Introduction

Even with the implementation of various theories and models intended for improving critical reading skills, research suggests that much remains unknown regarding the most effective practices. A few learning theories, like self-paced or experiential learning, tend to be deficient in providing learners with a socially interactive and collaborative context. For instance, a model that prioritizes personalization or teacher-centred instruction does not always sufficiently encourage students to evaluate and analyse the information given to them actively. Therefore, more effective ways of nurturing critical reading skills in students that enable collaboration must be identified.

How the cooperative learning model effectively develops learners’ critical reading skills.

The differences between teaching critical reading skills using a cooperative model and teaching critical reading using a cooperative learning model.

Methodology

A Comprehensive Guide to Literature Search, Identification, and Selection

A comprehensive electronic search was conducted using a variety of databases, including Scopus, Google Scholar, EBSCO, Emerald Insight, Science & Direct, SpringerLink, Taylor & Francis, and ProQuest. This search yielded a total of 489 articles from diverse studies. Subsequently, the study results were selected based on their relevance to teaching critical reading skills using a cooperative learning model. This selection process identified 145 relevant studies. This study aimed to select and focus on research using an experimental/quasi-experimental design, and 48 studies were obtained. The experimental design was chosen using a comparative design of the intervention group and the control group, and 28 studies using experimental designs were obtained from the results of this selection, which is further described in Figure 1.

Figure 1.Flow Chart for Meta-Analysis Process by PRISMA Guidelines

In the context of this study, a total of 23 journal articles and five proceedings were utilised as research data sources. The effectiveness of the cooperative learning model in enhancing critical reading skills was assessed by calculating a total of 28 effect sizes. These effect sizes were obtained by comparing the subjects’ pre-test and post-test scores in the experimental group or the experimental and control groups' scores. The meta-analysis also synthesises critical reading scores using a cooperative learning model, drawing from 2043 subjects. The descriptive data of the study and its characteristics are shown in Table 1.

Table 1. Data on Independent Variables of the Study

The characteristics of the 28 studies on critical reading using a cooperative model can be analyzed based on the level of essential reading application at the level of education, country, and type of publication. First, in terms of critical reading application levels, most studies focused on primary and secondary education levels. These studies identified that cooperative models are highly effective at increasing students' engagement in reading activities and discussions, which can hone their critical reading skills. At the primary and secondary education levels, skills such as text analysis, questioning information, and composing arguments cooperatively are strengthened through collaboration between students. Meanwhile, studies at the college level show that this model effectively deepens students' understanding of complex academic texts, especially in the context of critical analysis and argument evaluation.

Second, these studies are spread across different countries, with most of the research conducted in countries with advanced education systems, such as the United States, the United Kingdom, and Australia. These countries have better educational support, so the cooperative model can be implemented more comprehensively with adequate resources. Meanwhile, although there is less research from developing countries, some studies from regions such as Asia and Africa suggest that applying cooperative models in local education can improve students' critical reading skills. However, they often face the challenge of lacking facilities and access to adequate materials. Therefore, although countries with more resources were more dominant in this study, the application of cooperative models in developing countries also showed positive results despite specific barriers.

Third, the types of publications in this study are diverse, ranging from journal articles published in educational journals and educational psychology to conference proceedings and research reports. Published journal articles generally focus on the methods used in research and the results obtained from experiments conducted in various educational settings. Publications in conference proceedings often provide a preliminary view of the effectiveness of cooperative models in more specific countries or regions. In addition, several research reports and dissertations give an in-depth analysis of the impact of this model on critical reading skills, often focusing on teaching techniques and student interaction. This type of publication shows the diversity in the way research is published. It shows that the theme of critical reading with a cooperative model remains a growing and highly relevant area of education research.

Selection Criteria for Included Studies

The criteria for selecting the manuscript used by the researcher to include this study in the meta-analysis are as follows. First, the manuscript must be the result of empirical research relevant to the topic and purpose of this study and published in a peer-reviewed journal. Second, the selected manuscript must present quantitative data that can be extracted for statistical analysis, such as effect size, sample count, and standard deviation or p-value explicitly reported; the complete instrument is as follows.

Table 2. Selection Criteria for the Study Analyzed

Final Selection and Data Extraction

The selected study records were subsequently extracted with data to address the previously posed inquiries. The extracted data comprised the following: author, year of publication, average score of critical reading learning ability using cooperative learning model, standard deviation, sample size, t-count value, and effect sizes. All data collected through data collection instruments is then extracted using Comprehensive Meta-Analysis V3 (CMA) software. This extraction process is essential for analyzing and organizing data from the various studies that have been selected, allowing researchers to obtain more comprehensive and accurate results. Data were obtained based on the data extraction results from 28 studies, including complete mean, standard deviation, and sample size data.

Data Analysis

Findings/Results

The results of the analysis conducted by the researcher with the help of the Comprehensive Meta-Analysis (CMA) program have been compiled and presented in the form of a precise tabulation. This analysis involves various statistical calculations to ensure the accuracy and consistency of the findings of the multiple studies. Using CMA, researchers can systematically organise the data, provide a more comprehensive picture of the diverse research results, and assess the significance of the observed effects. The results can now be found in Table 3 below, which provides detailed and easy-to-understand information about the study's main findings.

Table 3. Meta-Analysis Results

The fixed model demonstrated a mean overall standard difference of 0.784 (95% CI, 0.689 to 0.880) with a p-value of 0.00 (<.05). The effect sizes derived from the random model revealed an overall standard difference in the mean of 0.939 (95% CI, 0.563 to 1.315), accompanied by a p-value of 0.00, which is also less than 0.05. When employing a cooperative learning model, these two meta-analyses suggest a positive association between effect sizes and critical reading learning. Consequently, critical reading learning utilising the collaborative learning model is more effective than critical reading learning that does not employ the cooperative learning model. Additionally, the study reported a substantial effect, given that the magnitude of both effect sizes exceeded 0.6. To observe the considerable impact of the two class models employed and the effect sizes greater than 0.6, the data from each study are presented in Table 4.

Table 4. Study Effect Measures Included in the Study Studied

Based on the description in Table 4, it can be seen that the effect size of the 28 articles analyzed showed two studies with negative results. In comparison, the other 13 articles had a lower limit that was also negative, and two articles showed negative results on Z-Value. As illustrated in Table 4, these results provide an overview of the effect size and p-values for each study analyzed. This data is an essential basis for conducting further analyses in this study. In addition, the data presented in Table 4 will be used to examine the heterogeneity between studies to understand how critical reading learning using a cooperative learning model affects the results obtained. The results of this heterogeneity test, which measures the variation between the studies used in the analysis, can be found in Table 5, which will be the primary reference in the subsequent analysis process.

Table 5. Results of Effect Size Heterogeneity Analysis

The I2 statistic for heterogeneity was determined to be 93.464 (93.46%), p = 0.000 (p < 0.05), indicating that the alternative hypothesis was accepted. The findings of this study demonstrate significant variation among the studies selected for this investigation. Consequently, estimating the average impact size of the 28 studies investigated using random effects is more appropriate. These findings imply that further investigation into the moderate factors influencing the Y variable may be advantageous.

As illustrated in Figure 2, a forest plot for critical reading learning employing a cooperative learning model is presented. Given the continuous variation in critical reading ability outcomes, this forest plot features an unaffected line designated as 0. The average effect sizes can be summarised on the treatment side by utilising critical reading learning and employing a cooperative learning model towards the positive side of the graph, according to the direction of the set effect sizes. The forest plot in Figure 2 reveals that, among the 28 studies conducted, two exhibit a less significant influence measure or negative z-value. The remaining 26 studies demonstrate a significance of <0.05 and a positive z-value. This observation suggests that the majority of studies exhibit consistent effect sizes.

Figure 2. Forest Plot

Publication Bias

As illustrated in Figure 3, the funnel plot for critical reading learning using a cooperative learning model demonstrates a symmetrical distribution of the estimated effect around the centerline. This suggests the possibility of including related trials in the analysis. The symmetry of the funnel plot findings on the axis remains uncertain, necessitating the application of Egger's test for further testing.

Figure 3. Funnel Plot of Precision by Std. diff. in the sense of and with Standard Error

The regression test findings demonstrate a substantial, albeit less pronounced, deviation from zero, as indicated by the P value 0.05 for both the one-tailed and two-tailed tests. The researcher will consider the P value of 0.05 for the two-tailed test, which supports the null hypothesis and suggests the presence of significant asymmetry, thereby ensuring the result is not biased. Furthermore, publication bias was identified by implementing the Begg and Muzambdar rating correlation and the classic fail-safe N, as demonstrated in Tables 6 and 7.

Table 6. Results of Begg and Muzambdar Rank Correlation

Table 7. Hasil N fail-safe classic

The missing study score was recorded in 2089; this figure indicates the existence of potential research that may not be published or detected in this analysis, which could lead to bias. This suggests that some relevant studies may not be represented in the data analyzed, which often influences the results of meta-analyses. However, the significance value of Kendall from the correlation of the Begg and Muzambdar ratings exceeding 0.5000 on both sides provides evidence that there was no significant bias in this study. Based on these parameters, it can be concluded that the studies used in this meta-analysis do not show substantial bias, so the results obtained can be considered valid and trustworthy.

Moderator Analysis

Education Level Moderator Analysis

Based on the analysis, the effect sizes obtained are known to be heterogeneous. The effect sizes included in the study are then calculated and compared according to the level of education, starting from elementary school, junior high school, high school, and university. The results of this comparison are given in Table 8.

Table 8. Results of Moderator Analysis with Random Effects Model for Education Level

The effect sizes (d = 1.307) of the study on critical reading learning using a cooperative learning model, in which student research subjects in elementary school (k = 2) were involved, were categorized as high and did not show statistical significance (z = 7.079, p <.05). Additionally, the effect size score (d = 1.142) from the study conducted with middle school students (k = 2) was also high and showed statistical significance (z = 8.066, p <.05). Conversely, the effect size score (d = 1.498) from the study conducted with students in high school (k = 2) was included in the high category and did not show statistical significance (z = 5.978, p <.05). The subsequent data, pertaining to the highest level of education, exhibited an effect size score (d = 0.649) for studies conducted with students at universities (k = 22), which fell within the intermediate range and demonstrated statistical significance (z = 11.784, p <.05). The results of the analysis between groups demonstrated that the score of the influence of research with different levels of education in critical reading learning using the cooperative learning model was statistically significant (QBG = 412.115, sd = 27, p = 0.000).

Analysis of the State Moderator Study

A quantitative analysis was conducted to calculate and compare the effect sizes for the data of each study included in the study, with the analysis performed according to the author's country of origin. The results of the comparison that were analysed are presented in Table 9.

Table 9. Results of Moderator Analysis with Random Effects Model of Countries Where the Study Was Conducted

An examination of the data presented in Table 9 reveals that the effect size (d = 1.453) of the study conducted in Colombia (k = 1) employing a cooperative learning model for critical reading instruction attained a high level, demonstrating statistical significance (z = 8.960, p < .05). Conversely, the study conducted in Ethiopia (k = 1) exhibited an effect sizes score (d = -0.403), which was classified as low and did not demonstrate statistical significance (z = -1.639, p <.05). Subsequent research undertaken in India yielded a score of effect sizes (d = 2.812) with (k = 1), and the findings were classified within the high category, demonstrating statistical significance (z = 11.488, p <.05). A parallel study undertaken in Indonesia yielded an effect sizes score (d = 0.859) with a total of five studies (k = 5). The results were classified within the high category. They demonstrated statistical significance (z = 7.382, p <.05). Research conducted in Iran demonstrated that the effect sizes score (d = 0.352) with the number of studies as many as 4 (k = 4) and the results were in the intermediate category and showed statistical significance (z = 3.107, p > 0.05). Subsequent research conducted in Libya yielded an effect size score (d = 0.296) with the number of 1 study (k = 1). The study results were included in the intermediate category and were not statistically significant (z = 1.016, p > 0.05). A similar study conducted in Egypt demonstrated that the effect sizes score (d = 1.969) with the number of studies as many as 2 (k = 2), the results of the study were interpreted in the high category and showed statistical significance (z = 8.475, p > 0.05).

Meanwhile, subsequent research in Taiwan demonstrated that the effect size score (d = 0.257) corresponded to several studies of 2 (k = 2). The study findings indicated that the intermediate category did not demonstrate statistical significance (z = 1.554, p > 0.05). Similar studies conducted in Thailand yielded an effect size score (d = 0.692) with five studies (k = 5). The results were included in the high category and showed statistical significance (z = 5.552, p>0.05). Research conducted in Turkey yielded an effect size score (d = 0.530) with two studies (k = 2). The data were interpreted into the intermediate category and showed statistical significance (z = 2.661, p > 0.05). Concurrent studies have been conducted in the UAE, yielding an effect size score (d = 1.042) of 2 (k = 2). The findings of these studies were classified within the high category and demonstrated a level of statistical significance (z = 4.510, p > 0.05). A similar study in Ukraine showed the effect size score (d = 1.041) with several studies of 1 (k = 1). The data were interpreted into a high category and showed statistical significance (z = 4.820, p > 0.05). In contrast, the most recent study undertaken in the United States yielded an effect size score (d = 1.041) with several studies of 1 (k = 1). The data were classified within the vertical category and did not demonstrate a level of statistical significance (z = 0.186, p > 0.05). The results of the inter-group analysis revealed that the effect sizes scores of studies conducted in various countries in critical reading learning using the cooperative learning model were statistically significant (QGB = 413.089, sd = 27, p = 0.000).

Some studies conducted in Ethiopia, Libya, and the United States showed a P value of 1,000, indicating that applying cooperative learning models in such contexts had a weak or negative effect on improving critical reading skills. This high P-value indicates no significant difference or substantial relationship between using cooperative models and improving critical reading skills in students in these regions. This phenomenon raises the question of why models proven effective in many studies show suboptimal results in these countries. To answer this, it is important to understand that the effectiveness of a learning model does not solely depend on the method itself but is also highly determined by the social, cultural, and educational context surrounding it.

Thus, high P values in studies from Ethiopia, Libya, and the United States do not solely reflect the ineffectiveness of cooperative learning models universally, but rather reflect the complexity of the local challenges faced in their implementation. Geographical, social, cultural, and educational policy contexts greatly influence the extent to which a model can work optimally. Therefore, the results of the meta-analysis must be comprehensively understood, taking into account contextual differences between countries and regions. This understanding is important so that the conclusions drawn are not misleading generalizations, but are reflective and adaptive to the reality on the ground.

Publication Type Moderator Analysis

Effect sizes for research data included in various forms of publications, such as scientific articles and proceedings, serve as a metric for conducting comparative analyses of research findings. The results of these comparative analyses are presented in Table 10.

Table 10. Results of Moderator Analysis with Random Effects Model for Publication Types

As illustrated in Table 10, the critical reading learning study employing the cooperative learning model, disseminated through scientific articles (k = 23), demonstrates a substantial effect size (d = 0.792). This study exhibits statistical significance (z = 15.274, p < .05), suggesting a high level of efficacy. The results of the effect sizes score (d = 0.666) from the study published in proceedings (k = 5) were included in the intermediate category. They showed statistically significant results (z = 5.278, p <.05). The findings of the inter-group analysis demonstrated that the score of the influence of research with different types of publications in critical reading learning using the cooperative learning model was statistically significant (QGB = 413.089, df = 27, p = 0.000).

Moderator Analysis of Types of Learning Models

The effect sizes for the research data incorporated into the learning model, which encompasses direct learning, conventional learning, cooperative learning, scientific learning, and traditional learning, serve as a metric for comparative analysis. Table 11 elucidates the outcomes of this comparative analysis.

Table 11. Results of Moderator Analysis with Random Effect Model on Learning Model

As demonstrated in Table 11, the critical reading learning study employing a cooperative learning model, which was taught using a direct learning model (k = 7), exhibited a substantial effect size (d = 0.562). This effect size was categorized within the high category and demonstrated statistical significance (z = 5.455, p < .05). The results of the effect sizes score (d = 0.798) from the research on critical reading learning using the conventional learning model (k = 5) were included in the high category and showed a statistically significant level (z = 7.665, p <.05). Additionally, the results of the effect sizes score (d = 0.848) from the research on critical reading learning using the cooperative learning model (k = 13) were interpreted in the high category and showed a statistically significant level (z = 11.364, p <.05). Conversely, the results of the research on critical reading learning taught using the scientific method yielded an effect sizes score (d = 0.073) with a total of 2 (k = 2). The analysis results of the effect sizes score were included in the intermediate category. They did not show a statistically significant level (z = 0.401, p <.05). On the other hand, results about the conventional model of teaching critical reading showed an effect size score (d = 1.453) with a total of 1 (k = 1). The effect sizes score analysis indicated a high category which showed significance at the level of (z = 8.960, p <.05). The results of the inter-group analysis revealed that learning model of the study and determination of the achievement of critical reading learning through the cooperative learning model was significant (QGB = 413.089, df = 27, p = 0.000).

Conclusion

The focal results of this research show that the cooperative learning model has a more significant relative impact on the development of critical reading skills than other models of learning. The meta-analysis study showed that using this model goes beyond helping students understand textual materials. It also helps actively engage them to argue, make inferences, and evaluate the materials received. Such meta-cognition processes of information comprehension supporting the notion of critical reading skills as sophisticated skills are best trained in inter-student cooperation. These outcomes suggest a new perspective regarding the strength of cooperative learning which has been overlooked.

The approach that integrates information technology into education as a science, specifically the theory of cooperative learning and critical reading skills, has dramatically advanced due to this research. Theoretically, this study enhances the existing knowledge on the effectiveness of cooperative learning models in the context of critical reading skills. This subject has been extensively explored in the realm of higher education. From a practical standpoint, this study offers empirical evidence that can serve as a reference for educators in designing more interactive and collaborative learning strategies, which, in turn, can enhance the quality of student literacy. Consequently, the cooperative learning model is applicable in formal education settings and can be adapted to various other learning contexts.

Nonetheless, this study's findings are very promising, but some limitations must be observed. For example, the study only examines previously published studies and has not directly tested the application of cooperative learning models in the field. Therefore, further development can focus on research that implements the model at different levels of education and cultural contexts. More in-depth studies on the inhibiting factors in the application of this model can also provide more comprehensive insights to improve its effectiveness in the future.

Recommendation

The findings of this study offer several significant recommendations for developing critical reading skills through a cooperative learning model. Educators are advised to adopt this model more broadly, given its effectiveness in improving student participation and critical reading and thinking skills. Moreover, teachers are required to undergo specific training on the implementation of cooperative learning models which includes practical approaches and evaluation of their effectiveness. It is also essential to further examine the different contexts within which different cooperative learning models are implemented to determine their applicability and effectiveness across all educational spheres.

Limitation

Notwithstanding, it is critical to note the limitations of the study. First, the analysis was based on existing literature which does not allow for inclusion of any unpublished work. Second, the diversity in research methods and designs of the provided literature make synthesizing an overarching conclusion infeasible. Moreover, the study did not incorporate all possible learning environments, so the findings may not be completely applicable across different settings. These shortcomings are bound to be encountered in the future and will hopefully allow the issue of the effectiveness of cooperative learning models to be more comprehensively understood.

Acknowledgements

This research received support from the Center for Higher Education Funding and Assessment (PPAPT) and the Education Fund Management Institute (LPDP) through the Indonesian Education Scholarship (BPI) of the Ministry of Education, Culture, Research, and Technology, with the scholarship recipient's identification number 202329112981.

Conflict of Interest

The authors declare that neither real nor perceived conflicts of interest exist. Artificial intelligence was not used in the preparation of this paper.

Authorship Contribution Statement

Setiawan: Conceptualization, design, analysis/interpretation, and writing. Suwandi: conceptualization, final approval, reviewing, and supervision. Rohmadi: reviewing and critical revision of the manuscript.