Trends and Patterns in Massive Open Online Courses: Review and Content Analysis of Research on MOOCs (2008-2015)

To fully understand the phenomenon of massive open online courses (MOOCs), it is important to identify and map trends and patterns in research on MOOCs. This study does so by reviewing 362 empirical articles published in peer-reviewed journals from 2008 to 2015. For the purpose of this study, content analysis and discourse analysis were employed to analyze the articles. Accordingly, the trend line showing the number of articles per year indicates that the extent of research on MOOCs is likely to increase in the coming years. In terms of research areas, the findings reveal an imbalance and three research areas out of fifteen constitute more than half of all research on MOOCs. With regard to types of MOOCs, related literature is dominated by research on xMOOCs. The discourse in MOOC articles takes a mostly neutral standpoint, articles with a positive outlook outweigh those that are negative, and there is an increase in a more critical discourse. Theoretical or conceptual studies are preferred by researchers, although MOOC research generally does not benefit from being viewed through theoretical or conceptual lenses.


Article abstract
To fully understand the phenomenon of massive open online courses (MOOCs), it is important to identify and map trends and patterns in research on MOOCs. This study does so by reviewing 362 empirical articles published in peer-reviewed journals from 2008 to 2015. For the purpose of this study, content analysis and discourse analysis were employed to analyze the articles. Accordingly, the trend line showing the number of articles per year indicates that the extent of research on MOOCs is likely to increase in the coming years. In terms of research areas, the findings reveal an imbalance and three research areas out of fifteen constitute more than half of all research on MOOCs. With regard to types of MOOCs, related literature is dominated by research on xMOOCs. The discourse in MOOC articles takes a mostly neutral standpoint, articles with a positive outlook outweigh those that are negative, and there is an increase in a more critical discourse. Theoretical or conceptual studies are preferred by researchers, although MOOC research generally does not benefit from being viewed through theoretical or conceptual lenses.

Introduction
The phenomenon of MOOCs has recently attracted considerable attention in the fields of higher education (HE), lifelong learning, and distance education (DE). In spite of the increasing demand and interest, many questions remain unanswered regarding what MOOCs really are and where they are heading in terms of their impact on educational institutions and educational opportunities. Among many published evaluations, researchers have used the following terms to refer to MOOCs: a disruptive innovation (Skiba, 2012;Billington & Fronmueller, 2013;Flynn, 2013); a digital tsunami (Auletta, 2012;Brooks, 2012;McKenna, 2012); an avalanche (Barber, Donnelly, Rizvi, & Summers, 120 Ebben and Murphy (2014) analyzed 25 peer-reviewed articles to identify aspects of the scholarly discourse on MOOCs. They identified two major phases of scholarship on MOOCs, namely Connectivist MOOCs, Engagement, and Creativity from 2009 to 2011/2012 (phase 1); and xMOOCs, Learning Analytics, Assessment, and Critical Discourses about MOOCs from 2012 to 2013 (phase 2).
Sa 'don, Alias, and Ohshima (2014) examined 164 papers published between 2008 and mid-2014 to identify emergent trends regarding MOOCs in higher educational institutions (HEIs). They reported that the top ten nascent research trends in MOOCs for HEIs (at that time) were pedagogical issues, assessment and accreditation, engagement or motivation, knowledge sharing, cultural diversity, technology, social interaction, participant retention, learning analytics, policy, and instructional design. Kennedy (2014) identified the characteristics of MOOCs in informal and post-secondary e-learning with a review of research conducted between 2009 and 2012. After the elimination of several articles, six articles were used to identify the characteristics of MOOCs. She found that openness, barriers to persistence, and MOOC models were the main characteristics that dominated MOOC research at that time. Veletsianos and Shepherdson (2015) conducted research by applying descriptive and inferential statistics to bibliometric data to investigate inter-disciplinarity in MOOC research. They examined 183 research papers published between 2013 and 2015. They reported that education and computer science disciplines were the most prevalent, with a trend towards more interdisciplinary approaches between (Veletsianos & Shepherdson, 2015 compared to MOOC research published between 2008 and 2012 (Liyanagunawardena et al., 2013). Raffaghelli, Cucchiara, and Persico (2015) discussed the methodological approaches in MOOC research between January 2008 and May 2014. Their analysis of 60 articles showed that the majority of research consisted of theoretical studies and case studies; and that there is a need for clear guidelines to identify research methodologies appropriate for the ontological and epistemological questions that address MOOCs. Sangrà, González-Sanmamed, and Anderson (2015) investigated 228 studies that focused on MOOCs between 2013 and 2014. They found that pedagogical strategies, learner motivation, and implications for HE systems were the most popular focus areas.
Veletsianos and Shepherdson (2016) examined 183 papers on empirical studies of MOOCs published between 2013 and 2015, in order to identify gaps in the related literature. They found that most of the contributions to MOOC literature come from North America and Europe. They reported that the selected papers had a focus on students (83.6%), design (46.4%), context and impact (10.9%), and instructors (8.2%). Bozkurt, Özdamar Keskin, and de Waard (2016) reviewed 51 theses and dissertations published between 2008 and 2015. They identified that MOOCs are on the verge of the "plateau of productivity" as described in the Gartner Hype Cycle. Additionally, they found that, though it is a multidisciplinary 121 research avenue, MOOC research is dominated by the field of education; and researchers used qualitative (49%), quantitative (21%), mixed (18%), review (8%), and other (4%) research methodologies. They also highlighted the finding that nearly half of the theses and dissertations ignored any possible benefits from employing theoretical frameworks by not using them. The MOOC research in the theses and dissertations that were analyzed, focused on (extended) xMOOCs rather than on (connectivist) cMOOCs. Bulfin, Pangrazio, and Selwyn (2014) investigated 371 news media headlines over the preceding 24 months within mainstream news media sources in the United States, Australia, and the UK to identify how MOOCs are perceived in these sources. In their analysis, they found that MOOCs are considered to be a portentous development for HE. Kovanović, Joksimović, Gašević, Siemens, and Hatala (2015) examined 3958 news articles, ranging from 2008 to the first half of 2014, to identify MOOC-related public discourse. By using topicmodeling technique, their research revealed that while the total number of news articles followed a declining trend, the quality of the discussions demonstrated an increasing trend. Deimann (2015) examined the MOOC movement by conducting a discourse analysis of 58 articles published in the New York Times between 2012 and 2013. He indicated that the MOOC phenomenon is fueled by a net of power-knowledge relations and MOOCs contribute to a deeper understanding that is beyond pedagogical or economical perspectives. Chen (2014) investigated 306 blog posts related to MOOCs published from January 2010 to June 2013, making use of text-mining. He reported that MOOCs provide opportunities to learners, faculty members, universities, and MOOC providers. He also found that challenges that MOOCs need to overcome include questionable course quality, high dropout rates, unavailable course credits, ineffective assessments, complex copyright issues, and necessary hardware required to join MOOCs.

Broadcast and Social Media
Finally, Shen and Kuo (2015) performed a sentiment and influencer analysis based on Twitter data from June 2013 to May 2014 to explore public sentiment on social media towards MOOCs. They found that positive tweets outweighed negative tweets, even though a slight increase in the number of negative tweets was evident over that time period.
When these articles are examined in terms of their scope, it can be noticed that they covered different aspects of MOOC research, which makes it difficult to compare research findings with each other and conduct follow-up studies. The range of above review studies differ from sample size to issues covered.
However, it is also observed that the methodological approaches, type of MOOCs, opportunities and challenges, use of technology in education, pedagogical approaches, social interaction, use of technology in education, HEIs, quality assurance, and dropout and retention rates were common interests in most of these MOOC reviews.
However, one of the common issues that was salient in MOOC review studies was the cultural relationship and geographical distribution of the participants or authors that were interested in MOOCs. Liyanagunawardena et al. (2013) reported that sampled studies in their research mostly 122 presented participant demographics, which demonstrated that a large majority of participants were from North America and Europe. Similarly, in other MOOC review studies (Ebben & Murphy, 2014;Gašević et al., 2014;Veletsianos & Shepherdson, 2016), it was reported that majority of the authors of MOOC studies were mainly originated from North America and Europe; followed by authors from Australia, Asia, or Africa. This indicates a geographical pattern for the interest in MOOC research and might further indicate a linguistic or cultural relationship.
Another interesting point highlighted in MOOC review articles was the need for new methodological approaches resulting from complex and new nature of networked learning spaces. Thus, approaches such as data-mining, learning analytics, or social network analysis in MOOC research (Ebben & Murphy, 2014;Gašević et al., 2014;Kovanović et al., 2015;Raffaghelli et al., 2015;Sangrà et al., 2015) would be helpful to analyze and interpret massive, sheer volume of data; in other words, big-data, distributed across the networks and globe.
The number of sampled articles analyzed in the reviews presented above ranges from 6 to 266 articles.

Classification of Research Areas
In a systematic review study, it is vital to reflect what has been done in previous research studies and what has been omitted. Therefore, a framework of research areas in distance education, developed by Zawacki-Richter (2009), was used to identify the most prominent and the most neglected areas in MOOC research. Zawacki-Richter's (2009, p.7-9) framework consists of the following levels (an extended version is presented in Appendix A).

Reliability
Articles included in the sample were coded by the first author of this paper, and re-coded by the second author, according to above-mentioned framework of research areas in DE. The extent of agreement between the two raters was calculated using the Kappa statistic proposed by Cohen (1960), which yielded an inter-rater reliability of κ =0.913. A value of between 0.81 and 1.00 reflects almost perfect agreement (Landis & Koch, 1977), or according to Altman (1991), a value within the same interval is regarded as being very good. Thus, the coding of the articles according to the DE research areas can be considered as being acceptable, with an inter-rater value of 0.913 for Cohen's Kappa statistic.

Classification of Research Method, Designs, and Models
Educational research is usually dominated by qualitative, quantitative, or mixed methods research.
However, the advent of network technologies has enabled some innovative research methods based on specific data collection and analysis techniques such as the use of "big data" in learning analytics. In this sense, a new schema of research methods and models/ designs was introduced in this research.
On these grounds, in addition to quantitative, qualitative, mixed and theoretical research methodologies, data mining and analytics was included. Additionally, two research methods-designbased research and action research-that don't fit into any of the standard research methodologies, were classified as "practice-based" methodologies.

Research Method and Design
This paper used the method of systematic review (research synthesis) to arrive at a comprehensive and reliable overview of MOOC research. Systematic reviews involve three key activities: identifying and describing relevant research, critically appraising research reports in a systematic manner, and synthesizing research findings into a coherent statement (Gough, Oliver, & Thomas, 2012). Such reviews can provide guidance for researchers in planning future studies, as well as convenient summaries of the literature on a particular issue (Petticrew & Roberts, 2008). Two basic systematic research methodologies are aggregative and configurative reviews (Gough, Oliver, & Thomas, 2012).
In this study, a configurative review was used, in which the synthesis is made predominantly by configuring data from the sampled studies to answer the review questions.

Sampling
The selected articles were found by searching for using the following keywords: MOOC, MOOCs, Massive Open Online Course, and Massive Open Online Courses. To screen the articles, multiple academic databases were used; however, EBSCO, ERIC, Google Scholar, and Scopus were found to provide the most comprehensive search results. Searches were conducted for each year separately, and recurring articles were removed from the list of sampled articles. The inclusion criteria for sampling were: published in a peer-reviewed journal between 2008 and 2015; written in English; online full-text accessibility; and searched keywords to appear in the title.
The search was limited to the time period from 2008 to 2015. The year 2008 was selected as a starting point since the first MOOC was run at this date, and the first example from the grey literature, that is to say non-conventional, non-commercial literature, was written in 2008 by Cormier (2008) who also invented the term "MOOC." Though there were some articles that used the searched keywords in their abstracts or list of keywords (or both), we deliberately selected only those that included the keywords in their titles, assuming that this would identify articles with MOOCs as their focal point.
After screening and examining 888 articles, a total of 362 articles ( Figure 1) that met the inclusion criteria were further examined according the research questions of the study.

Data Collection, Procedure, and Analysis
The study used document analysis to collect data, content analysis to identify research trends and patterns, and discourse analysis to identify the tone of the selected articles. The overall research flow is shown in Figure 2. Document analysis was used to collect data and create a valid corpus based on the research questions and inclusion criteria stated above. Document analysis is a technique that involves skimming (superficial examination), reading (thorough examination), and interpretation (Bowen, 2009). During the initial searching and screening processes, a total of 888 papers were identified. This first corpus was analyzed through skimming, which yielded that 526 papers were irrelevant (articles that have searched keywords in the title, but do not address MOOCs in the main text), or did not meet the inclusion criteria; these were then excluded. Following the document analysis process, 362 empirical articles that were published in peer-reviewed journals were selected for further analysis.
After the identification of the 362 articles, the researchers used content analysis, which can employ different methodological approaches (qualitative or quantitative or both) and a variety of data types (Banks, Louie, & Einerson, 2000). The use of such approaches allows researchers to make replicable and valid inferences from data within their context, with the purpose of providing knowledge, new insights, a representation of the facts, and a practical guide to action (Krippendorff, 1980). The 362 articles or related sections were coded based on explicit rules of coding (Berelson, 1952), and according to pre-set categories (e.g., research method, model/ design etc.) defined by the researchers, keeping in mind the purpose of the research, and according to predefined research areas of DE. Some sections were based on counting the findings (e.g., citation analysis). The results were reported using trendline graphs or descriptive analysis such as frequency or percentage values.

126
In the final step, the researchers applied discourse analysis to analyze the viewpoints, perspectives, and aims hidden in the text, in order to reveal the position taken by the authors of each paper (Van Dijk, 1993). For this purpose, the conclusion sections of the papers were coded according to preset categories of positive, negative, neutral, and critical discourses, so as to identify the relationship between the text and the underlying message.

Significance and Limitations of the Study
This study provides a 360-degree evaluation of research on MOOCs by identifying trends and patterns in the field over an eight-year period. The aim is to examine the phenomenon from different aspects and thus provide a complete map of the field. Articles included in this research study were collected through an open search that provided a significant corpus of 362 articles and enabled the researchers to present a holistic perspective from the advent of MOOCs in 2008 to 2015. This time span is a sufficient period to allow the field to mature and provide sufficient data to identify trends and patterns. Lastly, the study not only identifies trends and patterns in MOOC research, but also provides a research agenda for future directions, which is important for the improvement of MOOCs in particular, and open, distance, and distributed learning in general.
In addition to its significance, this study has some limitations. First of all the research corpus is limited only to peer-reviewed articles published in journals between 2008 and 2015. Other studies such as conference proceedings were not included with an assumption that not all proceedings are filtered through a review mechanism. However, it is thought that conducting a similar analysis would contribute to the literature. Secondly, as lingua franca, only articles written in English were included to the research to reach a global perspective. Thirdly, the articles which have online full-text access were included in the research corpus and those that required payment to access full-text were excluded. Finally, articles that included searched keywords in their titles were analyzed. The rationale for such an approach is that: articles that used defined keywords in their title would specifically focus on MOOCs.

Findings and Discussion
This section explains research areas, patterns (discourse analysis, MOOC types, and citation analysis) and trends (methodology, research design/model, and theoretical frameworks) of MOOCs respectively.

Research Areas
The classification of research areas in distance education developed by Zawacki-Richter (2009)

127
At macro level (overall 40.4%), theories and models (27%) is the strongest emerging research area, since more than a quarter of the articles studied MOOCs within angle of this research area. The inflation in this research area also increased the overall value of the macro level. Other macro level research areas such as research models in DE and knowledge transfer (4.2%), globalization of education and cross-cultural aspects (3.8%), distance teaching systems and institutions (2.8%), access, equity, and ethics (2.6%) were identified as research areas that are rather neglected.
At the meso level (overall 25.6%), quality assurance (6.1%) is the most emerging research area, which includes topics such as dropout rates, accreditation and quality standards. Cost and benefits (5.6%), educational technology (4.2%), innovation and change (4%), and professional development and faculty support (3.3%) are less investigated research areas. Following these, management and organization (1.6%) and learner support services (0.7%) appear to be the least examined research areas at this level.
The micro level (overall 34.4%) is the second most studied level of classification. At this level, learner characteristics (15.7%) is the most examined research area, followed by instructional design (11%), and interaction and communication in learning communities (7.3%).
The three most studied research areas (theories and models at the macro level; learner characteristics and instructional design at the micro level) constitute 53.7% of the overall corpus, which clearly identifies the remaining twelve research areas as those that need to be studied more. These findings related to research areas have revealed not only those areas that are most researched, but also those that are most neglected and require a special focus to improve MOOC practices. Based on these findings, researchers and institutions can develop a research agenda and adjust their research interests accordingly. According to this classification, the nature of the discourse across all the articles is 27.1% positive, 1.1% negative, 56.4% neutral, and 15.5% critical. Based on the positive trend lines, it is possible to say that MOOCs will remain on the research agenda of open and distance learning. It is also noteworthy that very few of the papers take a negative perspective, while a considerable number of research articles take a critical perspective. Since critical researchers don't take the promise of MOOCs for granted, they produce valuable information that can contribute to the improvement of this lifelong learning model. The overall outlook provides an insight regarding how MOOCs are perceived in academia. On examination of Figure 4, it can be seen that the number of critical articles has been increasing from 2012 onwards. This may indicate that MOOCs are now being evaluated from more realistic perspectives, which is significant in order to build a robust research foundation. In terms of positive and negative discourses, the findings of this study demonstrate similar patterns with Shen and Kuo (2015). In their sentiment analysis, they found that public opinion in microblogging services generally favoured learning through MOOCs. Critical discourse in MOOC articles also has a similar pattern with previous research. For instance, Adams (2013) reported that the critical discourse became more apparent by 2013 because it was thought that MOOCs failed to achieve their promises. Ebben and Murphy (2014) also stated that a critical discourse about MOOCs started by 2012. In a similar vein, according to an analysis of public discourse surrounding MOOCs, Kovanović et al. (2015) analyzed 3,958 news articles and in line with the thoughts and findings of previous research, found that while there is a decrease in the number of MOOC-related news articles, the quality of the discussions in news articles appears to be increasing. They also noted that the discourse about MOOCs changed significantly.

128
As stated in previous research, year 2012/2013 is the beginning of critical discourse that can be linked to high dropout rate that was diagnosed with the second generation xMOOCs which followed the first generation cMOOCs. We also see that how an open, free learning model, that is to say cMOOCs, was transformed into something semi-open, that is to say xMOOCs, and adopted freemium business model resulting with criticism by 2012 onwards.

MOOC Types
The selected articles were examined according to MOOC types ( Figure 5). For the analysis, research articles were coded if the data was gathered in a specific type of MOOC, and theoretical/conceptual articles were coded if they explained MOOC types in their literature review section and provided a synthesis accordingly. Interestingly, across all the articles, the number of articles (n=27; 7.5%) that focus on cMOOCs is relatively small since their first appearance in 2008, while there was a sudden increase in the number of xMOOC articles from 2013 onwards (n=116; 32.1%). The number of articles that did not distinguish between the MOOC types or reported on both c-and xMOOCs (n=53; 14.7%) shows a similar pattern to xMOOCs. Although the number is small (n=6; 1.7%), there are some articles that report on hybrid/dual-layer MOOCs.
The letters in the MOOC acronym clearly define its meaning, and the initial letters such as "c"  (Ebben & Murphy, 2014). The findings of this study confirm Ebben and Murphy and raise an interesting fact, which indicates that the MOOC research sphere is xMOOC dominant and it seems that xMOOCs are being the default type. There might be many reasons regarding this trend.
In an effort to explain the dominance of xMOOCs, Veletsianos and Shepherdson (2015) suggested that pedagogic approaches used in xMOOCs are well-known among the academics and very similar to those already have been used and widely adopted in HEIs and online learning practices, while the approaches used in cMOOCs are not much known and academics might be hesitant or skeptical about it. Accordingly, it is possible to suggest that among different MOOC types (cMOOCs, xMOOCs, or hybrid/dual-layer MOOCs), xMOOC research has matured and was adopted by mainstream education, while cMOOC and hybrid/dual-layer MOOC research needs more time to mature and potentially there are many issues to further explore.

Citation Analysis
A total of 11,520 references were collected from the 362 articles published between 2008 and 2015, thus providing a map of the most cited works in MOOC research (See the five most cited works in Table 1). The 74 works that are cited at least 10 times are listed in Appendix B. The citation pattern reveals some interesting results. The 74 works are cited a total of 1,696 times which means that they constitute 14.7% of 11,520 references. This distribution of references exhibits a similarity with the Pareto Principle (Juran, 1975), which suggests that approximately 80% of outcomes originate in 20% 132 Table 1 The Five Most Cited Works Note. An extended list is provided in Appendix B.

Methodology and Research Design/Model
The findings reveal that, of all the sampled studies (n=362), conceptual/descriptive studies constitute the most employed methodology (53.3%) ( Figure 6)-more than half of the studies that examined MOOCs used this methodology. They took the form of literature reviews (24.3%), position papers (8%), opinion papers (6.1%), reports (5.5%) and other research models. Furthermore, conceptual/descriptive studies constitute the majority of papers in almost each year.
Unfortunately, though conceptual/descriptive studies have value on their own, many of the studies using this type of the methodology were poorly reported with a lack of empirical data, and did not contribute much to the literature or synthesize current literature; on the contrary, many are superficial reviews. In addition, the number of position and opinion papers is obtrusive. This indicates that many of the researchers at that time were still discussing the phenomenon of MOOCs, and deciding whether they were for or against it.
Currently, most of the MOOCs are provided through learning platforms such as Coursera and the data needed for research are confined to these platforms. Thus, the distinctly outnumbered conceptual/descriptive papers further indicate the obstacles to access MOOC data, which, in turn, ends up with 193 conceptual/descriptive papers out of 362.
Quantitative methodology is important in being able to generalize research findings and improve the field horizontally; in particular, more correlational and experimental research studies are needed to explore the complex nature of MOOCs. Mixed methods ranked fifth (5.5%) among the research methodologies employed in the selected articles. Explanatory sequential (3%) and convergent parallel (1.9%) designs were the most used in this category. Considering the many aspects of MOOCS that may be studied, mixed methodology is a powerful approach for researchers to employ, building on the strengths of both quantitative and qualitative data. Nonetheless, the use of mixed methods in the selected articles is relatively low.
Practice-based studies constitute the least used type of research methodology (0.6%), with action research (0.6%) being the only model evident. Practice-based methodologies follow reflective, iterative, cyclical processes, which may be difficult for individual researchers to implement in studying MOOCs. Nevertheless, such studies would be favorable in contributing to promoting the success and sustainability of MOOCs.
These findings confirm those of Raffaghelli et al. (2015), who reported that theoretical/conceptual papers constitute the majority, that is to say 23.3% of MOOC research (n=60). They noted that mixed (20%), quantitative (15%), "others" (15%), qualitative (11.7%), and design-based research (8.3%) were 134 other methodologies used. In their analysis, they classified empirical articles under a research paradigm based on their explicit findings, or their own interpretation when the methodology was not explicitly stated in a study. Because no articles in the scope of the present study were classified as design-based research, our findings do not match those of Raffaghelli et al. (2015) in the category of practice-based research models. Gašević et al. (2014), who categorized methodological approaches as qualitative, quantitative, mixed and other, reported a contrasting pattern with this study among the proposals (n=265) submitted for funding to the MOOC Research Initiative (MRI), with preferred methodologies being mixed (36.2%), qualitative (27.9%), quantitative (30.2%) and other (5.7%).
On examination of Figure 6, it can be seen that there was an increase in MOOC research from 2013 onwards, yet more than half of the research is taken up by conceptual/descriptive studies, which is thought to be a handicap. Undoubtedly, some of these papers provide valuable insights, yet it is clear that there is an imbalance in the distribution of research methodologies among the selected articles (Table 2).  (2016), dependence on particular research methodologies may restrict our understanding of MOOCs. Raffaghelli et al. (2015) claim that it is necessary to define the main constructs needed to drive empirical research; however, the sparsity of empirical research on MOOCs indicates an inflated debate, poorly underpinned by empirical evidence.
In any scientific endeavour, researchers should interpret and draw conclusions from empirical data and a synthesis of previous findings (or both), so as to contribute to the analysis and development of their field.
These research findings trigger new issues regarding research in general and MOOC research in particular. It is salient that the MOOC phenomenon has been studied not only by means of quantitative, qualitative, or mixed paradigms, but other promising methodological paradigms such as data mining and analytics and practice-based methodology have also been used to better understand MOOCs. It is clear that research methodology in the field is evolving and new methodologies are emerging, mostly due to advances in technology and needs stemming from the transformational effect of MOOCs on our viewpoints. There is an increasing tendency to use data mining and analytics also in other research fields to better understand the practices that are mostly online. Innovative techniques such as learning analytics, social network analysis, and data mining are influencing the emerging methodological paradigms. These techniques have been used as a methodology, research model, and data collection and analysis tool on their own. We also see that practice-based methodologies such as action research and design-based research create a research realm of their own.
Another issue that requires attention, in terms of MOOC research, is similar to what has been found regarding research in the field of distance education (Bozkurt et al., 2015a). A great deal of MOOC articles is weak or superficial in reporting methodology and/or findings, which undermines the validity and reliability of their empirical research findings. In addition, abstracts and keywords, which are supposed to be an overall indication of the content of an article, do not necessarily reflect the key features of the selected studies, such as research aims, methodology, and conclusions.

Theoretical Frameworks
The theoretical or conceptual frameworks used in articles about MOOCs were used as a lens to interpret findings, and were counted and ranked according to their frequency (Table 3). It was seen that 64 articles (17.7%) out of 362 used employed theoretical/conceptual frameworks, while 298 (82.3%) articles didn't benefit from any theoretical or conceptual frameworks, or both. Those 64 articles used 43 different frameworks, and some used more than one framework in a single study.

136
On examination of Table 3, it can be seen that theoretical or conceptual frameworks used in MOOC research relate mostly to the educational area, including technology, innovation, social learning, and independent learning. Note: One study may employ more than one theoretical or conceptual framework.
According to Perraton (1988), research without a theoretical basis tends to be nothing more than data gathering, in many cases. Similarly, Bozkurt et al. (2015b) state that ignoring the use of theoretical or conceptual frameworks encourages researchers to report descriptive findings only. Anderson (2008) notes that theories allow researchers to make connections with other studies, provide deeper understanding of concepts, and facilitate the transformation of knowledge from one context to another. Referring to Immanuel Kant's saying "theory without practice is empty; practice without theory is blind,'' Morrison and van der Werf (2012, p.399)

Conclusion and Implications for Future Research
The results of this study reveal and describe research trends and patterns in MOOC research over the time period from 2008 to 2015. As can be seen in Figure 1, research on MOOCs started to expand from 2013 onwards, and the pattern indicates a positive trend in the coming years. The analysis regarding research areas in MOOCs has demonstrated that MOOC studies encompass three main research areas: theories and models (27%) (macro level), learner characteristics (15.7%), and instructional design (11%) (all at the micro level) (see Figure 3).
In terms of the types of MOOC that appear in the research, most of the selected articles deal with xMOOCs (32.1%), cMOOCs (7.5%), hybrid/dual-layer MOOCs (1.7%) or c/xMOOCs (14.7). However, many of the articles (44%) did not explain or clarify the type of MOOC that they were investigating.