Citation Analysis of Impact: Most Cited Articles

Table A3 in the Appendix provides the citation information of the top 20 most-cited L2 pronunciation articles in the 15 journals across the three periods (1977–1991, 1992–2006, 2007–2020; with 58, 106, and 465 articles, respectively). These articles were adjusted for the time effect and ranked by their normalized citations for the analysis (see also Table A4 in the Appendix for the top 20 most-cited lists ranked by the raw citations received by the articles). A number of most-cited articles in the first period underlay the introduction of some key concepts such as accentedness (e.g., Anderson-Hsieh & Koehler, Reference Anderson‐Hsieh and Koehler1988; Flege, Reference Flege1981; Patkowski, Reference Patkowski1990; Thompson, Reference Thompson1991; Van Els & De Bot, Reference Van Els and De Bot1987), comprehensibility (e.g., Gass & Varonis, Reference Gass and Varonis1984; Smith & Bisazza, Reference Smith and Bisazza1982), intelligibility (e.g., Eisenstein & Verdi, Reference Eisenstein and Verdi1985; Fayer & Krasinski, Reference Fayer and Krasinski1987), and the perception and production of some L2 sounds (e.g., Bohn & Flege, Reference Bohn and Flege1990; Borden et al., Reference Borden, Gerber and Milsark1983; Werker, Reference Werker1986). Naturally, some of these highly cited articles paved the way for the foundation of some related concepts of L2 pronunciation. For instance, intelligibility was conceptualized afterward by Munro and Derwing (Reference Munro and Derwing1995a), at the same time delineating comprehensibility and accentedness. Later on, intelligibility was introduced as a principle in its own right by Levis (Reference Levis2005) that has since then significantly guided L2 pronunciation teaching and research. Some of the most-cited studies in this period also made significant contributions to the field of L2 pronunciation by presenting alternative hypotheses or building on former theoretical arguments to shed light on, for example, interlanguage phonology (e.g., Eckman, Reference Eckman1981; Hodne, Reference Hodne1985), interlanguage phonetics (Flege, Reference Flege1980), and the critical period hypothesis (e.g., Flege’s Phonological Translation Hypothesis, Reference Flege1981). The high citation numbers of these theory-oriented studies imply that the first period served as a rotator to adapt some general SLA theories to phonetics and phonology.

In the second period, the focus on foreign accents and intelligibility became more transparent (e.g., Derwing & Munro, Reference Derwing and Munro2001; Field, Reference Field2005; Hahn, Reference Hahn2004; Koster & Koet, Reference Koster and Koet1993; Munro et al., Reference Munro, Derwing and Morton2006). Especially Munro and Derwing’s (Reference Munro and Derwing1995a) highly influential paper, with the most citations, alongside its reprint (Munro & Derwing, Reference Munro and Derwing1999), guided future studies with these foci. This makes sense as their seminal work led to a movement toward a standard terminology. As Munro and Derwing (Reference Munro and Derwing2020) has recently noted, the field of L2 pronunciation had witnessed inconsistencies among the concepts of accentedness, intelligibility, and comprehensibility before their 1995 study that showed these constructs to be related but partially independent. The following research on L2 pronunciation has consistently considered their work. A special issue of JSLP in 2020, “25 years of Intelligibility, Comprehensibility and Accentedness,” also refocused on its essential role in L2 pronunciation research. The second period was also the period when there was an increasing interest in the pedagogy of L2 pronunciation led by a number of most-cited articles (e.g., Derwing & Munro, Reference Derwing and Munro2005; Jenkins, Reference Jenkins2002; Levis, Reference Levis2005). These articles suggested informed insights and methodologies for L2 pronunciation instruction in consideration of the intelligibility principle, accentedness, and the concept of English as an International Language (EIL). This period also covered the most-cited studies in the nexus of L2 speech perception and production. These included vowel perception (Flege & MacKay, Reference Flege and MacKay2004) and production (Munro et al., Reference Munro, Flege and MacKay1996), the perception and acquisition of contrasting segments (Escudero & Boersma, Reference Escudero and Boersma2004), and some speech sounds (Hardison, Reference Hardison2003). The eminence of perception and production seems to have contributed a theoretical viewpoint to L2 pronunciation instruction, categorized as either being perception- or production-based (Lee et al., Reference Lee, Plonsky and Saito2020).

In the third period, the study of the effectiveness of instruction in L2 pronunciation development served as the main concern of several most-cited articles with the tools of form-focused instruction and corrective feedback (Saito & Lyster, Reference Saito and Lyster2012), ad hoc training (Wong & Perrachione, Reference Wong and Perrachione2007), podcasting (Ducate & Lomicka, Reference Ducate and Lomicka2009), and gestures (Smotrova, Reference Smotrova2017). A highly cited meta-analysis (Lee et al., Reference Lee, Jang and Plonsky2015) and a narrative review (Thomson & Derwing, Reference Thomson and Derwing2015) were also conducted to investigate the effectiveness of L2 pronunciation instruction. The eminence of these studies in this period is not surprising. With the start year of the third period being 2007, we concur with several authors in coming up with a date close to 2005 as a milestone when L2 pronunciation instruction began to generate more research as an important area of investigation (Isaacs, Reference Isaacs2018; Isaacs & Trofimovich, Reference Isaacs, Trofimovich, Isaacs and Trofimovich2016; Munro & Derwing, Reference Munro and Derwing2015; Thomson & Derwing, Reference Thomson and Derwing2015). Besides, accentedness and comprehensibility received wide attention among the highly cited articles in this period. The linguistic (e.g., Isaacs & Trofimovich, Reference Isaacs and Trofimovich2012; Saito et al., Reference Saito, Trofimovich and Isaacs2017) and nonlinguistic (Saito et al., Reference Saito, Dewaele, Abe and In’nami2018) influences on these concepts, their development (e.g., Derwing et al., Reference Derwing, Munro and Thomson2008; Derwing & Munro, Reference Derwing and Munro2013), and the judgments of them (e.g., Kang, Reference Kang2010; Kang et al., Reference Kang, Rubin and Pickering2010) were addressed in these papers. Finally, several highly cited articles with a focus on L2 speech perception and production also appeared in this period (e.g., Carlson et al., Reference Carlson, Goldrick, Blasingame and Fink2016; Pelzl et al., Reference Pelzl, Lau, Guo and DeKeyser2019; Saito et al., Reference Saito, Trofimovich and Isaacs2016; Shen & Froud, Reference Shen and Froud2019).

Cocitation Analysis: The Highly Influential References

Different from the citation analysis of impact that evaluated the effectiveness of the articles limited to the bibliometric dataset (i.e., the 629 articles) by the number of citations, cocitation analysis is based on the documents cited in the reference lists of the 629 articles. Therefore, cocitation analysis covers cocited documents located outside the available dataset and captures a much broader literature (Hallinger & Kovačević, Reference Hallinger and Kovačević2019). When put together, an amalgamation of these emic and etic perspectives to bibliometric analysis led us to more effectively reveal the intellectual structure and evolution path of the field of L2 pronunciation.

In the cocitation analysis, of the 1,327 unique references cited by the articles published between 1977 and 1991, the top 42 most-cited references (Table A5 in the Appendix) had citation counts of a minimum of three. The network map created by the VOSviewer produced six clusters (Figure 3). With regard to the second period (1992–2006), when the minimum number of citations of a cited reference was set to seven, of the 3,284 references, the top 55 most-cited references (Table A6 in the Appendix) created three clusters as shown in Figure 4. Among the 14,738 references in the 2007–2020 period, the top 50 most-cited references (Table A7 in the Appendix) had at least 25 citations and produced three clusters (Figure 5).

Figure 3.

The network map of the most-cited references (1977–1991).

Figure 4.

The network map of the most-cited references (1992–2006).

Figure 5.

The network map of the most-cited references (2007–2020).

Despite being quite an infertile period in terms of the number of cocitations, the first period provides some useful insights to shed light on the infancy of L2 pronunciation research. Some of the influential documents and clusters that emerged in the cocitation network map of this period set the scene and charted out directions for future L2 pronunciation research. This was the natural corollary of the emergence of a number of foundational theories and hypotheses. As a case in point, the cluster which we termed “interlanguage/interphonology” (Cluster 2 in green), highlighted as one of the biggest clusters with nine references, a total of 46 cocitations and total link strength (TLS), an attribute that indicates the total strength of the cocitation links. The top four most-cited references in this period were also in this cluster. This cluster was mainly represented by Selinker’s (Reference Selinker1972) highly influential paper Interlanguage, and the concomitant work on interlanguage phonology. Tarone (Reference Tarone and Richards1978) in this cluster, for example, the most-cited document in this period, elaborated on the phonology of interlanguage by making a case for L2 pronunciation errors, stating that variable rules of the interlanguage are among the possible causes of such errors. A relevant effort was also made by the two highly cited papers in the same cluster (Eckman, Reference Eckman1977, Reference Eckman1981). While in the former he considered applying the Contrastive Analysis Hypothesis (CAH) to English and German phonologies and proposed typological markedness as the alternative to CAH to explain the relative degree of difficulty, in the latter he provided data for the postulation of interlanguage phonological rules. The other documents in the interlanguage/interphonology cluster dealt with considerations along similar lines such as syllable structure of interlanguage phonology (e.g., Sato, Reference Sato1984; Tarone, Reference Tarone1980), built-in syllabus (e.g., Corder, Reference Corder1967), CAH (Lado, Reference Lado1957), language transfer and language universals (e.g., Gass, Reference Gass1979). Alongside these central theoretical concerns, the first period also established key research themes in the field of L2 pronunciation. These themes gained ground in several documents, such as age factors, the critical and sensitive periods for the acquisition of phonological systems (e.g., Flege, Reference Flege1987a; Krashen, Reference Krashen1973; Neufeld, Reference Neufeld1978, Reference Neufeld1980; Olson & Samuels, Reference Olson and Jay Samuels1973; Oyama, Reference Oyama1976; Snow & Hoefnagel-Höhle, Reference Snow and Hoefnagel-Höhle1977), foreign accents (e.g., Elsendoorn, Reference Elsendoorn1980; Flege, Reference Flege1981, Reference Flege1984; Hill, Reference Hill1970; Scovel, Reference Scovel1969), speech perception and production (e.g., Flege, Reference Flege1987b; Flege & Eefting, Reference Flege and Eefting1987; Flege & Hillenbrand, Reference Flege and Hillenbrand1984; MacKain et al., Reference MacKain, Best and Strange1981), and comprehensibility (e.g., Gass & Varonis, Reference Gass and Varonis1984; Varonis & Gass, Reference Varonis and Gass1982).

In the second period, the interlanguage/interphonology cluster grew bigger (Cluster 1 in red), and created the biggest cluster of this period with 22 documents and 188 cocitations to these documents that formed a TLS of 185. In addition to the theoretical insights presented in the first period’s interlanguage/interphonology cluster, the new members of this cluster in the second period proposed, or built on, a number of hypotheses. These include Eckman’s (Reference Eckman1991) Interlanguage Structural Conformity Hypothesis (which assumes that implicational universals operate in the structuring of interlanguage phonology), Major’s (Reference Major, Loup and Weinberger1987) Ontogeny Model (which claims that interference decreases over time whereas developmental processes increase, and then decrease), parameter setting in L2 phonology (Broselow & Finer, Reference Broselow and Finer1991), and autosegmental and metrical phonology (Goldsmith, Reference Goldsmith1990). Apart from this cluster, this period also generated two more meaningful clusters. The one that we called the “age/accent/speech perception-production” cluster (Cluster 2 in green) was a systematic coming together of these constructs as opposed to their positioning in the first period, and took the form of a cluster in real terms, with 21 documents, 178 citations, and a total of 175 TLS. The other meaningful cluster in the second period witnessed the debut of a new concept, that is, L2 pronunciation teaching. It emerged for the first time to form a cluster in company with comprehensibility and intelligibility. Therefore, we termed this cluster (Cluster 3 in blue) “pronunciation teaching/comprehensibility-intelligibility.” In terms of L2 pronunciation teaching, through a number of highly cited documents, this cluster suggested reference guides (Celce-Murcia et al., Reference Celce-Murcia, Brinton and Goodwin1996; Morley, Reference Morley1991; Pennington & Richards, Reference Pennington and Richards1986), the lingua franca core (LFC) for intelligibility-based pronunciation teaching (Jenkins, Reference Jenkins2000), as well as the results of comparative modes of L2 pronunciation instruction (Derwing et al., Reference Derwing, Munro and Wiebe1998). With respect to comprehensibility and intelligibility, the documents in this cluster established strong connections with accentedness (e.g., Derwing & Munro, Reference Derwing and Munro1997; Munro & Derwing, Reference Munro and Derwing1995a, Reference Munro and Derwing1995b, Reference Munro and Derwing2001). The reason for these terms to be salient together might be that speakers who are perceived as accented can also be highly intelligible (Derwing & Munro, Reference Derwing and Munro2015; Munro & Derwing, Reference Munro and Derwing1995a). That is to say, this cluster situated research on the concept of accentedness in the context of comprehensibility, intelligibility, and speaking rate whereas its word stem accent was mainly approached in terms of the perceived degree and strength in the related documents in Cluster 2 (e.g., Flege, Reference Flege1988; Flege & Fletcher, Reference Flege and Fletcher1992; Flege et al., Reference Flege, Munro and MacKay1995; Piske et al., Reference Piske, MacKay and Flege2001).

The third period was an established period with considerably higher numbers of cocitations to the references in comparison to the earlier periods. The interlanguage/interphonology cluster that swelled in the second period disappeared in this period. However, the pronunciation teaching/comprehensibility-intelligibility cluster expanded into the biggest cluster in this period (Cluster 1 in red), with 27 documents, 943 cocitations, and a TLS of 928. While two of the pedagogy-oriented documents in this cluster also existed in the related cluster of the second period (Derwing et al., Reference Derwing, Munro and Wiebe1998; Jenkins, Reference Jenkins2000), its expansion in the third period, no doubt, led to the inclusion of a great deal of new works. Levis (Reference Levis2005) for example, the most-cited document of this cluster with regard to the pedagogy of L2 pronunciation, shed light on the influences of competing ideologies (teasing apart the nativeness and intelligibility principles), identities, and contextual dynamics. Levis questioned the supremacy of native speaker models, challenging the earlier nativist paradigm that viewed nativelike L2 pronunciation to be possible and desirable to achieve, and instead placing the emphasis on helping learners become simply intelligible. Levis (Reference Levis2005) was followed by Derwing and Munro’s article (2005) that proposed a research-based approach to the teaching of L2 pronunciation. There were also two comprehensive reference books in this respect: Teaching Pronunciation (Celce-Murcia et al., Reference Celce-Murcia, Brinton and Goodwin2010) and Pronunciation Fundamentals (Derwing & Munro, Reference Derwing and Munro2015). The intellectual focus of the pronunciation teaching component of this cluster was furthered by the research syntheses on the effectiveness of L2 pronunciation instruction (e.g., Lee et al., Reference Lee, Jang and Plonsky2015; Saito, Reference Saito2012; Thomson & Derwing, Reference Thomson and Derwing2015). The expansion of the pronunciation teaching/comprehensibility-intelligibility cluster in the third period also extended to the work on the concepts of comprehensibility and intelligibility through newly added documents (e.g., Derwing et al., Reference Derwing, Munro and Thomson2008; Derwing & Munro, Reference Derwing and Munro2009; Field, Reference Field2005; Hahn, Reference Hahn2004; Isaacs & Trofimovich, Reference Isaacs and Trofimovich2012; Munro et al., Reference Munro, Derwing and Morton2006; Trofimovich & Isaacs, Reference Trofimovich and Isaacs2012).

Two other meaningful clusters in the third period were a result of the fission of the age/accent/speech perception-production cluster that emerged in the second period. Accordingly, “speech perception-production” and “age/accent” took the form of two separate clusters. While the age/accent cluster (Cluster 3 in blue) was a small formation devoted to the same concerns as in the second period with five documents, 192 citations, and 183 TLS, the speech perception-production cluster (Cluster 2 in green) was quite extensive, having 18 documents and 736 citations that received a TLS of 699. By far the most-cited document in the speech perception-production cluster was an instrumental book chapter by Flege (Reference Flege and Strange1995). Its guiding role in L2 speech perception-production research seems to result from the content laying out a model of L2 speech learning (SLM), hypotheses regarding the cause of accentedness in L2 speech production, and empirical findings on the perception and production of L2 vowels and consonants. The other highly cited documents in this cluster, among many others, were Best (Reference Best and Strange1995), Best and Tyler (Reference Best, Tyler, Bohn and Munro2007), Bradlow et al. (Reference Bradlow, Pisoni, Akahane-Yamada and Tohkura1997), Flege (Reference Flege1987b), and Flege et al. (Reference Flege, Bohn and Jang1997). In the speech perception-production cluster, we also encountered two seemingly irrelevant documents that found a seat with high citations. The first one was a paper from a scholar in the cognitive SLA family (Schmidt, Reference Schmidt1990). Having theorized the concepts of noticing and consciousness, this paper was cited in L2 pronunciation research when the researchers elaborated on the link between oral input, learners’ noticing, and speech perception (e.g., Trofimovich et al., Reference Trofimovich, Lightbown, Halter and Song2009). This interest in Schmidt echoes in Zhang’s (Reference Zhang2020) bibliometric analysis of the field of SLA between 1997 and 2018, which found his noticing hypothesis continued to influence the field even after three decades. The other one was Bates et al. (Reference Bates, Machler, Bolker and Walker2015) cited by L2 pronunciation researchers when they conducted mixed-effects logistic regression analyses by using the glmer() function in the lme4 package for R (e.g., García-Amaya & Lang, Reference García-Amaya and Lang2020), a package that provides functions to fit and analyze linear, nonlinear, and generalized linear mixed models (Bates et al., Reference Bates, Machler, Bolker and Walker2015).

Trend Analysis

Paper Titles

In another effort to trace the intellectual trajectories of the key terms and themes of L2 pronunciation research in the 629 articles, we first revealed the top ten most frequently occurring terms in the articles’ titles across the three periods in a heuristic manner rather than through a predetermined set of terms or themes. Naturally, a paper title could yield more than one key term with an intellectual magnitude, and this was the case in most of the paper titles in our dataset. For instance, we extracted “perception,” “production,” and “instruction” from the paper titled “The Effects of Perception- vs. Production-Based Pronunciation Instruction,” but disregarded “effects” having no significance for our research concern. Moreover, we left out generic terms such as English, language, pronunciation, phonology, EFL, ESL, and L2 while condensing derivatives into one term (e.g., accent|ed|ness). A cautious approach needs to be followed here while interpreting these results as paper titles may not always effectively cover or reflect research orientations.

As shown in Table 1, there are a number of evergreen terms and research themes that were principally addressed in all the periods. These include perception/production, segmentals and suprasegmental features, teaching/training/instruction, accent, speech/speaker, native/nonnative, and learners/learning. The terms perception/production, suprasegmental features, teaching/training/instruction, and native/nonnative made up more or less the same proportion of the top frequent terms across the three periods. Native was also the most typical collocate of speaker in all the three periods. Although nativeness is usually unattainable and unnecessary (Levis, Reference Levis2020), a possible reason why the nativeness principle remains alive is that L2 pronunciation has evolved separately from other language skill areas in the CLT era, and therefore it continues to stand as an engaging goal for many practitioners and learners (Levis, Reference Levis2020). Thomson and Derwing’s (Reference Thomson and Derwing2015) narrative review also reported most of the studies under scrutiny to have adopted the nativeness principle (63%), followed by the intelligibility principle (24%), while 13% had elements of both. Raising an objection to favoring the nativeness principle, Levis (Reference Levis2020) offers “to embrace the intelligibility principle and consign the nativeness principle to the past” (p. 323). With regard to segmental and suprasegmental(prosod|y|-ic), there was a diversity in cooccurrences in the third period, represented by their frequent collocates with features, cues, attainment/acquisition, perception/production, training/instruction, information, and accuracy. Collocates of perception/production in the second period were characterized by frequent cooccurrences such as vowel perception/production, accent perception, speech perception/production, speaker(s’) production, and accurate/accuracy of perception/production while in the third period, in addition, learner(s’) perception/production, tone and sound perception/production, perception/production task, perception/production training, production improvement, and segmental and suprasegmental(prosodic) perception/production became highly visible. Naturally, teaching/training/instruction collocated most often with pronunciation in all the periods. The coming together of explicit and instruction was also quite common in the third period. This could be accounted for by the promising findings on the effectiveness of explicit pronunciation instruction (Derwing et al., Reference Derwing, Munro, Foote, Waugh and Fleming2014; Kissling, Reference Kissling2013; Saito, Reference Saito2013) that seem to have contributed to the growth of research with such focus.

TABLE 1.

Top ten most common terms in the paper titles across the three periods

Note: Percentages were rounded and calculated within the top terms. In this way, each term could be assessed in terms of its specific period, without being affected by the widely differing number of articles in different periods.

While speech/speaker showed an undulating pattern in the paper titles across the periods, the ratio of learners/learning almost doubled and trebled, respectively, from the first and second periods to the third period, making it by far the most common title word in the last period. Cooccurrences such as speech learning and speech sounds were widespread in the third period. Despite its persistent presence in the paper titles, accent was proportionally in a constant state of decrease, and collocated most often with foreign in all the periods. Teachers/classroom appeared in the top ten list for the first time in the third period. Its emergence in the last period may provide evidence for the migration of L2 pronunciation research, more specifically, pronunciation instruction research, from laboratory settings to classrooms, which is also echoed by Lee et al. (Reference Lee, Jang and Plonsky2015). The collocates of teachers/classroom were not rich in diversity in the first two periods. In the third period, alongside teacher training/education, psychological constructs such as teacher cognition/beliefs/knowledge/identity exhibited eminence. This finding shows parallelism with the growing body of research on the learning of pronunciation pedagogy (Kochem, Reference Kochem2021) and the influence of psychological and social factors (Pennington & Rogerson-Revell, Reference Pennington and Rogerson-Revell2019a), which is also evident in general applied linguistics research (Lei & Liu, Reference Lei and Liu2019).

In this period, moreover, the typical collocates of classroom were L2, observations, learning, setting/environment, and instruction/training. When coupled with the notable increase in the ratio of learners/learning in the last period, this tends to indicate that the field of L2 pronunciation seems to evolve into a more learner- and learning-centered and instructional dimension. This trend may be due in part to the effects of student-centered learning in education and calls from influential scholars such as Oxford (Reference Oxford1996) to put the learners in the center of instruction. Previous research also acknowledged this change (Derwing & Munro, Reference Derwing and Munro2015; Murphy & Baker, Reference Murphy, Baker, Reed and Levis2015). Taken together, the collocates, along with the results of the citation, cocitation, title word, and Mann–Kendall analyses, tend to show that the field is now somehow structured by the research on L2 pronunciation instruction.

Comprehensibility/intelligibility became visible in the top ten list in the second period whereas it failed to show up in the third period despite its high frequency (n = 42). The other highly frequent key terms in the paper titles of the third period were the technologies (CAPT, ASR, computer, web, podcast, video, Twitter, distance, mobile, apps, mp3, video; n = 44), bilinguals/bilingualism (n = 36), listeners/listening (n = 32), experience (n = 23), and orthography (n = 21). One reason for the recently emergent technologies in the paper titles might be that technology seems to be a relatively recent tool of inquiry gaining rapid ground in L2 pronunciation research. Also, there is little reason to deny the role of the concomitant effects of using technological affordances in applied linguistics research, and by extension, in L2 pronunciation research. This technological uptrend in L2 pronunciation research also seems to be facilitated by the rise of several technology-oriented SLA journals such as CALL, ReCALL, and Language Learning & Technology. The years that CALL and ReCALL started to be included in the SSCI database are 2010 and 2011, respectively. The coverage of these two journals also plays a role in increasing the number of technology-oriented L2 pronunciation articles.

Mann–Kendall Trend Test

We also conducted a nonparametric Mann–Kendall test (Mann, Reference Mann1945) to find out whether the top ten key terms that occurred in the paper titles of the three periods, totaling 13 different terms, revealed significant trends in the historical framework. More specifically, on a yearly basis, this analysis included the time slot starting from the first year of appearance (1980–1983, for all the top terms) through 2020. By doing so, this time, we adopted a temporally holistic approach to the trend analysis of the top key terms as opposed to the period-based frequency analyses reported earlier. The trend analysis is illustrated in Table 2, showing the top key terms, levels of significance, and trend directions. The Mann–Kendall test results indicated significant trends for all the top key terms. There was a considerably increasing resort in the paper titles to most of the top terms at p < 0.01 level (i.e., perception/production, suprasegmental features, teaching/training/instruction, accent, speech/speaker, learners/learning, native/nonnative, word/lexi*/vocabulary, comprehensibility/intelligibility, teachers/classroom), and to some of them at p < 0.05 level (i.e., segmentals, adults).

TABLE 2.

Trend analysis of the top key terms in the paper titles

*↑ ↑, ↑ ↑ ↑; significantly increasing trends with p < 0.05, p < 0.01 respectively. ↓ ↓; significantly decreasing trend with p < 0.05.

Starting from Flege’s SLM (Reference Flege, Munro and MacKay1995), the terms perception and production are evergreen because production-based instruction methodologies go as far back as the ALM, which remains in practice today (Heinz, Reference Heinz2013). Despite its proportional decrease in the face of the other key terms in the paper titles in their specific periods, the term accent is also in a general increasing trend.

Citation Analysis of Impact and Cocitation Analysis: Influential Authors

A total of 908 unique authors contributed to the articles published in the 15 journals in the dataset. K. Saito contributed 39 articles. He is followed by P. Trofimovich, having 28 articles. T. M. Derwing, M. J. Munro, J. E. Flege, and T. Isaacs are among the top six, with the article counts of 14, 14, 13, and 12, respectively. R. I. Thomson (n = 8), J. M. Levis (n = 8), O. Kang (n = 7), D. M. Hardison (n = 7), I. R. A. Mackay (n = 5), R. C. Major (n = 5), E. Gatbonton (n = 5), C. T. Best (n = 5), and M. Wang (n = 5) also highlighted with document counts ≥ 5. When interpreting the productivity, it must be kept in mind that the authors’ article counts were only generated from the 629 articles in the dataset. Otherwise, they are no doubt in a more productive position. With respect to the authors’ citation performance revealed in citation analysis of impact, Table A10 in the Appendix displays the top 20 most-cited authors in L2 pronunciation research for the whole period in ranked order of normalized citations adjusted for the time effect and raw citations. The dominance of a number of authors (K. Saito, T. M. Derwing, P. Trofimovich, M. J. Munro) stands out in terms of the normalized citations they received. The other authors that make up the top ten are T. Isaacs, R. I. Thomson, O. Kang, J. E. Flege, R. Lyster, and C. T. Best. Although citation analysis led us to identify the prominent authors, this information needs to be approached with caution because citation analysis, as mentioned before, is limited to the documents in our dataset. Therefore, it is apt to resort to cocitation analysis that is based on authors cited in the reference lists of the reviewed documents, thereby covering a much broader literature. In this way, we can obtain a wider perspective on scholarly impact than basic citation analysis (White & McCain, Reference White and McCain1998).

Alongside citation analysis which portrayed the 44-year period holistically, we also conducted cocitation analysis with a view to identifying the influential authors in the field of L2 pronunciation across the three periods, and the potential intellectual affinities between the authors. In the author cocitation analysis, of the 898 unique authors cited between 1977 and 1991, the top 32 most-cited authors had citation counts of a minimum of six. The network map created by the VOSviewer produced four clusters (Figure 6a). Regarding the second period (1992–2006), the minimum number of citations of a cited author was set to 14, creating three clusters for the top 32 most-cited authors out of 2,042 (Figure 6b). Among the 8,011 cited authors in the 2007–2020 period, the top 33 most-cited authors had at least 60 citations and produced three clusters (Figure 6c). The list of the most-cited authors in the three periods is provided in Table A11 in the Appendix.

Figure 6.

The network maps of the most-cited authors in (a) 1977–1991, (b) 1992–2006, and (c) 2007–2020.

The top-cited author lists firstly show that J. E. Flege dominated the field with the highest citation scores throughout the three periods. This may be because the most well-known perception-based model, perhaps, is Flege’s SLM (Flege, Reference Flege and Strange1995). The model “was designed specifically with the L2 learner in mind, attempting to describe the process by which a learner perceives then comes to produce L2 speech sounds” (Lee et al., Reference Lee, Plonsky and Saito2020, p. 2). This model is directly germane to a large number of L2 pronunciation studies. There were two other scholars who kept their seats in the top lists in all the periods (S. Gass, R. C. Major). A number of theory builders such as L. Selinker, N. Chomsky, R. Lado, and S. Krashen, appeared in the top list of the first period only. Theory-building is a central plank of research in any field. This finding can be attributed to the fact that these theory builders are foundational in mainstream SLA research. A number of scholars (e.g., M. C. Pennington, M. Celce-Murcia, C. J. Sato) appeared in the second period for the first time, and then disappeared in the third period. Some of the scholars (e.g., E. Tarone, F. R. Eckman, T. Scovel, J. Morley, W. Labov) stayed in the top list in the first two periods but disappeared in the third period. A possible reason for this could be the natural theoretical evolution such as the increased use of new theories and paradigms (e.g., usage-based theories, EIL pedagogy) alongside the ones coming from the general field of education (e.g., individual differences, learning strategies, self-paced learning) and distant disciplines (e.g., music, computer science), which is also the case in general applied linguistics research (Lei & Liu, Reference Lei and Liu2019). A number of scholars entered the list in the second period, and remained increasingly influential in the third period, best exemplified by T. M. Derwing, M. J. Munro, C. T. Best, and A. Cutler. Of these, especially Derwing and Munro dominated the field by resolving the inconsistencies among accentedness, intelligibility, and comprehensibility. These authors inspired researchers publishing on these constructs and gave impetus to many of the studies conducted recently. The top list of the third period also welcomed newly added scholars, such as, in order of citation count, K. Saito, P. Boersma, P. Escudero, P. Trofimovich, J. M. Levis, T. Isaacs, P. K. Kuhl, Y. Wang, and R. I. Thomson. Several scholars from the cognitive-interactionist family, such as R. Ellis, R. DeKeyser, and M. Long also appeared together in the top list of the third period.

The network maps (Figure 6) visualize the intellectual structure of the field of L2 pronunciation by clarifying some meaningful clusters and intellectual links between some of the highly influential authors. For example, the 1977–1991 map identifies the “theorists” cluster (Cluster 1 in red), comprising the pioneers of the approximative system (W. Nemser), CAH (R. Lado), Markedness Differential Hypothesis (F. R. Eckman), Universal Grammar (N. Chomsky), and interlanguage phonology (E. Tarone). The 1992–2006 map shows some of the scholars taking center stage in the first cluster (M. J. Munro, T. M. Derwing), second cluster (R. C. Major, F. R Eckman, E. Tarone), and third cluster (J. E. Flege). The 2007–2020 map reveals the strongest intellectual affinities built by some of the highly influential authors in this period (e.g., in order of LS [link strength], J. E. Flege–C. T. Best with 64.96, M. J. Munro–T. M. Derwing with 62.04, T. M. Derwing–K. Saito with 51.30, J. E. Flege–T. M. Derwing with 41.40, and J. E. Flege–M. J. Munro with 40.83). The intellectual affinity of Derwing and Munro is long-standing, and they have been instrumental in the remarkable strength of Canadian researchers in L2 pronunciation (e.g., K. Saito, P. Trofimovich, R. Thomson). Saito worked with Munro as a postdoctoral researcher, and studied under Trofimovich. Many of these researchers also have long-standing connections through the annual “Pronunciation in Second Language Learning and Teaching (PSLLT)” conferences and serving on the editorial board of the JSLP.

In these lists of the most highly cited authors in L2 pronunciation research, we can see the effects of both the eclectic nature of the SLA field and its evolving development. The inclusion of applied linguists, theoretical linguists, educators, teacher education specialists, methodologists, cognitive psychologists, and those from relevant disciplines altogether present a field that values applied and theoretical perspectives toward phonetics, phonology, and the learning and teaching of L2 pronunciation.

Citation Analysis of Impact and Cocitation Analysis: Influential Sources

Table A12 in the Appendix displays the citation analysis performance of the 15 journals in the dataset in ranked order of normalized citations adjusted for the time effect. Language Learning is by far the most-cited journal (n = 121.85). It is followed by other influential journals, such as Studies in Second Language Acquisition (SSLA) (n = 77.53), TESOL Quarterly (n = 70.32), Applied Psycholinguistics (n = 70.07), Bilingualism-Language and Cognition (n = 63.16), and Applied Linguistics (n = 40.68). However, Language Teaching Research (n = 14.82), Computer Assisted Language Learning (n = 13.64), ELT Journal (n = 2.47), and ReCALL (n = 2.45), with normalized citation counts as few as below 15, seem to have limited intellectual impact on the work in L2 pronunciation research in the dataset (cf. the number of raw citations and document counts per journal in Table A12 of the Appendix for a fuller picture).

In the source cocitation analysis, to establish tangible outcomes, the 1977–1991 and 1992–2006 periods were integrated into a single period, and formed the earlier period. When the default setting of the VOSviewer algorithm was preserved (i.e., the minimum number of citations of a source being 20), out of 2,090 sources, 32 met the threshold in the earlier period whereas the latter period (2007–2020) generated 143 publication venues out of 5,129. Table A13 in the Appendix shows the list of the 32 most-cited sources in the earlier period and the top ten most-cited sources in each cluster in the latter period.

The intellectual structures of the top publication sources are visualized in Figure 7. The network map identified three clusters in the earlier period (Figure 7a). The largest cluster in this period (Cluster 1 in red) firstly included influential phonetics and speech-oriented journals such as Journal of Acoustical Society of America (JASA), Journal of Phonetics, and Language and Speech. There were also relatively less influential journals and books in this cluster with similar foci such as Phonetica, Journal of Speech, Language and Hearing Research, and Speech Perception and Linguistic Experience: Issues in Cross-Language Research. Moreover, this cluster included journals with a focus on psychology and psycholinguistics such as Applied Psycholinguistics, Journal of Memory and Language, Cognition, Brain and Language, and Attention, Perception, & Psychophysics. The other two clusters in this period (Cluster 2 in green and Cluster 3 in blue), as in Cluster 1, illustrate intertwinements, but this time in the sense that both of these clusters are an interweave of the sources centered on SLA/T (Second Language Acquisition and Teaching), on the one hand, and linguistics/phonological focus, on the other. Cluster 2, for example, includes SLA/T sources such as Language Learning, SSLA, and Applied Linguistics as well as those with linguistics/phonological focus such as the books Interlanguage Phonology: The Acquisition of a Second Language Sound System and First and Second Language Phonology. Likewise, in Cluster 3, TESOL Quarterly, IRAL–International Review of Applied Linguistics, and ELT Journal sit together with the book Current Perspectives on Pronunciation: Practices Anchored in Theory and the journal Language, which “publishes peer-reviewed articles dealing with issues in linguistic theory” as mentioned in the website. In the earlier period, overall, the most influential publication venues were Language Learning, JASA, TESOL Quarterly, SSLA, and Journal of Phonetics (with citation counts of 414, 259, 203, 183, 154 and TLS of 308.92, 207.57, 143.76, 157.08, 138.98, respectively). The most established links in this period, displaying intellectual affinities, were built between Language Learning–TESOL Quarterly (LS = 42.65), Language Learning–SSLA (LS = 35.10), and JASA–Journal of Phonetics (LS = 26.29).

Figure 7.

The network maps of the most-cited sources in (a) 1977–2006 and (b) 2007–2020.

The latter period (2007–2020) presents a well-defined three-clustered structure (Figure 7b). The first cluster (Cluster 1 in red, with 53 sources) was named the “Psycholinguistics” cluster and most effectively represented by the journals such as Journal of Memory and Language, Applied Psycholinguistics, Cognition, Bilingualism-Language and Cognition, and Attention, Perception, & Psychophysics. The second cluster (Cluster 2 in green, with 46 sources) was entitled the “Phonetics/Speech” cluster, strongly represented by the sources such as JASA, Journal of Phonetics, Language and Speech, focused theses, and Speech Perception and Linguistic Experience: Issues in Cross-Language Research. The third cluster (Cluster 3 in blue, with 44 sources) was labeled as the “SLA-T” cluster. The strongholds of this cluster were the journals such as SSLA, Language Learning, TESOL Quarterly, Applied Linguistics, and Modern Language Journal. Overall, the most influential sources in this period were JASA, SSLA, Language Learning, Journal of Phonetics, TESOL Quarterly, Journal of Memory and Language, Applied Psycholinguistics, Language and Speech, and Applied Linguistics (with citation counts of 1,323, 1,106, 1,096, 1,031, 630, 489, 487, 447, 442 and TLS of 1124.86, 990.74, 974.96, 913.87, 559.42, 448.62, 465.11, 430.76, 415.44, respectively). Starting from the earlier period, especially in the latter period, these journals hold the flagship position, that is, the most consistently visible sources in cocitation analysis resorted by researchers in their efforts to access and cite research on L2 pronunciation. While the sources show a tendency to receive more citations in the latter period considering the higher number of articles in comparison to the earlier period, some noticeable changes are evident in some of the journals’ ranking in the top-cited lists of the two periods (e.g., rises: Journal of Memory and Language from #14 to #6, SSLA from #4 to #2, Cognition from #16 to #11; falls: Language Learning from #1 to #3, TESOL Quarterly from #3 to #5, IRAL from #7 to #21). Also, some of the sources entered the top-cited list in the latter period (e.g., Bilingualism-Language and Cognition, Praat: Doing Phonetics by Computer, System, Foreign Language Annals), whereas some of the sources disappeared in the top-cited list in the latter period (e.g., Language, Journal of Child Language, Journal of Psycholinguistic Research, and Interlanguage Phonology: The Acquisition of a Second Language Sound System). As a result, it is possible to recommend the sources in the three clusters, in order of intellectual impact, to the researchers and readers of L2 pronunciation who would like to keep abreast of inspiring documents and experts in the heartland of L2 pronunciation research. The sources that became visible in the latter period such as Praat: Doing Phonetics by Computer and System point to the increasing role of technology in L2 pronunciation research. Praat’s manual is a highly cited source used by researchers in their efforts to record student responses and conduct phonetic analysis. This tool contributed to the implementation of CAPT practices, providing endless opportunities for repetition and exposure to a wide range of target language speech (Pennington & Rogerson-Revell, Reference Pennington and Rogerson-Revell2019b) and for the improvement of segmental and suprasegmental features (Fouz-González, Reference Fouz-González2020). Similarly, System is an international journal that mainly publishes articles on educational technology and applied linguistics. In the latter period, the strongest associations were observed between JASA–Journal of Phonetics (LS = 130.57), SSLA–Language Learning (LS = 113.55), JASA–SSLA (LS = 72.31), Language Learning–TESOL Quarterly (LS = 65.81), JASA–Language Learning (LS = 62.17), SSLA–Journal of Phonetics (LS = 59.75), Language Learning–Applied Linguistics (LS = 53.13), and SSLA–TESOL Quarterly (LS = 51.81).