Over the past two decades, the theoretical and review articles on CL in the field of PE have dealt extensively in both the goals and the outcomes of the CL model. Casey and Goodyear (2015), who analyzed 27 articles that aimed at exploring the empirical research on the use of CL in PE, reported on learning achievements in the physical, cognitive, and social domains, and even in the affective domains albeit to a lesser extent. The researchers reported that CL enhances academic learning through the physical and cognitive learning domains, as students acquire a certain level of physical competence and develop an understanding of movement techniques and tactics as a result of their engaging in CL activities. They claimed that one explanation for this enhanced academic learning is the increased opportunity for discussions and face-to-face interactions between students, which increases their opportunities for engaging in higher order thinking skills. The researchers also found social learning outcomes to include the development of interpersonal skills, interpersonal relations, the ability to listen and to speak coherently, while sharing ideas and constructing a new understanding together. Finally, they found that social learning outcomes also encompass students’ ability to exhibit caring, empathy, respect, and support.

Bores-García et al. (2021), who later reviewed 15 studies that were conducted during 2014–2019, pointed to the contribution of CL to all dimensions of the human personality, while stating that the affective domain, however, had been inadequately addressed in the examined studies. Both studies (Casey and Goodyear, 2015; Bores-García et al., 2021) assert that short implementation durations would not yield sustainable learning outcomes, and claim that teachers struggle with implementing the model because of its complexity, and that more time is needed to assimilate it in school. Chiu et al. (2014) conducted content analysis on 15 articles that dealt with CL in the physical education curriculum. In line with the findings of Bores-García et al. (2021) and Casey and Goodyear (2015), they demonstrate that the main educational value that CL brings to the physical education curriculum is social responsibility. However, a slightly different point of view is expressed in a recent theoretical article (Montoya et al., 2020), where motivation was found to be the main educational value that arises from CL. In addition to presenting the varied outcomes of CL, researchers are concerned that a significant proportion of interventions are short-lived, and that in some interventional studies, the specific intervention plan is unclear (Casey and Goodyear, 2015; Bores-García et al., 2021), rendering the fidelity of the study of CL in PE somewhat obscure (Casey et al., 2015).

The terms teaching model and teaching strategy are not synonymous. A teaching model presents achievable didactic teaching principles and defined teaching goals that are rooted in a defined educational approach (Joyce et al., 2015). A teaching strategy, on the other hand, is subject to the chosen teaching model and presents teaching methods, i.e., ways of teaching and techniques that makes it possible to comply with the didactic principles of the model (Joyce et al., 2015). Indeed, the CL model was originally created to deal with learning gaps in the heterogeneous classroom; its principles include creating a positive dependence on members of the cooperative group, creating a proactive interaction that allows each participant to contribute to the group, etc. One CL strategy is the Jigsaw model, in which the teacher divides the class into small heterogeneous home groups of four-seven children. For each topic, a different representative from each group is taught the material, so that they can then return to the home group and teach their new specialty to the other group members, thereby increasing the participation of each and every student in the group.

Bores-García et al. (2021) specify the CL teaching strategies used in the reviewed research studies (Jigsaw, Joint Action Studies in Didactics, and Learning Teams), yet most do not present such detailed data (e.g., Chiu et al., 2014; Casey and Quennerstedt, 2020; Montoya et al., 2020). Several researchers in the field of CL (e.g., Hastie and Casey, 2014; Casey et al., 2015; Dyson and Casey, 2016) present five fundamental CLM principles: (1) positive interdependence; (2) individual accountability; (3) group processing; (4) promoting face-to-face interaction; and (5) small groups and interpersonal skills. However, they claim that as implementing all five elements is too complex for teachers, the specific steps while applying the model should be further examined, to understand the outcomes. This argument, which points to the difficulty of applying CL principles, justifies the vast efforts that have been made by numerous researchers in designing detailed and interwoven teaching strategies in a manner that allows for expected outcomes to be achieved. The current study therefore strives to specifically examine the teaching strategies that are implemented in CL intervention studies in PE, their outcomes, and the relationships between the two.

In this study, we conducted a systematic review of 44 research studies (from 2,576 initial studies) that were published over a 20-year period, from January 2000 to April 2020. The search for sources focused on field research articles in which there was an examination of the impact of CL intervention programs that entailed physical activity on processes and achievements. Our review included quantitative studies, qualitative studies, and mixed-method studies. The search was conducted via seven electronic databases, including ERIC, Google Scholar, SPORTDiscus, EBSCO host, and Web of Science, ProQuest, and ScienceDirect. The descriptors used for the search included “cooperative learning” or “collaborative learning”; “school” or “class”; and “physical education,” “physical activity,” or “movement.”

The exclusion criteria included the following: (1) duplicated articles, (2) articles that were not published in journals that are indexed in the Journal Citation Report (JCR) or in the Scimago Journal Rank (SJR); (3) articles written in languages other than English; (4) articles that do not address an intervention program; and (5) articles that address an intervention program yet do not mention any of the CL teaching or learning process examined in this study. The search summary, using the Prisma framework (Page M. J. et al., 2020; Page M. et al., 2020), is presented in Figure 1 as flow chart.

The article sorting process resulted in our identifying 44 articles that met all of our inclusion criteria. As past PE teachers, we chose to gather data that could help teachers understand how CL research might help them while considering this method for their purposes. Therefore, the data were examined in relation to: (1) outcomes in social, physical, affective, and cognitive domains; and (2) type of CL teaching strategy – (a) Jigsaw (JIG); (b) Learning Team (LT); (c) Complex Instruction (CI); (d) Student Team – Achievement Division (STAD); (e) Performer and Coach Earn Rewards (PACER); (f) Cooperative Learning Model Teaching Strategy (CLM); and (g) combined teaching strategies (Combined). Data were also examined regarding the learning process, including the intervention duration and students’ previous experience in CL. Finally, we also examined the research method (qualitative, quantitative, or mixed methods) and the participants in the article (students vs. teachers; individual vs. group level).

Given the nature of the data that was reported in the reviewed articles, whereby conducting a meta-analysis was not suitable due to inconsistency of reported data between studies, we decided to focus on the presence or absence of various key features. As such, we recorded the data in a binary manner. The outcome of each intervention served as the dependent variable, whereby when the reported outcome was positive and statistically significant, the variable received a value of 1. When the reported outcome was either negative, not significant, or no outcome was reported, the variable received the value of 0. In addition, teaching strategies and additional discrete variables (e.g., type of study, class, grade, etc.) that may have impacted the outcome were similarly recorded, whereby their presence was marked with a value of 1 and their absence was marked with a value of 0. Compiling the data in this manner enabled us to address the following important generic question: Which teaching strategies that were reported as having been implemented in the studies are associated with the reported positive outcomes? Moreover, applying this type of analytical approach in research is the best option in cases where conducting a meta-analysis is not a possible option (Garson, 2012). Finally, the advantage of applying this approach is that it minimizes the number of articles that are excluded from the analysis due to missing data.

Discriminant Function Analysis (DFA) (Haas et al., 2004; Garson, 2012) was used to identify the possibility of classifying (or discriminating between) articles that report positive outcomes and those that do not report such outcomes through features of the interventions or relating to the specific studies. The DFA was conducted separately for each of the four domains of outcomes (social, physical, affective, and cognitive). The predictors applied in our analysis included: (1) type of research method (qualitative, quantitative, or mixed); (2) the level at which the student data and the teacher data were reported (individual or group average); (3) teaching strategy reported (yes or no); (4) duration of the intervention in academic hours (≥24 h, yes or no); and (5) whether the students had previously experienced CL prior to the study (yes or no). The results of the DFA in this review paper are reported using a standardized and unstandardized linear function for each outcome, the two centroids and the cut-off point for classification. We also report the percentage of correct classification, thereby indicating the predictive power of each DF.

It is important to note that the analysis reported in this paper is of articles (i.e., research reports) rather than of the actual studies. As such, most independent variables are classified as being present or lacking. Regarding the dependent variable of outcomes, we assumed that positive outcomes are more likely to be reported while negative or non-significant outcomes are more likely to be omitted (Marks-Anglin and Chen, 2020; Page M. J. et al., 2020; Page M. et al., 2020).

The prevalence of the four examined domains was similar among the articles. Some studies examined variables from two different domains, while a number of qualitative articles only examined processes, not outcomes. Among those that presented positive outcomes, 14 articles presented positive social outcomes, including seven relating to behavioral changes, two dealing with attitudes toward friends and the class, and two dealing with both behavioral changes and attitude changes. Ten articles addressed the physical domain, with a focus on motor abilities in skills, fitness, and games. In the affective domain there were a total of 12 articles, of which seven dealt with improving motivation for coping or succeeding and five dealt with changes in perceptions and attitudes. Finally, in the cognitive domain, 10 articles were identified, of which four showed improvements in academic achievements and six showed improvement in thinking skills.

When examining the strategies that were applied in the review research studies, only three teaching strategies were addressed in more than one article. Of the 44 articles reviewed, the JIG strategy was applied in five studies, LT was applied in five studies, and CI was applied in two studies. Ten articles applied general CLM principles, while 20 intervention programs (almost half of all programs) employed the Combined strategy. Moreover, in several studies, the researchers incorporated principles of non-CL strategies, such as the Sport Education Strategy (Montoya et al., 2020) or the Self-Learning Strategy (Benkhaled et al., 2015).

Table 2 shows that overall, the DFA yielded acceptable-to-high levels of correct classification (68.2–81.8%). The positive DF coefficients were found to be associated with four positive outcomes in the affective and cognitive domains, while negative DF coefficients were associated with positive outcomes in the social and physical domains. Meaningful results were those that yielded a medium or larger effect size.

When the JIG strategy was implemented, positive outcomes were reported in the affective domain, and to a lesser degree in the cognitive and social domains (standardized coefficients = 0.734, 0.404, and −0.403, respectively). When the LT teaching strategy was implemented, positive outcomes were reported in the physical and affective domains, yet with a low-medium effect size (standardized coefficients = 0.377 and 0.360, respectively). When the CI teaching strategy was reported, positive outcomes were seen in the cognitive domain (standardized coefficient = 0.586). When CLM was reportedly implemented, positive outcomes were reported in the affective domain, and to a lesser degree in the cognitive domain (standardized coefficients = 1.160 and 0.421, respectively). Finally, when Combined teaching strategies were reported, positive outcomes were seen in the affective domain (standardized coefficient = 0.797).

When the duration of the intervention was examined as being ≥24 h in terms of academic lessons, a negligible impact was seen on the positive outcomes, with the exception of positive outcomes being reported in the social domain (standardized coefficient = −0.367, i.e., low-medium effect).

When examining the students’ previous experience in CL that was reported in the reviewed articles, no significant associations were found with any positive outcomes, thereby suggesting that this variable has no impact on the positive outcomes reviewed in this study. However, student data that was reported on the individual level was associated with positive outcomes in the cognitive and social domains (standardized coefficients = 0.691 and −0.592, respectively).

It is important to note that the DFA for the affective domain only yielded 68.2% correct classification. Such levels are estimated to be equivalent to a medium effect in comparison to the DFA of the other three domains (i.e., social, physical, and cognitive) that yielded 79.5–81.8% correct classification, which is estimated to be equivalent to a high effect (Coe, 2002).

Further to our findings, some associations were seen between certain teaching strategies and outcomes. The social domain was found to be weakly associated with the JIG model, while the physical domain was found to be weakly associated with the LT strategy. The affective domain was found to be associated with JIG, CLM, and Combined strategies, and to a lesser effect with LT. Finally, the cognitive domain was found to be associated with the CI strategy, and to a lesser degree with JIG and CLM. As such, we cannot fully recommend a certain teaching strategy for teachers who wish to achieve social and physical goals through CL in their PE classes.

In more than half the studies, the intervention duration was relatively short. However, we did not find a relationship between the length of the intervention program and its outcomes. This differs from previous findings, whereby researcher suggest that teachers must invest time and effort in order to conduct effective CL in small groups (Baloche and Brody, 2017). Moreover, Bjørke and Mordal Moen (2020) argued that children who embark on CL, approach it with skepticism at first, only gradually developing a positive attitude toward the process, understanding its benefits, and actually beginning to learn through it. In our research, no association was found between the reported duration of the intervention programs and outcomes on physical, affective, and cognitive domains, yet a small impact was seen, however, on social outcomes.

Only four of the reviewed research studies in this paper addressed students’ previous CL experience and this aspect was not associated with any positive outcomes. Previous studies (e.g., Ghaith, 2018) mention that students who have learned to listen to each other in small group lessons, in classes such as math or language, may apply this CL skill in PE classes as well. In other words, when experiencing CL in one field, students could be expected to transfer this capability to other fields, thereby achieving more meaningful learning. It is therefore recommended that future research examine this aspect of previous CL experience among children.

A range of CL strategies are analyzed in the literature and applied in various teaching professions (e.g., Felder, 2001). Yet in our review, only the following three strategies—JIG, LT, and CLM—were found to have been used as intervention programs in more than one article. The question is therefore, what makes each of these strategies special and what is their contribution to the intervention outcomes in the various dimensions? To answer this, we will now present the main concept of these three teaching strategies and their unique association with PE classes.

Of the 44 research studies reviewed in this article, 10 were not based on a detailed or structured teaching strategy, but rather implemented the more general CLM that is based on five principles (positive interdependence; individual accountability; group processing; face-to-face interaction; and small groups and interpersonal skills). When examining the outcomes of this more general model, our results indicate positive outcomes in the affective domain and a positive yet weaker outcome on the cognitive domain. Yet to maintain fidelity and enable other researchers to replicate research findings, the methodology must be more detailed than it is in these 10 articles (Casey et al., 2015). As such, PE teachers who wish to implement CLM may find it difficult to do so. Future research could therefore benefit from specifying the CLM methodology and comparing studies that employ the same techniques.

Moreover, 20 studies combined principles from different strategies. These studies, which we referred to as Combined, presented outcomes in the affective domain. However, here too, PE teachers may find it difficult to replicate these studies, due to the inability to replicate the methodologies employed. Many studies with identical combination and comparison between different combinations and between different well-defined CL’s teaching strategies may change the picture.

As the articles reviewed in this study were only written in English, articles in other languages such as Spanish and Portuguese were excluded. Therefore, the review may be culture biased. Moreover, throughout this reviewed, we classified each article according to type of teaching strategy employed and outcomes on four domains. With regards to outcomes, our classification differentiated positive outcomes and all other outcomes (negative, unclear, or not reported)—based on the assumption that negative outcomes are less likely to be reported (Torgerson, 2006). That being said, this limitation does not compromise the quality of this literature review, and may even strengthen it, as we acknowledge the possible (or likely) bias that some of the literature may be subject to.