries/regions, institutions, or authors by simply clicking
on the nodes.

3.7. Findings concerning research themes and analysis of their evolution

3.7.1. Research keywords and analysis of their evolution


The top 20 keywords ranked by frequency are shown in Table 6. The keyword ‘classroom’ is ranked first with a frequency of 997.
Other very frequent keywords include ‘discourse’ (578), ‘student’ (543), and ‘education’ (486). The evaluation of the keywords in
terms of frequency with respect to particular periods of time is compared, as shown in Fig. 9. Over these periods of time, most
keywords have demonstrated a growth in frequency, in particularly, keywords such as ‘education’ and ‘knowledge’ have shown a
dramatic increase.

3.7.2. Research hotspot analysis


To further highlight the evolution of research topics over time in this research field, we calculated the frequency of all the
keywords in the publication on the basis of the previous four periods of time. For each such period, the keywords were computed and
visualized as a cloud map, as shown in Fig. 10. The size indicates the frequency count of a keyword. From the analysis, the research
topics have become more diverse with time while ‘classroom’, ‘student’, ‘education’, ‘discourse’ are always the core focuses of the
field. ‘Teacher’ has been focused on more in the last 10 years.

4. Discussion

Bibliometric analysis is a useful tool for analyzing and mapping the literature in the field of classroom dialogue. Based on the
3914 publications collected from the WoS database, this research provides a comprehensive overview of, and an intellectual structure
for, the field by recognizing prolific authors, institutions and journals, identifying geographical publication distribution, discovering
and visualizing scientific collaborations, showing research area distributions, and revealing the development of themes in the period

25
Y. Song, et al. Computers & Education 137 (2019) 12–31

Fig. 8. Collaborative network of authors with publications ≥ 5 (Different colors represent authors from different countries/regions, for example, the
purple nodes represent authors from South Korea). (For interpretation of the references to color in this figure legend, the reader is referred to the
Web version of this article.)

Table 6
Top 20 keywords ranked by total frequency for different periods of time.
Keyword 1999–2018 1999–2003 2004–2008 2009–2013 2014–2018

R(TP) % R(TP) % R(TP) % R(TP) % R(TP) %

Classroom 1(997) 25.47 1(83) 22.99 1(117) 23.59 1(300) 24.69 1(497) 26.98
Discourse 2(578) 14.77 2(46) 12.74 2(80) 16.13 2(198) 16.30 4(254) 13.79
Student 3(543) 13.87 3(35) 9.70 3(66) 13.31 3(180) 14.81 3(262) 14.22
Education 4(486) 12.42 5(30) 8.31 4(39) 7.86 4(137) 11.28 2(280) 15.20
Knowledge 5(353) 9.02 7(22) 6.09 5(39) 7.86 5(116) 9.55 5(176) 9.55
Language 6(328) 8.38 6(25) 6.93 12(25) 5.04 6(102) 8.40 6(176) 9.55
Instruction 7(290) 7.41 4(32) 8.86 6(32) 6.45 8(77) 6.34 8(149) 8.09
Teacher 8(283) 7.23 10(20) 5.54 9(29) 5.85 7(84) 6.91 7(150) 8.14
Science 9(252) 6.44 9(20) 5.54 7(32) 6.45 9(76) 6.26 9(124) 6.73
School 10(223) 5.70 23(12) 3.32 10(27) 5.44 13(64) 5.27 10(120) 6.51
Achievement 11(218) 5.57 16(13) 3.60 16(22) 4.44 11(68) 5.60 11(115) 6.24
Literacy 12(208) 5.31 12(18) 4.99 14(24) 4.84 12(64) 5.27 13(102) 5.54
Children 13(185) 4.73 11(19) 5.26 11(25) 5.04 31(40) 3.29 15(101) 5.48
Talk 14(180) 4.60 18(13) 3.60 8(29) 5.85 22(47) 3.87 18(91) 4.94
Collaborative learning 15(177) 4.52 15(14) 3.88 19(18) 3.63 10(72) 5.93 27(73) 3.96
Performance 16(166) 4.24 81(4) 1.11 21(18) 3.63 29(42) 3.46 14(102) 5.54
Classroom discourse 17(163) 4.16 27(10) 2.77 22(17) 3.43 20(49) 4.03 19(87) 4.72
Communication 18(163) 4.16 8(21) 5.82 27(15) 3.02 26(43) 3.54 22(84) 4.56
Motivation 19(161) 4.11 33(10) 2.77 24(16) 3.23 16(53) 4.36 23(82) 4.45
Computer mediated communication 20(156) 3.99 43(7) 1.94 17(20) 4.03 14(59) 4.86 29(70) 3.80

26
Y. Song, et al.

27
Fig. 9. The trend of the top most used keywords in the publications.
Computers & Education 137 (2019) 12–31
Y. Song, et al. Computers & Education 137 (2019) 12–31

Fig. 10. The visualization of keyword evolution during 1999–2003, 2004–2008, 2009–2013, and 2014–2018.

1999–2018.
The publications and citations in relation to classroom dialogue have grown consistently across the 20 years, which can be further
divided into a steady development between 1998 and 2006 and a flourishing since 2006. This finding was compatible with that
reported by Howe and Abedin (2013) based on a review of 225 publications from 1972 to 2011. According to analysis of countries/
regions, the large increase in the number of studies published, on one hand, is attributable to the growing interest of researchers from
non-English speaking countries/regions (e.g. Taiwan, Singapore, Hong Kong) and, on the other hand, due to important contributions
made by scholars in the USA and the UK. In particular, the USA has contributed dramatically more publications, especially since the
year 2012, with seven out of the 10 most prolific institutions, and with authors ranked the most prolific, which indicates its dominant
position in the field. In addition, contributions from the UK were quite numerous, with several authors (e.g. Neil Mercer, Christine
Howe) playing a prominent role in dialogic research: many of them are from University of Cambridge. Collaborations between
authors and institutions have been strengthened, with interior and trans-regional collaborations going hand in hand. The journal

28
Y. Song, et al. Computers & Education 137 (2019) 12–31

analyses identified a list of journals which were likely to publish articles within the scope of classroom dialogue. The publications
were diversely distributed and there was broad interest from multiple research perspectives. Keyword analyses depicted major
themes in the articles published over four periods of time in the past 20 years, namely, 1999–2003, 2004–2008, 2009–2013,
2014–2018. This is helpful for understanding the focuses of research in each period and how the field has developed across the years.
Thematic features in relation to research on classroom dialogue were revealed by analyzing the keywords and a few distinguishing
ones are illustrated in the following paragraphs.
There were several recurring keywords and they encapsulate broad and common thematic areas canvassed throughout the whole
period of time. Firstly, the term ‘classroom’ was the top most frequent keyword in publications across the four periods, which
indicated that most studies were concerned with the implementation of dialogic methods in classroom settings. Secondly, the use of
terminology tended to be diverse, and terms such as ‘discourse’, ‘talk’, ‘dialogue’, ‘discussion’ constantly appeared in publications,
with ‘discourse’ being used most frequently. Because of the overlapping of conceptual differences in terminology, there is a lack of
universally agreed conceptualization with respect to the key concepts amongst researchers, which has the potential to cause con-
fusion. As indicated by Howe and Mercer (2017), the field needs to find common analytical frameworks in order to adequately
compare findings across a rapidly growing number of studies. Thirdly, a majority of the studies targeted teacher–student interaction,
while relatively fewer were concerned with students working in groups without a teacher. Fourthly, science education has con-
sistently been the research focus of research on the past 20 years, with ‘science’ ranking amongst the top ten most frequently used
keywords. This is probably because the process of communication and negotiation is helpful when it comes to reasoning and thinking,
which is particularly important for scientific learning (McNeill, González-Howard, Katsh-Singer, & Loper, 2017). In addition,
mathematics and language learning were the subjects in which authors showed most interest. Fifthly, the role of language is high-
lighted, with the term ‘language’ being listed within top 10 most frequently found keywords, probably due to its role as a medium for
transmitting knowledge from one mind to another (Littleton & Mercer, 2013). The linguistic features which relate to vocabulary,
grammar, and rhetoric were commonly studied, while the ‘physical’ aspects, such as gestures, vocal fluency, and clarity, were less
discussed. Lastly, regarding the educational sectors addressed in the publications, emphasis was placed on schools, while dialogue in
preschool or higher education was studied to a less extent, which can be seen by comparing the frequency counts for ‘school’,
‘preschool’, ‘university’ and ‘kindergarten’.
Sequences between some of the keywords changed a great deal, and new keywords emerged, reflecting the shifting trends in the
field. First and foremost, keywords such as ‘computer-mediated communication’ and ‘technology’ started to appear in 2004–2008 and
have continued to increase since then. This indicates that there is growing interest among authors in communication mediated by
computers, media and technologies. As suggested in Zawacki-Richter and Latchem (2018), digital tools are cognitive tools and
intellectual partners in the knowledge construction process”, which can be adopted both on and off campus. In terms of on-campus
dialogue, the use of information and communication technologies (ICT) and digital media in schools has become an important area of
research in the last 10 years. For example, González-Howard, Marco-Bujosa, McNeill, Goss, and Loper (2018) reported a technology,
toolkit, and how it was used as a source for integrating argumentation into science classroom. More and more researchers have
acknowledged that the creation of an interactive and authentic hypermedia environment is helpful for the implementation of dialogic
pedagogy, which may consequently facilitate learning. Also, there has been a tendency for research into dialogue to make use of
online learning and web-based systems with a reference to keywords such as ‘flipped’ ‘classroom’ and ‘online’. For example, Mori,
Omori, and Sato (2016) published an article about the impact of flipped online Kanji instruction on written vocabulary learning. Epp,
Green, Rahman, and Weaver (2010) reported student-instructor interaction patterns in real-time, scientific online discourse. We can
probably deduce that incorporation of digital tools into dialogic teaching and learning is and will still be a prominent feature of
research on classroom dialogue in the near future.
What students can achieve from participating in classroom dialogue had always been a concern of authors. Three relevant
keywords (i.e. ‘knowledge’, ‘achievement’ and ‘acquisition’), were captured between 1999 and 2008, and new keywords have sub-
sequently joined these since the year 2009, for example, ‘creativity’, ‘thinking’, ‘comprehension’ and ‘outcome’. This indicates that
authors have not only paid attention to general academic attainment, but have also shown increasing interest in particular aspects of
learning achievement relating to classroom dialogue. Moreover, papers have shown a growing concern with what teachers might do
to improve the quality of dialogue (e.g. Sandilos, Goble, Rimm-Kaufman, & Pianta, 2018). This is reflected in the appearance of
professional development during the period 2004–2008, and there has been consistent increase since then. Teacher professional
training is designed and carried out with the aim of encouraging dialogic pedagogy and supporting teachers to implement it in a more
prolific way (e.g. Howe et al., 2015; van der Veen & van Oers, 2017). A few other aspects, including students' participation in
classroom dialogue, have received attention recently; studies of relationships between personal factors (e.g. gender) and dialogic
performance were quite popular in period 1999–2003, but have declined in recent years.
This study maps the developmental trends and topic tendencies in classroom dialogue research in a general sense. Further in-
vestigations would be needed to explore and compare the tendencies and hot research topics in relating to classroom dialogue
research in different disciplinary contexts (e.g. mathematics, second language), and in various specific research areas, such as lin-
guistics (Loewen & Sato, 2018; Sato & Ballinger, 2016). Such analysis would reveal the intellectual structure of classroom dialogue
research in specific disciplines or research areas, and to what extent and by what means research in one discipline or area influences
research methods and findings in other disciplines or areas. This will help establish the connections between the theory, the fra-
mework or methodology, and practice in the various fields.

29
Y. Song, et al. Computers & Education 137 (2019) 12–31

5. Conclusion

This study presents a comprehensive overview of the publications relating to classroom dialogue during years 1999–2018, using
bibliometric analyses. It should be acknowledged that the findings are rooted in one database, WoS, and only publications in journals
were considered. Further investigations would need to extend to more databases, and books, conference papers and dissertations
should also be counted. Furthermore, it would be interesting to investigate how the keywords will relate to each other in future,
which would provide a more in-depth understanding of the field. Despite the limitations, our work still makes prominent con-
tributions. It is the first in-depth study to keep track of advances in the field from a quantitative perspective. Analyses of publication
and citation trend indicate a promising prospect for classroom dialogue research and its great value in applied settings. Findings
concerning prolific authors, institutions, countries and journals will help people to identify marking research, and suitable channels
for sharing their studies, as well as the potential opportunities for collaboration. Our work also contributes by displaying research
status and the development of the field over the past 20 years, which will enable scholars to become more aware of the research
hotspots while making decisions about which topic to address.

Declarations of interest

None.

Acknowledgement

The work is funded by the grants from the Guangdong Philosophy and Social Science Foundation (No. GD18XJY23) and the
National Natural Science Foundation of China (No. 61772146).

References

Alexander, R. J. (2001). Border crossings: Towards a comparative pedagogy. Comparative Education, 37(4), 507–523.
Assefa, S. G., & Rorissa, A. (2013). A bibliometric mapping of the structure of STEM education using co‐word analysis. Journal of the American Society for Information
Science and Technology, 64(12), 2513–2536.
Budd, J. M. (1988). A bibliometric analysis of higher education literature. Research in Higher Education, 28(2), 180–190.
Burns, C., & Myhill, D. (2004). Interactive or inactive? A consideration of the nature of interaction in whole class teaching. Cambridge Journal of Education, 34(1),
35–49.
Campbell, D., Picard-Aitken, M., Côté, G., Caruso, J., Valentim, R., Edmonds, S., ... Laframboise, M. C. (2010). Bibliometrics as a performance measurement tool for
research evaluation: The case of research funded by the National Cancer Institute of Canada. American Journal of Evaluation, 31(1), 66–83.
Chen, X., Liu, Z., Wei, L., Yan, J., Hao, T., & Ding, R. (2018a). A comparative quantitative study of utilizing artificial intelligence on electronic health records in the
USA and China during 2008–2017. BMC Medical Informatics and Decision Making, 18(5), 117.
Chen, X., Wang, S., Tang, Y., & Hao, T. (2019). A bibliometric analysis of event detection in social media. Online Information Review, 43(1), 29–52.
Chen, X., Xie, H., Wang, F. L., Liu, Z., Xu, J., & Hao, T. (2018b). A bibliometric analysis of natural language processing in medical research. BMC Medical Informatics
and Decision Making, 18(1), 14.
Cobo, M. J., López-Herrera, A. G., Herrera-Viedma, E., & Herrera, F. (2011). An approach for detecting, quantifying, and visualizing the evolution of a research field: A
practical application to the fuzzy sets theory field. Journal of Informetrics, 5(1), 146–166.
Cobo, M. J., Martínez, M. A., Gutiérrez-Salcedo, M., Fujita, H., & Herrera-Viedma, E. (2015). 25 years at knowledge-based systems: A bibliometric analysis. Knowledge-
Based Systems, 80, 3–13.
Dallimore, E. J., Hertenstein, J. H., & Platt, M. B. (2010). Class participation in accounting courses: Factors that affect student comfort and learning. Issues in Accounting
Education, 25(4), 613–629.
Ding, Y., Rousseau, R., & Dietmar, W. (2014). Measuring scholarly impact : Methods and practice. Measuring scholarly impact: Methods and practice. Incorporated:
Springer Publishing Company.
Epp, E. M., Green, K. F., Rahman, A. M., & Weaver, G. C. (2010). Analysis of student–instructor interaction patterns in real-time, scientific online discourse. Journal of
Science Education and Technology, 19(1), 49–57.
Falagas, M. E., Karavasiou, A. I., & Bliziotis, I. A. (2006). A bibliometric analysis of global trends of research productivity in tropical medicine. Acta Tropica, 99(2–3),
155–159.
Fejes, A., & Nylander, E. (2014). The Anglophone International (e) A Bibliometric Analysis of Three Adult Education Journals, 2005-2012. Adult Education Quarterly,
64(3), 222–239.
Geng, Y., Chen, W., Liu, Z., Chiu, A. S., Han, W., Liu, Z., ... Cui, X. (2017). A bibliometric review: Energy consumption and greenhouse gas emissions in the residential
sector. Journal of Cleaner Production, 159, 301–316.
Gillies, R. M. (2016). Cooperative learning: Review of research and practice. Australian Journal of Teacher Education, 41(3), 39–54.
Gimenez, E., Salinas, M., & Manzano-Agugliaro, F. (2018). Worldwide research on plant defense against biotic stresses as improvement for sustainable agriculture.
Sustainability, 10(2), 391.
González-Howard, M., Marco-Bujosa, L. M., McNeill, K. L., Goss, M., & Loper, S. (2018). The argumentation toolkit: A resource for integrating argumentation into your
science classroom. Science Scope, 42(3), 74–78.
Hao, T., Chen, X., Li, G., & Yan, J. (2018). A bibliometric analysis of text mining in medical research. Soft Computing, 22(23), 7875–7892.
Henderson, J. B., McNeill, K. L., González-Howard, M., Close, K., & Evans, M. (2018). Key challenges and future directions for educational research on scientific
argumentation. Journal of Research in Science Teaching, 55(1), 5–18.
Hennessy, S., Rojas-Drummond, S., Higham, R., Márquez, A. M., Maine, F., Ríos, R. M., ... Barrera, M. J. (2016). Developing a coding scheme for analysing classroom
dialogue across educational contexts. Learning, Culture and Social Interaction, 9, 16–44.
Heradio, R., de la Torre, L., Galan, D., Cabrerizo, F. J., Herrera-Viedma, E., & Dormido, S. (2016). Virtual and remote labs in education: A bibliometric analysis.
Computers & Education, 98, 14–38.
Hirsch, J. E., & Buela-Casal, G. (2014). The meaning of the h-index. International Journal of Clinical and Health Psychology, 14(2).
Howe, C. (2017). Advances in research on classroom dialogue: Commentary on the articles. Learning and Instruction, 48, 61–65.
Howe, C., & Abedin, M. (2013). Classroom dialogue: A systematic review across four decades of research. Cambridge Journal of Education, 43(3), 325–356.
Howe, C., Ilie, S., Guardia, P., Hofmann, R., Mercer, N., & Riga, F. (2015). Principled improvement in science: Forces and proportional relations in early secondary-
school teaching. International Journal of Science Education, 37(1), 162–184.

30
Y. Song, et al. Computers & Education 137 (2019) 12–31

Howe, C., & Mercer, N. (2017). Commentary on the papers. Language and Education, 31(1), 83–92.
Hunt, G. E., Jackson, D., Watson, R., & Cleary, M. (2013). A citation analysis of nurse education journals using various bibliometric indicators. Journal of Advanced
Nursing, 69(7), 1441–1445.
Kosmützky, A., & Krücken, G. (2014). Growth or steady state? A bibliometric focus on international comparative higher education research. Higher Education, 67(4),
457–472.
Lewison, G., & Devey, M. E. (1999). Bibliometric methods for the evaluation of arthritis research. Rheumatology, 38(1), 13–20.
Littleton, K., & Mercer, N. (2013). Interthinking: Putting talk to work. Routledge.
Loewen, S., & Sato, M. (2018). State-of-the-arts article: Interaction and instructed second language acquisition. Language Teaching, 51(3), 285–329.
Martínez, M. A., Cobo, M. J., Herrera, M., & Herrera-Viedma, E. (2015). Analyzing the scientific evolution of social work using science mapping. Research on Social
Work Practice, 25(2), 257–277.
Mathews, J. P., Miller, B. G., Song, C., Schobert, H. H., Francois, B., & Finkleman, R. B. (2013). The EBB and flow of US coal research 1970-2010 with a focus on
academic institutions. Fuel, 105, 1–12.
Mazloumian, A. (2012). Predicting scholars' scientific impact. PLoS One, 7(11), 149–246.
McNeill, K. L., González-Howard, M., Katsh-Singer, R., & Loper, S. (2017). Moving beyond pseudo argumentation: Teachers' enactments of an educative science
curriculum focused on argumentation. Science Education, 101(3), 426–457.
Mercer, N. (2010). The analysis of classroom talk: Methods and methodologies. British Journal of Educational Psychology, 80(1), 1–14.
Mercer, N., & Dawes, L. (2014). The study of talk between teachers and students, from the 1970s until the 2010s. Oxford Review of Education, 40(4), 430–445.
Merigó, J. M., Gil-Lafuente, A. M., & Yager, R. R. (2015a). An overview of fuzzy research with bibliometric indicators. Applied Soft Computing, 27, 420–433.
Merigó, J. M., Mas-Tur, A., Roig-Tierno, N., & Ribeiro-Soriano, D. (2015b). A bibliometric overview of the journal of business research between 1973 and 2014. Journal
of Business Research, 68(12) 2645-2.
Milesi, C., Brown, K. L., Hawkley, L., Dropkin, E., & Schneider, B. L. (2014). Charting the impact of federal spending for education research: A bibliometric approach.
Educational Researcher, 43(7), 361–370.
Mori, Y., Omori, M., & Sato, K. (2016). The impact of flipped online Kanji instruction on written vocabulary learning for introductory and intermediate Japanese
language students. Foreign Language Annals, 49(4), 729–749.
Mortimer, E. F., & Scott, P. H. (2003). Meaning making in secondary science classrooms. Maidenhead, UK: Open University Press.
O'Connor, C., Michaels, S., Chapin, S., & Harbaugh, A. G. (2017). The silent and the vocal: Participation and learning in whole-class discussion. Learning and Instruction,
48, 5–13.
O'Connor, C., & Michaels, S. (2007). When is dialogue ‘dialogic’? Human Development, 50(5), 275–285.
Peng, B., Guo, D., Qiao, H., Yang, Q., Zhang, B., Hayat, T., ... Ahmad, B. (2018). Bibliometric and visualized analysis of China's coal research 2000–2015. Journal of
Cleaner Production, 197, 1177–1189.
Prell, C., Hubacek, K., & Reed, M. (2009). Stakeholder analysis and social network analysis in natural resource management. Society & Natural Resources, 22(6),
501–518.
Resnick, L., Asterhan, C., & Clarke, S. (2015). Socializing intelligence through academic talk and dialogue. American Educational Research Association.
Sampson, M., Horsley, T., & Doja, A. (2013). A bibliometric analysis of evaluative medical education studies: Characteristics and indexing accuracy. Academic Medicine,
88(3), 421–427.
Sandilos, L. E., Goble, P., Rimm-Kaufman, S. E., & Pianta, R. C. (2018). Does professional development reduce the influence of teacher stress on teacher–child
interactions in pre-kindergarten classrooms? Early Childhood Research Quarterly, 42, 280–290.
Sato, M., & Ballinger, S. (2016). Understanding peer interaction: Research synthesis and directions. Peer interaction and second language learning: Pedagogical potential and
research agenda. 1–30.
Schwab, G. (2011). From dialogue to multilogue: A different view on participation in the English Foreign Language classroom. Classroom Discourse, 2, 3–19.
Serrat, O. (2017). Social network analysis. Knowledge solutions (pp. 39–43). Singapore: Springer.
Sinclair, J., & Coulthard, M. (1975). Towards an analysis of discourse: The English used by pupils and teachers. Oxford: Oxford University Press.
Svensson, G. (2010). SSCI and its impact factors: A “prisoner's dilemma”? European Journal of Marketing, 44(1/2), 23–33.
Tobler, W. R. (1970). A computer movie simulating urban growth in the Detroit region. Economic Geography, 46(sup1), 234–240.
van der Veen, C., & van Oers, B. (2017). Advances in research on classroom dialogue: Learning outcomes and assessments. Learning and Instruction, 48, 1–4.
Vygotsky, L. (1978). Mind and society: The development of higher mental processes. Cambridge, MA: Harvard University Press.
Wiebe Berry, R. A., & Kim, N. (2008). Exploring teacher talk during mathematics instruction in an inclusion classroom. The Journal of Educational Research, 101(6),
363–378.
Xian, H., & Madhavan, K. (2014). Anatomy of scholarly collaboration in engineering education: A big‐data bibliometric analysis. Journal of Engineering Education,
103(3), 486–514.
Zancanaro, A., Todesco, J. L., & Ramos, F. (2015). A bibliometric mapping of open educational resources. International Review of Research in Open and Distance Learning,
16(1).
Zawacki-Richter, O., & Latchem, C. (2018). Exploring four decades of research in Computers & Education. Computers & Education, 122, 136–152.
Zhong, S., Geng, Y., Liu, W., Gao, C., & Chen, W. (2016). A bibliometric review on natural resource accounting during 1995-2014. Journal of Cleaner Production, 139,
122–132.
Zupic, I., & Čater, T. (2015). Bibliometric methods in management and organization. Organizational Research Methods, 18(3), 429–472.

31

You might also like